Beruflich Dokumente
Kultur Dokumente
VOLUME 2
The titles published in this series are listed at the end of this volume.
EMERGING MARKET
CAPITAL FLOWS
Proceedings of a Conference held at the Stern
School of Business, New York University on
May 23-24, 1996
Edited by
RICHARD M. LEVICH
S'f'ER:N.
Leonard N. Stern School of Business ,
Acknowledgments . ix
Introduction
Richard M. Levich Xl
This volume presents the proceedings of a conference held at the Stern School
of Business, New York University on May 23-24, 1996. It is a pleasure to
thank Edward Altman and Jianping Mei for their assistance in planning the
conference, and also to thank Ingo Walter, the Director of the NYU Salomon
Center, for encouraging us to undertake this conference. We also thank Brinson
Partners, Inc. and Salomon Brothers Inc for their financial sponsorship of the
conference. Thanks are also due to Mary Jaffier who efficiently handled the
numerous administrative arrangements for the conference. Finally, we thank
the authors of the papers and the discussants for contributing their energy to
this project, and bearing with us until the completion of this conference volume.
IX
RICHARD M. LEVICH
New York University
Introduction
In a little over one decade, the spread of market-oriented policies has turned
the once so-called lesser developed countries into emerging markets. In 1982,
the 32 developing country stock markets surveyed by the International Finance
Corporation had a market capitalization of $67 billion representing about
2.5% of world market capitalization. By the end of 1995, the market capitaliza-
tion of emerging stock markets had grown 28-fold, exceeding $1.9 trillion or
10.7% of world equity market capitalization.
Many forces have been responsible for the tremendous growth in emerging
markets. Trends toward market-oriented policies that permit private ownership
of economic activities, such as public utilities and telecommunications, are part
of the explanation. Corporate restructuring, following the debt crisis of the
early 1980s, has permitted many emerging-market companies to gain interna-
tional competitiveness. And an essential condition, a basic sea-change in eco-
nomic policy has opened up many emerging markets to international investors.
The growth in emerging markets has been accompanied by volatility in
individual markets, and a sector-wide shock in 1995 after the meltdown in the
Mexican Bolsa and Mexican peso. This experience led some economists to
question the economic fundamentals underlying emerging markets, suggesting
instead that the surge in emerging-market capital flows was the sign of a
speculative asset bubble - the sort of fad, or mania that cannot be sustained.
At the same time, other economists argued that emerging markets were a fully
credible and permanent element of international capital markets, albeit a sector
with special risk considerations.
Emerging-market capital flows continue to be the subject of intense discus-
sion around the world among investors, academics, and policymakers. This
book examines the issues of emerging-market capital flows from several distinct
perspectives. The research papers in this volume address a number of related
questions about emerging markets, including:
What have we learned from the history of emerging-market capital flows
with respect to market performance, stability and impact on economic
growth?
What are the return characteristics of emerging-market equities and how
do they affect capital flows? How do emerging-market risks and diversifica-
tion opportunities change over time?
What are the benefits of international cross-listing of securities?
How closely are emerging and developed equity markets integrated?
What are the impacts of political risk in emerging markets?
Xl
xii Introduction
Ross Stevens analyze the relationship between political risk variables and
returns in emerging equity markets.
In Part IV, we turn to examine lending to emerging markets through bank
lending and sovereign debt issues. Peter Aerni and Georg Junge survey the
developments in cross-border bank lending to emerging markets. The pricing
and hedging of interest rate risk inherent in Brady bonds is analyzed in a paper
by Jacob Boudoukh, Matthew Richardson, and Robert Whitelaw. Estimates
of the country and currency risk premia represented in Mexican sovereign debt
are the focus of a paper by Ian Domowitz, Jack Glen and Ananth Madhavan.
And Robert Bernstein and John Penicook explore the practical aspects of
portfolio investment in emerging-market debt securities.
Finally, in Part V we present two papers dealing with aspects of corporate
debt in emerging markets. Edward Altman, John Hartzell, and Matthew Peck
analyze the methods for assessing risk ratings on corporate debt from emerging
markets. And Oliver Hart, Raphael La Porta Drago, Florencio Lopez-de-
Salinas, and John Moore describe an alternative, market-based approach to
bankruptcy proceedings in emerging markets, and assess the pros and cons of
their approach relative to existing bankruptcy procedures.
This conference volume, in the tradition of the New York University
Salomon Center, presents original thematic papers written by academic, govern-
ment, and industry observers of emerging markets around the world, with
discussants drawn from international banking and securities practitioners and
government regulators. We hope that this book offers the reader a valuable
window on the economic and financial research and policies issues that effect
this important, fast growing segment of the global financial market.
PART ONE
INTRODUCTION
The financial crisis in Mexico in late 1994 and early 1995 came as a surprise
to most observers, not so much because there was a major devaluation of the
peso, but because the aftermath of this devaluation left the Mexican financial
system and economy in a crisis from which it only in 1996, more than a year
later, started to recover. The conventional wisdom, as presented by, for example,
Dornbusch and Werner (1994), was that a devaluation was exactly what Mexico
needed to spur exports and growth. Instead, the devaluation occurred more or
less simultaneously with (and perhaps touched off) a debt crisis in which the
Mexican government found itself unable to roll over its debt. Fears of a default
of one sort or another totally paralyzed the economy in late December 1994
and January 1995. An explanation of the Mexican crisis that focuses on
Mexico's government debt has a puzzling aspect, however: in 1994 Mexico had
a very low ratio of government debt to national product by international
standards (see Table 1).
Source: International Monetary Fund and Organization for Economic Cooperation and
Development.
* The research reported here has been supported by a grant from the Air Force Office of
Scientific Research, Air Force Materiel Command, USAF, under grant number F49620-94-1-0461.
The US government is authorized to reproduce and distribute reprints for government purposes
not withstanding any copyright notation thereon. The views expressed herein are those of the
author and not necessarily those of the Air Force Office of Scientific Research or the US government.
3
R. Levich (ed.), Emerging Market Capital Flows, 3-22.
< 1998 Kluwer Academic Publishers.
4 T.J. Kehoe
This paper traces the events in 1994 that left Mexico vulnerable to a debt
crisis - the steady conversion of government debt into short-term, dollar-
indexed tesobonos. This conversion seems to have been the result of an
agreement, both implicit and explicit, between the Mexican government and
members of the international financial community.
The paper also proposes a theoretical framework, based on a model devel-
oped by Cole and Kehoe (1996b), for identifying situations in which a debt
crisis can occur. The essential feature of this framework is that there is an
interval of levels of government debt, called the crisis zone, which depends
heavily on the maturity structure of the debt, for which a crisis can occur.
Furthermore, the framework formalizes the idea of 'herd behavior' of investors
often discussed in the popular press: investors feared that Mexico would be
unable to honor its commitments on government bonds becoming due. This
made these investors unwilling to purchase new bonds. Since these fears were
widespread, Mexico was unable to sell new bonds and, consequently, was in a
position where default of some sort seemed inevitable. This situation then
justified the expectations that Mexico would be unable to honor its commit-
ments. Had these expectations not been present, however, no crisis would have
occurred.
This theoretical framework suggests an accounting methodology for calculat-
ing the size of the crisis zone. Identifying situations in which debt crises can
occur potentially goes a long way towards eliminating the possibility of such
a crisis, since governments would then face strong pressures from financial
markets to stay out of the crisis zone. The theoretical framework also suggests
a significant potential role for an international lender oflast resort which would
sharply limit the possibility of debt crises in a manner similar to that in which
a central bank can limit the possibility of runs on private domestic banks. As
recent banking crises in such countries as Japan and the USA have illustrated,
however, it is essential that a central bank serve as regulator as well as lender
of last resort. Similarly, any international agency that would serve as lender of
last resort should be willing to provide this service only at the price of being
able to regulate government financial policy.
Carlos Salinas de Gortari made two decisions that resulted in the financial
crisis: first, it allowed the Mexican peso only a small devaluation (a nominal
12%) against the US dollar over the course of the year, and in maintaining
the value of the peso, without adjusting its monetary policy, it lost most of
Mexico's foreign reserves. Second, as the Salinas administration refinanced
Mexico's government debt during 1994, it allowed the debt to become mostly
short-term and dollar-indexed.
The combination of these two decisions left Mexico open to a speculative
attack, when investors realized that the Banco de Mexico did not have enough
reserves to continue supporting the peso, and shortly afterwards, a 'bank run',
when bond holders realized that the Banco de Mexico did not have enough
reserves to meet the payments becoming due on the dollar-indexed debt.
In 1994, as it had in 1992 and 1993, Mexico ran a large current account
deficit. What changed in 1994 was the level of foreign portfolio investment
(Figure 1). 1994 was a difficult year politically for Mexico: there was an uprising
in Chiapas in January; the presidential candidate of the ruling Partido
Revolucionario Institutional (PRI), Luis Donaldo Colosio Murrieta, was assassi-
nated in March; the Secretary of the Interior, Jorge Carpizo McGregor, who
had been encharged with ensuring honest elections in August, threatened to
resign in June; the Secretary General of the PRI, Jose Francisco Ruiz Massieu,
was assassinated in September; Ruiz Massieu's brother Mario resigned as
assistant attorney general in November, charging a high-level coverup of the
35,---------------------------------------------------------,
30
DIRECT PLUS
PORTFOLIO INVESTMENT ___ "
25
,
<I> 20
,
~
0'5
ui
::i
~
II) '0
"Of-'9-e-,-+-'-98-2-+-'-98-3-+---'984---">-'98-5-+--'98-6-+-'9-87-+-'-98-8-+-'-98-9-+---'990-i--'99-'-+-":"'9":"92-+--:-'99:::3::-+-:'994:::-:-'
assassination within the PRI; and there were threats of new uprisings in Chiapas
in November and December.
The political uncertainty generated by these events, combined with rising
interest rates that made the USA a more attractive investment target, resulted
in a substantial drop in foreign investment: foreign portfolio investment in
Mexico fell from US$ 28.4 billion in 1993 to US$ 8.2 billion in 1994. (It is
worth noting, however, that foreign direct investment actually rose from
US$ 4.9 billion to US$ 8.0 billion.)
Perhaps even more significantly, there were presidential elections in August,
with the new president, Ernesto Zedillo Ponce de Leon who had replaced
Colosio as the PRI candidate, taking office in December. The change of
government was, as it has been every 6 years in Mexico since 1928, a time of
great uncertainty. At the end of each of the previous three administrations -
in 1976, 1982, and in 1987 - there had been large devaluations. Mexicans and
foreign investors had come to associate ends of presidential terms with
devaluations.
In the face of the drop in foreign investment, the Salinas administration
continued to maintain the value of the peso against the dollar. There were
good reasons to do so, at least during the first half of 1994. A series of social
pacts negotiated between leaders of government, business, and labor had, since
1987, set a policy of a maximum allowable rate of depreciation of the peso
against the dollar. This policy had resulted in a decline in the rate of inflation
in Mexico from 159.2% in 1987 to 7.1 % in 1994. At the same time real wages,
which had fallen sharply following the 1982 financial crisis, rose by more than
20% between 1987 and 1994.
To the extent to which the Salinas administration believed that the shocks
that buffeted Mexico in 1994 were transitory, it was justified in selling the
Banco de Mexico's foreign reserves to insulate Mexico from these shocks. At
the same time that Mexicans and foreigners were selling pesos for dollars, the
Banco de Mexico was sterilizing by reissuing the pesos. This policy was designed
to promote a stable money supply and interest rates. With elections due in
August, it is easy to understand why these sorts of policies were attractive
during the first three-quarters of 1994.
Policy judgements often involve calculated risks, and poor judgements are
far easier to identify if there is a run of bad luck than if there is not. As political
shocks continued to hit Mexico during the fall of 1994, foreign reserves fell to
dangerously low levels. November was a crucial month: it was in November
that foreign reserves fell below the Mexican monetary base, and on
November 18 alone the Banco de Mexico had to sell US$ 1.7 billion to maintain
the value of the peso.
Figure 2 traces out the behavior of foreign reserves held by the Banco de
Mexico during 1994. It is worth noting that the Banco de Mexico made
Can debt crises be prevented? 7
~r-------~~------------------------------------------.
25
DEC JAN FEB MAR APR MAY JUN JUI. AUG SEP OCT NOV DEC
1993 1994
1ro,~--------------------------------------------------------,
~---------~,~----------------------~-
140
;,-...-- ....
," --,
........ " "
,,
....
\ RESERVES
~---.---.---.---.--- '"
40
MONETARY BASE
20
o
"
D lot lot A S N D
1993 1994
Figure 3. Mexican international reserves vs. money supply. December 1993-December 1994.
Source: Secretaria de Hacienda y Credito Publico.
Mexican government debt can be divided into two broad categories: domestic
and external. This division has nothing to do with who holds the debt; rather
it depends on where it is sold. Domestic debt is sold at auctions held by the
Banco de Mexico, while external debt is sold abroad. The debt crisis was
caused by a run on domestic debt. Although yields on such external debt
instruments as Brady bonds increased sharply on secondary markets during
the crisis, Mexican external debt has a long maturity structure. The immediate
Can debt crises be prevented? 9
danger of default was the result of the short maturity structure of the domes-
tic debt.
Table 2 traces the evolution of composition of Mexican domestic government
debt during 1994. There were four types of debt instruments: certificados de fa
tesoreria de fa federaci6n (cetes), peso-denominated bonds with maturities of
28,91, 182,364, and 728 days; tesobonos, dollar-indexed bonds with maturities
of91, 182, and 364 days; bonos de desarrollo (bondes), peso-denominated bonds
with maturities of 1, 2, and 10 years; and ajustabonos, inflation-indexed, peso-
denominated bonds with maturities of 3 and 5 years.
Following the assassination of Colosio in March, the Mexican government
steadily converted its domestic debt from peso-denominated cetes, bondes and
ajustabonos into dollar-indexed tesobonos, as depicted in Figure 4. In the second
week of March 1994, due to uncertainty about the situation in Chiapas and a
possible independent presidential campaign by Manuel Camacho Solis, who
had been edged out as the PRI candidate by Colosio, the peso had begun to
fall against the dollar. The assassination accelerated this fall, and the peso
moved from the bottom to the top of its trading band, devaluing by almost
8% over a month. This drop in the value of the peso led to a sharp increase
in Mexican interest rates with a resulting drop in the prices of Mexican bonds
and equities. According to Torres and Vogel (1994), much of this movement
into tesobonos was the result of discussions between representations of the
Weston Forum, a group of New York investment funds, and officials of the
Mexican Finance Secretariat and the Banco de Mexico.
To understand the importance of the move of Mexican debt into tesobonos,
it is worth understanding how these bonds worked. Tesobonos were sold by
the Banco de Mexico in weekly auctions. The Banco received bids in US
Source: Banco de Mexico (1995) and International Monteary Fund, International Financial
Statistics, various issues.
Market values converted to $US (in the case of cetes, bondes and ajustabonos) using the exchange
rate at the end of the month.
10 T.]. Kehoe
~r------------------------------------------------------'
35
PESO DENOMINATED BONDS
., ....
....., , ,
,,
.... _.. RESERVES '"
,,
,
.. --- .. _--
10
TESOBONOS
D M A M A s o N D
11193 1994
Figure 4. Mexican international reserves vs. Government bonds. December 1993-December 1994.
Source: Secretaria de Hacienda y Credito Publico.
M,--------------------------------------
30
25
MEXICAN 91 OA Y CETES
~~----
'0 MeXICAN 364 DAY TESOSONOS ./ I
______ ____ __
....
o~~~ ~ ~ ~~
D J A A o D
.993
made sense only if the Mexican government had private information that the
risk of devaluation was lower than financial markets thought that it was. The
sale of tesobonos demonstrated a confidence on the part of the government
that later events proved unjustified and that, in any case, financial markets did
not share.
12 T.J. Kehoe
Cole and Kehoe (1996b) present a formal model that captures the intuition
that the Mexican debt crisis was self-fulfilling in the sense that the failure of
the government's auctions of new debt put the government into a position
where default seemed inevitable, thereby justifying the panic that led to the
failure of the auctions. This model and relevant results are described in the
appendix.
The model has three types of actors: domestic consumers, who make con-
sumption and investment decisions; foreign investors, who purchase govern-
ment debt and are risk neutral, reflecting the small size of the country relative
to world capital markets; and a government, which taxes, spends on public
goods, offers new bonds for sale, and decides whether or not to honor commit-
ments on old bonds. The central actor in the model is the government. Cole
and Kehoe (1996b) model the government as benevolent in that it seeks to
maximize the welfare of the domestic consumers; they show, however, how it
is also possible to model the government as more impatient than consumers
or international investors. The welfare of consumers and governments depends
both on private consumption and on provision of the public good.
The government cannot commit to repaying its debt; all of the actors know
that the government resolves its maximization problem every period. If the
expected present value of defaulting exceeds that of repaying old debt, the
government will default. If the government defaults, the country is subject to
a penalty that results in a decline in domestic productivity. This penalty reflects,
for example, the large distortion created by the imposition of dual exchange
rates. In the model, for some levels of government debt, a crisis can occur
depending on the realization of a random event that is extrinsic to the funda-
mentals of the model, a sunspot variable. An unfavorable realization of this
sunspot variable can lead to a panic in which the international investors are
unwilling to purchase new government debt. This panic is rational if the failure
of the new debt auction puts the government in a situation where it prefers to
default. At the same time, however, the panic is somewhat arbitrary because a
favorable realization of the sunspot variable would not lead to a panic, the
government would be able to sell its new debt, and no crisis would occur.
In this model a self-fulfilling crisis is possible if the government would choose
to default if no new borrowing were possible, but would choose to honor its
commitments if new borrowing were possible. Cole and Kehoe (1996b) show
that, if a crisis is possible, the probability of its occurence is arbitrary: for any
probability of an unfavorable realization of the sunspot variable, there is a
different equilibrium. Although Cole and Kehoe model the crisis as dependent
on a sunspot variable, it is also possible to model it as dependent on a random
event connected to the fundamentals, such as a political shock. The essential
point is that there are multiple equilibria: there is an equilibrium in which the
shock touches off a crisis and there is an equilibrium in which it does not.
Can debt crises be prevented? 13
The crucial insight of the model is that the government finds itself in a far
different position if it cannot sell its new bonds than if it can. If the level of
government debt is low compared with its ability to raise revenue, however,
these positions are not very different: the government will choose to repay its
debt and to avoid the default penalty whether or not new borrowing is possible.
Similarly, if the maturity structure of the debt is long enough, these positions
are not very different: with government debt of long maturity little new borrow-
ing is needed in anyone period. Figure 6 depicts the size of the crisis zone for
a simple model described in the appendix that has been calibrated to match
the general features of the Mexican economy in 1994. For levels of the debt
too high, a crisis occurs immediately; for low levels of the debt, no crisis is
possible; and, for intermediate levels of the debt - levels in the crisis zone - a
crisis can occur with fairly arbitrary probability. Notice how fast the crisis zone
shrinks as the maturity of the debt increases. For a variety of reasons not
explained by the model, bonds with a short maturity may be less risky than
those with a long maturity if the maturity structure itself is constant.
Nevertheless, one lesson that can be drawn from the model, a lesson that
financial markets did not seem to understand in 1994, is that letting the whole
maturity structure shorten is very risky.
Although the Cole-Kehoe model is only a simple first step at modeling
events like the Mexican financial crisis, it provides a framework that may
eventually be capable of identifying countries in danger of a crisis like that
which hit Mexico in 1994-95. A country is in danger of a crisis if the amount
~O.8
'a"
i
a:: 0.6
~
'&
1
-10.4
0.2
0~0----~----+---~-----+4----~5-----6~---+-----r----~--~10
Average Maturity of the Government Debt in Years
of government debt becoming due during, say, one quarter is large compared
with its ability to raise taxes or to cut expenditures and compared to its foreign
reserves. It is worth noting that a government might find itself in danger of a
self-fulfilling debt crisis even though it would have no problem meeting the
interest payments on its debt, or even in eventually reducing this debt to a
level below the crisis zone. In terms of the literature on bank runs, the govern-
ment may be illiquid even though it is solvent.
The first step towards resolving the Mexican debt crisis was taken on 31 January
1995, when US President Bill Clinton announced a US$ 48.8 billion loan
package put together primarly by the US Exchange Rate Stabilization Fund,
the International Monetary Fund, and the Bank for International Settlements.
It was not until 9 March, however, that the Zedillo administration was able
to put together a complete economic plan for the year. Essential features of
the economic program enacted in Mexico during January-March included a
rapid opening of the banking sector to foreign competition and foreign owner-
ship, an increase in taxes, and measures to aid domestic banks left in precarious
positions by the devaluation. In general, the period from late December through
early March was a period of great uncertainty and economic paralysis in
Mexico.
It is plausible that most ofthe uncertainty and paralysis that afflicted Mexico
in early 1995 could have been eliminated if the loan package had been made
available, say, a month earlier. This has led commentators like Eichengreen
and Portes (1995) and Sachs (1995) to propose international institutional
mechanisms for dealing with crises like that in Mexico. An international lender
of last resort conceivably would have prevented the run on Mexican govern-
ment debt and allowed the Mexican economy to recover very quickly after the
December devaluation.
A crucial question, of course is, When does a lender of last resort lend? It
is one thing to lend to support a budget deficit or to support an exchange rate
that is subject to repeated speCUlative attacks. It is another thing to lend to
stop a run on government debt. Much the same as Bagehot (1873) proposed
a lender of last resort to eliminate domestic bank runs, an international lender
of last resort could, by its very existence, eliminate the possibility of a crisis
like the Mexican crisis: international investors, knowing that the government
has access to an international lender of last resort, would not find it rational
to panic. As shown in the appendix, an international lender of last resort can
eliminate the possibility of a crisis in the Cole-Kehoe model.
It is worth recalling Bagehot's conditions for a central bank to provide
credit to an illiquid domestic bank: the domestic bank should be solvent; it
should provide good collateral; and there should be a penalty interest rate. All
Can debt crises be prevented? 15
of these three features were present in the Mexican crisis and its subsequent
resolution: the debt service on Mexican government debt was very small in
relation to Mexican government revenue; the Mexican government was able
to use its oil export revenues as collateral; and the loan package involved
substantially higher interest rates than those the Mexican government was
paying before the crisis. (See US General Accounting Office 1996 for details
on the agreements.)
The concept of international lender of last resort is one deserving more
attention. It is worth making two related observations, however: First, serving
as lender of last resort without regulating creates a potential moral hazard
problem. Governments will undertake riskier policies if they know that an
international lender of last resort stands ready to bail them out. Second, an
international lender of last resort indirectly inherits much of the responsibilities
of the domestic central bank in serving as lender of last resort to domestic
commercial banks. A country like Mexico in 1994 and 1995, with a fragile
domestic banking system, will use its monetary and debt management policies
to insulate these banks from unfavorable shocks. To properly regulate a central
bank, an international lender of last resort needs to make sure that the central
bank is properly regulating domestic commercial banks.
ApPENDIX
The Model
This section lays out the model utilized by Cole and Kehoe (1996b) to analyze
the 1994-95 Mexican debt crisis. There is a single good in each period,
t = 0,1, .... This good can either be consumed or be saved as capital.
Production utilizes capital and implicitly, inelastically supplied labor. There
are three types of people in the model: domestic consumers, international
investors, and the government. We describe each in turn.
There is a continuum with measure one of identical, infinitely lived domestic
consumers. The consumers' utility function is
co
E L P'(c, + v(g,}}
'=0
where Ct is private consumption and gt is government consumption. We assume
that 0 < P< 1 and that v is continuously differentiable, concave, and monotoni-
cally increasing. The consumer's budget constraint is
,=0
The assumption of risk neutrality of investors captures the idea that the
domestic economy is small compared to world financial markets. Each investor
is endowed with x units of the consumption good in each period and faces the
budget constraint
x, + q,bt +1:::; X + Ztbt.
Here qt is the price paid for one-period government bonds that pay bt +1 in
period t+l if Zt+l=l, but pay if Zt+l=O; ZtE(O, I} is the government's
default decision. We can choose the endowment x to be large enough that we
can ignore corner solutions to the investor's utility maximization problem.
Initially, each investor holds bo units of government debt. There is a constraint
b,+ 1 ~ - A, where A can be chosen large enough so that it rules out Ponzi
schemes but does not otherwise bind in equilibrium.
There is a single government. In every period it chooses its new borrowing
level, B t +1 ; whether or not to default on its old debt, Zt; and the level of
government consumption, gt. If the government defaults by setting Zt = 0, then
productivity drops from OCt = 1 to OCt = oc < 1 from period t onward. In period 0,
the government debt is Bo = boo
The government is benevolent in that its objective is to maximize the welfare
of domestic consumers. Its budget constraint is
gt + ZtBt :::; ()(oc,f(k t ) - t5k t } + qtBt + 1
We do not need to impose a borrowing constraint on the government to rule
out Ponzi schemes because, if the government tries to sell too much new debt
B t +1, the price qt falls to zero. That the tax rate is constant captures the idea
that changes in tax policy occur much more slowly than do debt crises.
In each period there is an exogenous sunspot variable (t, whose value is
realized. This variable is independently distributed on the interval [0, 1]. In
the next section we describe equilibria where, if the level of government debt
B t is above some crucial level and (t is below another crucial level, then
international investors are not willing to buy new government debt and the
Can debt crises be prevented? 17
Cole and Kehoe (1996b) carefully define a recursive equilibrium for their model.
The crucial elements of this definition are policy functions that describe the
optimal actions for each actor given the actions of other actors that have
already occurred and taking into account the optimal responses that will follow.
For the domestic consumers, there is an optimal consumption policy Ct and
an optimal investment policy kl + l . Each is a function of the aggregate state
variable St = (Btl K t , at-I> 'I)' the individual holdings of capital entering the
period kl (which in equilibrium is, of course, equal to the aggregate capital
stock K r), and the actions that have already occurred in the period relevant
for the consumer's decision in the current period or for the determination of
the aggregate state in the subsequent period, B1 + 1 , gtl and ZI' Thus, consumers
actions are determined by the policy functions
't
Cole and Kehoe (1996b) demonstrate that there is always an equilibrium in
which agents ignore the realizations of the sunspot variable as long as the
government prefers to repay its debt by setting Zt = 1. Furthermore, this equilib-
rium is stationary in that it satisfies Bt = B, Ct = cn and kt = k n , where
(1- lJ)(f'(kn) - <5) = 1/P-1
cn = (1 - lJ)(/(kn ) - <5kn).
To calculate the maximum level of debt for which the government prefers to
repay its debt, we need to calculate the levels of consumption and investment
that will take place if there is a default:
For the government to prefer not to default it must satisfy the constraint
that the utility of not defaulting is greater than or equal to that of defaulting:
Cole and Kehoe (1996b) refer to this constraint as the participation constraint.
(They also show that an equilibrium without default may be possible even
when this constraint is violated if the government reduces its debt immediately;
this case is not very interesting, however, because there is no explanation of
how the government could start with a level of debt high enough to violate
the constraint.)
The government is in a very different position if it is not able to borrow
from the international investors because qt = O. In this case it is optimal for
the government to default if the utility of not defaulting is less than that of
defaulting:
E L 0.97'(c, + log(g,)).
,=0
The discount factor of 0.97 corresponds to a yearly discount factor of 0.955
(=0.97 3/2 ), which implies a yearly yield of 0.047 (=0.955- 1 -1) on risk-free
bonds - to be thought of as US Treasury bills. If we set the probability of
default to be 1t = 0.02, then the yearly yield on Mexican government bonds -
to be thought of as tesobonos - would be 0.079 (=[(0.97)(0.98)r 3/2 -1).
Consequently, there would be a 3% risk premium on the Mexican government
bonds. These numbers roughly match the average yields on 90-day US T-bills
and 91-day tesobonos during 1994 (see Figure 5).
The choice of the funtional form v(g) = log(g) is somewhat arbitrary. In
comparison with a function that displays more curvature, such as v(g) = - g- 1,
this function allows the government a fair amount of substitutability in govern-
ment consumption over time. This, in turn, allows the government flexibility
in reducing its expenditure to be able to payoff old debt in the event of no
lending. Consequently, setting v(g) equal to log (g) rather than to - g -1 increases
the upper level of the debt for which no crisis can occur. The main point of
the model is to show that crises can occur for fairly low levels of debt, however,
and our choice of v(g) biases, if anything, the results against crises.
The technology is given by the feasibility constraint
c, + g, + k'+1 - 0.95k, + z,B, ~ 2k?4 + Q,B'+I'
The choice of b = 0.05 corresponds to a yearly depreciation rate of 0.074
(= 1 - 0.95 3/2 ). The capital share 0.4 is lower than that found in Mexico's
national income and product accounts, but the published numbers include
income of self-employed workers in capital income. The constant in the pro-
duction function is a scaling factor that only influences the results because the
utility function c + log (g) is not homothetic.
If consumers expect default to occur nest period with probability 1t = 0.02,
they set k'+1 = k" where k" solves
(1- 0)[(0.98 + 0.21X)0.8(k,,)-0.6 - 0.05] = 1/0.09 - 1.
Setting 0 = 0.2 and IX = 0.95 - which implies that Mexico would incur a perma-
nent drop in productivity of 5% if it were to default - we obtain a capital
stock of k" = 39.04 and a yearly GDP of (3/2)2(k")0.4 = 12.99, for a capital/
output ratio of 3.00. The investment/GDP ratio is 0.05k"/(2(k,,)0.4) = 0.23
Can debt crises be prevented? 21
REFERENCES
Bagehot, Walter (1873). Lombard Street: A Description of the Money Market. New York: Scribner,
Armstrong.
Banco de Mexico (1995). Informe Anua11994. Mexico: Banco de Mexico.
Cole, Harold L. and Timothy J. Kehoe (1996a). Self-Fulfilling Debt Crisis and Capital Flight,
in progress.
Cole, Harold L. and Timothy J. Kehoe (1996b). A self-fulfilling model of Mexico's 1994-95 debt
crisis, Journal of International Economics, 41,309-30.
Dornbusch, Rudiger and Alejandro Werner (1994). Mexico: stabilization, reform, and no growth,
Brookings Papers on Economic Activity, I, 253-297.
Eichengreen, Barry and Richard Portes (1995). Crisis? What Crisis? Orderly Workoutsfor Sovereign
Debtors. London: Centre for Economic Policy Research.
Sachs, Jeffrey (1995). Do We Need an International Lender of Last Resort? Harvard University,
unpublished manuscript.
Torres, Craig and Thomas T. Vogel (1994). Market forces: some mutual funds wield growing clout
in developing nations. Wall Street Journal, 14, June, 1.
US General Accounting Office (1996). Mexico's Financial Crisis: Origins, Awareness, Assistance,
and Initial Efforts to Recover. Washington: US G.A.O.
ANDRES VELASCO! and PABLO CABEZAS 2
1 New York University and NBER 2 New York University
INTRODUCTION
Starting in the late 1980s, and as a response to the economic slowdown and
the fall in interest rates in industrialized economies, private capital returned to
Latin America. Tens of billions of dollars started pouring in, with the flow to
some countries exceeding the records set during the petro-dollar euphoria of
the late 1970s. Policy makers in Latin America welcomed this at first. Foreign
capital could finance much needed imports and investment, reigniting growth
after years of stagnation. Perhaps, thanks to the newly arrived dollars, the 'lost
decade' of the 1980s would be finally over.
But the capital inflow soon created problems. The incoming money tended
to strengthen Latin American currencies to the point where many new exporting
firms (an important creation of the trade reforms of the 1970s and 1980s) found
it difficult to compete abroad. External deficits, not a bad thing in themselves,
tended to grow beyond what prudence dictated. Banks had to recycle the funds
domestically, sometimes over-extending themselves into risky areas such as
real estate and consumer credit.
Optimists minimized the extent of such dangers. With reformed governments
running fiscal surpluses, they stressed, private firms were doing the borrowing.
Capital arrived to finance productive investment, not government-owned white
elephants. Currency appreciation was simply a sign of growing productivity
and economic strength. Nothing could possibly be wrong with this.
Since the Mexican meltdown of December 1994, the tone of this debate has
changed. What most of Latin America learned in the early 1980s has been
rediscovered recently: foreign capital can leave just as quickly as it arrived.
Debt-led expansion can easily turn into financial panic. As a result, the
'Washington consensus' no longer claims that "the more foreign capital, the
merrier." Spokesmen for institutions as venerable as the IMF and the US
Treasury have endorsed arbitrary rules of thumb, arguing that any current
account deficit (and the accompanying capital inflow) over 3% of GDP is
dangerous. Economists with solid neoclassical training have found themselves
defending mild forms of capital controls. Heterodoxies that seemed long
discredited are suddenly back in fashion.
23
R. Levich (ed.), Emerging Market Capital Flows, 23-47.
1998 Kluwer Academic Publishers.
24 A. Velasco and P. Cabezas
What are the proper lessons from Latin America's most recent boom and
(in some places) bust cycle? This paper revisits the debate over the consequences
of capital inflows and the appropriate policy response to them by comparing
the experience of Chile and Mexico. These countries are interesting for two
reasons. The first is that, thanks to their strong record of economic reforms,
both countries have been the most successful in the region in earning the
confidence of foreign investors. Table 1 shows that, relative to the size of the
economy, capital inflows were especially high for Chile and Mexico. For Chile,
during the period 1990-1994, capital inflows averaged more than 6.7% of
GOP. In the Mexican case, net capital inflows averaged 6.2% of GOP during
the same period. Other Latin American economies also experienced some
capital inflows, but not of the same magnitude received by Chile or Mexico:
for the period 1990-1994, Argentina received inflows that averaged 1.2% of
GOP, while Brazil's capital inflows averaged 1.8% of GOP.
The second is that both countries differed substantially in their policy
response to the capital inflows. As is well known, Mexico embraced foreign
capital warmly (skeptics will say that, given the size of external deficits, it had
no choice), while Chile has endeavored to keep it at arms' length. But neither
country is run by wild-eyed populists taking advantage of the inflows to embark
on an unsustainable fiscal expansion, nor by grim-faced bureaucrats denouncing
the evils of foreign investment and closing the door to all inflows. On the
Table 1.
Reserves
Trade Current Capital
balance account account (flow) (stock)
contrary, both countries have often been hailed as paragons of sound macro-
management. Insofar as Mexico's and Chile's policy stances have differed, they
have differed in respects about which 'reasonable people can disagree'. That is
why they provide especially fertile ground for extracting useful policy lessons.
What do (or should) policy makers worry about when contemplating the
possible effects of large-scale capital inflows? Any shopping list, influenced not
only by recent events, but also by the experience of Latin America in the early
1980s, must include the following:
Real exchange rate appreciation, which might be especially costly if it has
to be reversed under conditions of hysteresis.
Loss of control over monetary aggregates in cases where a fixed exchange
rate policy is pursued.
Volatility of flows, which may cause unexpected and sharp required swings
in the current account.
Intermediation of the flows through the domestic banking system, which
might give rise to a spurt of high-risk lending and sow the seeds of an
eventual banking crisis.
Of course, many well-known caveats are applicable. Real exchange rate
appreciation may well be an equilibrium phenomenon (a response to reform-
induced productivity changes, for instance) and not something for policy makers
to agonize over; loss of control of the money supply may be a good thing, in
that it brings with it monetary discipline; some of the consequences of volatility
are hedged in today's financial markets; potential banking instability may be
the result of poor regulation or outright fraud, not the capital inflows them-
selves. Nonetheless, the recent experience, not just of Mexico, but of countries
such as Argentina, Brazil and the Philippines, suggests that these concerns may
be real indeed.
Policy makers' worries can be tackled by an array of possible measures.
Choices made in the following areas are particularly important in determining
how the capital inflows will affect the macroeconomy:
The exchange rate regime, and the choice among fixing, flexing, crawling,
or some combination of these.
The extent of sterilization, and the effects sterilization has both on the
exchange rate and on fiscal accounts.
Other supporting measures, especially fiscal policy, which can be especially
important in limiting real appreciation and moderating the external deficit.
Capital controls of one kind or another.
In what follows we review the Chilean and Mexican experience in the light
of these policy issues and the policy responses they elicited. The next section
is largely descriptive, in that it summarizes the macroeconomic environment
in each country and the size and nature of the capital inflows it received. The
following three sections deal, in a more analytical vein, with policy options in
the areas of exchange rates, money, debt management, bank behavior and
capital controls. A final section concludes and offers some policy suggestions.
26 A. Velasco and P. Cabezas
Chile and Mexico were often seen from the outside as 'twin' economies: two
highly market-oriented economies that had liberalized internal and external
markets, both went through successful stabilization, deregulation and privatiza-
tion of state owned enterprises. These similarities, however, can obscure other
very fundamental differences. Some of these differences became extremely
important when the countries had to confront the capital inflow, and later
outflow, of the 1990s.
A first contrast has to do with the timing and maturity of the reforms. While
Chile implemented a first wave of structural reforms (trade and fiscal) in the
mid-1970s and a second (further privatization and deregulation) in the
mid-1980s, Mexico only started its reforms by 1983 and some of the most
important ones (deregulation, trade) were in fact delayed until after 1987. This
meant that by the 1990s most of the initial cost of the reforms were for Chile
a thing of the past, and the country was poised to begin enjoying their growth
and efficiency benefits. In Mexico, by contrast, the painful part of the process
is still under way, and growth has been very slow in coming. The numbers are
telling: while Chile enjoyed an average rate of growth of 7.1 % from 1988 to
1994, Mexico displayed a barely acceptable average of 2.8% per annum for
the same period (Table 2). Chilean economic growth was also reflected in an
enormous increase in employment. Unemployment rates came down from an
average of 22% for 1980-1985 to a comfortable 4% during 1993 and 6.2%
during 1994. Mexico's unemployment figures are notoriously unreliable, but
there is evidence of substantial unemployment and especially underemployment
during the period. There was greater similarity between the two countries in
the trajectory of wages. Together with the decrease in the unemployment rate,
in Chile real wages showed a steady recovery, increasing 24% between 1988
and 1994, while in Mexico real wages increased by slightly more than 25%
(Table 3).
A second difference is that in Chile inflation stabilization was achieved in
the late 1970s, while in the early 1990s Mexico was still undergoing the effects
Chile Mexico
Source: IFS, IMF, Chilean 1992-1994 figures come from new revisions made by Chilean Central
Bank.
Mexico and Chile compared 27
Chile Mexico
1 Annual averages. If we compute December to December rates, Chile's inflation also achieved
single digits in 1994.
28 A. Velasco and P. Cabezas
almost balanced figures. The Chilean experience in terms of fiscal policy has
been similar. The participation of government consumption over GDP has
been stable around 10% since 1988. The Chilean government has shown a
steady surplus on its accounts, even though as a share of GDP this surplus
has decreased slightly in the last 3 years (Table 5).
Even greater differences between the two countries emerge from an analysis
of the external sector. Even though real exchange rates appreciated in the two
countries, the magnitudes of the appreciation differed. Use of the exchange rate
as a nominal anchor produced a real appreciation of 22.2% for the Mexican
peso (Table 6) from 1989 to 1994. In Chile, partially as the result of a deliberate
government policy to limit real revaluation, the Chilean peso appreciated by
only 8.8% between 1989 and 1994 (Table 6).
In terms of the balance of payments, the two countries presented quite
different results. In the 1990-94 Chilean case, current account deficits showed
a stable and decreasing trend, averaging 1.6% of GDP per year. That, coupled
with a big capital account surplus, averaging 6.8% of GDP, implied a massive
increase of international reserves held at the Central Bank: almost 8 billion
dollars in 1990-94, which amounted to 5.2% of GDP annually (Table 1). This
Chile
Fiscal deficit -0.2 4.8 3.5 2.5 3.0 2.2 2.2
Public consumption 10.4 9.9 9.8 9.5 9.4 9.7 9.3
Mexico
Fiscal deficit -10.3 -5 -2.8 -0.2 1.5 0.4 -0.9
Public consumption 8.6 8.5 8.4 9.0 10.1 10.8 NA
Chile Mexico
Chile"
Foreign direct investment 63.5 103.3 33.3 57.1 57.5
Medium and long run credit 9.4 23.3 22.2 25.0 31.5
Portfolio investment 12.5 26.7 20.4 41.1 27.4
Short run credit 30.2 63.3 44.4 30.4 9.6
Others -16.7 -113.3 -18.5 -53.6 -26.0
Total 100.0 100.0 100.0 100.0 100.0
Mexico b
Foreign direct investment 11.5 56.7 29.6 25.0 32.9
Medium and long run credit 28.1 90.0 35.2 41.1 30.5
Portfolio investment -22.9 106.7 79.6 94.6 8.2
Short run credit 12.5 40.0 33.3 3.6 1.4
Others 7.3 -10.0 -27.8 -5.4 -2.7
Total 100.0 100.0 100.0 100.0 100.0
increase in reserves posed a big problem for the conduct of monetary policy,
as we will see later. The Mexican experience is rather different. As a response
to the appreciated peso, and boosted by inflated expectations about future
growth fueled by entry into NAFTA, the current account began worsening in
the late 1980s, reaching an annual average of 6.1 % of GDP in 1990-94 and
peaking at 8.6% of GDP in 1994. Capital inflows therefore largely went to
finance this deficit, and the impact on international reserves was moderate:
between 1987 and the reserve peak of December 1993, reserves grew by an
annual average of 0.7% of GDP (Table 1).
In terms of the composition of net capital inflows, Chile and Mexico show
yet another important difference. For Mexico (Table 7), the principal compo-
nent of the capital in flow was short term: in the strong capital inflow period
of 1990-93 short-term flows - portfolio investment and loans under a year of
maturity - accounted for approximately 55% of the total, with the remaining
45% being long term (FDI and long-term loans). For Chile, the equivalent
breakdown is 45% short-term and 55% long-term (Table 7).
Finally, Chile also outperforms Mexico in terms of indicators of long-term
growth potential. Throughout the period 1989-1994, Chile's investment rate
surpassed Mexico's by an average of more than 3% of GDP a year. Differences
in savings were even more dramatic: Chilean savings rates were higher by an
average of 7% of GDP a year (see Table 8). Although under perfect capital
flows a temporarily low savings rate need not be a problem for economic
growth - as long as the country can finance its desired increase in the capital
stock in the international credit market - a different story arises once inter-
national capital dries out. In an environment with no external funding, high
30 A. Velasco and P. Cabezas
Chile
Current account -2.5 -1.8 0.3 -1.7 -4.6 0.0
Investment 25.5 26.3 24.5 26.8 28.8 26.8
Savings 23.0 24.5 24.8 25.1 24.2 26.8
Mexico
Current account -2.8 -3.0 -5.1 -7.3 -6.4 -8.6
Investment 22.2 22.8 23.4 24.4 23.2 23.5
Savings 19.4 19.8 18.3 17.1 16.8 14.9
domestic savings rates are crucial to maintain the pace of investment, and in
that sense they represent a key requirement for a stable growth path.
Policy makers in Chile and in Mexico have faced some standard tradeoffs in
the design of an optimal exchange rate regime. A fixed or semi-fixed rate is
effective in bringing down inflation and (almost by definition) in minimizing
short-term nominal fluctuations; as a result, both countries relied on fixing in
the early stages of their stabilization processes. Over the longer haul, however,
greater degrees of flexibility may be desirable in dealing with capital movements
and in achieving some small degree of monetary independence. As a conse-
quence, in the late 1980s and early 1980s both moved toward mixed regimes
of crawling pegs plus bands. The similarities were sufficient to lead observers
to hail Chile and Mexico (along with Israel) as having devised a novel and
successful exchange rate regime (Helpman et al., 1994). Again, these apparent
similarities were offset by seldom stressed but nonetheless important differences.
After describing the regime policies implemented by each country, this section
speculates on how these differences may have determined macro performances
in each country.
Since 1983 Chilean exchange rate policy has explicitly attempted to target
the real exchange rate. For this purpose, the 'reference' nominal exchange rate
is adjusted periodically by the Central Bank to reflect the difference between
domestic and international inflation. In addition, in the early days the exchange
rate could float in a band of plus or minus 5% with respect to the reference
level. Such a system was adopted after an important real devaluation, and
worked well for the remainder of the 1980s: it offered the export sector a
competitive and reasonably predictable rate of exchange.
nounced peg system: the peso/dollar rate was allowed to depreciate by 1 peso
per day in 1989, 80 cents per day in 1990 and 40 cents per day in 1991. Finally,
during November 1991 the system evolved to a crawling peg with adjustable
bands. The moving band system involved a lower bound fixed at a level of
3.05 new pesos per dollar, while the upper band was devalued at a rate of
0.0004 new pesos per day. Hence, the band was designed to widen slowly over
time: by September 1994 the band was (plus or minus) 6% wide around a
central parity of 3.2438 pesos per dollar. Much as in Chile, Mexican authorities
also intervened within the band. Unlike Chile, this was undertaken to minimize
fluctuations in the politically delicate period of NAFTA negotiation and ratifi-
cation. Dirty floating produced an increase in the level of international reserves,
from 6.5 billion in 1989 to 25.4 billion by the end of 1993.
Over time, real appreciation became an increasing worry. Between 1990 and
1993 the nominal exchange rate depreciated by 17%, but because of the big
disparity between domestic and international inflation the real exchange rate
appreciated by 25-35%, depending on the actual index used. By early 1994
Mexican inflation had declined sufficiently for the extent of overvaluation to
have stabilized. Moreover, in March 1994 the nominal exchange rate experi-
enced a depreciation of 10%, and the US dollar itself was depreciating in real
terms against the European currencies and the Japanese yen, so that Mexico's
multilateral real exchange rate was less appreciated than its bilateral real rate
vis-a-vis the US dollar. Regardless of any of this, however, concerns over the
unsustainability of the overvalued rate (and the huge external deficits that it
had helped create) were to playa major role in making investors skittish and
in prompting the December 1994 panic.
The choice of exchange rate regime is often described as a tradeoff between
flexibility and credibility. The description is particularly revealing in the case
of Mexico and Chile. Think of possible regimes as a continuum with totally
fixed rates at one end, crawling and adjustable pegs and target zones somewhere
in between, and clean floating at the other end. The more fixed the rate, the
argument goes, the greater the commitment against inflation and therefore the
greater the credibility3; the more flexible the rate, on the other hand, the greater
the ability of the government to offset shocks.
Chile's crawling peg has emphasized flexibility at the expense of credibility.
There is a degree of automatic accommodation of domestic inflation, given
that the central parity is readjusted on a daily basis to offset inflation differenti-
als. Moreover, the authorities have not been shy to move the central parity
(and thereby the edges of the band) in a surprise manner in order to accommo-
date perceived changes in market conditions. In particular, as evidence has
mounted that an important share of the capital inflow is in fact long term
(DFI and related categories), real appreciation has been allowed for. Mexico's
3 The standard caveat is that credibility could also be attached to a montary rule under flexible
exchange rates. Arguing that credibility is more easily obtained under fixing requires (a) the
assumption that there are large political costs attached to the abandonment of a fixed parity or
(b) the belief that monetary targets are too costly to observe by agents.
Mexico and Chile compared 33
regime, by contrast, has put the emphasis on credibility. The edges of the band
are predetermined rather than responding endogenously to inflation differenti-
als. And when facing pressure at one of the bounds, as happened in March
1994 after the assassination of Luis Donaldo Colosio, the Banco de Mexico
has been unwilling to move the band in order to curtail reserve changes.
Mexico's stress on credibility at first seemed to payoff. Inflation rates
converged relatively quickly, edging toward dollar inflation in 1994. Chile, by
contrast, spent over a decade (from the early 1980s to the early 1990s) with
inflation in the 20-30% range, until attaining single-digit inflation for the first
time in 30 years in 1994. In addition, there were serious fears that Chile lacked
a nominal anchor (monetary policy has aimed at targeting real interest rates),
so that an exogenous shock could seriously de-stabilize the price level.
However, it was not Chile's lack of monetary and exchange rate discipline,
but Mexico's lack of exchange rate flexibility, that has proven costly over the
longer haul. The real test for Mexico came with the shocks of 1994: first the
increase in US interest rates and then the political turmoil caused by peasant
uprisings and political assassinations. In the days that followed the March
assassination the exchange rate went all the way to the top, in what constituted
a nominal devaluation of around 10%.4 The exchange rate spent the rest of
the year at or very near the ceiling. Both marginal and inframarginal interven-
tion led to large reserves losses. The upshot is that between and March and
December Mexico operated an essentially pegged exchange rate, in that only
the top of the band was relevant. 5
Having allowed the real exchange rate to become overvalued in previous
years, Mexico's unwillingness to lift the ceiling of the band in response to the
shocks in March and latter in the year was to prove costly. Opponents of
moving the band ceiling emphasized the possible credibility costs: a devaluation
might lead investors not just to doubt the commitment to low inflation, but
also to revise their assessments about the commitment to property rights of
the reformers in power. In the end the issue was moot. The combination of
loose money (in the course of 1994) and a fixed rate allowed reserves to
dwindle, making it increasingly likely that the peg could not be sustained
regardless of policy makers' wishes. When investors ran on Mexican assets at
the end of the year, there was little the government could do but let the
exchange rate float.
At least since David Hume, economists have understood that fixed exchange
rates require that the money supply be mainly determined by the balance of
4 The fact that in the same period the Central Bank also lost around $9 billion in reserves only
on overall domestic public debt, however, was different to that in Chile: because
of the fall in real interest rates in the early 1990s and the amortization of
substantial amounts (often using privatization proceeds), internal public debt
as a share of GDP decreased from almost 28% in 1988 to 12% in 1993.
What about broad sterilization? Here the policy course of both countries
also diverged. Chile, in keeping with its more activist reputation, increased
reserve requirements on foreign currency deposits held by commercial banks
by 50%, using a structure such that the required reserve decreased as the rate
as the maturity of the deposit increased. Together with this increase in reserve
requirements, the Central Bank introduced a 30% marginal reserve requirement
on interbank deposits. Mexico, by contrast, engaged in reverse broad steriliza-
tion: a zero legal reserve requirement for banks was instituted, sharply increas-
ing the money multiplier.
Sterilization is not without perils. Start with the broad kind: increases in
reserve requirements for banks can gradually cause a process of disintermedia-
tion, with the consequent loss in efficiency. Matters are more problematic in
the case of narrow sterilization, for here the list of possible ills is long:
It may increase interest rates (relative to what they would have been with a
capital inflow but without sterilization) and hence perpetuate the inflow. 6
It may deteriorate fiscal and quasi-fiscal accounts, for the domestic bonds
typically used to sterilize carry higher interest rates than is earned by reserve
holdings.
Also in the fiscal realm, it may dangerously increase the stock of domestic
debt.
Of these, the most serious problems are the fiscal ones. Calvo (1991), an
early source of concern on the issue, has even claimed that these policies, aimed
at curtailing monetization and enhancing the credibility of the anti-inflation
stance, may actually increase expected and realized inflation. This would occur
if a perverse monetarist arithmetic is at work, so that the large increase in debt
leads to expectations that it will be eventually monetized. But without going
to such an extreme, the adverse fiscal impact of sterilization was clearly felt in
Chile and Mexico. In Chile, preliminary estimations suggest that in 1990-93
the Central Bank incurred annual losses of about 0.5% of GDP because of the
interest rate differential between domestic bonds issued by the bank and interna-
tional reserves. In Mexico the same loss averaged an annual 0.25% of GDP
over the same period.
6 Frankel (1994) has sorted out the conditions under which this is so.
36 A. Velasco and P. Cabezas
chosen to carry out his sterilization, both in terms of their denomination and
maturity.
Mexico used an array of instruments, induding the peso-denominated cetes
and Bondes, the inflation-indexed ajustabonos and the dollar-indexed tesobonos.
Until 1993, however, it was the peso assets that predominated, accounting
roughly for three-quarters of domestic government interest-earning liabilities.
Chile, by contrast, relied heavily on indexed debt: all liabilities over 30 days in
maturity were (and are still) denominated in UF, an inflation-linked unit of
account. The structure of maturities was also different. Early in the decade
Mexico relied rather heavily on short-term cetes, perhaps gambling that, as the
stabilization process was consolidated rates would come down, which made it
fiscally unadvisable to borrow long. By the end of 1993, the average maturity
of Mexican internal public debt was 290 days (see Table 11), and short term
debt (of less than a year of maturity) accounted for 60% of total domestic debt
(Table 9).
The Chilean experience was quite different. When at the end of 1989 the
newly independent Central Bank launched its tight money policy, it attempted
to establish credibility as swiftly as possible. One policy chosen was to offer
10-year indexed bonds with a high real interest rate. Throughout the first year
of heavy sterilization (1990) the Central Bank kept on issuing long term debt.
The policy was abandoned at the end of the year. During 1991 short term
bonds (with maturities of less than a year) were used to sterilize, once again
reflecting expectations about the future course of interest rates. Between 1992
and 1994, the average maturity of the sterilization bonds was increased steadily,
showing an underlying preference to balance the average maturity of the central
bank's stock of debt around a target of about two years (see Table 11). As a
consequence of this maturity management, from 1990 to 1994 Chilean short
term stock of domestic debt was kept on average at 33% of total liabilities -
equivalent to 31 % of international reserves (see Tables 9 and 10).
In terms of macroeconomic performance, these differences in debt manage-
ment between the two countries were to prove important. It is likely that the
presence of a large stock of non-indexed debt kept alive in investors the fear
Chile"
Short -term debt b 18.0 24.0 53.0 34.0 35.0
Long-term debt 82.0 76.0 47.0 66.0 65.0
Mexico
Short -term debt" 45.3 43.0 45.2 62.6 78.4
Long-term debt 54.7 57.0 54.8 34.7 21.6
Table 10. Short-term internal public debt, 1994 (share of total reserves)
January 25 104
February 22 93
March 23 106
April 22 141
May 20 157
June 21 173
July 21 182
August 22 180
September 22 183
October 24 172
November 29 280
December 29 410
Table 11. Average maturity of new debt subscribed by the central bank (in days; 1994)
Chile Mexico
7 Calvo (1988) offers a model that emphasizes this point, and which shows that non-indexed
During 1994, however, three things happened that were to set Mexico far
apart from Chile. First, total Mexican government short term domestic debt,
regardless of currency denomination, grew both in absolute magnitude and as
a multiple of reserves. Expressed in dollars, domestic debt amounted to 1.7
times reserves in December 1993 and 2.6 times reserves in September 1994.
Second, and more important, the average maturity of Mexican domestic debt
shrank drastically during 1994. This was the result of a deliberate policy choice.
With the increased turmoil in 1994, the yield curve turned steeper, and issuing
long-term debt became increasingly expensive. Conjecturing that the shock was
transitory, the Mexican government followed the correct policy of borrowing
short in order to get over the hump (until the end of the year, say) without
wrecking public finances. Ex post, we know that this strategy had two weak-
nesses: (a) the shock could turn out not to be transitory, in which case a real
fiscal adjustment would have been needed to compensate for the higher interest
payment burden and (b) the shorter maturities rendered the government largely
defenseless against any circumstance in which investors refused to roll over
their government bonds. Third, after the March assassination of leading presi-
dential candidate Colosio and the accompanying perceived increase in devalua-
tion risk, the Mexican government began rolling over its short-term peso-
denominated debt into short-term dollar-indexed debt. Starting at $1 billion
at the beginning of the year, by the end of September (before the last great
decline in Central Bank foreign assets), the stock of tesobonos outstanding
equaled the amount of reserves. In December the stock of tesobonos reached
$18 billion. This large stock of short-term public debt created, as Calvo (1994)
stressed, an important source of financial fragility.
Such fragility was soon to prove lethal. The December 20th devaluation of
the Mexican peso provoked an investor panic. As the tesobonos matured,
investors were unwilling to roll them over. Several bond auctions found no
takers. Mexico was on the brink of default.
When confronted with an inflow (and often concerned with a possible future
outflow) a number of countries have been tempted recently to resort to capital
controls. In this regard, Chile and Mexico stand as polar opposites in the range
of Latin American experiences. Chile has resorted to a battery of policies
intended to limit short-term inflows. Mexico, by contrast, has been largely
opposed to such moves, with the exception of a couple of (relatively mild)
policies intended to discourage interest arbitrage by Mexican banks.
In Chile the major policy initiative was the imposition in June 1991 of an
implicit tax on capital inflows. s The tax took the form of a non-interest-bearing
8 In addition, there was the long standing requirements that aU OFI stay in the country for at
least one year (which applies only to capital - profits can be repatriated immediately.
Mexico and Chile compared 39
deposit in the central bank equivalent to 20% of total investment. This reserve
requirement was to be maintained in the central bank for a period that varied
from 90 days to one year depending on the maturity of the loan. In addition,
a stamp tax of 1.2% a year, previously applied only to domestic currency loans,
was extended to foreign currency loans in operations of up to one year. In July
1991 an alternative to the reserve requirement was created: instead of maintain-
ing a cash reserve, borrowers were permitted pay up front the equivalent of
the financial cost of the reserve requirement. In May 1992, the reserve require-
ment was extended to 30% of the foreign loan, and the period required to be
held in the central bank was fixed to one year no matter the maturity of the loan.
The combination of reserve requirement (whether paid directly or indirectly)
and stamp tax imposed no marginal cost on medium and long term lending,
but it imposed a heavy financial cost to short term loans (less than one year).
Agosin and Ffrench-Davis (1995) calculate that for a foreign loan of one-year
maturity, the value of the implicit taxes have fluctuated from an annual 2.6%
in the second semester of 1992 to a high of 3.9% in 1994. The percentages are
substantially higher for loans of shorter maturities.
In Mexico, by contrast, the monetary authorities have continued to liberalize
the capital account. Since 1990, foreign investors have been allowed to invest
in government bonds and in shares in the financial system (not voting shares).
Foreign investors can invest in basically any project developed in Mexico.
Capital mobility is, in essence, unimpeded.
The only exception to the de-regulating trend came in 1992. Concerned over
the big exposure to exchange risk that the banking system was undertaking,
Mexican authorities imposed a minimum liquidity coefficient to foreign exchange
transactions: 15% of foreign exchange borrowing must held in liquid securities
denominated in the same currency. During 1992, a regulation that limited foreign
currency liabilities of commercial banks to 10% of their total loan portfolio was
approved. Both policies limit the scope and profitability of cross-border interest
arbitrage - borrowing abroad in dollars and lending at home in pesos, earning
the differential- by banks, which had been an important source of capital inflows.
But they could also be justified on prudential grounds as ways of diminishing the
currency risk exposure of deposit-taking institutions at home.
What are controls expected to do and do they do it? International economics
commonly assert that capital controls are not very effective because they can
(relatively easily) be evaded. It is regrettably much less common to be clear
what it is that controls are ineffective at. Any evaluation of the usefulness of
controls must start by specifying what it is that they are expected to accomplish.
At a most basic level, controls of the sort we have described have as a basic
aim to discourage speCUlative capital inflows. Somewhat more subtly, however,
this can be split into three distinct tasks:
Partially to sever the link between domestic and foreign interest rates,
creating a wedge the endows domestic monetary policy with a modicum of
independence .
To reduce the total size of the inflow, presumably to reduce the extent of
real currency overvaluation.
40 A. Velasco and P. Cabezas
To change the composition of the inflows (while leaving the size of the flow
pretty much unchanged), encouraging short-term (hot) flows to become
longer-term (cooler) flows.
How well have capital controls performed these three tasks? Chilean evidence
on the first point is mixed. On the one hand, Chile has been able to keep
nominal and real rates above the relevant US benchmark to a greater extent
than simple risk premia calculations would suggest. Laban and Larrain (1993)
present evidence suggesting that the 'offset coefficient' has declined since 1991
- precisely the year when capital controls were tightened - thereby endowing
the Chilean central bank with a greater degree of monetary independence. On
the other hand, the scope to exploit this degree of independence is severely
limited. There have been numerous episodes since 1990 where rates of interest
that were deemed 'excessively high' by agents brought forth a surge in inflows,
forcing as a counterpart the relaxation of the domestic monetary stance.
On the other two points (effects on the size of inflows and their composition)
the evidence is also controversial, but a consensus view increasingly holds that
controls have been quite effective at changing the composition of flows toward
longer term maturities, while relatively ineffective at reducing the overall level of
the flow. 9 Econometric evidence from Chile presented by So to (1995), for instance,
suggests that a combination of evasion and substitution toward longer maturities
has meant that, after controlling for other determinants of flows, there is little
discernible impact of the tax has left on overall volumes of funds entering the
country. The composition of the flows, however, seems sharply responsive to the
tax. Similar evidence is provided for the Colombian case by Cardenas and
Barrera (1995).
Does this mean that controls should be discarded because they are incapable
of keeping capital out? Not necessarily: affecting only the composition of flows
may be just what policy makers need. Sachs et al. (1996b), in a study of 20
emerging markets in the aftermath of the Tequila shock, find the degree of
financial and currency vulnerability is not systematically correlated with the
size of the previous capital inflow: many Asian countries, along with Chile and
Colombia, absorbed large amounts of foreign capital in 1989-94 and felt no
effects from Tequila. It is, however, correlated with the composition of such
flows: the larger the share of short-term flows in the total, other things equal,
the greater the disarray in the local financial markets in the first half of 1995.10
The different behavior of local banks in 1989-1994 offers the last piece of
evidence on why Mexico found itself in financial disarray in late 1994 and
9 See Ffrench Davis and Agostn (1995) and Soto (1995) for differing views on these issues.
10 Technically, that study measures financial and currency pressures by means of an index that
is a weighted average of nominal exchange rate depreciation and reserve losses. The behaviour of
local stock markets in early 1995, not reported in that paper, also fits this pattern of correlation.
11 This section dra,,:s on Sachs et al. (1996b).
Mexico and Chile compared 41
early 1995, while Chile did not. Banking and currency difficulties often go hand
in hand: the link has been present in crises ranging from that of the US in the
1930s to that of Chile in the early 1980sY But the sheer magnitude of both
bank and currency crises in the recent case of Mexico, and also in Venezuela
and Argentina to cite just two additional examples, has brought back the issue
with a vengeance (see Kaminsky and Reinhart (1995) for a detailed analysis of
a number of such episodes).
Theoretically, the link between these 'twin crises' is not hard to ascertain.
Abrupt changes in the demand for money (caused, for instance, by expectations
of devaluation and an incipient speculative attack) can cause a sharp fall in
bank deposits. Under a fractional reserve system, however, banks do not have
the cash in hand; in the absence of an injection of liquidity from the outside
(typically from the Central Bank), cessation of payments and a bank panic can
readily occur. Even if banks could simply wait until loans mature in order to
satisfy depositor's demands (something that would take time, given banks'
essential role as maturity transformers), the ensuing adjustment would not be
easy or painless. The resulting credit squeeze on borrowing firms would send
interest rates sky-high. In emerging markets banks are the main sources of
corporate credit, and most firms cannot simply turn around and borrow from
the world market, no matter how de-regulated the capital account may beY
The need to avoid a wave of bankruptcies and serious economic disruption
provides yet another rationale for the authorities to step in.
One upshot is that the monetary base is not the only claim on the Central
Bank that can be called in at times of trouble. The reality is that, with bank
liabilities covered by implicit or explicit government guarantees, all M2 is
potentially a liability of the government, to be redeemed with dollars at the
time of a crunch. A second upshot is that links to the financial system make it
even more likely that self-fulfilling runs against a currency will succeed, as
Obstfeld (1994) has stressed. If fears of devaluation prompt a bank run which
in turn causes an expansion of liquidity with which Central Bank reserves can
be bought, the circle is fully closed. And, as the experience of Argentina recently
suggests, the problem is not eliminated by the adoption of a rigid peg or
currency board system, where the Central Bank is prevented from issuing
domestic credit. Sachs et al. (1996a) argue that in fact the problem may be
worsened, for the absence of a lender of last resort can magnify fears of bank
illiquidity and turn them into problems of insolvency.
The discussion so far implies that such a vicious cycle can affect all banking
systems. But banks, like Orwell's animals, are not all equal. The kind of
vulnerability that we have been describing depends crucially on two features
12 Wigmore (1987) has argued that the failure of the Fed to protect the US banking system in
the winter of 1932-33 was the result of the Fed's fears that lender of last resort credit to the banks
would undermine the link of the US dollar to gold. In Chile in 1982, high interest rates under a
fixed exchange rate helped precipitate a banking collapse; the associated expansion of domestic
credit helped precipitate the end of the exchange rate peg. See also Velasco (1991) for details.
13 Calvo (1996) stresses this point.
42 A. Velasco and P. Cabezas
Argentina 23 38
Brazil 21 51
Chile 8 -12
Mexico 207 361
of a banking system. First, the size of the liquidity cushion: the greater are the
ratios of reserves and of bank capital to deposits, the larger the likelihood a
bank can withstand a shock. Second, the quality of its portfolio: weak borrowers
mean weak banks, because a small tremor in interest rates can lead to a large
share of non-performing loans. Both are of course related: a growing share of
doubtful credits eats into capital and reduces the cushion available to cover
additional shocks.
What determines bank weakness? Portfolios can be made suddenly weak
by an exogenous shock - think of oil prices and Texas banks in the mid-1980s-
but bad luck is not the only cUlprit. Portfolios are more often than not
endogenously made weak by swift expansions of credit, with boom leading to
bust. As Hausmann and Gavin (1995) persuasively argue, the empirical link
between lending booms and financial crises is very strong. Rapid growth in the
ratio of bank credit to GDP preceded financial troubles in Argentina (1981),
Chile (1981-82), Colombia (1982-83), Uruguay (1982), Norway (1987),
Finland (1991-92), Japan (1992-93), and Sweden (1991). Their data even
shows a lending boom in the US at the time of the S&L crisis. The rationale
for this link is simple. When lending expands very quickly, lenders' ability to
screen marginal projects declines, and they are more likely to end up with a
large share of weak borrowers in their portfolios. High risk areas, such as
credit cards and consumer and real estate loans, tend to grow more than
proportionately. In addition, regulators (particularly in developing countries)
soon find their limited oversight capacity overwhelmed. Thus, a bank portfolio
that is extremely vulnerable to the vagaries of the business cycle is the most
likely product of a lending boom.
Table 12 shows the evolution of the claims by the domestic financial system
on the domestic private sector (as a share of GDP) for four selected Latin
American countries. 14 If there is one single indicator that sets Chile and Mexico
14 This variable was constructed in the following way. For each year we calculated the ratio of
claims on the private sector by deposit money banks and monetary authorities (lFS line 32d) to
GDP (IFS line 99b). When inflation is high this ratio is biased upward because the available
annual figure for claims on the private sector corresponds to the figure for December, while
nominal GDP reflects the average price level for the entire year. To correct for this bias we
multiplied the biased ratio by the ratio of the average price level to December's price level. When
inflation is low this factor is basically one.
Mexico and Chile compared 43
dramatically apart, this is it. Between 1989 and 1994, bank credit to the private
sector grew by 207% in Mexico, while it grew by only 8% in Chile.
Two caveats concerning lending booms are in order. A first one is that it is
extremely important to distinguish levels from rates of increase. Many successful
developing countries show very high ratios of private sector credit to output.
This is of course nothing but financial deepening, and in it of itself not
something to be concerned about. What is worrisome is the occurrence of
sharp increases in lending to the private sector within a short period of time,
which are likely to lower average loan quality. Another caveat is that lending
activity is likely to be closely related to the stabilization cycle. On the other
side of the bank balance sheet are the deposits that finance the loans. Deposits
are highly correlated with money demand, and therefore with expected inflation.
When a policy turnaround puts an end to hyperinflation, deposits swell and
so do bank loans. This effect, which is nothing but a beneficial payoff (greater
financial intermediation) from stabilization, probably explains some of the
sharp increase in lending in a country like Argentina. It does not, however,
explain an episode like Mexico's, in which stabilization occurred in 1988-89
and the increase in credit in 1991-94.
Why did a lending boom happen in Mexico (and to some extent in Argentina)
and not in Chile? A commonly cited culprit is swift liberalization of the capital
account, followed by a surge in inflows which presumably get intermediated
by the banking sector. The are two problems with this explanation. The first
one is that both countries have had reasonably open capital accounts for a
long time. The second is that there is no obvious correlation between the size
of the capital inflow and the ensuing behavior of bank credit. As we saw above,
both countries experienced similarly large capital inflows, and bank lending
grew tremendously in Mexico but not in Chile. 1s
But if capital account liberalization does not seem to have mattered, the
same is not true of domestic financial liberalization. Already in the late 1970s
and early 1980s the cycle from bank privatization and deregulation to lending
boom to eventual bust was clearly visible in a number of countries. In Latin
America, Argentina, Chile and Colombia, plus Uruguay, went through the
cycle. 16 The point, of course, is not that deregulation is bad per se, but that it
can lead to rapid credit expansion. For instance, financial liberalization in
1988-90 was followed by a lending boom in Indonesia. In Mexico, privatization
and deregulation of the banking system in the early 1990s had a similar effect.
A key difference is that in Mexico the capital inflow took place simulta-
neously with a decrease in banks' required reserve ratios, while in Chile such
ratios were increased (at least for dollar deposits). As Rodriguez (1993) has
stressed, this can set the stage for the inflow to cause a large increase in
domestic bank lending. The stance of bank supervisors probably also made a
ISSachs et al. (1996b) also find no such correlation for their sample of 20 developing countries.
16 See Baliilo and Sundarajan (1991) for studies from a set of countries, including Argentina,
Chile, Uruguay and the Philippines.
44 A. Velasco and P. Cabezas
17 See Rojas-Suarez and Weisbrod (1995) for evidence on the recent Argentine and Mexican
banking troubles and the just-launched restructuring packages.
Mexico and Chile compared 45
would agree that, ifthe capital inflow and therefore the change in the 'fundamen-
tal' real exchange rate are more or less permanent, then in the long run real
exchange rate targeting cannot succeed: repeated nominal devaluations would
simply elicit repeated increases in prices, failing to affect the real exchange rate.
As usual, definitions of what constitutes 'the long run' vary widely. If there is
enough price stickiness over plausibly short periods, and if capital inflows are
also short-lived, so that a brief period is all that is at stake, then nominal
exchange rate policy may well have some ability to prevent real appreciation.
But this policy is not costless: in the 1990s Chile and Colombia (both of which
have targeted the real rate) had higher inflation than the other Latin American
countries, in spite of virtuous fiscal policies. In short, an accommodating
nominal exchange rate policy may be able to limit real appreciation over the
short-to-medium run, though probably at some expense in terms of inflation.
Capital controls
Most economists are understandably weary of capital controls, for the state
can easily put up an iron curtain of costly and inefficient regulations. But such
skepticism must confront the fact that several of the more successful developing
countries, such as Korea, Malaysia and Indonesia in Asia, Colombia and Chile
in Latin America, have on occasion created disincentives to short term inflows.
And, most important, it was precisely these emerging markets that passed the
Tequila shock with the smallest hangovers.
46 A. Velasco and P. Cabezas
How does one ensure that such controls do more good than harm? The
experience of the Chiles and Malaysias of the world suggests two lessons. First,
get the cosmetics right: liberalize outflows as you restrict inflows, in order to
make sure that investors do not come to fear a return to the populist policies
of yesteryear. Second, do not expect controls to do more than they can: the
amount of monetary independence a Tobin tax affords is, almost by definition,
limited by the size of the tax wedge. Attempts at fixing unrealistically high
interest rates (relative to world rates) can only cause, as Chile has occasionally
found out, additional large and destabilizing inflows.
ACKNOWLEDGMENTS
Support from the c.V. Starr Center for Applied Economics at NYU is gratefully
acknowledged.
REFERENCES
Alesina, A., A. Pratti and Tabellini (1990). Public confidence and debt management: a model and a
case study of Italy. In Dornbusch and Draghi (eds.), Public Debt Management: Theory and
History.
Balino, Tomas and Sundarajan Vasudevan (1991). Banking Crises: Cases and Issues. Washington
DC: International Monetary Fund.
Calvo, G. (1988). Servicing the public debt: the role of epectations. American Economic Review.
Calvo, G. (1991). Temporary stabilization policy: the case of flexible prices and exchange rates.
Journal of Economic Dynamics and Control, IS( 1).
Mexico and Chile compared 47
Calvo, G. (1996). Capital Flows and Macroeconomic Management: Tequila Lessons. Mimeo,
University of Maryland.
Calvo, G. (1994). Comment on Dornbusch and Werner. Brookings Papers on Economic Activity, 1.
Calvo, G. (1995). Varieties of Capital Market Crises. Mimeo.
Calvo, G. and Reinhart Leiderman (1993). The Capital Inflows Problem: Concepts and Issues.
Mimeo.
Calvo, G. and P. Guidotti (1990). Indexation and maturity of government bonds: an exploratory
model. In Darnbusch and Draghi (eds.), Public Debt Management Theory and History.
Cardenas, M. and F. Barrera (1995). On the Effectiveness of Capital Controls in Colombia. Mimeo,
Fedesarrollo.
Cole, H. and P. Kehoe (1996). Reputation Spillover Across Relationships: Reviving Reputation
Models of Debt. NBER Working Paper 5486.
Dornbusch, R. and A. Werner (1994). Stabilization, reform and no growth. Brookings Papers on
Economic Activity, 1.
Drazen, A. and Masson (1994). Credibility of policies and credibility of policy makers. Quarterly
Journal of Economics.
Ffrench Davis, Ricardo and Manuel Agosin (1995). Managing Capital Inflows in Latin America.
Mimeo, University of Chile.
Frankel, 1. and C. Okongwu (1995). Have Latin America and Asian countries so liberalized portfolio
capital inflows that sterilization is now impossible?
Frankel, 1. (1994). Sterilization of Capital Inflows: Hard (Calvo) or Easy (Reisen)? Mimeo.
Hausmann, Ricardo and Gavin Michael (1995). The Roots of Banking Crises: The Macroeconomic
Context. Inter-American Development Bank. Prepared for the Conference on Banking Crises in
Latin America, October 1995, Washington, DC.
Helpman, E., L. Leiderman and G. Bufman (1994). A new breed of exchange rate bands: Chile,
Israel and Mexico. Economic Policy, 19.
Kaminsky, Graciela and Carmen Reinhart (1995). The Twin Crises: The Causes of Banking and
Balance-of-Payments Problems. Working paper No. 544, International Finance Paper, Board of
Governors of the Federal Reserve System.
Laban, R. and F. Larrain (1993). The Chilean experience with capital mobility. In Bosworth,
Dornbusch, Laban (eds.), The Chilean Economy, Policy Lessons and Challenges.
Obstfeld, M. (1993). The logic of currency crises. CaMers Economiques et Monetaires, 43.
Rodriguez, C. (1993). Money and credit under currency substitution. IMF Staff Papers, 40(2).
Rojas-Suarez, Liliana and Steven R. Weisbrod (1995a). Managing Banking Crises in Latin America:
The Do's and Dont's of Successful Bank Restructuring Programs. Inter-American Development
Bank. Prepared for the Conference on Banking Crises in Latin America, October 1995,
Washington, DC.
Rojas-Suarez, Liliana and Steven R. Weisbrod (1995b). Banking Crises in Latin America: Experience
and Issues. Inter-American Development Bank. Prepared for the Conference on Banking Crises
in Latin America, October 1995, Washington, DC.
Rojas-Suarez, Liliana and Steven R. Weisbrod (1995c). Achieving Stability in Latin American
Financial Markets in the Presence of Volatile Capital Flows. Inter-American Development Bank.
Working Paper Series 304 (4).
Sachs 1., Tornell and Velasco (1996a). The collapse of the Mexican peso: what have we learned?
Economic Policy, 22.
Sachs 1., Tornell and Velasco (1996b). Financial crises in emerging markets: the lessons from 1995.
Brookings Papers on Economic Activity, 1.
Soto, M. (1995). Encaje a los Creditos Externos: La Evidencia Empl'rica. Mimeo, Catholic University
of Chile.
Velasco, A. (1987). Financial crises and balance of payments crises. Journal of Development
Economics.
Velasco, A. (1991). The Chilean financial system: liberalization, crisis, intervention. In Baliiio, T.
and Sundarajam, V. (eds.), Banking Crises: Cases and Issues. Washington, DC: International
Monetary Fund.
Wigmore, B. (1987). Was the banking holiday of 1933 a run on the dollar rather than the banks?
Journal of Economic History, 47,739-56.
BARRY EICHENGREEN
University of California at Berkeley
INTRODUCTION
Foreign lending to sovereigns and companies has a long and colorful history.
This fact is both easy to remember and easy to forget. Each time difficulties
arise in international capital markets, observers invoke the precedent of history.
Each debt crisis occasions the publication of scholarly studies citing parallels
with debt crises past, making historical studies like this one a cottage industry.2
At the same time, the repetition of events suggests that each additional lending
burst reflects decisions by investors who, if they are not ignorant of history,
fail to use it to inform their actions. In the age of 24-hour trading, the typical
investor complains, who has time to reflect on history books? The relevance
of history is limited, moreover, by changes in the structure of international
financial intermediation, the role played by bond markets in the 1920s and
bank finance in the 1970s having given way to equity finance, fundamentally
altering the lending process.
As is predictably the case when sharp distinctions are drawn, reality lies
between the extremes. Important changes have indeed occurred over the course
of the century in the structure and organization of lending. The most prominent,
no doubt, is the evolution of intermediation: the rise and retreat of the New
York market in international bonds in the 1920s and 1930s, the triumph and
tragedy of bank lending in the 1970s and 1980s, and the growth of emerging
equity markets as a vehicle for capital transfer in the last 10 years. One might
also point to the establishment and evolution of the International Monetary
Fund and Group of Ten as mechanisms for crisis management.
These changes should not be allowed to obscure important elements of
continuity unifying the century of historical experience. Throughout the period
lending has been shaped not just by economic and political conditions in the
debtor countries but by monetary policies in the creditors' markets. Large-
J Prepared for the NYU Conference on Emerging Market Capital Flows, May 23-24, 1996.
The author is John L. Simpson Professor of Economics and Political Science at the University of
California, Berkeley, Research Associate of the National Bureau of Economic Research, and
Research Fellow of the Centre for Economic Policy Research. The present paper draws on work
with Albert Fishlow and Richard Portes (Eichengreen and Fishlow, 1995; Eichengreen and Portes,
1995), whose collaboration is acknowledged with thanks. Helpful comments were provided by
Michael Dooley and Philip Suttle.
2 The burst of lending in the 1920s and the debtservicing difficulties of the 1930s prompted the
publication of classic works by Feis (1930) and Lewis (1938). The lending boom of the 1970s and
the debt crisis of the 19805 had their counterpart in Edelstein (1981) and Fishlow (1985).
Presumably the same will be true of 1994-95 once the Mexican crisis is behind us.
49
R. Levich (ed.). Emerging Market Capital Flows. 49-74.
1998 Kluwer Academic Publishers.
50 B. Eichengreen
3 This fact was emphasized by Simon Kuznets, for whom shifts in foreign lending were a
prominent characteristic of the long swings in economic activity that subsequently came to be
known as Kuznets Cycles.
International lending in the long run 51
Table 1. Foreign dollar loans annually taken in the USA, by classes of borrowers, 1923-31 (in
millions of dollars). Aggregate loans and repayments
National and
provincial
government Municipal Corporate Total
Long term
1923 231.9 54.0 16.7 5.9 80.9 12.9 329.5 72.8
1924 676.9 57.7 66.7 2.0 165.0 43.9 908.6 103.6
1925 551.6 114.2 88.3 7.7 278.5 17.1 918.4 139.0
1926 436.7 105.5 92.8 17.8 354.6 37.0 884.1 160.3
1927 584.8 63.5 198.1 18.3 456.0 76.0 1,238.9 157.8
1928 486.3 256.1 111.7 43.4 567.1 105.0 1,165.1 404.5
1929 97.4 380.6 48.0 13.8 227.4 45.7 372.8 440.1
1930 432.7 120.1 63.8 29.4 260.7 136.3 757.2 285.8
1931 75.1 217.9 22.4 33.7 101.3 93.7 198.8 345.3
Short term
1923 90.7 68.5 1.0 10.3 2.0 7.1 93.7 85.9
1924 175.5 109.0 7.5 11.3 57.1 14.1 240.1 134.4
1925 152.4 264.4 0.6 9.5 70.4 16.4 223.4 290.3
1926 98.6 156.2 7.2 2.2 41.2 26.6 147.0 185.0
1927 75.7 135.3 8.9 5.2 46.6 61.4 131.2 201.9
1928 9.2 59.2 1.8 1.0 10.0 44.5 21.0 104.7
1929 7.5 13.2 2.4 15.2 40.8 22.7 56.4
1930 162.6 99.6 25.6 14.2 17.4 26.2 205.6 140.0
1931 36.6 37.0 24.7 7.5 12.1 44.1 73.8
Private
Official Direct
development private Export
Years Total assistance Total investment Portfolio credits
for direct investment, were completed only with delay. The post-war boom in
commodity prices encouraged direct investment in extractive industries and
processing capacity in Latin America and Asia and bond issues to finance
railways and port facilities linking resource-rich states and provinces to interna-
tional markets.
52 B. Eichengreen
Table 3. Aggregate net long-term resource flows to developing countries, 1989-94 (US$ billions)
- Not available.
Data provided in this table cover countries included in the World Bank's Debtor Reporting System
and non-DRS developing countries. Foreign direct investment includes reinvested profits. Grants
exclude technical cooperation grants. Officially guaranteed export credits are included under
private loans, and direct export credits under bilateral loans.
Projected.
Source: World Bank (1995), p. 7.
The rise in bank finance in the 1970s responded to economic growth and
liberalization in the developing world. Growth worldwide fluctuated from 4 to
6% in the 1960s, an impressive pace by historical standards. Aspiring borrowers
in Asia and Latin America shared fully in this prosperity. Indeed, they ranked
at the upper end of the growth leagues: growth in Latin America averaged
5.7% per annum in the 1960s, while growth in East Asia was even faster. This
contrasted with the more troubled 1950s, when growth had been less uniform
and persistent. Tariffs were reduced, albeit more gradually in Latin America
than East Asia. Export promotion was embraced in both regions, albeit with
less impressive results in Latin America. This record of trade and growth served
to attract international bankers to the developing world.
From the debt crisis of the 1980s Latin America emerged with a firmer grasp
on fiscal conditions and with inflation under growing control. Consumer price
inflation in Latin America, excluding Brazil, fell to 14% in 1994, down from
1400% in 1989. The ethos of liberalization encouraged the deregulation of
domestic markets, foreign trade and capital transactions. Tariffs were reduced,
encouraging trade-related inward investment. Privatization created new oppor-
tunities for financial capital. The economies of East Asia, for their part, survived
the debt crisis largely unscathed and moved up the technological ladder toward
International lending in the long run 53
~ ~----------------------------------------------~
(8)
10
-5
1974 1975 1976 1977 1971 1979 1910 1911 1912
12
II
10
2
1916 1987 1911 1919 1990 1991 1992 1993 1994
...-Nominal -+-Real
Figure 1. US Interest rates. (a) Prime lending rate, 1974--82; (b) prime lending rate, 1986-94.
International lending in the long run 55
foreign lending (Figure 1). They were reduced by an Open Market Committee
seeking to stimulate the economy's recovery from the recession of the early
1990s and concerned by the weakness of the California economy. By mid-1994
it was clear that recovery was secure, and the Fed turned its attention to
inflation. By raising interest rates, it increased the yield and attractiveness of
domestic securities and heightened the debt-servicing burdens of countries like
Mexico with large amounts of short-term debt. Monetary policy in the creditor
countries is far from a complete explanation for the most recent wave of lending
to emerging markets, but Calvo et al. (1992), Chuhan et al. (1993), Fernandez-
Arias (1994), and Dooley et al. (1996) unanimously conclude that fluctuations
in international interest rates account for a sizeable share of the variation in
capital flows.
In all three cases, capital flows had a major impact on the recipient countries.
The increased availability of liquidity and downward pressure on interest rates
stimulated domestic spending. The supply of non-traded goods being inelastic
(compared with tradeables, the supply of which is all but perfectly elastic),
inflows drove up real exchange rates. This caused resources to shift toward the
production of nontradables, widening the trade deficit. In all three episodes,
but especially in the 1920s and 1970s, the increased availability of credit
financed budget deficits which proved difficult to rein in once foreign funds
dried up. In all three instances, foreign funds were channelled through banks
which were left with a mismatch in the maturity structure and currency com-
position of their assets, rendering the banking system vulnerable to
disturbances.
One important difference between the three episodes is currency manage-
ment. In the 1920s, virtually every borrowing country was on the gold standard;
foreign investors regarded membership as a signal of financial probity and
made it a prerequisite for lending. 4 In the 1970s the transition to floating had
begun, but many developing countries, especially in Latin America, continued
to peg. By the 1990s, fully half of large developing countries had gone over to
floating, up from 27% in 1982.
In the 1920s, a capital flow automatically raised the reserves of the recipient
country. The borrower, be it a company, bank or government, converted the
foreign exchange it obtained into domestic currency, depositing the latter with
the central bank. The additional liquidity put upward pressure on prices,
worsened export competitiveness, and fueled financial activity. If the central
bank sought to mop up that liquidity by raising the discount rate, the rise in
rates relative to world levels only attracted more capital from abroad. When
the German Reichsbank kept interest rates high in the second half of the 1920s,
for example, it only stimulated a flow of foreign deposits into German banks.
4There were considerable short-term flows to countries like Germany before they returned to
gold, but these were limited to the early period during which the quick resumption of gold
convertibility was still widely anticipated. Once it became clear in 1922 that quick resumption was
not in the cards, this capital flow dried up (see Holtfrerich, 1986).
56 B. Eichengreen
In the 1970s, exchange rate policies were more diverse. Thailand, Korea,
Indonesia and Malaysia all switched from dollar to basket pegs, varying the
weights on different currencies to provide scope for exchange rate flexibility.
They followed austere fiscal policies to limit the inflow-induced appreciation
of their real exchange rates. In contrast, most Latin American countries contin-
ued to peg to the USA, causing their exchange rates to strengthen when the
dollar appreciated after 1979, and adopted accommodating fiscal policies
(Figure 2). The budget deficits of Asian countries receiving large capital inflows
averaged only 1.1 % and 3.0% of GOP in 1973-76 and 1977-82, compared
with 2.4% and 4.3% in Latin America and the Caribbean (IMF, 1994; p. 58).
Real exchange rates appreciated by 3% between 1976-78 and 1979-80 in Latin
America while depreciating by 11 % in East Asia (Sachs, 1985; Figures 3 and 4).
Latin American governments supported their overvalued rates by rationing
foreign exchange; substantial black-market premia emerged virtually every-
where but Colombia and Venezuela. 5
By the 1990s this distinction between Latin America and Asia had largely
disappeared, a growing number of Latin American countries having elected to
float. Both floaters and peggers had to cope with the tendency for capital
inflows to raise the real exchange rate, undermining the competitiveness of the
traded-goods sector. A variety of instruments were deployed to address this
problem, foremost among them fiscal policy. This time fiscal retrenchment was
more dramatic in Latin America than East Asia: while the budget deficits of
the high-inflow Asian countries fell from 3% to 2% of GOP between 1983-89
and 1990-93, those of the high-inflow countries of Latin America and the
Caribbean fell even more dramatically, from 5% to essentially zero (Figure 2).
Fiscal adjustment by itself did not suffice. A frequently cited success story
is Thailand, whose sharp fiscal correction in 1988-91 limited upward pressure
on its real exchange rate. But even that draconian adjustment did not prevent
the current account deficit from widening to 5% of GOP in 1993 or insulate
the country from the fallout from the Mexican crisis in 1994. There as elsewhere
it proved difficult to adjust fiscal policy with the speed and to the extent
required to accommodate shifting capital flows. Countries seeking to manage
their exchange rates and capital accounts resorted to other instruments, notably
sterilized and un sterilized intervention. The upward pressure that open market
sales placed on interest rates only encouraged capital inflows, as in the 1920s;
this response was evident in Brazil and Malaysia in 1991-93.6
Exchange rate flexibility mattered most when capital flows turned around.
Countries seeking to defend their pegged rates were forced to adopt drastic
measures of monetary retrenchment. Under the gold standard this implied
allowing the monetary base to decline by an amount commensurate with the
(a)
(b)
6
o ~--------------------------~__-,~----------~~-;
1
1986 1987 1988 1989 1990 1991 1992 1993 1994
Figure 3. Real exchange rate index: Latin America. Wholesale prices relative to US (1990 = 100).
--- unadjusted; -+- adjusted by output/worker (Summers-Heston); ---.- adjusted by
output/person (lFS).
~~-----------------------------------------------------,
ICD
150
ICD
1m 1m I_ 1"1 1m 191:1 I'" ItIS 1911 1917 1911 19It I!IIO 199\ 1992 1991
Figure 4. Real exchange rate index: Asia. Wholesale prices relative to US (1990 = 100).
___ unadjusted; -+- adjusted by output/worker (Summers-Heston); ---.- adjusted by
output/person (IFS).
loss of gold and foreign exchange reserves. The deflation that ensued could
threaten macroeconomic stability. These dynamics were evident in the late
1920s: when US interest rates rose and foreign lending declined in 1928, the
balances of payments of capital exporters like the United States strengthened,
obviating the need to increase exports. This shift had as its counterpart a
deterioration in the balance of payments accounts of capital importers in Latin
International lending in the long run 59
10 ~-------------------------------------------------------,
;!t 0
IS
20
~ ~------~------~------~------~------~----~------~~
1924 1925 1926 1927 1921 1929 1930 1931
Figure 5. Trade deficit/imports, 1924-31: Latin America and Central and Eastern Europe.
(Excludes Uruguay for 1926-27. Sources: International Historical Statistics: The Americas (1993)
and European HIstorical Statistics (1975) by B.R. MitchelL)
America and elsewhere, forcing such countries to boost exports and curb
imports in order to sustain their pegged currencies (Figures 5 and 6).
Stimulating exports required lowering the relative prices of the goods they sold
internationally in order to price them into world markets. In other words, for
the maintenance of external balance it was necessary for price levels to decline
more rapidly in capital-importing than in capital-exporting economies.
Eventually this adjustment was achieved, but at considerable macroeconomic
cost in countries wedded to their pegged rates. In contrast, countries which
chose to boost exports by devaluing their currencies, also halting the con-
traction of domestic credit, could slow the deflationary spiral and more quickly
restore macroeconomic stability.7
While it is still too early to assess the relative success of countries with
pegged and floating currencies in adjusting to the Mexican shock, there is some
reason to think that the policy independence conferred by exchange rate
flexibility has been beneficial. In Argentina, the archetypical example of a
country with a pegged rate, the decline in capital inflows associated with the
Mexican crisis plunged the economy into recession. Venezuela, which pegged
7 Eichengreen and Sachs (1985) and Campa (1990) make these points for Europe and Latin
America, respectively.
60 B. Eichengreen
~ r-------------------------------------------------------,
20
10
30
.~ ~------~----~----~~----~----~------~----~----~~
1929 1930 1931 1932 1933 193< 1935 1936 1937
Figure 6. Growth rate of exports, 1929-37: Latin America and Central and Eastern Europe.
(Excludes Spain for 1936-37. Sources: Exports from International Historical Statistics: The
Americas (1993) and European Historical Statistics (1975) by B. R. Mitchell. Nominal exchange
rate from BMS (1943).)
its currency before devaluing at the end of 1995 and subsequently adopted a
fluctuation band, also experienced an adjustment crisis. In contrast, floaters
like Chile and Brazil have weathered the storm relatively well. Even in Mexico,
where the policy was handled badly, currency depreciation has boosted exports.
The other place where the advantages of exchange rate flexibility are evident
is the management of financial instability. After 1928 the withdrawal of foreign
deposits and the deterioration of domestic macroeconomic conditions com-
bined to undermine the liquidity and solvency of banking systems. By the early
1930s, banking crises in debtor countries had become the norm. The most
serious crises, in the summer of 1931, originated in Austria, Hungary and
Germany, three debtor countries with fixed exchange rates. Each had suffered
serious recessions starting in 1929, reflecting the compression of demand caused
by the global economic downturn and the curtailment of capital inflows. Banks
in all these countries had loaned heavily to industry, which was battered by
the slump. And their central banks had their hands firmly tied by gold standard
statutes. Bernanke and James (1991) and Grossman (1994), analyzing samples
of developing and industrial countries, conclude that bank failures were most
prevalent in countries whose central banks were prevented by the gold standard
constraints from engaging in lender-of-last-resort intervention.
Similar contrasts are evident in the 1980s. Again the typical sequence was
International lending in the long run 61
8 This statement should be tempered to take into account the limited capacity of the central
bank to extend credit to the financial system within the limits of the convertibility law and the
decision to permit commercial banks to utilize their own reserves by relaxing the high reserve
requirements that had been in place on the eve of the crisis. See Gavin and Hausmann (1996).
9 See Eichengreen and Portes (1989) and Klug (1993). Buybacks were also widespread in the
1980s; Peru again resorted to the practice in 1995.
62 B. Eichengreen
Historical evidence points to the need for institutions to deal with the problems
of coordination and collective action that arise in the wake of debt-servicing
difficulties. In the 19th century and 1930s these institutions took the form of
bondholders committees; in the 1980s they took the form of the London and
Paris Clubs and bank steering committees. No analogous mechanism existed
for dealing with the Mexican crisis, through which the IMF and G-7 countries
muddled, adopting an exceptional bailout package. There is every reason to
think that the response to the Mexican crisis will not be generalized. Crises
that take place further from US borders will lack the same salience in
Washington, DC. Now that the Congress has discovered the existence of the
Exchange Stabilization Fund, it will be loath to permit the President to again
use it for similar purposes. The personal connections between the US Treasury
and IMF that facilitated assembly of the Mexican package cannot be taken
for granted. Even a doubling of the General Arrangements to Borrow may not
provide the Fund with the resources needed to deal with several crises at once.
If a Mexico-style crisis again ensues, there may be no way to halt the creditors'
scramble for the exits, restructure debts in an orderly way, and inject new
money.
Elsewhere, Richard Portes and I have suggested a menu of institutional
reforms to address this problem (Eichengreen and Portes, 1995). Our first
recommendation is that the IMF more actively transmit signals about the
advisability of unilateral suspensions of debt service (temporary stays or stand-
stills). Debtor governments have the capacity to impose the equivalent of a
creditor standstill by suspending debt service payments but hesitate to do so
for fear of jeopardizing their credit market access. Encouraging the IMF to
advise the debtor and issue opinions on whether or not there is justification
for a stay of payments would allow the Fund to carry out an important
signalling function; a government which received approval for its standstill
would suffer relatively little damage to its reputation, while the possibility that
the Fund would not approve would discourage debtors from utilizing the
option strategically. Naturally, the IMF should limit its advice before the fact
to the debtor government and share its opinion with the markets only ex post
so as to avoid inciting a panic. A definitive reinterpretation of Article VIII(2)(b)
of the IMF Articles of Agreement would support the Fund in this role even if
it was not legally binding in national courts.
Creating a bondholders steering committee would eliminate uncertainty
about the locus of authority in negotiations. That committee would be responsi-
ble for restructuring bonded debts, while the London and Paris Clubs would
retain their responsibility for bank loans and official credits. Discussions
between the debtor and the Paris Club, the London Club and the bondholders
committee could rely on a specially constituted conciliation and mediation
service to prevent extended deadlocks.
Changes in bond covenants to permit a majority of creditors to alter the
terms of payment would prevent rogue investors from holding up the settlement.
To make this palatable to the lenders, dissenting creditors should have recourse
to an arbitral tribunal. To prevent a negotiated agreement or the findings of
the arbitral tribunal from being disputed in court, loan agreements would
specify that objections by minority creditors were subject to the tribunal's
arbitration.
64 B. Eichengreen
financial meltdown should take into account the possibility that financial
instability can spread to other emerging markets.u
The alternative to a procedure for orderly workouts may not be the absence
of intervention, of course. The alternative, as in Mexico, may be a bailout. And
the prospect of a bailout creates moral hazard for lenders as well as borrowers.
Would not changes in bond covenants permitting a majority of creditors to
alter the terms of payment, like those we propose, be regarded by bondholders
as weakening their bargaining position? (In other words, might the terms at
which developing countries could borrow deteriorate if bond covenants carried
such clauses?) The same question can be asked, of course, of a corporation's
creditors in the context of a bankruptcy procedure: why do countries' statutes
permit 'cramdown' by the courts or allow two-thirds of each class of creditors
to ratify a restructuring agreement? The need for unanimous consent creates
incentives for particular creditors to behave strategically by withholding con-
sent; cramdown and majority rule can be seen as countering these incentives.
Once again the question arises of why, if such clauses have merit, they have
not already been adopted by the markets. An answer is that bargaining power
between a country's debtors and its government may be distributed more
asymmetrically than the bargaining power between a corporation and its
creditors. Courts can seize a corporation's collateral and dismiss its manage-
ment; there are no analogous mechanisms for containing moral hazard in the
context of sovereign debt. There may be a temptation for bondholders to cave
in to a government's strategic default; allowing the least compromising bond-
holders to set the terms of bargaining (or requiring them to be bought out by
the government or by other creditors) can be thought of as a counterweight.
Hence, if they are to accept majority rule, bondholders may require other
compensation. This is why we recommend that dissident creditors should have
recourse to an arbitral tribunal. To prevent a negotiated agreement or the
findings of the arbitral tribunal from being disputed in court, loan agreements
would specify that objections by minority creditors were subject to the tribunal's
arbitration.
How would the Bondholders Council be organized? History suggests that
a confusing proliferation of committees can spring up in the absence of official
accreditation. It may be desirable therefore for the governments of the creditor
11 In fact, there are reasons to worry that moral hazard is more severe in the sovereign than
the corporate setting. In the case of sovereign debts, no mediation and conciliation service or
arbitral tribunal for dissident creditors could seize collateral in the manner of a bankruptcy court.
It is not possible to 'throw out' the government of a country whose debts are restructured in the
way that bankruptcy judges in the United States can replace the management of firms in Chapter
11. (Note, however, that sovereign financial crises often lead to replacement of the finance minister,
which should have some ex ante deterrent effect on the relevant policymakers. Moreover, bank-
ruptcy procedures in other countries do not always empower the courts to replace the management
of firms undergoing reorganization.) That there can be a moral hazard problem in this context
provides the basis for arguing that IMF surveillance before the fact and conditionality afterwards
need to be strengthened.
66 B. Eichengreen
The G-10 Working Group on Sovereign Liquidity Crises (the Rey Committee
after its chairman Jean-Jacques Rey) has finally come forth with its proposals
for handling financial crises in emerging marketsP To facilitate restructuring
and reorganization it proposes modifying the provisions of loan contacts. It
recommends incorporating into all loan agreements a 'collective representation
clause' clearly designating the creditors' representative and making provision
for a bondholders meeting. 13 That representative (the fiscal agent or indenture
trustee) would be empowered to deal with governments and international
llGroup of Ten (1996). At the Halifax Summit of June 15-17, 1995, the Heads of State and
Government of the G-7 countries encouraged G-lO Ministers and central bank governors to
consider new procedures for the orderly resolution of sovereign debt crises. In response, the G-I0
established a working party consisting of representatives of ministries of finance and central banks.
In a parallel initiative it has also been exploring the feasibility of expanding the General
Arrangements to Borrow in order to provide the liquidity needed to respond to financial
emergencies.
13 Currently, many sovereign bond agreements, most notably Brady Bonds, do not provide for
bondholders meetings at the initiative of the bondholders or the debtor.
68 B. Eichengreen
14 Under the United States Trust Indenture Act 1939, the indenture trustee takes responsibility
for the affairs of the bondholders in a default situation. The trustee acts as a communications
center and provides coordination services for the bondholders. He is obligated to carry out
instructions voted by the bondholders. Absent such directions, he may unilaterally accelerate the
balance due, recover a judgement against the obligor, or sue to enforce the bond covenants. See
Macmillan (1996). However, the framers of the 1939 Act exempted sovereign debt from this
provision, instead giving more limited responsibilities to the fiscal agent (who, for example, has no
fiduciary responsibility to the bondholders). The implication of the Rey Committee's recommenda-
tion is that the Trust Indenture Act should be amended to allow for indenture trusteeship in the
case of sovereign debt. Similar legislation would also have to be adopted in the other principal
creditor countries, like the UK, where the fiscal agent is also used.
International lending in the long run 69
In its absence, new clauses will work their way into the markets only
gradually, as old loans are retired and new ones are issued. Since bond issues
run 10, 20 or 30 years to maturity, it will be well into the next century before
much of an effect is felt. Another approach would have been for the Rey
Committee to encourage governments to convert old issues into new ones prior
to maturity and even to recommend that G-10 governments and the IMF
provide some finance for the purpose.
While noting that bondholders often organize representative committees
and suggesting that this is an adequate way of solving the representation
problem if and when the markets find it congenial, the report does not recom-
mend that governments themselves promote the establishment of standing
committees for this purpose. The Rey Committee may have been impressed
that the Costa Rican, Guatemalan and Panamanian restructurings of the 1980s
and 1990s were accomplished without the intervention of a bondholders com-
mittee; rather, negotiations proceeded through direct contacts between the
government and members of the main groups holding the securities. Is But it
is not clear that this mechanism will work so smoothly for larger countries
with more numerous creditors. The G-10 Committee may have anticipated
that defaulted debts will be bought up by 'vultures' whose small number will
allow governments to negotiate with them relatively easily.I6 While this is not
unrealistic given sufficient time, the process of consolidating bond holdings in
the hands of a small number of investment professionals will not be completed
in a matter of days or months. In the absence of a representative committee
authorized to speak for the bondholders, negotiations will remain messy for
some time. Again, the Rey Committee may have erred on the side of inaction
in order to appear as uninterventionist as possible.
The Rey Report discusses the idea of an independent entity to act as
coordinator, conciliator and arbitrator in negotiations interchangeably with
proposals for an international bankruptcy court, apparently on the grounds
that both would involve the creation of a new international agency. But the
idea of an agency for voluntary mediation and conciliation is more limited. Its
involvement could be subject to the express approval of both debtors and
creditors; provision for this might be made in the bond contract itselfY It
would be a low-cost way of giving official agencies like the IMF a way of
recognizing and supporting the authority of particular representatives of the
bondholders. The Rey Committee objects that existing arbitration procedures
CONCLUSION
18 Thus, Allan Meltzer has argued that the market solution would have been better for Mexico.
"Default and renegotiation was the lower cost solution. Mexico would not have been the first
country to default, and the default would not have been the first for Mexico. The record of recent
defaults shows that, once there is an agreement on the terms of the renegotiation, capital flows to
the country soon resume" (Meltzer, 1995; p. 11).
International lending in the long run 73
REFERENCES
Bernanke, Ben and Harold James (1991), The gold standard, deflation and financial crisis in the
great depression: an international comparison. In R. Glenn Hubbard (ed.), Financial Markets and
Financial Crises. Chicago: University of Chicago Press, pp.33-68.
Chuhan, Punam, Stijn Claessens and Nlandu Mamigni (1993). Equity and Bond Flows to Latin
America and Asia: The Role of External and Domestic Factors. Policy Research Working Paper
no. 1150. Washington, DC: The World Bank.
Calvo, Guillermo A., Leonardo Leiberman and Carmen M. Reinhart (1992) Capital Inflows to Latin
America: The 1970s and 1990s. Unpublished paper, International Monetary Fund.
Campa, J.M. (1990). Exchange rates and economic recovery in the 1930s: an extension to Latin
America. Journal of Economic History, 50, 677-682.
Cuddington, John T. (1989). The extent and causes of the debt crisis of the 1980s. In Ishrat Husain
and Ishac Diwan (eds), Dealing with the Debt Crisis. Washington, DC: The World Bank, pp.l5-42.
Dooley, Michael, Eduardo Fernandez-Arias and Kenneth Kletzer (1996). Recent private capital
inflows to developing countries. World Bank Economic Review, 10,27-50.
Edelstein, Michael (1981). Overseas Investment in the Age of High Imperialism, New York: Columbia
University Press.
Edwards, Sebastian (1989). Real Exchange Rates, Devaluation and Adjustment. Cambridge: MIT
Press.
Eichengreen, Barry and Albert Fishlow (1995). Contending with capital flows: what is different
about the 1990s? Council on Foreign Relations Occasional Paper, New York: Council on Foreign
Relations.
Eichengreen, Barry and Richard Portes (1989). After the deluge: default, negotiation, and readjust-
ment during the interwar years. In Barry Eichengreen and Peter Lindert (eds), The International
Debt Crisis in Historical Perspective. Cambridge: MIT Press, pp. 12-47.
Eichengreen, Barry and Richard Portes (1995). Crisis? What Crisis? Orderly Workoutsfor Sovereign
Debtors. London: Centre for Economic Policy Research.
Eichengreen, Barry and Jeffrey Sachs (1985). Exchange rates and economic recovery in the 1930s.
Journal of Economic History, 45, 925-946.
Feis, Herbert (1930). Europe, The World's Banker, 1870-1914. New Haven: Yale University Press.
Fernandez-Ansola, Juan Jose and Thomas Laursen (1995). Historical Experience with Bond
Financing to Developing Countries. Unpublished manuscript, IMF.
Fernandez-Arias, Eduardo (1994). The New Wave of Private Capital Inflows: Push or Pull? Policy
Research Working Paper no. 1312. Washington, DC: The World Bank.
Fishlow, Albert (1985). Lessons from the past: capital markets during the 19th century and the
interwar period. In Miles Kahler (ed.), The Politics of International Debt. Ithaca, New York:
Cornell University Press, pp. 37-94.
Gavin, Michael and Ricardo Hausmann (1996). The Roots of Banking Crises: The Macroeconomic
Context. Working Paper Series 318, Office of the Chief Economist, Inter-American
Development Bank.
19 The enormous cost of support packages, plus doubts on the part of creditor-country politicians
and financial institutions that they have much at stake, make it unlikely that 'bridge loans' (read
'bail-outs' or 'emergency assistance') on the same scale will be offered to the next Mexico.
74 B. Eichengreen
Grossman, Richard (1994). The shoe that didn't drop: explaining banking stability during the great
depression. Journal of Economic History, 54, 654-682.
Group of Ten, Working Party on the Resolution of Sovereign Liquidity Crises, The Resolution of
Sovereign Liquidity Crises.
Holtfrerich, Carl-Ludwig (1986). US capital exports to Germany: 1919-1923 compared to
1924--1929. Explorations in Economic History, 23,1-32.
International Monetary Fund (1994). World Economic Outlook. Washington, DC: IMF, October.
Klug, Adam (1993). The German buybacks, 1932-1939: a cure for overhang? Princeton Studies in
International Finance, no. 75. International Finance Section, Department of Economics,
Princeton University.
Lary, Hal B. (1943). The United States in the World Economy. Washington, DC: GPO.
Lewis, Cleona (1938). America's Stake in International Investments. Washington, DC: The Brookings
Institution.
Macmillan, Rory (1995). The next sovereign debt crisis. Stanford Journal of International Law,
31, 305-358.
Meltzer, Allan H. (1995). A Mexican Tragedy. Unpublished manuscript, Carnegie Mellon
University.
Pinon-Farah (1996). Private bond restructurings: lessons for the case of sovereign debtors. IMF
Working Paper no. 96/00 (February).
Sachs, Jeffrey D. (1985). External debt and macroeconomic performance in Latin America and East
Asia. Brookings Papers on Economic Activity, 2, 523-564.
Sachs, Jeffrey D. (1995a). Alternative Approaches to Managing Financial Crises in Emerging Markets.
Unpublished manuscript, Harvard University.
Sachs, Jeffrey D. (1995b). Do We Need an International Lender of Last Resort? Unpublished manu-
script, Harvard University.
Securities and Exchange Commission (1937). Report on the Study and Investigation of the Work.
Activities. Personnel and Functions of Reorganization Committees. Part V: Protective Committees
and Agenciesfor Holders of Defaulted Foreign Bonds. Washington, DC: GPO.
Truman, Edwin M. (1996). The risks and implications of external financial shocks: lessons from
Mexico. International Finance Discussion Paper no. 535, Board of Governors, Federal Reserve
System (January).
United States Department of Commerce (1976). Historical Statistics of the United States.
Washington, DC: GPO.
World Bank (various years). World Debt Tables. Washington, DC: The World Bank.
PHILIP SUTTLE
J.P. Morgan
Professor Eichengreen's paper covers two sets of issues, with his usual degree
of straightforwardness and clarity. The first is that of the similar patterns
evident in cycles in lending to developing countries, emerging markets, or 'high-
risk far-off places' (call them what you will). He compares the 1920s, the 1970s
and 1990s and emphasizes that the period of surging lending shared three
common characteristics:
Apparent improvements in the long-run economic prospects of borrowing
countries;
Easy monetary conditions in the lending economies;
Ensuing problems of macro-management in borrowing countries in both
boom and, especially, bust periods. Particularly difficult throughout is the
choice of an appropriate exchange rate regimen.
I agree that this is the right way to think about the swings in capital flows to
emerging economies, and I do not have a great deal to add to what Barry has
written. I would like to make a couple of points, however, from the perspective
of the markets today. First, since February, we have had (in market terms) a
significant shift in thinking on U.S. interest rates; 2-year note yields have risen
by over 100 basis points. The last two times this happened were February 1994
and OctoberjNovember 1994. And we all know what happened in emerging
markets - specifically Mexico - on both occasions. This time, however, the
peso has actually appreciated against the dollar, Mexican foreign exchange
reserves have edged up, and short-term interest rates have fallen from 50% to
25%. These moves have a lot of market participants talking about emerging
markets decoupling from U.S. rates. I noted, for example, an article on that
very theme in the Wall Street Journal yesterday.
What should we make of this? One point to make is that it illustrates the
short time horizon of market thinking. Let us wait and see how well emerging
markets do in the late summer when the Fed is likely to tighten policy. Another
point, which is being a bit more generous to the markets, is that the rate link
is particularly important when the maturity of the lending is short term. This
was certainly the case in 1994-95 (not just in Mexico), and was also true of
These comments reflect the personal opinions of the author and do not necessarily reflect the
views of J.P. Morgan.
75
R. Levich (ed.), Emerging Market Capital Flows, 75-78.
CI 1998 Kluwer Academic Publishers.
76 P. Suttle
the late 1970s and early 1980s. The bottom line is that the countries that get
into trouble when rates shoot up are those with a lot of short-term debt.
Second, concerning this point, we need to think globally and not just about
U.S. interest rates. U.S. short-term interest rates are almost double their lows
of 1993, but German and, especially, Japanese interest rates remain very low.
This helps to explain, I think, why capital flows into Emerging Europe and
Emerging Asia have been so strong recently (in some cases too strong, illustrat-
ing a point that Barry makes about policy problems in dealing with capital
inflows). To take one illustration: The liabilities of Thailand to BIS-area banks
(mainly G-10 banks) rose by $28 billion in the first 9 months of 1995. Much
of this was short term and much came from Japan, looking to take advantage
of the short-term spread between Japanese short-term rates of 0.5% and Thai
short-term rates of 10%. Indeed, you could even go as far as to say that the
next 'bubble' being created by the Bank of Japan's current excessively easy
monetary policy is not to be seen in Japan, but in the extended boom that is
going on in its smaller neighbors to the southwest.
One lesson to bear in mind looking ahead is thus: Beware rising Japanese
and German interest rates. I am not saying they would have the same devasta-
ting effect on Emerging Asia and Emerging Europe, respectively, that rising
U.S. rates in 1994-95 had on Latin America, but my view is that markets
currently underweight this risk.
The second set of issues that the paper then goes on to consider is that of
how to handle crises, on the assumption that they periodically occur in the
area of sovereign lending. The Mexican crisis was the latest in the long line of
such accidents, and it would be foolish to think that it will be the last of this
round. The key proposition of the paper is that we have a classic case of market
failure where what is good for efficiency in the vast majority of cases (i.e. a
capital market that takes risk) is a poor outcome for all when the bad event
occurs. Barry thus suggests that the world would be a better place if two key
institutional reforms were adopted:
The IMF were put in the position of being able to declare that a moratorium
of payments by a country is justified;
New bond contracts were written (and, if possible, old ones rewritten) in a
way that would allow a bondholder committee to be formed quickly with
the power to restructure bonded debt.
I am skeptical that such proposals would do much good. I would not wish the
role of default arbiter on my worst enemy, let alone on the IMF. As we all
know, default is as much about willingness to pay as ability to pay, and I am
not sure how to judge when it was objectively fair for a country to cease
making payments. Moreover, rewriting bond contracts may not do much good.
Bonds are already flexible enough: They go down in price to reflect the fact
that I am not being paid, or may not be paid. Moreover, most emerging market
bonds start off being non-investment grade - that is subject, in the eyes of the
rating agencies, to considerable default risk. If these price mechanisms fail to
work - as in the case of Mexico - then there arises a public policy choice of
Comment 77
either doing something ad hoc, surprising and innovative, as was the case in
Mexico, or relying on the creation of some form of bondholders committee
which, whether constituted ex ante or ex post, would hold its first meeting
soon after any formal default (and probably sit for many years after).
In fact, almost 18 months on from the onset of the Mexican crisis, I think
distortion arises in an attempt to compare the Mexican crisis with the debt
crisis of the early 1980s. Then, Mexico was a domino (not even the first) in a
process that did bring down a whole set of countries. Rescheduling became an
industry, and institutions - steering committees, Paris Clubs, London Clubs,
the IIF - were created left, right and center. Last year, Mexico suffered a
sudden loss in market access, which certainly spread to Argentina but not
much further. Mexico and Argentina were forced into very painful, deep adjust-
ments, as small open economies occasionally are when they overextend
themselves.
Of course, support packages were put together for both countries, and may
well have helped (I think probably did) the spread of contagion effects to other
countries and regions. But growth in many Latin American countries remained
strong in 1995, and growth in all other parts of the emerging world actually
picked up in 1995. This was fueled by stronger, not weaker, capital flows. At
the start of the year, it looks as if the Mexican crisis would make the overall
emerging world's growth much weaker in 1995; in fact, the growth surprises
were generally on the weak side in the OECD, not in the emerging markets in
1995. We did not have another generalized debt crisis and I do not think we
are on the edge of one today.
I also think that it is wrong to call the support that Mexico enjoyed last
year a 'bail-out.' For the U.S. Treasury, the assistance extended has turned out
so far to yield a significant financial return (although not as good as if they
had bought Brady's), and clearly helped stabilize the Mexican economy, which
is now nicely moving into a phase of recovery. Those creditors who were bailed
out in the early months of 1995 were classic sellers at the lows. Of course, they
got more back than they might have done, but the point is that the package
(from today's vantage point at least) seems to have been a pareto improvement:
I can see plenty of winners, but I have a hard time spotting losers.
Remember that Mexico returned to the markets this time within 7 months,
rather than the 7 years that it took last time. To me, this suggests that a flexible
approach that involved as little market disruption as possible worked wonders
compared to the institutional "getting everyone together in a smoke-filled
room" approach that Barry quotes Ted Truman as looking back at wistfully
in his paper.
In conclusion, I do not want to give the impression either that a bottomless
pit of cash from the public sector is the right approach to every problem, or
that the market works perfectly. There are cases of market failure which make
it highly appropriate for public sector institutions to play a role in dealing
with incipient crisis as decisively as possible. In some cases large packages may
be important; in others, a different approach may be warranted. Last year, for
78 P. Suttle
example, the support enjoyed by Argentina and Mexico from the IMF was
radically different, even though their quotas as the Fund, which should govern
access, are very similar. I would say that famous 'case-by-case' approach that
guided the early response to the debt crisis of the early 1980s is probably the
most adept way to deal with international financial market difficulties that
inevitably lie ahead. Remember, that for all the regularities in international
finance there seem to be, we are often hit by the next crisis because we are
trying too hard to look backwards to solve the last one.
MICHAEL P. DOOLEY
University of California at Santa Cruz
This is a challenging paper. It puts forth a clear and forceful argument for
systematic government intervention in international capital markets. First let
me layout the bare bones of the argument.
Financial contracts do not, in practice, eliminate the risk that debtors will
be unwilling to meet their obligations under some circumstances. Contracts
that set out all the contingencies under which both debtors and creditors might
be willing ex ante to adjust repayment terms would involve enormous transac-
tions costs and would reduce, but not eliminate, moral hazard and the tempta-
tion for ex post self-interested bargaining. For these reasons real world contracts
are quite simple. Contingent conditions are strictly limited to easily verified
factors such as international interest rates that are beyond the control of
the debtor.
It follows that a proportion of countries will default on their obligations. If
international interest rates rise, then this proportion increases. The point made
in the paper that I think is the most useful is that an efficient private debt
market will, nevertheless, make such loans with positive default probability
and charge an appropriate default premium. Financial crises are a feature of
an efficient system. I agree with the author that we must learn to live with this
feature of financial markets.
A good system would be one in which the parties agree costlessly to enter
into a new agreement that reflects conditions realized after the contract was
entered into. In the best case the new contract is the same as would have been
agreed upon if the original contract had been contingent on the state of the
world that in fact occurred.
A domestic bankruptcy court can be thought of as a place where all the
contingent contracts are recorded and an impartial judge decides which con-
tract to enforce. When the actual contract generates insolvency the judge
impartially applies well known criteria to rewrite the contract and compels all
creditors and the debtor to comply with its interpretation of the facts. Thus, a
bankruptcy court mimics a complete menu of contingent contracts.
For international contracts there is no court to enforce a judgment. This
leads to strategic behavior by debtors and creditors to protect their interests.
Eichengreen focuses on self-interested bargaining among private creditors as
an important factor in generating costly settlements. In the background is the
79
R. Levich (ed.), Emerging Market Capital Flows, 79-81.
C> 1998 Kluwer Academic Publishers.
80 M.P. Dooley
sensible fear that this leads to prolonged wars of attrition among private
creditors that are very damaging to debtor countries.
The solution therefore has two building blocks. The first requirement is that
a settlement mechanism must have some power to enforce a settlement. Since
coercion is not possible for private creditors from different legal systems, an
alternative is a cash side payment from an international organization or its
equivalent. In practice this might take the form of new multilateral loans that
are junior to the existing private debt. As Eichengreen points out, the terms
on which new loans are made can be a powerful source of incentives for both
debtors and creditors to agree to a settlement. Moreover, he argues that because
the cost of failure to attain a settlement is so high that creditor governments
cannot credibly commit to not providing side payments.
The obvious problem with side payments are that they will feed back to the
first part of the game. An important feature of the domestic bankruptcy court
is that it does not offer cash to buy creditor agreement. To the contrary lawyer
fees are recognized as an important dead weight loss that must be shared
among the debtor and its creditors. (I might add that in this light lawyers are
actually useful since they provide a commitment mechanism.) This is equivalent
to the IMF appropriating some of Mexico's reserves for the service of enforcing
an agreement among private creditors rather than making new loans on gener-
ous terms. The original contracts are obviously distorted by the expected value
of the costs or payment. The domestic distortion is usually too little debt; the
international distortion is too much debt.
There is a domestic counterpart for a mechanism that involves positive side
payments, the lender of last resort (LOLR) to the banking system. Since the
LOLR is also a creditor the game with investors inevitably involves the expecta-
tion that risky investments involve private benefits and public costs. The well
known solution to the distortion associated with lenders of last resort, or more
generally side payments to buy agreement, is supervision and regulation. For
this reason Eichengreen proposes an expanded role for the IMF to provide
just such a supervision and regulation for governments. He points out, I think
correctly, that the Fund would be happy to expand into this very labor intensive
role in the system. But I have serious doubts about the Fund's ability to play
such a role.
Part of the problem is identified by Eichengreen. Solvency crises need not
reflect bad economic management, at least not bad management that can be
identified ex ante. Simple, noncontingent debt contracts mean that a certain
number of 'no fault' insolvencies are a feature of an efficient system. The Fund
cannot see these coming any better than creditors. The Fund as an institution
did not see Mexico coming even though some staff members provided very
clear warnings about the likely impact of rising international interest rates.
Why not? We may know that certain shocks to the system will expose weak
creditors. A rise in international interest rates is the dominant example. But
there is no evidence that we can guess which debtor will attract speculative
interest and which will be unable or unwilling to fight off an attack.
Comment 81
We do know that the initial capital inflow will go to countries most likely
to use up their own resources and borrow from the Fund or friendly govern-
ments in order to defend the regime. I would argue that this is likely to be
countries with fixed exchange rates, weak banking systems and access to
political good will. The lesson I draw from history is that solvent governments
that offer an exchange rate guarantee and a banking system guarantee suffer
from private inflows just large enough to exhaust the governments' net worth.
Increasing the access of such governments to an international lender of last
resort will increase the scale of private capital flows necessary to exploit the
expected insurance.
My conclusion is that extension of even greater incentives for another round
of insurance seeking capital inflows is not as Eichengreen argues "the only
game in town".
The debtor countries themselves are the players in the game that have the
most to lose from such 'solutions'. Their alternative is to break the circle of
insurance. First, fixed exchange rates must be replaced by more flexible forms
of exchange rate management. Second, domestic banking systems must be
effectively regulated. Third, private investors must be given no hope for a side
payment in the event of a crisis. It may be that creditor governments cannot
resist the temptation to save the world. After all, they are correct in arguing
that the bailout process does not cost them much, if anything. But debtor
countries are the big losers in this game and this means that they can and
should refuse to participate in the bailout of international investors.
PART TWO
The Mexican financial crisis of late 1994 and early 1995 resulted in a linked
collapse of stock market values in almost all developing countries, regardless
of economic policies or performance. The contagion effect was clear, and much
commented on, if not fully explainable by either theory or past experience.
Other Latin American equity markets, being close to ground zero, universally
declined by 15-30% in less than a month, as did markets in Asia, where equity
indexes in Hong Kong, Singapore, Taipei, Seoul and Bangkok dropped 10-15%
in a matter of days. Markets in Poland, Hungary and the Czech Republic
fell by similar amounts see (Figure 1). Overall, the International Finance
Corporation's IFCI Latin America Index lost 19% during the calendar year
1995; IFCI Asia index lost 7% while the IFCI EMEA Index (Eastern Europe,
Middle East and Africa) gained 20%, but this was due almost entirely to gains
by the heavily-weighted South African equity index (Littler and Malouf, 1996).
Meantime, the U.S. S&P 500 index had risen over 30%.
These dramatic price movements occurred against the backdrop of an ongo-
ing decline in emerging market share prices. 2 Together with sizeable losses in
emerging market bonds, the investment performance of sophisticated hedge
funds, country funds and regional funds that had made major commitments to
emerging equity markets was savaged. These developments, in turn, triggered
sharp portfolio adjustments and mutual fund redemptions and caused a general
loss of confidence in developing-country stock markets as serious investment
vehicles. Apparently portfolio managers all rushed for the exits, whether or not
the local markets were directly related to Mexico, to avoid performance punish-
ment and to anticipate fund redemptions, as well as to maintain their port-
folio weights.
1 This paper is based in part on an earlier paper by the authors (Smith and Walter, 1996).
2 The Morgan Stanley Group Emerging Market Index dropped nearly 20% in the calendar
year 1994, having risen by 20% in 1993, and 48% in 1992. Stock prices in individual countries
that had attracted particularly large portfolio equity inflows were hit much harder than the index
as a whole. As measured in dollar terms at the end of January 1995, as compared with January 1,
1994, the stock market in Turkey was down 57%, Mexico 56%, China 54%, Poland 50%, Hong
Kong 41 %, Israel 40%, and 10% or more in just about all the rest of the emerging market
countries. The principal exceptions were Chile where the market was up 42%, Brazil 39%, South
Africa 10% and South Korea 9% over the same period.
85
R. Levich (ed.), Emerging Market Capital Flows, 85-105,
1998 Kluwer Academic Publishers.
86 R. C. Smith and I. Walter
LATIN AMERICA
Mexico -22.2
Peru -19.2
Brazil -10.2
Chile -e.g
Argentina .Q.8
Venezuela 4 .8
EUROPE
Hungary -21 .1
Poland 13
Turkey -12.9
Czech Repub lic 7
ASIA
Pakistan -13.4
Philippines -13.2
China -12.5
India -12.2
South Korea -11 .4
Taiwan -11 .3
Hong Kong -10.3
Thailand -10.3
Malaysia oQ.2
Indonesia -8.3
Singapore -6.5
Sri Lanka -2.3=
-25 -20 -15 - \0 ..s 0
3 For an excellent review of the history of developing country debt defaults and reschedulings,
see Salomon Brothers, 1993.
Rethinking emerging market equities 87
50
40
30
c:CD ,'"
rf 20
0
10
Figure 2. History of Sovereign debt defaults, expropriation of investment, and bank debt reschedul-
ings, 1823-1989.
Nation-states in default on foreign bonds
- ---- Nation-states involved in reschedulings
- - -- Major country foreign investment in default
Note: Major countries refer to Great Britain (1823-1939), France (1833-1939), Germany
(1880-1939) and the United States (1914-1939).
Source: Salomon Brothers, Inc.
4 World Bank data, as cited in The Economist, January 28,1995, and Financial Times, February
20, 1995.
88 R. C. Smith and I. Walter
The end of the 1980s saw a collapse in the market for high-performance
investment vehicles such as leveraged buyouts, takeovers financed with junk
bonds, speculative real estate development in North America, Japan and
Europe, as well as Japanese stocks, a market which had enjoyed nearly 20 years
of continuously rising prices. Institutions had been the principal investors in
these instruments which, having served their time, needed to be replaced.
Despite the difficulties of the late 1980s, financial markets looked ahead
optimistically to a continued expansion of global investment activities as a
result of the considerable deregulation in Western Europe following efforts to
implement the EU's single market initiative, the rebirth of Eastern Europe as
a collection of nascent market-driven economies, and the recognition that
liberalizing reforms in countries like Chile and Mexico could bring to Latin
America much of the potential realized in the market-oriented economies of
Asia. Indeed, in the gloomy recession-bound environment of the early 1990s
in New York, London and Tokyo, the lure of the emerging markets of Latin
5 A term coined by John Williamson of the Institute for International Economics as the
conventional wisdom among opinion-leaders that free markets and price stability are preconditions
for viable economic development, comprising such elements as removal of internal and external
market distortions, balanced budgets, pegged exchange rates and privatization of state-owned
enterprises.
Rethinking emerging market equities 89
America and Asia became self-evident. Attention was attracted away from the
conventional international investments of the past to zero-in on the promise
of undervalued, high-growth and diversification opportunities in parts of the
world that had not materially participated in international portfolio invest-
ment before.
The first equity markets to 'emerge' during the 1980s were actually France,
Germany, Spain and Italy - all countries in which capital market instruments
appeared in some quantity for the first time following deregulation efforts and
market reforms. After the fall of the Berlin Wall in 1989, the excitement shifted
to Eastern Europe, where investment opportunities were pursued aggressively,
especially in Poland, the Czech Republic and Hungary, although such invest-
ments were severely constrained by a shortage of securities. Country-specific
investment funds such as the Morgan Stanley Germany Fund, which was
intended to identify choice investments in the newly reunified nation, quickly
rose in the market to trade at substantial premiums above their net asset values.
Also during this time, the major international banks were unwinding the
last of their Latin American loans (and some in other parts of the world) which
had been in arrears for many years. Some were able to sell their loans at deep
discounts in the secondary market to market-makers who would resell them
to others. The Brady debt restructurings beginning in 1989 created a new
supply of debt instruments for loan conversion (about $250 billion at par value)
which were collateralized as to principal by US government securities and
offered partial interest payment guarantees. These soon became actively traded
as a new form of low grade, high-yield bond, offering equity-like returns to
investors. Bond yields of 25-40% were available on many Latin American
issues, yields which continued to reflect bank loan losses that many investors
felt were not likely to apply to the post-Brady situation. The Brady restructur-
ings were coupled to government promises of immediate and significant regula-
tory reforms and improved economic management. The first country subject
to Brady debt restructuring was Mexico under President Carlos Salinas de
Gortari, soon followed by Argentina, where the new government of Carlos
Menem, a former radical Peronista, had aggressively pursued market economics
and privatization.
Early success with Brady bonds, recognition of beneficial effects of serious
economic policy reforms, evidence of substantial foreign direct investment by
multinational corporations, and the return of massive amounts of flight capital
by residents of several Latin American countries began to be reflected in the
stock markets of those countries. Private capital flows to major Latin American
countries exceeded $72 billion in 1993, a substantial turnaround from the late
1980s when capital flows were significantly negative. This activity attracted
American, British and other fund managers, who began to invest aggressively,
90 R. C. Smith and I. Walter
and the sudden flow of funds into comparatively illiquid markets began to
demonstrate the expected effects. Stock prices rose rapidly, encouraging invest-
ment bankers and money managers to augment their offerings of emerging
market funds to smaller institutions and to the general public. On June 30,
1993 the 100 largest among such funds had an aggregate net asset value of
$26.3 billion. By the end of that year, the combined net asset value of these
funds had more than doubled to $54.1 billion as a result of both market
appreciation and new money raised (Micropal, 1994).
It was not so much the availability of attractive investment returns that
caused the capital stampede: these had been available before. Rather, it was
the changed perception of these countries from backward, hopeless places in
which money could only be made by those on the inside, to bright beacons of
hope for revival based on the long delayed discovery of economic truth. This
was not just a reading of the situation in terms of the Washington Consensus;
it was a view shared by those with serious money in New York, London,
Zurich and Tokyo.
The buoyant demand for emerging market securities encouraged countries
that were undecided about the free market policies to join the bandwagon. The
obviously positive market-effects of the new policies were a testimony, many
felt, to their wisdom. Rising equity prices made possible many important
privatization issues. New issues by emerging market countries, principally
privatizations, reached $17.5 billion in 1994, about $2 billion more than the
year before. These included nearly $3 billion in Peruvian, $2 billion in Chinese,
and over $1 billion each in Pakistani and Indonesian privatization issues.
Argentina issued $900 million of privatization equities in 1994, bringing total
proceeds since 1992 to more than $11 billion. Such activity was vastly in excess
of any prior usage level of capital markets on the part of these countries.
Indeed, by historical standards these amounts were astonishing.
A popular explanation offered by investors to explain their aggressive
involvement in emerging markets was the idea that corporate assets were
available very cheaply in some countries relative to the value of comparable
assets in the USA or other developed country markets, because expected values
were attractive and/or the political and economic risks were so high. But what
if the expected returns were boosted and/or the risks dropped dramatically
because of sensible economic practices along the lines of the Washington
Consensus? Argentina was often used as an example: once the government's
new economic policies were known and appeared to be both credible and
durable, the Buenos Aires stock market shot up to correct the undervaluation
6
of the corporate assets for the new environment embodying dramatically lower
perceived risk. Thus, equity investors' search should not only be for potentially
attractive companies, but also for countries whose economic policies were
about to change. Early investments in Peru, Turkey, Poland and China were
made on this basis. An associate of George Soros noted that they liked to
invest when things have gone from "utterly hopeless" to "just plain desperate."
Rethinking emerging market equities 91
"Peru has come back to the world," he said. It had dug itself out of "an
economic mess second to none," and was now a suitable place to invest
(Powers, 1993).
6'Major market" in this case is defined as the US dollar-based Morgan Stanley Capital
International World Index, while 'emerging market" is defined as the US dollar-based International
Finance Corporation Emerging Markets Composite Index, each weighted by the capitalizations
of the various markets.
92 R. C. Smith and I. Walter
Rlsk(%)
.. ..
Figure 3. Investor gains from emerging market equity portfolio allocation, 1987-1994 (year-end).
Our proxy for the major global equity markets is the US dollar-based MSCI World Index which
is composed of securities in the major markets worldwide, including the US. For emerging markets,
we use the US dollar-based IFC Emerging Markets Composite Index, an aggregate of activity in
emerging markets globally. Risk is defined as absolute volatility using the standard deviation of
month-to-month total returns. Average return is the annualized total return performance over the
designated period. We assume no adjustment for withholding taxes.
Source: Merrill Lynch Global Investment Strategy.
major and emerging markets, as well as less than perfectly correlated exchange
rate movements against the dollar, ensuring 'different baskets' for the invest-
or's eggs. 7
The data certainly looked good, but how reliable were they? Many of the
more popular emerging markets had only a few years of history, most of which
covered a time of global speculative boom. In such markets, statistical informa-
tion provided by governments and by corporations is often late or unreliable,
or both. Net returns require subtracting from gross equity returns transaction
costs (bid-offer spreads, fees, commissions, custody charges, clearance and
settlement costs and delays, etc.), which in emerging market countries are often
enormous in comparison to the more developed markets. And large risk adjust-
ments are required for market illiquidity, where even moderate buying or selling
pressure often leads to massive price changes. Such conditions were certainly
far removed from what many investors were used to - adequate and timely
information disclosure and dissemination, low transactions costs and liquid
7IPD gains tend to be greater across global equity markets than across global bond markets,
where they derive solely from less than perfectly correlated interest rate and exchange rate
movements. Moreover, unlike the global bond markets, stocks are highly differentiated and subject
to local trading conditions, although listings on foreign stock exchanges through depository receipts
have made some emerging market equities considerably more accessible to foreign investors.
Rethinking emerging market equities 93
Speculative excesses
8 Nor was it much easier to assess investment parameters in emerging debt markets, comprising
bond issues by sovereign governments and corporations and sold to international investors. These
securities, being rated below investment grade by the principal rating agencies, were looked upon
as 'junk bonds'. International investors usually evaluate junk bond opportunities by comparing
the price of the bonds to the risk-free rate of interest for the credit involved. This means adding
the default loss rate (i.e., default rate adjusted for recoveries) for foreign bonds to the risk-free rate
(e.g. US Treasuries of comparable maturities), so that a dollar bond issued by the government of
the Philippines, for example, might be priced at 2)1.% over comparable US Treasuries to compensate
the investor for the higher default risk - commonly called a 'haircut'. In the case of emerging
market corporate bonds, very little default rate data existed. Instead, investors would look at the
country's prevailing sovereign bond yields and would add to it a default rate expectation based
on the corporate credit risk involved, a sort of 'crew cut' against the risk-free rate.
94 R. C. Smith and I. Walter
'~, --j
0-1'0 ~-- . _1
I
i).l~ ~ ....... _............."._. . !
I b:merging Market Composite Index
I
'" I
Relativr to Major M.rket Composite Index
.......................................................j
I
i
.. I
...._.... .. .......................1
0.50 .
~I~--~--~----+----+----~--~---+----~--~---+--~--~
.1..,.81 Jun-81 000411 Mar-41Z Kov-8Z Apru Oct93 M..... Sop-Q4 F.WS A"lI.Q5 ".,96
RI.k('lfo)
..
Figure 5. Investor gains from emerging market equity portfolio allocation, 1987 (year-end) to 1995
(first quarter).
Our proxy for the major global equity markets is the US dollar-based MSCI World Index which
is composed of securities in the major markets worldwide, including the US. For emerging markets,
we use the US dollar-based IFC Emerging Markets Composite Index, an aggregate of activity in
emerging markets globally. Risk is defined as absolute volatility using the standard deviation of
month-to-month total returns. Average return is the annualized total return performance over the
designated period. We assume no adjustment for withholding taxes.
Source: Merrill Lynch Global Investment Strategy.
Professional fund managers have become somewhat used to this, and take
their punishment quietly when it comes, because the punishment tends to affect
many if not most of their rivals in the same way. After major price changes,
most fund managers will announce that they are looking for opportunities.
This occurred in the case of US junk bonds, mortgage-backed securities, REITs,
venture capital and other forms of relatively liquid high risk investments, but
it has not occurred in the case of large American LBOs or Japanese leveraged
investments. With the exception of bottom-fishing in specific 'oversold' issues,
this has not occurred in emerging market equities either, where the period of
abnormal risk-adjusted returns is unlikely to reappear within the next several
years.9 There are a number of reasons for this:
9 For example, the Mexican stock market reached a low of 1447.52 on the Bolsa index on
February 27, 1995 and had by mid-July recovered to about the same level (2231.11) it held on
December 20, 1994, the day before the initial peso devaluation that set off the crisis. The recovery
was mainly due to buying of undervalued shares by local investors, in addition to international
funds selectively rebuilding their Mexican portfolio weights and hedge funds trying to take advan-
tage of low share values. Lower US interest rates, lower Mexican inflation rates, and a new
Mexican sovereign debt issue may have contributed to market confidence as well. In dollar terms,
however, the converted Bolsa index at that point had only recovered from 2231.11 to about 1400,
due to the weakness of the peso.
96 R. C. Smith and 1. Walter
Investors who took large hits and lost considerable confidence in many
emerging equity markets are not likely to plunge back in again soon. They
will certainly be cautious about investing new money in such markets, and
will endeavor to get some of it out. They may exploit special 'buying
opportunities', but not in large volume. This has shown-up in a reduced
level of transactions in country funds and in emerging market manager
selections by pension funds and other institutional investors. Indeed, it was
already appearing in the US during the first three quarters of 1994 when,
according to the Securities Industry Association, combined purchases of
foreign securities had dropped nearly 90% from a year earlier, to the lowest
level in 4 years. The decline had been progressive, with the third quarter
down 69% from the second quarter, which in turn was down 43% from the
first quarter of 1994. For the full year 1994, combined US purchases of
foreign securities, net of sales, declined by more than 60% from the record
levels achieved in 1993.
Investors will be careful not to assume low correlations with the major markets.
As an investment class, emerging markets demonstrated a much higher correla-
tion with each other during the early part of 1995 than had been previously
experienced. The Mexican crisis clearly spilled-over onto many other emerging
equity markets, adversely affecting their performance as well. This demon-
strates a substantial erosion of the IPD gains associated with emerging market
investments, at least in the short-term. As inter-market correlations shot up,
it was revealed that the eggs being placed in different baskets that were in
fact all tied together. The risk-return characteristics thus appeared to be
substantially altered over a period of just 3 months and triggered by events
in only one emerging market country, albeit an important one. This may
have done lasting damage to the confidence of investors and fund managers in
the underlying value of emerging market equities in international portfolio
diversification, and was repeated in 1997 in Asia.
Investors now appear to realize that there are more risks in securities traded
in inefficient emerging markets than were originally considered. These
include vulnerability to shock-effects from sudden political or economic
developments, lack of hedging vehicles to manage risk, and the effect of low-
quality information disclosure by companies and governments upon which
to base investment decisions. Indeed, disclosed information in many emerg-
ing markets is often nearly useless, being late or highly misleading. Auditing
and accounting standards are often well below OECD standards, and in
many cases few regulations protecting investors' rights exist or are enforced.
Accordingly, the best investment information comes from insiders or specula-
tors or those 'in the know', which makes these markets much more suscepti-
ble to market rigging, fraud and corruption - as one Hong Kong investor
was quoted as saying: "Invest in anything where I don't have some kind of
inside information? Don't be silly!" (Smith and Walter, 1997).
Liquidity in many emerging markets is also well below the minimum stan-
dards prevailing in developed countries. Often there are virtually no investing
Rethinking emerging market equities 97
The simple lesson for investors in emerging equity markets appears to be that
asset allocation to these markets can be justified only when the acquisition
price is low enough to provide a margin to cover the true incremental risks,
and only when the investor has the ability, the confidence and the patience to
ride out the market panics or periods of great illiquidity that occur from time
to time. Surely there will be many investors who will continue to participate
actively in these markets, but they will tend to be specialists who are realistic
about the prices they are willing to pay. Those who are not specialists are most
likely to stay away or reduce their commitments.
We thus expect that global investors will be much more tough-minded about
emerging equity markets in the future. High-performance fund managers will
look elsewhere for assets, and emerging market country funds will continue
trading at substantial discounts from their net asset values, to which they
reverted in 1994-95.11 Foreign portfolio equity capital will revert to being a
scarce commodity in many developing economies, and the term 'emerging
markets' itself may fade away, its particular time in the colorful lexicon of
finance having passed. Countries too, therefore, will have to adjust to new
circumstances. What should equity investors look for in the next phase of
the process?
Sound economic policies. The easy-money conditions of the early nineties
are over, so foreign capital must be must be competed-for by countries on
the basis of good, long-term, risk adjusted investment returns. Providing an
economic environment that holds out adequate prospects of such returns is
a big job that includes many changes and reforms, and one that by no
means should be shrugged off with the notion that legislation is planned to
take care of this or that. Investors care about what actually happens, not
whether legislative bills are passed or not. Nor is there much use complaining
about foreign investors serving as self-appointed 'judge and jury' of the
efficacy of national economic policies. These are real people with hard-
earned money invested in assets they hope make sense for them, or entrusting
others to do so on their behalf relying on real political and economic analysis
by people whose careers are on the line.
Building financial infrastructure. Governments that want to attract interna-
tional portfolio investment must be clear about the priority need to create
some basic preconditions for viable capital markets - an obvious point
honored as much in the breach than in practice (Walter, 1993b).
Fundamental is a functional financial system that embraces a viable banking
industry,12 insurance and securities industries,13 pension and mutual funds. 14
11 On July 4, 1994 the average premium over net asset value of 32 emerging market mutual
funds was 28.4%. A year later, the same funds traded at an average discount of 6.9% to net asset
value. Barron's, price quotations for closed-end funds, July 4, 1994 and June 20, 1995.
12 Banks in many emerging market countries are all too often large, subsidized bureaucratic
institutions that possess few skills in finance and drive customers to transact in parallel (unofficial)
markets. Many are loaded-down with nonperforming loans from state-owned enterprises or large
domestic corporate combines deemed 'too big to fail'. The worst of this debt ultimately will have
to be separated from the banking system and put into 'bad banks', from which future recoveries
might someday be paid. The bad bank in such a country, possibly a subsidiary of the central bank,
can 'purchase' impaired loans from commercial banks using government bonds. Thus recapitalized
and solvent, banks can begin again to develop a viable lending business.
13 This involves providing a central market place, a trading system that includes rules for price
disclosure and settlements, and rules providing for the fitness and capitalization of securities firms
dealing with the public. The role of banks in the securities industry must also be determined, as
well as the extent to which the participation of qualified foreign firms is to be permitted. Foreign
firms (often through joint ventures) can contribute considerably to the training of employees and
management of local firms, and to the general professionalism and efficiency of national financial
systems.
14 Some developing countries have also created short-term markets in government securities
and commercial paper, in tandem with banking activities, as a competitive alternative borrowers
and depositors. Countries like Korea, the Philippines and Colombia have had domestic commercial
paper markets in operation for twenty years or more, while Poland has recently created one.
Rethinking emerging market equities 99
1~ For example, much of the recent distress in the Chinese securities market, particularly that
involving the Shanghai bond futures market, have been associated with insufficient regularity and
enforcement powers.
16 Others, such as Russia and the former Czechoslovakia, rushed-through privatization programs
in the interest of quick reform, but on a basis that may ultimately prove to be self-defeating. None
of the foregoing conditions for public ownership were in place, few of the enterprises were
economically viable in their own right - or depended on continued government subsidies or public
procurement to continue in business - management was not substantially improved, and the
process of ownership-distribution through vouchers was rife with fraud, corruption and racketeer-
ing. It is difficult to see how ordinary citizens will benefit from such privatization efforts if left with
worthless shares while the valuable ones fall into the hands of the well-connected or corrupt.
100 R. C. Smith and 1. Walter
inflows on many countries has often been a glut of foreign exchange and
liquidity, which can have severely adverse effects. Principal among these is
inflationary pressure, caused by a sudden, substantial increase in the money
supply, and appreciation of real exchange rates. Imports in some countries
subjected to such inflows consequently increased, exports declined, and trade
balances deteriorated. Some governments, such as Chile, South Africa and
several Asian countries, have limited portfolio capital inflows in various
ways to avoid the problem of excess liquidity and to maintain a competitive
exchange rate. Without such controls, the impact of massive portfolio flows
is hard to counteract. I7 It may also be worth considering whether foreign
portfolio equity investments via mutual funds should be tapped using closed-
end rather than open-end funds. In closed-end funds, the shock of investor-
demand shifts is taken by secondary-market prices of the funds in the
developed-country stock markets rather than by massive, destabilizing,
cross-border financial flows.
What about the IPD benefits of emerging market equity investments?
Despite the dramatic spike in inter-market correlations in the period immedi-
ately before and after the Mexican crisis, long-term correlations with major
markets remain extremely low. Table 1 shows the weekly total return correla-
tions in dollar terms between the Standard & Poors 500 index and various
emerging market indexes during the 3-year period February 1993 to January
1996 period. The efficient frontier mapping risks and returns between the
Morgan Stanley major market index and the IFC emerging index snapped-
back to the same contour it had before the 1994-95 Mexican crisis, as shown
in Figure 6.
17 In Chile, such controls in effect seek to increase the cost of investment by imposing reserve
requirements on loans, stamp taxes on securities transactions, and widening the bands within
which the currency can fluctuate .. Of the countries which experienced increased equity market
prices in 1994, as against emerging market trends at the time, most maintained restrictions on
capital inflows.
Rethinking emerging market equities 101
Table 1. US-emerging market correlations, February 1993-January 1996 (weekly price returns in
dollar terms)
Asia
China -0.06
Hong Kong 0.23
India 0.03
Indonesia 0.05
Korea -0.01
Malaysia 0.08
Pakistan -0.01
Philippines 0.01
Singapore 0.14
Sri Lanka -0.09
Taiwan 0.08
Thailand 0.20
Latin America
Argentina 0.36
Brazil 0.25
Chile 0.18
Colombia -0.04
Mexico 0.26
Peru 0.22
Venezuela 0.18
Africa
Nigeria 0.06
South Africa 0.07
Zimbabwe -0.04
come into play, and investors respond to revised expectations against current
asset prices. These observers know that those who get in early are the ones
who make money, but for them the risks were substantial at the time because
no respectable consensus had yet been formed in Washington, New York or
anywhere else to make everyone feel comfortable and, indeed to pave the way
for many new investors. They also knew that the formation of broad-gauge
agreement about such complex and elusive subjects as economic growth
102 R. C. Smith and I. Walter
15.0 ..----....------.-----.-----.-I-----r------clr----"I
No Change In Major
~::~~~;gkln~";!~ Major
I !
.j
~ _ .I!! LLll
_~.j... ~l~:w_w____ 100% In Emerging ___ _
-EFJc"" . . .~
13.0
,0 i
, !
I! Equity Markets
;;
:r
.2'
;;
~
~
11.0 'K"",' i ____-'-
+-- 1100% in Major Global Equity Markets 1
9.0~---~---~---~---~---_4---~---~
3.6 4.0 4.4 '.8 5.2 5.6 6.0 6.4
Risk (%)
Figure 6. Investor gains from emerging market equity portfolio allocation, 1987-1995 (year-end).
The proxy for the major global equity markets is the US dollar-based MSCI World Index which
is composed of securities in the major markets worldwide, including the US. For emerging markets,
we use the US dollar-based IFC Emerging Markets Composite Index, an aggregate of activity in
emerging markets globally. Risk is defined as absolute volatility using the standard deviation of
month-to-month total returns. Average return is the annualized total return performance over the
designated period. We assume no adjustment for withholding taxes.
Source: Merrill Lynch Global Investment Strategy.
extent. The hard work is in sticking with the free-market policies until they
produce some of the expected results. Its a long term process, usually accompa-
nied by some policy backsliding and victimization by world economic events
outside the control of individual countries. The new paradigm, if there is one,
is to adopt the policies and simultaneously to build the infrastructure necessary
for them to provide the greatest value, i.e., to attract foreign capital from a
competitive marketplace and to retain it.
Recent World Bank evidence (Levine, 1996) suggests that the development
of local equity markets plays a critical role in the economic growth process:
Countries that had more-liquid stock markets in 1976 tended to grow much
faster over the next 18 years than those which did not.
High levels of stock market liquidity, measured by the turnover ratio (trading
volume divided by market capitalization) tends to be associated with more
rapid growth over the same period.
Countries with high trading-to-volatility ratios likewise tended to grow
faster, after controlling for conventional economic, political and policy vari-
ables associated with growth differentials for various periods and country
samples. Volatility per se does not seem to be related to growth, but rather
the ease with which stocks can be traded.
Stock market development seems to complement (rather than substitute for)
bank finance, both of which seem to promote growth independent of each
other. Higher levels of development of the banking system are associated
with faster growth no matter what the state of development of the stock
market, and vice-versa, for reasons that are not yet well understood.
Although most corporate investment in developing countries is finance
through bank loans and retained earnings, both (along with the debt-equity
ratio) are positively associated with stock market liquidity.
Such findings suggest that international portfolio capital flows may playa
substantially more important role in the emerging market growth process than
previously thOUght. They can contribute disproportionately to market liquidity,
especially in the presence of 'noise traders' such as open-end mutual funds
which must sell in response to new client investments and redemptions and
maintenance of portfolio weights. They can force securities prices into line with
those prevailing on global markets. They can encourage upgrading of the legal
infrastructure, trading systems, clearance and settlement utilities, information
disclosure and accounting standards, and custody services. They can improve
the process of corporate governance, perhaps in association with significant
shareholdings by banks. They can also serve as a bellwether for local portfolio
investors, who may find encouragement from a significant foreign presence in
the marketplace.
Very few emerging market countries have the economic capacity or the
political will to adopt far-reaching free-market policies all at once. Who does?
Gradual approaches work, however, perhaps best of all. Successfully rebuilt,
former developing countries like Japan, Germany, South Korea, Taiwan,
Singapore, Spain, Chile at no point adopted a totally free-market approach.
104 R. C. Smith and I. Walter
They moved purposefully in that direction, but only at a pace that could be
accommodated by the accompanying political thinking and infrastructure-
building. Other countries that have tried hard to accept the new policies
(Mexico, Argentina, Brazil, perhaps India) have had considerable success
despite some recent disappointments. They need more time for their efforts to
bear the fruit that their more successful peers have enjoyed, but so far it appears
there is little likelihood of a reversion to the pre-Washington Consensus era in
any of these countries. The most powerful part of the new paradigm, however,
is that countries must know that it all depends on them. A consistent barometer
of their efforts, flawed as it may be from time to time, is in the capital they are
able to attract. IS
While it may be true that foreign capital can be attracted by offering assets
for investment at extremely low valuations, that is not the way most countries
are going to want to do it. Earning higher valuations will depend on being
able to create a credible environment for open market-driven institutions and
practices to develop. Some may instead attract investment simply by encourag-
ing GNP growth so that the funds will flow in despite other factors. This
appears to be the current Chinese approach. But it will certainly be a fragile
one, and regardless of the economy's impressive growth investors are not going
to take the country seriously until China changes its basic investment climate.
Investors have other places they can go instead. India today may well be one
of these. In the end, China may need foreign capital so badly that it will adjust
its approach.
For now, global equity investors have retreated from emerging markets after
price collapses shattered what may have been excessive confidence in their near
term future. From now on, money may be harder to come by, but it will still
be available to those developing countries which follow basic economic strate-
gies consistent with long-term growth and build sound and durable domestic
financial infrastructures. Countries want to attract long-term, patient investors
who are interested in harvesting suitable risk-adjusted returns as a reward for
careful selection of countries, industries and companies, and bearing the risks
over substantial periods of time.
Governments of emerging market countries have to face reality. To attract
capital one has to compete for it in a world market that offers many opportuni-
ties to those with funds to invest. To compete successfully, countries have to
do what is good for themselves in the long run, yet often difficult and politically
controversial in the short term. They have to fully adopt market economics,
satisfy a number of preconditions and then rebuild or build de novo the
banking and non banking financial institutions that are the core of a workable
18 It also depends on where a country is starting from - how much market infrastructure already
exists. In Eastern Europe, very little existed and without it large institutional change has been
almost impossible to undertake despite some promising signs of entrepreneurial development. Even
so, major differences exist between the available infrastructure in the Czech Republic and Russia,
between Poland and East Germany, and between countries within this region and China
Rethinking emerging market equities 105
REFERENCES
Corbo, Vittorio and Leonardo Hernandez (1994). Macroeconomic adjustment to capital inflows.
World Bank Policy Research Working Paper No. 1377.
Krugman, Paul (1995). Dutch tulips and emerging markets. Foreign Affairs, July/August.
Levine, Ross (1996). Stock markets: a spur to economic growth. Finance and Development. A
summary of twelve papers presented at a World Bank conference on Stock Markets, Corporate
Finance and Economic Growth.
Littler, Graeme and Ziad Malouf (1996). Emerging stock markets in 1995. Finance & Development,
March.
Micropal (1994). World's 100 largest emerging market equity funds. Micropal, April 1994.
Powers, Mary (1993). Soros fund manager says Peru has turned around. Reuters Asia-Pacific
Report, July 23.
Salomon Brothers (1993). Emerging Market Borrowing: Lessonsfrom History. New York: Salomon
Brothers Inc.
Smith, Roy C. and Ingo Walter (1993). Bank-industry linkages: models for Eastern European
economic restructuring. In Christian de Boissieu (ed.) The New Europe: Evolving Economic and
Financial Systems in East and West. Amsterdam: Kluwer.
Smith, Roy C. and Ingo Walter (1996). Rethinking emerging markets. Washington Quarterly.
January 1996.
Smith, Roy C. and Ingo Walter (1997) Street Smarts. Boston, Harvard Business School Press.
Walter, Ingo (1993a). The Battle of the Systems: Control of Enterprises in the Global Economy. Kiel:
Institut fUr Weltwirtschaft.
Walter, Ingo (1993b). Cross-border equity flows: tapping into global markets. ASEAN Economic
Journal, October.
GEERT BEKAERTl, CLAUDE B. ERB2, CAMPBELL R. HARVEy3
and TADAS E. VISKANTA 4
1 Stanford University, 2.4 First Chicago Investment Management Co., 3 Duke University
ABSTRACT
The behavior of emerging market returns differs substantially from the behavior of developed
equity market returns. We show that these differences have persisted in the period ending
March 1996 but, at the same time, document how some salient characteristics of emerging
markets vary through time. Finally, we offer some ideas on the forces that drive the cross-
section of returns, volatility, skewness, kurtosis and correlation in emerging milrkets and
detail the implications for asset allocation.
INTRODUCTION
Second, we have learned that the emerging market returns are more predict-
ble than developed market returns. Harvey (1995) details much higher explana-
tory power for emerging equity markets than developed market returns. The
sources of this predictability could be time-varying risk exposures and/or time-
varying risk premiums, such as in Ferson and Harvey's (1991, 1993) study of
US and international markets. The predictability could also be induced by
fundamental inefficiencies. In many countries, the predictability is of a remark-
ably ~imple form: autocorrelation. For example, Harvey (1995) found a 0.25
autocorrelation coefficient for Mexico in a sample ending in June 1992. An
investor who followed a strategy based on autocorrelation in this country
would have lost 35% like everyone else in December 1994. However, the
investor would have been completely out of the market in the next 3 months
(or shorter if possible). Momentum appears to be important for many of these
markets.
Third, we have learned that the structure of the returns distribution is
potentially unstable. Garcia and Ghysels (1994) reject the structural stability
of the prediction regressions presented in Harvey (1995). These regressions
allow for the influence of both local and world information. Bekaert and
Harvey (1995, 1996a) present a model which explains the results of Ghysels
and Garcia. The Bekaert and Harvey model allows for the relative influence
of local and world information to change through time. They hypothesize
that as a market becomes more 'integrated' into world capital markets, the
world information becomes relatively more important. Bekaert and Harvey
(1996a) found that the changing relative importance of world information also
influences volatility.
Fourth, the Bekaert and Harvey (1996a) framework suggests that the increas-
ing influence of world factors on emerging expected returns may manifest itself
in increased correlation with developed market benchmarks.
The goal of this paper is to explore three aspects of the emerging markets
data. First, we examine the behavioral characteristics beyond the volatility, the
skewness and kurtosis. Second, the paper explores the relation between risk
variables and expected returns. Harvey (1995) and Bekaert (1995) find that
higher values of beta (from a capital asset pricing framework) are associated
with lower expected returns. This is the opposite from what we would expect
from theory, however, it is consistent with these markets being segmented. That
is, the countries with the higher beta values are more likely to be integrated,
and hence have lower expected returns relative to the segmented countries.
Third, we examine the time-varying correlation of these markets with developed
markets. Longin and Solnik (1995) and Erb et al. (1994) describe how correla-
tions change through time in developed markets. Harvey (1993,1995), Errunza
(1994) and Bekaert and Harvey (1996a) show some evidence that correlations
are changing in emerging markets. Finally, we examine what is important for
explaining both the cross-section of expected returns and volatility in emerging
markets. Following Erb et al. (1996b), we try to link political, economic, and
The behavior of emerging market returns 109
The two main sources of emerging market benchmarks are the International
Finance Corporation (IFC) and Morgan Stanley Capital International
(MSCI).5 Both provide country benchmark indices which are based on a value
weighted portfolio of a subset of stocks which account for a substantial amount
of the market capitalization within each emerging market.
The IFC produces two types of indices: Global (IFCG) and Investable
(IFCI). For nine countries, data exists back to 1976. Currently, the IFC provides
data on 27 countries. 6 MSCI also produces both Emerging Markets Global
(EMG) and Emerging Markets Free indices (EMF) which resembles the IFCI.
Our paper focuses on the global indices. Part of the interest in studying
emerging markets is the impact capital market liberalizations have on the
returns. Hence, we study markets before and after they are accessible by
international investors. 7
IFC and MSCI use a different hierarchical process in the company selection
for the country indices. MSCI follows the same technique that it uses in its
popular developed country indices. First, the market is analyzed from the
perspective of capitalization and industry categories. Next, a target of 60%
coverage of the total capitalization of each market, with industry weightings
approximating the total market's weightings is established. Finally, companies
are selected based on liquidity, float, and cross-ownership to fulfill these goals.
In contrast, the IFC's order of preference is size, liquidity and industry. The
IFC primarily targets the largest and most actively traded stocks in each
market, with a goal of 60% of total market capitalization at the end of each
year. As a second objective, the index targets 60% of the trading volume during
the year. Industry is of tertiary priority. Although there is some hierarchical
differences in the structure of construction, there is little difference in the
behavior of the IFCG and the EMG. Table 1 details the difference between the
IFCG and the EMG returns over identical samples for each index. Of the 22
countries where there is MSCI and IFC data, the returns indices have greater
5 Barings also provides the Barings Emerging Market Indices (BEMI). However, we choose to
EMF indices is 91.8%. For the period April 1991-March 1996, the correlation between the IFCI
and the MSCI EMF is 97.2% The correlation between ENG (EMF) and the MSCI World-All
Countries is 41% (49%).
.....
.....
Table 1. Comparison of IFC and MSCI emerging market global indices. 0
(1:0
Argentina Jan. 88 -10.6 -2/.1 61.9 0.76 0.01 -0.06 0.19 0.04
Brazil Jan. 88 1.4 2.3 19.4 0.96 0.22 0.23 0.28 0.34 tl
--
:-
Chile Jan. 88 1.7 -0.8 7.6 0.96 0.08 0.05 0.32 0.36
Colombia Jan. 93 -4.5 -2.0 7.8 0.96 -0.04 0.04 0.16 0.05
Greece Jan. 88 7.0 0.8 11.9 0.96 0.18 0.13 0.05 -0.01
India Jan. 93 5.0 -1.5 6.6 0.98 0.02 -0.13 0.38 0.16
Indonesia Jan. 90 -1.4 2.2 9.1 0.96 0.11 0.15 0.37 0.45
Jordan Jan. 88 -4.2 -0.1 11.9 0.75 0.09 0.22 0.16 0.14
Malaysia Jan. 88 0.6 0.1 4.8 0.98 0.49 0.46 0.48 0.48
Mexico Jan. 88 1.4 -1.3 10.9 0.96 0.28 0.26 0.45 0.48
Pakistan Jan. 93 0.1 2.5 4.5 0.99 0.01 0.02 0.51 0.15
Peru Jan. 93 -0.1 2.6 7.0 0.99 0.45 0.44 0.46 0.46
Phillippines Jan. 88 2.3 0.0 9.7 0.95 0.37 0.36 0.50 0.47
Poland Jan. 93 -7.3 3.3 15.1 0.99 0.38 0.46 0.35 0.41
Portugal Jan. 88 -0.5 0.4 6.4 0.96 0.48 0.49 0.12 0.14
South Africa Jan. 93 -1.4 -0.2 3.2 0.99 0.35 0.35 0.48 0.53
South Korea Jan. 88 -0.2 0.3 6.7 0.97 0.37 0.35 0.38 0.38
Sri Lanka Jan. 93 2.5 -0.8 5.1 0.99 0.00 0.00 0.36 0.35
Taiwan Jan. 88 1.0 -1.1 7.2 0.99 0.21 0.22 0.82 0.81
Thailand Jan. 88 0.4 0.6 5.3 0.99 0.36 0.33 0.48 0.47
Turkey Jan. 88 2.8 -2.5 22.1 0.94 0.02 0.01 0.20 0.24
Venezuela Jan. 93 -4.0 -2.0 9.1 0.98 -0.03 -0.12 0.19 -0.15
Average -0.4 -0.8 11.5 0.95 0.20 0.19 0.35 0.31
Composite: (FC global composite; AC world: MSC( all country world index.
Source: (FC global indices, MSCI EM indices. Monthly returns in US dollars.
112 G. Bekaert et al.
Latin America
Argentina 35 22307.8 2.0 31 22161.1 3.5
Brazil 86 93939.6 8.5 68 63813.7 10.2
Chile 47 39421.3 3.5 43 39019.3 6.2
Colombia 28 6658.9 0.6 15 5334.2 0.9
Mexico 81 65162.4 5.9 65 58686.5 9.3
Peru 36 7421.7 0.7 20 6910.0 1.1
Venezuela 16 2652.3 0.2 5 1930.8 0.3
East Asia
China 171 29494.8 2.7 23 3005.5 0.5
Korea 151 125037.1 11.2 145 17314.7 2.8
Philippines 46 39729.2 3.6 35 19314.9 3.1
Taiwan, China 83 114474.5 10.3 83 17504.9 2.8
South Asia
India 131 71141.3 6.4 76 14792.2 2.4
Indonesia 45 54570.7 4.9 44 27724.6 4.4
Malaysia 123 162134.5 14.6 123 135326.0 21.5
Pakistan 68 6646.6 0.6 25 4951.0 0.8
Sri Lanka 44 1314.9 0.1 5 456.5 0.1
Thailand 73 95035.9 8.5 72 30821.1 4.9
EMEA
Czech Republic 69 12346.3 1.1 5 5206.7 0.8
Greece 53 11199.7 1.0 47 10623.8 1.7
Hungary 16 2957.4 0.3 8 2544.8 0.4
Jordan 51 3276.3 0.3 8 1107.4 0.2
Nigeria 35 1712.4 0.2 0 0.0 0.0
Poland 23 3892.8 0.4 22 3874.5 0.6
Portugal 30 11404.5 1.0 26 9012.0 1.4
South Africa 63 105981.4 9.5 63 105981.4 16.9
Turkey 54 20641.3 1.9 54 20641.3 3.3
Zimbabwe 24 1677.0 0.2 5 370.4 0.1
Regions
Composite 1682 1112232.6 100.0 1116 628429.1 100.0
Latin America 329 237564.0 21.4 247 197855.5 31.5
Asia 935 699579.5 62.9 631 271211.4 43.2
EMEA 394 175089.1 15.7 238 159362.2 25.4
is sharply lower than the average returns detailed in Harvey (1995) reflecting
the 80% drop in the Mexican market over the period December 1994-February
1995. Over the past 5 years, the average return in Venezuela was negative.
Over this period, the average return of the IFC Composite was similar to the
MSCI World and the MSCI World-All Countries. The difference between the
emerging markets is in the volatility. The IFC Composite had a volatility of
Table 3. Summary statistics, April 1991-March 1996.
GMM Bera-
Arithmetic Geometric Standard normality Jarque Kolomorogov- First Beta Beta Beta
Start return return deviation test test Smirnov order MSCI MSCI lFCG
Country date (%) (%) (%) Skewness Kurtosis p-value p-value .<ttatistic autocorrel world AC world compo8ite
Argentina Apr. 91 35.5 25.6 56.7 3.09 16.92 <0.01* <0.01* 1.17* 0.06 1.53 1.69 0.91
Brazil Apr. 91 44.3 37.1 52.2 0.89 1.51 <0.01 <0.01* 1.06* 0.09 1.19 1.43 1.38
Chile Apr. 91 24.7 23.6 26.6 0.35 -0.43 0.Q2 0.41 0.73 0.23 0.09 0.20 0.64
China Jan. 93
Colombia Apr. 91 40.5 39.0 40.1 1.33 2.02 <0.01* <0.01* 0.95* 0.54 -0.05 -0.01 0.25
Czech Republic Jan. 95
Greece Apr. 91 -2.7 -5.5 24.1 -0.34 0.06 0.41 0.57 0.64 0.16 0.52 0.54 0.28
Hungary Jan. 93
India Apr. 91 12.6 6.3 36.8 0.67 1.47 0.07 0.02* 0.91* 0.25 -0.60 -0.49 0.69
Indonesia Apr. 91 9.5 5.5 29.2 0.15 0.09 0.84 0.89 0.49 0.18 0.58 0.72 1.04
Jordan Apr. 91 10.0 9.4 14.3 0.34 -0.80 <0.01* 0.24 0.95* 0.07 0.1I 0.13 0.11 ;;!
Malaysia Apr. 91 20.4 19.1 23.9 -0.06 1.03 0.31 0.40 0.48 -0.16 0.57 0.67 0.87 <:)-
""
Mexico Apr. 91 \5.7 8.5 37.5 -1.01 2.04 <0.01 <0.01* 0.54 0.33 0.83 1.08 1.40
Nigeria Apr. 91 37.5 12.7 69.5 U9 11.92 0.07 <0.01* 2.02* -0.04 U5 U4 -0.05 ""::s-
~
Pakistan Apr. 91 22.4 18.1 35.3 1.03 1.92 <0.01 <0.01* 0.88* 0.30 0.06 0.15 0.57 '"c
Peru Jan. 93 .....
Philippines Apr. 91 24.1 22.4 28.7 1.37 4.47 0.08 <0.01* 0.94* 0.Q2 0.54 0.69 1.l0 <Q.,
Poland Jan. 93
Portugal Apr. 91 8.3 6.6 19.6 0.41 1.55 0.25 0.05* 0.51 om 0.98 Lot 0.25 ""~
South Africa Jan. 93 ""
~
South Korea Apr. 91 7.8 4.7 26.0 1.07 1.75 <0.01 <0.01* 0.95* 0.04 0.41 0.51 0.68 S
~
Sri Lanka Jan. 93
Taiwan Apr. 91 7.3 1.4 37.1 2.24 7.65 <0.01* <0.01* 1.12* 0.08 0.69 0.88 1.42 ~
~
Thailand Apr. 91 20.1 17.2 30.0 1.08 1.85 <0.01 <0.01* 0.87* 0.03 0.20 0.35 1.14 .....
;>;-
Turkey Apr. 91 13.6 -3.9 61.2 0.64 0.42 <0.01 0.13 0.68 0.08 -0.12 -0.01 0.87 ....
Venezuela Apr. 91 -7.0 -16.8 45.8 -0.45 1.91 0.15 0.01* 0.65 -0.25 0.45 0.54 0.57 "".....
Zimbabwe Apr. 91 4.8 -l.l 34.9 0.38 0.87 0.39 0.27 0.65 0.32 0.84 0.90 0.55 ....
""
:::
.....
MSCI world Apr. 91 11.2 lU 10.7 -0.32 -0.34 0.01 0.50 0.48 -0.25 1.00 1.01 0.20 ;::!
MSCI AC world Apr. 91 11.0 11.0 10.5 -0.23 -0.39 0.03 0.59 0.44 -0.22 0.98 1.00 0.25 '"
IFCG composite Apr. 91 11.4 10.6 16.4 0.93 3.47 0.03 <0.01* 0.71 0.40 0.47 0.62 1.00 ......
......
w
Composite: IFC global composite; AC world: all country world index.
Source: IFC global indices, MSCI EM indices. Monthly returns in US dollars.
* Significant at 0.05 level based on empirical distributions. Kolomorogov-Smirnov empirical critical value: 0.835.
114 G. Bekaert et al.
16.4% compared to the MSCI World-AC volatility of 10.5%. Hence, over the
past 5 years, the contribution of a diversified emerging markets investment to
a diversified global portfolio must have come from the correlation properties.
Figure 1 presents rolling 5-year average returns for the 20 emerging markets
and the three benchmark indices. The 're-emergence' effect of Goetzmann and
Jorion (1996) appears evident for six countries in particular: Argentina, Chile,
the Philippines, Portugal, Taiwan and Turkey. For these countries, the average
returns in the 5 years after emergence in the IFC database are much higher
than the subsequent 5 years. However, there are a number of exceptions, with
no such pattern in Brazil, Greece, Colombia, Mexico, Nigeria, Pakistan, South
Korea, Thailand, Venezuela, and Zimbabwe. Overall, the evidence for the
re-emergence effect is mixed. These figures suggest that the mean returns are
time-varying. The evidence presented in Harvey (1993, 1995), Bekaert (1995)
and Bekaert and Harvey (1995) suggest that emerging market returns are more
predictable than developed market returns. While rolling 5-year mean returns
are useful descriptors of the data, the evidence on predictability suggests that
time-varying means are best captured by regression models.
Bekaert and Harvey (1995, 1996a) suggest that care must be taken in
specifying the prediction model. In particular, if a market experiences increased
(or decreased) integration into world capital markets, it is likely that the
parameters of the prediction model change through time. Bekaert and Harvey
propose models where the influence of world versus local information changes
with the degree of integration. That is, as a market become more integrated
into world capital markets, it is more likely that world information will have
a greater impact on the time-varying mean returns.
The final panel of Figure 1 shows mean returns in the 1980s and 1990s.
Most of'the capital market liberalizations took place before 1992. The graph
shows that the mean returns in many countries are much lower in the 1990s
compared to the 1980s. For example, the four countries who had greater than
65% returns in the 1980s all had less than 25% returns in the 1990s.
Changing volatility
Argentina
Total Return
2' 140%
~ 120%
J100%
.I...
80%
j 60%
40%
f 20%
~
f
C 0%
-20%
-40% ,...
CXl CXl ;b CXl CXl CXl CXl
N
i <Xl 0> N
C') It) 0 C') It)
0> Cii 0> 0> ~
<C <C
CXl 0> 0>
0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0>
..- ..- ..- ..- ..- ..- ..- ..- ..- ..- ..- ..-
Brazil
Total Return
r 80%
~
Iii 60%
f.
....~
j
40%
:8 20%
E
~ 0%
~
l-20%
i N C') -.:t It) ,...
<Xl 0> 0 N -.:t C') It) <C
<Xl <Xl <Xl <Xl en Cii en en en en en
<C
<Xl <Xl CXl CXl
'"
..- '" '" '"
..- .- '" en
.- '"en
..- .- .- .- '" '"
en .- .- en
en .- .- '" '" '"
Chile
Total Return
2' 100%
~ 80%
!
....I 60%
j 40%
20%
i
f
-20% -
0%
1&
~ -40%
.- N C')
;b It) <C ,... <Xl en 0 ... N C') -.:t It) <C
<Xl
en .-
.- en
<Xl CXl
~.~
CXl
~ .- .-
CXl
'"
en .-
<Xl
Indonesia
Total Return
:i'"
15%
~
Iii 10%
~
~,
5%
J
~
~
u 0% I
II
E
~
II
-5%
'"
-
J: -10%
co co '"
~ Nv
co co co '"
r-- co
co co co '"
L()
c;; '" v
0 N to
'" '" '" '" '" '" '" '" '" '"'" '" '"'" '"'" '"'" '"'" '"'"
co
~
L()
Jordan
Total Return
.~20%
&
~ 15%
~
.;, 10%
Iu 5%
1l
E
:5
.,
~ 0%
~
J: -5%
a; '" co co '"co cor-- coco '"co 0'" c;; '" '"'" '" '" '"'"
N
co co
V L() N V L()
'" '" '" '" '" '" '" '" '" '" '" '" '" '" '"
~
'" ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Malaysia
Total Return
~35%
~
Iii 30%
~
~ 25%
E
~ 20%
u
""~ 15%
lf 10%
Iii
J: 5%
v
co co '"
to r-- co
'" v
'" '" '" '" '" '" '" '" '"'" '"'" '" '"~ '"'" '"'" '"'" '"'"
~ N L()
co co co co co co co
0
c;; N to L()
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Figure 1. (Continued.)
.."
~.
00
--
....~
o
~g.
Average Arithmetic Return - Five Yeer RaUlng A_age Arithmetic Return - Five Ye.. Railing
~
;>;-
Average Arithmetic Return - Five Ye. RoIling , I:)
! c.n 0 tn ~ ~ ~ ~ ~ o 0 0 ~ ~ B ~ o B ~
"#. "#. "#. "#. "#. "#. "#. "#. "#. "#. "#. "#. "#. "#. "#. "#.
'" ~ ~ "#. ~ "#. ""......
~ll:I" ~I 1981 ...""
1981rr- I:)
1982 1982 1982 :-
1983 1983 . 1983
1984 1984
1985 1985
1986 1986 ~ ~
~
1987 Iit"tl 1987 litz 1987 lit 3:
1~1 -CD
1988 i~ 1988 C' 1988 ~ ~.
CD
!. CD ~ CD n
1989 -AI 1989 c:
-i _. AI 1989. i:o
5; ::s ~
...,
1990 ::s 1990 - ::s 1990 . ::s
1991 1991 . 1991
1992 1992 1992
1993 1993 1993
1994 1994 1994
1995 1995 1995
1996 1996 1996
..,
~.
;;1
~
~.
I:
Average Arithmetic Return - Five Vear Roiling Average ArithmeUc Return - FlveVe.. Roiling
, Average Arithmetic Return - Five Vear Roiling
I\.) to.> ..,. U1 1\.)..,. '"
~ ~ I\J U) ::.. 01 en ...... Q)
~ 0 0 0 0 0 0 0 0
'" o~ 0 0 000 0000000 0
*- *- *- *- *- *- *- *- *- *- *- *- *- *- *- *- *- *- *- *- *- *-
1981 1981 1981
1982 1982 1982 i_
1983 1983 1983
1984 1984 1984
1985 1985 1985
1986 o'g' 1986 0' 1986 ~ ~
1987 -I: 1987 --u
III 0
o'-u '='"
e!.g: -.., 1987 [E;
1988 1988 ~
:::U" :::ue- 1988 R :::u is" s:::.
CD CO <:::
1989 Soo 1989 I:
- I l_l ~ ~"
~ CiJ .., 1989 .,o
:::::I III :::::I ... CD
1990 1990 1990 :::::I rn .sa.,
1991 1991 1991 ~
~
1992 1992 1992 ~
1993 1993 ( 1993 ~ ~.
1994 1994 t 1994 ::!
1995 1995 s:::.
1995
~
1996 1996 L I_ _ _- - ' ' - -_ _ _ _ _ ~
1996 L-'_ _ _ _ _ _ _ ~--..-J
....,~
~
...
;:
~
'"
.....
.....
\C
..., .....
~. tv
~
o
.....
~
~ b::J
'" ~
Average Arithmetic Return - Five Vear Roiling ;0:;-
, Average Arithmetic Return - Five Vear Roiling Average Arithmetic Return - Five Vear Rolling $:)
~ ~ N W .::. c.n C') ...... ~
, ~
~
N .j>. C1I
0 0 0 000 0 0 0 0 0 0 0 0 0 0 '" N.J:!o. m [X)
....
0
'" a 0 0 00 0 ....
I ?ft- ::.e ?ft- '* '#. t/!. *' t/!. '#. ?ft- ?ft- ?ft- ~ ~ ?ft- ?ft- ?fl. "if!. '#. ;Ie '#. ?f!. ~
....
1981 1981 1981 $:)
1982 " :-
1982 1982
1983 " 1983 1983
1984 1984 1984
1985 1985 1985
1986 [ ~ 1986 ~ 1986
1987 ; ~
~i? 1987 iii;!
-Ill
1987 ~~
1988 ' :;0;- 1988 :;0= :;0 ~.
CD CD CD III 1988
1989 CD III
C'<
... 1989 -~
c: Q, 1989 C~
1990 ~
... ..,
1990 ~ 1990 ~
Venezuela
Total Return
I 80%
,.
~
;;;
60%
f
ii:
r:: 40%
.a
~
u 20%
i
fII
0%
f
-20% ,...
~
ex>
N
ex> '"ex> ~
ex>
II)
ex>
<C
ex> ex>
ex> en 0
ex> ex> en
N
c;;
en en '" it II)
en
<C
en
en en en en en en en en en en en en en en en
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Zimbabwe
Total Return
... 60%
~;;;
,. 40%
~,
c: 20%
.a
~
u 0%
i
E
E
<-20%
II
-
'"
j -40% ;;; Nex> ex> ;1i II) <C ,... ex> en 0
'" ex> ex> ex> ex> ex> en en en en it en en
N <C
'"
II) ~
en en ~ en en ~ en en en
~ ~ ~ ~
en en en en en en en
~ ~ ~ ~ ~ ~ ~ ~
MSCI World
Total Return
r 35 %
,.~ 30%
II
.~ 25%
II.
E20%
.a
~ 15%
1
,ho%
~
.
II
OJ
iii
5%
~ 0% ,...
ex> ex> ex> ex> ex> ex> ex> ex> ex> en c;; en CI) CI) en en
N
'" <C ex> 0 N
'" <C
~ ~ II) CI) ~ II)
en en en en en en en en en en en en
~ ~ ..- ~ ~ ~ en en
~ ~ ~ ~
Figure 1. (Continued.)
122 G. Bekaert et al.
i
S12%
~
, 10%
j
i 8%
If 6%
! 4%
N <"l It) co .... co (I) 0 a; N <"l ~ It) co
~ co co co co (I)
~
co
(I)
co co co (I) (I) (I) (I) (I) (I)
(I) (I)
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
~
II
35%
,! 30%
!
'725%
E
1 20 %
1 15%
! 10%
& 5%
II
~ 0%
a; N <"l It) co ....
co 0 a; N <"l It) co
...
~ (I)
~
co co co co co co co co
...'"
(I) (I) (I) (I) (I)
~
(I)
~
'"
~ ~ ~
(I)(I)
~
'"
~
(I)
~
'"
~
(I)
~ ~
(I)
~
(I)
~ ~
~100%
~
1i
80%
60%
40%
II: 20%
&
j O%rr~~~----------~
0% 20% 40% 60% 80% 100%
Average Raturn (1990:01-1996:03)
Figure 1. (Continued.)
~~
!""
~
Amuallzed Volatility - Five Year RoIling Annualized Volatility - Five Yea, Roiling Annualized Volatility - Five Yaa, Roiling
l ~ m m ~ ~ ~
~ 1:7: ~ !l: ~ ~ g: 8l o~ 0 0 0 0 0 0 .j>. en 00
[ '#. '#. '#. '#. '#. '#. '#. '#. '#. '#. '#. '#. '#. '#. '#. 0 0 0 ~
;- '#. '#. '#. '#.
OQ
1981 1981
1981
s 1982 1982
1982
1983 1983 ~'
fr 1983
a 1984 1984
1984 c
~
1985 c 1985
~, 1985
1986 1986
1986
;;1
(\)
g" c::::
sa, 1987 2,(") 1987 ~llJ 1987 ~~
_cc <::r
-~ ~
1988 I:Il =r 1988 I:Il I:Il I:Il CD
~ _. =N 1988 s::.
~ ~
Q.
1989
-
=Ci'
~ 1989 r ~= 1989
a~
~~ ...0'
1990 1990 1990
~ <Q.,
1991 1991 (\)
!! 1991 ;:I
1992 . (\)
1992 1992
~ ~
1993 1993 1993 ~'
1994 1994 1994 ;:I
1995 1995 1995 ...s::.
1996 fL'_ _ _ _ _ _ _ _ _--1 1996tL _ _ _ _ _ _ _ _~ 1996 ~
~
-
E'
~
'"
.-.
N
W
.." ..-
00' tv
... .j:>.
'"'"
!'>
~ 0
c
'"
5" Annualized Volatility - Five Yea, RoIling
~
;>:-
Annualized Volatility - Five Yea, RoIling Annualized Volatility - Five Yea, Roiling I:>
... NNW W ~
~ I\J W ".. 01
a 0'1 0 r.n 00'10 J\..) I\.) W W ~ ~
''""
.t: af!. -;;. fI!. tf!. if!. tf!. oe. a a a a a '"a o c.n 0 0'1 0 ::l
'"af!."'" af!. af!. af!. af!. af!. af!. af!.
'" af!. af!. af!. af!. af!. '"af!."'" ~
1981 1981 ....
1981
I:>
1982 1982 1982 .~
:-
1983 1983 1983
1984 1984 1984
1985 1985 1985
1986 1986 1986
1987 1987
~-
-:::I ~G> 1987 ~g>
1988 !!l.a. - ..,
I CD iii 0
1988 r~ 1988
== iii =CD ~3
1989 =0 _. C"
~ 1989 ~CD 1989
1990
~iii
1990 1990
1991 1991 1991
1992 1992 1992 .
1993 1993
1994 1994 Co
1995 1995
~::!
1995
1996 1996 1996
~~
~
~
'"g. Annualized Volatility - Five Year Rolling Annualized Volatilily - Five Year ROiling Annualized Volatilily - Five Year Rolling
N ~ ~ NNW W ~ ~ ~ ~ ~ ~ ~ hl N NNW W W W W
o ~ ~ m ~ 0 N W ~ rn m ~ ~ ~ 0 ~ W woo ~ ~ N
~ ~ ~ ~ * ~ ~ ~ * ~ # ~ ~ ~ ~ ~ ~ ~ *' ~ 'cf!. #. ?J? rf!.
1981 1981 1981
1982 " 1982 1982
1983 1983 1983
1984 1984 1984
1985 1985 1985
1986 1986 1986 ~ ~
tI>
1987 <s: 1987 1987 <5" <:r-
Q.e!..
a.c..
-0
o c.
-0
tI>
til .., ~
1988 til til 1988 t 1988 . ~::::l :::.
ct. s:g. =(\) <;
=(/1
1989 ~iii' 1989 ~::::l 1989 ~ ~. ...o
1990 1990 1990 <Q.,
tI>
1991 1991 1991 ~
tI>
1992 1992 1992
~
1993 1993 1993 ~.
1994 .~ 1994 Co
1994 ~
1995 ~ 1995 1995 ...:::.;.;-
1996 1996,~____________________~ 1996 ~ ......:::::; ~
....
tI>
....
;:
...
;::
'"
N
VI
-
..., ......
~.
N
s: 0\
~
~
r:; ~
c
:: ~
g. Annualized Volatility - Five Va"r RoIling
;>;-
0: Annualized Volatility - Five Year Rolling Annualized Volatility - Five Year RoIling l:)
~ ....a. l\) f\.) W w .J::!r. '" W ".. c:n Q) .....,
'"
.!:: 01 o 01 0 01 0010 ~ ~ ~ ~ ~ ~ ~ o 0 000 0 ......."'
0~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
1981 1981 1981 ...."'
1982 1982 1982 !?--
1983 " 1983 1983
1984 1984 1984
1985 1985 1985
1986 1986 1986
1987 ~"tI
-I
1987 ~z 1987 ~3:
I 25: iii cC' -CD
1988 cr.1/I 1988 cr.CD 1988 !!.><
=. C::;"
1989 ~~ 1989 ~ ~. 1989 ~O
1990 ~ 1990 1990 "
1991 1991 1991
1992 1992 1992
1993 1993 1993
1994 1994 1994
1995 1995 1995
1996 " 1996 fL...-_ _ _ _ _ _ _ _ _- - - ' 1996 "
The behavior of emerging market returns 127
Philippines
Volatility
45%
'"
.5
~
: 40%
)'
~
~35%
.
;;;
~
~ 30%
11
:>
~
25%
a; co '"
N
co """
It') I'- co 0>
'"
co co co co co co
0
0> c;; N
0> '"0>0> """0>0> It')
0>
'"
0>
0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0>
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Portugal
Volatility
60%
,
~ 50%
~
.
~40%
~
ii 30%
~
i
!.1:1 20% -
~
10%
a; '"
co co ~ co co I'-
N It') co 0>
'"
co co co
0
0> c;; N
0> '"en'" """'"
It')
0> '"en
0>
0>
~
0>
'" '" '"
~ ~ ~
0>
'" '"
~ ~
0>
~
0> 0>
~
0>
'" '"
~ ~ ~
South Korea
Volatility
45%
.~
~40%
~
~ 35%
.
~
;;;
~ 30%
::>
~25%
!!I
c
.l
20% ,... co 0> 0 c;; N
a; N
'"
co co """
It')
'"
co co co co co co '" c; It')
'"enen
0>
~
'"
~ ~ 0> ~ 0> 0> ~ 0> en ~ en en
~ ~ ~
0> 0>
~ '" ~ ~
'" ~
0>
0>
~
Figure 2. (Continued.)
...,
~. IV
00
~
!'-'
-
() 0
t:J:j
;:::
'"
:;. ~
Annualized Volatility - Five Year Roiling Annualized Volatility - Five Year Roiling ~
Annualized Volatility - Five Year Rolling !:l
~ --. J\J f\) W W
rn m m ~ ~ 00 ~ w ~ ~ (.h tn en m ..... ~
rn 0 ~ 0 ~ a 0 c.n 0010 01 0 CJ1 o en 0 U't 0 c.n 0 ....
! fF. 'of!. (/!. ~ 'if!. ?fe. :.e
0 ~ ~ ~
...
"#. *' *' ?/!.. *' *' ;f!. *' cf.. cI! '#. ~
1981 1981 1981 ...
!:l
1982 ' 1982 1982 c. :-
1983 1983 1983
1984 1984 1984
1985 1985 1985
1986 1986 1986
1987 ~~ 1987 <-I
o :r 1987 ~~
_t: -
II)I I__
)
II) ..,
1988 ,. =:~ 1988 r 1988 21:i-
=11) =11)
-- :::I
1989 ~~ 1989
--
~Q. 1989 ~:::1
1990 1990 1990
1991 1991 1991
1992 1992 1992"
1993 1993 1993
1994 i 1994 l
1994 1-
1995 . 1995 1995-
1996L[____________________~ 1996 1996
.."
""....'"
'"
!"
~
Annualized Volatility - Ave Yaar Rolling Annualized Volatility - Five Year Rolling
"5" Annualized Volatility - Five Year Rolling
~ ~
t\,) I'\) W W ~ .J::to. CJ'1
'"'" 0 ~ 0> 0> o
'" ( 11 o en 0 c.n 0 c.n 0 ~ ~ ~ t ~ &; g; ~ ~
~ 'iJ!. 'iJ!. 'iJ!. 'iJ!. 'iJ!. :.!!
o
-
'" 'iJ!. ~ ';Je '#. ?f!. ';Je '#. ?fi 'iJ!. ';Ii!. ';/t! ~ '* ';j!. '*' ?J!. '*
1981 1981 1981
1982 1982 1982
1983 1983 1983
Hl84 1984 1984
1985 1985 ~
1985
1986 " s:: 1986 1;c 1986 ~
t\)
1987 c:::::cn 1987 C <~ 1987 ! <~ <::l"
2.0 2.3 2.~
Q) - I C" Q) CD ~
1988 1988 =::r.1 1988 =::r. N \:>
_. 0
5!::e =:C" =:c <;
1989 1989 1989 (3
.:<" ... .:<"~ .:<"~ ...
CD I
1990 c:: 1990 1990
~
1991 1991 1991
~
1992" 1992 ,~ 1992 E
1993 1993 t
~
1993 ~.
1994 1994 1994
~
~
1995 1995 1995
;':.
1996 1996 1996L[__________~________~ ~
...
t\)
....
;:
~
'"
N
\0
-
130 G. Bekaert et al.
::!
~100%
I.. 80%
:::::. 80%
. .. .
J
~
I
40%
20%
.:
I 0,% 20% 40% 60% 80% 100%
Annualized VoIaIIllty (1990:01-1996:03)
Figure 2. (Continued.)
The behavior of emerging market returns 131
It is well known that emerging market returns depart from the normality. Tests
presented in Harvey (1995) and Bekaert and Harvey (1996a) show substantial
deviations from normality. Part of the goal of this paper is to examine which
countries show large deviations from normality and how those deviations
change through time.
There are a number of reasons why we observe non-normality in the equity
market returns. First, the presence of limited liability in all equity investments
may induce option-like asymmetries in returns (Black, 1976; Christie, 1982;
Nelson, 1991). Second, the agency problem may induce asymmetries in index
returns (Brennan, 1993). That is, a manager has a call option with respect to
the outcome of the firm's investment decisions. Managers may prefer high
positive skewness. Third, conditional heteroskedasticity may induce fat tails.
Fourth, regimen shifts, such as those detailed in Bekaert and Harvey (1995)
may induce both skewness and kurtosis. Finally, thinly traded securities' returns
may appear non-normal. The behavior of conditional skewness is studied in
Harvey and Siddique (1997) for a sample of developed countries.
Emerging market returns have more positive skewness than developed
market returns, with a coefficient of skewness greater than zero in 16 of 20
emerging equity markets. The highest skewness is found in Argentina and
Taiwan's equity returns. Emerging markets also present more excess kurtosis
than the world benchmark: Chile and Jordan are the only two countries where
the excess kurtosis is lower than the world benchmark.
We present three tests of normality: one based on Hansen's (1982) general-
ized method of moments (GMM), Bera-Jarque (1982) test and Kolomogorov-
Smirnov. Normality is generally rejected. Based on the empirical distribution,
the GMM test rejects normality in 4 countries, the Bera-Jarque in 13 countries
and the Kolomogorov-Smirnov in 11 of the 20 countries. All tests are based
on the last 5 years of data. These results are consistent with those in Harvey
(1995) and Bekaert and Harvey (1996a). Bekaert and Harvey (1996a) calculate
132 G. Bekaert et al.
Conditional correlations
Figure 5 presents rolling 5-year correlations with the MSCI World-AC for the
20 emerging markets and the IFC Composite index return. As detailed by
Harvey (1995), these correlations are generally small. By March of 1996, the
5-year trailing correlation with the world is less than 40% in all countries
except Portugal. Interestingly, there has been no clear trend in the correlations
across the emerging markets over the past 5 years.
Bekaert and Harvey (1996a) present a model of conditional correlation
where the means, volatilities and co variances are influenced by both local and
world information. Their model predicts that as a market becomes more
integrated with world capital markets, the relative influence of world and local
information changes. This change will affect the conditional correlation between
the emerging market and the world benchmark. Bekaert and Harvey offer
evidence that conditional correlations increase after capital market lib-
eralizations.
Many of the major liberalizations occurred before 1992 (Bekaert, 1995). It
is also clear in the data that correlations generally have increased over the
."
ciQ'
...s:
'"~
~
<:
~ " Skewnesa - Five Year Rolling Skewness - Five Year Rolling Skewness - Five Year Railing
,
~ b a b a a a a a ~ a a I'\) '" tAl VJ
"'...."
.... a i\l ~ ~ in Co a o (n 0 tn 0 u,
0 a'" '" o a
~ 1981 '" '" '" 1981 1981 '"
fI<I
iO<"
1982 1982 1982
'"
~" 1983 1983 1983
:>
"'"'" 1984 1984 1984
0
-, 1985 1985 ~ 1985
c::
0 1986 1986 19 ~
P- en en <I:>
1987 ~ 1987 1987 en
~.., <::l-
1?:: CD C1 CD<C <I:>
....
'" 1988 ~ 1988 ::e~Ill
.., 1988 CD ::s-
(l
::e~ ~ ~
::e~ ~ ::.
~
1989 CD
f/)
CD 1989 CD
f/)
== 1989 CD C!: :So
C f/) ~
.... f/) f/) f/) DI ...
:> 1990 1990 1990
~
<Sa.,
1991 1991 Ie <I:>
:=i
<I:>
1992 1992
~
1993 1993 ~.
1994 1994 1994
:=i
1995 1995 1995 ...::.
1996 1996 1996,L-____________________~ ...."'<I:>"
...
<I:>
;::-
...
~
'"
w
w
-
t..>
.j::.
~~ -
~
0
~ \:tI
~
'"~. Skewness - Five Vee' Roiling Skewness - Five Vee, Rolling ~
s: Skewness - Five Vee' Roiling I:l
660 0 0 0 0 o '"
600 ~ ~ U. o~ c.n (::, '"
U. ~
~ ~ NON ~ m ~ 0 N ~ 0 m 0 ~ (::, ~
....
1981 .. 1981 , ~
1981
1982 I:l
1982 1982 :-
1983 1983 1983
1984 1984 1984
1985 1985 1985
1986 1986
C/J C/J C/J(")
1987 t 1987 1987 "'0
'"~ :::I
~G> CD -
1988 Co ~ ; 1988 ~ 0
1988 :::I
CD :::I 3
1989 iir CD (') 1989 CD !:r.
1989
ien
en en CD g: I\)
1990 1990
en 1990
1991 1991 1991
1992 1992 I"
1993 [ 1993 !
1994 1994 1994
1995 1995 1995
1996 t 1996 1996L[____________________~
~--~------------~
.."
~"
;;:
!""
r)
c
's"
;:: Skewness - Five Year Railing Skewne&s - Five Year Roiling Skewne&S - Five Year Rolling
, , ~
.:., ~ ~ 6 0 0 0 o
'" 6 0 0
~ 0 (n 0 (n 0 (n (n 0 (n 0 (n 0 ~ ;"
--_.-
1981 1981 1981
1982 1982 1982
1983 1983 1983
1984 1984 1984
1985 1985 1985
1986 1986 1986 ;;!
en ~
ens:: 1987 ~ C- 1987 en::J
1987 ~O <;J-
~III CD 0 ~
~ ..... CD 0 ;::-
1988 ~ iii" 1988 ::J 0- 1988 ~ ::J I:l
::J'< CD III ::J CD <:j
1989 !!!. 1989 rn ::J 1989 CD rn
rn _. 0"
g:CD III rn ...,
1990 1990 1990 rn III
~
1991 1991 1991
1992
~
~
1992 1992
1993 ~
1993 1993 ~"
1994 1994 1994 ~
1995 1995 I:l
...,
1995
1996 1996 1996 "':'!:.: ir
....
...,
~
....;::
~
'"
....
W
VI
.."
QQ' v.>
0\
"...
'"~
-
r:; o
c
::s O:l
g. II>
Skewness - Five Yea, Roiling Skewness - Five Yea, Rolling ;>;-
0 r-J .... Skewness - Five Yea, Rolling l:l
./. N o r-J .... II>
a a a a'" a ~ ~ 0
N 0
"'"
.t:: a a a a a 6 ~
1981 a in a in a in a in
II>
1981 1981 .....
1982
1982 1982 ~
1983 1983
1 1983
1984
1984 1984
1985
1985 1985 ~
1986
1986 1986
1987 en1J en en
~III 1987 ~z
CD
~
~ _. 1987 ~s:
1988 1988 ~ cc' CD CD
::l III ::l CD 1988 ~ )(
::l _.
1989 CD
III -III 1989 CD ::::!. CD (')
III ::l gJ III 1989 III 0
1990 1990 III
1990
1991 .
1991 ; L 1991
1992 1992 1992
1993 1993 <. 1993 .-
1994 1994 l
1994
1995 1995 1995
1996 1996 ~ 1996
."
~.
~
"
~
1992 1992
3<ll
1993 1993 1993.- ~
~.
1994 1994 1994
3
1995 1995 1995 i :::.
...
~
1996 ,'--_ _ _ _ _ _ _ _ _ _------' 1996 1996 c
~
<ll
.........::::
;:::=
en
......
VJ
-.]
...,
aq. w
I: 00
-
....
~
r;
c
o
~
~
Skewneaa - Five Year Roiling Skewneaa - Five Year Rolling
I: Skewness - Five Year Rolling , tl
0 6 0 0 ~
p p . . ~ o ~
0) ():) 0 I\,) i.n i.n
.1::3"'" ~ ~ a a a a a'" a a i.n '"(::, '"in ....
1981 1981 1981
"'....
I:>
1982 1982 1982 :-
1983 1983 1983
1984 1984 1984
1985. 1985 1985
1986 1986 1986
en C/)-l 1987
en
1987 1987 ~~ ~iii'
CD _.
~ct (1) III
1988 ~ ., 1988 1988 ~ ~
~ ~ ~
~
==
III ~ III
1989 (1) CD 1989 (1) ~ 1989 m~
UJ'< UJ Co
UJ UJ UJ
1990 1990 1990
1991 1991
1991,
1992 1992 1992
1993 1993 . 1993
1994 1994 1994
1995 1995 1995
1996 1996 1996
The behavior of emerging market returns 139
Venezuela
Skewness
1.5
,.
~
~
'"
1.0
0.5
0.0
1- 0 .5
"'"
CIl
-1.0
Zimbabwe
Skewness
3.0
f
2.0
li
~ 1.0
~
Ii. 0.0
J -1.0
MSCI World
Skewness
0.4
0.2
:5'" 0.0
~
Iii -0.2
~
~ -0.4
i
-0.6
-0.8
ill -1.0
-1.2
-1.4
a; ..., LO co .... ..., co
....'" .... .... .... .... .... ....
N
00 00 00 00 co "-
00
co 00 co
0) 0
Ol (l)
N
Ol '"
Ol Ol
I()
Ol Ol
....
Ol Ol (l) (l) Ol Ol (l) Ol Ol Ol (l) Ol
.... ....
(l) (l) Ol
....
Ol
Figure 3. (Continued.)
140 G. Bekaert et al.
0.1
'"
.5
fUll
~ 0.0
, -0.1
~ -0.2
y~
1-0.3
~
-0.4
,'"
~
~ 0.0
'7
1.0
0.5
1-. 0.5
~ -1.0
3.0
I~ 2.0
";'
o 1.0
i -: .-...
~ 0.0 f-------,""-c=--------1
i
1-1.0
~
Figure 3. (Continued.)
.."
Qq'
s:
~
:to-
'T1
:.;;'
...:,'"
Kurtosis - Five Year Rolling Kurtosis - Five Year Rolling Kurtosis - Five Year Roiling
~ ~
.,...'"
~
... ~ o I\J 00 I\J &. U1 o U1 ~
~ a a a a a a a a a a a
s 1981 19B1 1981
""i'I" 1982 1982 1982
:l
'"
0 1983 1983 1983
~.
0 1984 1984 1984
.....
c:: 1985 1985
in 1985
c:>- 1986 L 19861- 1986 ~
o t't>
~ 1987 ' 1987 ~ (J-
1987 c::ca
...&f c:: (') ..,ellJ
.., t't>
1988 1988 :::'CD ;:s-
~ 1988 :::':1" -m N o ;:, !:>
oIII = III -. en <:::
CD _. ;:,e..
o_. -
...2 1989 1989 III
1989 c
t:S (ii" III m ....
~ 1990 1990
1990 -s.,
1991 1991 r- 1991 t't>
3
1992 1992 1992
1993 1993 ~
1993 ~.
1994 ~ 1994 1994 ~
3
1995 1995. !:>
1995
1996 1996 1996 ~
....
....
t't>
....
;::
....
;:s
'"
......
~
......
.."
0<;' ~
IV
~
'"
-!'-
-
~ ~
<:>
::s t:I:l
s Kwtosls - Five Year Rolling ~
"'
Kurtosis - Five Ye.. Railing Kurtosis - Five Ye.. RDlUng $:l
6 0 ... P I\J W ,J:Io. en .....
.t; o '" {..,
''"" 0 0> 0 0 0 o 0 0 0 0
0 in 0 i.n 0'" c.n ~ '"0 ....
"'....
0 0 '"
0 0 0
'" 0 0
1981 '" 1981 "'....
1981 $:l
1982 1982 :-
1982
1983 1983
1983
1984 1984
1984
1985 1985
1985
1986 1986
1986
1987 1987 ,,0
c: -
... ::J
1987 "(j)
c: ... c: Q..
1988 -0. ::lCD 1988 ... 0
1988
"o -'
til DI o CD 0'3
1989 iii' 1989 1989 se. 2:
;2 til DI
1990 1990 1990
1991 1991
1992 1992
1993 1993 F
1994 1994
1995 1995 1995
1996~1__- L_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _- J 1996 1996
..,
0;;'
s::
..
'"f>.
r;
c
::s
g.
s:: Kurtosis - Five Vear RoIUng Kurtosis - Five Vear Roiling
0 .... I\,) W ,J:It.
,Kurtosis - Five Vear Roiling
01 ,(, ~ 0 ~ <.0>
'" ~ o o
000 ~
~ 0 o 0 0 0 0 0 0 0 o .0 ;"
1981 1981
'" 198.1
1982 1982 1982
1983 1983 1983
1984 1984 1984
1985 1985 1985
1986 1986 1986 ~
(I>
1987 ,,3: 1987 1987 ,,:;- <::r
C I\)
"c...
C 0 C Q. (I>
., - 1988 ., 0 ;:,-
1988 -I\) ::La. 1988 -:;, I:l
0,< 51 I\) CD ~
(II (II 1989 51
1989 iii" :;, 1989 _" (II O
iii" iii" til iii" ...
1990 1990 c
1990
~
1991 1991 (I>
1991
::!
(I>
1992 1992 1992
~
1993 1993 1993 ~.
1994 1994 1994 ::!
1995 I:l
1995 1995 ...
;>;-
(I>
1996 1996 - 1996 ....
~
....
:::
...::s
'"
.....
...
~
..., ....
0;;"
...'" t
'"
~
r;
c
o
~ ~
;;" Kurtosis - Five Year Rolling ;0:-
Kurtosis - Five Year Roiling Kurtosl. - Five Year Roiling
~ ~
~ ~
s::.
'"'" &. 0 01 0 01 0 01 .j>, 00 0 .j>, 00 0 N 00 o
.!:: '" '" '" '"
0 0 0 0 0 0 0 0 '"0 0 0 '"
0 0 0
'" 0 0 0 o 0.j>, o o ...~
1981 1981 1981 ""...
1982 1982 1982 " ~
1983 1983 1983
1984 1984 1984
1985 1985 1985
1986 1986 1986
1987 1987 C __ 1987 ~3:
..,~~~ "Z .., CD
1988 1988 "'0 1988 -)(
8"
__ iii'
I o-CD
.., oen --
en - en -- 1989
(')
1989 r en ::::I 1989 iii- I iii" 0
1990 1990 1990
1991 1991 1991
1992 1992 1992
1993 1993 1993
1994 " 1994 1994
1995." ( 1995 1995 r
1996 1996 1996 "
.."
oq.
s:
;;:
"!'-
~
0 0
Kurlosls Five Year Rolling
0
s Kurlasls Ave Year RaIling Kurlosls Five Year Rolling
, ~ ~ I\J hl W W ~ .e:.. 01
~ ~
!=' ~ !-II 0 en o en o
'" o m 0 ~ 0 ~ 0 ~ 0
! o o o 0
'" 0
'" 0 o 0 o o o o
1981 1981 1981
1982 1982 1982
f
1983 1983
1983 F,
1984 1984 1984
~
1985 '0 1985 1985
1986 1986 1986 ;2
en 1J ~
,,1J 1987 ,,~
1987 "g 1987 c: 0 0-
.... .... c: = ~
.... :7
c: - 1988 Co ::'-0 ::r-
1988 o c: 0-0 I::l
UI _. <:::
0" Ulec ~
!e. Q
UI CD
1989
--
_. I\l
UI _ 1989 _.
UI CD
UI ...5
I\l 1990 1990
~
1991 1991 'e 1991
~
~
1992 1992., 1992
1993
~
1993 1993 ~.
1994 1994
;;!
1995 I::l
1995
1996 tL'_..1..-_ _ _ _ _ _ _ __ 1996 f - ~
~
*...
;::
...;::s
'"
......
Vo
"""
...,
~.
... 0\
'"
"!>0-
-"""
r:; ~
c
"
5" ~
Kurtosl. - Five Yr Railing Kurtosis - Five Ye.. Railing Kurtosis - Five Ye.. Railing
0 I\) ~
o """" ~ I\J ~
I\)
"'" 0 -1>0 01
,!:; 6 en b tn b u. po o ,:, 0 I\) -1>0 01 0> 0
6 6 6 6 o 6 (:, (:, (:, 6 66(:,
1981 1981
r......
1981 "'
1982 I:l
1982 1982 :-
1983 1983 1983 &
1984 1984 1984 Co
iil'"
~
r;c
,.
S KlJrbsis - Five 'Year Roiling Kurtosis - Five Year Roiling Klriosis - Five Year Rolling
o M W ~ m m
~ f\) fA .1:10.
'"
'J::" .
....
0 0 0 0
en
0
en
0
~ o M ~ m CD 0 ~ ~ o 0 0 0 0 0 o
o 0 o 0 0 0 (:, 000
1981 [: 1981 1981
1982 1982 1982
1983 1983 1983
1984 1984 1984
1985 1985 1985
1986 1986 1986 ~
s: (\)
,,~ 1987 ,,~
1987 1987 c: 3 c: ::l 0-
c: 0
"CIJ .. C"
1988 1988 1988 ~I!! ~
6"Dl t:::>
UI C" 5l c: ~
1989 ~i 1989 iii" ~
1989 iii" !! c
-" 0
UI .. CD Dl -;
1990 c:: 1990 1990
~
1991 1991 1991
1992
~
(\)
1992 1992
1993 ~
1993 ~.
1994 1994 1994
;;
1995 t:::>
1995 1995~ -;
;0:-
1996 (\)
1996 ...
...~
;::
-;
;::
'"
~
-...I
-
..., .....
~.
00
;;: """
:""
0
~ b:l
(\)
;>0:-
s Kurtoela - FiIIe Yea, Roiling Kurbsis - Five 'Yea, Roiling I:l
1 (\)
...., 0 N .... 6 6 boo 0 0 0 .......
.e b b b b '"b b i:n ~ i\JbN~CnCoO
1981 1981 ~
Kurtoal. (1980:01-1989:12) 1982 ~
1982
in P ?I ~ ~
~o 0 0 0 0 1983 1983
oK
1984 1984 ~
:;; "T1
0 m en
~~ 1985 0 1985
Gl 3 0
00 (j)
1986 1986
1 - CO _.
~-~
1987 ,,0 1987 ,,=
i~~ c: C" c: 0
i~ UJ
... DI
i < i: III 1988 1988 :::'0
g I!IJ ;. o 0
-- o c:
III :::::I
1989 ~. 0 1989- _.
~~ i a; l!- III 3 III ..,
-
lIlo 1990 't:I 1990 '<
- CO "
;DOC
~ UJ
0
! ::1 1991 . III 1991 :E
"'.-
b il ~ it ..,0
UJ
1992 . 1992 a:
...!ii 1993 1993
1994 1994
1995 1995
1996 1996
."
0<;'
.,...
'"
:.n
::rl
<
CD
-<
~
CD Ave Year Rolling Correlation Five Year Rolling Correlation Five Year RoIling Correlation
...... 6 6 0 0 0 0 0 6 6 a a 000 6 6 000 0 0
:".. iv 0 iv :".. m i:x> 0 :".. iv b r...:, ~ mCD (:) :".. iv 0 iv :".. m i:x> 0
~
5' 1981 1981 " 1981 ......
OQ
.., ~~~:1 .:::." .~.
1982 .1 1982 1982
......0 r /:~
~ 1983 1983 1983
~
:;:.
o 1984 ,J 1984 \ 1984 ,-
" .......
::l t!" 'ii 'ii 'ii
rA (') (')
1985 (') 1985 1985
0 ~ ...;.v ..-:
...., G'l
1986 5' ! ....."
~
e 1986 n 1986 [ ~
n ~
I ; ;~I'-..;
.,''
[(') n>
in 1987 o 1987 0 1987 ,,'
.. ~~ o ., III ~ca c:r
0-
~ 1988
>
I . -~
i ~
CD
n
::T 1988
...~ ..?
~ i i 1988 g
~ Cil CD ~
~
-~ ~
iii
_CD '. ~ a !:!, III C!: <:;
,,'
...... 1989 ,J I = 1989 ..
'0;. ;;:: 0'- 1989
"'/ it C!:~
~ ~ 0' III
~ Ch ~
1990 ,. ..' l7i ~ 1990 -'0:." Q 1990 ~;'" It- l7i
Q
o~ ...0'
...c: ~ Q
~ ~ ::. <Q.,
1991 , 1991 :'" ~ 1991 ~
~. ~ .~
..... '\ ( ~
;. 1992 1992 1992 ~
\ ~ ~ ~
:E .......,.. 5: ~.. c. 5: ~
0 1993 1993 1993 L--
::!.
/ ~'
0- 1994 ~~ 1994 1994
r'{'o
<T 3
...
::l
..,
::r
1995
'.'\" 1995 r 1995 r
fa, ~
...
~ 1996 '.
1996 rc ..... 1996 c
E!
~
I~ ......~
~
...i'I"
l" ...;::...
::!
'"
. j::.
\0
-
...,
~' v.
0
i;l
-
~
nc 0
:: t;x:,
~
5' Five Ye. RoIling Comoletlan Five Y. . RoIling Correlellon ;>;-
s:: Five Yeat RoIling ComoIetian I:l
0 0 0 0 0 0 ~
~ 0 0
'" 6 6 0 0 0 0 0 6 6 p b b
~ :,... ;., b ;., :,... a. Co b :,... ;., b :,... a. a. b :,... ;., b ;., :,... a. a.0 b ::l.
, 1981
'" ~
1981 " 1981 t ...
.~i" I:l
1982 "..... 1982 1982 :-
1983 ..",' 1983 1983
"
1984
!
,i' 1984
/:
}
;;
1984
;;
1985 .] 0 1985 .:' 0 1985 0
.,..
'" II: G1 G1
~, .. 1986 '
1986 r("" 1986
0 I,. 0 00
1987 0 1987 r: .. .. .l o (j) 1987 ~ 0
3
1988
/! ~ CD -
=I :::I
1988
-.~~.
"
II 1988
I!
'8 Cil 0'
III
- Q_..
~
i-CDi iii 3
III (')
if , J ..if
1989 cr. 1II 1989 gCD 1989 o=:C"_.
0 .~. :::I III
~ :::I ~ :::I 1990 ~
1990 0 1990 ..' ,Q ,Q
Indonesia
Correlation
1.0
0.8
I IFC Glo~omPOSite MSCI~~~.~.World
IS0.6
I!' 0.4
r
".',
~ 0.2 / ".
~ 0.0
1II.
-0.2
-0.4 .....
N <') v It) CD ,... co 0> 0 ..... N <')
;;r; It) CD
...
co co co co co co co co co
..... ..... ..... ..... ..... ..... .....
0> 0> 0> 0> 0> 0>
..... ..... .....
0> 0> 0> 0>
0> 0>
0>
0>
~ ~
0>
.....
0> 0> '"..... .....0>
0>
Jordan
Correlation
1.0
IFC Global Composite MSCI (AC) World
0.8
c:
l
a
11
0.6
I!' 0.4
~ 0.2 ........
II ~......!"\
~ 0.0
1II.
-0.2
-0.4
;;; N <') V It) CD ,... CO 0> 0
C;; N <') V It) CD
0>
..... ...
CO
0>
CO
~ ...
CO
0>
CO
0>
.....
CO
0>
co co
..... ..... .....
0> 0>
CO
0>
... ...
0>
0> 0>
0> 0>
0> 0> 0> 0>
.....
0>
..... ..... ..... '" ... 0>
0>
Malaysia
Correlation
1.0
c:
0.8
i!
~
0.6 ..f :........
8r 0.4 .
:' !..........
.
~'\
~ 0.2
II
~ 0.0
1II.
-0.2 I IFC GIO~mposlte MSCI.. ~~~.WOrJdl
-0.4 ...co N <') It) CD ,... co 0>
~ co co co co co
0
C;; N <')
;;r; It) CD
'".....
co
'" '"
co
..... '"
..... '" ..... 0>
'"..... '"... 0> 0>
..... '"'" '"...
'"..... 0>..... ..... 0>
~ '"..... ...'"
0> 0>
Figure 5. (Continued.)
.." .....
~' Ul
...s:: IV
..
!J'
(:j ~
~ b:l
E' Five Year Roiling Correlation Five Year Roiling Correlation ~
'"
s:: Five Ye.r RoIling CorrelaDon , , s::.
6 6 0 0 0 0 0 6 6000 0 0 !=>?PPPPP.
......
~ ~ N 0 N ~ in 00 0 ~ NON ~ in 00 0
'".......
.,. '" 0 '" "'" '" ex> 0
1981 1981 1981 \ ~
1982 1982 1982 ~:~;: l2-
1983 1983 1983 1 ..
t
1984 1984 1984 ,I"
"i
.--
"i .~" "i
1985 0 o 1985 o
1985 ::~ ..
!il ~
1986 1986 ~ 1986 ....
0-0 ~ o ~ 0
1987 o I 1987 "
"If
I~ 3 ~ ~
1987 ~ g z li3 g s::
.., CD
CD _, 3
t'...............:~
1988 I
- I I_I 1988 a; cO'
-CD 1988 ~ /l ~ ><
I .., ~ !!t 0'
1989 i =1 i 1989 ~. .. -, 0
~ 0 ::I 1989 g: iii' .:-. ~ 0
In ::I In ::I
1990 0 1990 ~ ::I 1990 Q
Q
1991 ~): 1991
~~ 1991 ,9 L~
1992 ':IE '~ 1992 Co
1992 ~
1993 '
~ i'i: 5:
1993 L- 1993
1994 1994 1994
1995 - 1995 ..... 1995
1996 '."
1996 1996
.."
~ .
...
'"
~
:s
.,g. Five Year Roiling Correlation Five Year Rolling Correlation Five Year Roiling Correlation
6 6 o 0 0 a 0 6 6 0 0 p p p . 6 6 0 p 0 o 0
'"
.!::: ~ ;., o ;., ~ en ix> 0 :,. ;., 0 ;., ~ 0'> 00 0 ~ ;., 0 ~ CD ex, 0
1981 1981 1981
'"
'
.......
1982 1982 1982
,i
,
1983 ~: 1983 1983
1984 f 1984 1984
.....i 'ii
()
'ii 'ii
1985 ,_.r 1985 () 1985 ()
Gl
1986 ........... ~ 1986 1986 ~ ~
.... :.~. [ ()CIJ I[
" ()lJ ~
1987 o., 0c: 1987 b'lJ 19871,
-g
() o :7 0-
.;;' ., 0 0
, ::::: ~
: ; :f ~ 3 ::r-
1988 1988 1988 CD -C'
~.
;-c:::. 15 -"0 :::.
!!!.CC <:::
1989 \ I ~6 1989 ar o_. I
_ 1989 ~ ~ :i" O
'~'t ;:: CD ....
1990 ~ gm ~ :I en
o:I UI
"" Q 1990 Q 1990 :Q .sa,
1991 1991 :> ~
~ ~ 1991 ~.9 ~
1992 ~
~ 1992 ~ 1992 ~
5: 5: a: ~
1993 1993 1993 ~.
1994 1994 1994 ~
1995 :::.
1995 1995 ....
;>;-
~
1996 1996 1996 .......
~
l::
.......
::!
'"
......
Vl
W
.." .-
~. VI
., ~
'"~
?i
<:>
o
::s
g. Five Veer RoIling Correlation Five v.. RoIling Correlation ~
Five V RoIling CorreleOon tl
''"" b !=' !=' !=' !=' !=' ,.. 6 boo 0 0 0 .... 6 0 0 0 0 0
~ :... P :... NON :... ;" Co 0 P ~
.....
'" 0 '" .co. Ol CD 0 .co. N (:) N :,. 0, CD 0
1981 1981 - ~
.....
1981
.\ tl
1982 1982 1982 :-
"'1
1983 1983 '. 1983 "
~.,
1984 ,--- 1984 1984
;; ;;
1985 o 1985 \',..,. ~ 1985 o
1986 (")
1986 ....... 1986
I.. (")
1987 1987 bl-l 1987 g iii'
~
CD
ct
... i. ...... i. I! 3 ; ~
1988 ~ 1988 '" 1988 !. i'
I\)
- CD i'i'
cr.:::J
I I
III
1989 1989 I
~'
III
:,.
1989
gcr.,< g Co ocr.:::J
1990 1990 ".-';" ~ :::J
.~ ~ 4'1 990 " .0 .'.:'......\
1991 ~~ 1991 ~~ 1991 ~~
1992 'j 1992 'j 1992 'j .i
a. is: -::.~~
1993 L..- 1993 1993
1994
"
'i.T':=~ 1994
:... ""
1994 '0,:
1995 .'} 1995 1995 ....
......
...: ......
1996 i' 1996 1996 ,"
The behavior of emerging market returns 155
Venezuela
Correlation
Zimbabwe
Correlation
1 --;:::::::=========;--,
lIFe GIo~ompoaIte MSCI..I:~~.WorId I
1.0
1.0 1 --r==========;---i
IMS~ ~v.brld I MSCIAlI..
<.~
fLY' ---
i*
~ 0.8
O.
ii 0.2 .:
t
j 00
S'O~~.2--;;0.-;;-0--;;0-;;-.2--;;.':;-.--;;0.';" ...J
CCIn'.won ..... World (1.0:01-1_:03)
Figure 5. (Continued.)
156 G. Bekaert et al.
Arithmetic Return
50%
R Sq: 6.6%
T-Stat: 1.13
40%
30%
20%
10%
0%
-10%
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0
Beta vs. MSCI AC World
Figure 6. Risk and return - IFCG indices. Sample: April 1991-March 1996.
longer horizon (see final panel of Figure 5). For example, the correlation of
Argentina and Brazil with the world was zero or slightly negative in the 5 years
ending in 1981. By March 1996 the correlations were above 30%. Long horizon
increases are also evident in South Korea and Thailand. However, in the last
5 years there has been little overall change. In the 5 years ending in 1991 the
correlation of the IFC Composite and the MSCI World was 30%. In March
1996, the correlation was 35%. Slightly higher correlations are found comparing
the IFC Composite to the MSCI World-AC, which has a higher emerging
market composition.9
9The MSCI World-AC begins in 1988. Before 1988, we splice the MSCI World index to the
AC index.
The behavior of emerging market returns 157
Arithmetic Return
50% r---------------------------------------------------i
R Sq: 21.4%
T-Stat: 2.21
40% -
30%
20%
10%
0% r---------------------------------------------------~
-10% L-~ __ ~ __ ~ ___ L_ _ ~
_ _~_ _~_ _L __ _~~_ _~_ _~_ _~_ _~
Figure 7. Risk and return - IFCG indices. Sample: April 1991-March 1996.
Following Ferson and Harvey (1994), Erb et al. (1995a, 1996b) and others, we
examine the relationship between some country-specific risk attributes and the
distribution of returns.
Survey-based measures
The first of the measures used to group these attributes is Institutional Investor s
Country Credit Rating (IICCR). Institutional Investor country credit ratings
are based on a survey of leading international banks who are asked to rate
each country on a scale from zero to 100 (where 100 represents the maximum
creditworthiness). Institutional Investor averages these ratings, providing greater
weights to respondents with higher worldwide exposure and more sophisticated
country analysis systems. These ratings have appeared in the March and
September issues of Institutional Investor since 1979 and now cover over 135
countries (for additional details see Erb et al., 1996a).
Whenever a surveyor expert panel is used to subjectively rate credit-
worthiness, it is hard to define exactly the parameters taken into account.
At any given point in time an expert's recommendation will be based upon
those factors the expert feels are relevant. In a recent survey of participants,
the most important factors for assessing emerging markets' credit rating were
(i) debt service, (ii) political outlook, (iii) economic outlook, (iv) financial
reserves/current account and (v) trade balance/foreign direct investment.
The next four measures are from Political Risk Services' International
Country Risk Guide. They include the political risk index (ICRGP), economic
risk index (ICRGE), financial risk index (ICRGF) and the composite risk index
(ICRGC). The political index is studied in Harlow (1993) and Diamonte et al.
(1996). Erb et al. (1996b) examine the information in all four of the ICRG risk
indices. On a monthly basis, ICRG uses a blend of quantitative and qualitative
measures to calculate risk indices for political, financial and economic risk, as
well as a composite index. Five financial factors, 13 political factors and six
economic factors are used. Each factor is assigned a numerical rating within a
specified range. A higher score represents lower risk (for additional details see
Erb et al., 1996b). The composite index is simply a linear combination of the
three subindices. The political risk is weighted twice that of either financial or
economic risk. ICRG, as well as many of the other providers, think of country
risk as being composed of two primary components: ability to pay and willing-
ness to pay. Political risk is associated with a willingness to pay, while financial
and economic risk are associated with an ability to pay.
We also include Euromoney's country credit risk (EMCCR). Euromoney's
rating system is based on both qualitative and quantitative methods. The
political component is a qualitative survey of experts. The economic component
is quantitative and based on Euromoney's global economic projections. The
financial component is also quantitative and based on (i) debt indicators,
(ii) debt in default or rescheduled, (iii) credit rating (Moody's or Standard and
The behavior of emerging market returns 159
Poors), (iv) access to bank finance, (v) access to short-term financing and
(vi) access to international bond and syndicated loan markets.
Macroeconomy
The survey-based measures gauge indirectly the future macroeconomic condi-
tions in each country. One of the primary economic measures that influences
these ratings is the inflationary environment. Ferson and Harvey (1993, 1994)
argue that asset exposure versus world inflation helps explain both the cross-
section and time-series of expected returns in 18 developed markets. Erb et al.
(1995b) examine the interaction of inflation and asset returns in emerging
markets. We use a trailing 6-month measure of inflation represented by the
consumer price index reported in the International Financial Statistics database
of the International Monetary Fund. In the case of Taiwan, which is not a
member of the IMF, we use inflation reported in their national accounts.
Demographics
Bakshi and Chen (1994) propose a life-cycle investment hypothesis. Younger
investors have a higher demand for housing than for equities. As age increases,
more investment is allocated to the stock market. As a result, a rise in average
age should be accompanied by a rise in the stock market. Bakshi and Chen
(1994) found support for this hypothesis using data from the USA. Erb et al.
(1997) found that average age growth explains the risk premiums in a number
of developed countries. We examine three variables: population growth, average
age and average age growth. All of these data are based on annual statistics
compiled by the United Nations.
Market integration
Bekaert and Harvey (1997a) argue that the size of the trade sector relative to
the total economy is a reasonable proxy for the openness of both the economy
and the investment sector. They use exports plus imports divided by GDP as
an instrument for market integration. This variable, along with other proxies
for market integration, is used in a function which assigns time-varying weights
to world versus local information. Bekaert and Harvey found that increases in
this ratio are associated with the increased importance of world relative to
local information for both the mean and the volatility of the country's stock
returns.
Persistence
A number of researchers have pointed to momentum as an important firm-
specific attribute (Jegadeesh and Titman, 1993; Asness et aI., 1996; Ferson and
Harvey, 1997). We examine two measures of momentum: the lagged monthly
return and the lagged quarterly return from 4 months ago to one month ago,
i.e. the quarterly return lagged by an extra month.
160 G. Bekaert et al.
Size
We follow a number of studies, beginning with Banz (1981) that document a
relationship between firm size and expected returns. Recently, Berk (1996a,b)
has argued that size measured by market capitalization should be a proxy for
risk. This attribute has recently been studied on a country-level basis by
Keppler and Traub (1993) and Asness et al. (1996) who found that size helps
explain the cross-section of expected returns in a sample of developed markets.
Summary statistics
Some summary measures for many of these attributes are included in Table 4.
The March 1996 value of the attribute is reported. In the lower panel, the
rank-order correlation of all of the attributes is reported. Most of the correla-
tions follow from intuition. Consider the ICRG indices. These indices are highly
correlated with the Euromoney and Institutional Investor country credit risk
measures. All of the survey measure are negatively correlated with inflation
(high inflation means low rating). The most negative correlation with inflation
is found for the ICRG economic risk index. Average age is positively correlated
with the survey risk indices, indicating that low average age is associated with
a low rating. Size is positively related to the ICRG ratings (smaller markets
appear more risky). There is also a positive relation between the size of the
trade sector and the ICRG ratings. The lowest correlations are found for the
ICRG indices and the fundamental attributes.
Portfolio approach
Country ICRGC ICRGP ICRGF ICRGE ICCR EMCRR INFLATE TRDGPD POPGR AAGEGR AVEAGE MKTCAP PIE PIB PID
(%) (%) (%) (%)
Argentina 72.5 76.0 35.0 34.0 38.4 57.2 0.7 12.8 1.2 0.3 30.9 22308 16.7 1.4 29.2
Brazil 65.5 64.0 34.0 33.0 35.8 55.4 29.2 13.6 1.7 0.8 27.1 93940 40.3 0.5 28.9
Chile 80.5 76.0 43.0 41.5 59.2 79.8 7.6 42.1 1.6 0.6 29.0 39421 15.9 1.9 26.1
China 72.0 68.0 38.0 38.0 56.4 70.8 1.0 0.9 29.6 29495 31.8 2.0 37.6
Colombia 66.0 58.0 39.0 35.0 46.7 62.6 19.1 61.4 1.6 0.9 26.2 6659 12.0 1.0 36.8
Czech Republic 82.5 82.0 42.0 40.5 60.1 74.6 8.5 12346 13.4 1.0 87.7
Greece 75.0 76.0 38.0 36.0 49.8 73.3 8.5 60.2 0.3 0.6 38.9 11200 10.8 2.1 24.6
Hungary 76.0 79.0 40.0 32.5 43.6 67.7 29.6 49.7 -0.5 0.2 37.7 2957 21.4 1.1 125.0
India 67.0 62.0 36.0 36.0 45.8 66.7 9.7 17.2 1.9 0.5 26.0 71141 14.3 2.3 65.8
Indonesia 70.5 65.0 39.0 37.0 51.8 73.2 10.5 43.9 1.5 0.8 26.2 54571 26.6 3.5 1124 ~
(I:>
Jordan 74.5 73.0 38.0 38.0 30.5 54.3 7.0 130.1 4.6 0.3 21.4 3276 15.6 1.7 50.0 <:::r'
(I:>
Malaysia 79.5 75.0 43.0 41.0 68.4 84.5 3.3 166.4 2.3 0.6 24.8 162134 28.4 3.7 83.3 ;:-
!::>
Mexico 69.5 66.0 40.0 33.0 41.2 58.8 43.8 37.2 2.0 0.9 24.8 65162 18.6 1.7 117.6 ~
ICRGF 1.00 0.83 0.88 0.88 -0.59 0.43 -0.50 0.42 0.44 0.58 0.31 -0.03 0.30
...
~
ICRGE 1.00. 0.79 0.81 -0.76 0.52 -0.39 0.43 0.33 0.54 0.29 0.03 0.17
nCCR 1.00 0.97 -0.61 0.39 -0.58 0.54 0.53 0.66 0.36 -0.02 0.13
EMCRR 1.00 -0.65 0.38 -0.54 0.44 0.48 0.71 0.45 0.06 0.21
INFLATE 1.00 -0.31 0.24 -0.10 -0.27 -0.50 -0.28 0.02 0.06
TRDGDP 1.00 0.02 0.18 -0.11 0.06 -0.02 0.06 -0.01
POPGR 1.00 -0.47 0.95 -0.09 -0.05 0.35 -0.07
AAGEGR 1.00 0.39 0.36 0.26 -0.12 0.11
AVEAGE 1.00 0.07 0.02 -0.36 -0.09
MKTCAP 1.00 0.68 0.26 0.27
PIE 1.00 0.16 0.37
P/B 1.00 0.27
P/D 1.00
ICRGC, political risk services: international country risk guide - composite. ICRGP, political risk services: international country risk guide - political. ICRGF, political risk services:
international country risk guide - financial. ICRGE, political risk services: international country risk guide - economic. nCCR, institutional investor country credit ratings. EMCRR.
Euromoney country risk ratings. INFLATE, annual consumer inftation: IFS database. TRDGDP, trade openness (exports + imports)/GDP. POPGR, annual growth in total
population - UN data. AAGEGR, annual growth in average age or population - UN data. AVEAGE, average age or population - UN data. MKTCAP, IFC global market
capitalization (US$ millions). PIE, IFC global price/earnings ratio. P/B, IFC global price/book ratio. P/D. IFC global price/divident ratio.
The behavior of emerging market returns 163
level itself. These portfolios are investible with respect to the attribute: that is,
lagged attribute information is used to determine which countries are in the
portfolios and the analysis is conducted out of sample. Given the small number
of emerging markets, we examine only two portfolios: high attribute and low
attribute. In each case, we track the returns to portfolios that are equally
weighted by country, and those that are weighted by each country's equity
market capitalization. To reduce potential transactions costs, we only consider
quarterly rebalancing.
Table 5 presents the results of the portfolio strategies. Of the ICRG indices,
the composite index (ICRGC) produces the greatest separation of expected
returns. With the equally weighted investment scheme, the high attribute port-
folio (low risk) presents 29.7% average return with a 25.7% volatility. The
low attribute portfolio (high risk) delivers a 36.6% average annual return
with a slightly lower volatility, 23.5%. Interestingly, the fl-value against the
World-AC is much lower for the low attribute portfolio. Hence, the alpha of
this strategy is quite large.
Of the family of ICRG indices, the financial and economic risk appear to
be the most important and in the equally weighted portfolio strategies. The
political risk measure is only important in the capitalization weighted invest-
ment strategies. These results are consistent with those presented in Erb et al.
(1996b). This implies that pure political risk is diversifiable and not priced.
Both the Institutional Investor and the Euromoney credit ratings also are able
to discriminate significantly between high and low expected return securities.
As with the ICRG Composite, the low attribute portfolios have higher means
and lower volatilities than the high attribute portfolios. Inflation also appears
to be an important instrument in portfolio selection. In this case, the high
attribute portfolio has much higher expected returns than the low attribute
portfolio. However, in contrast to the ICRG, EMCCR and nCCR, the high
attribute portfolio has much higher volatility than the low attribute portfolio.
Trade to GDP has only marginal ability to distinguish between high and
low expected returns. The low attribute portfolio has higher expected returns
than the high attribute portfolio. However, the volatility of the low attribute
portfolio is greater. Nevertheless, the fl-value of the low attribute portfolio is
close to zero leading to a very high ex-value. Caution needs to .be exercised
here: the fl-value of the low attribute portfolio may be low because the market
is not intregrated. The idea of Bekaert and Harvey (1997a) is that trade to
GDP is a proxy for integration. Indeed, the low fl-value of the low trade to
GDP portfolio is consistent with their results.
The three demographic variables - population growth, average age growth
and average age - offer no ability to discriminate between high and low
expected return countries. The demographic asset pricing theory presented in
Chen and Bakshi (1994) is most appropriate for time-series analysis of devel-
oped countries. That is, holding other factors constant, an increasing average
age will be associated with higher demand for equities. It is difficult, if not
Table 5. Country risk level portfolio strategy, January 1985-March 1996.
~
-
Risk attribute High attribute Low attribute Low-high attribute
Capitalization weighted
ICRGC 12.0 33.0 1.06 0.61 30 29.8 28.6 0.47 0.38 58 15.~ 40.1 35.4 30.9b
ICRGP 13.7 34.3 1.12 0.72 22 20.3 23.4 0.39 -0.04 38 5.7 35.9 26.6 27.3 b
ICRGF 15.7 33.6 1.05 0.74 23 22.2 31.5 0.49 0.03 49 5.6 44.0 27.1 b 29.6b
ICRGE 16.8 31.8 0.96 0.60 25 18.0 38.5 0.91 0.73 45 1.0 42.0 10.2 5.9
IICCR 16.4 32.0 0.99 0.64 22 24.7 35.9 0.82 0.60 34 7.2 37.0 16.5 13.3
EMCRR 17.1 31.9 0.99 0.64 23 19.4 36.9 0.84 0.59 46 2.0 37.2 11.2 8.8
Table 5. (Continued.)
INFLATE 19.6 34.0 0.82 0.67 43 17.0 31.8 0.98 0.61 23 -2.2 34.2 1.7 4.2
TRADEGDP 20.1 35.0 1.06 0.72 19 15.3 31.1 0.71 0.44 30 -4.0 38.8 9.1 7.8
POPGR 16.7 27.1 0.60 0.50 21 14.3 27.1 0.63 0.65 20 -2.1 30.3 3.1 0.6
AAGEGR 15.8 31.8 0.82 0.91 17 15.7 20.9 0.37 0.07 17 -0.1 29.0 12.6 18.0b
AVEAGE 16.2 27.7 0.63 0.57 20 18.5 26.6 0.62 0.59 15 2.0 30.7 5.8 3.8
~
~
MKTCAP 15.1 32.5 1.07 0.63 22 27.7 26.3 0.14 0.43 50 11.0 44.7 38.3 c 25.6"
<:J-
MOM-1 23.8 29.7 0.72 0.67 237 19.3 34.3 0.99 0.58 232 -3.6 38.0 1.9 7.3 ~
~
MOM-2-4 15.2 36.7 1.01 0.72 130 19.4 30.7 0.73 0.41 139 3.7 39.7 17.4 17.1 :::.
PlEa 14.5 35.5 1.09 0.98 33 21.9 32.4 0.59 0.29 69 6.4 43.0 25.1" 26.2" '"c
p/Ba 10.5 38.1 1.18 0.92 40 24.6 26.1 0.48 0.57 64 12.7 42.3 33.9c 25.7"
....
P/D 17.8 36.4 1.12 0.68 42 32.9 26.3 0.41 0.39 85 12.8 41.0 34.6c 25.0' .sa.,
~
Table 6. Estimated transaction costs in the emerging markets - Baring Securities emerging market
index spread analysis
Weight BEMI
+ standalones
Spread in Weight BEMI (%)
Country basis points (%)
offer and bid price. Barings uses the midpoint in the divisor in order to avoid
the problems caused by large fluctuations in the current price. The percent
spreads in Table 6 are based on snapshots of individual stocks during the weeks
of July 17 and July 24, 1995. The country spreads are calculated by capitaliza-
tion weighting the percentage spreads of the individual firms within each
country. The percentage spreads are, in many countries, much larger than one
would expect in developed markets. The spread in Chile is close to 400 bp. In
both Argentina and Turkey, the percentage spread is more than 150 bp. These
high transactions costs reinforce the need to minimize trading. Indeed, many
investment managers do not practice active stock selection strategies in emerg-
ing markets because of the massive transactions costs. 'Active' management in
emerging markets is often interpreted in the context of country selection rather
than stock selection.
While the portfolio analysis shown in Table 4 does not explicitly account
for transactions costs, we do include a measure of average turnover. The highest
turnover is found with the momentum strategies. The turnover is so high that
it is unlikely that these strategies could be successfully implemented in the form
specified here. The lowest turnover is found with the demographic variables.
This is not unexpected given that the data is only available annually and there
is little variation over the years.
The most impressive ratios of low-high portfolio returns to turnover are
found for the survey risk attributes, in particular the ICRGC, EMCCR and
IICCR. For example, in the low attribute of EMCCR portfolio, the turnover
is 46%. With 10 countries in that portfolio, four or five would change over the
entire year (four rebalancing opportunities).
Another important constraint is investibility. The portfolio analysis is init-
iated in 1985 and includes all the countries in the IFC-Global database.
However, for much of this period many of the returns were not attainable due
to investment restrictions (Bekaert and Urias, 1996).
The portfolio exercise suggests that many of these risk attributes can discrimi-
nate between high an low expected returns environments. What about the
other moments? Erb et al. (1996a), and Bekaert and Harvey (1997a) argue that
attributes like trade to GDP and credit rating affect conditional volatility.
Ferson and Harvey (1994, 1997) establish a link between country attributes
and time-varying risk exposures. It makes sense to examine the relation between
these some of these attributes and the other moments of returns.
Figure 8 focuses on three attributes: ICRGC, trade to GDP, and PB. The
average values of these attributes over the past five years is plotted against the
moments of the country returns: mean, standard deviation, skewness, excess
kurtosis and correlation with the World-AC. For each of the attributes, there
is a negative relation between the value of the attribute and the average returns
which is consistent with the portfolio exercise. For the ICRGC and the
The behavior of emerging market returns 169
AnnualtedVolatillty
80%
''''
""
50%
.,%
,,% "
~~------------------ 20"
'O%50~----eo::::-----=-o':'O""----~80----"" ,~
I
...... ...--
::f
4.0
3.0
,~
20 r 10.0 ~
'.0 l----~~--!-----------__j
5.0 ~
0.0 f----------'----~-----
0.0 ~--------=------------
0.0 ~--=----------------_!
.(l.'
-1.050'----~BO-----,LO- - - -..
-----" .().250L--~--60=-----:":-----;80~---~"
-...
AVaragII leRG CompoeIte Rdnp Averaga leRG Compoeita Rdngs
-.. AnnllllllzedVoAatllIty
,,%
-I
-' ,,%
60%
"" ,,%
20%
40%
,~
,-
3D"
.,:~
-
20"
, .,,, ,,,,, ,.". ,""
.........-
""0%
--
~ 20% 40% 60%
""
_TnodeIGNP
'20% 140% ,eo%
"", 20% 60%
Average l'nIdeIGNP
'20% '40"
4.0 20.0
3.0 15.0
~O
10.0
'.0 "
5.0
0.0
,
0.0
-1.0
-2'00%
20" 40" 60" ,,% ,,,% '20"
Average TradelGNP
'40" '60" '60%
-5.0
'" 20% .", "" 80% 100%
Average TradafGNP
'20% 140% 160" 180%
I..
0.'
1.0
0 -
o ,
0.2
r-
o.of---~-----
o.ol---~--------------j
.(l.'
--- --
"'0ow. 20% 40'4 .,.., 80% 1~ 12m' 1~ 16O'Mo 1110% ~.2,", 20% 4(fIJI, 6()% 80% 1(1()% 120% 140% 160')1, 180%
~~ AwnIgeTnIdIIIGNP
'''' "'"
""
_r----_--;-___,
10"
-
--- ---
~r------~-------~
--
.10%0';;-.0---;1.';;"0---::2.::-0---;;',.o;:----:':,.o:-----},.o '~0~.0---;'.':-O---::2.':-0----::'.O;:---~'.::-0--~'.'0
--
'.0
3.0 -
20.0
15.0
10.0
10
f-~-------- ..-~-~
0.0t---------''----~-----1
---
00
---
-1.0
.0 ---:27.0 ---:,':".0---,':".0--..J,.o
-2.0';;-.0- - - : ,7 -s.OOL.
O- - - :'-=-.O---::2'::-0---::,-=-o- - - :.'::-0---!,.o
1.'
OA
0.2
0.0
---
0.01--------------=-..,
---
.(l.5
.1.00.';;-0---;,.70 ---:;2.';;"0---:;,.7
0 ---;,.':-0---:~o .(l~.LO---:,~.0---::2':".0---::3'::-.0---:,7.0---:~0
Figure 8. (Continued.)
Trade/GDP, there is a sharp negative relation between the rating and volatility.
There is no significant association between PB and volatility. None of these
three measure does well in explaining the cross-section of skewness. There is a
negative relationship between the Trade/GDP variable and skewness, but this
is not significant at conventional levels. All three measures show a negative
The behavior of emerging market returns 171
association with kurtosis. Low composite rating, low Trade/GDP and low PB
are associated with higher kurtosis. PB has considerable success in explaining
the cross-section of fJ-values which is consistent with Ferson and Harvey's
(1997) evidence that PB is an instrument for time-varying risk exposure. Low
PB is associated with higher risk. Finally, there is a positive relationship
between the ICRG composite risk and correlation with the world market.
Consistent with Bekaert and Harvey (1997a), there is a positive relationship
between Trade/GDP and correlation with the world.
CONCLUSIONS
This chapter has further explored the behavior of emerging market returns.
The first goal of the paper was to go beyond the mean and variance and
investigate skewness and kurtosis. Indeed, there is considerable research interest
in asymmetric variances, semi-moment analysis and down-side optimization.
These advances make a lot of sense in emerging markets where the returns
distributions depart from the usual normal assumption. We showed that the
deviations from normality are persistent in the emerging market returns, show
no evidence of disappearing in the near future, and are time-varying in nature.
We then explored what matters for emerging market investment. The tradi-
tional fJ-risk paradigm is problematic in emerging markets because a number
of the markets are unlikely fully integrated into world capital markets. Indeed,
in a completely segmented market, country variance (which is usually consid-
ered idiosyncratic) is the appropriate measure of risk. We explored a group of
risk attributes that have been successfully applied in developed markets. We
found that a number of these attributes such as the International Country Risk
Guide's composite risk, trade to GDP and price to book value are useful in
identifying high and low expected return environments.
Finally, we tried to link our attribute analysis to the behavior of the emerging
market returns. The risk attributes, not only discriminate among the mean
returns, they also offer information about other moments. For example, we
found that low ICRG composite ratings are strongly associated with high
volatility, high excess kurtosis and low correlations with the world benchmark.
This analysis suggests that models of formal asset allocation in emerging
markets need to go beyond both the attribute sorting portfolio approach and
the simple mean-variance analysis. In these markets, the attributes contain
information about volatility, correlation, skewness and kurtosis as well as
expected returns.
REFERENCES
Asness, Cliff S., John M. Liew and Ross L. Stevens (1996). Parallels Between the Cross-sectional
Predictability o/Stock and Country Returns. Working paper, Goldman Sachs Asset Management,
New York, 1996.
172 G. Bekaert et al.
Bakshi, Gurdip S. and Zhiwu Chen (1994). Baby boom, population aging, and capital markets.
Journal of Business, 67, 165-202.
Banz, RolfW. (1981). The relationship between return and market value of common stocks. Journal
of Financial Economics, 9, 3-18.
Bekaert, Geert (1995). Market integration and investment barriers in emerging equity markets.
World Bank Economic Review, 9,75-107.
Bekaert, Geert and Campbell R. Harvey (1995). Time-varying world market integration. Journal of
Finance, SO, 403-444.
Bekaert, Geert and Campbell R. Harvey (1997a). Emerging equity market volatility. Journal of
Financial Economics, 43, 29-78. Duke University and Stanford University.
Bekaert, Geert and Campbell R. Harvey (1997b). Foreign Speculators and Emerging Equity Markets.
Working paper, Duke University and Stanford University.
Bekaert, Geert and Michael Urias (1996). Diversification, integration and emerging market closed
end funds. Journal of Finance, forthcoming.
Bera, Anil and Carlos Jarque (1982). Model specification tests: a simultaneous approach. Journal of
Econometrics, 20, 59-82.
Berk, Jonathan (1996a). A critique of size related anomalies. Review of Financial Studies,
forthcoming.
Berk, Jonathan (1996b). An Empirical Re-examination of the Relation Between Firm Size and Return.
Unpublished manuscript, University of British Columbia.
Black, Fischer (1976). Studies of stock market volatility changes. Proceedings of the 1976 meetings
of the American Statistical Association, Business and Economics Statistics Section, 177-181.
Brennan, MJ. (1993). Agency and Asset Pricing. Unpublished manuscript, UCLA and London
Business School.
Campbell, John Y. and Yasushi Hamao (1992). Predictable bond and stock returns in the United
States and Japan: a study of long-term capital market integration. Journal of Finance, 47, 43-70.
Chan, K.c., G. Andrew Karolyi and Rene Stulz (1992). Global financial markets and the risk
premium on US equity. Journal of Financial Economics, 32,137-168.
Chan, K.C., Yasushi Hamao, and Josef Lakonishok (1991). Fundamentals and stock returns in
Japan. Journal of Finance, 46, 1739-1744.
Chen, Nai-fu, Richard R. Roll and Stephen A. Ross (1986). Economic forces and the stock market.
Journal of Business, 59, 383-403.
Christie, A.A. (1982). The stochastic behavior of common stock variances: value, leverage, and
interest rate effects. Journal of Financial Economics, 23, 407-432.
Diamonte, Robin, John M. Liew and Ross L. Stevens (1996). Political risk in emerging and devel-
oped markets. Financial Analysts Journal, forthcoming.
Erb, Claude B., Campbell R. Harvey and Tadas E. Viskanta (1994). Forecasting international equity
correlations. Financial Analysts Journal, November-December, 32-45.
Erb, Claude 8., Campbell R. Harvey and Tadas E. Viskanta (1995a). Country risk and global equity
selection. Journal of Portfolio Management, 21, 74-83.
Erb, Claude B., Campbell R. Harvey and Tadas E. Viskanta (1995b). Inflation and world equity
selection. Financial Analysts Journal, November-December, 28-42.
Erb, Claude B., Campbell R. Harvey and Tadas E. Viskanta (1996a). Expected returns and
volatility in 135 countries. Journal of Portfolio Management, 32-48. Also in http://www.duke.edu/
- charvey/Country _risk/couindex.htm.
Erb, Claude B., Campbell R. Harvey and Tadas E. Viskanta (1996b). Political risk, economic
risk and financial risk. Financial Analysts Journal, 52, 28-46. Also in http://www.duke.edu/
- charvey/Country _risk/couindex.htm.
Erb, Claude B., Campbell R. Harvey and Tadas E. Viskanta (1997). Demographics and Inter-
national Investment. Financial Analysis Journal, 53, 14-28. Also in http://www.duke.edu/
- charvey/Country_risk/couindex.htm.
Errunza, Vihang R. (1994). Emerging markets: some new concepts. Journal of Portfolio Management,
Spring, 82-87.
Errunza, Vihang R. and Etienne Losq (1985). The behavior of stock prices on LDC markets. Journal
of Banking and Finance, 9, 561-575.
The behavior of emerging market returns 173
Fama, Eugene F. and Kenneth French (1992). The cross-section of expected stock returns. Journal
of Finance, 59, 427-465.
Ferson, Wayne E. and Campbell R. Harvey (1991). The variation of economic risk premiums.
Journal of Political Economy, 99, 285-315.
Ferson, Wayne E. and Campbell R. Harvey (1993). The risk and predictability of international
equity returns. Review of Financial Studies, 6,527-566.
Ferson, Wayne E. and Campbell R. Harvey (1994). An exploratory investigation of the fundamental
determinants of national equity market returns. In Jeffrey Frankel (ed.), The Internationalization
of Equity Markets. Chicago: University of Chicago Press, 59-138.
Ferson, Wayne E. and Campbell R. Harvey (1996). Fundamental determinants of national equity
market returns: a perspective on conditional asset pricing. Journal of Banking and Finance,
forthcoming.
Garcia, Rene and Eric Ghysels (1994). Structural Change and Asset Pricing in Emerging Markets.
Unpublished working paper, Universite de Montreal.
Goetzmann, William N. and Philippe Jorion (1996). Re-emerging Markets. Working paper, Yale
University.
Hansen, Lars P. (1982). Large sample properties of generalized method of moments estimators.
Econometrica, 50, 1029-1054.
Harlow, W. Van (1993). Political Risk and Asset Markets. Fidelity Investments Monograph,
Boston: MA.
Harvey, Campbell R. (1991). The world price of covariance risk. Journal of Finance, 46,111-157.
Harvey, Campbell R. (1993). Portfolio enhancement using emerging markets and conditioning
information. In Stijn Claessens and Shan Gooptu (eds.), Portfolio Investment in Developing
Countries. Washington: The World Bank Discussion Series, pp. 110-144.
Harvey, Campbell R. (1995). Predictable risk and returns in emerging markets. Review of Financial
Studies, 8, 773-816.
Harvey, c.R. and A. Siddique (1997). Conditional Skewness in Asset Pricing Tests. Working paper,
Fuqua School of Business, Duke University.
Heston, Steven L, K. Geert Rouwenhorst and Roberto E. Wessels (1995). Capital market integration
and the international cost of funds. Journal of Empirical Finance, 2, 173-198.
Jegadeesh, Narasimhan and Sheridan Titman (1993). Returns to buying winners and selling losers:
Implications for stock market efficiency. Journal of Finance, 48, 65-92.
Kandel, S. and R.F. Stambaugh (1995). Portfolio inefficiency and the cross-section of expected
returns. Journal of Finance, 50,157-184.
Keppler, A. Michael (1991). The importance of dividend yields in country selection. Journal of
Portfolio Management, Winter, 24-29.
Keppler, A. Michael and Heydon D. Traub (1993). The small country effect: small markets beat
large markets. Journal of Investing, Fall, 17-24.
Lintner, John (1965). The valuation of risk assets and the selection of risky investments in stock
portfolios and capital budgets. Review of Economics and Statistics, 47, 13-37.
Longin, Francois and Bruno Solnik (1995). Is the correlation in international equity returns con-
stant: 1960-19901 Journal of International Money and Finance, 14, 3-26.
Merton, Robert C. (1973). An intertemporal capital asset pricing model. Econometrica, 41, 867-887.
Nelson, Daniel B. (1991). Conditional heteroskedasticity in asset returns. Econometrica, 59,
347-370.
Roll, Richard and Stephen A. Ross (1994). On the cross-sectional relation between expected returns
and betas. Journal of Finance, 49, 101-121.
Ross, S.A. (1976). The arbitrage theory of capital asset pricing. Journal of Economic Theory, 13,
341-360.
Sharpe, William (1964). Capital asset prices: a theory of market equilibrium under conditions of
risk. Journal of Finance, 19,425-442.
Solnik, Bruno (1983). International arbitrage pricing theory. Journal of Finance, 38, 449-457.
Stulz, Rene (1981a) On the effects of barriers to international investment. Journal of Finance,
36, 923-934.
Stulz, Rene (1981b). A model of international asset pricing. Journal of Financial Economics, 9,
383-406.
IAN DOMOWITZl, JACK GLEN2 and ANANTH MADHAVAN 3
I Northwestern University, 2 International Finance Corporation, 3 University of Southern California
ABSTRACT
We examine empirically the impact of international cross-listing on stocks traded on the Mexican
Stock Exchange. We focus on the Mexican market because equity issues in Mexico differentiate
between foreign and domestic investors, permitting tests of hypotheses about the effects of cross-
listing on different shareholder classes. For shares open to foreign ownership, ADR introduction is
associated with higher volatility unrelated to volume and lower liquidity, consistent with the
hypothesis that cross-listing induces market fragmentation. However, the implicit bid-ask spreads
for shares open to foreign ownership decrease although there are no such changes for shares
restricted to domestic ownership. This finding is consistent with the hypothesis that cross-listing
also creates in greater competition for order flow in some segments of the market. These effects
vary with firm characteristics such as market capitalization. Internal capital market segmentation,
induced by foreign ownership restrictions is an important factor in these results. Cross-listing is
thus more complex than previously believed, with differential impacts on foreign and domestic
investors.
INTRODUCTION
Trading offoreign stocks in the Untied States has grown exponentially as investors
recognize the need for international diversification while seeking higher expected
returns. Accompanying this growth is an increase in the number of American
Depository Receipt (ADR) listings on US exchanges. 1 While foreign corporations
presumably view ADR listing as value enhancing, it is not clear that all classes of
domestic investors are better off as a result. Indeed, relatively little is known about
the effect of foreign listings on the domestic market. 2 Changes in liquidity, volatility,
I Foreign companies wishing to trade equities (and debt) in the US market often achieve this
goal by issuing American Depository Receipts (ADRs). The ADR represents shares held by a
trustee in the local market. The receipts, issued by a US depository (usually a large bank), are
then treated as US securities for clearance, settlement, transfer, and ownership purposes. The
depository is responsible for the payment of dividends (in dollars) to investors, for the distribution
of information about the issuer, and for conversion between shares and ADRs. Foreign stocks may
also be traded exclusively in the 'pink sheet' or over-the-counter market.
2 Chan et al. (1994) analyze the intraday price behavior of dually listed stocks; Forster and
George (1993, 1994) study the pattern of return variances for international cross-listing; Alexander
et al. (1987, 1988), and Howe et al. (19.93) discuss the impact of international listing on US stocks;
Kleidon and Werner (1993) examine the effect of cross-listing on UK stocks; Jorion and Schwartz
(1986) and Foerster and Karolyi (1993, 1994) analyze cross-listed Canadian stocks; Jayaraman
et al. (1993) examine the effect of US ADR listing on domestic stock prices; and Freedman (1990)
provides a theoretical model of international lisitng.
175
R. Levich (ed.). Emerging Market Capital Flows, 175-192.
CI 1998 Kluwer Academic Publishers.
176 I. Domowitz et al.
and the cost of trading associated with order flow migration following cross-listing
may adversely affect the domestic market. In a segmented market, the benefits and
costs of cross-listing may not be shared equally by all investor classes, even if cross-
listing is value enhancing. These issues are especially important for smaller
European capital markets facing new competition resulting from economic integ-
ration and for emerging markets that are less liquid than their foreign competitors.
Debates over consolidation and fragmentation also make the potential
impacts of international cross-listing controversial in academic settings. Models
of multi-market trading (e.g., Pagano, 1989) suggest that trading gravitates to
the most liquid market centers. In an international context, these theories
suggest that cross-listing, by providing another venue for trading, may induce
order flow migration to overseas markets that offer lower transactions costs.
Faced with a loss of order flow, liquidity in the domestic market may decrease,
thereby increasing bid-ask spreads and price volatility. The presence of traders
with private information could exacerbate such migration (Chowdhry and
Nanda, 1991) since traders would prefer to trade in markets that offered the
most transparency, i.e., information on prices, quotes, and trades. Alternatively,
if markets are transparent and the total number of investors in both markets
increases after cross-listing, liquidity and price efficiency can increase. In
addition, competition for order flow may also lead to the dissipation of dealer
rents in the local market arising from new foreign competition, as in Madhavan
(1995), thereby lowering spreads.
Recent evidence provided by Foerster and Karolyi (1994) for Canadian
securities listed on US exchanges, however, suggests that a strict dichotomy
between market consolidation through enhanced competition and fragmen-
tation through order flow migration is inappropriate. Rather, cross-listing may
result in fragmentation as well as greater inter-market competition in those
segments of the equity market facing order flow migration. This is likely to be
especially important in emerging markets where investment barriers as restric-
tions on foreign equity ownership are relatively common. 3 We refer to this
occurrence as market segmentation.
We address these issues using new data on daily prices and volumes from
1989-1993 obtained from the Mexican Stock Exchange, the Bolsa Mexicana
de Valores (BMV). The Mexican market is of particular interest for this study.
Previous studies of international listing have focused almost exclusively on
developed economies, especially Britain, Canada, and Japan, for which data
are readily available. However, little is known about the impact of cross-listing
on an emerging market such as Mexico. International listing may have a
significant impact on an emerging market because these markets are generally
much smaller and less liquid than the foreign markets where cross-listing
occurs. Foreign listing is also likely to have a significant impact on the domestic
3 Examples include China, Thailand, and Mexico. See Eun and lanakiramanan (1986) for a
more detailed description of these investment barriers.
Cross-listing, segmentation and foreign ownership restrictions 177
Mexican !Darket, since foreigners (mostly US nationals) account for over 27%
of holdings and up to 75% of trading. Further, the BMV is open for trading
during much of the time that the United States' equity markets are open.
Consequently, an analysis of trading in Mexico provides a way to better
understand the effects of inter-market competition. Finally, the Mexican data
include information on multiple series of stock for some individual companies,
with alternative forms of foreign ownership rights. As noted above, the effects
of market segmentation are most likely to be observed in markets with such
distinctions.
We find strong evidence for the market segmentation hypothesis. In particu-
lar, for those series open to foreign ownership, cross-listing is associated with
an increase in volatility unrelated to volume, as well as a reduction in liquidity.
These findings are consistent with a fragmentation hypothesis emphasizing the
diversion of order flow. However, such shares also experience a decline in
implicit bid-ask spreads following ADR introduction. The decline in spreads
may be explained by increased competition among domestic liquidity providers
as they attempt to retain order flow in these issues following cross-listing. No
such effects are present in a control group of matched stocks, indicating that
our results are not driven by market-wide phenomena or macroeconomic
factors.
Interestingly, there is no discernible pattern to the changes in liquidity
and the implied bid-ask spread for shares whose ownership is restricted to
domestic nationals, although there is an increase in the price volatility
unrelated to volume for such shares. This suggests that the changes in
liquidity observed in other series were the result of a migration of foreign
investors to US markets. Finally, although ADR listing is associated with
positive excess returns, these benefits accrue largely to those series open to
foreign investors prior to cross-listing. Our results thus provide strong
support for the market segmentation hypothesis, which acknowledges both
the costs of order flow fragmentation following the cross-listing as well as
the benefits of increased inter-market competition. This implies in turn that
the benefits of cross-listing are not evenly spread among all classes of
investors. In addition, we find some evidence that the effects of order flow
migration are concentrated in larger stocks, possibly because these stocks
are favored by foreign investors prior to cross-listing.
foreign brokers or traders are allowed in Mexico. In 1993, there ,were 884
equity instruments traded on the BMV. The market, however, is highly concen-
trated in a few issues, with Telefonos de Mexico stock alone accounting for
18% of trading volume. In 1993, 26 equity instruments were also traded as
ADRs in the United States. This set of 26 series, issued by 16 of the largest
firms on the BMV, is the focus of our study.
Mexican equity instruments are of particular interest for our study in that
each company may issue several different series of shares, each with different
rights or shareholder bases. All series have full participation in the earnings of
the firm.4 The major divisions are into series A, B, and C (or L). Series A
shares may legally be held only by Mexican nationals and account for at least
51 % of company voting rights. s Series B shares are open to foreigners and
institutional investors and are limited to 49% of the ownership.6 Both A and
B shares can be further divided into fixed (lor F) and variable (2 or V)
components, all of which carry equal rights. Under a company's statutes, it
must carry a minimum level of capital, the fixed part, which can only be
changed by vote of shareholders at an extraordinary meeting. The variable
part changes whenever rights are issued or the company decides to buy back
shares. Series C shares are open to all investors, but are limited to 30% of
total capital. Series Land 0 are also available to all investors, but carry limited
or no voting rights. These arrangements ensure that Mexican nationals control
the voting of listed companies. Similar divisions are found in other countries
where public policy seeks to restrict or control the foreign ownership of capital.
Alternatively, foreign ownership restrictions may arise endogenously, as shown
by Stulz and Wasserfallen (1995) in the case of Switzerland. A variety of other
series designations exist, but are not relevant to the sample considered here.
The exchange operates from 8:30 a.m. to 4:00 p.m. local time from Monday
to Friday, so that the market first opens at the same time as the New York
Stock Exchange, where most Mexican ADRs are traded, but closes an hour
earlier. Orders are either sent to the BMV computer systems or worked by
floor brokers through open outcry.7 Preopening orders can be entered by
computer and crossed automatically subject to certain price limits. Block trades,
like all transactions, must occur on the trading floor and are subject to the
rules governing the auction process. In summary, the trading mechanism on
the BMV closely resembles a standard continuous auction market.
4 The exception is a new class of 'D' shares issued by very few companies. Such series are not
represented in our data.
5 Foreigners may participate only through the Mexican national development bank (NAFIN)
trust, if at all. In this case, the series designation changes from A to N or CPO. Voting rights then
are held by NAFIN.
6 If the B shares are for a financial group, they can be owned only by Mexican institutional
investors. Financial groups are not among the companies in the subset of shares experiencing an
ADR introduction.
7 The computer systems handled little volume and small orders over the time period of this study.
Cross-listing, segmentation and foreign ownership restrictions 179
The data
The data used in this study were obtained from the BMV, and until now have
not been publicly available. The sample used here consists of daily observations
on prices, returns, and share volumes of 26 equity series. These series represent
16 firms over the period September 1989 to July 1993, all of which experienced
an ADR listing over the sample period. In addition, we gathered data on 10
non-ADR companies with similar trading frequencies and characteristics to
form a control sample to check the validity of our results. On average, there
were 828 trading days per series, with some variation due to the introduction
of new series types over the full period.
The data were screened and cross-checked in a number of ways to detect
potential errors. The major problem encountered was the fact that companies
routinely change series designations after corporate actions, although the under-
lying changes often have no effect on ownership or voting rights. For example,
the A series of company CIFRA underwent seven changes in series designation
between January 1988 and July 1993. Companies typically took at least one
action per year that had some effect on series designation, few of which were
meaningful. Each such change was investigated using information from the
Anuario Bursatil (the BMV's year book) and data supplied from Mexican
brokerage houses.
Returns were adjusted based on information from the BMV and from the
Anuario Bursatil, as were volumes. This process is more complicated than the
usual simple adjustments for dividends and stock splits. For example, the BMV
would occasionally announce the simple cancellation of outstanding tradeable
shares in the A series. This happened to CMA (which was used as a control in
some of our analyses but is not in the ADR sample), for example, on 9/11/90,
when 3.6% of the shares traded disappeared. Share holdings were also adjusted
based on reciprocal buy/sell offers between companies under what seemed to
be the same parent umbrella. Again for CMA, this happened on 8/13/94, when
such an agreement between Compania Mexicana de Aviacion and Corporacion
Mexicana de Aviacion increased the CMA holdings by 2.6%. Corrections for
these types of problems were made manually. Screening of the data indicated
a few outliers for most series. Except for a formulaic error that affected all
series for a few months early in the sample, the outliers did not seem to be a
result of keypunch error. This error was rectified, but we did not alter the data
otherwise.
Company names, acronyms, and series identifiers for the 26 stocks our
sample are provided in Table 1. Of the 26 series, 11 are A class (including fixed
and variable) shares, 9 are B class shares, and the remaining 6 are C, L, or 0
(non-voting) shares. The majority of the ADRs in our sample are traded over-
the-counter or in the 144A market. Of the stocks in our sample, only three are
listed on an exchange and for these stocks we have data on trading in the USA
as well as in the domestic market.
180 1. Domowitz et al.
Table 1. Stock series abbreviations, names of companies, stock exchange keys (board keys), and
series by company in our sample for stocks traded on the Bolsa Mexicana de Valores for which
an ADR was introduced
Table 2 shows the data of ADR listing, the average daily trading volume (in
shares), return volatility (in percentage), and the number of trading days before
and after ADR listing available for analysis for each stock. Note that stocks
are listed by their abbreviation followed by series identifier and type. Thus,
TTOBV denotes the B (Variable) series for company Tolmex (TTO). The dates
of ADR listing vary widely across the sample, with the earliest being company
EPN (AF series), listed in December 1989 and the most recent being
Transportacion Maritima Mexicana, listed in June 1992. It is clear that there
is a wide range in trading activity and volatility across stocks. Similarly, it is
difficult to discern a pattern to the changes in volume and volatility following
ADR listing. We turn now to a more detailed examination of this issue.
ESTIMATION ISSUES
Empirical hypotheses
Consider first the case where a security is domestically traded. Market partici-
pants, both domestic and foreign nationals, buy and sell the stock according
to auction principles. In this market, it is intuitively clear that the greater the
number of participants, the deeper the market, where market depth is measured
in terms of the ability to accommodate order flow shocks without substantial
price movements.
Suppose now that the risky asset is internationally cross-listed, providing
an alternative venue for trading. Cross-listing may result in a net increase in
Cross-listing, segmentation and foreign ownership restrictions 181
Summary statistics for all stock series in our sample of stocks traded on the Bolsa Mexicana de
Valores for which an ADR was introduced. The month and year of the ADR introduction are
given under 'ADR listing'. Volume prior and Volume after are average daily shares traded, in
thousands, before and after the listing, respectively. Volatility prior and Volatility after denote the
average standard deviation of daily percentage returns, before and after the listing. Days prior and
Days after are the number of trading days in the sample before and after the listing.
an increase in the total number of traders in both markets. It follows that there
is more liquidity and greater depth. Participation by new investors should
result in more efficient aggregation of investor beliefs, increasing signal precision
and lowering the subjective variance of the price forecast. Although these effects
arise because of increased participation in the foreign market, price competition
among market centers may further reduce the costs of trading in the domestic
market (Madhavan, 1995) as domestic liquidity providers seek to retain
order flow.
An alternative hypothesis, which we term the fragmentation hypothesis,
maintains that international listing results in a migration of investors away
from the domestic market and that this results in less efficient pricing and
lower market quality. Formally, suppose intermarket price competition is
imperfect or there is a delay in disclosing transactions resulting in reduced
transparency. If the costs of trading abroad are lower in the foreign market
than in the domestic market for some current investors, then the number of
domestic investors decreases (as these investors migrate abroad), leading to a
decrease in liquidity following cross-listing. In this scenario, the foreign market
(with lower transaction costs) receives not only a portion of the pre-listing
order flow but also the order flow arising from new participants. Fragmentation
may also lead to higher variance in public beliefs because information aggrega-
tion is less efficient with fewer market participants. As a result, the subjective
variance of investors' beliefs regarding the asset's fundamental value may
increase following international listing. Factors ignored in our simple analysis
are likely to exacerbate these effects. In particular, the analysis of Chowdhry
and Nanda (1991) suggests that order flow migration may be more likely to
occur under asymmetric information. In turn, this implies that trading costs
(e.g., the bid-ask spread) are likely to increase with market fragmentation.
An intermediate view, which we term the segmentation hypothesis, combines
aspects of the two extreme theories, recognizing that the effects they seek to
highlight are not necessarily mutually exclusive.s In this view, the local market
may experience differential effects to international listing ifthere is segmentation
between the assets held by foreign and domestic customers. If foreign investors
currently trading in the domestic market obtain the greatest reduction in
trading costs by moving abroad (as seems likely), the loss of order flow is most
likely in securities where foreign ownership is significant. As with the fragmenta-
tion hypothesis, the loss of order flow is associated with a decrease in market
liquidity and an increase in volatility if intermarket arbitrage is not perfect.9
However, even in the absence of current price information, imperfect intermar-
ket competition may still benefit the domestic markets in the form of lower
transactions costs for those securities facing the greatest order flow loss.
8 In the international asset pricing literature segmented markets are markets in which either the
price of risk differs or prices are determined by different risk factors (Jorion and Schwartz, 1986).
9 However, as in Madhavan (1995), this phenomenon may be associated with a narrowing of
bid-ask spreads as domestic liquidity providers respond to the loss of order flow.
Cross-listing, segmentation and foreign ownership restrictions 183
Empirical tests of the hypotheses described above is difficult given the available
data. Our approach to testing for changes in liquidity and volatility is based
on the idea that price volatility has two components: the first component
represents the volatility arising from changes in fundamentals and imperfect
information signals, while the second component represents the volatility
related to the changes in investor holdings, i.e., to trading volume. As discussed
above, any effects of international listing on liquidity and volatility should be
associated with changes in base-level volatility and market liquidity, and can
thus be estimated. To see this, consider a simple microstructure model (motivated
by Kyle, 1985) the 'daily' price change on day t is given by
APr = pX r + Gr
In this representation, Ap is the price change, x is the net (signed) order
imbalance, G is the innovation in public beliefs due to public news announce-
ments, and p captures the sensitivity of prices to order flow. The coefficient p
reflects the effects of asymmetric information as well as, perhaps, inventory
control costs, and is hypothesized to be an inverse function of the number of
market participants. Taking the variance on both sides, we obtain
(J2(Ap,) = W, + p2(J2(X,)
where Wr is the variance of, G" the random innovation in fundamental values.
Assume further (omitting time subscripts for notational ease) that the order
imbalance, x, is the sum of N normally distributed order shocks, q, where N
represents the number of traders and q represents each trader's (signed) order
size. Then, the standard deviation of x is proportional to E [I x I] = N E [I q I] =
V, where V represents the expected volume for the day. Intuitively, higher
volumes are associated with larger expected absolute order imbalances, and
hence greater price volatility.
It follows that the price change from period to period can be represented
as
(1)
where A. (proportional to p2) is a parameter that is inversely related to the
number of market participants, and V, is the expected volume period t. We can
interpret Wr as an inverse measure of the precision of public information, which
in turn is likely to be positively related to the number of traders who choose
to gather information about the stock. More precise public information implies
that conditional expectations are more accurately centered on fundamentals
so that the changes in expected beliefs have smaller variance. Hence, WI is
smaller the greater the precision of public information. Similarly, the parameter
A. is an inverse measure of market liquidity. More active trading implies lower
values of ),.
184 I. Domowitz et at.
i't = Po + PI ADRt
CLOt = Yo + Yl ADR t (3)
CL lt = 150 + 15 1 ADR"
where ADRt is a dummy variable taking the value 0 if date t is before ADR
listing and 1 otherwise. Under the competitive hypothesis ADR listing increases
liquidity and price volatility so that PI and Yl are negative. Under the fragmen-
tation hypothesis, ADR listing decreases liquidity and price volatility so that
PI and Yl are positive. Segmentation allows for mixed effects depending on the
asset involved.
Our model assumes that transactions occur at a single market clearing price,
but in reality, transactions occur at bid or ask prices. The results of Foerster
and Karolyi (1994) suggest that cross-listing is associated with changes in the
cost of trading, as measured by the bid-ask spread. Unfortunately, even though
it is straightforward to incorporate a spread into our model, we cannot directly
make inferences about the bid-ask spread without quotation data. In this
section, we discuss an alternative approach to making indirect inferences about
changes in the cost of trading accompanying ADR listing.
Models of dealer trading imply that any change in transaction costs
(measured by the implicit spread) following ADR listing should be in the
opposite direction to the change in liquidity and is likely to be in the same
direction as the change in volatility. If foreign listing increases market fragmen-
tation (competition, market quality should decrease (increase) and transaction
costs (and hence implicit spreads) should increase (decrease) as well. Under the
segmentation hypothesis, both liquidity and spreads may decrease as the result
of cross-listing.
To see this, suppose that ADR listing results in a reduction in the number
of investors participating in trading, so that A. rises and liquidity falls. In a
segmented market, this loss is most likely to occur in stocks with significant
Cross-listing, segmentation and foreign ownership restrictions 185
foreign ownership, since foreign investors are more likely than domestic inves-
tors to obtain a reduction in their trading costs by trading in the US ADR
market. However, since dealers in these issues also face more price competition
for smaller orders from the foreign market, spreads may narrow after cross-
listing.
Roll (1984) provides a technique using the covariance in successive price
changes to make inferences about the size of the bid-ask spread when bid-ask
quotations are unavailable. The idea is simple: bid-ask bounce causes the serial
covariance to be negative, and this can be computed using transaction prices
alone. Roll's model has been extended by George et al. (1991; hereafter referred
to as GKN) among others.
The GKN procedure builds on the intuition of the Roll (1984) model but
offers several extensions including adjustments for potential autocorrelation.
Our approach to estimation differs slightly from GKN in the way we adjust
for autocorrelation and by our use of generalized method of moments (GMM)
to jointly estimate the underlying parameters of the implicit spread. The GMM
technique is attractive for several reasons, but most importantly because we
can construct a new statistic to test the change in implicit spreads following
ADR listing. Domowitz, Glen and Madhavan (1996) provide a formal treat-
ment of this test statistic.
EMPIRICAL RESULTS
We estimate the time-varying parameter model in equations (4) and (5) using
Hansen's generalized method of moments. The GMM procedure produces
consistent parameter estimates under very general distributions for the stochas-
tic processes generating the data, and the standard errors correct for hetero-
skedasticity and autocorrelation. Further, with GMM the parameters for
individual series within a company can be estimated jointly. This is particularly
important because the cross-correlation in errors across different series for the
same company are likely to contain valuable information.
Table 3 presents the GMM estimates, with corresponding autocorrelation
and heteroskedasticity consistent standard errors in parentheses, for all 26
series in the sample. The estimates confirm the general findings in the literature
with respect to the positive price variability-volume relationship. The coefficient
Yo is positive in all cases and A.t is positive in 22 of 26 cases, with any negative
values being very close to zero.
Overall, volatility increases following the ADR listing. Averaging across all
series, the volatility coefficient OCot rises from 0.105 prior to listing to 0.226
afterwards. This increase is observed in 21 of the 26 series. The result is
consistent with empirical evidence (e.g. Jayaraman et at., 1993), which suggests
that the return variance increases following international listing. The sensitivity
186 I. Domowitz et al.
Series Yo YI Do DI Po PI
CEMA 0.122 0.463 0.270 -0.110 1.998 6.868 0.233
(0.012) (0.043) (0.060) (0.076) (0.393 ) (3.228)
CEMB 0.161 0.355 0.033 0.189 0.013 0.611 0.167
(0.026) (0.047) (0.063) (0.078) (0.288) (0.294)
CIFA 0.034 0.097 0.200 -0.093 -0.002 0.000 0.096
(0.005) (0.011) (0.079) (0.090) (0.001) (0.001)
CIFB 0.030 0.121 0.311 -0.049 0.047 -0.038 0.212
(0.005) (0.013) (0.109) (0.123) (0.056) (0.064)
CIFC 0.022 0.028 0.085 0.119 0.002 -0.003 0.073
(0.005) (0.006) (0.144 ) (0.149) (0.001 ) (0.001)
COMBV 0.032 0.175 0.190 -0.024 0.130 0.017 0.084
(0.004) (0.006) (0.078) (0.093) (0.038) (0.094)
SANA 0.043 -0.021 0.032 0.068 0.004 0.450 0.044
(0.006) (0.006) (0.061 ) (0.075) (0.001 ) (0.078)
SANAV 0.040 -0.011 0.183 -0.021 0.002 0.055 0.047
(0.005) (0.006) (0.068) (0.082) (0.001 ) (0.023)
EPNAF 0.002 -0.002 0.026 -0.029 0.016 0.900 0.240
(0.001) (0.001) (0.095) (0.094) (0.023) (0.430)
LIVO 0.Q78 0.104 0.170 0.034 1.157 1.892 0.144
(0.008) (0.023) (0.053) (0.094) (0.085) (1.341)
LIVC 0.072 0.218 0.283 -0.214 0.987 0.222 0.192
(0.008) (0.029) (0.065) (0.087) (0.042) (0.175)
FEMBV 0.082 0.128 0.158 0.066 -0.065 0.056 0.184
(0.007) (0.017) (0.054) (0.077) (0.066) (0.067)
GCAAF 0.094 0.371 0.067 0.252 144.7 -144.6 0.248
(0.023) (0.054) (0.066) (0.097) (74.89) (74.89)
MASAV 0.131 0.077 0.281 -0.167 1.103 3.783 0.094
(0.013) (0.024) (0.066) (0.091 ) ( 1.307) (2.447)
MASBV 0.106 0.113 0.322 -0.263 2.176 0.413 0.129
(0.012) (0.021 ) (0.060) (0.090) ( 1.260) (1.566)
SYNA 0.013 0.001 0.168 0.128 -0.Q15 0.302 0.088
(0.003) (0.004) (0.099) (0.149) (0.024) (0.043)
SYNB 0.113 0.007 0.145 0.063 -0.945 15.91 0.084
(0.003) (0.004) (0.093) (0.114) (0.435) (7.196)
CERB 0.006 0.044 0.104 0.042 24.97 -16.21 0.085
(0.002) (0.006) (0.089) (0.107) (5.833) (6.491 )
PONA 0.001 0.0008 0.067 -0.018 635.1 -583.2 0.152
(0.001 ) (0.0002) (0.061) (0.089) (296.5) (295.8)
PONB 0.006 -0.001 0.270 -0.031 -0.001 0.082 0.085
(0.001) (0.001) (0.051) (0.081 ) (0.004) (0.032)
TELA 0.043 0.174 0.389 -0.324 0.000 -0.048 0.195
(0.005) (0.014) (0.085) (0.095) (0.000) (0.012)
TELL 0.068 0.008 0.122 -0.083 0.004 0.003 0.105
(0.007) (0.011) (0.071) (0.096) (0.002) (0.003)
TTOBV 0.114 0.201 0.312 -0.187 0.011 7.063 0.185
(0.012) (0.027) (0.074) (0.092) (0.037) (3.125)
TMMA 0.327 0.007 0.029 0.207 86.17 -74.67 0.064
(0.065) (0.074) (0.109) (0.122) (7.459) (16.73)
TMML 0.409 -0.060 -0.048 0.253 383.3 -139.9 0.077
(0.067) (0.078) (0.079) (0.100) (295.5) (309.5)
VITO 0.658 0.743 0.217 -0.084 82.09 -82.29 0.090
(0.054) (0.130) (0.052) (0.076) (51.22) (52.23)
Coefficient estimates and heteroskedasticity-consistent standard errors (in parentheses) for the
liquidity model given by equations (4) and (5). The model is estimated by generalized method-of-
moments, with equations for multiple stock series within a single company estimated jointly.
Volume is measured in tens of millions of shares. The series represent all stocks traded on the
Bolsa Mexicana de Valores for which an ADR was introduced.
Cross-listing, segmentation and foreign ownership restrictions 187
of current volatility to past volatility shocks does not appear to change in any
systematic fashion. The base coefficients imply that ;, is higher, therefore liquid-
ity is lower, following ADR listing in a majority of the 26 series. On average,
liquidity as measured by;, is lower by 26% after the listing. These findings
combined with those pertaining to volatility support the fragmentation
hypothesis.
This summary of results obscures an important feature of the problem. The
nature of the individual series determines the degree of foreign ownership in
the Mexican market, and A series cannot be owned by foreigners. Foreign
cross-listing cannot therefore result in a migration of foreign investors to the
ADR markets for these shares. Unless cross-listing is associated with substantial
migration of domestic investors, it is unlikely that domestic market liquidity
will suffer in the A series shares. On the other hand, series B, C, L, and 0
shares are traded by foreigners in the home market. Consequently, if foreign
investors migrate to US markets (as appears likely since most foreign investors
in Mexico are US nationals), then liquidity may decrease in these shares.
Indeed, in the B shares, liquidity tends to fall (i.e., /31 rises in seven cases
and falls in two cases) while volatility rises in all but one case. Series 0 shares
uniformly show an increase in volatility, and the single estimated increase in
liquidity is estimated with a large standard error. C shares also exhibit higher
volatility, while the L shares experience no significant change in volatility or
liquidity. The L shares are similar to C, but carry limited voting rights, and
the effect of cross-listing appears to have no effect on the home market trading
of such instruments. On the other hand, series A shares show very mixed
results, as measured by changes in ;,. Volatility rises significantly in six cases,
but falls for three series, with the remaining results being statistically
insignifican t.
We also note that simple increases in trading volume after ADR introduction
are distributed rather randomly, regardless offoreign ownership rights. Overall,
62% of the series experienced some increase in volume, but this figure is not
statistically significantly different from 50%. The figure is slightly higher for
the A shares, at 67%. Higher volume is observed for 63% of the B shares, and
for about 60% of all shares with foreign ownership rights prior to interna-
tionallisting.
In summary, our results thus far suggest that ADR listing is associated with
an increase in base-level volatility unrelated to volume, as well as an increased
sensitivity of price variability to volume. The latter finding is indicative of less
liquidity following ADR listing. These findings are consistent with the fragmen-
tation and segmentation hypotheses.
We also computed formal tests of changes in implicit spreads after ADR listing.
At first glance, the change in the spread is largely negative. The estimated spread
decreases in 17 of 23 series that experienced some change. The impression
188 I. Domowitz et al.
Cross-sectional evidence
Percentage
y=1 CONS P P/BV MCAP LR predicted
This table contains maximum likelihood estimates for logit models of the form PrE y = 1] = F(XfJ);
F(z) = eZ /1 + eZ The dependent variable, y, takes on the value of 1 if i. (a liquidity parameter,
higher values of which indicate lower liquidity) increases; if s (the bid-ask spread) decreases; and
if i. increases and s decreases for the same stock series. All changes are measured relative to the
introduction of the ADR for any given series. The explanatory variables are: a constant term
(CONS), price relative to book value (P/BV), share price (P), and market capitalization (MCAP),
all measured as of the date of ADR introduction. LR is the significance level of a test of the joint
significance of p, P/BV, and MCAP Percentage predicted is the percentage of cases correctly
predicted by the model. The sample consists of 17 series for which corporate data were available.
volatility and liquidity in a control sample of 10 matched stocks that did not
have ADRs. They found no changes in base-level volatility and liquidity in the
control group over the period during which ADRs were introduced in other
comparable securities. Similarly, the control group exhibited no significant
price effects around any of the ADR listing dates relative to the behavior of
the series for which ADRs were listed. This leads to the conclusion that
economy-wide changes over the period are not responsible for the changes we
observe in the market for securities that experience cross-listing. The effects of
ADRs did not spread to the remainder of the market, ruling out negative
externalities along the dimensions examined here.
CONCLUSION
ACKNOWLEDGMENTS
This paper was prepared for the New York University Salomon Center
Conference on Emerging Markets. We thank Rene Garcia and other seminar
participants for their helpful comments. Financial support from the World
Bank is gratefully acknowledged. Expert research assistance and several helpful
suggestions were provided by Mark Coppejans. The comments and opinions
contained in this paper are those of the authors and do not necessarily reflect
those of the International Finance Corporation, or the World Bank.
REFERENCES
Alexander, G., C. Eun and S. lanakiramanan (1987). Asset pricing and dual listing on foreign capital
markets: a note. Journal of Finance, 42,151-158.
Alexander, G., C. Eun and S. lanakiramanan (1988). International listings and stock returns: some
empirical evidence. Journal of Financial and Quantitative Analysis, 23, 135-151.
Bailey, W. and V.P. Chung (1994). Exchange Rate Fluctuations, Political Risk, and Stock Returns:
Some Evidencefrom an emerging Market. Working paper, Cornell University.
Chan, K.C., W. Fong and R. Stulz (1994). Information, trading, and stock returns: lessons from
dually-listed securities. Journal of Banking and Finance, forthcoming.
Chowdhry, B. and V. Nanda (1991). Multi-market trading and market liquidity. Review of Financial
Studies, 4, 623-656.
Domowitz, I., 1. Glen and A. Madhavan (1996). International Cross-Listing, Foreign Ownership
Restrictions, and Order Flow Migration: Evidence from an Emerging Market. Working paper,
University of Southern California.
Domowitz, I., 1. Glen and A. Madhavan (1997). Market segmentation and stock prices: evidence
from an emerging market. Journal of Finance, 52,1059-1086.
Eun, C. and S. lanakiramanan (1986). A model of international asset pricing with a constraint on
the foreign equity ownership. Journal of Finance, 41, 897-914.
Foerster, S.R. and G.A. Karolyi (1993). International listings of stocks: the case of Canada and the
US. Journal of International Business Studies, 24, 763-784.
Foerster, S.R. and G.A. Karolyi (1994). Multimarket Trading and Liquidity: A Transaction Data
Analysis of Canada-US Interlistings. Working paper, Ohio State University.
Forster, M.M. and T. George (1993). Volatility, Trading Mechanisms, and International Cross-
Listing. Working paper, Northwestern University.
Forster, M.M. and T. George (1994). Trading Hours, Information Flow, and International Cross-
Listing. Working paper, Northwestern University.
192 I. Domowitz et al.
Freedman, R. (1990). A Theory of the Impact of International Cross-Listing. Working paper, Stanford
University.
George, T., G. Kaul and M. Nimalendran (1991). Estimation of the bid-ask spread and its compo-
nents: a new approach. Review of Financial Studies, 4, 623-656.
Grossman, S (1976). On the efficiency of competitive stock markets when traders have diverse
information. Journal of Finance, 31, 573-585.
Hansen, L. (1982). large sample properties of generalized method of moment estimators.
Econometrica, SO, 1029-1084.
Harvey, A.C. (1981). The Econometric Analysis of Time Series. New York: John Wiley & Sons.
Howe, J., J. Madura and A. Tucker (1993). International listing and risk. Journal of International
Money and Finance, 12, 99-110.
Jayaraman, N., K. Shastri and K. Tandon (1993). The impact of international cross-listings on risk
and return: the evidence from American depository receipts. Journal of Banking and Finance,
17,91-103.
Jorion, P. and E. Schwartz (1986). Integration versus segmentation in the Canadian stock market.
Journal of Finance, 41, 603-616.
Kyle, A. (1985). Continuous auctions and insider trading. Econometrica, 53,1315-1335.
Kleidon, A. and I. Werner (1993). Round the Clock Trading: Evidence from UK Cross-Listed
Securities. Working paper, Stanford University.
Madhavan, A. (1995). Consolidation, fragmentation, and the disclosure of trading information.
Review of Financial Studies, 8, 579-603.
Pagano, M. (1989). Endogenous market thinness and stock price variability. Review of Economic
Studies, 56, 269-288.
Roll, R. (1984). A simple implicit measure of the effective bid-ask spread. Journal of Finance, 39,
1127-1139.
Stulz, R. and W. Wasserfallen (1995). Foreign equity investment restrictions, capital flight, and
shareholder wealth maximization: theory and evidence. Review of Financial Studies, 8,1019-1057.
WILLIAM N. GOETZMANW and PHILIPPE JORION 2
I Yale School of Management. 2 University of California at Irvine
INTRODUCTION
Recent research shows that emerging markets (EMs) are distinguished by high
returns and low covariances with global market factors. These are striking
results, because of their immediate implications for international investors.
Such findings have been reported in a stream of papers, including those of
Errunza and Losq (1985) and Harvey (1995). In parallel, practitioners have
generally advocated EMs on the basis of their performance. For instance,
Divecha et al. (1992) report that, over the 5-year period ending in March 1991,
an EM Index returned 7.1 % more than the World Index. With a correlation
of 0.35 between the two indices, a mean-variance analysis reveals that investing
40% in EMs apparently would have increased returns by 4% annually relative
to the World Index, with no greater risk. This apparent 'free lunch' may help
explain the stampede into the stock markets of developing countries, where
net portfolio inflows reached $22 billion in 1995, up from only $138 million
in 1985. 1
Such results can be given differing interpretations. One view is that the
mispricing is due to market segmentations. Segmentations can occur because
of legal barriers, due to restrictions on capital movements or on the ownership
offoreign assets, or information effects, possibly due to differences in accounting
rules, language barriers and so on. Whatever the reason, segmentations imply
that assets in different countries do not display the same risk-adjusted expected
returns. In other words, foreign assets appear mispriced from the viewpoint of
domestic investors.
This first interpretation formalizes the 'free lunch' story. In practice, the
story must rely on a pricing model to adjust for risk - not an easy matter
given all the controversy surrounding the appropriate pricing model for US
equities. Further, as Bekaert and Harvey (1995) have shown, relationships may
evolve over time, as the flow of foreign investment into EMs coincides with a
financial restructuring of the market that eventually leads to close integration
with developed markets.
193
R. Levich (ed.). Emerging Market Capital Flows. 193-197.
I 1998 Kluwer Academic Publishers.
194 w.N. Goetzmann and P. Jorion
Survival has been examined in a number of papers. Brown et al. (1995), for
instance, provided analytical solutions for a process where all firms start at the
same time, but disappear as soon as they hit a lower threshold. They found
that average returns are biased upward, but that the volatility is not affected.
Volatility does affect the magnitude of the bias, however. The greater the total
risk of the series, the higher the ex post conditional mean. This is especially
important for emerging markets due to their high volatility. In addition, survival
induces all kinds of spurious relationships; Goetzmann and Jorion (1995), for
instance, examined the predictability of stock returns based on dividend yields
Table 1. Time line stock market founding dates. This table compiles the founding dates of exchanges
in cities currently within the borders of the identified countries, in chronological order up to 1975,
when standard databases started.
and found that survivorship biases the results toward finding spurious evidence
of predictability.
This particular survival process, however, may not adequately represent the
actual selection of emerging markets. Typically, emerging markets become
included in standard databases when a market has become large enough. So,
instead of the 'down-and-out' process analyzed in previous work, the actual
selection resembles an 'up-and-in' process.
Goetzmann and Jorion (1996) analyzed patterns in equity markets that
follow the latter rule for emergence. They provide simulations of a simple
model where all markets are priced fairly (according to their global p-values),
but are only retained in the sample when their capitalization exceeds a size
threshold. This allows them to compare the apparent performance of emerged
markets with their true expected returns and to develop estimates for the extent
of survivorship biases, which are found to be surprisingly high; the bias in
average return can be 5-10% for recently emerged markets. Furthermore, the
return bias increases with residual variance, as if local factors were 'priced'.
Apparently, markets that have 'emerged' appear to be mispriced solely due to
the selection process.
These simulation results are complemented by an examination of the histori-
cal performance of emerging markets. The simulations reveal empirical regulari-
ties which should be present in recent data if survivorship is an issue. In
particular, they demonstrate that returns immediately after emergence are
196 w.N. Goetzmann and P. Jorion
greater than later on, which is consistent with the simulations. They also show
that the performance of not-yet-emerged markets is typically inferior to that
of emerged markets.
INDIRECT EVIDENCE
CONCLUSIONS
Survivorship effects are akin to the 'peso problem' in the foreign exchange
market. The term was first used in the early 1980s, when forward rates on the
Mexican peso appeared to be systematically biased forecasts of future spot
rates. This was because forward rates rationally anticipated the probability of
devaluation that was not observed in the test sample period.
More generally, peso problems can be interpreted as a failure of the paradigm
of rational expectations econometrics, which requires that the ex post
Risk in emerging markets revisited 197
REFERENCES
Bailey, Warren and Peter Chung (1995). Exchange rate fluctuations, political risk and stock returns:
some evidence from an emerging market. Journal of Financial and Quantitative Analysis, 30,
541-561.
Bekaert, Geert and Campbell Harvey (1995). Time-varying world market integration. Journal of
Finance, 50, 403-444.
Bekaert, Geert, Claude Erb, Campbell Harvey and Tadas Viakanta (1996). The Behavior of
Emerging Market Returns. Working Paper, Stanford University.
Brown, Stephen J., William N. Goetzmann and Stephen A. Ross (1995). Survival. Journal of Finance,
50, 853-873.
Divecha, A., J. Drach and D. Stefek (1992). Emerging markets: a quantitative perspective. Journal
of Portfolio Management, Fall, 41-50.
Erb, Claude, Campbell Harvey and Tadas Viskanta (1995). Country risk and global equity selection.
Journal of Portfolio Management, Winter, 74-83.
Errunza, V. and E. Losq (1985). The behavior of stock price on LDC markets. Journal of Banking
and Finance, 9, 561-575.
Goetzmann, William and Philippe Jorion (1995). A longer look at dividend yields. Journal of
Business, 68, 483-508.
Goetzmann, William and Philippe Jorion (1996). Re-emerging Markets. Mimeo.
Harvey, Campbell (1995). Predictable risk and returns in emerging markets. Review of Financial
Studies, 8, 773-816.
International Finance Corporation (1995). The IFC Indexes: Methodology, Definitions, Practices.
Washington, DC: IFC.
STIJN CLAESSENS
World Bank!
This is a very nice and interesting paper which reviews and sets out the
rationales for investing in emerging markets, the experiences with investing in
emerging market equities over the last few years, and draws some lessons from
the experience in December 1994 in Mexico. The issues raised by the authors
are indeed the important ones, for investors as weIl as policy makers in emerging
markets. At the same time, this is a difficult paper to comment on: one is more
likely to disagree with the tone of its conclusions than with any of the issues
raised or the analysis done.
In general I have no problems with most of the paper, particularly since
much of it is more of a descriptive nature. I do not share the authors' somewhat
pessimistic views, however, on emerging markets. One way to sum up the
paper's and my view of emerging markets is that the authors see the glass as
half-empty, while I see it as half-full. I may be more optimistic, and, we all
have to wait and see what is the right tone for investing in the emerging
markets. But, in my opinion there is some support in the facts for a more
optimistic view. Four facts have influenced my 'half-fuIl' vision.
The tequila effect was limited and did not last long, except in Mexico. The paper
rightly points out that the Mexico crisis resulted in a overhang, the tequila
effect, in Mexico and many other emerging markets in the first quarter of
1995. But there has been a substantial correction since then (Figure 1, data
from the IFC's Emerging Markets Data Base). Stock markets in Argentina,
Indonesia, Malaysia, and Thailand, for example, soon recovered and were
at the same level or higher at the end of 1995 compared with January 1994.
Since then, equity markets have further recovered in many emerging markets.
Yes, portfolio equity flows declined in 1995, but they had already been declining
in 1994 from their record level in 1993. On the other hand, total private flows
are very resilient. There was indeed a further decline in portfolio flows in
1995 and equity flows were half of those in 1993 (Table 1, data from World
Debt Tables, 1996b, World Bank). Overall, however, private flows (portfolio
(debt, equity, FDI, commercial bank lending and other private flows) to
all developing countries actually increased between 1994 and 1995, from
$159 billion to $167 biIlion, up slightly from $154 billion in 1993 (Table 2,
data from World Debt Tables, 1996b, World Bank). FDI to developing
199
R. Levich (ed.). Emerging Market Capital Flows. 199-205.
~ 1998 Kluwer Academic Publishers.
200 S. Claessens
1,000 1,200~-----------~
Chile Argentina
~~--------------------~ 4OOL---------------------~
JM Mar May Jul Sep New Jan Mar May JiJ Sep New JM Mar May Jul Sep New Jan Mar May Jul Sep Nov
1994 1995 1994 1995
1~r---------------------~ 5OOr-------------------------~
Indonesia Brazil
60~------------------------~ 200L-------------------------~
Jan Mar May Jul Sep New JM Ma May Jul Sep New Jan Mar May Jul Sep Nov Jan Mal May Jul Sep Nov
1994 1995 1994 1995
~Or------------~ 1.200 ~------------~
Malaysia Mexico
500r-------------------------~ ~O~------------------------~
Thailarld Philippines
~L---------------------~ 200L---------------------~
Jan Mar May Jul Sep New Jan Ma May Jul Sep Nov Jan Mal May Jul Sep New Jan Mar May Jul Sep Nov
1994 1995 1994 1995
Table 1. Aggregate net private capital flows to developing countries, 1990--95 (US$ billions)
* Preliminary.
Source: World Bank, Debcor Reporting System.
Table 2. Net private capital flows to developing countries, 1990-95 (US$ billions)
" Preliminary.
b Country rankings are based on cumulative 1990--95 private capital flows received.
Source: World Bank, Debcor Reporting System and staff estimates.
fundamentals and less likely to pull out when faced with short-term uncer-
tainty or volatility.
These four factors lead me to the following explanation of what we observed
over the last few years. There was an initial portfolio adjustment in the early
1990s as emerging markets suddenly became credit unconstrained due to the
Brady debt reductions and, very importantly, due to their better macro policies
and structural reforms. Combined with low international interest rates and
initial euphoria, both in New York and Washington, private capital flows
expanded rapidly and expectations of continued high returns on emerging
stock markets became the accepted wisdom.
The Mexico crises in December 1994 has led to a sense of realism in the
market. Investors now evaluate countries and companies better, and they look
closer at the fundamentals and risks involved. The initial euphoria, and, impor-
tantly, the effects of the initial portfolio adjustment of investors in developed
Table 3. World growth summary, 1966-2005 (annual percentage change in real GDP)
Forecasts
Note: GDP measured at market prices and expressed in 1987 prices and exchange rates. Growth rates over historical intervals are computed using least g
squares regression. ;:!
a. Estimated. ;:!
~
Source: OECD national accounts statistics; World Bank data, staff estimates, and projections. ~
tv
o\;.)
204 S. Claessens
countries are now over. In addition, a different class of investors, more sophisti-
cated and with a much longer time horizons, have now started to invest (or
are looking at investing) in emerging (equity) markets.
This is overall good news, both for investors and emerging markets. The
Mexico crisis and the involvement of more sophisticated and more demanding
investors has led to increased reform efforts in emerging markets which will
make these markets more attractive, transparent, reliable and less volatile. At
the same time, the diversification benefits for investors 2 and the growth opportu-
nities in emerging markets remain. Neither do I sense that there is a great lack
of interest in emerging markets among institutional investors. If anything, these
investors are only now getting ready to commit serious long-term money.
There is just more realism on both sides. The countries now understand better
that they need to earn these flows. And the investors realize that there are no
20% or more returns at low risks forever.
What does my analysis imply for the lessons and recommendations of this
paper? The paper has four main messages, mainly for emerging markets: pursue
sound macroeconomic policies; build the institutional infrastructure for securi-
ties markets; overhaul the management of corporations; and use, selectively,
capital controls.
On sound macro: As (almost) all economists, I fully agree and, as I have argued,
this has been reinforced by the Mexico crisis.
Building financial infrastructure: Here I also agree that emerging markets need
to speed up their efforts at building better environments for investing and
trading in securities. I would actually like to reinforce one of the messages
of the paper. As the paper points out, building financial infrastructure is
partly demand-driven, so foreign influence and inflows may be very useful
to stimulate this. We have seen this in particular in transition economies,
for example in Russia, but also in other emerging markets. Opening up to
foreign capital may thus be a way of improving the infrastructure, even
though at the same time the infrastructure can be a bottleneck when attract-
ing foreign capital.
Overhauling corporations: I agree that no sustainable capital market will emerge
unless a country has well-managed and solid firms. Additionally, I agree
with the way the paper proposes this should be done: corporatize state
enterprises, and then privatize in a sensible, careful manner, preferably to a
strategic investors. I do not share the views on transition economies, how-
ever, because many, such as the Czech and Slovak Republics and Russia,
did not have a choice but to privatize quickly. These countries' political
economy would have made the alternative - poor management under state
ownership combined with massive asset stripping and profit transfers - even
worse than the admittedly imperfect privatization which they pursued.
2 The mean-variance efficient frontier of the authors is not the way to do it; the paper by Harvey
provides a much better methodology for looking at risk and return in emerging markets.
Comment 205
Capital controls: I would disagree. Yes, for some countries which have had
capital controls on in- and outflows for a long time and where these controls
have been reasonable effective (some East Asian countries fall in this cate-
gory), capital controls may continue to benefit them in terms of guarding
against some aspects of international capital flows. But, capital controls on
inflows will come at a price. We know, for example, that capital controls
are more often motivated by desires to protect domestic financial markets
than by good macro or micro-economic arguments (see, for example, Alesina
et al., 1993; Dooley, 1996 for a review). Capital controls can thus stifle
competition and innovation in the financial sector. For these countries, there
will thus be a tradeoff between better macro control and a more efficient
financial sector. This tradeoff will mainly exist for capital inflows (and then
only for some countries); controls on outflows have in the past not prevented
a massive capital flight from many developing countries. For many develop-
ing countries, capital controls have also not limited overall capital inflows.
Particularly in Latin America and Africa, capital controls on in- and outflows
have not been very effective. At most, capital controls on inflows have
changed the composition of inflows (there is some evidence on Chile which
shows that they are inefficient; see Valdes-Prieto and Soto, 1996; see also
Cardenas and Barrera, 1995 for Colombia). For these countries, capital
controls could be just distortionary, with limited net gains.
REFERENCES
AJesina, Alberto, Vittorio Grilli and Gian Maria Milesi-Ferretti (1993). The Political Economy of
Capital Controls. Working paper 4353, National Bureau of Economic Research, Cambridge, MA.
Cardenas, Mauricio and Felipe Barrera ( 1995). On the Effectiveness of Capital Controls for Colombia.
Mimeo, FEDESARROLLO and University of Chicago. Paper presented at the 8th
InterAmerican Seminar on Economics organized by the NBER and FE DESARROLLO
(Bogota, November 16-18, 1995).
Dooley, Michael P. (1996). A Survey of Academic Literature on Controls Over International Capital
Transactions. Working paper 5352, National Bureau of Economic Research, Cambridge, MA.
Valdes-Prieto, Salvador and Marcelo Soto (1996). New Selective Capital Controls in Chile: Are they
Effective? Mimeo, Universidad Cat6lica de Chile, Santiago, Chile.
World Bank (1996a). Global Economic Prospects, 1996. Washington, DC.
World Bank (1996b). World Debt Tables, 1996. Washington, DC.
RENE GARCIA
University of Montreal
This paper is important since it sheds some light on an unexplored issue: the
impact of cross-listing on an emerging market, which is potentially smaller and
less liquid than a market in a developed economy. Previous studies of interna-
tional listing had focused on countries such as Britain, Canada, and Japan.
The emerging market studied by the authors is Mexico, during the period from
the end of 1989 to the middle of 1993. This period is chosen as to avoid two
main turmoils that have rocked the Mexican economy. It starts just after an
hyperinflation period and ends just before the monetary crisis of 1994-1995.
However, 1989-1993 was a period of restructuring of the economy and bank
privatization. The rate of inflation fell from 20% per annum in 1989 to a yearly
rate of 10% in 1993, but the exchange rate remains relatively stable during
the period.
The data set used by the authors is new and made up of daily observations
on prices, returns and share volumes of 26 equity series, issued by 16 large
firms having experienced an ADR (American Deposit Receipts) listing. One
interesting feature is that we can observe the returns and volumes of various
classes of shares having different restrictions regarding foreign ownership. The
main difference is between securities from the A series, which can be held
legally only by Mexicans, and securities from the B series, which can be owned
by foreigners and institutional investors.
Methodologically, the paper differs from previous studies attempting to
discriminate between segmentation and integration of capital markets. For
example, in Jorion and Schwartz (1986) markets are segmented if the price of
risk differs between markets or if prices are determined by different risk factors.
Hence, the authors focus on restrictions imposed by a certain asset pricing
model on the pricing of assets. They carry therefore joint tests of the integration
hypothesis and of the chosen asset pricing model. Instead, the authors of this
paper examine the impact of Mexican ADRs on volatility, liquidity and bid-
ask spreads of domestic shares. In this context, they analyze three hypotheses:
the competitive hypothesis, in which the foreign listing provides for more
liquidity and greater depth (less volatility) in the domestic market; the fragmen-
tation hypothesis, in which the migration of investors away from the domestic
market provides less liquidity and more volatility and the bid-ask spread is
likely to increase; finally, the segmentation hypothesis where the effects are
mixed, e.g. the bid-ask spread might decrease. This concept of segmentation is
207
R. Levich (ed.). Emerging Market Capital Flows, 207-210.
ro 1998 Kluwer Academic Publishers.
208 R. Garcia
therefore different from the previous definition found in the international asset
pricing literature and appears more as a way to qualify mixed empirical results.
To test the aforementioned hypotheses, the authors rely on two models, one
estimating and testing changes in volatility and liquidity, the other computing
and testing changes in spreads. In the first model given by equations (2) and
(3), support will be found for the competitive hypothesis if PI < 0 and YI < 0,
while fragmentation will be consistent with PI > 0 and YI > O. If mixed effects
are obtained, the authors conclude that markets are segmented. For the second
model on changes in the bid-ask spread, the authors rely on a technique
developed by Roll (1984) to infer the size of the bid-ask spread based on the
covariance of successive price changes.
In analyzing the results, the authors compare the parameter estimates of the
A and B series for the two models taken separately. From one model to the
other, the criteria for assessing the effects seem to differ. For the first model,
the authors argue that liquidity falls (PI> 0) in seven of nine cases, but the
coefficient is significantly different from zero in only three cases at the 5%
level, while five coefficients are positive and significant for the A series. For the
change in spread model, the sign is negative for most of the securities of the A
series, but the authors argue that only two have a p-value of less than 10%.
Two additional comparisons involving the two models taken together appear
useful. Using the first, one can compare the results in terms of sign and
significance for companies that issue both A and B series (Table 1).
It is hard to conclude from Table 1 that the results are very different for the
A and B series. Another useful comparison consists in looking jointly at results
from models 1 and 2 for securities of the B series (Table 2).
A series B series
CEM +* +* * +* + *
CIF +* + +* +* *
MAS +* + - (12%) +* + *
PON +* + +* +
SYN + +* + +* *
* Significance of 5% for the coefficients and a p-value of 10% for llS as selected by the authors.
i', +* +* +* +* +* +* + +*
p, + + + + +* +* +*
llS -* +* + * + * *
Comment 209
It is only for TTO that one finds statistically significant evidence for the
fragmentation hypothesis based on YI and PI, and for segmentation when
adding the change in spread result.
Two individual cases for the A series are also worth mentioning. For both
GCA and TEL, which had respectively the highest increase and the highest
decrease in volume before and after the ADR introduction, the results show
that YI > 0, PI < O. For GCA, Po and PI are of opposite sign and of the same
magnitude and have equal standard errors. For both GCA and TEL, the
coefficient C>l is very significant and of the same sign as the change in volume.
These results suggest some identification problems caused by the correlation
between the absolute value of the price change and volume.
Regarding estimation issues, the authors never mention the instruments used
in the estimation nor the p-value of the overidentification l-statistic, making
it hard to evaluate the overall adequacy of the model. By using the absolute
value of price changes to measure the standard deviation of price changes, the
authors rely implicitly on a normality assumption and therefore lose one of
the advantages of using the distribution free GMM method. Therefore, the
authors could gain some efficiency by using maximum likelihood techniques
to estimate a model of the ARCH-type family. This could have the added
advantage of modeling the mean return with an asset pricing model, and make
the link with the previous literature on segmentation versus integration.
Within the GMM framework, the authors could add some useful tests to
their results. Since different series for the same company are estimated jointly,
the LR test of Eichembaum et al. (1988) could be used to test whether the
orthogonality conditions for the B series hold with possibly different parameters
than the A series. The authors could also use predictive tests of structural
stability proposed by Ghysels and Hall (1990) to test whether the orthogonality
conditions hold after the ADR introduction given the estimates of the parame-
ters obtained before the ADR listing. This test is potentially more powerful,
since it does not involve the estimation of the coefficients after the ADR listing.
On the more general issue of controlling for macroeconomic factors, since
non-ADR firms tend to be smaller firms, cross-sectional evidence showing that
fragmentation is more likely to occur in larger firms might be worrisome if one
considers that inflation might affect smaller and larger firms differently and
that the inflation rate changed during the period.
In conclusion, this seminal paper on the issue of cross-listing in the context
of an emerging market shows that ADR introduction tends to be associated
with higher volatility, lower liquidity and a reduction in the bid-ask spread.
What is less clear than suggested by the authors is the fact that these effects
are limited to the shares open to foreign ownership prior to the international
cross-listing.
REFERENCES
Eichenbaum, M.S., L.P. Hansen and K.J. Singleton (1988). A time series analysis of representative
agent models of consumption and leisure choice under uncertainty. Quarterly Journal of
Economics, 103, 51-78.
210 R. Garcia
Ghysels, E. and A.R. Hall (1990). A test for structural stability of Euler conditions parameters
estimated via the generalized method of moments. International Economic Review, 31,355-364.
lorion, P. and E. Schwartz (1986). Integration vs. segmentation in the Canadian stock market. The
Journal of Finance, 41, 603-616.
Roll, R. (1984). A simple implicit measure of the effective bid-ask spread. Journal of Finance, 39,
1127-1139.
VI HANG R. ERRUNZA
McGill University, Montreal, Canada
I tend to agree with a number of issues and ideas put forward by Smith and
Walter. Given their emphasis on the current state of the markets, it would be
useful to restate the two fundamental principles of emerging market (EM)
investing in an effort to put their arguments in a richer perspective. First, EM
investors must have a long-term perspective. The focus on the short term on
the part of media and some fund managers is not only inconsistent but has
damaged the whole concept of EM investing. The emphasis on long term is
and should be the basis of all investments, domestic or international, and is
critical for the mutual benefits to investors and recipients. Second, the primary
motivation for investment in EMs is risk reduction through diversification. In
the long run, investors may expect higher returns assuming that the EMs will
continue to grow at a higher rate and that the economic growth will be reflected
in equity returns. Thus, correlations with developed markets (DMs) and return
volatility playa critical role. We review each in turn.
It is often stated that EM correlations with DMs will increase due to
increasing economic integration. This is a myth. Despite economic integration
and internationalization of stock trading, diversification benefits have persisted
among DMs. In addition, by and large, the EM correlations have remained
low in the long run. This result has withstood the test of time regardless of
data sets, sample countries, test periods and methodology used to measure
benefits to international portfolio diversification (see Errunza, 1994, and refer-
ences therein). Over time, some markets have become more correlated, many
have remained stable and some have shown a decrease in correlations. Figure 1
plots Thai correlations with the US, which show an uptrend, Figure 2 plots
Argentinean correlations with the US, which suggests no clear trend and
Figure 3 plots Indian correlations with the US which show a downtrend.
Indeed, EMs have provided valuable hedging services during major global
declines. Figure 4 plots returns for a group of EMs and DMs during the
October 1987 market crash. Economic and industrial structures along with
other nation specific attributes guarantee this. In summary, the case for risk
211
R. Levich (ed.). Emerging Market Capital Flows. 211-215.
~ 1998 Kluwer Academic Publishers.
212 v.R. Errunza
-0.4
-0.6
-0.8L-~~-----~~~~----------~-----~--------------------~----------~
78 80 82 84 86 88 90 92 93
IFC TOTAL RETURNS INDICES
Figure 1. Correlations with US - Thailand.
-0.2
-0.4L-----------~-----~~=-~-------~~--~~~--------~
78 80 82 84 86 88 90 92 93
IFC TOTAL RETURNS INDICES
Figure 2. Correlations with US - Argentina.
0.4
0.2
-0.2
-0.4 ~ ..........
,"",""",",,'---'--_~--'--:~_---L _ _ _ _ _.!'----_ _ _ _.....J
78 80 82 84 86 88 90 92 93
IFC TOTAL RETURNS INDICES
20
10
o
-10
-20 :
-30
-40
-50~~~~~~~~~~~~~~~~~~~=-~~
~ a:l ~
IFC TOTAL RETURNS FOR EMs
MSCIP GD RETURNS FOR DMs
NOle: Returns are for Mexico, Malaysia, Brazil Philippines, Korea, Pakistan, India, Venezuela,
Zimbabwe, Japan, United States, United Kingdom, Australia, and the World market.
Figure 4. Percentage returns - October 1987.
214 v.R. Errunza
REFERENCES
Errunza, Vihang (1994). Emerging-markets: some new concepts. Journal of Portfolio Management,
20,82-87.
Errunza, Vihang and Etienne Losq (1985). The behavior of stock prices on LDC markets. Journal
of Banking and Finance, 9, 561-575.
Errunza, Vihang, Arthur Moreau and Jin-Chuan Duan (1993). The Pricing of American Depository
Receipts: Theory and Evidence. Working Paper #93-05-06, McGill University.
Eun, Cheol, Stijn Claessens and K wang Jun (1993). International trade of assets, pricing externali-
ties and the cost of capital. In Claessens, Stijn and Sudarshan Gooptu (eds), Portfolio Investment
in Developing Countries. Washington, DC: World bank Discussion Papers.
Goetzmann, William and Philippe Jorion (1996). Re-emerging Markets. Working Paper, University
of California, Irvine.
PART THREE
INTRODUCTION
The confluence of several trends has radically altered the nature of capital
flows to developing countries over the last decade. Securitization, privatization
and capital account liberalization have boosted the supply of assets available
to foreign investors. Substantial improvement in market institutions, ranging
from insider trading regulations to efficient settlement and registration mecha-
nisms, have reduced micro-risk, while more stable macroeconomic policies have
lowered country risk. On the external side, eased restrictions on emerging
market investment for institutional investors from developed markets! and the
attractive perceived risk/return tradeoff led to substantial investor interest in
emerging markets. In combination, these trends caused a significant shift in
the composition of capital flows from public to private and from FDI and
bank loans to portfolio, in particular equity, investment. 2
Despite these developments, investment advice for emerging markets contin-
ues to be frequently prefaced with the warning 'caution, high risk'. Part of this
perceived risk is homemade, reflecting the vagaries of the economic, and at
times political development processes unfolding in many of the recipient coun-
tries, and is compensated by the higher average expected returns yielded by as
yet unexploited profit opportunities in the newly open markets. However, any
reader of the recent business press might be tempted to conclude that a
significant part of the risk reflects not so much changing fundamentals in the
recipient countries but rather the fickleness of foreign investors. Judging from
the tone of the press, investment in emerging markets is largely undertaken by
novice investors and mutual fund managers with little understanding of funda-
mentals, 3 ready to rush 'lemming like'4 into and out of markets at the slightest
1 Notably the introduction of rule 144A in the United States on the equity side and the revised
guidelines for the Samurai bond market in Japan on the bond side.
2 While official flows increased slightly from US$ 41 bn in 1988 to US$ 54 bn in 1993, private
flows dramatically accelerated from 33 to $159 bn over the same period.
3 "Yet one can still find mutual-fund managers investing in the region whose knowledge of Latin
America scarcely goes beyond steaks and sombreros." Economist, Survey on Latin American
Finance, December 9th, 1995, page 3.
4 Economist, May 13th, 1995: 71-73.
219
R. Levich (ed.), Emerging Market Capital Flows, 219-235.
o 1998 Kluwer Academic Publishers.
220 H.C. Wolf
resolutions all but impossible, thus enhancing the scope for contagious
volatility.
While this simple model of an expectations driven market provides an
example of 'spurious volatility', it skirts the deeper issues, notably, why all
emerging markets should be viewed as a single entity, why investors should
think in terms of 'the market' rather than individual securities and, most
importantly, the cause for the change in expectations without an accompanying
change in fundamentals. Brennan (1990) provides a partial answer to the first
question: with costly learning, information gathering and processing will only
be undertaken if the expected benefits exceed the costs. As these benefits depend
upon other investors also acquiring, processing and, crucially, acting upon this
information to remove the pricing inefficiency, a mispricing may exist both
across and within markets.
Principal-agent problems in asset management (Scharfstein and Stein, 1990;
see also Grinblatt et al., 1995), roughly built around another observation
attributed to Keynes, that it is better for reputation to fail conventionally than
to succeed unconventionally, provide a second potential explanation for herding
behavior. If managers are evaluated not with respect to the efficient frontier,
but either relative to a market average or to an index of other managers in the
same asset class, low returns shared with most other managers carry little
penalty relative to an ex-post unsuccessful pursuit of an unorthodox strategy.
Given the reward structure, even above-average quality managers may opt to
hide in the herd rather than follow strategies which (ex-ante) may dominate
on risk-return basis. Applied to emerging markets, the principal-agent problem
might thus lead to correlated withdrawals even if (a subset of) mangers is
(ex-post correctly) confident about fundamentals.
Aside from these 'rational contagion effects' reflecting information issues,
contagion has also been attributed to institutional features, notably forced
redemption and two stage investment strategies. A significant fraction of the
inflows into emerging equity markets has been through open ended mutual
funds. 7 Large scale withdrawals from these funds in excess of cash reserves,
coupled with reductions in inflows and limits on permissible borrowing, may
force partial liquidation to honor redemptions. Table 1 provides an illustration
for a sudden end to capital inflows in the aftermath of the Mexican crisis. In
the case of multi-country funds, these liquidations are likely to be concentrated
in the most liquid markets. If these markets were not initially adversely affected,
the redemption pressures may thus generate a contagion effect. In this case, the
contagion reflects not so much a reduction of confidence in these markets, but
rather, perversely, their quality, reducing the costs of forced liquidations.8
7 As of mid 1995, some 519 open-ended mutual funds managed some US$ 39.7 bn, compared
with 237 closed end funds with US$ 37.4 bn under management.
8 The same outcome would of course be observed if global mutual aim to exploit perceived
mispricing via purchases in the most downtrodden markets financed through sales of equities in
less affected markets.
222 H.C. Wolf
Net equity
Emerging Share in purchase by
markets global mutual funds
9 Again, a self-enforcing dual equilibrium lurks in the background: to the degree that decisions
based on viewing emerging markets as a single asset class generate contagion across these markets
on a significant scale, emerging markets do, in fact, resemble a single asset class.
Determinants of emerging market correlations 223
reflect the presence of yet another excluded fundamental. Put differently, the
presence of contagion can only be established conditional on a given set of
explanatory variables and a given postulated relationship between these vari-
ables. Finally, while little direct work on emerging market contagion has been
done, the very substantial literature on establishing factors driving emerging
market returns is of course directly linked with the topic: the residual variance
left unexplained by this literature poses an upper bound on contagion effects.
Is contagion in fact as prevalent as the recent business press suggests? A
first stab at the data suggests otherwise. During the three most recent 3-year
periods, returns in the top twenty emerging markets differed dramatically, with
a total range between the highest and lowest return of above 300 % in every
period, and a range of 100 % even if the five top and the five bottom performers
are excluded. Nor were monthly returns particularly highly correlated. Table 2
reports, for individual emerging markets, statistics on the correlations of
monthly returns over the period 1988 to 1993 vis-a-vis a set of some twenty
other emerging markets. The results are in line with the general tone of the
literature (see Claessens et al., 1993, 1995a,b; Claessens, 1995; Errunza and
Losq, 1985): the average correlation equals 0.13, for nine markets the correla-
tion does not significantly differ from zero, and the maximum average correla-
tion of one equity market with all other equity markets amounts to just 0.25
Table 2. Correlations
for Malaysia. On first sight, evidence for contagion, at least over extended
periods, is thus slim, judging from the unconditional correlation of stock market
indices, the preferred (if implicit) measure of much of the recent writing on
contagion.
The motivation of this paper is to argue that these unconditional correlations
of market indices however provide quite poor measures of contagion, for two
reasons. First, index returns are partly determined by sectoral composition and
partly obscured by idiosyncratic noise. For example, consider two emerging
markets dominated by equities in a single sector, say petroleum. An increase
in the world demand for oil may lead to a substantial increase in the equity
prices of oil companies in both economies. Given the market weight of the
sector, a high correlation of the two overall indices results, yet this correlation
has little to do with "contagion" as the term is used. To the degree that returns
differ across sectors and sectoral composition differs across markets, it thus
becomes necessary to correct market indices for composition effects to obtain
meaningful measures of co-movements (Roll, 1992). Table 3 throws some light
on these two issues. The first column reports, for each market, a similarity
index defined as the sum of the absolute differences between the fraction of
total market capitalization in that market accounted for by each of 25 sectors
and the percentage of the market capitalization in all emerging markets
accounted for by the same sectors. A value of zero implies an equal distribution
Weight Weight
Overall Least equal in in Sectoral
similarity weighted country world stock returns Cumulative
index sector market market 1984-93 percent
10 Table 3 reports results at a particular moment. New issues and price changes can change the
relative ratings significantly. However, as long as these changes do not equalize the weighting
structure across markets, and there is no a priori reason to expect that they will, composition
effects will remain important.
11 See e.g. Cline (1995): "Rationally, spillover to the other big emerging markets from the
Mexican crisis should be limited, because few have the explosive combination of low reserves,
sizable short-term government debt held by foreigners, a large current-account deficit and a fixed
exchange rate:
226 H.C. Wolf
of the analysis, it varies between 400 and 1300 stocks. Since every observation
belongs to both one country and to one sector, it is not possible to uniquely
identify the country effect. The identification problem can be addressed in a
number of ways, we use benchmarking against a reference country (India) and
a reference sector (Food). The reference points were chosen since at least one
observation was available for both in every time period. The alternatives are
use of a generalized inverse (Zervos, 1994) and referencing relative to the
sample mean (Heston and Rouwenhorst, 1994). The appropriateness of the
identification method used depends upon the issue examined. Since our focus
is solely on the correlations, rather than the magnitudes, of the identified
country effects, the precise reference point chosen is of secondary importance
as long as it is common for all observations.
Country and, to a lesser extent, sector effects together explain between one-
third and one-half of the variance of returns on individual stocks, as graph 1,
plotting the 12-month rolling average of the R2 of the fixed effects regression
throughout the sample period time illustrates. The result is in line with earlier
studies (Divecha et al., 1992; Zervos, 1995) finding country effects to dominate
sector effects, with total explanatory power in the 33-50% range. Comparing
results for emerging and developed markets reveals the primacy of c~untry
228 H.e. Wolf
12See Beckers et al. (1992), Lessard (1974, 1976), Adler and Dumas (1983), Solnik and de
Freitas (1988), Grinold et al. (1989), Drummen and Zimmermann (1992), Roll (1992) and Heston
and Rouwenhorst (1994) for a small sample of work on developed markets. Heston and
Rouwenhorst (1994) is closest in approach to this paper.
13 See Adler and Simon (1986), Eun and Resnick (1988), Jorion (1991), Levy and Lim (1994),
Stockman and Dellas (1987) inter alia for discussions of the exchange rate in portfolio
diversification.
Table 5. Correlation matrix: total return and country effect
Arg Bra Chi Col Jor Kor Mal Mex Nig Pak Tai Tha Tur Yen Zim
Arg -0.22 -0.11 0.02 0.10 -0.08 -0.13 0.27 0.11 0.07 -0.05 0.08 0.15 -0.03 -0.06
Bra -0.19 0.14 0.08 -0.10 -0.06 0.00 -0.02 -0.11 -0.11 0.11 -0.10 0.16 -0.23 0.02
Chi -0.02 0.21 0.09 0.28 0.29 0.22 0.27 -0.03 0.21 0.31 0.22 0.01 -0.08 0.13 ...""t:l
Col -0.08 0.14 -0.10 0.33 0.20 0.24 0.21 0.38 0.63 0.23 0.25 0.27 0.25 0.17 "OJ
Jor -0.14 -0.08 om 0.07 0.53 0.52 0.50 0.32 0.52 0.41 0.49 0.03 0.23 0.40 3""
;;.
Kor -0.\3 -0.\3 -0.03 -0.06 0.04 0.55 0.49 0.38 0.45 0.45 0.37 0.07 0.27 0.50 ~
~
,~
Mal -0.11 0.12 0.11 0.19 0.19 0.30 0.54 0.37 0.42 0.52 0.60 0.33 0.08 0.54
Mex 0.35 0.08 0.07 -0.02 -0.05 0.27 0.42 0.28 0.42 0.51 0.41 0.08 0.16 0.46 '"
.Q.,
Nig 0.07 -0.11 -0.21 0.09 0.05 0.\3 0.06 -0.02 0.33 0.10 0.21 0.28 0.30 0.29
Pak 0.04 -0.03 0.04 0.43 0.10 0.09 0.11 0.04 -0.03 0.31 0.36 0.14 0.20 0.38 ""3
Tai -0.07 0.14 0.22 0.11 0.22 0.24 0.44 0.30 -0.15 0.11 0.38 0.20 0.07 0.36 ""
~
Tha 0.06 0.02 0.20 0.16 0.19 0.16 0.65 0.28 -0.02 0.22 0.28 0.23 0.03 0.36 ~.
Tur 0.16 0.15 -0.13 0.10 -0.07 -0.03 0.17 -0.13 0.20 0.04 0.18 0.18 -0.08 0.09
3
Yen 0.06 -0.19 -0.19 0.17 -0.12 -0.05 -0.18 -0.11 0.19 0.04 -0.20 -0.16 -0.09 0.38 ~
"OJ
Zim -0.12 0.04 -0.14 -0.05 -0.15 -0.05 0.27 0.08 0.06 -0.02 0.07 0.06 0.03 0.17 ;>;-
~
Above diagonal: country effects. Below diagonal: total returns. ~
"OJ
'"
"OJ
""~
g.
~
'"
tv
tv
\0
230 H.C. Wolf
Mean Mean
correlation correlation Mean
equity depression absolute
Country R2 Coeficient returns rates difference
more correlated across markets than are country effects, implying the presence
of significant idiosyncratic factors. Overall, exchange rate movements thus
account for at most a small portion of the country effects and their correlations,
a finding consistent with other work in this area.
Tables 7 and 8 examine the presence of causality effects between markets.
Table 7 applies Granger causality tests for the 1985-1995 period to examine
temporal linkages between the estimated country effects. The Philippines,
Pakistan and Malaysia are found to be temporally prior to three and two
Lag
2 4 6 12
other markets, out of 14 markets in the sample. 14 Brazil and Zimbabwe, being
Granger-caused respectively by six and four other markets, seem most sensitive
to developments elsewhere. Overall, there is little evidence for global leaders,
nor for the type of regional spillover effects from larger to smaller markets
found for capital flows (Calvo and Reinhart, 1994). Table 8 reports the contribu-
tion of own shocks to the variance of the estimated country effects at lags
between two months and one year, estimated by a VAR including the
13 markets for which complete data from 1985 to 1995 were available. A lower
contribution signals a greater dependence on other markets. Thailand, Malaysia
and Venezuela are seen to be least independent of shocks to other markets,
Jordan, Taiwan and the Philippines most independent. Again, however, no
clear outliers emerge, the proportion of the variance explained by own shocks
varies in the fairly narrow range from 26 to 42. Neither table thus suggests the
presence of dominant 'leading markets'.
"For those that think of Latin America in terms of generals, jungles and
sackfuls of worthless currency, it may be time to overhaul some myths.
Things have changed. (S)oldiers have long since goose-stepped back to
the barracks, their power usurped by squadrons of technocrats and battal-
ions of economic miracle makers."
FT August 27-28, 1994.
"Mexico's currency crisis has dimmed expectations for economies
throughout Latin America. The crisis and the border war .. , between
Peru and Ecuador have raised some fundamental questions in the minds
14The sample was restricted to countries with data beginning in 1990 or before.
232 H.C. Wolf
CONCLUSION
A century ago, Bagehot noted that "the same instruments which diffused capital
through a nation are gradually diffusing it among nations" and warned that
while "the effect of this will be in the end much to simplify the problems of
international trade ... for the present, as is commonly the case with incipient
causes whose effect is incomplete, it complicates all it touches.,,15 The assess-
ment remains valid today. Over the last decade, capital flows to emerging
markets have dramatically risen and, for the first time since the late 19th
century, been dominated by private-to-private flows. The inflows have generally
been cautiously greeted by governments aiming to enhance integration with
international financial markets. Yet the transition from closed to integrated
financial systems has not been without cost. In particular, policy-makers have
been concerned with the potentially disruptive consequences of capital flow
reversals. To the degree that such reversals reflect internal decisions, adoption
of stringent fiscal and monetary policies, possibly augmented with restrictions
on some types of capital flows, could be used to mitigate the likelihood and
extent of reversals. More recently, in particular in the wake of the Mexican
crisis, the possibility of reversals largely unrelated to any domestic fundamentals
under the influence of policy makers has attracted increased attention. To the
degree that such 'contagion effects' - capital flow reversals unrelated to funda-
mentals - are indeed present, financial integration, even though desirable long
term, may impose significant costs (Williamson, 1993; Gooptu, 1993).
The presence of contagion has been widely inferred from high correlations
of aggregate indices, mostly over very short durations. We argued above that
such unconditional correlations do not, in fact, provide good indicators of
contagion effects, for two reasons. First, aggregate returns were shown to be
subject to very significant composition effects potentially leading to spurious
correlations. Second, contagion, at least in the sense used here, requires
co-movements not attributable to changes in joint movements in fundamentals.
We asked whether the scope for contagion effects remains sizable once these
two problems have been corrected. Using a regression of individual equity
returns on country and sector dummies, we isolated a market effect purged of
sectoral composition effects and idiosyncratic noise. The market effects exhibit
higher correlations across emerging markets compared to the correlation of
market indices, a correlation which can not be attributed to exchange rate
movements. However, there do not appear to be strong causality patterns
among emerging markets, little evidence emerges in favor of either local or
global 'leading markets'. In sum, the potential scope for contagion effects is
enhanced by removing composition effects. However, as argued above, a con-
vincing argument that the observed correlation in fact reflects contagion must
also establish the absence of a matching correlation between fundamentals, a
task left for future.
ACKNOWLEDGMENT
I thank Maha Ibrahim for excellent research assistance and the IFC for kindly
making the Emerging Market Data Base available.
REFERENCES
Adler, Michael and Bernard Dumas (1983). International portfolio choice and corporation finance.
Journal of Finance, 38, 925-984.
Adler, Michael and David Simon (1986). Exchange rate surprises in international portfolios. Journal
of Portfolio Management, 12, 44-53.
Andrews, David and Shogo Ishii (1995). The Mexican Financial Crisis: A Test of the Resilience of
the Marketsfor Developing Country Securities. IMF Working Paper No. 95/132.
Beckers, Stan, Richard Grinold, Andrew Rudd and Dan Stefek (1992). The relative importance of
common factors across the European equity markets. Journal of Banking and Finance, 16, 75-95.
Bekaert, Gert and Campbell Harvey (1995a). Time-varying world market integration. Journal of
Finance, 50, 403-444.
Bekaert, Gert and Campbell Harvey (1995b). Emerging Equity Market Volatility. NBER Working
Paper No. 5307.
Black, Fischer and Robert Litterman (1991). Global Portfolio Optimization. Mimeo, Goldman Sachs.
Brennan, Michael (1990). Latent assets. Journal of Finance, 45, 709-730.
Buckberg, Elaine (1996). Institutional Investors and Asset Pricing in Emerging Markets. IMF
Working Paper No. 96/2.
Calvo, Guillermo (1995a). Varieties of Capital Markets Crisis. Mimeo.
Calvo, Guillermo (1995b). Managing Economic Reform Under Capital Flow Mobility. Mimeo.
Calvo, Sara and Carmen Reinhart (1995). Capital Flows to Latin America. Mimeo.
Cashin, Paul and John McDermott (1995). Informational Efficiency in Developing Equity Markets.
IMF Working Paper No. 95/58.
Cashin, Paul, Manmohan Kumar and John McDermott (1995). International Integration of Equity
Markets and Contagion Effects. IMF Working Paper No. 95/100.
Chuhan, Punam (1994). Are Institutional Investors an Important Source of Portfolio Investment in
Emerging Markets? WPS 1243, The World Bank.
Claessens, Stijn, Susmita Dasgupta and Jack Glen (1995a). Return behavior in emerging stock
markets. World Bank Economic Review, 9,131-151.
Claessens, Stijn, Susmita Dasgupta and Jack Glen (1995b). The Cross Section of Stock Returns:
Evidencefrom Emerging Markets. World Bank Discussion Paper No. 1505.
Claessens, Stijn (1995). The emergence of equity investment in developing countries: overview.
World Bank Economic Review, 9, 1-17.
234 H.e. Wolf
Cline, William (1995). Managing international debt. The Economist, February 18th, 17-19.
Devenow, Andrea and Ivo Welch (1995). Rational Herding in Financial Economics. IFA Working
Paper No. 218.
Drummen, Martin and Heinz Zimmermann (1992). The structure of European stock returns.
Financial Analyst Journal, July/August, 15-26.
Errunza, Vihang, Etienne Losq and Prasad Padmanabhan (1985). Tests of integration, mild segmen-
tation and segmentation hypotheses. Journal of Banking and Finance, 16,949-972.
Eun, Cheol and Bruce Resnick (1988). Exchange rate uncertainty, forward contracts, and inter-
national portfolio selection. Journal of Finance, 43, 197-215.
Frankel, Jeffrey (1993). The Internationalization of Equity Markets. NBER Working paper No. 4590.
Gooptu, Sudarshan (1993). Portfolio investment flows to emerging markets. In Stijn Claessens and
Sudarshan Goopto (eds.), Portfolio Investment in Developing Countries, World Bank Discussion
paper No. 228.
Grauer, R. and N. Hakansson (1987). Gains from international diversification. Journal of Finance,
42,721-741.
Grinblatt, Mark, Sheridan Titman and Russ Wermers (1995). Momentum investment strategies,
portfolio performance and herding: a study of mutual fund behavior. American Economic Review,
85,1088-1105.
Grinold, Richard, Andrew Rudd and Dan Stefek (1989). Global factors: fact or fiction. The Journal
of Portfolio Management, Fall, 79-88.
Heston, Steven and Geert Rouwenhorst (1994). Does industrial structure explain the benefits of
international diversification? Journal of Financial Economics, 36, 3-27.
Howell, Michael (1993). Institutional investors and emerging markets. In Stijn Claessens and
Sudarshan Gooptu (eds.), Portfolio Investment in Developing Countries, World Bank Discussion
Paper No. 228.
International Finance Corporation. Emerging Stock Market Factbook, various years, Washington,
DC.
Jorion, Phillipe (1991). The pricing of exchange rate risk in the stock market. Journal of Financial
and Quantitative Analysis, 26,363-376.
Lessard, Donald (1974). World, national and industry factors in equity returns. Journal of Finance,
24,379-391.
Lessard, Donald (1976). World, country and industry relationships in equity returns. Financial
Analysts Journal, 32, 2-8.
Levy, Haim and MarshaIl Sarnat (1970). International diversification of investment portfolios.
American Economic Review, SO, 668-675.
Levy, Haim and Kok Chew Lim (1994). Forward exchange bias, hedging and the gains from
international diversification of investment portfolios. Journal of International Money and Finance,
13, 159-170.
MacKay, Charles (1841). Memoirs of Extraordinary Popular Delusions. London: Richard Bentley,
Reprint.
Ripley, D. (1973). Systematic elements in the linkage of national stock market indices. Review of
Economics and Statistics, 55,356-361.
RoIl, Richard (1992). Industrial structure and the comparative behavior of international stock
indices. The Journal of Finance, 47, 3-41.
Scharfstein, David and Jeremy Stein (1990). Herd behavior and investment. American Economic
ReView, SO, 465-479.
Solnik, Bruno (1974). Why not diversify internationally rather than domestically. Financial Analyst
Journal, 30, 91-135.
Solnik, Bruno and A. De Freitas (1988). International factors of stock price behavior. In S. Koury
and A. Gosh (eds.), Recent Developments in International Banking and Finance, Lexington:
Lexington Books.
Stockman, Alan and Harris Dellas (1987). International portfolio nondiversification and exchange
rate variability. Journal of International Economics, 26, 271-289.
Determinants of emerging market correlations 235
Williamson, John (1993). Issues posed by portfolio investment in developing countries. In Claessens
and Gooptu, 11-17.
Zervos, Sara (1994). Industry and Country Components in International Stock Returns. Mimeo,
Rochester University.
ROBERT E. CUMByl and ANYA KHANTHAVIT2
1 Georgetown University: 2 Thammasat University
ABSTRACT
In this paper we estimate a bivariate two-state Markov switching model of excess returns
on both domestic equities and a world index of equities for Thailand, Taiwan, and Korea.
Our reason for doing so is to determine if changes in the behavior of equity returns can be
linked to changes in policies governing the integration of these economies and their capital
markets with world markets. We find clear evidence of two regimes: one characterized by
a low variance of domestic equity returns and a low {3 of domestic equities relative to the
world index, the other by a high variance of equity return and a high {3, in all three countries.
The differences across states in the covariance of the local market returns with the world
index is consistent with greater integration of goods and capital markets in the high-
covariance, high-{3 state. We find, however that the even in the 'integrated' state, equity
returns are not consistent with a simple, single-{3 model of an integrated world capital
market. For all three countries our estimates suggest that stock returns are higher than
would be predicted by a simple CAPM. Only for Thailand is the temporal behavior of the
probability that the high-covariance state is generating the data consistent with a change
in government policies leading to greater goods and capital market integration. The esti-
mated probabilities indicate quite clearly that a regime change occurred in the mid-1980s.
The early 1980s are characterized by the low covariance state, 'segmented' state and
thereafter, the data are generated by the high covariance, 'integrated' state.
INTRODUCTION
1500 -------------------------------------- -
8
,...;
~ 1000
~
,...;
500 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - --
o~~==~~====~--------~
1977 1979 1981 1983 1985 1987 1989 1991 1993
600
~~~~.~: ::::::::~~~::
200 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - --
O~------------------------------~
1977 1979 1981 1983 1985 1987 1989 1991 1993
1200 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
1000-- -----------------
800 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
II
~
400 -- - - - - - - - - - - - - - - - - - - --
200 -- - - - - - --
o1985
--------------------------------------
1986 1987 1988 1989 1990 1991 1992 1993 1994
Figure 1. (Continued.)
moved in tandem. The early 1990s slump was both shorter lived and milder in
Thailand, where it has been followed by another dramatic rise in equity values.
Figure 1 suggests that equity returns might be characterized by more than
one regime. We therefore model the distribution of equity returns as subject to
(stochastic) changes in regime. We then ask whether the differences in the
distribution of returns across regimes is consistent with the effects of greater
integration of both financial and goods markets with those abroad. Because
investors can sometimes circumvent legal or regulatory restrictions, effective
capital market integration need not coincide with legal changes. Our approach
is therefore to let the data determine when regime changes occur rather than
using legislative or regulatory changes to impose a particular date (or set of
dates) for the regime changes.
The plan of the paper is as follows. In section I, we describe a two-state
Markov switching representation of equity returns and present estimates for
equity returns in the three markets. In section II, we expand the univariate
models of section I and examine bivariate models of equity returns in each
market and the return on a world index of equities. We present estimates of
constrained and unconstrained versions of the bivariate models that allow us
to examine the effects of goods market and capital market integration and to
test the restrictions of a simple version of a capital asset pricing model. In
section III, we compare the results described in section II with estimates of the
same bivariate Markov switching representation applied to countries with long-
240 R.E. Cumby and A. Khanthavit
standing links with world goods and equity markets. Section V offers some
concluding remarks.
1 The random variable z, can follow an autoregressive process but the return data we analyze
do not appear to have an autoregressive component.
A Markov switching model of market integration 241
Monthly data from January 1977 to December 1994 (216 observations) are used for Thailand and
Korea and January 1985 to December 1994 (120 observations) for Taiwan. The Q(12) statistic is
the Ljung-Box test for autocorrelation and is distributed as X2 ( 12). Under the null hypothesis that
returns are independently and identically distributed, the autocorrelations have an asymptotic
standard error of T-I/2, which is 0.068 for Thailand and Korea and 0.091 for Taiwan.
The results for Thailand and Korea are based on monthly data from January 1977 to December
1994 (216 observations) and the results for Taiwan are from January 1985 to December 1994 (120
observations). For each market, the excess returns on the national market index (r,) is computed
as continuously compounded returns in local currency in excess of local short-term interest rate.
A non-linear filter is used to estimate the equation, r, = I/(s,) + Z" where I/(s,) is the mean return
of the national index in state s,. z, is distributed normally with a mean zero and a variance 0-2(s,).
s, follows a first-order two-state Markov process with transition probabilities I/;,j = Prob(S, =ilS'-1 =
i). Standard errors are in parentheses.
markets have volatilities above 8% per month. In order to verify that infrequent
trading is not a severe problem, we also report autocorrelations of excess
returns. If infrequent trading is a problem, the month-end prices used to
compute the index will be averages of prices observed at different points in
time. This may induce serial correlation in the computed returns. 3 No evidence
of serial correlation is found in any of the excess returns.
Table 2 presents maximum likelihood estimates of the parameters of the
two-state Markov switching representation of equity returns from the three
markets that we are examining. Two states are clearly distinguished in all three
equity return series. In all three cases, the variance of excess returns is signifi-
cantly higher in state 1 than in state O. Only for Thai equity returns do we
find a statistically significant difference in mean returns, although the point
estimates of the mean returns in the two states are quite different for Taiwanese
equities. Interestingly, the diagonal elements of the transition probability matrix
are around 0.9 or higher in all three sets of estimates. Since the probability of
switching states is low, the expected duration of a regime is quite long. The
3Bailey et al. (1990) found evidence of autocorrelation in daily returns computed from local
market indexes for all three of the countries we consider, The absence of any autocorrelation in
our data may be due to the monthly sampling interval or due to differences in the composition of
the indexes,
A Markov switching model of market integration 243
large and highly significant estimates of nu and noo also suggest that equity
returns are highly dependent on the realization of the state variable in the
previous period. The null hypothesis that the data can be adequately described
by a random walk for equity values (noo = 1 - nu) can be rejected in all three
markets at any reasonable significance leve1. 4
Although a two-state description of the data is superior to a simple random
walk representation, the evidence does little to shed light on the causes of the
change in regime. Our interest in these data and in the procedures we are using
derives from hypotheses about the implications of changes in policies governing
market integration for the behavior of equity values. There is nothing in the
estimates that confirms or contradicts any of these hypotheses. As a result we
expand our model in the next section and consider the link between equity
returns in the three countries and the return on a world index of equities.
The integration of domestic goods and capital markets with those abroad
imposes restrictions on the bivariate stochastic behavior of domestic and world
equity returns. In this section we examine a bivariate two-state Markov switch-
ing model of equity returns and use the estimates to draw inferences about the
extent to which integration of goods and capital markets is behind the behavior
of equity returns that we describe above. In doing so we allow the data to
dictate both when regime changes are observed and how the behavior of equity
returns differs across regimes.
Let r w.t be the return on a world index of equities expressed in local currency
terms and in excess of a short-term local currency interest rate. We will assume
that the distribution of r w.t is independent of the realization of the state variable,
St, so that,
where the diagonal elements of L(St) are u 2 (St) and O"~ as above. The off-
diagonal element, Uiw(St = 0) = O"iw(O) and Uiw(St = 1) = uiw(I).
If the policies concerning increased market integration are the cause of the
change in the stochastic process governing domestic equity returns, there are
4 The series of short-term interest rates that are available from the three countries we are
examining may diverge from short-term market clearing rates to an extent that varies across
countries and over time. In order to determine if these problems with the short-term interest rate
series were affecting the results in an important way, we have also estimated two-state Markov
switching models for real rates of return on the equity indexes. The results are very similar to
those obtained using excess rates of return.
244 R.E. Cumby and A. Khanthavit
two sets of restrictions placed on the joint process set out above. First, if
domestic equity markets are segmented in state 0 but integrated with world
equity markets in state 1, a simple version of the capital asset pricing model
implies,
5 As is well known, a sufficient condition for the simple CAPM to be a valid description of
equilibrium real rates of return on equities is that relative purchasing power parity holds. A useful
survey of the main results in international asset pricing is found in Stulz (1992).
A Markov switching model of market integration 245
The results for Thailand and Korea are based on monthly data from January 1977 to December
1990 (168 observations) and the results for Taiwan are from January 1985 to December 1990 (72
observations). For each system, the excess returns on the national market index (r,) and the world
index (r w,,) are computed as continuously compounded returns in local currency in excess of local
short-term interest rate. A non-linear filter is used to estimate the following system, r, = II(S,) + z"
r w,' = IIw + zw,l' where IIw is the mean return of the world index and II(S,) is the mean return of the
national index in state s,' Z, == {z" zw,,} is distributed normally with a zero-mean vector and a
variance covariance matrix L(S,), Vec(L(S,)) = {u 2(s,), Uiw(S,), u~)}, s, follows a first-order two-
state Markov process with transition probabilities 1li.j = Prob(S, = jlS'_1 = i), Standard errors are
in parentheses, In the constrained system, II(S, = 1) = IIwO"iw(S, = 1)/O"~,
246 R.E. Cumby and A. Khanthavit
6 Turner et al. (1989) find negative expected excess returns in one state of their two-state Markov
switching model of US equity returns.
7 Legal restrictions on the proportion of foreign ownership of shares issued by Thai companies
vary across industries and across firms in an industry. In addition, some firms impose restrictions
more stringent than the legal restrictions. In September 1987, the Stock Exchange of Thailand
created an 'Alien Board' for trading by foreigners in the shares of Thai firms with binding foreign
ownership restrictions. Bailey and Jagtiani (1994) study the variation of the premium on Alien
Board shares over time. The Far East Economic Review (2/26/87) reports that the lack of meaningful
English-language research impedes foreign investment in the Thai stock market even when foreign
ownership restrictions are not binding.
S In August 1985, the first of what are now seven closed-end Thai equity funds began trading
in London. This was followed by Thailand fund, which also trades in London, in December 1986.
These two funds are currently the smallest of the closed-end Thai funds and were followed in 1988
by five more funds, one of which trades on the NYSE. At the end of 1990, the net asset value of
A Markov switching model of market integration 247
Both this sizable premium over identical shares available only to domestic
investors and the difference between the constrained and unconstrained esti-
mates of J1( 1) point to substantial barriers to international investment.
The evidence that investment barriers are important in state 1 is also
consistent with the results of Khanthavit and Sungkaew (1993) who estimate
a latent variables version of a conditional capital asset pricing model with
investment restrictions. They find statistically significant evidence of investment
barriers using Thai stock returns from 1986 through 1989 and their estimates
of the magnitude of implied tax rate are even larger than those obtained here. 9
If Thai equity markets are (at least partially) segmented in both states, the
higher expected equity return in state 1 should be reflected in a larger variance
of returns in state 1 (or in a higher market price of risk).l0 The estimates do
point to highly volatile returns in state 1, with the estimated variance exceeding
the variance of returns in state 0 by a factor of nearly 8 and exceeds the
variance of returns on the world index by a factor of 6. Thus, the very high
estimated mean return in state 1 is associated with a similarly large estimated
variance of returns.
The third and fourth columns of Table 3 present the estimates for the
unconstrained and the constrained bivariate two-state Markov models for
Korean and world equity returns. Again, we will focus on the unconstrained
estimates. As was true with the estimates for Thai equity returns, two states
are clearly distinguished in the data, the probabilities of changing states are
both less than 10%, and again, the null hypothesis that equity prices follow a
random walk is rejected at any reasonable significance level. Unlike the Thai
estimates, the mean return in state 0 exceeds the mean return in state 1,
although both are imprecisely estimated and neither is significantly different
from zero.
Moving from state 0 to state 1 there is a striking increase in the covariance
of Korean and world equity returns just as is the case with the Thai data. In
state 0 the estimated covariance is small and insignificantly different from zero
while in state 1 the estimated covariance is large and highly significant. Once
again it is useful to look at the f3 of Korean equities and the correlation of
Korean and world equity returns as a means of assessing the change in the
these funds amounted to $600 million or 2.5% of the Thai market capitalization. Bosner-Neal
et al. (1990) report that during 1988, a year that is classified by the filter to be, with high probability,
in state 1, the closed-end Thai fund traded in the USA at an average premium over net asset value
of 25.46%. While this premium is consistent with the existence of barriers to foreign investment,
the evidence in Bailey and lagtiani (1994) suggests that there are substantial differences in the
behavior of the premium on the closed end Thai funds and the premium on Alien Board shares.
9 As we discuss below, this period is one that is generally classified by our estimates as
characterized by state 1.
10 An increase in growth rates of output, which might lead to an increase in expected earnings
growth, cannot alone explain a higher average return on equity. Without an increase in the rate
at which future earning are discounted, an increase in the growth rate of future earnings should
lead to a discrete jump in equity values.
248 R.E. Cumby and A. Khanthavit
11 While two closed-end Korea funds were created in November 1981, both are small funds and
are not listed on any exchange. The largest of the seven closed-end Korea funds that are now
traded is the Korea Fund, which was created in August 1984 and trades on the NYSE. The second
largest, the Korea Europe Fund was created in March 1987 and trades on the London exchange.
At the end of 1990, the net asset value of the seven funds was $413.6 million or 0.37% of Korean
market capitalization. The evidence presented in Bailey and Lim (1992) and in Bailey and Jagtiani
(1994) suggests that the premium on the closed-end funds might be an unreliable measure of the
degree of investment restrictions, however.
A Markov switching model of market integration 249
or less and the null hypothesis that Taiwanese equity prices follow a random
walk is rejected at any reasonable significance level. The estimates from the
Taiwanese system have a great deal in common with those from the Thai
system. The estimated mean excess return in state 1 exceeds the estimated mean
excess return in state O. Unfortunately, possibly due to the small sample, neither
mean is estimated sufficiently precisely to allow us to distinguish it statistically
from zero at standard confidence levels.
As is the case in the systems estimated using excess returns on Thai and
Korean equities, the estimated covariance of Taiwanese equity returns with
returns on the world index is sharply higher in state 1. Once again, the
covariance is small and insignificant in state 0 and large and significant in
state 1. The estimated P for Taiwanese equities rises from 0.11 in state 0 to
1.20 in state 1 and the correlation between excess returns on Taiwanese equities
and the excess returns on the world index rises from 0.08 in state 0 to 0.27 in
state 1. The similarities across the three markets in the behavior of the estimated
covariance is striking. In all three markets, there is strong evidence of consider-
ably greater covariance between local and foreign equity returns in state 1 than
in state O. This evidence is consistent with greater integration of goods and
possibly capital markets in state 1 in all three countries.
The restrictions on the mean excess return in state 1 that are implied by a
CAPM do not lead to a sufficiently large increase in the log-likelihood values
to lead to rejection of those restrictions. A number of aspects of the estimates
suggest that this failure to reject is due to the relatively small sample and the
consequent imprecision with which the parameters, especially the means, are
estimated. The value of p( 1) that is implied by the unconstrained estimates of
O"iw(1), and, O"~, along with the estimate of Pw is 0.46% per month. While this
value is well within a 95% confidence interval around the unconstrained point
estimate of p(1), it falls short of that point estimate by 3.38% per month.
Although this tax rate equivalent of the barriers to investment is not statistically
significantly different from zero, it is so large that the failure to reject the
CAPM restrictions should not be interpreted as evidence that they, in fact, hold.
Because the evidence does not support the hypothesis that Taiwanese capital
markets are integrated in state 1, the change in the mean return between
states 0 and 1 should reflect a higher variance of excess returns in state 1. The
estimated variance is substantially larger in state 1. In fact, given the size of
the increase in the estimated variance, the increase in the estimated means is
remarkably small. The much greater proportional increase in the variance of
returns might suggest some partial integration of capital markets in state 1,
although the imprecision with which the means are estimated makes it difficult
to draw any firm conclusions.
Figure 2 presents the estimates of the probabilities that Sf = 1 based on data
through period t, P(St I rp r~ , rt - 1, r~-1' ... ). Three episodes are clearly visible in
the filter probabilities for Thai equity returns. The probability that the begin-
ning of the sample is drawn from state 1 is high. This period corresponds to a
late 1970s stock market boom and subsequent crash that is also reflected in
250 R.E. Cumby and A. Khanthavit
0.8
~0.6
~
J::J
0.4
0.2
O-'------------------.----.J
1977 1979 1981 1983 1985 1987 1989 1991 1993
0.8
~0.6
~
0.4
0.2
Figure 1. This period preceded the liberalization measures that have motivated
our examination of the behavior of stock prices, and is commonly attributed
to market manipulation in a what was at the time a rather small and thin
market. During the first part of the 1980s the probabilities that the data are
drawn from state 1 are generally low (apart from a spike in September 1982),
and then rise sharply in mid-1986. In the mid-1980s, the Thai authorities
removed several barriers to foreign investment and pursued policies of deregula-
tion. At the same time they adopted export promotion policies that included
exchange rate changes as well as tax and other incentives for export manufactur-
ing. 12 Exports both grew rapidly and expanded beyond traditional commodity
0.8
&,0.6
~
.D
0.4
0.2
Figure 2. (Continued.)
exports. In the second half of the 1980s, Thailand was the world's fastest
growing economy. Rapid growth was accompanied by a boom in stock prices.
This period of high growth in output and exports as well as high average stock
returns is classified by the data as being drawn from state 1 with high prob-
ability. The filter probabilities and the parameter estimates from the two-state
model provide a consistent and convincing case that changes in behavior of
equity values are consistent with greater integration and these changes coincide
with changes in the economic policies of the authorities. The measures that
began the process of integrating Thai equity markets with those abroad do
not, however, appear to have led to full integration.
These results are also consistent with those of Bekaert and Harvey (1995)
who estimate a model of partially segmented markets (expected returns depend
on both the covariance with a world index and on the variance of returns)
with changes in regimes. They find that probability that the data are consistent
with integration is low in the early 1980s and rises in the mid-1980s.
The estimated filter probabilities for Korean equities exhibit somewhat more
puzzling behavior. The probability that the high variance (and low mean) state
is governing returns is low until the second quarter of 1979 when it rises
sharply. The probability remains high until the fourth quarter of 1982, when
it rises before falling again in the summer of 1988. A final rise and fall in the
probability occurs in the second quarter of 1990 and the first half of 1993,
respectively. It is difficult to fit these estimated probabilities with the conse-
quences of market integration policies. At least three explanations are plausible.
First, political developments have been important in Korea during the sample
and may account for some of the behavior of equity returns. Second, the
parameter estimates seem to suggest that partial capital market integration
played a role in Korean equity values in the sample. Partial capital market
252 RE. Cumby and A. Khanthavit
Are the changes in the behavior of excess equity returns described above due
to changes in policies concerning market integration or are they some feature
of the data that we mistakenly attribute to economic liberalization? To help
13 The Korea trust began in 1981. The number of Korea funds has grown and by the end of
the sample more than 30 Korea funds were trading.
14 Far Eastern Economic Review (1/1/87, 3/26/87, and 5/21/87) and Moreno and Yin (1992).
15 From 1983 until 1991 foreign investors were restricted to four approved funds.
16 Bekaert and Harvey (1995) estimate a fairly stable probability of integration that is close to
one throughout the sample both for Taiwan and for Korea.
A Markov switching model of market integration 253
The results are based on monthly data from January 1977 to December 1990 (168 observations).
For each system, the excess returns on the national market index (r,) and the world index (r",.,)
are computed as continuously compounded returns in local currency in excess of local short-term
interest rate. A non-linear filter is used to estimate the following system, r, = JI.(s,) + Z, r ".' = JI. .. +
z""" where JI. .. is the mean return of the world index and JI.(s,) is the mean return of the national
index in state Sr' Z, == {Z" z.,.,) is distributed normally with a zero-mean vector and a variance
covariance matrix 1:(S,). Vec(1:(S, = {(f2(S,), (f;.,(s,), (f~)}. s, follows a first-order two-state markov
process with transition probabilities T;.} = ProbeS, = jl S'-1 = i). Standard errors are in parentheses.
A Markov switching model of market integration 255
market returns with the world index is consistent with greater integration of
goods and capital markets in the high-covariance, high-P state. We find, how-
ever that the even in the 'integrated' state, equity returns are not consistent
with a simple, single-p model of an integrated world capital market. For all
three countries our estimates suggest that stock returns are higher than would
be predicted by a simple CAPM.
Only for Thailand is the temporal behavior of the probability that the high-
covariance state is generating the data consistent with a change in government
policies leading to greater goods and capital market integration. The estimated
probabilities indicate quite clearly that a regime change occurred in the
mid-1980s. The early 1980s are characterized by the low-covariance state,
'segmented' state and thereafter, the data are generated by the high-covariance,
'integrated' state. The estimated probabilities for Korean and Taiwanese equi-
ties fail to show a sustained change from the low-covariance state to the high
covariance state. But, unlike Thailand, both of these countries adopted export-
oriented trade polices before our sample begins. And both liberalized capital
markets later than Thailand. The temporal behavior of the estimated probabili-
ties for these two countries suggests that differences in regimes are directly
linked to policy changes leading to liberalization. The results seem to be more
consistent with partial integration of capital markets, perhaps with investment
barriers more important in some periods than in others.
ACKNOWLEDGMENTS
Warren Bailey, Mario Crucini, Rene Stulz, John Ammer, seminar participants
at Ohio State University, and participants in this conference on the future of
emerging market capital flows provided helpful comments on earlier versions
of this paper. Those comments along with financial support from the Salomon
Brothers Center for the Study of Financial Institutions and research assistance
from Simi Kedia are acknowledged with thanks.
REFERENCES
Bailey, Warren, Rene M. Stulz and Simon Yen (1990). Properties of daily stock returns from the
pacific basin stock markets: evidence and implications. In S.G. Rhee and R.P. Chang (eds), Pacific
Basin Capital Markets Research. Amsterdam: North-Holland, 155-171.
Bailey, Warren and Julapa Jagtiani (1992). Time-varying premiums for international investment:
some empirical evidence. Journal of Financial Economics, 36, 57-87.
Bailey, Warren and Joseph Lim (1992). Evaluating the diversification benefits of the new country
funds. Journal of Portfolio Management, Spring, 74-80.
Bekaert, Geert and Campbell R. Harvey (1995), Time-varying world market integration. Journal of
Finance, 50, 403-444.
Bosner-Neal, Catherine, Greggory Brauer, Robert Neal and Simon Wheatley (1990). International
investment restrictions and closed-end country funds. Journal of Finance, 45, 523-547.
Bowring, Phillip (1987). While others falter. Far Eastern Economic Review, June 25, 68-70.
256 R.E. Cumby and A. Khanthavit
Bowring, Phillip and Carl Goldstein (1987). Profitable pragmatic politics. Far Eastern Economic
Review, May 21, 97.
Carter, Liz (1987). Bangkok beckons despite the hurdles. Far Eastern Economic Review, February
26,68-69.
Engel, Charles and James D. Hamilton (1990). Long swings in the dollar: are they in the data and
do markets know it? American Economic Review, SO, 689-713.
Goldstein, Carl (1987). Their own worst enemy. Far Eastern Economic Review, January 1, 59.
Goldstein, Carl. (1987). A cash-stuffed case for investment. Far Eastern Economic Review, March
26,74-75.
Goldstein, Carl (1987). A promise offreedom. Far Eastern Economic Review, May 21, 66.
Hamilton, James D. (1988). Rational expectations econometric analysis of changes in regime: an
investigation of the term structure of interest rates. Journal of Economic Dynamics and Control,
12, 385-423.
Hamilton, James D. (1990). A new approach to the economic analysis of non stationary time series
and the business cycle. Econometrica, 57, 357-384.
Khanthavit, Anya and Jirat Sungkaew (1993). Measuring Thailand's barriers to investment. Pacific
Basin Finance Journal, 1, 355-367.
Moreno, Ramon and Norman Yin (1992). Exchange rate policy and shocks to asset markets: the
case of Taiwan in the 1980s, Federal Reserve Bank of San Francisco. Economic Review, 1, 14-30.
Stulz, Rene M. (1981). On the effects of barriers to international investment. Journal of Finance,
36, 923-934.
Stull, Rene M. (1992). International Portfolio Choice and Asset Pricing: An Integrative Survey.
Working paper, Ohio State University.
Turner, Christopher M., Richard Startz and Charles R. Nelson (1989). A Markov model ofhetero-
skedasticity, risk, and learning in the stock market. Journal of Financial Economics, 25, 3-22.
World Bank/International Finance Corporation, Emerging Stock Markets Factbook, various issues
DATA ApPENDIX
ABSTRACT
This chapter examines the response of stock prices in emerging capital markets to the
announcement of events highlighting the opening of capital markets. The announcements
analyzed in this research are the private corporate Eurobond offerings and launching of
emerging-market country mutual funds. Corporate Eurobond issues in emerging markets
are associated with positive abnormal returns to the underlying stock around the successful
offering. In addition, the launching of an emerging-market closed-end country fund is
preceded by sharp increases in the underlying stock market index as well as an equally
weighted portfolio of the top firm holdings of the country fund. There are sharp increases
in the month of the fund launching but stock prices decline in the months following the
offering. There is a marked reduction in volatility of stock returns following the introduction
of the country fund, implying that better and more critical monitoring by fund managers
and analysts results in the emerging markets becoming less speculative and more disciplined.
INTRODUCTION
of companies are realizing that overseas offerings can make up for the shortfall,
as well as enhance their exposure in the international markets, despite some
initial bumps for some. The developed world's appetite for global diversification
has made it easier for many to raise offshore funds, while financial deregulation
and economic liberalization aimed at creating strong domestic capital markets
has expanded the list of potential overseas issuers from emerging markets. The
dismantling of domestic barriers to foreign investment, booming stock markets
in Brazil, India, Turkey, Mexico and others, and an impending wave of priva-
tizations are creating worldwide investor demand for the increasing amount of
new paper from these countries. The 'torrid pace' of Eurobond offerings is
expected to continue well into the nineties, with nearly 40 offerings in 1994
from India alone, totalling over $5 billion.
Eurobond offerings started to take off in the emerging markets only in 1990,
while the international equity-related issuances (in the form of global and
ADRs) are also becoming increasingly popular. Such Eurobond offerings had
been popular among the developing economies in the mid-1970s, but private
issuers in emerging markets were locked out in the 1980s, regaining access to
this source of funding in the 1990s. Eurobond offerings in the USA have been
researched by Kim and Stulz (1988,1992) and Kidwell et al. (1985). However,
no study has analyzed such issues by corporations in emerging capital markets.
This study attempts to fill that gap by examining the announcement effect of
such offerings on shareholder wealth.
We also examine the effect on shareholder wealth of gaining access to an
emerging market through a newly created investment instrument like a closed-
end country mutual fund. Sponsored country mutual funds from emerging
markets have become increasingly popular in the last 5 years. Recent research
by Bailey and Lim ( 1992), Diwan, Errunza and Sen bet ( 1993) and Hardouuvelis
et al. (1993) have analyzed the post-offering performance of closed-end country
funds from emerging and developed markets by studying their market prices,
net asset values, the premium/discount patterns and benefits of diversification
to US investors. However, none of them analyze the impact of the fund offering
on the underlying local market index.
Recent research analyzes a number of policy issues pertaining to the macro-
economic effects of large capital inflows and their effect on stock market
behavior (Claessens and Gooptu, 1993). Policy makers are also grappling with
the issue of the effect of portfolio flows on the cost of capital of the issuing
firm, though this research does not address this issue. If these issues are wealth-
generating for shareholders, we may expect more of them.
Eurobonds
are called domestic bonds, while bonds issued in the international markets are
called foreign bonds or Eurobonds. Eurobonds are underwritten by an inter-
national syndicate of commercial and/or investment banks and sold principally
in countries other than the country in whose currency the bond is denominated.
Eurobonds are unregulated and economically more efficient. Being bearer
securities they provide the holders with anonymity, and their tax considerations
and no withholding tax features make them attractive to several foreign invest-
ors. In addition, bond covenants in Eurobonds are less restrictive than on
domestic issues. With short maturities averaging 3-10 years, they are increas-
ingly popular among small investors as well as with institutions, who prefer
them because of liquidity needs and foreign exchange risk. They are very liquid
in the secondary markets, trading efficiently on Euroclear and Cedel. Given
their unregulated nature, Eurobonds are usually issued only by reputed and
highly rated firms. These special features make them attractive to issuers,
especially since they may reduce borrowing costs by 25-100 basis points relative
to domestic dollar bonds or dollar term loans.
Kim and Stulz (1988) find that US shareholders benefit from Eurobond
issues. They find a small positive average abnormal return of 0.46%
(z-statistic of 3.38) associated with the offering announcement. 1 This result
differs from the literature on stock-price effect of domestic bond issues, which
reports a negative or zero wealth effect. Eckbo (1986), using a sample of 459
straight-debt issues from 1964 to 1981, finds a negative stock-price announce-
ment effect of -0.11 %, with a z-statistic of -0.96 (statistically insignificant).
Mikkelson and Partch (1986), using a sample of 171 straight bond issues from
1972 to 1982, find a negative announcement effect of -0.23%, with a
z-statistic of -lAO (insignificant). However, James (1987) finds the average
announcement effect for a sample of 80 bank loan agreements to be 1.93%,
with a z-statistic of 3.96 (statistically significant). For convertible domestic
bonds, Dann and Mikkelson (1984) find an average (significant) abnormal
return of - 2.31 % between 1969 and 1979. Kim and Stulz (1992) find a
significant average return of -1.66% for domestic convertibles and a significant
-0043% for Eurobond convertibles between 1965 and 1987.
In international markets, emerging economies choose to issue Eurobonds
over foreign regulated bonds like Yankee bonds and Bulldogs, which might be
inaccessible to them anyway. Lack of disclosure requirements and absence of
registration and prospectus make it easier for firms from emerging markets to
issue Eurobonds. With fewer covenants, their bearer nature and no pressure
for rating, emerging markets are increasingly becoming regular players in the
international market vying for private capital. However, only the most presti-
gious firms are able to access the Eurobond market. In addition, being unregu-
lated it is easier to issue the (more popular) convertible Eurobonds by these
firms, since their equity is not internationally listed. Several Eurobonds are
1 Abnormal returns are excess returns over and above risk-adjusted returns. According to semi-
strong efficient market hypothesis, abnormal returns should be zero.
262 K. Tandon
Country funds
2 It might be interesting to compare the effects of Eurobond issues to those of GDRs and ADRs
from emerging economies.
3 Such performance is not linked in any way to agency theory or changes in cost of capital of
the top tier firms. Research on financial innovations and cost of capital in emerging countries is
beyond the scope of this research.
4 Kim and Singhal (1993) have analyzed the impact of market openings in South Korea. Bosner-
Neal et a1. (1990) analyzed market openings in five developed and emerging markets on their
country fund prices.
External financing in emerging markets 263
DATA
5 Our data set ends in 1992. In 1993 and 1994 there were an additional 75 Eurobond offerings
from private firms in emerging countries. Preliminary results do not indicate any major changes
when data are updated to mid-1994.
6lt is difficult to obtain the detailed composition of the fund at the time of the launch data. We
obtained the composition of the top ten holdings at a time as close as possible to the launch date.
Some old prospectuses have been made available by Chris Persichetti of Lipper Analytical Services.
Some have been obtained from a recent research report by Smith Barney.
264 K. Tandon
included in the EMDB data base. In addition to the market index, this
research analyzes the pre- and post-launching performance of an equally-
weighted portfolio of the top holdings in the fund, for which we have data
in our IFC data base. For a listing of the mutual funds in our data base
and their composition of top ten holdings, see Appendix 2.
Table 1 offers information about Eurobond offerings by country and some
statistics about our sample. Our final sample of 49 firms included firms from
nine countries over the period 1990-1992: Argentina (4), Brazil (11), Mexico
(7), Venezuela (3), Portugal (2), Indonesia (2), Thailand (1), Korea (16) and
Taiwan (3). There have been additional corporate offerings of Eurobonds in
1993 by firms from India, Chile, Colombia, Malaysia and Philippines, in
addition to the above nine countries. The average size of the Eurobond offering
in our sample is $88 million and the average maturity is 6 years. As evidenced
in Table I, there is a big increase in the number of offerings in 1993. This indi-
cates the increasing popularity of corporate Eurobonds by emerging markets
as a source of private sector funds for growth and development. Several of the
Eurobonds from Asian countries are convertible, while some of them from
Latin America are callable.7 Most have maturities between 2-15 years, with
5-year maturity being preferred by emerging markets.
Table 1. Frequency distribution of corporate Eurobond offerings by country, amount, number and
year of offerings
7 Due to our small sample, we do not analyze the sub-samples for convertibles, straights and
callables separately in this paper.
External financing in emerging markets 265
METHODOLOGY
Eurobonds
The methodology used in the first part of this study about Eurobonds is similar
to a standard event methodology, often used in the finance literature. This
methodology has been successfully applied to similar problems in the papers
by Kim and Stulz (1988, 1992), Eckbo (1986), Mikkelson and Partch (1986)
and Dann and Mikkelson (1984). It involves measuring the abnormal return
on the firm's stock surrounding the date of issuance.
The impact of the Eurobond offering on common stock of the issuing firm
is estimated using the following market model for stock returns:
where, Rjt = return on stock j over week t, R mt = return on market index over
week t, ejt = error term on security j in week t with the properties, E(ejt) = 0
and E(ejt-l. ejt) = o.
The above equation is estimated for each offering announcement. For each
country, that country's own market index is used. The estimation period is
weekly from t = - 56 to t = - 5, relative to the initial announcement date as
recorded in the Euromoney Bondware data base and The World Bank's World
Debt Tables. The estimated parameters, aj and bj , and the realized (actual)
return on the market index in week t are used to construct predicted returns
around the event date. Excess returns are then computed as the deviation
between realized returns and predicted returns. Specifically, the abnormal return
for firm j in week t is computed as:
Weekly abnormal returns are calculated for each firm over the interval weeks
t = - 3 to t = + 3. For a sample of N firms, the weekly average abnormal
return (AR) for each week is obtained: 8
1 N
AR, = N j~l AR j ,
8 We first compute the average abnormal return over all firms for weeks - 3 to + 3 surrounding
the announcement date. Cumulative abnormal returns (CARs) are the abnormal returns summed
over weeks - 3 to + 3.
266 K. Tandon
where
and where, (1j = standard deviation of the residuals in the estimation period,
T = number of weeks in the estimation period, Rm = mean return on the market
portfolio over the estimation period, and Rmk = return on the market portfolio
in the kth week of the estimation period.
Assuming normality and independence across securities, the Z-statistic is
computed as follows:
Zt=VNoSAR t
The Z-statistic is then used to test the hypothesis that the average standardized
abnormal return equals zero.
Country funds
The second part of this research examines the impact of the launching of
closed-end country funds (emerging markets only) on the market index of the
reference country and on an equally-weighted portfolio of the top holdings of
the fund at the time of the fund launching. We are interested in an equally
weighted portfolio of the top ten holdings but are often constrained by data
availability of these top tier firms in the Emerging Market Data Base (EMDB).
As a result, our portfolio of top holdings ranges from a low of only four firms
to a high to ten for other funds, except for one of the two Brazil funds with a
high of 14 firms.
We compute the mean and variance of the returns of the underlying market
index and the above constructed portfolio of top holdings for a period of
12 months before and 12 months after the launching of the mutual fund. We
also compute the mean adjusted excess returns on the market index and the
portfolio of top holdings for up to 2 months after the launching of the fund.
To calculate the mean adjusted excess return, we first take the mean of the
returns for 12 months preceeding the launching of the fund. To arrive at the
mean adjusted excess return post-launching, we deduct the mean from the
following month's return after the launching of the country fund. 9
EMPIRICAL RESULTS
9 For the portfolio of top holdings, we adjusted the portfolio's return by the mean of the market
index also, but no significant differences are noticed.
External financing in emerging markets 267
ment week and weeks surrounding it. Abnormal returns are computed as
market-model residuals as outlined above. Table 2 presents the abnormal
returns for three small windows for the entire sample of 49 offerings as well as
for each country. Cumulative abnormal returns are presented in column 3 for
week zero (the event date), in column 4 for weeks 0 and + 1 around the event
and in column 5 for weeks -1 and + 1 around the Eurobond offering.
As can be seen from Table 2, the Eurobond offering for the entire sample is
associated with a positive significant excess return on the underlying stock
over windows (0, + 1) and over (-1, + 1). Most of the excess return is driven
by return in week + 1, following the succesful completion of the Eurobond
offering.
A breakdown of the cumulative abnormal returns by country of origin
suggests that the positive returns are all driven primarily by the subsample of
Korean firms. Though positive returns are associated with five countries, only
returns in Korea are statistically significant. However, individual country analy-
sis should be treated with caution, given the small sample size associated with
several of these countries, except Korea and Brazil. .
Interpretation of results
A number of studies have examined the returns on common stock around the
bond offerings in the USA. To summarize, Eckbo (1986) and Mikkelson and
Partch (1986) find a negative stock-price announcement effect associated with
straight-debt issues, while Kim and Stulz (1988) find a small positive announce-
ment effect for straight Eurobond issues. For domestic and Eurobond convert-
ibles respectively, both Dann and Mikkelson (1984) and Kim and Stulz (1992)
find negative abnormal returns.
Table 2. Weekly excess returns on an equally weighted portfolio of stocks around Eurobond
offerings
Sample
No. of offerings CAR (0) (%) CAR (0, 1) (%) CAR (-1, +1) (%)
Cross-sectional regressions
We test the source of this gain by estimating a cross-sectional regression of the
abnormal returns on two variables: size and maturity of the Eurobond offering.
Though other variables may be relevant in explaining the abnormal returns,
given the data available, we regress standardized cumulative abnormal returns
(SCARs) against size of the offering (AMT) and maturity (MAT). We find the
following relationship:
SCARs (-1, + 1) = 0.586 - 0.045 AMT - 0.05 MAT
(1.53)(-0.99) (-1.14)
There is no significant relationship between abnormal returns and size or
maturity. (Results for SCARs for other windows are similar.) Although not
statistically significant, abnormal returns are inversely related to size and
maturity, a result similar to that found by Kim and Stulz (1988).
t= -12 t= +1 t= -12 t= +1
Mutual fund to -1 t=O to +12 to -1 t=O to +12
India data are only for 4 months before and 4 months after.
b Indonesia data are for 3 months before and 3 months after.
270 K. Tandon
launching on the volatility of returns on the underlying market index and the
equally weighted portfolio of top holdings.
Table 3 Panel A shows that the ratio of post- to pre-launching mean return
is greater than one in only four of 15 funds for the market index and in only
two cases for the portfolio of top holdings. In other words, the mean return
decreased in 11 of 15 cases for both the market index and the equally weighted
portfolio of top holdings after the country fund is launched. In fact, the mean
return is positive in all 15 cases before the launching of the country fund but
became negative in nearly half the cases in the months after the fund is launched.
In the month of the fund launching (i.e. t = 0), when overseas fund managers
are busy buying securities in the local market, the mean return' is higher than
the pre-launch return in nine of 15 cases for the market index and in seven
cases for the portfolio of top holdings. Post-launching mean returns declined
sharply. This may be an indication of the so called over-reaction hypothesis in
the capital markets. While stock prices rise some time before and at the time
of the launching of the closed-end country fund on indications of a one time
influx of foreign money in the country's stock market, the stocks tend to cool
down once the fund is launched. The presence of foreign fund managers in an
emerging economy leads to a greater foreign monitoring, making the markets
less speculative.
Table 3 Panel B presents the results of the effects of the country fund
launching on the returns volatility of underlying stocks. The volatility compari-
son is based on the ratio of post- to pre-launching variances computed from
monthly returns 12 months before and 12 months after the fund launching.
Column 2 presents the variance ratio for the market index and column 3
presents the ratio for the equally weighted portfolio of top tier firms in the
country fund. Variance tended to decrease in nine of the 15 cases for both the
market index and the equally weighted portfolio of top holdings. This lends
further support to the above hypothesis that the presence of foreign fund
managers make the emerging capital markets less speculative and that greater
and more experienced monitoring and/or financial analysis makes the emerging
markets less speculative.
In addition to the mean return and the variance of stock prices, we
compute the mean-adjusted excess returns for both the market index and
the portfolio of the top holdings (Table 4). These estimates are obtained
from the mean adjusted returns model, where the estimation period is month
-12 to month -1 relative to the launching of the country fund (month 0).
Similar to Table 3, mean adjusted excess returns are positive in nine cases
for both the market index and the portfolio of top holdings in the month
t = O. However, cumulative excess returns for 3 months from t = 0 to t = + 2
are positive in only half the cases. The mean across all the country funds in
month t = 0 is 2.44% for the market index but 6.20% for the portfolio of
the top holdings. However, the mean for cumulative mean-adjusted excess
returns for months t = 0 to t = + 2 are negative for both the market index
External financing in emerging markets 271
Table 4. Mean adjusted excess returns on market index and portfolio of top holdings post-
launching of country funds (%)
Country fund CAR (t=0) CAR (0, +2) CAR (t=O) CAR (t = 0, +2)
as well as the portfolio of top holdings, being more negative for the overall
market than the portfolio of top holdings.
This research examines the behavior of stock prices in emerging capital markets
to two announcements related to the opening of that country's capital market.
Eurobond offerings have increased substantially in the emerging markets in
the last 3 years and their offerings are associated with positive abnormal returns
to the underlying stocks around the offering, leading to unexpected gains
for their shareholders and maybe to a lower cost of capital for the firm. This
also implies that there is value associated with the successful completion
of such offerings in emerging markets, since they enhance the exposure and
creditworthiness of the firm in international markets.
We also analyzed an aspect of the opening of emerging capital markets: the
launching of their closed-end country funds. Mean returns in the month of the
fund launching and months preceding it are high and positive in anticipation
and influx of new foreign money in the market; however, the trend reverses
and returns decrease in the months following the launching of the fund.
Increased monitoring and advanced financial research/analysis by fund manag-
ers and analysts leads to a decrease in the speculative behavior in the market,
as evidenced by a decrease in stock market volatility following the launching
of the country fund. This decrease in volatility may ultimately bring discipline
and more responsible development of the emerging stock market by decreasing
speculative bubbles.
Emerging markets are quickly discovering the benefits of the global market
as a source of raising private capital funds and most are now doing it through
equity issues in addition to the Euro-debt issues. They are doing this increas-
ingly through new issues of Global and ADRs. The economic benefits to
shareholders and a greater discipline in the domestic capital market accompa-
nying market openness is encouraging other emerging economies, like China,
Philippines, Pakistan, Sri Lanka, etc,. to open their capital markets to foreign
investors. As a logical extension of this research, it would be interesting to
analyze the economic effects of GDR and Rule 144A ADR offerings on the
stock prices of emerging market firms.
ACKNOWLEDGMENTS
This research has been supported by the Debt and International Finance
Division and the Research Support Budget (RPO 678-49) of The World Bank.
The author thanks Stijn Claessens and Shan Gooptu of IECDI, World Bank,
and Nusret Cakici (Rutgers) and Jahangir Sultan (Benteley) for comments.
I also thank Euromoney for access to their Bondware Data Base and to Chris
Persichetti of Lipper Analytical Services. The findings, interpretations and
conclusions are the author's own and should not, in any way, be attributed to
the World Bank. This research is also printed in Emerging Markets Quarterly
(Summer 1997), New York, 63-73.
External financing in emerging markets 273
REFERENCES
Bailey, W. and 1. Lim (1992). Evaluating the diversification benefits of the new country funds.
Journal of Portfolio Management, 18, 74-80.
Bosner-Neal, c., G. Bauer, R. Neal and S. Wheatley (1990). International investment restrictions
and closed-end fund prices. Journal of Finance, 45,523-548.
Claessens, Stijn and S. Gooptu (eds) (1993). Portfolio Investment in Developing Countries.
Washington, DC: World Bank.
Dann, 1. and W. Mikkelson (1984). Convertible debt issuance, capital structure change and
financing-related information. Journal of Financial Economics, 13, 157-186.
Diwan, I., V. Errunza and 1. Senbet (1993). National index funds - empirical perspectives. In Stijn
Claessens and S. Gooptu (eds), Portfolio Investment in Developing Countries. Washington, DC:
World Bank.
Eckbo, E. (1986). Valuation effects of corporate debt offerings. Journal of Financial Economics,
IS, 119-151.
Gooptu, S. (1993). Portfolio investment flows to emerging markets. In Stijn Claessens and
S. Gooptu (eds), Portfolio Investment in Developing Countries. Washington, DC: World Bank.
Hardouvelis, G., M. La Porta and T. Wizman (1993). Closed-end country funds. NBER Conference,
June 1993.
James, C. (1987). Some evidence on the uniqueness of bank loans: a comparison of bank borrowing
agreements, private placements and public debt offerings. Journal of Financial Economics, 19,
217-235.
Kidwell, D., W. Marr and G. Thompson (1985). Eurodollar bonds: alternative financing for US
companies. Financial Management, 14, 18-27.
Kim, E. Han and Vijay Singhal (1993). Opening up of stock markets by emerging economies: effect
on portfolio flows and volatility of stock prices. In Stijn Claessens and S. Gooptu (eds), Portfolio
Investment in Developing Countries. Washington, DC: World Bank, 393-402.
Kim, Y. C. and R. Stulz (1988). The Eurobond market and corporate financial policy: a test of the
clientele hypothesis. Journal of Financial Economics, 17, 189-205.
Kim, Y. C. and R. Stulz (1992). Is there a global market for convertible bonds? Journal of Business,
65,75-91.
Mikkelson, W. and M. Partch (1986). Valuation effects of security offerings and the issuance process.
Journal of Financial Economics, 15,31-60.
World Bank. World Debt Tables 1992-93, 1993-94.
274 K. Tandon
* Though we have information on the top ten holdings in each fund at the time of launching, we
do no have data on all ten holdings. This column states the number of firms for which we have
data to form our portfolio of top ten holdings.
ROBIN L. DIAMONTE,I JOHN M. LIEW 2 and ROSS L. STEVENS3
IGTE Investment Management, 2Goldman Sachs Asset Management, 3Integrity Capital
Management
ABSTRACT
Using analyst estimates of political risk, we show that political risk represents a more
important determinant of stock returns in emerging than developed markets. Average
returns in emerging markets experiencing decreased political risk exceed those of emerging
markets experiencing increased political risk by approximately 11 % per quarter. In contrast,
the difference is only 2.5% per quarter for developed markets. Further, the difference
between the impact of political risk in emerging and developed markets is statistically
significant. We also document a global convergence in political risk. Over the last 10 years,
political risk has decreased in emerging markets and increased in developed markets. If this
trend continues, the differential impact of political risk in emerging vs. developed markets
may narrow.
INTRODUCTION
Does political risk affect stock returns? The often observed link between dra-
matic political events and large market moves clearly suggests that it can.
However, because quantifying political risk is difficult, little beyond anecdotal
evidence exists which examines its systematic impact on stock returns. We
provide direct evidence on this issue by exploiting analyst estimates of politi-
cal risk.
First, we show that changes in political risk have a bigger impact on returns
in emerging markets than in developed markets. In emerging markets, political
risk changes represent an economically and statistically significant determinant
of stock returns. Average returns in emerging markets experiencing decreased
political risk exceed those of emerging markets experiencing increased political
risk by approximately 11 % per quarter. Changes in political risk represent a
less important determinant of stock returns in developed markets. There is no
statistically significant difference between average returns in developed markets
experiencing decreased political risk and developed markets experiencing
increased political risk.
Second, we document a global convergence in political risk. Over the last
ten years emerging markets have become politically safer while developed
markets have become riskier. If this trend continues, our results suggest that
1 A similar version of this paper was published in the May/June 1996 Financial Analysts Journal.
277
R. Levich (ed.), Emerging Market Capital Flows, 277-289.
1998 Kluwer Academic Publishers.
278 R.L. Diamonte et al.
DATA
Political Risk
Low 75
I=~II
Risk
65
55
45
35
25
ffiIh Iraq Invasioo of Kuwait
Riok 15
..,
VI
Figure 1. Political risk in Kuwait and Iraq around the Gulf War.
Political risk in emerging and developed markets 279
decreased more slowly. The level of political risk in Iraq has yet to return to
its pre-invasion level.
Figures 2-4 show summary statistics of the political risk measure for the 21
developed and 24 emerging markets for which we have both stock return and
political risk data. We present the average risk, average quarterly risk change,
and standard deviation of the quarterly risk change. 2 Figure 2 shows that
emerging markets have been politically more risky than developed markets. In
fact, the riskiest developed market, Hong Kong, has been politically safer than
all but five emerging markets. ICRG's political sub-component data (described
in the appendix) indicates that relative to the other developed markets, Hong
Kong has suffered from weak political leadership and high external conflict
risk. Within the emerging markets, the two riskiest countries, Pakistan and Sri
Lanka, help make Asia the riskiest region. Pakistan has been plagued by
corruption in government and Sri Lanka has suffered political terrorism and
substantial civil war risk.
Figure 3 documents a global convergence in political risk over our sample
period. Emerging markets have become politically safer while developed mar-
kets have become riskier. The average change in political risk is negative
(riskier) for 19 out of 21 developed markets and positive (safer) for 21 out of
24 emerging markets. Among the emerging markets, Chile has experienced the
largest average decrease in political risk due primarily to its strong political
leadership and consistency of free market reforms. However, in spite of Chile's
gains, the Latin America region has become safer at a slower rate than either
the Aisa or the Europe/Mideast/Africa regions. The Philippines (Asia) and
Zimbabwe (Africa) have been the largest contributors to their respective regions'
decreasing political risk. The Philippines has benefited most from strengthening
political leadership while Zimbabwe has enjoyed a sharp drop in external
conflict risk.
Figure 4 documents large and volatile political risk changes in emerging
markets compared to developed markets. However, emerging market regions
exhibit less volatile political risk changes than most individual developed
countries since political risk changes are not perfectly correlated across coun-
tries. The fact that the changes are larger and more volatile in emerging markets
does not necessarily imply that they represent a more important determinant
of stock returns in these markets.
Stock return data
2We used quarterly, as opposed to monthly, changes for two reasons. First, ICRG's political
risk measure changes slowly over time and it is often the case that a country may not experience
a political risk change over a single month. However, over any quarter virtually all countries
experience changes. Second, the substantial first-order autocorrelation observed in monthly stock
returns for many emerging markets suggests that non-synchronous trading contaminates the
monthly return data. The use of quarterly return data mitigates this problem.
280 R.L. Diamonte et al.
-
, ,
""",JOZliMS
I I I
I """'"'"
,
I I
"""""1l0N
.........,00
I
A_
, I
_s
pl.IlIfJlIIB2 ..... BN
, utdllJf"
kOlWO()
I
..... ":l
, I I
,
,
I
"".... ,
o,ooofto,rs
UIlrNljV.1.
AI""
I
A-JoOunH
, ,IIIlUJOd
",ods
001101'1
, '!$A ...I'I
iluo)<OuoH
I
"IO~
XopuJ .....,
'i'\4'ln2'8UOA
XIlPUIO:II
]
'0
X8pU1 Ur:lV'l
.,c..
..c
'I'"IS
E-
GOoeJD
.... uoII""
""fIOd
""tI!l\U
IlOO
XeQllIVW3
u.".,..
'IIrcOlJOPUI
-,
...o6iN
I
M\l!dd!I!4d
""'d
""".,O'S
oo o o<0 o o o o o
'" <0 ID
'"
:E .."
rD ~.
qQ' ;:
::r'" Average Quarterly Change in Political Risk (1985-1995)
C> W
'" 6 6 o o o o o
p...
- :... N o N :... 0> CD
Q~
Austrau...l-iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii~---II---t---1--
:s
SwllZ.er'll.nd
~
m
NOI'WIY
5' USA
l Jopan
c. Swodon
g, I>onmatk
i;l. A\Jslria
!'r Firde: nd
3 ,....
~. UK
:s Singflpelf.
(JQ
< New2elilland
0.
'" Nel:hert4l~
~ Iklgl_
~
Moo.~liCo
~0. MSCllndox
S Co .....
~
;;:- German)'
Ireland
F GobnbiO
;! VenelUeLa
C>
F,lnce
l e"uu
[ Tatwll'l
lJ~rMe.ll
,..a Hungary
0.
TuOtev
a
~ Spain
~
@ As1a4ndc!1lC
Malay:u:
3'
3 IFClndllC
Thailand
n
,:l EMA Ind!X
p Hong Kong
-l PoI1ug.,
::r
C>
Nlgo'"
5' G,,,,,,,,,
0.
~ Korea
"0
S: ""'"
lndonll!!ls.
5'
e:.. SrI lanka
:I. Pa kb,'a n
m
,..
0.
,....
~
Argemlni
S-
Ol Jof'l1an
@ ZII'T'ltblilbwtl
PoIoIId
'2.
[ Phllpplno.
N' C/oij.
ao
:;'
18~ SJiJ)(.ItJUl pado/aaap pUtJ 3u!-8.1aUla U! >[S!.1 /V;J!J!l0d
tv
00
tv
3.5 TI-------------------------------------------------------------------------------- ~
r-.
t:::l
II ~.
i;;; 3.0 I I
I
:::::. ~
....!I
ib
~2.5 rl----------------------------------------------------------~ ~
....
~ \:)
;f :-
i. 2.0 I r
~
~
J 1.5
"0
i 1.0 1 r.J IH-I-il 1
">
c!
I 0.5 -t....u-....n I ~H-I-tf-H-I l-I- 1
GI
00 II ,U. I ,Y, . 1 ,1 ,11, 1 ,1 rI ,I ,. 11 ,1 ,1 ,1 ,1 1 ,1 ,11, . ,11,1 1.1 1,11, 1 ,11, 1 ,11 1111 11 ,11 ' ' ,' 11 ,11 ,' ,111111, 11 ,11 ,11, 11,11.11,1 1
. ~11~jiJtJl:j!llf~ Ijj1iJfJ~j!1IJ!i}lj~ili~~Jj!I~~!1
Figure 4. Volatility of political risk changes: emerging vs. developed markets. The political risk data are from ICRG. The index political risk data are
capitalization-weighted.
Political risk in emerging and developed markets 283
Test procedure
We divided our sample of countries into two categories (emerging markets and
developed markets) based on their classification by IFC and MSCI. For each
category and each calendar quarter, we form two portfolios. The first portfolio
contains the countries that experienced political risk increases and the second
portfolio contains those that experienced risk decreases. We then calculated
portfolio returns by weighting each country's return by the absolute value of
its contemporaneous percentage risk change. Thus, countries experiencing large
risk changes receive more weight than those experiencing small risk changes. 3
We repeated this procedure each quarter to obtain a time-series of returns to
each portfolio.
"""
6 ,0 ~
t--
in 5,0 t:::l
$on is'
~ ~:lI
:'c: 4,0 -
~
0
E ~
! 3.0
i"0 ?t
.,:::>c 2.0
,5
E 1.0 - - - - - - -
~
i
a:
;; n
0 0.0 Il~ ll il
t-
I!
"'"
~ -1.0 IU'II
<II
-2.0 . --
j
.. J:
I i ~~ 1
~
~ ~ z ~
J!I~il~~ jji~~~~af~~ ..~ i~ -E
~ o !!o ~ ~s
i5~~ < z
~ - ~
~ .. IIJil}jl
~! J ~ ~
1!lji ,~j] ~ilf!:~
~
~ ~ ~ .. ... s t- }
~
z
Figure 5. Average stock returns: emerging vs. developed markets. All returns are in US dollars and include dividends. Developed market returns are from
Morgan Stanley Capital International (MSCI). Emerging market returns are from the International Financial Corporation (lFC). Monthly returns start in
1/85 except for Portugal (2/86), Turkey (1/87), Finland (\/88), Ireland (\/88), New Zealand (\/88), India (2/90), Sri Lanka (10/93), Hungary (1/94), and
Poland (1/94). The index returns for the emerging market regions are capitalization-weighted. The MSCI and IFC index returns are based on their
published indices.
30.0,--------------------------------------------------------------------------------------
I25.0 +1------------------------------------------------------------------------~
1E
~ 20.0 I n I
!!
~
!S 15.0 ~
.E ;:;.
..,:5 e..
...
"> c:;;.
c! 10.0 ~
;;.
<JI
I ~
~
E 5.0 ~
:>
~.
a:
" s:::.
~
0.0 l:>...
.
.., ,! ~ ~ .8 'll ~
:>
~ ] 8. :i! ~- II f ~
'" "" J! en j j .~ Ol
E ~
.f
-" g ::>' {! ~
0 Co .:s ~
J ~
0..
t I 1
Z
r" i .. J i i
Z ~
~
Figure 6. Volatility of stock returns: emerging vs. developed markets. All returns are in US dollars and include dividends. Developed market returns are from 3
Morgan Stanley Capital International (MSCI). Emerging market returns are from the International Financial Corporation (IFC). Monthly returns start in ...s:::.
~
1/85 except for Portugal (2/86), Turkey (1/87), Finland (\/88), Ireland (\/88), New Zealand (\/88), India (2/90), Sri Lanka (\0/93), Hungary (1/94), and ~
Poland (1/94). The index returns for the emerging market regions are capitalization-weighted. The MSCI and IFC index returns are based on their t:i
published indices. N
00
Vl
286 R.L. Diamonte et al.
Results
Tables 1 and 2 contain our results. For both emerging and developed markets,
Table 1 shows the average return, t-statistic, and the average political risk
change for four portfolios:
Portfolio 1: decreasing risk
Portfolio 2: benchmark index
Portfolio 3: increasing risk
Portfolio 4: decreasing risk - increasing risk (Portfolio I-Portfolio 3)
The average return on the decreasing risk portfolio exceeds that of the bench-
mark and the increasing risk portfolio for both developed and emerging mar-
kets. In emerging markets, returns on the decreasing risk portfolio exceed those
of the benchmark by over 8%/quarter and those of the increasing risk portfolio
by over 11 %/quarter. These differences are far less dramatic in developed
markets where returns on the decreasing risk portfolio exceed those of both
the benchmark and the increasing risk portfolio by less than 3%/quarter.
The risk decreasing minus risk increasing portfolio returns (portfolio 4) are
strongly statistically significant in emerging markets (t-statistic = 3.90), but
only marginally significant in developed markets (t-statistic = 1.84). This evi-
dence suggests that changes in political risk clearly help explain the cross-
section of country returns in emerging markets, but only marginally help in
developed markets.
We also see that emerging markets produce larger average political risk
changes than developed markets. However, this difference (4.02 emerging vs.
Table 1. Changing political risk and stock returns (January 1985-June 1995)
Average Average
return Average return Average
(%/qrt) t-statistic L1 risk (%/qrt) t-statistic L1 risk
All returns are in US dollars and include dividends. Developed market returns are from Morgan
Stanley Capital International (MSCI). Emerging market returns are from the International
Financial Corporation (IFC). Monthly returns start in 1/85 except for Portugal (2/86), Turkey
(1/87), Finland (1/88), Ireland (1/88), New Zealand (1/88), India (2/90), Sri Lanka (10/93), Peru
(10/93), Hungary (1/94), and Poland (1/94). The political risk data are from Political Risk Services:
International Country Risk Guide (ICRG). The benchmark for the developed markets is the MSCI
World Index. The benchmark for the emerging markets is the IFC index. The decreasing (increasing)
risk portfolio consists of countries whose political risk decreased (increased) over the period.
Average ~ risk is the equal-weight average quarterly change in political risk for each country in
the portfolio over the full sample period. Within each portfolio, returns are weighted by their
percent change in political risk.
Political risk in emerging and developed markets 287
Table 2. Comparative effects of changing political risk on emerging and developed stock returns
(January 1985-June 1995)
All returns are in US dollars and include dividends. Developed market returns are from Morgan
Stanley Capital International (MSCI). Emerging market returns are from the International
Financial Corporation (IFC). Monthly returns start in 1/85 except for Portugal (2/86), Turkey
(1/87), Finland (1/88), Ireland (1/88), New Zealand 0/88), India (2/90), Sri Lanka (10/93), Peru
00/93), Hungary (1/94), and Poland (1/94). The political risk data are from Political Risk Services:
International Country Risk Guide (ICRG). The benchmark for the developed markets is the MSCI
World Index. The benchmark for the emerging markets is the IFC index. The decreasing (increasing)
risk portfolio consists of countries whose political risk decreased (increased) over the period.
Average .1 risk is the equal-weight average quarterly change in political risk for each country in
the portfolio over the full sample period. Within each portfolio, returns are weighted by their
percent change in political risk.
3.35 developed) does not fully explain the larger average quarterly return to
portfolio 4 in emerging vs. developed markets (11.28% vs. 2.46%). Emerging
market returns are also more sensitive to a given change in political risk. In
portfolio 4, the average quarterly return per unit of political risk change in
emerging markets (11.28%/4.02=2.81%) exceeds that in developed markets
(2.46%/3.34 = 0.74%). Thus, the magnitude of political risk changes and the
sensitivity of stock returns to a given change in political risk are both larger
in emerging than developed markets.
Table 2 tests the differential impact of changes in political risk between
emerging and developed markets for statistical significance. We subtract the
developed market portfolio return from the emerging market portfolio return
for each set of portfolios in Table 1. For the set of decreasing risk portfolios,
emerging market average returns exceed those in developed markets by almost
6%/quarter (t-statistic = 2.36). For the set of increasing risk portfolios, emerg-
ing market average returns fall below those in developed markets by almost
3%/quarter, although the difference is not statistically significant (t-statistic =
-1.37). Our bottom line result from Table 2 shows that the differential impact
of political risk changes between emerging and developed markets is economi-
cally and statistically significant. The average return difference between port-
folio 4 for emerging markets and portfolio 4 for developed markets is almost
9%/quarter with a t-statistic of 2.72.
Taken together, our results in Table 1 and 2 support the hypotheses that
political risk represents an important determinant of stock returns in emerging
markets, and political risk affects stock returns more in emerging than devel-
oped markets.
288 R.L. Diamonte et al.
CONCLUSION
This paper quantifies the importance of political risk in emerging and developed
markets.
Our two main results are easily summarized:
1. Changes in political risk have a bigger impact on emerging market returns
than on developed market returns. The average difference in returns between
emerging markets experiencing decreased political risk and those experienc-
ing increased political risk is approximately 11 % per quarter. In developed
markets the difference is only 2.5%.
2. Over the last 10 years we have seen a convergence in global political risk.
Emerging markets have become politically safer while developed markets
have become politically riskier.
These results suggest several prescriptions. In emerging markets, if one can
forecast changes in political risk, one can forecast stock returns. Therefore,
emerging market analysts are well advised to devote considerable resources to
forecasting political risk changes. In developed markets political risk is less
important. Developed market analysts are better off devoting resources to
forecasting other sources of return such as changes in expected future economic
conditions. However, if global political risks continue to converge the large
differential impact of political risk changes between emerging and developed
markets may fade.
ApPENDIX
The International Country Risk Guide (ICRG) Model for forecasting financial,
economic, and political risk was created in 1980 by the editors of International
Reports. Used by banks, multinational corporations, importers, and exporters,
among others, the ICRG model is used to determine the risks of operating in,
investing in, or leading to particular countries using an evaluation system that
examines current political, financial, and economic risk.
The combined political risk score rates countries on a scale from 1 to 100
and consists of the following thirteen components, each assigned a maximum
numerical weighting.
INTRODUCTION
The dramatic increase in private portfolio flows into emerging market countries
at the beginning of the 1990s has received a lot of attention. Less well-known
are the changes in cross-border bank lending to these countries. Net short-
term bank flows to major emerging market countries averaged US$ 35 billion
a year from 1993 to 1995, a more than 4-fold increase over the 1988-1990
period. Over the last two years, long-term bank lending to major emerging
market countries also picked up vigorously after years of stagnation. This is
shown in Figure 1, and one might be tempted to conclude that we are facing
a renaissance in lending to developing countries. Today the volume of net bank
lending to major borrower countries exceeds that of 1982, i.e. the year when
such lending peaked and the so-called international debt crisis broke out. Even
when measured in terms of lending as a percentage of GDP, cross-border
lending is higher today than it was in 1982.1 Of course, this raises questions as
to the sustainability of these flows and the efficiency of international lending.
An initial question suggested by Figure 1 is whether history is repeating itself.
Could it even be that the recent Mexican peso crisis was a forerunner of more
general fragilities in the financial systems, similar to what we experienced at
the beginning of the 1980s?
With respect to the major theme of this conference on the history and the
future of emerging market capital flows, two other questions present themselves:
what has changed between 1982 and today? What are the stylized facts and
how do they fit into economic theory? To answer these questions, we first
present the stylized facts of what has changed in cross-border bank lending to
emerging markets over the last decade. We then go on to see how these facts
fit into economic theory. In doing so, we will concentrate on enforcement
problems associated with international loan contracts and leave out the ques-
tion of risk sharing. Finally, combining facts and theory, we might be able to
say something about the future of cross-border bank lending. Will bank lending
* The views expressed in this paper are those of the authors, not necessarily those of SBC.
1In 1982, the ratio of net cross-border lending to GDP for major borrowing countries was
0.7%. In 1994 this ratio jumped to 0.87% and in 1995 it reach a temporary peak of 1.22%.
293
R. Levich (ed.), Emerging Market Capital Flows, 293-305.
~ 1998 Kluwer Academic Publishers.
294 P. Aerni and G. Junge
STYLIZED FACTS
2 Today we may face another type of major risk arising mainly from the sheer magnitude of
short-term dollar-linked debt.
296 P. Aerni and G. Junge
such as the UK, France and Germany underwent similar adjustment processes.
For individual banking groups, such as the nine biggest US banks or the three
biggest Swiss banks, which typically were forerunners in syndicated lending to
developing countries, the adjustment was even more dramatic than the averages
in Table 1 suggest.
Fact 2: The exposure of the international banking system to emerging market
countries has decreased considerably.
At the same time as the international banking system's exposure to emerging
market countries decreased, the creditworthiness of these countries, as mea-
sured, for example, by their debt-to-export ratios, has improved substantially
(see Table 1). This change was caused not least of all by growth in exports well
above the world average. The new export strength is one of the fruits of the
liberalization programs in the real economy and the financial sector launched
by most developing countries in an attempt to reinstate their creditworthiness
in the aftermath of the international debt crisis.
Fact 3: The creditworthiness of the emerging market countries has greatly
improved.
Besides fostering international trade, the liberalization programs pursued by
many emerging market countries in the financial sector also lowered the cost
disadvantage of external funds over domestic finance as perceived by the private
sector. This resulted in an increase in the demand for external funds in general
and offered new investment opportunities for international banks. The interna-
tional banking industry accordingly became more interested in securities-
related cross-border activities. More importantly, the financial liberalization
process also allowed banks to shift their policy focus from public to private
borrowers.
Fact 4: International banks concentrate their cross-border lending on the private
sector and to a large extent on the domestic banking sector.
The rediscovery of the private sector in developing countries might be the most
important structural change of recent years. As shown in Figure 3, bank lending
is headed increasingly in the direction of private borrowers and domestic banks.
Specifically, the non-bank private sector's share of total external bank claims
has increased in Asia from 28% to 46% and in Latin America from 28% to
44%. This increase reflects the new boom in trade finance, the large-scale
privatizations and the forgiveness of sovereign debt under Brady-type
agreements, mainly in Latin America. The preference for lending to private
corporations in emerging markets is presumably driven by the better commer-
cial prospects. It also represents a shift away from the old-fashioned, general-
purpose lending to sovereigns and public entities. In the past, lending was all
too often conducted for general purposes under central government guarantees,
witlr important behavioral implications for both borrowers and lenders. For
the borrowing government, a loan to finance a good project often had no
Cross-border emerging market bank lending 297
Latin
America 1985 1995
countries implemented market reforms in the late 1980s and opened up their
economies, the new flows of lending went to the private sector.
Summarizing the main findings above, we conclude that over the last to-15
years, cross-border lending has lost its pre-eminent role in financing emerging
market countries. Consistently, the exposure of the international banking
system to emerging markets has decreased considerably. This is true despite
the fact that lending in absolute and relative figures is larger today than in
1982. Today, the role of cross-border loans has to be seen and explained within
a mix of other investment instruments, most of them tradable on a market,
such as bonds or equities. It seems to be more and more the case that interna-
tional banks treat their cross-border lending positions as part of an overall
foreign investment portfolio with corresponding implications for exposure
levels. In contrast to the early 1980s, today's international financing of emerging
markets appears to have a better footing: lenders are diversified and there is
no widespread economic borrower fragility. Moreover, the altered relevance of
cross-border lending to emerging markets has been accompanied by a general
shift in the banks' lending policy. Whereas in the 1980s loans were mostly
granted to sovereigns, most notably in Latin America, in the 1990s they have
been redirected to private borrowers in emerging market countries.
How do these findings coincide with economic reasoning and what is to be
expected in the future?
the case with multinational firms for instance, the threat that physical assets
will be confiscated in the event of an opportunistic contract violation is only
credible within an international legal framework that enforces such actions.
Accepting that cross-border loans are hard to collateralize, banks can
increase self-enforceability by writing short-term loans. In the short run, regula-
tory and political uncertainty is lessened and there is less room for opportunistic
behavior. Empirical evidence suggesting that banks prefer short-term loans is
given in Figure 4.
Short-term lending also provides a natural device for disciplining borrowers
whenever the loan serves as interim financing for long-term projects. In this
case, the bank not only has the option of periodically renegotiating the terms
of the contract but also commits itself to regularly screening the project. This
latter point is especially important when information about the soundness of
the borrower is scarce, creating potential moral hazard problems on the side
of the borrower. In this case, the bank should check whether the borrowing
firm actually invests as contracted or gambles on investment strategies with a
very low expected return but a small probability of exorbitantly large returns
in some states of the world. As loan repayments are fixed and thus independent
of actual returns where contractual obligations are met, pursuing such 'roulette'
strategies might be attractive to the borrower, but not to the lender.
Claim 3: Short-term cross-border lending as interim financing increases self-
enforceability and provides a credible commitment to monitor the borrower,
thereby reducing potential moral hazard problems.
Monitoring by banks is not only valuable in the presence of moral hazard, but
also in the context of adverse selection, where the need to monitor emerges
from an information advantage for the borrower with regard to the prospects
1980-1982 1993-1995
of the investment. In such a setting, a bank should only sign a loan agreement
when it has an advantage over its competitors in terms of monitoring. This
could be the case for, instance, if a bank has already financed a lot of similar
projects making it better at evaluating the risk on such loans. A bank with a
monitoring disadvantage will most likely only get to finance very risky, less
profitable projects, since borrowers with highly attractive projects prefer to
negotiate with a bank that does a better job of screening and is therefore able
and willing to offer the loan on better terms.
One way to bridge a 'gap' in monitoring is to finance particular investments
indirectly through better informed competitors. When such indirect financing
involves a cross-border loan, the self-enforceability problem reappears.
However, self-enforceability is not much of a problem if the domestic banking
industry is adequately integrated into the international financial markets and
would therefore face high external penalties in the event of opportunistic
behavior.
Which banks can we expect to have monitoring advantages with respect to
cross-border loans? In general, domestic banks in most industries are better
positioned to outperform their foreign counterparts in monitoring borrowers
as they are more integrated into the relevant information networks. However,
this argument loses weight when the value of a project depends more on world
market conditions than on home market conditions. When world market
information is valuable, it seems reasonable to assume that international banks,
which by definition are integrated into the international flows of trade and
finance, have at least no information disadvantage. Furthermore, as far as self-
enforceabilty is concerned, world markets only matter when potential borrow-
ers are somehow dependent on international trade and thus face high external
penalties for opportunistic behavior.
Claim 4: In the presence of adverse selection, we expect domestic banks to give
loans to home-market-oriented firms whereas foreign banks will either finance
loans indirectly through domestic banks or give loans to world-market oriented
firms.
intermediaries of loans are able to reduce monitoring costs through the econo-
mies of scale inherent in any systematic monitoring. From this theory it follows
that the distinguishing feature of a bank loan as compared to other means of
financing, for instance, foreign direct investment or portfolio investments in
bonds or equities, is the bank's commitment to monitoring the borrower's
activities.
Portfolio investments in bonds and equities lack this monitoring commit-
ment owing to free-rider problems among small investors. An atomistic investor
has little incentive to monitor the company any more seriously than by watching
market prices, since the benefits are enjoyed by all investors while the costs
are borne only by those actively carrying out the monitoring.
The distinction between foreign direct investment and cross-border loans
where the bank's monitoring role is concerned is not so obvious. A direct
investor is typically interested in control, an activity implying monitoring by
its very nature. However, the incentives to take appropriate action in case of
underperformance are quite different for a direct investor than for a lender. In
general, lenders are more able to commit themselves credibly to a tough
liquidation policy than direct investors. A direct investor shares the profits of
a financial enterprise on a pro rata basis and is therefore more vulnerable to
renegotiation in the event of non-performance than a lender sharing in the
profits on fixed terms. Hence, the difference between foreign direct investment
and cross-border loans lies in the credible commitment of the lender to taking
appropriate action in case of opportunistic behavior.
This admittedly narrow view of banks and bank loans seems to fit the empirical
evidence on the financing behavior of firms reasonably well. In general, internal
finance in the form of retained earnings constitutes the greater portion of net
funds, whereas the importance of the various external investment instruments
differs depending on the firm's size. In general, the funding preferences of firms
seem to follow a familiar pattern.
Claim 6: When deciding how to finance new investment projects, firms' funding
preferences follow a pattern akin to a 'pecking order' in which internal finance
(retained earnings) ranks well above external sources. In the latter category,
bank loans rank ahead of tradable securities for small firms, and vice versa for
large firms.
What does the future hold for cross-border bank lending? Will this form of
financing be crowded out as the theory of financial intermediation seems to
suggest or will it continue to boom as in the recent past? The theories of
sovereign lending and financial intermediation can offer only partial answers.
Nevertheless, they do provide a basic framework and, coupled with empirical
observations, permit us to draw the following conclusions.
First, general purpose long-term lending to sovereigns in emerging markets
does not seem to have such a rosy future. Barring major changes in international
law and contract enforcement mechanisms, lenders will most likely be reluctant
to channel massive sums to sovereigns in the form of general-purpose lending
no matter how great the demand in emerging markets might be.
Second, long-term cross-border lending to private borrowers will prob-
ably continue to play an important role in international finance. With the
development of further self-enforcing contract mechanisms, direct cross-border
304 P. Aerni and G. Junge
ACKNOWLEDGMENTS
The comments of Cliff Asness, Rick Buckholtz, Kent Clark, Britt Harris, Brian
Hurst, Antti Ilmanen, Bob Krail, Burt Porter, Rebecca Runkle, and Ingrid
Tierens are gratefully acknowledged. We also thank Laurel Fraser, Uwe
Ketelsen, Bettina Kessel, and Taro Harano for research assistance.
Cross-border emerging market bank lending 305
REFERENCES
Fazzari, Steven M., R. Glenn Hubbard and Bruce C. Petersen (1988). Financing constraints and
corporate investment. Brookings Papers on Economic Activity, 1, 141-195.
Lessard, Donald R. (1993). Country risk and the structure of international financial intermediation.
In Dilip K. Das (ed.), International Finance. Contemporary Issues. London, New York: Routledge.
Mayer, Colin (1988). New issues in corporate finance. European Economic Review, 32, 1167-1189.
Singh, Ajit (1995). Corporate Financial Patterns in Industrializing Economies. A Comparative
International Study. International Finance Corporation, Technical Paper No.2, Washington DC:
World Bank.
DONG-HYUN AHW, JACOB BOUDOUKH 2 , MATTHEW
RICHARDSON 2 and ROBERT F. WHITELAW 2
1 University of North Carolina. Chapel Hill. 2 New York University
ABSTRACT
While there is significant interest in investing in Brady bonds, the source of attraction is
often the exposure to sovereign risk (and its yield compensation), while the exposure to US
interest rate risk is a 'necessary evil'. This paper addresses the problem of determining the
interest rate sensitivity of Brady debt. We show that the most relevant state variables in
determining the duration of a Brady bond are US interest rates and the bond's strip spread.
Motivated by the difficulty of using structural models to price and hedge Brady debt, we
provide a model-free approach to estimating the hedge ratio. Using our approach to hedge
the Argentinian Par and Discount Brady bonds, we find that only a small fraction (15%
or so) of the total risk is hedgeable, but our hedged portfolio exhibits little covariation with
US interest rates.
INTRODUCTION
Brady bonds are sovereign debt issued to replace commercial bank loans made
to developing countries over the past two decades.! This emerging market debt,
which includes debt of Argentina, Brazil and Mexico, among many others, has
several unique features. First, some of the bonds are highly liquid, giving
investors the opportunity to invest in sovereign credit. Second, the majority of
the Brady bonds are denominated in US dollars, and, credit risk aside, are
therefore closely related to fixed and floating rate US government debt. Third,
most of the Brady Bonds include some type of credit enhancement, usually in
the form of the entire principal and some interest payments being collateralized
by zero coupon US Treasury securities.
To the extent that there are ample opportunities to invest in US government
debt, the primary reason for the success of the Brady bond market is the
opportunity to invest in the sovereign debt of emerging markets. Thus, isolating
the sovereign component of Brady bonds by hedging out the US interest rate
risk of these bonds is especially important to market participants (Telljohann,
1 Since 1990, a substantial number of Brady bonds have been issued, with billions of dollars in
principal currently outstanding. Brady bonds get their name from the then Secretary of the Treasury
Nicholas Brady, who emphasized a market-based approach in providing a plan for reducing
emerging market debt.
307
R. Levich (ed.). Emerging Market Capital Flows. 307-317.
1998 Kluwer Academic Publishers.
308 D.-H. Ahn et ai.
1994}. Informal evidence for the appetite investors have for securities which
afford liquid sovereign risk without the US interest rate risk can be found in
the common practice of 'stripping off' Brady bonds (i.e., shorting the guaranteed
component), and the enthusiasm which met the recent plan by the Mexican
government for swapping Brady bonds for purely sovereign debt. This lack of
appetite for US interest rate risk is not surprising since most financial institu-
tions and pension funds consider this risk systematic, while the sovereign risk
is often considered non-systematic. Hence, there is a significant interest in
understanding, quantifying and hedging the interest rate risk of Brady bonds.
In the case of a fixed-rate Brady bond, estimating the bond's interest rate
sensitivity involves knowledge of both the bond's characteristics (e.g., maturity,
coupon and any embedded options) and the level of interest rates. It is impor-
tant to note, however, that even if interest rates are independent of the Brady
bond's default rate, hedging the interest rate risk of the Brady bond requires
additional information regarding default probabilities. This result is true even
in the simple world of a flat term structure and an equal probability of default
each period. Ahn, Boudoukh, Richardson, and Whitelaw, 1997, show that,
under these assumptions, the Macaulay duration of the bond not only changes
as the default probability increases, but that this change may be positive or
negative, depending on the coupon, the interest rate level, and the default
probability.
Given the current values of the interest rate and default probability, one
could dynamically adjust a hedged position by duration matching using, say,
T -note futures, thus, completely isolating the Brady bond from instantaneous
interest rate risk. The problem with this approach is that the assumptions
underlying these theoretical duration measures are generally poor (Cumby and
Evans, 1995; Nadler et ai., 1996). The goal ofthis paper is to develop a different
approach for hedging Brady bonds using interest-rate instruments, such as
T-note futures, following the methodology in Ahn et ai., 1997. The idea is to
estimate a conditional hedge ratio between returns on a Brady bond and
returns on T-note futures. The hedge ratio is conditional in the sense that we
account for relevant current information. This is important for Brady bonds
because, as interest rates and default rates change, the interest rate sensitivity
of Brady bonds will change in a highly nonlinear fashion. In order to estimate
this conditional hedge ratio, a structural model is usually required (as with
most fixed-income valuation approaches). Unfortunately, this requirement
involves making a number of assumptions on the underlying processes which
mayor may not be reasonable.
In this chapter we take an empirical approach to estimating a conditional
hedge ratio for Brady bonds using T-note futures. We take a stand only on
what the relevant state variables are, namely, the level of interest rates and the
strip spread, but not on the precise functional form of their relation to duration.
We implement an out-of-sample experiment, in which we re-estimate the hedge
ratio using a moving window of past data, and compare various methods. We
show that there is some, albeit limited advantage, to the use of conditioning
Hedging the interest rate risk of Bradys 309
Theoretical background
How sensitive are Brady bond returns to interest rate changes? Ahn et al.,
1997, address this issue in a formal setting; however, to gain some intuition,
consider a simple economy in which there is a flat term structure with interest
rate r and a constant probability of default p (with no recovery). A fixed-rate
bond paying a coupon c with underlying principal F has a present value equal
to
where PVt (') equals the present value of the cash flow in period t, adjusted for
the probability of default. Thus, the Macaulay duration is simply a value-
weighted sum of the maturities of the bond's cash flows. Both high interest
rates and a high probability of default tend to substantially lower the present
value of payoffs further in the future, thus reducing the duration of the bond.
In fact, changes in the probability of default have much the same effect as
interest rate changes, so that the standard convexity results follow.
Although Brady bonds face default, part of their cash flows are generally
collateralized by US Treasuries. For example, it is standard to guarantee the
principal, and sometimes the more immediate coupon payments (usually 12-18
months worth), using US Treasury strips. As an illustration, consider a fixed-
rate bond with guaranteed principal. The valuation equation analogous to
equation (1) is
The above discussion considers a fixed rate Brady bond. However, a number
of existing bonds offer floating rates. It is well known that floating rate bonds
are relatively insensitive to interest rate changes, as long as the probability of
default is not correlated with interest rates. Here, the guaranteed component
of Brady bonds dramatically changes this result. As the probability of default
increases, only the duration of the guaranteed component matters; thus, in the
above example, the Macaulay duration is the maturity of the bond's principal.
Therefore, in contrast to the fixed rate bond, the duration of the floating rate
note can vary from six months to thirty years, depending on the likelihood
of a default. Consequently, hedging floating rate Brady bonds may require
substantial positions in the bond market, a vastly different result to that
normally found.
METHODOLOGY
Since the interest rate sensitivity of Brady bonds changes with both the term
structure of interest rates and the term structure of default probabilities, it is
crucial to take these into account when forming the hedged position. The
problem of implementing this hedged position is substantial. Specifically, the
assumptions of a flat term structure and constant default probability are clearly
inconsistent with the data. While models have been developed which generalize
this base case (e.g., Cumby and Evans, 1995), these models involve strong
parametric assumptions. Thus, the results can be difficult to interpret to the
extent that these models are forced to fit the current interest rate and default
probability environment.
An alternative method is to allow the model to have more flexibility and to
take an empirical approach to estimating the hedge ratio. The difficulty is that,
as we have seen, this hedge ratio varies substantially with important economic
variables, such as interest rate levels and default probabilities. Thus, standard
regression-based hedges will not be sufficient. Here, we take a different approach
towards estimating the conditional hedge ratio. Using estimates of conditional
comovements between Brady returns and the hedging instrument, we estimate
the conditional hedge ratio directly. That is, we estimate the conditional relation
between the rate of return on a Brady and the return on, say, a T-note futures,
conditional on relevant information available at any point in time. Suppose
we are given T observations, Zl, Z2, ... , ZT, where each Z, is an m-dimensional
vector. Specifically, let Z, == (R~,+ j ' R[i'+ j, x,), where R~,+ j and R[i'+ j are the
j-day returns on the Brady bond and T-note futures, respectively, and x, is an
(m - 2)-dimensional vector of relevant factors known at time t. Given the
discussion above, two prime candidates for the (m - 2)-dimensional set of
variables are the current level of interest rates and some measure of the
probability of default (such as the strip spread between the Brady bond's yield
on the non-guaranteed portion and US Treasury rates).
Hedging the interest rate risk of Bradys 311
Pt=-
oE B I TN
05
[R"'+J R'.'+J,x,1 _
OR TN . -.,
t,t+ J
then for every $1 of a Brady bond held, the investor should short $0.50 worth
312 D.-H. Ahn et al.
of T-note futures. This hedge ratio will change dynamically, depending on the
current economic state described by X" In our specific example, the hedge ratio
should change in response to changes in the interest rate level and the prob-
ability of default.
EMPIRICAL ANALYSIS
Data description
In this study we employed several data sources over the period July 1992 to
March 1996: (i) Brady bond prices of the Argentinian Par bonds (fixed rate)
and the Argentinian Discount bonds (floating rate) from a major investment
bank in the emerging markets area,2 (ii) strip spreads for both of these bonds
from the same investment bank, and (iii) lO-year T-note futures prices and
various term structure information, including the 3-month, I-year, 5-year and
10-year yields from Datastream.
With respect to the Brady bond data, we collected daily data on two
Argentinian bonds: (i) the dollar denominated, $12.7 billion 30-year Par bond,
with fixed rates building up to 6%, and principal and 12 months coupon
interest guaranteed by US Treasury strips, and (ii) the dollar denominated,
$4.3 billion 30-year Discount bond, with an floating rate of LIBOR plus
0.8125%, and principal and 12 months coupon interest (at 8%) guaranteed by
US Treasuries. Both bonds mature on March 31, 2023. (For more information
see Chase, 1995.)
US interest rates and futures returns are taken from Datastream. Specifically,
the I-year and lO-year rates are daily fixed-maturity series, compiled by the
Federal Reserve Board of Governors. The 10-year T-note futures return series
is a series of nearest to maturity futures, spliced at the last day of the month
prior to the expiration month to avoid liquidity-related effects. The maturity
effect in the futures contracts (which are issued in a 3 month cycle) may cause
a seasonal pattern in the data, although this effect is secondary.3
While the Brady market is extremely liquid, data concerns related to non-
synchronous quotes across the markets can be somewhat alleviated by using
a longer measurement period. Throughout the study we analyzed weekly
returns, which strike a reasonable compromise between measurement issues
and the relatively small size of our data sample. We used overlapping data in
2 These results differ somewhat from Ahn et al., 1997, due to the nature of the strip spread
calculation. Here our primary interest is in hedging a particular fixed and floating rate issue. For
this application, the data from the investment bank is of high quality.
3 A simple way to see this is by considering the corresponding forward contract, which is,
essentially, a long position in a long bond financed by a short position in a short bond (for a long
futures position). The first order effect on changes in this forward/futures price will be due to
changes in the long rate due to its much higher duration.
Hedging the interest rate risk of Bradys 313
Di:.,+s Rf.'+5 i: ./
Rr.~~s ,
spar
R~~s+s ,
Sdis
"
Mean -0.004 0.033 4.824 6.495 0.227 8.780 0.122 8.733
Std. dev. 0.126 0.842 1.230 0.740 3.386 3.103 3.430 3.207
Di!,t+s 1.000 -0.943 -0.094 -0.082 -0.493 -0.240 -0.235 -0.238
Rf.t+5 -0.943 1.000 0.093 0.082 0.461 0.225 0.222 0.225
s
-0.094 0.093 1.000 0.705 -0.051 0.523 -0.081 0.603
"./ -0.082 0.082 0.705 1.000 -0.100 0.238 -0.127 0.276
"
Rf.~~s -0.493 0.461 -0.051 -0.100 1.000 0.210 0.830 0.198
,
spar
-0.240 0.225 0.523 0.238 0.210 1.000 0.121 0.983
R~~s+5 -0.235 0.222 -0.081 -0.127 0.830 0.121 1.000 0.147
,
Sdis -0.238 0.225 0.603 0.276 0.198 0.983 0.147 1.000
Summary statistics are provided regarding the Argentinian Par and Discount Brady bonds, and
relevant state variables. Return data are for weekly changes, overlapping daily. The sample period
is 10/15/1992 to 2/22/1996. The variables are: Di:.'+5 the weekly change in the 10-year US par
i: i:
yield, Rf.,+, the return on 10-year T-note futures, the I-year US par yield, the lO-year US par
yield, Rf.~~5 the return on the Brady Par bond, sF" the Brady Par bond strip spread, R~;'+s the
return on the Brady Discount bond, s~;' the Brady Discount bond strip spread.
order to use all available information, and where necessary, we adjusted for
the overlap.
Table 1 provides basic summary statistics for our dataset. During the sample
period (October 15 1992 to February 22 1996), there was an average weekly
gain in both the Par and the Discount bonds of 0.23% and 0.12%, respectively.
Some of this return can be attributed to the decline in US interest rates, and
the rest to an improvement in the perceived credit worthiness of Argentina's
debt.
The standard deviations of returns on both bonds are approximately 3.4%
per week, about four times larger than the volatility of the corresponding
T-note futures contract (with a return standard deviation of 0.84% per week).
These standard deviations give us a preview of the results to follow, namely,
that most of the variation in Brady bond prices will not be explained by
variation in US Government bond returns or related derivatives. To see this,
we simply need to recognize that the typical asset underlying a T-note futures
contract is a par coupon-bearing US government bond, with cash flows which
are fairly similar to the promised cash flows of the Argentinian Par bond.
However, this Brady bond is much more volatile, and its correlation with the
T-note futures contract is only 0.46. At the same time, the Argentinian Par
bond is as volatile as the Argentinian Discount bond, and the correlation
between the two Brady bonds is 0.83. This immediately indicates that sovereign
risk is causing most of the variation, not dollar interest rate risk.
The volatility of Brady bond returns over our sample period can be largely
attributed to the events that took place at the end of 1994. The collapse of the
Mexican peso, and with it the decline in the perceived credit worthiness of
Mexican sovereign and Brady debt, had large spillover effects throughout South
314 D.-H. Ahn et al.
CD
N
...
N
Itlllllliltli Ihi
............. 110
~ -- .Cp..
- .Cparl",
lIo
.CdlelN
~ ........Cdlello
I
UJ
!!!
!d\.
~ ~
:g
~
CD ." ~."" ~"
.,............. '-0 ,", ,
\...A........- .................... ..:.~ .....~.~t'
...
o~ ____ ~ ______ ~ ____ ~ ______ ~ ____ ~ ______ ~ ____ ~ ____ ~
Figure 1. The strip spreads on Argentinian Par and Discount Brady bonds and the to-year US
Government par yield. The thick lines denote that the series is above its sample mean.
America. Specifically, during the half year surrounding the collapse of the
Mexican Peso (most of which occurred during last weeks of 1994 and January
1995), the credibility of the Argentinian Peso's peg to the dollar became
doubtful. The resultant impact on Argentinian Brady bonds was due to a
perceived increase in the default probability. This is apparent in Figure 1, which
depicts the path of the strip spreads of the Argentinian Par and Discount
bonds. 4 The strip spread rose from about 6% prior to the event, to an average
level of over 12%, with strip spreads as high as 20% or more at times.
Out-oj-sample hedging
4The strip spread is obtained in the following manner. Using a zero curve imputed from prices
of US government bonds, the guaranteed payments of a given Brady bond are stripped off. The
strip spread is then the difference between the default-free yield on the promised payments, and
the defaultable yield given the value of the stripped bond. Some subtle issues arise with respect to
the proper calculation given the rolling guarantee, which gives rise to some differences in strip
spread calculations across methods. Irrespective of the method of calculation, however, the
strip spread is the single best summary statistic for the market-perceived average default probability
throughout the remaining life of the Brady bond.
Hedging the interest rate risk of Bradys 315
par bond, and the Brady's time to maturity. We did so repeatedly using the
250-day window until the last day in our sample.
We experimented with a number of specifications for the functional form
and the relevant conditioning variables in estimating the conditional hedge
ratio, Pt. Our benchmark is a state-independent (although not time invariant)
hedge ratio, which is estimated repeatedly but without any conditioning on the
relevant state variables. We then assume that the function Pt = g(it> St> r) is a
linear function of these variables, and consider hedging first using it only, then
adding St to the set of conditioning information, and last, the time to maturity
r. We also allowed for non-linearities by considering a second order Taylor
expansion on the most interesting set of conditioning information, namely it
and St. This expansion involves the addition of squared and interaction term
of these variables (three additional variables in total) to the original set of two
conditioning variables.
Table 2 documents results for the Par and Discount Brady bonds respec-
tively. For each of the hedging procedures considered we provide two summary
statistics. First we document the volatility of the returns on a hedged portfolio
which involves a long position in the relevant Brady bond, hedged by the
appropriate short position in lO-year T-note futures. The goal is to minimize
the volatility of hedged returns. The second statistic is the correlation coefficient
between hedged returns and the contemporaneous change in the 10-year rate.
Since our experiment is conducted as an out-of-sample experiment, we are not
guaranteed orthogonality between the hedging errors and interest rate changes.
As pointed out in the introduction, one might consider the ability to hedge a
specific source of risk, interest rate risk in our case, important.
With respect to hedging the Par bond, the total unhedged standard deviation
of returns was 3.68% per week, and the correlation between returns and interest
Table 2. Hedging results for the Par and Discount Brady bonds respectively, using 10-year T-note
futures contracts
Conditioning information
f f
f
't,S,
f
't,Sf 't, S" T
Unhedged None "linear linear Taylor linear
Out-of-sample hedging errors are analyzed for the hedge regression R~'+5 = ex + P,R[{':.s + '.'+5'
At any given date starting the 251st date, the previous 250 daily observations are used for
estimation. The hedge ratio p, is allowed to depend linearly or nonlinearly on conditioning
information. Two summary statistics are recorded: (i) hedged return volatility, and (ii) ex post
correlation of the hedging error with changes in the 10-year par yield.
316 D.-H. Ahn et al.
rate changes was - 0.56. These numbers correspond closely, although not
exactly, to the numbers in Table 1, due to the 'missing' 250 days upfront in the
results of this section. The standard error using the state independent hedging
method was reduced to 3.09%, and the correlation reduced to -0.10. With
respect to the state-dependent hedge ratio results, the results overall can be
considered disappointing. Using the to-year rate as a single conditioning vari-
able yields a standard deviation of 3.14% and the correlation of hedging errors
with interest rate changes is -0.01. On the positive side, conditioning on the
level of interest rates reduces whatever correlation with interest rate changes
was left over from the unconditional hedge. This comes at the cost of increasing
the variability of hedged returns.
Adding the spread to the set of conditioning information should, according
to our theory, help to pin down the appropriate duration/hedge ratio. Indeed
this bivariate hedge does yield the lowest return volatility (3.087%), but the
improvement is hardly significant from an economic perspective. A Taylor
expansion of the two-variable conditioning set provides no benefit, and, in fact,
increases the return volatility (3.28%). This latter result is simply an artifact
of estimation error, exacerbated by the highly volatile spread.
The results for the Discount bond were worse. There was very little, if any,
reduction in the return volatility, and the only reduction was in the correlation
coefficient between hedged returns and interest rate changes. This outcome
would not be too surprising if the default probability was very low, since a
default-free floater should exhibit very little volatility. However, the magnitude
and variation of the strip spread suggest that conditioning on this variable
should produce a viable hedge. Consequently, the empirical results are
disa ppoin ting.
The results for both the Par and the Discount bonds are qualitatively robust
to (i) the length of conditioning period, (ii) the hedging horizon, (iii) the specific
subperiods, and (iv) the addition of the I-year US interest rate to the set of
conditioning variables. We also experimented with smoothing methods, which
allow the state-dependent hedge ratio to vary more 'sluggishly'. Specifically,
we considered an exponentially smoothed hedge ratio, with various smoothing
parameters. Another refinement which we considered was the use of Stein
estimators. 5 The results were not remarkably different for either of these
attempted improvements.
CONCLUSION
Our empirical results and their weak link to the theory illustrate the difficulty
in hedging Brady bonds. From an asset pricing perspective the results pose a
5 Stein estimators are also known as 'shrinkage' estimators. They adjust for the signal to noise
ratio by appropriately shrinking the hedge ratio. Stein estimators are often used by practitioners
in the context of fixed income securities hedging and portfolio analysis (for a technical discussion
see Judge et al., 1985).
Hedging the interest rate risk of Bradys 317
REFERENCES
Ahn, D.-H., J. Boudoukh, M. Richardson and R.F. Whitelaw (1997). The Interest Rate Risk of Brady
Bonds. Working paper, University of North Carolina, Chapel Hill and New York University.
Chase Manhattan (1995). The Emerging Markets Handbook.
Cumby, R. and M. Evans (1995). The Term Structure of Credit Risk: Estimates and Specification
Tests. Stem School of Business, Economics Department working paper EC-95-14.
Judge, G., W.E. Griffiths, C. Hill, H. Lutkepohl and T.-C. Lee (1985). The Theory and Practice of
Econometrics. New York: John Wiley and Sons.
Nadler, D., P. Tsoucas and C. Wierzynski (1996). Brady Bond Valuation and the Universe of Credit
Term Structures. Goldman Sachs Fixed Income Research Series.
Telljohann, K. (1994). Quantifying and Isolating the US Interest Rate Component of a Brady Par
Bond. Chicago Board of Trade working paper.
IAN DOMOWITZl, JACK GLEN 2 and ANANTH MADHAVAN 3
1 Northwestern University 2 International Finance Corporation 3 University of Southern California
INTRODUCTION
With hindsight, the events following the collapse of the Mexican peso in 1994
suggest a major misjudgement of the potential reaction of foreign investors to
a devaluation. This view is clearly ennunciated by Lustig (1995), who also
notes that it is difficult to accept that investors were not aware of the exchange
rate risk in a country with a high current account deficit and low level of
domestic savings. Although there has been extensive factual investigation of
circumstances leading up to and beyond the devaluation in the form of event
chronologies and macroeconomic statistics, little evidence has been presented
with respect to risk and expectations issues.
In Mexico, as elsewhere, interest rates on short-term debt issued by the
government can provide evidence on the beliefs that investors hold with respect
to risk. We examined the interest rates paid to investors by the Mexican
government for short-term debt instruments over the 1993-1994 time frame.
The nature of the data allowed us to decompose risk premia demanded by
investors into two components. The first is based on local dollar-denominated
debt, which we term a country risk premium. This premium is compensation
for the risk that the Mexican government might default on its obligations either
to repay the debt or to allow movement of capital outside the country. The
second is based on both dollar-denominated and peso-denominated debt, rep-
resenting compensation for the risk associated with movements in the exchange
rate, a currency risk premium.
Studies relating to either of these concepts are certainly not new. Work
relating to country premia is manifested in analysis of the spread between
dollar-denominated rates and LIBOR that countries are charged for loans
(Edwards, 1986). Foreign exchange risk premia have been extensively examined;
a recent survey is provided by Lewis (1995). Most work has concentrated
directly on exchange rates and on bank deposits or short-term securities issued
by different entities and subject to different default risk.
By employing only instruments issued by the Mexican government, we
avoided confusing diverse default risks with country or currency risk. Following
Frankel and Okongwu (1996) and Domowitz et al. (1996), we derived country
and currency risk premia directly from information on Mexican sovereign debt,
and analyzed the behavior of those premia over time. Particular attention was
319
R. Levich (ed.), Emerging Market Capital Flows, 319-334.
Ii:> 1998 Kluwer Academic Publishers.
320 I. Domowitz et al.
paid to the behavior of the premia and their underlying instruments around
unusual events in the Mexican political and economic climate, including the
Chiapas uprising, the Colosio assassination, and the enormous increase in the
supply of dollar-denominated debt in the last three-quarters of 1994. With
respect to the devaluation, we found no evidence, based on interest rate move-
ments, the structure of risk premia, or the growth in dollar-denominated debt,
that investors anticipated the devaluation and the crisis that ensued.
Our approach to the expectations issue differed from that of Frankel and
Okongwu (1996), who employed survey data on exchange rate depreciation.
Viewing the risk premia as synthetic tradeable instruments, we applied the
perfect foresight version of the expectations hypothesis (Campbell and Shiller,
1991). We documented the time series behavior of the residuals of that model,
which represent forecast errors under the expectations hypothesis, once again
concentrating on the Mexican political and economic changes over the period.
There is clear evidence from this analysis that investors did not anticipate the
events of December, 1994. On a qualitative basis, reaction to the devaluation
in terms of the forecast errors mirrors that to other unanticipated events, such
as the uprising and assassination.
I The Mexican government also issues ajustabonos, which are indexed to Mexican inflation, but
these were not as popular as cetes during our same period.
Country and currency risk premia 321
and
2 Tesobonos are a reincarnation of a previous instrument, pagafes, which the switchover between
the two instruments taking place in 1991 when exchange controls were eliminated and the controlled
and market exchange rates were unified. For an examination of the links between the cetes and
pagafes markets see Khor and Rojas-Suarez (1991).
3Capital controls were imposed in 1982 following the debt crisis and subsequent devaluation.
See Melvin and Schlagenhauf (1985) for a description of the effect those controls had on eurodollar
interest rates paid by Mexican borrowers. Following the December 1994 devaluation, the govern-
ment offered investors a choice between payment in dollars and payment in indexed pesos in an
attempt to reduce investor concerns over repayment.
322 I. Domowitz et al.
We assume that the default risk premium, y~" is the same for both instru-
ments because they are issued by the same entity. One could argue that the
default risk on cetes is lower than for tesobonos because they are peso-denomi-
nated and the government can print an unlimited amount of pesos for
repayment purposes. In fact, Lustig (1995) identified the dollarization of
Mexican short-term debt as an important contributing factor to the peso crisis
in 1994, on the grounds that "it vastly increased the risk of default." However,
tesobonos are, in effect, also peso-denominated, albeit with a link to the
exchange rate. Printing pesos should have an eventual impact on the exchange
rate, but given a willingness to accept high domestic inflation, repayment of
tesobonos would be possible for any reasonable level of indebtedness. From
another perspective, the real default risk inherent in these two instruments is
more political than economic. Even under very severe economic circumstances,
repayment would be possible if the political willpower existed. Default decisions
are more likely to be political than economic and with each of the two
instruments being held by both foreign and domestic investors, any decision
to default would likely apply to both instruments.
In an international context, the currency risk premium, y~" can be further
decomposed into two separate factors: one for expected depreciation of the
peso, and one for unexpected currency movements. The first of the two has
been the subject of some debate because of its expectational nature; Frankel
and Okongwu (1996) discussed three different approaches to estimating this
premium and examine the premium in the Mexican context. The second com-
ponent, commonly called the foreign exchange risk premium, has been studied
extensively, albeit primarily for developed countries. Lewis (1995) reviewed
much of the relevant literature, which concentrates on the foreign exchange
risk premium by subtracting the forward exchange rate from Y~,.
From a purely domestic perspective, however, an alternative decomposition
of the currency risk premium would include a term for expected peso inflation,
with another term for the risk of unexpected peso inflation. To the extent that
purchasing power parity holds, then these two alternative decompositions yield
equivalent results, but to the extent that there is variation in the real exchange
rate, the different investor groups can have divergent opinions on the appro-
priate value of Y~,. The domestic decomposition views interest rates as being
determined by domestic investors who are concerned with the domestic
purchasing power of the peso. In the international decomposition, investors
are concerned with the international purchasing power of the peso.
Unfortunately, we know of no theoretical model that permits one to differentiate
empirically between these two potentially different perspectives. Frankel and
Okongwu (1996) sidestep the issue by concentrating exclusively on the inter-
national decomposition in their analysis, which is equivalent to assuming that
purchasing power parity holds, at least on average in an expectational sense,
over the sample period. Our analysis also evaded the issues by looking only
at the value of Y~, and ignoring the decomposition.
Country and currency risk premia 323
Sample statistics for the 91- and 182-day cetes, tesobonos, currency premia and
country premia are presented in Table 1. Observations correspond to weekly
auction prices for the period July 1993 to the end of November 1994. Some of
the discussion that follows will also include observations from the period that
followed the December 1994 devaluation. Those observations are excluded
from the sample statistics, however, because they are markedly different from
the remainder of the sample period. Separate sample statistics are reported for
each of the 2 years in order to give an idea of the effect of the sample period
Table 1. Summary statistics on traded instruments and derived risk premia in the Mexican debt
market for 91-day and 182-day maturities
Standard
Series Minimum Maximum Mean Median deviation
Cetes (91-day)
1993 0.107 0.159 0.136 0.139 0.012
1994 0.091 0.180 0.142 0.145 0.027
Cetes (182-day)
1993 0.106 0.154 0.134 0.137 0.013
1994 0.099 0.176 0.140 0.143 0.022
Tesobonos (91-day)
1993 0.047 0.055 0.051 0.051 0.002
1994 0.040 0.082 0.066 0.068 0.010
Tesobonos (182-day)
1993 0.049 0.059 0.053 0.052 0.003
1994 0.048 0.088 0.071 0.074 0.011
Peso premium (91-day)
1993 0.056 0.104 0.085 0.089 0.012
1994 0.033 0.112 0.076 0.075 0.020
Peso premium (182-day)
1993 0.047 0.095 0.080 0.084 0.013
1994 0.042 0.105 0.069 0.065 0.015
Country premium (91-day)
1993 0.005 0.040 0.022 0.020 0.007
1994 0.005 0.040 0.023 0.024 0.008
Country premium (182-day)
1993 0.014 0.049 0.026 0.024 0.007
1994 0.014 0.050 0.028 0.027 0.008
Data for cetes (governments securities denominated in pesos) and tesobonos (government securities
denominated in dollars, payable in pesos as the official exchange rate) are annualized effective
yields calculated from the Mexico government weekly primary auctions, from the beginning of
July 1993, through the end of November 1994. The currency risk premium ('peso premium') is
calculated as the arithmetic difference between cetes and tesobonos yields. The country risk premium
('country premium) is calculated as the difference between the tesobonos yield and the yield of a
3-month US Treasury security. Data reported are the minimum and maximum yields and premia,
the mean and median yields and premia, and the standard deviation of the yields and premia all
in percentage/l00.
324 I. Domowitz et al.
on the results, something that might be important given the political and
economic events that took place in Mexico in 1994. The instruments are issued
with maturities ranging from 7 to 360 days, but not all maturities are issued
every week. For the sample used in this study, 91 and 182 day were the most
common maturities and the analysis is restricted to those two maturities.
As might be expected given the currencies involved, the cetes rates were
substantially larger than the tesobonos rates on average. They were also much
more volatile, although the volatility of both instruments increased markedly
in 1994 relative to 1993. Tesobonos rates rose sharply in 1994 relative to 1993,
with an increase of roughly 34% at the longer maturity and just over 29% at
the shorter maturity. Cetes rates also increased in 1994, but by a much smaller
amount than the tesobonos and without any substantive difference across
maturities. The average growth rate was approximately 4.5% for both
maturities.
Average cetes rates and their volatilities were roughly double their tesobonos
counterparts in 1994, a reflection of the risk that investors associated with peso
inflation and the exchange rate. Given the differences in level and volatility of
rates, it is useful to consider comparisons based on the ratio of mean return
to volatility, however. For example, the 1993 91-day cetes ratio was 11.3,
compared with 25.5 for the tesobonos, in sharp contrast to the simple compari-
sons of yields. Further, the ratios of return relative to risk were virtually
equalized for the 91-day instruments in 1994, with values of 5.3 and 6.6 for
cetes and tesobonos, respectively. Similar results hold for the 182-day maturities,
with 1994 return-risk ratios of approximately 6.5 for both instruments. It is
notable that while rates were rising across years, yields relative to volatility fell
from 38% to 74% across instruments and maturities, with an average drop
of 57%.
Figure 1 displays the time series behavior of the 91-day cetes and tesobonos
yields for the period July 1993-February 1995. The figure confirms the basic
story offered by the sample statistics, but provides additional insight as well.
Of particular note is the behavior of the yields subsequent to the peso devalu-
ation that occurred on December 21, 1994, with no obvious increase in yields
since about April prior to that event. This suggests the success that the Mexican
government had in convincing the market that no devaluation was forthcoming.
In fact, the Bank of Mexico followed a policy of sterilized intervention during
the March-April period, and again in November and December. The decline
in foreign exchange reserves was offset by an increase in the Bank's net domestic
assets. As a result, relatively low interest rates were maintained early in the
year and later, during a period of selling pressure on the peso. Although there
were other signals of potential internal imbalances, there also were positive
moves, including a reduction in inflation and the maintenance of balanced
government fiscal accounts (International Monetary Fund, 1995).
Less obvious, perhaps, is the behavior of the rates at the time of several
other notable events during the sample period. In particular, January 1, 1994
was the date on which the Chiapas rebels occupied that state: no apparent
Country and currency risk premia 325
0.45
0.4
0.35
0.3
g
.... 0.25
~
0.05
0
.....
N
"~
0)
change in either yield followed the occupation. On March 23, 1994, (the ruling
party) PRI presidential candidate Colosio was assassinated in Tijuana. The
yield on both debt instruments moved up at that time, but the increase in the
tesobonos rate actually preceded the assassination, whereas the increase in the
cetes yield followed immediately after that important political event. Finally,
Ernesto Zedillo, the candidate of the ruling party, won the presidential election
on August 21. The cetes rate moved down in anticipation of that event and
stayed low until the devaluation; no obvious change in the tesobonos rate
occurred during the period leading up to or after the election.
Table 1 also contains summary statistics for the two risk premia. The peso
premia were roughly four times the level of their country premia counterparts
for both maturities and both years; they were also about twice as volatile.
Volatility increased for both premia in 1994 relative to 1993, but whereas the
average country premium increased in 1994, average peso premia actually
declined for both the shorter- and longer-term bills.
Figure 2 contains time series graphs for the 91-day premia. With the excep-
tion of March 1994 and a brief period in early 1995, the currency premium
greatly exceeded the country premium. The currency premium declined substan-
tially over the 8 months that preceded the March 1994 assassination of presiden-
tial candidate Luis Colosio. In contrast, the country premium remained flat
over most of this time, falling only shortly before the assassination. Both premia
increased at the time of the assassination, although, like the tesobonos rate on
which it is based, the increase in the country premium took place prior to the
assassination and the increase in the currency premium was much larger and
followed the assassination. After the assassination, the country premium
declined until, by the time of the August 1994 presidential election, it had
reverted to its pre-assassination level. The increase in the currency premium
was more persistent, and never quite fell back to its previous low.
326 I. Domowirz er al.
0.25 r------------------------,
g
02
0.15
1==::1 !~\~. ji
,.
;; ; \
e
c
&
-./""
N ...
cc
...... ... ...
<"l 0 cc N N en .... N 0
....... ... ...... ...en... ....... ......cc
N
It) It)
Also notable in the graph is the effect of the December 1994 devaluation
which led to dramatic increases in both premia, but with the increase being
more spectacular, and more short-lived, for the country risk premium. In fact,
investors retained their holdings of resobonos immediately following the devalu-
ation (International Monetary Fund, 1995). In particular, foreign holdings
increased about 5.5% in the week following the devaluation, and did not
decline until the end of January, 1995.
The term structure for each of these bills is expressed using the following
notation:
(3)
(4)
Employing equations (1) and (2), the following term structure variables are
derived for the two risk premia:
(5)
(6)
Equation (5) indicates that the term structure of the peso premium can be
measured using only the observable rates of return from the ceres and tesobonos.
In equation (6), the term structure of the Mexican country risk premium
Country and currency risk premia 327
0.02
0.015
0.01
0.005
o
-0.005 1-~~~.,_\/_..!._,____:_<=_L-V_-4.:._--7""!_'r__~-'I----"-..c..v_,----t
-001 . ... I ,I
. \... I .. 1 "1
-0.015 ii,
" 1
-0.02 , '
-0.025 1/
-0.03 ~...........,....,...~.,....,...,......,..,..,..,...,........,..,..,..,...,..............,....,...,......,..,~............,...,...,...........~,......,..,~.,...,..,..,...........,...,..,..~
E ~
o ~
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
~ ~ ~ ~ 0 ~ g g ~ 0 ~ ~ ~ 0
;~ ~ ~ ~
~ ~ ~
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ m
Figure 3. Cetes and Tesobonos term premia (182 day-91 day).
4 One alternative would be to employ US T bills as proxies for risk-free rates, but it is difficult
to believe that 6-month T bills are viewed as truly risk-free by the investing public because of the
inflation risk that they contain. As a result, the term structure of T bill rates would contribute to
our country risk measure and could actually introduce more noise than it removes. Moreover, as
shown below, under the expectations hypothesis the term premium should be constant over time.
328 I. Domowitz et al.
The downward sloping cetes term structure revealed itself again in the
currency/peso premium term structure, which was downward sloping, on
average, in both 1993 and 1994. Like the tesobonos, however, the country risk
premium term structure retained its positive slope in both periods. Volatility
of the peso premium was substantially higher in 1994 than in 1993, whereas
volatility was roughly constant across the two years for the country risk
premium.
Figure 4 contains time series graphs for the term structure of the country
and currency risk premia, defined as the spread between the value of the premia
at 91 and 182 days. The country term premium behaves exactly like its related
tesobonos term premium, which is to be expected under our working assumption
that there is no difference between the short and long-term risk free rate.
Differences between the term structure of the cetes and the currency risk
premium, however, are notable. Even though the correlation between the cetes
and peso risk term premium was 0.86, there were times, especially during the
period just prior to the 1994 devaluation, when the cetes term structure was
upward sloping while the currency term structure was downward sloping.
0.03
0.02
0.01
0 0
....
. ...
0
'. ...
.. . .. ..
";:J
c: -0.01 ' "" ,. \:'
~
' "
" ,
-0.02
1=:r:1
Q.
-0.03
-0.04 ...
-0.05
.... It) 01 C') co ....
N
....
<0 C') ..... .... <0 0
.... 001 N NNco
.... 01
~ C')
.... 0....
N N N 0 0 N N 0 0
..... co 01 N N C') ~ <0 ..... co .... .... ....
0 0 0 0 0 0 0 0 0 0
C') C') C') C;; C')
01 01 01 01 01 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
rate is a function of the levels of current and expected future shorter-term rates.
In the most simple case where the long instrument has a maturity exactly equal
to the sum of two short instruments, the relationship takes the following form:
6r---------------------------------------------------,
5
4
;-
3
"
"
'. . ..... .. --
0 "
-1
-2
.. .., .., ..v; .., ..,
... g ..,~
II I ~ "~ i8 iS
cD
N
It)
I Si Si
('oj It) It) cD cD ('oj
0 ~ ~ N 00 0 ('oj
~ ~ ;::
~ ~
0> ~ Si Si ~ ~ ~ 0
~
0
~ I 0
~
0
~ ~ ~
the values which would have obtained with perfect foresight. This provides
clear evidence that the expectations that investors held prior to the devaluation
did not anticipate the large increase in the short-term risk premia that followed
the devaluation.
Closer examination reveals that the period just preceding the devaluation
was somewhat unique with respect to the similarity of the two series. For most
of the remainder of the sample period, their behavior was much different. Two
subperiods merit additional comment because they provide information on the
factors that influence the behavior of the markets. First consider the behavior
of the currency premium residuals over the period January-May 1994. During
that period, the long premium was underpriced, indicating that investors under-
estimated the future short-term rates that were subsequently realized. What
happened is largely a result of a single important political event. On March
23 the ruling party presidential candidate, Colosio, was assassinated and short-
term (and long-term) interest rates on cetes increased sharply, where they stayed
until August. This upturn in rates (and the currency risk premium) led to a
substantial error in the expected value ofthe future short-term rate as measured
by the perfect-foresight model.
Consider now the behavior of the country risk residuals over the period
January-August 1994. In order to understand the behavior of these residuals,
one must also consider Figure 6, which presents the outstanding amounts of
cetes and tesobonos issued by the Mexican government, in millions of new
pesos.6 In the figure, one can readily see that this was a period during which
the outstanding supply of tesobonos increased sharply. Relative to 1994 overall,
and the January-August period in particular, growth in tesobonos supply started
100000
90000
80000
70000
60000
50000
40000
30000
20000
_ _ Celes
10000 ....... Tesebonos
....... - ..
0
~ ~ ~ ~
....co ..
;J;
.:.
::E
!
OJ
::E
1 ;J;
.,Q.
U)
;J;
:>
z0
~..,
OJ
~
.:.
~
CONCLUDING REMARKS
7 For the Nafta ratification, the calculations use the one-month period prior to ratification, but
including the week of the ratification. For the Chiapas uprising and Colosio assassination, the
mean-squared error is produced using data four weeks prior to the events, but including the week
of the event. The full-sample mean-squared error uses all data from 7/93 through 8/94.
Country and currency risk premia 333
short rates. We have used the model in a rather unorthodox fashion, namely
to model the market's reactions to differences in risk premia over different
horizons. The philosophy is that since these premia can be generated from the
returns on market-traded instruments, the term structure of the premia should
be explainable by the same economic models of the term structure that are
used for other fixed-income securities. In fact, since the premia can be extracted
directly from the data, they themselves could be traded, in principle. In other
words, they are tradeable synthetic instruments. The model residuals provide
considerable insight into the extent to which the markets anticipated events
and their reactions to those events. The analysis indicates that there was no
anticipation of the major political developments, as would be expected. It
characterizes, however, the markets' reactions to the supply shocks that took
place over the sample period. Finally, the residual analysis provides additional
evidence that there was no expectation of the devaluation that subsequently
took place.
Was there any other observable indicator of the pending devaluation in
terms of market activity? Lustig (1995) suggested that the answer to this
question is in the affirmative. In particular, she writes that "the systematic
increase in tesobonos held by the public ought to have been seen as an unequivo-
cal sign of the lack of credibility of the exchange rate policy." We disagree, at
least insofar as the claim implies that the markets should have recognized and
acted on this phenomenon. Beyond any evidence presented in terms of our risk
premia constructs, the argument is based on a simple risk-return tradeoff.
The supply of tesobonos, both relative to that to cetes and in absolute terms,
was very limited in 1993. During the last half of that year, the ratio of yield to
risk, measured by the standard deviation of returns, was over 25 for tesobonos
and about 11 for cetes. It is intuitively obvious that tesobonos represented a
good value, relative to risk, and compared to the peso-denominated instrument.
These ratios were not substantially different in early 1994. As the supply of
tesobonos deepened, investors were happy to buy them for reasons unrelated
to the credibility of exchange rate policy. Their holdings of cetes were already
large, and tesobonos, representing a different sort of risk, allowed for portfolio
diversification. More importantly, perhaps, the return-risk tradeoff was such
as to make the tesobonos the more desirable instrument, on a risk-adjusted
basis. Abstracting from exchange rate risk and the diversification issue, one
would expect that investors would buy tesobonos until the return-risk tradeoff
was equalized between instruments. That is exactly what happened. Prior to
the devaluation in 1994, the tesobonos ratio for the longer maturity was 6.5,
compared with 6.4 for cetes. If exchange rate credibility were really the problem,
we also might reasonably expect that the ratio should be larger for the cetes;
i.e. investors would demand more return relative to volatility for the peso-
denominated notes. This phenomenon is not observed in the data. We are left
with the conclusion that despite the difficulty in accepting the fact that sophisti-
cated investors were not aware of the increasing exchange rate risk as develop-
334 1. Domowitz et al.
ments unfolded in 1994, their actions, as expressed in market statistics and risk
measures, indicate that the devaluation was rather a complete surprise.
ACKNOWLEDGMENTS
Funding from the World Bank is greatly appreciated. Expert research assistance
was provided by Mark Coppejans. We are grateful to Michael Pettis for his
help in obtaining the interest rate data and to Robert Hodrick for helpful
discussions.
REFERENCES
Campbell, J.Y. and R. J. Shiller (1991). Yield spreads and interest rate movements: a bird's eye view.
Review of Economic Studies, 58, 495-514.
Cox, J.e., J.E. Ingersoll and SA Ross (1981). A reexamination of traditional hypotheses about the
term structure of interest rates. Journal of Finance, 36, 769-799.
Domowitz, I., J. Glen and A. Madhavan (1996). Identification and Testing of a Term Structure
Relationship for Country and Currency Risk Premia in an Emerging Market. Working paper,
University of Southern California.
Edwards, S. (1986). The pricing of bonds and bank loans in international markets: an empirical
analysis of developing countries' foreign borrowing. European Economic Review, 30,565-589.
Frankel, J.A. and C. Okongwu (1996). Liberalized portfolio capital inflows in emerging markets:
sterilization, expectations, and the incompleteness of interest rate convergence. International
Journal of Finance and Economies, I, 1-28.
International Monetary Fund (1995). Evolution of the Mexican Peso Crisis. Background paper,
International Capital Markets Division.
Khor, H.E. and L. Rojas-Suarez (1991). Interest rates in Mexico. IMF Staff Papers, 38, 850-870.
Lewis, K. (1995). Puzzles in international financial markets. In R.W. Jones and P.B. Kenen (eds),
Handbook of International Economics, Volume 3. Amsterdam: North-Holland.
Lustig, N. (1995). The Mexican Peso Crisis: The Foreseeable and the Surprise. Brookings Institution
Discussion Paper No. 114.
Melvin, M. and D. Schlagenhauf (1985). A country risk index: econometric formulation and an
application to Mexico. Economic Inquiry, 22, 601-619.
ROBERT J. BERNSTEIN and JOHN A. PENICOOK Jr
Brinson Partners. Inc.
INTRODUCTION
Emerging market debt (principally the Brady bond market) warrants consider-
ation in diversified portfolios based upon its normal return potential, risk
characteristics and portfolio diversification benefits. While a realistic review of
sovereign credit history discloses occasional periods of extreme duress, there
are several reasons to be optimistic regarding future credit performance.
Foremost is that the Brady plan provided a market response to the debt crisis
of the 1980s, and the resulting bond market, in turn, provides constant market
feedback on fiscal, monetary and exchange rate policies. Market feedback,
combined with a general trend towards more participative government, bodes
well for the overall economic development process.
Fundamental investment analysis of this market requires an understanding
of sovereign credit risk and the compositional complexities of the Brady bonds
themselves. The normal return potential of the market, in conjunction with its
low correlation to other bond and equity markets, offers the opportunity to
improve a portfolio's risk/reward profile. The remainder of this chapter reviews
emerging market debt generally, examines the Brady market specifically and
provides some perspective on the long-term and current attractiveness of this
asset class.
Terminology
In the broadest sense, the group of 'emerging' countries includes all nations
not considered industrialized or already developed. Since the latter group has
only two dozen or so members, the emerging country universe encompasses
most of the world's population and geography. However, because the majority
offer no investable debt securities, only a subset of these countries comprises
the emerging country debt universe. Hence, the more precise terminology is
emerging market, rather than emerging country.
335
R. Levich (ed.), Emerging Market Capital Flows, 335-370.
o 1998 Kluwer Academic Publishers.
336 R. J. Bernstein and J. A. Penicook Jr
While convention and market terminology lump all of these countries into
one market there are profound and fundamental differences among them. Many
Latin American countries have a poor history of macro-economic management
and suffer from deep social inequality, but their recent economic performances
have largely improved. Eastern Europe is recovering from decades of central
planning, but some countries have pre-war histories of success with capitalism
and opening them up to the rest of the world has the potential of producing
outsized growth rates. Africa is generally income-poor, but commodity-rich.
Finally, a number of Southeastern Asian countries have very high savings rates,
resulting in their exporting rather than importing of capital.
Economic significance
Emerging country economic growth rates have far exceeded those of the devel-
oped world recently. Real GDP growth for the major emerging markets has
averaged well in excess of 5% annually over the past decade, versus 2.5% for
the industrialized economies. Although their share of world output and exports
is still relatively small, recent economic reforms, global trade liberalization and
access to the global capital markets have allowed developing economies to
more efficiently utilize their vast natural endowments and to partially realize
their economic potential (see Figure 1). Continued trade integration and a
relatively stable global economic background suggest that developing econo-
mies as a group will continue to grow faster than industrialized nations.
Market growth
Along with the size of their economies, the importance of developing countries'
debt securities in the international marketplace has grown as well. While still
small in comparison to that of industrialized country debt, the overall size of
the market (see Figure 2) has grown to become an important sector of the
global capital markets. Over the last four years alone, total trading volume of
emerging market debt securities has increased by six hundred percent; about
half of this volume was in the Brady bond market (Figure 3).
There are four segments of the emerging debt market: bank loans, Eurobonds,
Brady bonds and local issues.
Bank loans
GDP EXPORTS
$27.8 Trillion $5.1 Trillion
Emerging
MRrkets
23% Emerging
Markets
18~/O
De"eloped
Economies
72%
Developed
Economies
77~'o
Figure 1. Emerging economies: share of World economy (Source: World Bank, World Development
Report, 1997, Institute of International Finance).
Eurobonds
Developing countries have issued bonds to foreigners for at least 100 years.
(Eurobonds are internationally issued securities.) These bonds were serviced
even during the 1980s bank loan crisis. One possible motivation behind such
an admirable repayment history may have been that these obligations were
small compared to bank debt (and now compared to the Brady issues), so that
338 R. J. Bernstein and J. A. Penicook Jr
DCorporale
Sovereign
Olher Bonds USS
20%
Local Currency
5%
8%
Figure 2. Emerging market debt universe as of December 1996 (Source: Salomon Brothers).
1992
LouJ lnstnnnenu
I~,
llXans Irlilrumen~
)1"0
Figure 3. Emerging market debt trading volume (Source: Emerging Markets Traders Association).
Emerging market debt 339
Argentina Characteristics
29%
Face Value $22 billion
Market Value $23 billion
Average Maturity 7.7 years
Duration 4.5 years
Yield.to-Maturity 9.5%
Mexico
55% Spread vs. U.S. Treasury 3.3%
Sovereign Component 61%
Corporate Component 39%
15%
10%
5%
12191 6192 12192 6193 12193 6194 12194 6195 12195 6/96 12/96
Figure 5. Return volatility. Latin Eurobond Index and Emerging Markets Bond Index Plus
(EMBI +) (trailing 12-month basis). Annualized standard deviation of monthly logarithmic return
premia. Latin Eurobond index, J.P. Morgan: data from 12/31/91-12/31/96.
* Index shown reflects EMBI from 12/31/90-12/31/95 and EMBI+ from 12/31/95-12/31/96.
Moreover, the illiquidity of the Eurobond sector actually induces higher volatil-
ity in the Brady market as Eurobond investors attempt to hedge risk exposures
via Brady short sales. In sum, the Eurobond market has had lower return
volatility than the Brady market because it is less sensitive to interest rate
volatility (lower duration), and because it has been less liquid. Because new
emerging market bonds are issued in the Eurobond format, Eurobonds will
eventually dominate the Brady sector.
Discount bonds
Debt principal reduced by 35-50% of face value
Coupon floats at a spread over LIBOR
Principal collateralized with US Treasury 'zeros'/rolling interest guarantee
Par bonds
Debt exchanged at par
Coupon fixed at below market rate
Principal collateralized with US Treasury 'zeros'/rolling interest guarantee
Debt conversion bonds
Exchange of old bank loans (DCBs) contingent upon 'new money' bonds (NMBs)
NMB and DCB coupons float at a spread over LIBOR
No collateral; pure sovereign risk
Table 2 Debt reduction achieved through Brady plan (in $US, billions)
* Pre-Brady Plan debt figures equal face amount plus interest arrears.
** Sum of Pre-Brady Debt and Equal-weighted average of Effective Discount Rates.
Source: Salomon Brothers, ANZ"
342 R. J. Bernstein and J. A. Penicook Jr
Local issues
Restructuring plan Issue Instruments issued Maturity (yrs) Rate Principal Interest Principal Interest Debt Interest
exchange Grace (yrs) type collateral collateral reduction reduction Conv/New arrears
data money
Fixed
36%
Non-Dollar
U.S. Dollar 7%
93%
Non-
Perfonning
12%
UncoJlateralized
62%
Collateral ized
38%
The J.P. Morgan Emerging Markets Bond Index (EMBI), was initiated on
December 30, 1990 and became the most widely used index. The EMBI is a
capitalization weighted index of US dollar-denominated Brady and 'Brady-
like' bonds. The latter predates the official Brady program, but exhibit similar
characteristics to Brady bonds and trade in the same market.
In July 1995, J.P. Morgan introduced an index that is broader than the
EMBI, creatively named the EMBI + (Figure 7). The goal was to include a
wider range of markets and debt instruments than the Brady market. The
EMBI + includes five additional markets - Morocco, Panama, Peru, Russia
and South Africa - and expands the December 31, 1995 market capitalization
from $76 to $11 0 billion. The EMBI +, like the EMBI is limited to issues that
are readily accessible to institutional investors and is published daily.
In addition to Brady bonds, the EMBI + contains several similar debt
instruments. The December 1996 EMBI + comprises Brady bonds (70%), loans
(14%), Eurobonds (10%), and local issues (6%). Because most of these loans
346 R. J. Bernstein and J. A. Penicook Jr
Venezuela
South Africa 8.8% Characteristics
0.5%
Ecuador Bulgaria
2.5% 1.6%
Figure 7. Emerging markets bond index (J. P. Morgan), as of December 31, 1995.
have not yet been restructured into 'Brady-like' bonds, these securities are
treated differently than performing bonds for statistical purposes. In short,
while inferences can be made with regard to the loans' yields and spreads based
upon the projected restructuring, the index does not do so in order to remain
objective about the restructuring process. That is, the index treats the pre-
restructured loans as having no yield until the restructuring actually occurs.
Because of the EMBI + 's expanded breadth, it is now the most widely used
index for analytical purposes. We have constructed a composite index compris-
ing the EMBI from 12/31/90 to 12/31/95 and the EMBI + from there forward.
The myriad of bond types in the Brady market-fixed rate, floating, step-up-to-
fixed, step-up-to-floating coupons-together with different collateralization
structures and amortization schedules make standard yield calculations and
comparisons inapplicable. For example, comparing the yield on a collateralized
bond of country A to the yield on an uncollateralized bond of country B does
not provide information as to the yield on the risky sovereign cash flows, which
is important to the investor. As a result, a number of conventions have been
established by market participants in order to facilitate relative value compari-
sons among the different developing country bonds, and between them and
other fixed income securities. Indeed, a new vocabulary has been created for
the developing country debt market. The purpose of these analytical tools is
to evaluate the sovereign portion of the bond independent of the collateral.
'Stripped' yield and 'stripped' spread
Brady bonds are unique in that for a number of these bonds, two (or more)
semi-annual coupon payments are collateralized with money market securities,
Emerging market debt 347
Risk measures
1 The coupon collateral is a contingent, or rolling, interest guarantee in that if the sovereign
borrower makes its coupon payment, the collateral remains in place and 'rolls forward' to cover
the next scheduled coupon payment. In the event of default, interest collateral would be paid to
the bondholder in lieu of the sovereign's payments. Market participants disagree on the exact
valuation methodology for the interest guarantee because of differing opinions on contingent
valuations. Some use probability models to estimate the timing of the contingent guarantee, others
assume that the immediate coupon payments are riskless, and yet others ignore the rolling interest
guarantee altogether for yield calculations. Just as differing assumptions in option-adjusted mort-
gage models make cross-model comparisons problematic, 'stripped spread' comparisons of various
Brady issues are only meaningful in the context of a particular model.
348 R. J. Bernstein and J. A. Penicook Jr
South 1\ rriC3
l'hilippillC'S _Tolal "Ulcodc"\J- Yicl~,
P(lltUHJ CSove",ign "Slrippc~ - Yidds
M ...\ico
V,,ne7.ucla
Nigeria
Argentina
Panama
Ur.v.il
I'ern-
Mtlrocco
Ecuador
~u.s: s ia
ULJIl:!-~ifi3
EMIlI+
0 2 6 10 12 14 16 18 20
I'uunt
Figure 8. Emerging Markets Bond Index Plus (J.P. Morgan), as of December 31, 1996.
1 Yields for non-performing loans reflect market estimates.
2EMBI+ - Emerging Markets Bond Index Plus, J.P. Morgan (December 31,1996).
3 Note: EMBI + composite blended and sovereign yield calculations exclude non-performing loans.
Cash
Flow SI,::orL~~~~~~'~n~~~~LLLLLLLLLLLLLL~~~ll-ll-ll-U-U-~~~~~~~J
D D
Co!ta ler,'
non n n 0
5 10
0 0 non
15
000 non
20
0 DOD 0 0
25
no
30 Yoo,s
'1 ',
...
1
Sow.,elgn V . ~I d
SO\lefelgl"
l2 - .. Sp,na
Discount U,S. Tre.,sury Yield
Fili;lDJ t tI 'A t
...
20 25
l
30
+
'fUrS
'0 '5
Bond : 7
Cc ll ~lII.ral : Sfj.'5 59.5516
Sovereign: ""ffi
Figure 9. Brady bonds valuation - an illustration. Note: Valuing the collateral by discounting the
collateral cash flows at the appropriate US 'spot' interest rate, allows the present value of the
collateral ($16) to be subtracted from the market price of the bond ($47) to derive the present
value of the sovereign cash flows ($31). Given the timing of these sovereign cash flows, a yield-to-
maturity on the stripped sovereign portion can then be calculated.
Emerging market debt 349
Percent
2 5 , - - - - - - - - - - - - - - - - - - - - -________________________________- ,
20
15
.,
10 ...
~ ... ..
'
...
"
,",
,.,
....
12190 6191 12191 6192 12192 6193 12193 6/94 12194 6/95 12195 6/96 12/96
Figure 10. Emerging Market Brady Index Plus, stripped yield and stripped spread history. Index
shown reflects EMBI from 12/31/90--3/31/96 and EMBI+ from 3/31/96-12/31/96.
to sovereign credit risk (in some cases as little as 50%), a change in stripped
spread will result in the repricing of only a subset of the cash flows (the
sovereign cash flows).
Thus, in addition to the standard interest rate duration measure, 'spread
duration' measures the bond's price responsiveness to movements in the
stripped spread. If an overall widening of credit spreads is expected, the portfolio
manager now has the tool to estimate which bonds will be more or less
adversely affected. A Brady bond's spread duration is a function of the collater-
alization level, stripped yield level and maturity. Since most Brady bonds were
issued in the last 5 years, maturity (typically 20-30 years) has less impact upon
spread duration than does collateralization and stripped yield levels.
Spread duration is a particularly important portfolio management tool
because 'stripped spreads' are themselves so volatile. The historical volatility
of EMBI + stripped spread changes is approximately 42% and is higher for
individual countries (Table 4). Again, this volatility is due both to the perceived
default risk of the countries, as well as the limited investor base of these
securities.
Attribution of returns
ing total returns across Brady countries because of the significantly different
bond formats. Because investors can manage or hedge general US interest rate
exposure outside of the Brady market, a Brady bond's total return can most
usefully be separately attributed to US interest rate exposure and to sovereign
spread performance.
To illustrate these distinct effects, Table 5 reviews December 1996 EMBI +
returns. December's rising US interest rates (approximately 35 basis points)
had varying impacts on returns due to differing interest rate sensitivities across
countries. For example, the rise in US interest rates had a relatively large
negative return impact on Nigeria because of its low, fixed rate coupon. In
contrast, the same rise had a positive impact on Bulgaria since its bonds are
predominantly floating rate issues. Moreover, evolving assessments of credit-
worthiness also had differing return impacts on each country. For example,
during December, investors became much more optimistic regarding Nigeria's
creditworthiness and consequently bid up prices aggressively to reflect perceived
lower credit risks. In contrast to the large positive return due to credit spread
effects in Nigeria and other countries, the market's assessment of Peru's credit-
worthiness deteriorated slightly and modestly decreased returns.
Many economic measures are relevant to assessing the credit risk of a develop-
ing country. One manner of organizing economic and financial considerations
is to compartmentalize measures into three categories: structural, solvency and
serviceability. In addition to making the analysis more manageable by removing
redundancies, this categorization produces a 'term structure' of credit risk, akin
to the well-known notion of the term structure of interest rates. Political
analysis also is required to assess policy stability.
Structural
Measures belonging to this category describe the long-term fundamental health
of the country. They include economic variables such as reliance on a particular
commodity for export earnings, welfare indicators such as per capita GNP,
and social/economic measures such as income distribution. These variables
generally are not directly linked to default, but countries with poor structural
fundamentals are likely to develop economic problems. Further, given two
countries with similar other variables, the one with the inferior structural
measures will likely have a lower tolerance to adverse economic shocks.
Solvency
In contrast to the structural variables, the solvency class contains intermediate
term measures of a country's economic health. In particular, these variables
should reflect the country's ability, over time, to meet its central government
352 R. J. Bernstein and J. A. Penicook Jr
Table 5. Emerging Markets Bond Index Plus factor returns - December 1996
Sovereign credit spread (incremental income) plus spread change effect (principal)
debt obligations. Both internal and external debt are included. Countries with
inferior solvency measures, all else being equal, have higher default risk because
international debt service competes with local economic constituencies for
resources.
Serviceability
The factors in this category are of short-term, if not immediate, concern. They
reflect the country's foreign exchange reserve position relative to its immediate
obligations (and are therefore usually presented in ratio form). Despite good
or improving fundamentals and strong solvency measures, a developing country
may be forced into a crisis if its reserves are (or will become) deficient, or if
alternative reserve sources, such as the International Monetary Fund, are
circumscribed. Serviceability then would be analogous to a company's ability
to meet payroll or lease payments with sufficient working capital.
Political considerations
in place to serve as anchors to these policies. The resignation or death of one key
policy maker may be enough to alter economic policy. In sum, political factors
can cut both ways: the politics of individual countries are often fragile, but interna-
tional politics often act as counterbalances.
While nascent representative governments may suffer from institutional
instability, it is important to recognize that these countries have undergone
profound political change in a short time period. Several countries have moved
from military rule to competitive, multi-party democracies within the decade.
For example, in 1982, 92% of the EMBI + countries' popUlations were under
communist or military rule; now 80% are governed by democratic rule.
Willingness to pay
Some argue that sovereign risk analysis is doomed to failure because, notwith-
standing its ability to pay, a country may be unwilling to pay. Distinguishing
sovereign risk from corporate or municipal credit risk on this basis alone exposes
a deficient understanding of default risk. Borrowers default when their competing
economic interests override the damage done by default, and default is never a
casual decision. Corporations and municipalities are faced with the same decision
as sovereign borrowers: at what point are you willing to capitulate and damage
your reputation? Few wait until they are completely destitute to make this deci-
sion. For example, Columbia Gas Systems found the burden of high-priced, long-
term gas 'take or pay' supply contracts of the 'energy shortage era' so damaging
to its future that management declared bankruptcy and forced its suppliers to
re-negotiate the supply contracts to lower prices. Similarly, Orange County
California viewed the financial implications of their failed investment scheme so
negatively that they too declared bankruptcy. Orange County taxpayers perceived
little 'ownership of the problem' because of the obscure nature of the investment
scheme and its genitor. Most important, both borrowers were willing to default
even though the debtors had substantial resources available to pay creditors and
suppliers. The point is that economic strain creates a 'willingness' issue for borrow-
ers of all types.
Sovereign credit perspective
Total External debt is equal to net external debt for developed economies and gross external debt for emerging economies.
Inflation is measured on an annual average basis for Panama; Inflation is measured as CPI price change year-over-year for all other countries.
Emerging market debt 355
PORTFOLIO CONSIDERATIONS
Since the beginning of 1991, when Brady bonds became viable assets for
institutional investors, the EMBI has outperformed the broad global and US
bond markets by an extremely wide margin. Figure 11 displays EMBI return
premia (defined as total return less cash return) along with return premia for
other market indices. Of course, given the nature of the risk inherent in these
bonds and the immaturity of the market, volatility is also greater. Sovereign
credit spreads themselves are volatile, as noted in Table 4, and have low
correlations to US interest rates (Table 7). Consequently, this unique credit
risk translates into low return correlations with the other major markets
(Table 8) and suggests potential portfolio benefits. These portfolio benefits are
illustrated in Figure 12, which shows a simplified 'efficient frontier', including
emerging debt markets. The efficient frontier including emerging debt markets
dominates the developed markets alternative since 1990. Other major indices
are also plotted for comparative purposes.
Return volatility
Monthly log returns are presented for the EMBI in Figure 13. Figure 14 pro-
vides a perspective on return premia volatility. Returns are volatile for two
356 R. J. Bernstein and J. A. Penicook Jr
18%
12%
10%
8%
6%
4%
2%
0%
EMSI... WGBI BIG GIM
Figure 11. Market indices, return premia and volatility - 5 years (December 31, 1990-December
31, 1995). EMBI, Emerging Markets Bond Index Plus, J.P. Morgan. Index shown reflects EMBI
from 12/31/90-12/31/95 and EMBI + from 12/31/95-12/31/96. BIG, Broad Investment Grade
index, Salomon Brothers. WGBI, World Government Bond Index, Salomon Brothers. GIM, Global
Investable Markets index, Brinson Partners. Return premia equals total return less cash return
calculated on a logarithmic basis in US dollars. Volatility equals annualized standard deviation
based on monthly logarithmic return premia.
Table 7. Emerging markets bond index spread changes and US treasury yield changes correla-
tion table
Table 8. Market indices return premia correlation matrix, 5 years (December 31, 199O-December
31, 1995)
EMBI+ LOO
WGBI 0.43 LOO
BIG 0.37 0.86 LOO
GIM 0.54 0.61 0.53 LOO
EMBI +, Emerging Markets Bond Index Plus. Index shown reflects EMBI from 31/12/90-31/12/95
and EMBI+ from 31/12/95-31/12/96, J.P. Morgan.
BIG, Broad Investment Grade Index, Salomon Brothers.
WGBI, World Government Bond Index, Salomon Brothers.
GIM, Global Investable Markets Index, Brinson Partners.
Return premia equals return less cash return.
Correlations based on monthly logarithmic return premia calculated on a logarithmic basis in
US dollars.
10.0% , - - - - - - - - -_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _- ,
8.0%
7.0%
o
3.0%
1.0%
--
0~'-0%-----3.....0 % - - - - -
.-. ...-----5.....0 % - - - - -.. . . , - - - - - 7.....0%----~8.0%
10%
10%
Figure 13. Emerging markets bond index, monthly returns, 5 years (December 31, 1991-December
31, 1996), calculated on a logarithmic basis in US dollars.
~%,------------------------------------------------------------,
Averago 14,24%
25%
20%
15%
10%
5%
Figure 14. Emerging markets bond index, return volatility, 5 Years (December 31, 1991-December
31, 1996), volatility equals the annualized standard deviation of monthly logarithmic return premia.
market in 1994. In both cases, investors had taken on more risk than they
understood or appreciated (corporate credit risk and interest rate option risk,
respectively) and had enjoyed very good absolute returns for a significant
period of time. However, these securities were particularly ill-suited for their
respective owners. In the case of high yield bonds, the investment horizon of
retail mutual fund investors was a poor match for this illiquid and volatile
market. In the case of the mortgage market, portfolio managers of retail-
Emerging market debt 359
oriented, short-term government bond funds (and others) were heavily invested
in securities with extreme prepayment sensitivity. When interest rates began to
rise, reversing a 3-year trend, homeowner prepayments slowed abruptly. The
value of many sensitive securities declined precipitously as prepayment assump-
tions were reassessed. The emerging debt market mirrored these two examples
in two respects: it had produced uniformly stellar returns over a preceding
three year period, and mutual fund managers had placed billions of dollars of
these securities in retail-oriented mutual funds that have liquidity and risk
tolerances inconsistent with the nature of the securities. The panic liquidation
of these investments after the Mexican peso devaluation caused tremendous
illiquidity in the market, pushing prices down well below fundamental value.
Figure 15 outlines the changing participation in the emerging debt market.
While this data is estimated, long-term investors such as pension funds and
insurance companies are gradually replacing banks, flight capital and hedge
funds as the primary holders of these bonds. As the investor base evolves,
volatility due to liquidity panics should lessen.
Correlations
An attractive feature of the Brady market is its low correlation to other asset
classes, including other US bond markets, as illustrated in Table 8. The limited
performance history of individual Brady countries, however, blunts the oppor-
tunity to distinguish among countries with statistical confidence.
Asset allocation
Expected return
1990
165 billion US$
Banks
97%
on-Banks
30/0
1995
179 billion US$
Insurance Companies
3%
Dealers
5%
Local Banks
4%
Figure 15. Global distribution of total emerging market debt (in SUS) (Source: Salomon Brothers).
Emerging market debt 361
o Equity Markets
o Bond Markets Emltrgin, EIIUi1W
Emerging Debt
J~P~O
~--------~----~--~----~----~--------~~--~--~
o0.0 0.5 1.0 1.5 2.0 2.5
RiSk (8eta)
Figure 16. Equilibrium risk/reward. Global market beta as risk equilibrium risk premium.
~
~
In the short run, the discount rates applied to bond cash flows will drive
prices and returns; but in the long run, the cash flows themselves determine
the return performance. That is, for long-term horizons, the actual receipt of
cash flows is the critical issue in realizing expected returns. Presuming that the
recent Brady plan debt relief would be sufficient to resume debt repayments,
cash flows can then be estimated by projecting the probability and timing of
any future default/renegotiation. Holding the discount rate constant, a cash
flow reduction roughly translates into a proportional return reduction. Thus,
if a debtor immediately reduced its coupon payments on a 30-year obligation
by 40% (and then made all future payments), the expected holding period
return would decrease by approximately 43%. For example, if the stated cash
flows were priced to return 18%, the 40% coupon reduction would drop
expected returns to approximately 10.3%. Each year that passes without a
rescheduling significantly boosts return: the expected return climbs to 14.8%
if rescheduling does not occur until year seven.
Figure 17 examines return scenarios given varying recovery rates and varying
pricing yields. (The pricing yield is simply the yield to maturity given the price
of the cash flows.) The top graph illustrates that at the currently high pricing
yields, holding period returns climb rapidly as renegotiation is avoided. Also,
after 10 years of payments, returns are relatively insensitive to recovery assump-
tions, due to high compounding rates. The bottom graph indicates that for a
given principal recovery assumption of 60%, returns are more variable at
higher pricing yields.
Another manner of assessing prospective returns is to consider the implied
conditional probability of default (conditional upon no previous default and
equally probable in any year). That is, given market pricing, recovery estimates
and required return, what is the implied annual probability of default? Figure 18
plots probability-weighted return versus annual default probability (for three
pricing scenarios), assuming a 60% recovery value. Also, for each conditional
default probability, a corresponding cumulative 5-year equivalent is plotted on
the right scale. For example, if the sovereign cash flow is priced to yield 18%,
the recovery assumption in the event of default is 60% and the required return
is 10%, then the implied conditional default probability assumption is 15%;
the corollary assumption is that the probability of default within 5 years is
55%. Framed in this manner, the investor can judge the reasonableness of this
default assumption in light of intermediate global economic trends and portfolio
diversification effects.
In summary, sovereign default episodes are relatively rare and no easier to
anticipate than large macro-economic shifts. Pricing yields, recovery value
assumptions and default probabilities determine expected returns.
All fixed income sectors (mortgage, investment grade corporates, high yield
and Brady bonds) are exposed to the risk that spreads over US Treasury rates
364 R. J. Bernstein and J. A. Penicook Jr
Long-Term Returns
Varying Rate of Recovery (40%. 50%. 60%)
Pricing Yield 016%
20%
18%
16%
14%
12%
E
i 10%
~
II:
8% %
- - - -- - 50%
6%
-40%
4%
2%
0%
0 4 8 12 18 20 24 28
y_untII~
Long-Term Returns
Varying Pricing Yield (10%. 14%. 18%)
Recovery Rate .60%
20%
18%
18%
8%
b3
6%
4% - - - -- - 14%
-10%
2%
0%
0 4 8 12 16 20 24 28
y .... until Renegotiation
18%r-__~----------------------~~~~~~~-.90%
16% 80%
1 14% 70%:a ~
-ij, 12% 60%~~
li 11)111
:= ~10% 50%111 '8
> ..
~1i
= II: 8% 40%~~
.D ::J::J
11 6% 30%E,!
e
0.. 4%
::Jill
20%OQ
2% 10%
0% +-~~+-----~----~----~-----+----~-----+O%
0% 5% 10% 15% 20% 25% 30%
Conditional Default Probability (annual)
Figure 18. Return vs. conditional default probability, recovery rate assumption = 60% (varying
pricing yield).
change. That is, despite a constant Treasury curve, incremental yield (spread)
above this curve may increase or decrease for a number of reasons related to
perceived risks in the sector. The change in spread impacts total returns just
as a change in the underlying Treasury curve does.
Active management opportunities exist because spreads across countries are
less than perfectly correlated and because bond structures vary within a country.
Table 10 shows the correlation of stripped spreads across countries. Again, for
some countries, the performance history is brief. Nonetheless, each country's
perceived creditworthiness does not move in lock-step with the overall market,
as noted in the discussion of return attribution. This variation translates into
active management opportunities as investors compare the price of sovereign
credit risk (spread) versus their own assessment of that sovereign credit risk
(value).
Opportunities also arise due to the variety of Brady bond structures. For
instance, Argentine issues differ in that the Par bond is a fixed rate instrument,
while the FRB is a floater. Hence, the spreads on the two Argentine issues may
move differently as investors use the bonds to manifest opinions on general
interest rate trends. For example, if investors anticipate declining interest rates,
they may overwhelmingly prefer the fixed rate Par bond over the FRB, notwith-
standing the relative spreads on the two issues, which are illustrated in Figure 19.
The point is that investors focusing solely on relative spread characteristics of
the issues may capitalize on these market phenomena by swapping into the
higher credit spread issue and hedging differing interest rate sensitivities.
RELATIVE VALUE
~
~
Table /0. Emerging Markets Bond Index stripped spreads, correlation matrix five years (December 31, 1990-December 31, 1995)
Index
f
~
'";:S.
Start Date EMBI Argentina Brazil Bulgaria Ecuador Mexico Morocco Nigeria Panama Philippines Poland S. Africa Venezuela Latin NonLatin
I:>
0.5
-0.5
.,
".5
2.5
-3
:;: .,
~ ~ <a ~ m ~ m ~ ~ ~
Figure 19. Argentine brady bonds, stripped spread comparison - floating rate bond vs. par bond
(data through December 31, 1996).
CredIISP*(IOifIerenCe('IIo)
4.5 ,----'-'-------=--:-::~-c:__----------------__,
Repubtic of Argentina
Spread Minus YPF Spread
3.5
2.5
1.5
0.5
Figure 20. Argentine Eurobonds, sovereign vs. corporate yield spread comparison (Republic of
Argentina 2003 vs. YPF 2004) (data through December 31, 1996).
of 2004 ($350 million issue), respectively. YPF was the Argentine national oil
monopoly and remains its largest corporation after being privatized. Due to
the sovereign risks noted above, the corporate issue should be priced at a
higher credit spread than its sovereign counterpart; currently, the situation is
grossly reversed and this is observed in other countries as well. This is the
equivalent of accepting a lower yield for IBM, Ford or J.P. Morgan than the
US Treasury. Hence, sovereign Eurobonds are far more attractive than their
corporate counterparts given a longer term, more fundamental view.
Some market participants argue that, under duress, countries may choose
to service Eurobond obligations over Brady obligations, just as they discrimi-
nated against the old bank loans. Others believe that since the preceding bank
loans were already renegotiated, the resulting Brady bonds suffer some genea-
logical defect and are more vulnerable than Eurobonds. This concern misses
the more important dynamic that Eurobond investors and bank creditors of
the 1980s were distinct investor groups. Today, the same creditor community
holds both Eurobonds and Brady bonds, so sovereign issuers no longer benefit
from discriminating among their repayment priorities. In short, the borrowers
have less leverage with their creditors now.
If default risk and volatility distinctions do not warrant higher credit spreads
on Brady issues, perhaps the sheer complexity of the Brady formats may explain
the relative mis-pricing. Just as US High Yield bond portfolio managers are
often unfamiliar with sovereign credit analysis, others are equally uncomfortable
with the unique analytical aspects of the Brady market, as discussed earlier.
Figure 21 plots the sovereign (stripped) spread of an Argentine Brady bond
versus the spread of the previously mentioned Republic of Argentina Eurobond.
Emerging market debt 369
Figure 21. Argentine sovereign yield spread, Brady vs. Eurobond (Argentine discount vs. Republic
of Argentina 2(03) (data through December 31, 1996).
Despite having identical sovereign credit risk, the Brady issue's sovereign
exposure is currently priced much more attractively.
Sovereign credit risks are obviously different in nature from corporate credit
risks. Nonetheless, they are viewed as similar in degree by the credit rating
agencies (generally rated BB/B). Figure 22 outlines the spread history of the
Brady market versus the US High Yield market. On this basis, Brady spreads
are currently attractive not only relative to Eurobonds but also to US High
Yield alternatives.
CONCLUSION
The Brady bond market is the largest and most actively traded US dollar
segment of the emerging debt market. While developing countries began adopt-
ing the Brady plan in 1989, institutional investors have just begun taking an
active interest in the resulting bonds. In order for interest to expand, investors
must become familiar with this market's important diversification benefits, its
unique analytics and the peculiar aspects of sovereign credit analysis.
As presented here, new analytical methods are required to accurately mea-
sure the pricing of Brady bonds' credit risk. Sovereign credit analysis, or the
assessment of fundamental value, also requires an understanding of the unique
serviceability risk of US dollar-denominated, emerging market debt. Due to
currency denomination, investor segmentation, liquidity and analytical com-
plexities, the Brady sector is currently the most attractive segment of the
emerging debt market - more attractive than foreign corporate debt and
370 R. 1. Bernstein and 1. A. Penicook lr
Figure 22. Brady market vs. high yield market (data through December 31, 1996. Source:
Salomon Brothers).
REFERENCES
Conybeare, John A.C. (1990) On the repudiation of sovereign debt: sources of stability and risk.
Columbia Journal of World Business, Spring/Summer.
Dym, Steven (1992). Global and local components of foreign bond risk. Financial Analysts Journal,
March/April,83-91.
Dym, Steven (1994). Identifying and measuring the risks of developing country bonds. Journal of
Portfolio Management, Winter, 61-66.
Eichengreen, Barry and Richard Portes (1988). Setting Defaults in the Era of Bond Finance. Centre
for Economic Policy Research, Discussion Paper No. 272.
Fridson, Martin S. (1994). International emerging markets debt in the asset allocation process. High
Yield Securities Research, Merrill Lynch, October 28.
Fridson, Martin S. (1994) Political Risk versus Corporate Risk: A Phony Comparison. Merrill Lynch.
Grief, Avner (1994). Cultural beliefs and the organization of society: a historical and theoretical
reflection on collectivist and individualist societies. Journal of Political Economy, 102,912-950.
Johnson, Bryan T. and Thomas P. Sheehy (1994). The Index of Economic Freedom. The Heritage
Foundation.
Purcell, John F. H. and Jeffrey A. Kaufman (1993). The Risks of Sovereign Lending: Lessonsfrom
History. Emerging Markets Research, Salomon Brothers.
KENNETH ROGOFF
Princeton University
1 For overview of the theory of sovereign debt see Obstfeld and Rogoff (1996, chapter 6), and
Eaton and Fernandez's (1995).
2 For a model of debt renegotiation that encompasses both debtor and creditor governments,
see Bulow and Rogoff (1988). That model focuses on creditors' ability to interfere with the gains
from trade that both debtor and creditor-country taxpayers enjoy.
371
R. Levich (ed.). Emerging Market Capital Flows. 371-373.
o 1998 K luwer Academic Publishers.
372 K. Rogoff
3 Bulow et al. (1992) argued that the market value of World Bank loans was well below book
value for the reason given in the text.
Comments 373
at best, however, so investors cannot rely too much on this mechanism (see
Obstfeld and Rogoff, 1966; chapter 7). I conclude only by restating the obvious;
these new loans may produce very high profits, but they are risky.
REFERENCES
Bulow, Jeremy and Kenneth Rogoff (1988). Multilateral negotiations for rescheduling developing
country debt: a bargaining-theoretic framework. International Monetary Fund Staff Papers, 35,
644-657.
Bulow, Jeremy, Kenneth Rogoff and Afonso Bevilaqua (1992). Official creditor seniority and burden
sharing in the former Soviet bloc. Brookings Papers in Macroeconomic Activity, 1,195-222.
Eaton, Jonathan and Raquel Fernandez (1996). Sovereign debt. In Gene Grossman and Kenneth
Rogoff (eds.), Handbook of International Economics. Amsterdam: Elsevier Science Publishers,
1995.
Obstfeld, Maurice and Kenneth Rogoff (1996). Foundations of International Macroeconomics.
Cambridge: MIT Press.
MARTIN D. EVANS
Georgetown University
INTRODUCTION
POTENTIAL PITFALLS
In view of the admirable clarity with which the paper is written, I have few
criticisms of the analysis. Therefore, in these comments, I will focus on the
interpretation of the paper's empirical results and place them in perspective
against other Brady bond research. In particular, I wish to point to some
potential pitfalls of the methodology employed and consider whether these
could impact upon the central conclusion concerning the importance of sover-
eign risk.
The identification of state variables
namely, credit risk as measured by a single default probability, and the level
of an appropriate US interest rate. Although this squares with the bond pricing
model, other state variables may also impact upon the bond's duration in more
general models (see, for example, Claessens and Pennarchi, 1992; Longstaff and
Schwartz, 1993; Cumby and Evans, 1995).
In general, the price of a Brady bond will depend on the probability that it
will default at any future point in time. These probabilities describe the term
structure of credit risk for the bond. In the author's model, the term structure
of credit risk has a simple structure because current and anticipated future
credit risk are assumed to be identical. That is to say, the probability at period
t of a default between periods t + hand t + h + 1 is the same at every horizon
h > O. Under these circumstances, when new information about the credit risk
comes to the market, the resulting variations in the term structure can be
captured by movements in a single factor. Alternatively, it is possible that the
market responds to new information that has implications for current or future
credit risk alone. Under these circumstances, there could be changes in both
the 'shape' and 'level' of the term structure that cannot be represented by
movements in a single factor. In this case, the duration of the bond will depend
on all the factors needed to represent the movements in the term structure.
With this perspective, it seems to me that the theoretical justification for
focusing on just two state variables is a little weaker than authors suggest. It
may well be that the choice is entirely appropriate for the Argentinean bonds
given the behaviour of their term structures, but this cannot be determined
purely on theoretical grounds.
Could the possible omission of relevant state variables significantly affect
the author's results? While it is hard to answer this question definitely, I doubt
that this is the case for two reasons. First, the authors demonstrate that their
results a quite robust to changes in the dependency between the hedging ratio
and the included state variables. Such robustness is unlikely if an important
state variable has been omitted. Second, their results appear consistent with
other research that uses quite different methods to study the Brady bond data.
I shall discusses these findings below.
A second potential pitfall concerns the use of the strip spread (the difference
between the yield on a Brady bond stripped of its guarantees and the yield on
a US Treasury) to proxy the probability of default. The accuracy of this proxy
depends in a large part on the method used to value the guarantees. Although
a number of different methods are used by financial institutions for this purpose,
they generally assume that current and anticipated credit risk are identical.
While this implies a simple structure for the term structure of credit risk that
considerable simplifies valuation, there is little a priori reason to believe that
this restriction holds true for a particular bond. Thus, in the absence of empirical
evidence supporting these restrictions, and hence the valuation model, there
Comment 377
must be some question over the accuracy of the strip spread as a proxy for the
probability of default.
Some recent research by myself and Robert Cumby (Cumby and Evans,
1995) examined this issue. After estimating alternative valuation models that
used different specifications for the behaviour of the term structure of credit
risk, we concluded that market participants differentiated between current and
anticipated future credit risks when pricing five of the six bonds in our sample.
For these bonds, our findings imply that the strip spreads calculated from
standard models contain valuation errors.
Figure 1 gives some visual evidence on the importance of these valuation
errors. The market price of the Mexican Discount bond between the beginning
of 1990 and the end of 1992 is plotted against the left hand axis. According to
our tests, the behaviour of these prices is inconsistent with a simple valuation
model in which current and anticipated credit risk are identical. The second
series, plotted against the right hand axis, is an estimate of the error in valuing
the guarantee. This is the difference between the estimated values calculated
from our preferred model (that allows for differences between current and
anticipated future credit risk) and a standard model (where the risks are
identical). As the figure shows, the difference between the value of these guaran-
tees is quite variable and economically significant. From these results it appears
that the standard way of calculating the strip yield for the Mexican Discount
bond would introduce a significant measurement error.
The plots in Figure 1 are quite representative of our results for Venezuelan
and Costa Rican bonds. Thus, while I have no direct evidence on the accuracy
of the Argentinean bond strips, it seems reasonable to doubt their accuracy in
the light of these findings. In principle this could have quite an impact on the
results. For if the authors' regression methodology is unable to pick up much
of the true variations in the dynamic hedge ratios because the state variables
are inaccurately measured, it is possible that their results understate the true
sensitivity of the Brady's to interest rate risk. Fortunately, the evidence cited
below suggests that this assessment is unduly pessimistic.
60 .
0.2
~". "'"90.06"""""
5590.03 90.09 "'"90.12."""91.03
,,,. '" 91.06
" " '''' 91.09
,,,,,. " 91.12
" "" .. " 92.03
,,' "" .. 92.06
'''' ", .. """, ... """,,)0
92.09 92.12
Non-fundamental factors
The authors interpret their findings to mean that sovereign risk is the dominate
factor in explaining Brady bond returns. Strictly speaking, this interpretation
says more about their priors than about the data because in effect we are
labelling a regression residual. It is quite possible that other non-fundamental
factors unrelated to the factors present in bond pricing models have a significant
impact on actual returns. In other words, the finding that Brady bond returns
are insensitive to interest rate variations does not in itself demonstrate that the
returns primarily compensate for sovereign risk.
To gain some insight into the possible importance of non-fundamental
factors it is useful to consider how well structural models for the pricing of
Brady bonds perform. For the case of the Mexican Discount bond shown in
Figure 1, the preferred model in Cumby and Evans (1995) implies that 96%
of the price variations are explainable in terms of fundamentals, i.e. the term
structure of credit risk and US interest rates. Since similar results were found
for the other Brady bonds studied, it appears that fairly simple pricing models
can relate a very large proportion of the price variations to fundamentals
factors. This evidence suggests that non-fundamental factors do not have a
dominant influence on Brady returns.
The structural model estimates can also be used to directly examine the
importance of sovereign risk for price variations in Brady bonds. In the case
of the Mexican Discount bond, changes the term structure of credit risk account
for approximately 95% of the explainable price variations. Once again, similar
results apply to the other Brady bonds studied. Thus, the evidence from these
structure models appears to support the authors' conclusion that Brady bond
returns primary reward investors for sovereign risk.
CONCLUSION
REFERENCES
Claessens, Stijn and George Pennacchi (1992). Deriving Developing Country Repayments for the
Market Prices of Sovereign Debt. Working paper, World Bank.
Comment 379
Cumby, Robert E. and Martin D. Evans (1995). The Term Structure of Credit Risk: Estimates and
Specifications Tests. Financial Markets Group discussion paper, London School of Economics.
Longstaff, Francis A and Eduardo S. Schwartz (1993). Valuing Risky Debt: A New Approach.
Working paper, The Anderson Graduate School of Management, UCLA.
RICHARD CANTOR
Federal Reserve Bank of New York
This study analyzes the term structure of interest rates in Mexico. 1 Domowitz,
Glen and Madhaven (DGM) find that, during the 18 months leading up to
the devaluation in December 1994, the term structure was broadly consistent
with the expectations hypothesis. In addition, they find that market expectations
of default on Mexican securities were fairly stable, and currency risk premia
were actually declining immediately preceding the devaluation.
This study should attract a broad audience, including both economists who
study term structure relationships in other countries and those interested in
improving their understanding of market conditions leading up to the peso's
devaluation. DGM's finding that the expectations hypothesis is supported by
the data is of particular interest because many studies find that the US data is
not entirely consistent with the theory. Once again, we are reminded that the
volatile time series data from emerging markets sometimes provide the best
testing grounds for our most important economic propositions. DGM's findings
regarding market expectations are also important since they provide some
support for the view that the attack on the peso was not fully grounded in
observable fundamentals and therefore somewhat arbitrary.
My comments are limited to three concerns I have with the paper. First,
the study's sample size is quite small and may bias the test results in favor of
the expectations hypothesis. Second, the available data on the Mexican term
structure have a number of shortcomings that may distort their information
content. Third, the Campbell-Shiller (1991) tests of the expectations hypothesis
used by DGM may not be valid in an environment with 'peso problems', where
the market expects large, but infrequent, asset price movements. More generally,
the Campbell-Shiller tests (which are normally applied to US Treasury securi-
ties) may be inappropriate for testing term structure for securities with default
and devaluation risks.
1 Babatz and Conesa (1994) and Lederman and Sod (1996) have also written on this subject.
381
R. Levich (ed.), Emerging Market Capital Flows, 381-384.
(I 1998 Kluwer Academic Publishers.
382 R. Cantor
Since secondary market data are not available, DGM use the average yields
at auctions for 3- and 6-month tesobonos and cetes. The authors' methodology
requires that these yields reflect true market clearing rates, embodying the
marginal investor's expectations of both expected peso depreciation and poten-
tial default. A number of factors, however, suggest that these auction yields do
not represent true market rates.
Bidder collusion may have contaminated some of the observed data. During
the sample period, the government conducted both discriminatory price and
unifonn price auctions. Umlauf (1993) showed that collusion has been common
in Mexican discriminatory price auctions but has been absent in unifonn price
auctions.
Government interventions may have contaminated the observed spreads
between cetes and tesobonos rates, the so-called currency premium. During
1994, the government sharply scaled back cetes issuance and increased teso-
bonos issuance. Moreover, in November 1994 alone, the authorities spent
almost $5 billion buying up peso-denominated securities. Both actions likely
suppressed the observed cetes-tesobones spread and created a false impression
that the expectations of a devaluation were declining going into December
1994. In fact, these actions themselves suggest that concern about a devaluation
were more likely increasing.
Government actions may have also contaminated observed tenn spreads.
On 11 occasions when the cetes yield curve was inverted, long-tenn debt
offerings were increased relative to short-tenn offerings. On another 11 occa-
sions, actual cetes sales were less than originally scheduled because of weak
investor demand at auction. Similarly, on 17 occasions, actual tesobonos sales
were less than originally announced. Finally, on 10 occasions the government
exercised its 'green shoe' option and sold more tesobonos than originally
scheduled.
Comment 383
Given the enormous devaluation and near default that occurred just after the
sample period, one might reasonably expect that interest rate fluctuations
during the preceding months reflected changing expectations about the timing
and magnitude of the potential devaluation or default. DG M essentially ignore
these concerns and simply end their sample period one month prior to the
devaluation. A pessimist might despair that an accurate test of the expectations
hypothesis was impossible: one would never try to test uncovered interest rate
parity with this dataset. DGM's approach, however, can be justified if the
devaluation or potential default were wholly unexpected. Moreover, their
approach is satisfactory if the market placed equal probability of devaluation
and default on the first 3-months and the next 3-months throughout the sample.
It seems more reasonable, however, that observed term spread fluctuations
were dominated by changing expectations about the timing and magnitude of
devaluation and default risks. Around the December devaluation, the peso fell
by 50%. To keep investors whole, quarterly cetes rates would have needed to
be 100%, or 1500% measured at an annual rate! If the market had expected
that this event might occur with just a 1% probability, that alone would have
justified a 15% currency premium. Expectations of a default would have had
similar consequences.
Bekaert et al. (1995) showed that for the US, standard expectations hypothe-
sis tests are biased against acceptance. Markets may infer from rising short
rates that a long-term shift to a high interest rate regime has become more
likely. Because such regime shifts occur less frequently in the data than in the
market's expectations, the long-rate appears to overreact to changes in the
short-rate.
Standard expectations hypothesis tests for Mexico may however be biased
in favor of acceptance. Investors may have required extra compensation to
hold Mexican assets going into a devaluation or default, but may not have
expected extraordinarily high interest rates to follow these events. In DGH's
dataset, inverted yield curve observations may reflect large near-term default
and devaluation risks. When these events did not occur, short rates fell back
and validated the prior inverted yield curve. But this validation was driven by
unfulfilled rather than fulfilled expectations! Properly accounting for the peso
problem should increase estimated standard errors and affect interpretations
given to term structure variations.
Even if the sample were large enough so that peso problem were not present
(if the large, infrequent events occurred in the sample with the correct fre-
quency), the direct application of the Campbell-Shiller tests to Mexican data
might still be inappropriate. These tests were designed for use on US Treasury
data where default and devaluation risks are assumed irrelevant. For Mexican
data, this assumption is clearly inappropriate. When studying the term structure
of the currency term premia, the efficiency of the term structure imposes a
384 R. Cantor
CONCLUSIONS
DGM are right to test an important proposition with the volatile data from
emerging markets. The data are interesting but have shortcomings. The sample
is small but corrections can be made for the resulting small sample bias. More
troubling, however, is the possibility that the auction yield data may poorly
reflect true market yields due to bidder collusion and government interventions.
Lastly, it is unclear whether the Campbell-Shiller tests of the expectations
hypothesis should be applied to this sample. The implications of the peso
problem need further exploration, and the tests need to be restructured to
explicitly take into account default and devaluation risks.
REFERENCES
Babatz, G. and A. Conesa (1994). The Term Structure of Interest Rates: An Empirical Analysis for
Mexico. Mimeo, Harvard University.
Campbell, 1. and R. Shiller (1991). Yield spreads and interest rate movements: a bird's eye view.
Review of Economic Studies, 58, 495-514.
Bekaert, G., R. Hodrick and D. Marshall (1996). On Biases in Tests of the Expectations Hypothesis
of the Term Structure of Interest Rates. NBER Technical Working Paper #191, January.
Bekaert, G., R. Hodrick and D. Marshall (1995). Peso Problem Explanationsfor the Term Structure.
Mimeo, Federal Reserve Bank of Chicago.
Lederman, A. and G. Sod (1996). Risk Premia and the Term Structure of Interest Rates in Mexico.
Mimeo, Yale University.
Umlauf, S. (1993). An empirical study of the Mexican treasury bill auction. Journal of Financial
Economics, 33, 313-340.
LAWRENCE GOODMAN
Salomon Brothers Inc.
The paper conceptually links sovereign credit analysis to spreads, relative value,
and portfolio dynamics. Sovereign credit analysis begins with economic and
financial dynamics, which are broken into elements of structure, solvency, and
serviceability. Politics and willingness are identified as further complicating
sovereign analysis. The authors cite Columbia Gas Systems and Orange
County's movement toward bankruptcy based on willingness. It is important
to note that recent parallels exist in the emerging markets. Coincident with the
Brady negotiations, debt reduction often exceeded the nation's capacity to
service external obligations.
At Salomon, we take a multipronged approach to the measurement of
sovereign risk. In terms of structural and solvency issues, we begin with an
assessment of key macro-variables including growth and interest rates in the
developed world as well as commodity prices. Major assumptions regarding
future developments in the industrialized world and commodity prices are
integrated into a country-by-country modeling process. The analytic frame-
work evaluates the nation's ability to meet internal and external financing
requirements in the short and medium term. Likewise, we evaluate short-term
385
R. Levich (ed.), Emerging Market Capital Flows, 385-387.
o 1998 Kluwer Academic Publishers.
386 L. Goodman
Bernstein and Penicook also tackle the great debate relating to the 'sovereign
ceiling' or whether corporate credits should trade through their sovereign
counterparts. The authors suggest the discrepancy relates to market segmenta-
tion, namely high yield investors demonstrating a greater comfort level with
corporate credits relative to less familiar sovereign credits. An interesting area
for further study would be theoretical work examining the potential effect of
the globalization of corporations i.e. the move toward nationless entities on
credit quality. In other words, the study could determine the impact on corpo-
rate credit quality of the geographic distribution of earnings and assets.
The authors draw our attention to the distinction between Brady and Euro-
bonds. Raising the question of why Brady yields exceed comparable Euro's?
One answer is the 'genealogical defect' or the fact that they were previously
tarnished with restructuring. Moody's recently eliminated their credit distinc-
tion rating Brady's inferior to Euros as: (1) Brady trading volumes have
advanced dramatically, (2) the small denominations of Brady bonds ($250,000)
results in wide distribution complicating potential debt restructuring, (3) instru-
ments held by bond funds are less malleable creditors, and (4) the amount of
Brady's outstanding will decline relative to Euro's over time.
CONCLUSION
The implications are clear that, over time, with a heightened and broadened
analysis as presented in the Bernstein and Penicook paper, emerging-markets
debt will represent an increasingly viable adjunct to a bond portfolio.
Nonetheless, volatility will remain omnipresent, as the risk of default fails to
vanish. In fact, no instruments are immune. In other words, an altered classifi-
cation e.g. loans, Brady's, or Euros will fall short of fully-insulating creditors.
The most likely instruments for restructuring will be represented by a combina-
tion of factors such as the relative seniority of the obligations and the cash
flow burden or potential economic gain to the debtor from restructuring.
PART FIVE
ABSTRACT
In this article we discuss a scoring system (EMS model) for emerging markets corporate
bonds. The scoring system provides an empirically based tool for the investor to use in
making relative value determinations. The EMS model is an enhanced version of the
statistically proven Z-score model (Altman, 1968) designed for US companies. Unlike the
original Z-score model, our approach can be applied to non-manufacturing companies and
manufacturers, and is relevant for privately held and publicly owned firms. The adjusted
EMS model incorporates the particular credit characteristics of emerging markets compa-
nies, and is best suited for assessing relative value among emerging markets credits. The
EMS model combines fundamental credit analysis and rigorous benchmarks together with
analyst-enhanced assessments to reach a modified rating, which can then be compared with
agency ratings (if any) and market levels. We have included a summary of Mexican
companies for which we have applied the EMS model.
The emerging market scoring model (EMS model) for rating emerging markets
credits is based first on a fundamental financial review derived from a quantita-
tive risk model and second, by analyst assessments of specific credit risks in
order to arrive at a final analyst modified rating. This rating can then be
utilized by the investor, after considering the appropriate sovereign yield spread,
to assess equivalent bond ratings and intrinsic values. The foundation of the
EMS model is an enhancement of Altman's Z-score model, described in the
body of this report, resulting in an EM score and its associated bond rating
equivalent.
The EM score's rating equivalent is then modified based on four critical
factors including: the firm's vulnerability to currency devaluation, its industry
affiliation, its competitive position in the industry and its market to book
equity value. Unique features of the specific bond issue should also be consid-
ered. These modifications are an important complement to the EM score.
The resulting analyst modified rating is compared to the actual bond rating
(if any). Where no agency rating exists, our modified analyst rating is a means
to assess credit quality and relative value both to credits within a country and
to US corporates. These results are listed in Table 1 for Mexican corporates
based on year-end 1994. The results have since been updated as of mid-1996
financials. The implied yield spread based on the analyst modified rating can
be observed from the US Corporate bond market. Steps 1-7 outline the process
391
R. Levich (ed.), Emerging Markee Capital Flows, 391-400.
I 1998 Kluwer Academic Publishers.
Table 1. Mexican corporate issuers - EM scores and modified ratings based on end 1994 w
1.0
tv
EM Bond-rating Modified Ratings
Company Industry score equivalent rating M/S&P/D&P
tTl
~
Aeromexico Airlines -4.42 D D NR/NR/NR ~
Apasco Cement 8.48 AAA A Ba2/NR/NR -~
CCM Supermarkets 4.78 BB- B+ NR/NR/NR s::.
Cemex Cement 5.67 BBB- BBB- Ba3/BB/BB ;:s
Cydsa Chemicals 4.67 BB- B+ NR/NR/NR ~
DESC Conglomerate 4.23 B BB+ NR/NR/NR s::.
:--
Empresas ICA Construction 5.96 BBB BB Bl/BB-/B+
Femsa Bottling 6.37 A- BBB+ NR/NR/NR
Gemex Bottling 5.40 BB+ BB+ Ba3/NR/NR
GIDUSA (Durango) Paper and forest products 4.61 B+ BB Bl/BB-/NR
GMD Construction 4.85 BB B- B3/NR/NR
Gruma Food processing 5.56 BBB- BBB+ NR/NR/NR
Grupo Dina Auto manufacturing 5.54 BBB- BB+ NR/NR/B
Hylsamex Steel 5.51 BBB- BB- NR/NR/NR
IMSA Steel 5.45 BBB- BB- NR/NR/NR
Kimberly-Clark de Mexico Paper and forest products 8.96 AAA AA NR/NR/NR
Liverpool Retail 9.85 AAA A+ NR/NR/NR
Moderna Conglomerate 5.28 BB+ BB+ NR/NR/NR
Ponderosa Paper and forest prod ucts 6.64 A BB NR/NR/NR
San Luis Autoparts 2.69 CCC CCC- NR/NR/NR
Sidek Conglomerate 4.68 BB- B NR/NR/CCC
Simec Steel 4.42 B+ B- NR/NR/CCC
Situr Hotel and tourism 5.17 BB+ B NR/NR/CCC
Synkro Textile/apparel 1.59 CCC- CCC NR/NR/NR
TAMSA Steel pipes 3.34 CCC+ B NR/NR/NR
TELMEX Telecommunications 9.57 AAA AA- NR/NR/NR
Televisa Cable and media 7.29 AA BBB+ Ba2/NR/NR
TMM Shipping 5.34 BB+ BB+ Ba2/BB-/NR
Vitro Glass 5.18 BB+ BB Ba2/NR/NR
NR: No rating. M: Moody's. S&P: Standard and Poor's. D&P: Duff and Phelps. Note: Ratings are for senior long-term foreign debt unless otherwise specified.
EM scores were calculated using fiscal year end 1994 financials.
Source: Salomon Brothers Inc.
Emerging market corporate bonds - a scoring system 393
by which we use the EM score to reach an analyst modified rating. Note that
our analyst modified rating is not constrained in any manner by the so-called
'sovereign-ceiling'. We do advocate, however, factoring in the appropriate cur-
rent sovereign yield spread differential between the emerging market country
and comparable duration US Treasuries (Step 7). In most cases, the full
sovereign 'haircut' should be added to the stand-alone issuer spread. There are
instances, however, where the full sovereign spread is not appropriate because
of unique attributes of the issuer or the issuer's sovereign affiliation. For
example, investor portfolio considerations may swell the demand for a firm in
a key industry, such as telecommunications. Or, the sovereign's huge supply of
outstanding debt relative to investor demand may widen that security's spread
vis-a-vis a more modest supply of a particularly attractive corporate bond. The
resulting spread may indeed be below that of the sovereign as a result of
technical factors rather than fundamental credit characteristics.
Score each bond by its EM score and classify it relative to its stand-alone US
bond rating equivalent. Emerging market corporate credits should initially be
analyzed in a manner similar to traditional analysis of US corporates. This
involves the examination of measures of performance in such a manner as to
establish a rating equivalent of the particular issuer. Instead of using a new
ad hoc system, which may not be based on a rigorous analytical examination
of credit worthiness, we will use an established and well-tested system. Since it
is not yet possible to build such a model from a sample of emerging market
credits, we suggest testing the applicability of a modified version of the original
Z-score model. It is based on a comparative profile of bankrupt and non-
bankrupt US manufacturers, however, our modification can be applied to non-
manufacturing, industrial firms and for private and public entities.
The original Z-score model is based on at least two data sources that make
it difficult to use for all emerging markets corporates: it requires the firm to
have publicly traded equity and it is primarily for manufacturers. In more than
25 years of experience in building, testing and using credit scoring models for
a variety of purposes, the original model has been enhanced to make it applica-
ble for private companies and non-manufacturers. The resulting model, which
is the foundation for our EMS model approach, is of the form:
(Xl) (X 2 ) (X4 )
working retained (X3 ) stockholders
Bond Capital/ earnings/ oper. income/ equity/
rating total assets total assets total assets total liabilities
Major accounting differences between the emerging market country and the
United States must be factored into the data used in the calculations of our
measures. For example, our calculation of retained earnings is based on the
sum of retained earnings and capital reserve, the surplus (deficiency) on restate-
ment of assets, and the net income (loss) for the current period.
The model has been tested on samples of both non-manufacturers and
manufacturers in the US and its accuracy and reliability have remained high.
We have also carefully calibrated the variables and the resulting score with US
bond rating equivalents. These equivalents (Tables 2 and 3) are based a sample
of more than 750 US firms with rated bonds outstanding.
US Average Sample
equivalent rating EM score size
AAA 8.15 8
AA+ 7.60
AA 7.30 18
AA- 7.00 15
A+ 6.85 24
A 6.65 42
A- 6.40 38
BBB+ 6.25 38
BBB 5.85 59
BBB- 5.65 52
BB+ 5.25 34
BB 4.95 25
BB- 4.75 65
B+ 4.50 78
B 4.15 115
B- 3.75 95
CCC+ 3.20 23
CCC 2.50 10
CCC- 1.75 6
D 0.00 14
Source: In-Depth Data Corp. Average based on over 750 US industrial corporates with rated debt
outstanding; 1994 data.
If the firm has high (weak) vulnerability, that is, low or zero non-local
currency revenues and/or low or zero revenues/debt, and/or a substantial
amount of foreign currency debt coming due with little cash liquidity, then the
bond rating equivalent in Step 1 is lowered by a full rating class, such as, BB +
to B +. There is no upgrade for a low (strong) vulnerability and we apply a
one notch (BB + to BB) reduction for a neutral vulnerability assessment.
Telecommunication High A
Independent finance High A
Natural gas utilities High A
Beverages High A
High quality electric utilities High A
Railroads High A
Food processing Mid A
Bottling Mid A
Domestic bank holding Low A
Tobacco Low A
Medium-quality electric utilities Low A
Consumer products industry Low A
H.G. diversified Mfg./conglomerate Low A
Leasing Low A
Auto manufacturers Low A
Chemicals Low A
Energy Low A
Natural gas pipelines High BBB
Paper/forest products Mid BBB
Retail Mid BBB
P&C insurance Mid BBB
Aerospace/defense Mid BBB
Information/data technology Mid BBB
Supermarkets High BB
Cable and media High BB
Vehicle parts High BB
Textile apparel High BB
Low-quality electric utilities MidBB
Gaming Mid BB
Restaurants MidBB
Construction MidBB
Hotel/leisure MidBB
Low quality manufacturing MidBB
Airlines Low BB
Metals High B
Source: Adapted from six-month credit quality overview, Salomon Brother Inc., January 18, 1995.
Step 3's rating is adjusted up (or down) one notch if the firm is a dominant
(or not) company in its industry or if it is a domestic power in terms of size,
political influence and quality of management. It is possible that the consensus
competitive position result is neutral (no change in rating).
Emerging market corporate bonds - a scoring system 397
From time to time we consider modifying the system to consider other impor-
tant factors. One such ingredient that we now feel is relevant to evaluating the
credit risk of a company is the market to book value of the firm's equity.
Despite the inefficiencies in emerging markets equity valuations, a company
whose stock is valued highly by the financial community can usually borrow
more easily and raise new equity or sell assets at better prices than one which
is being discounted by investors. Since the corporate bonds of emerging market
companies are, by rating agency definitions, almost all non-investment grade,
their yield and volatility patterns at times are more correlated with equity
market activity than are investment grade corporates.
There are two ways that we can introduce a market value of equity factor
into our system. First, a new variable reflecting the market to book value of
equity, or some similar measure, could be added to the existing four variables.
Since the original data base used to construct the EM system did not contain
that variable, it is impractical to re-estimate the equation using a new data
base. The second approach is to add an additional phase to our modified
equivalent bond rating process - one that incorporates a comparison of the
bond rating equivalent using the book value of equity to total liabilities (X4 in
the model) versus the same variable with the market value of equity (number
of shares outstanding times the stock price) substituted for the book value.
This second approach is what we actually have done. The procedure we
followed is to calculate the bond rating equivalent in the traditional manner,
which involves the initial bond rating based on the multi-variate model, and
modifications based on currency devaluation vulnerability, industry affiliation
and competitive position. The final phase now is to compare the bond rating
equivalent using book equity to the rating equivalent using the market value
of equity.
If the two systems give the identical rating or are different by only one
notch, then the modified rating is unchanged. If, however, the two versions
result in a two-notch differential, then we increase or decrease the final modified
rating by one notch. Finally, if the difference is a full rating class (three notches)
or more, the modified rating is changed by two notches.
We first applied this further modification in 1996. Based on six months
ended June 1996 data, 17 of 38 Mexican firms had higher bond rating equiva-
lents when using the market value compared to the book value equity. Of the
17 firms, six had the same modified rating since the difference was one notch;
one had a one notch upgrade and 10 had a two notch upgrade. Seven firms in
total had lower EM scores using market value compared to book value of
equity, with five resulting in a one-notch downgrade and the other two not
changed. Fourteen firms had the identical rating using the book and market
value of equity measures.
The impact of using the market value of equity versus the book value can
have a major impact in the final modified rating for a company. Such an
398 E.1. Altman et al.
impact reflects the often inefficient market for Mexican companies' equity. The
volatility of the Mexican peso and its impact of the Mexican equity market
can mask the intrinsic values of Mexican equities. In addition, inflation account-
ing can distort the book value of equity of Mexican firms because of income
statement non-cash charges and the consequent changes in retained earnings
and stockholders equity. We believe that despite these inefficiencies, the
Mexican equity market has rallied sufficiently to begin to incorporate the
market equity value in our model.
If the particular debt issue has unique features, such as collateral, a bona fide,
high-quality guarantor, or a restricted cash trust to payoff bondholders, then
the issue should be upgraded accordingly.
An important distinction must be made between this model and the original
Z-score model. First, this model, referred to by Altman (1993)1 as the Z"-score
model, is applicable for non-manufacturers and private firms in addition to
manufacturers and public firms. Second, the model applies our analysts' sub-
jective measures of credit strength as outlined in Steps 2-4, and the relationship
between the equity value based on market prices vs. book values.
It is important to remember that the stand-alone rating generated in Step 1
is based on the specific operating performance and financial characteristics of
the company. The analyst modified rating likely will change with the operating
environment within which a company functions. For US firms in mature
industries, this environment does not typically change dramatically. For
Mexican firms, however, their respective operating environments are subject
to major changes, such as the peso crisis of December 1994. Outlined below
1 See E. Altman, Corporate Financial Distress and Bankruptcy, 2nd edition, John Wiley & Sons.
N.Y., 1993, Chapter 8.
Emerging market corporate bonds - a scoring system 399
Accounting anomalies
The high inflation environment in Mexico precludes Mexican firms from the
standard credit analysis applied to US companies (the original Z-score model
was for US companies only). For example, the impact of noncash foreign
exchange losses on pre-tax earnings is dramatic for Mexican firms. Analysis of
retained earnings and the book equity, and therefore leverage ratios of Mexican
firms, is subject to more careful analysis and appropriate adjustments.
Government intervention
The Mexican Government has recently pro-actively supported certain sectors
of the Mexican economy in order to prevent default. Examples of this include
the support of the banking system through programs like Procapte; the facilita-
tion of providing short-term financing for Grupo Sidek (the Mexican conglom-
erate which defaulted, and subsequently made payment, on its commercial
paper); and the government's renegotiation of construction sector concessions.
Despite its recent support of the private sector, the Mexican Government's
continuing presence in crisis situations cannot be assumed with certainty.
the late 1980s and subsequently privatized, or they were controlled by wealthy
families for decades. Therefore, most Mexican Eurobond issuers have typically
dominated their respective markets. With the advent of NAFTA, and the
economic weakness brought about since the devaluation, we expect Mexican
firms will see greater competition and shrinking market shares in the future.
Our analyst modified rating embodies these particular Mexican credit fea-
tures. Together with timely sovereign and economic research, we can adjust
the analyst modified rating to incorporate changes in the Mexican economic
and corporate landscape.
Applying the analyst modified rating to the current market
CONCLUSION
The EMS model should be used to assess relative value among credits in the
inefficient trading environment for emerging markets credits. The model is
flexible, allowing for future modifications depending on the operating and
financial environment and sovereign risk. Early empirical results indicate that
the model has been extremely accurate and continued testing and use would
seem to be a reasonable conclusion.
2The four Mexican defaults involved Grupo Synkro, Situr, Sidek and GMD. The first three
defaulted in 1996 and the latter in 1997.
OLIVER HART, RAFAEL LA PORTA DRAGO,
FLORENCIO LOPEZ-DE-SILANES and JOHN MORE1.t
* Harvard University, t London School of Economics
ABSTRACT
There is widespread dissatisfaction with existing bankruptcy procedures around the world.
We propose a fast and cheap bankruptcy mechanism which leaves little room for court
discretion and meets the desiderata of a good procedure. Our scheme has four essential
building blocks. The first part consists of the reorganization offers for the firm and/or its
pieces. Reorganization offers can be made not only in cash but also in non-cash securities,
which facilitates the best allocation of assets in the presence of capital market imperfections.
Additionally, since reorganization offers may be placed by both firm insiders and outsiders,
our mechanism preserves the ex-ante bonding role of debt by penalizing poor management
while granting debtor protection when financial difficulties are the result offortuitous events.
In the second part of our scheme, the insider cash auction, the various claims on the firms'
assets are transformed into one common security, reorganization rights (RRs), thereby
aligning the objectives of all claimants. The insider cash auction has the goal of offering
preferential treatment to existing claimants, protecting them in cases where bad functioning
or non-competitive auction markets may allow third parties to reap benefits at their expense.
The third part of our procedure consists of a public cash auction for RRs, which reduces
the probability that credit constraints will result in a bankruptcy allocation that is unfair
to claimants. The final part ofthe procedure consists of a simple vote by the newly assembled
group of shareholders on which reorganization offer to accept.
INTRODUCTION
1 We are grateful to Lemma Sembit, Andrei Shleifer and the participant of the conference on
"The Future of Emerging-Market Capital Flows" for their comments. We are also grateful to
Fernando Salas for helpful conversations. Part of this paper draws on our proposal to modify
bankruptcy law in Mexico (Hart et al., 1995).
2 The liquidation procedure involves closing down the firm's operations and organizing a cash
auction for its assets. The receipts from the auction are distributed among claimants according to
absolute priority.
401
R. Levich (ed.), Emerging Market Capital Flows, 401-419.
~ 1998 Kluwer Academic Publishers.
402 O. Hart et al.
be able to raise the cash necessary to buy the firm. Capital markets imperfec-
tions may have dire consequences if firms are inefficiently dismantled and
their assets sold off cheaply at fire sale prices.
2. Reorganization (e.g., Chapter 11 in the US):3 Existing reorganization pro-
cedures address two questions simultaneously: (1) Who should get what?;
and (2) What should be done with the firm? The coupling of these issues
introduces conflicts of interest and may cause assets not to be put to their
most productive use. For example, a senior creditor may press for a speedy
liquidation (since he or she will then be paid off for sure), whereas junior
claimants may encourage protracted bargaining (since they enjoy upside
changes in the firm's value, but not the downside risk).
3 The reorganization process encourages creditors and shareholders to bargain about the future
of the company. A plan is implemented if it receives approval by a suitable majority of each
claimant class.
4 The resolution of a bankruptcy procedure may take anywhere from 3 to 7 years in Mexico
and even decades in Thailand.
Proposals for a new bankruptcy procedure 403
5 During the period between 1993 and July 1995, the total number of decisions made by the
Court of Arbitration in all of Russia totaled 613 (data provided by the Organizations Department
of the Court of Arbitration). This number is very low compared to OECD countries. In England,
for example, the number of court decisions for a similar period of time would be close to 15,000.
Another example is Mexico, where during the period 1989-1992, the average annual number of
firms filing for bankruptcy in Mexico City was less than 120. There have been virtually no major
Mexican bankruptcies since AHMSA (1986) and Aeromexico (1988), both of which were govern-
ment-owned firms. The number of firms successfully reorganized through court procedures is
surprisingly low. In fact, only thirty Mexican reorganizations have taken place in court in the last
twenty years.
6 In fact, the experience of the few litigated cases of firms filing for bankruptcy in Mexico shows
that debtors reap the benefit of the law's weaknesses (Lipkin, 1995; McHugh, 1995).
7 The procedure is a (significant) modification of the one developed by Aghion et al. (1992,
1995) and draws on the proposal made for bankruptcy reform in Mexico (see Hart et al., 1995).
404 O. Hart et al.
The advantages of this proposal can be divided into eight different areas:
1. RRs eliminate conflicts between different classes of claimants regarding the
future of the firm. All holders of RRs are equal and have only one objective:
to maximize the value of firm.
2. The ability to make offers in cash and/or non-cash securities, for the firm
as a whole or for parts of it, makes it more likely that the assets of the firm
will be put to their most productive use.
3. The procedure pays particular attention to capital market imperfections.
The proposal allows outsiders to participate in the process by placing
reorganization bids for the firm or its parts. This feature ensures that debt
acts as a bonding device since it provides a tool to discipline poorly perform-
ing management. Debt can be a particularly valuable way of mitigating the
corporate governance problem and imposing discipline on management in
8 It is not suggested that our procedure, if adopted, should be mandatory. Anybody who wishes
to deviate from it and design their own procedure should be aJlowed to do so.
Proposals for a new bankruptcy procedure 405
Rcorpni ...atinn Process Swts IReorpniulion Off.:" AnnounceJ Holders of RRJi VOle Reorganiution Plan
Figure 1. Schedule for the different stages of the proposed reorganization procedure (months).
preserves absolute priority while taking into account that capital market imper-
fections may be pervasive. Figure 1 provides a summary of the discussion.
A debtor who is unable to meet its obligations may seek the protection offered
by the bankruptcy law. Alternatively, a firm may be pushed into bankruptcy
by unpaid creditors. Either way, our procedure is triggered by the declaration
of a firm's bankruptcy. The initial step involves canceling all debts and placing
an automatic stay on the firm's assets that prevents secured creditors from
seizing their collateral. 9 Claimants, however, do not walk away empty handed.
Instead, as described below, they receive options to participate in the reorgani-
zation process. The essential feature of the process is that the firm is transformed
into an all-equity company.
In the context of an emerging market, it may be optimal to allow the debtor
to retain the power to manage, dispose of his estate and carryon his business
pending reorganization under the supervision of a court appointed receiver.
Alternatively, the firm's day-to-day operations may be conducted by the
appointed receiver.lO
The receiver analyzes the financial statements of the firm to determine whether
the level of its assets is sufficient to cover reorganization expenses. If this is the
case, two parallel processes are started:
9 An application by a debtor for reorganization can be made even after foreclosure proceedings
have begun.
10 Lsck of specialized managerial talent as well as the absence of a receivership profession, as
in England, may make it hard to find qualified individuals to substitute for their existing pending
reorganization.
Proposals for a new bankruptcy procedure 407
fractions thereof, within a ninety day period. Reorganization offers can be made
in any combination of cash and non-cash securities. In a non-cash bid someone
offers securities in the post-bankruptcy firm. Some illustrative examples of
acceptable bids are:
the old management team proposes to continue running the firm and offers
claimants equity in the post-bankruptcy firm;
the same financial arrangement can be offered by an outside management
team;
management (old or new) may offer claimants a combination of shares and
bonds in the post-bankruptcy firm.
Reorganization rights
The judge settles the list of admitted claims which result in the identification
of n classes of creditors who are owed (in total) the amounts D 1 , , D n ,
respectively, with class 1 having the most senior claims, class 2 the next most
senior claims, and so on. The firm's shareholders form the (n + 1)th class, with
a claim junior to all others.
Having identified these classes, the receiver can proceed to issue 100 reorgani-
zation rights (RRs), each one representing a 1% ownership stake in the post-
bankruptcy (all-equity) firm and conferring the right to one vote at the meeting
of holders of RRs that will later be assembled to decide the future of the firm.
11 The fact that contentious claimants do not participate in the latter stages of the procedure is
not too serious, since they probably share the same objectives as the rest of the claimants of their
class and would probably have voted in the same way as recognized creditors when deciding on
the future of the firm. Contentious claims may be sorted out in court as the reorganization
procedure continues, or they may be set aside and dealt with, together with late claims, once the
company has emerged from insolvency. At this stage, validated claims can be discharged with cash
and/or securities in line with the average receipts of other former claimants of the same class.
12 The receiver has the task of combining partial offers to assemble offers for the whole firm.
To package a whole bid it may be necessary to request supplementary bids for parts of the firm.
408 O. Hart et al.
Insiders are allowed to exercise their options to buyout more senior claimants
(insider auction)
Once the 3 months are up, the receiver makes public all outside reorganization
offers (described above) placed for the firm. Insiders are given an additional
month to analyze the reorganization offers received for the firm and to decide
whether they are going to exercise their options to purchase RRs. The insider
auction is designed to assure that no member of a class j is allowed to retain
cash unless all members of class j - 1 have been fully paid. Our procedure
deviates from Bebchuk's (1988) scheme only in the case that some but not all
members of a class decide to exercise their option to buy RRs.13
A detailed description of the insider auction is provided in Appendix 1. Here
we give a rough idea and some illustrations. We start the insider auction by
allowing the most junior class of claimants (class n + 1) to exercise their option
to buy RRs at the insider price of P,,+l (call this round 1). If shareholders
exercise all their rights to buy RRs, then the insider auction generates proceeds
that are sufficient to retire all the existing debt of the firm. The insider auction
then terminates. Of course, proceeds from the insider auction are going to be
insufficient to retire all the debt if shareholders choose to exercise only some
of their options. Suppose shareholders exercise their right to buy q" + 1 < 100
RRs. Then the funds generated (equal to C,,+l = P,,+lq,,+l) are handed to class n
claimants in return for an obligation to supply q" RRs to class n + 1. The
class n claimants are then compelled to use the cash received to exercise their
options to acquire RRs at the inside price p" from class n - 1. In turn class
n - 1 claimants use the cash they receive to exercise their options to acquire
RRs from class n - 2.
13 Bebchuk's scheme has the unpalatable feature that junior claimants may receive cash from
the insider auction even when senior claimants have not been fully paid. Our scheme avoids such
an outcome.
Proposals for a new bankruptcy procedure 409
Thus, there is a cascading effect: cash moves up the hierarchy from junior
to senior claimants and RRs move down the hierarchy from senior claimants
to junior claimants (cash and RRs are always transferred on a pro rata basis).
At the end of round 1, either class n - 1 claimants have been fully paid off
in cash or they have not. If they have, the insider auction ends. If not, in a
second round of the insider auction members of class n are given an opportunity
to exercise any remaining options to buy RRs from class n - 1. We stop when
we reach a class k of claimants such that everybody senior to them has been
fully paid off in cash. (If k = 1, no class is fully paid off.)
To illustrate the proposed procedure consider a firm which has three classes
of creditors that are owed $100 each; that is Db D2 and D3 are all equal to
$100 (Figure 2 illustrates this example). The receiver has solicited reorganiza-
tion offers in cash and/or non-cash securities for the whole firm or its parts.
Assume that the best reorganization offer received is perceived to be worth
$250. The receiver issues 100 RRs and asks shareholders if they are willing to
exercise their right to buy RRs at a unit price of $3. No shareholder is willing
to exercise her option to buy RRs in the insider auction since the value of the
best reorganization offer falls short of the face value of the debt of the company.
In contrast, class 3 claimants would like to exercise their option to purchase
all RRs since they are able to do so at a unit price of $2. Class 2 claimants
receive a total of $200 from class 3 claimants in exchange for the promise to
deliver to them the 100 units of RRs. In turn, class 2 claimants are compelled
to exhaust the cash received by calling RRs held by class 1 claimants at a call
price equal to $1 each. Class 2 claimants, therefore, purchase the 100 RRs held
by class 1 creditors in exchange for the $100 that are owed to class 1 claimants
REORGANIZATION OFFERS:
Assume the best reorganization offer received/or the!"m is worth $250.
INSIDER CASll AUCTION:
Step 1: Shareholders Do NOI Exercise the Option to B1I)' RRs since the firm is worth $250.
Step 2: Class 3 Exercise their Option to Buy 100 RRs:
$200 $100
Figure 2.
410 o. Hart et al.
and retain $100 in cash for themselves. After having paid class 1 claimants in
full for $100, class 2 claimants hand over all RRs to class 3 claimants. This
concludes the inside auction part of the procedure as both class 1 and 2
claimants have been fully paid.
Note that in the above example, each class 3 claimant needs to raise cash
only to cover his or her pro rata share of Dl + D2 This feature of the proposed
procedure may be especially attractive in emerging markets to the extent that
it mitigates liquidity constraints. 14 Although the proposed procedure alleviates
liquidity constraints it probably does not eliminate them. This is the motivation
for calling a public cash auction to settle the final allocation of RRs.
The receiver will rank the cash bids from highest to lowest forming a demand
schedule for RRs by outsiders. To form the supply curve of RRs the receiver
will rank the RRs held by insiders by their call price. The call price of an RR
held by a class 1 claimant is Pi + 1. We arbitrarily assign a call price of infinity
to any RR held by a shareholder (class n + 1) as they are the residual claimants.
Figure 3 illustrates the discussion. The figure corresponds to a situation where
ql RRs (held by class 1) have a call price of P2, q2 (held by class 2) have a call
price of P3 and the remainder, 100 - ql - q2 (held by class 3) have a call price
of P4.
The receiver will determine the equilibrium price for RRs where supply and
demand intersect (P* = P4)Y Suppliers of RRs will be paid their respective call
prices and in this fashion will receive satisfaction of their claims. The surplus
cash generated by the difference between P* and the call price paid to suppliers
(area A) is used by the receiver to payoff impaired claimants in accordance
with absolute priority. In Figure 3 this means that nonparticipating class 2
claimants (i.e., those who did not participate in the inside auction) are paid off
with the surplus cash first, followed by nonparticipating class 3 claimants.
To illustrate the usefulness of the outside public auction, consider once more
the firm that was presented in our previous example, keeping all assumptions
in that example but imagining that class 3 claimants are unable to raise any
of the $2 in cash needed to bid for RRs. However, the class 2 claimants are
not liquidity constrained. In the absence of an outside cash auction, all RRs
will end up in the hands of class 2 creditors who will have paid $100 for
securities that are valued at $250. The receiver hands the 100 RRs to the class 2
claimants and pays off class 1 debt in full. Therefore, the supply of RRs is
14 The fact that the insider auction is initiated at the top may also mitigate liquidity constraints.
If, for example, all shareholders exercise their options then more senior claimants are not required
to raise cash. This feature may make it less likely that liquidity constraints become binding.
15 If there are multiple equilibria, the receiver chooses the smallest price at which supply equals
demand. An alternative allocation rule would use price discrimination to capture the entire
consumer surplus. However, price discrimination may distort the incentives of outsiders to bid high.
Proposals for a new bankruptcy procedure 411
p Supply
Demand
100 Q
Figure 3.
perfectly elastic at $2 (i.e., the call price for class 2 claimants) for quantities
that are smaller than 100 RRs and becomes perfectly inelastic at 100 RRs.
To continue with the example, the receiver assembles the demand for RRs
by inviting outsiders to place bids for them. For simplicity, assume that demand
is perfectly elastic at $2.5 and that the receiver is presented with bids for 100
RRs at $2.5 dollars each. Outside bidders pay the equilibrium unit price of
$2.5 in exchange for each of the 100 RRs. To meet the demand for RRs, the
receiver calls the 100 RRs held by class 2 claimants, each at a price of $2.
Class 2 creditors are now paid in full. The surplus cash of $50 is prorated
among all class 3 claimants of the firm. Without the public cash auction, class 3
creditors facing liquidity constraints would have been penalized since the RRs
would have been left in the hands of class 2 claimants. Through the public
cash auction, class 3 creditors fare as well as in our previous example where
they are not liquidity constrained.
In more realistic settings, some members of the impaired class, but not all,
may choose to exercise their options. The appendix illustrates how an efficient
outside market for RRs makes it possible for all members of the impaired class,
whether they are liquidity constrained or not, to fare equally well.
The reader may wonder why the public cash auction does not make the
inside auction superfluous. The insider auction is unnecessary if outsiders bid
the true value of the firm in the public cash auction. However, imagine that
412 O. Hart et al.
REORGANlZATION OFFERS:
Assume the best reorganizDtkm offer received for thefirm is worth $150.
INSIDER CASH AUCTION:
SUp 1: Shareholders Do Not Exercise the Oplion 10 BIIJI RRs since the firm is WOI1h $150.
Step 2: Class J Cannot Exercise lheir Oplion 10 Bvy RRs as they are liqllidily constrained.
Stq J: Sen/or, Exercise lhelr Option 10 BllJlloo RRs:
$100
lOORRs
@I
Figure 4.
outsiders are able to place winning bids at only a fraction of the true value of
an RR possibly as a result of lack of competition in the public cash auction
market. Then, in the absence of an insider auction, outsiders would be able to
buyout the RRs of liquidity-constrained claimants at a fraction of their true
value. The insider auction allows a shareholder, for example, to buy an equity
stake in the post-bankruptcy firm by bidding her pro rata share of the debt
rather than by exceeding the bid placed by outsiders. Similarly, the insider
auction allows a member of the most senior class of creditors, for example, to
retain her RRs by bidding more than $0 rather than by surpassing the bid
placed by outsiders (assuming the outside bid is less than her call price).16
The allocation of RRs eliminates conflicts across classes of claimants and gives
new investors in the surviving entity the sole goal of maximizing the firm's
value. Holders of RRs can now proceed to meet and vote, on the basis of one-
16 We have constructed the supply curve for RRs for the public auction under the assumption
that no holder of RRs wishes to relinquish them for less than their call price. This need not be the
case, however. Consider Figure 3. One could imagine that some class 3 creditors, who are either
risk averse or relatively short of cash, may be willing to part with their RRs for a price between
P3 (the price they paid to get them) and P4 (their call price). The procedure could easily be modified
to accommodate this. Creditors could be asked to communicate to the receiver the lowest price at
which they are willing to sell and the receiver could construct a modified supply curve that
incorporates this information. The receiver would then use this modified supply curve to compute
the equilibrium price. Suppliers of RRs would be paid the smaller of the equilibrium price and
their call price.
Proposals for a new bankruptcy procedure 413
Clus 1
Debt=SO
RRs=O
Cash=Sloo
S2oo Class 1
Debt=SO
J 100 RRs RRs=O
Outside IDvestors Reeeiver @2 Casb=$loo
RRs=loo Casb=$SO
Cash 5-250 Class 3
S2S0
Debt=Sloo
looRRs RRs=O
Cash=$O
Shareholders
Shares=loo
RRs=O
SettUnmrt:
Class 1
Debt=SO
RRs=O
Casb=$loo
Class 1
Debt=SO
RRs=O
Cash=$loo
Receiver
Cash=$O Class 3
SSO Debt=SO
RRs=O
Cash=$SO
Shareholders
Shares=O
RRs=O
Cash=SO
Figure 4. (Continued.)
17 Where there are more than two competing reorganization offers, there are several possible
ways in which the holders of RRs may select among competing restructuring plans. One possibility
is to conduct two polls. In the first poll, all restructuring proposals are submitted to a vote, while
only two restructuring proposals which have received the most votes on the first round will be
voted on in the second round.
414 O. Hart et al.
REORGANIZATION OFFERS:
A.r.sume the best reorganization offer received/or thefirm is worth $350.
INSIDER CASH AUCTION:
Step I: Shorehclders Exercise the Option to BIIJI60 RRs:
$20
10RRs
@2
Figure 5.
CONCLUSION
REORGANIZATION OFFERS:
Assume the best reorganization offer receivedfor the firm is worth $250.
INSIDER CASH AUCTION:
Stq.l: Shareholders Do Not Exercise the Option to Buy RRs since theftrm~' worth $250.
Stq 2: Class 3 Cannot Exercise their Option to Buy RRs as they are liquidity constrained
Stq 3: Semors Exercise their Option to Buy 100 RRs:
$100
IOORRs
@1
Figure 6.
and non-cash reorganization offers for the firm or its parts. When measured
against existing reorganization procedures (Chapter 11), our scheme is less
cumbersome, since the complicated voting systems of these procedures is substi-
tuted by a simple vote by a homogenous class of security holders whose sole
objective is the maximization of the firm's value. Finally, the public cash auction
for RRs increases the likelihood that the claimants of the firm will receive a
fair value in the presence of capital market imperfections.
Many specific characteristics and rules of our proposal may be adapted to
suit particular concerns or legal issues in different countries, as the results
achieved by our scheme rest on only four essential building blocks. The first
part of our scheme consists of the reorganization offers for the firm and/or its
pieces. Reorganization offers can be made not only in cash but also in non-
cash securities, which facilitates the best allocation of assets in the presence of
capital market imperfections. Additionally, since reorganization offers may be
placed by both firm insiders and outsiders, our mechanism preserves the ex-
ante bonding role of debt by penalizing poor management while granting
debtor protection when financial difficulties are the result of fortuitous events.
In the second part of our scheme, the insider cash auction, the various claims
on the firm's assets are transformed into one common security, reorganization
rights (RRs), thereby aligning the objectives of all claimants. The insider cash
auction has the goal of offering preferential treatment to existing claimants,
protecting them in cases where bad functioning or non-competitive auction
markets may allow third parties to reap benefits at their expense. The third
part of our procedure is the public cash auction for RRs, which reduces the
416 O. Hart et al.
REFERENCES
Aghion, P., O. Hart, and J. Moore (1992). The economics of bankruptcy reform. Journal of Law,
Economics and Organization, 8, 523-546.
Aghion, P., O. Hart and 1. Moore (1995). Insolvency reform in the UK: a revised proposal.
Insolvency Law and Practice, 11, 67-74.
American Bar Association (1989 and 1993). Multinational Commercial Insolvency. Wisconsin: MG
Publishing.
Bebchuck, L. (1988). A new approach to corporate reorganizations. 101 Harvard Law Review,
775-804.
Center for International Financial Analysis and Research, Inc. ( 1994). International Accounting and
Auditing Trends.
Cutler, D. and L. Summers (1988). The cost of conflict resolution and financial distress: evidence
from the Texaco-Pennzoillitigation. Rand Journal of Economics, 19, 157-172.
Gilson, S., K. John and L. Lang (1990). Troubled debt restructurings: an empirical study of private
reorganization of firms in default. Journal of Financial Economics, 26, 315-353.
Hart, 0., R. La Porta Drago, F. Lopez-de-Silanes, J. Moore and F. Salas (1995). Proposal for a New
Bankruptcy Procedure in Mexico. Mimeographed, Harvard University.
John, Kose (1993). Managing financial distress and valuing distressed securities: a survey and
research agenda. Financial Management, 2, 60-78.
Keefer, P. and S. Knack (1993). Why Don't Poor Countries Catch Up? A Cross Country National
Test of an Institutional Explanation. Working paper No. 60, Center for Institutional Reform and
the Informal Sector, University of Maryland at College Park.
La Porta Drago, R., F. Lopez-de-Silanes, A. Shleifer and R. Vishny (1996). Law and Finance. NBER
working paper #5661.
Lipkin, A. (1995). Legal framework for bankruptcy and reorganization proceedings in emerging
markets. In M. McHugh (ed.), The 1995 Bankruptcy Yearbook and Almanac. New Generation
Research, Inc. Boston, MA: George Putnam, III, publisher.
Martinez, Gabriel (1993). Propuesta de Reforma a la Ley de Quiebras y Suspension de Pagos.
SECOFI, Mexico, D.F. Mexico.
McHugh, M. (ed.) (1995). The 1995 Bankruptcy Yearbook and Almanac. New Generation Research,
Inc. Boston, MA: George Putnam, III, publisher.
Price Waterhouse (1995). Bankruptcy Law in the Russian Federation. Report to the Russian
Federation, Moscow, Russia.
Proposals for a new bankruptcy procedure 417
ApPENDIX 1
ApPENDIX 2
Maintain the assumptions of our first example in the text where a firm has a
total debt of $300 equally divided among three classes of creditors. As in that
example the receiver solicits reorganization offers in cash and/or non-cash
securities for the whole firm or its parts. Assume now that the best reorganiza-
tion offer for the firm is $350 and that some shareholders are liquidity con-
strained and therefore cannot place bids for RRs. As before, the receiver issues
100 RRs and proceeds to conduct the insider cash auction.
Assume that only 60 shareholders (out of 100) are able to exercise their options
for RRs since liquidity constraints bind for the rest of them. Each exercising
418 O. Hart et al.
The receiver organizes a public cash auction and 100 bids come in at a price
of $3.5 each resulting in a perfectly elastic demand for RRs at that price. As a
result of the insider cash auction, the supply of RRs is perfectly elastic up to
the point where it reaches 40 RRs at which point it becomes perfectly inelastic,
as there are no more RRs in the possession of creditors of the firm. The receiver
accepts 40 bids and calls the 40 RRs still held by class 3 claimants at a price
of $3 per unit. The receiver gives $120 in cash to class 3 claimants who have
been fully paid off, since their debt was $100 and they paid $20 from their
pockets to exercise their options. Finally, the receiver has $20 left which he
distributes among the 40 non-participating shareholders, each receiving $0.50.
Proposals for a new bankruptcy procedure 419
Exchange rate changes have two impacts on the emerging market firm's credit-
worthiness, one that can be generalized and the other that cannot. First, exchange
rate changes can affect the local currency value of the US dollar debt. Real
appreciations, which generally occur during fixed nominal exchange rates regi-
mens, have no impact on the local currency value of the US dollar-denominated
debt. Real devaluations, typically due to a large nominal devaluation of the local
currency, increase the local currency obligation of the firm's US dollar-denomi-
nated debt, raising the firm's debt burden in local currency. The accounting
response to the devaluation records the full increase in the local currency value
ofthe US dollar obligation as part of income for the current period, thus generat-
ing serious accounting losses for the firm in the period immediately following any
devaluation. This partial effect of exchange rate changes on creditworthiness,
common to all firms with foreign currency debt, is easy to measure and, everything
else fixed, either does not affect the firm's creditworthiness (and possibly improves
it slightly) in the case of a real appreciation, or worsens the firm's creditworthiness
in the case of a real devaluation.
Second, and often less carefully analyzed, is the fact that exchange rate
changes affect the firm's ability to generate current and future cash flows, and
thus also affect the ability of the firm to repay its debt. The impact of this
effect on the firm's creditworthiness, unfortunately, is not generalizable and
requires a careful analysis of the individual firm's cash flow sensitivity to foreign
currencies values, as it may be either positive or negative as well as large or
small, for a given exchange rate change. As a first approximation it is useful
to know the current extent of the firm's foreign currency activities such as
exporting, foreign sales, and importing, as well as the currencies in which these
activities are based. This provides some idea of the likely current period impact
of an exchange rate change on the current cash flows of the firm. However, it
is also necessary to have some understanding of the structure of the firm's
input and output markets to determine how the firm's ability to generate cash
flows will be affected by the exchange rate change.
This is not an easy task, it involves a detailed analysis of each firm carried
out as part of the 'economic audit' that should be done diligently by the credit
analyst. While accounting data may provide some help for determining the
current foreign currency cash flow situation of the firm, it provides little help
in determining the changes in these cash flows in response to a devaluation. A
simple cash flow model of the firm must be developed in order to estimate the
effects of exchange rate changes on the firm's ability to generate cash flows,
both currently and in the future periods.
The overall exchange rate exposure of the emerging market firm then is a
combination of these two effects: a direct and easily observable valuation
impact on the US dollar liability and more difficult to determine impact of the
exchange rate change on the firm's ability to generate cash flow. Ultimately
the creditor is worried about the ability to the firm to meet debt service
requirements in various states of the world, and such an evaluation requires
analysis of the firm's ability to generate cash flows relative to its obligations
424 G.M. Bodnar
for different exchange rate realizations. Thus while the credit analyst should
be interested in the current foreign currency cash flow coverage of the interest
expense, she should also be concerned about how this cash flow coverage will
change when the exchange rate changes. 1
While the effect of exchange rate exposure effect may be difficult to evaluate,
its impact on creditworthiness in reality is not difficult to observe. Compare
the results of two Mexican companies that are part of Mexican holding com-
pany Grupo Sidek following the December 1994 Mexican peso devaluation.
One company, Situr, which in involved in real estate and hotels, experienced
significant accounting losses as a result of the devaluation (as did all Mexican
firms with dollar debt as the entire capital loss on the dollar debt is passed
through income for that period), but more importantly, it also suffered a serious
decrease in it ability to generate cash flow because of the devaluation. As a
result, it ended up defaulting on some of its dollar denominated debt earlier
this year (1996). On the other hand, Simec, which is a steel company, also
experienced a significant accounting loss due to the devaluation; however,
because of its ability to export its production, and the competitive nature of
the world market for steel, the devaluation increased its ability to generate
cash flows. In fact for the first quarter of 1995 Simec reported record results,
and has had little trouble making payments on its debt. Prior to the peso
devaluation Simec had been rated two notches below Situr.
Thus determining the credit rating of firms from emerging markets is a
formidable task. Aside from the issue of sovereign risk, credit analysis of firms
from emerging markets face two additional forms of complexity. Foreign firms
do not always report financial figures based upon US GAAP and financial
ratios based upon foreign GAAP may not provide an accurate rankings of the
firm's credit risk when compared to guidelines based upon US GAAP-based
financial ratios. Thus whenever possible foreign firms' figures should be made
consistent with US GAAP measurements before comparing financial ratios.
However, efforts to adjust financial ratios should not be done at the expense
of a careful analysis of the exchange rate exposure of emerging market firms
and the implications that exchange rate depreciations have for the future cash
flows of the firms. Given the increased importance of exchange rates for firms
in emerging markets, the issue of exchange rate exposure in credit analysis,
while less important domestically, becomes an issue of primary importance
when considering the creditworthiness of emerging market firms.
REFERENCES
Sondhi, A.C. ( 1995). Analyzing the credit risk of emerging market debt securities. In: Credit Analysis
of Nontraditional Debt Securities (A.C. Sondhi (ed.)). Charlottesville, VA: AIMR.
1 Such an analysis would also want to take into account the cash flow implications of any
currency hedging that the firm is doing.
WILLIAM J. CHAMBERS
Standard & Poor's
Standard & Poor's believes that there is a strong potential for modeling to
complement more detailed assessments of credit quality. We do not see this as
competition for traditional, full-fledged debt ratings, since models serve to
analyze one portion of the wide array of information surveyed in producing a
full rating. While potentially easier for insurance companies and banks, where
there is more homogeneity across the industry, the task is much more difficult
for corporates where sectoral differences make using a single standard regres-
sion equation more problematic, particularly if the model encompasses only
financial information. In different industries, various factors take on greater or
less importance and should weigh more or less heavily on the determination
of creditworthiness.
Most of these general issues regarding modeling have been discussed pre-
viously, and the focus of this discussion will be on the applicability of modeling
to credit risk determination in emerging markets. The Altman/Hartzell paper
employs something of a hybrid, using the purely quantitative model and then
consciously adjusting the results by applying certain qualitative factors.
well, though, as discussed below, the rapid and frequent changes in its values
suggest that it might be more useful to short-term traders than longer-term
investors.
Even more fundamentally, there are several essential questions that must be
addressed.
1. Is it possible to model credit risk? We believe that it is, but we're still some
considerable distance from that objective.
2. Is it possible to model emerging market credit risk? If modeling is difficult
in mature markets, it represents a much harder objective to attain in emerg-
ing markets. Almost by definition, the length of experience on which to
build a credible model is much shorter; the similarities to better established
markets is still doubtful; more volatile business conditions and the varied
elements which might affect creditworthiness in these markets are still much
more poorly understood.
3. Finally, and perhaps the most important of all these questions, is it possible
to develop a single comprehensive model which would deal with all emerging
markets, without having to be completely recalibrated each time for each
country and each industry? The present model makes an attempt at this
objective, but my view is that it is currently not possible to achieve this
objective.
used as arguments in the equation apart from any real impact on the firm's
operations or underlying creditworthiness. If the inflation rate thus affected
all modeled companies equally, it could be argued that, apart from obvious
differences from results obtained in the USA in absolute terms, the relative
creditworthiness of the Mexican firms would remain the same, and a simple
recalibration between model results and implied ratings would be possible.
However, the relative effect of inflation on the list of rated companies, which
vary significantly in the nature of their business and resulting asset bases and
the types and amounts of various liabilities incurred by each firm, e.g. the
relative amounts of local currency versus foreign currency denominated debt,
makes this assumption untenable.
QUALITY OF DATA
I Editor's note: Professor Ashwinpoul Sondhi presented his paper 'Accounting Perspectives in
Emerging Market Debt' at the conference, but it is not included in this book.
428 w.J. Chambers
this approach, there can be no recognition of, or adjustment for, cyclical swings.
In markets such as these, and with many of the companies focused on resource
based industries, such cyclical swings are significant and can substantially affect
the results for any given period without necessarily affecting the position of
the company longer term. As well, obviously in 1995 the test year for the model
there were a lot of 'one-off' effects in Mexico which hopefully will not be
repeated. Again, there may be the need to adjust for these.
To the numerically derived score, the authors make four key adjustments or
modifications to amend the model's results. These partially address some of
the concerns noted above.
This factor reflects both the nature of the company's sales in terms of currency
exposure and foreign currency costs, both interest and operating. This has
obviously proven to be a key factor in how different companies have fared
over the past year in Mexico with the position of some having been enhanced
and some having deteriorated. The difficulty with using this factor for scoring
or modeling purposes is the detailed information which it requires. Indeed,
public disclosure, particularly in emerging markets, but elsewhere as well,
seldom contains sufficient information to properly assess varying degrees of
susceptibility to foreign exchange gains or losses. This is especially true for
operating revenues and expenditures. Proxies, such as proportion of goods sold
abroad, exist, but these are subject to significant error. Hence, while this element
is critical in cases such as Mexico the lack of adequate information makes its
use extremely difficult.
How to weight this factor is also problematic. Ideally, the weighting should
reflect the likelihood and potential extent of devaluation in that particular
country. A high susceptibility to devaluation is much more important if the
risk of devaluation is high, and, indeed, if the chances of currency appreciation
are high, might actually become a positive rather than negative element. Should
a company located in, say, Malaysia be equally affected by this factor as a
company in Brazil? While this latter concern is less relevant to comparisons
only within a country, comparing the results with full-fledged ratings in the
US and the extension of the model to other markets such as Argentina as
planned, make cross-border comparisons inevitable.
Industry risk
Again, this is clearly a key factor. However, Mexican industry risk is quite
different from that in the US or other countries. One example is the cement
Comment 429
Competitive position
This adjustment factor takes into account a myriad of elements, including the
company's size, the degree of political influence (and implicitly the amount of
political support it may enjoy) and the company's quality of management. All
are important factors and are taken into account in performing a full-fledged
rating. As noted in the discussion of devaluation above, however, it is extremely
difficult to model or score some of these elements in the absence of a large
amount of detailed information. Nor is there an indication in the discussion
as to how these elements can be quantified, monitored or adjusted over time.
This adjustment factor encompasses collateral, guarantees, and the like accruing
to specific debt issues. We would argue that these are critical elements to
examine when adducing the creditworthiness of a particular debt issue.
However, the whole rest of the model seems designed to provide general
indications of the issuer's creditworthiness, not an analysis of a particular debt
issue. As a result, while important, it is unclear whether, how, and to what
degree such considerations should be factored into the model generated ratings.
The overall magnitude of these four adjustments can be quite significant. 2 For
Apasco and Ponderosa, the May adjusted rating is two full rating categories
below that which the ME score alone would indicate. For DESC the adjusted
2 The difference between, say, 'BBB' and 'A' is a category difference. The difference between a
'BBB -' and a 'BBB' is a notch difference, where there are three notches to each category.
430 w.J. Chambers
rating is 4 notches higher and for Tamsa the rating is 2 notches higher. All in
all, five of the sample of 29 ratings are higher than that indicated by the raw
score, while 19 are lower. The distribution of adjusted ratings vis-a.-vis the
unadjusted is quite different. The quantitative measures alone predicted only
82% of the final rating decision, leaving 18% to the other adjustment factors.
Could the adjustments be modeled? To a greater or lesser degree, the first
three, devaluation risk, industry risk and special debt features, could be. It
would be much more difficult with the fourth item. While size, market share,
and similar elements falling within the ambit of this category can be quantified
and hence modeled, the remaining elements such as management and political
influence are much more complex.
It is not clear how other factors such as group membership are factored in.
For example, the model produces ratings for Tolmex and Cemex, where the
former is a subsidiary of the latter and clearly is subject to the strategic decisions
of its parent. However, Tolmex's rating of A - is significantly different from
Cemex's BB +. Similarly, Sidek and Situr are both rated CCC - while another
member of the group, Simec is rated at B +. While certainly different members
of a group could enjoy different ratings, the substantial difference in these
ratings raises again the question of what, exactly the model is attempting to
measure.
Results of the Altman/Hartzell model have been published three times, in
May, July and December, 1995. The ratings for individual companies have
gyrated considerably from one period to the next. While not surprising given
the tumultuous economic environment in Mexico during the year, extreme
volatility in the ratings may affect their usefulness to investors. It should be
remembered that financial results incorporated in the May 1995 iteration of
the model included the first big effect of devaluation. Thus, one might have
expected less volatility in the model's output during the remainder of the year.
However, of the 29 original, adjusted ratings published in May, only three
remained constant as of the December, 1995 iteration. Between May and
December, 10 ratings were upgraded by between one and seven notches, while
16 were downgraded by up to 4 notches. The changes seem to be independent
of the original rating. Of the top-rated 8 companies in May, 6 were subsequently
downgraded while two were stable. At the other end of the spectrum, of the
10 lowest rated companies, 5 were up-graded while 5 were downgraded. The
original May rating results after adjustment predicted only 58% of the final
results at the end of the year. Frequent rating adjustments may be useful for a
trader who actively and frequently adjusts her portfolio, but are much less
useful to a medium and longer-term investor.
notes, in undertaking his analysis Standard & Poor's does not adjust the
company's reported results automatically to US GAAP. While we would review
US GAAP numbers if available, we believe that it is preferable to examine the
company's finances using figures prepared under its own, local accounting
standards.
On the surface, adjusting all numbers to a standard accounting framework
would make comparisons easier and eliminate the need to adjust for accounting
differences. However, the results of this, and even the process, are potentially
misleading. Clearly, companies located outside the US are not operated to
maximize results under US GAAP. They are operated in a manner to bring
forth best results under their own system. Standard & Poor's, in undertaking
an analysis, wishes to understand how companies are run, how decisions are
made within their own context.
Analysts must also be cautious in regarding US GAAP as necessarily the
only or even the best system. Other accounting systems are in some regards
more conservative, such as the UK insistence on amortizing all goodwill arising
from an acquisition immediately upon purchase, or more reflective of the actual
environment in which emerging companies operate, such as the inflation-
adjusted accounting utilized in much of Latin America. There are long and
serious debates as to how to properly reflect a company's financial position,
as evidenced by the extended and contentious debates of FASB and its equiva-
lents around the world. Nevertheless, few would argue that the policy in the
UK affects how a UK acquirer would view and values an acquisition or that
the use of historical accounting values properly reflects a company's position
in an environment of hyperinflation. Indeed, around the world, the US GAAP
is often the odd man out since it is one of the few, if not the only one, which
totally ignores real asset revaluation.
While examining companies in their own context, Standard & Poor's must
nevertheless make comparisons across systems to ensure the comparability of
our ratings. Generally, we have found that cash flow numbers are the best in
making comparisons, and, indeed for undertaking our analysis at all. In any
context, we've found that dynamic indicators such as cash flow provide superior
insight into the ability of a debt issuer to meet its debt obligations than static
indicators such as balance sheet ratios. In terms of comparisons, cash flow
information is also less susceptible to the peculiarities of a particular accounting
system than are income numbers, for example, which can be affected by depreci-
ation and other non-cash items or balance sheet information where asset
valuation, decisions to display certain items on the balance sheet as opposed
to off-balance sheet, such as pension obligations, can materially affect the
results.
SUMMARY
reliable data, both in terms of the business and financial positions of the
companies to be examined, and sophisticated quantitative tools, the develop-
ment of such models will require close definition of their objectives and conse-
quent interpretation of their results. Inclusion of 'soft', subjective variables adds
useful input to the realism of such models in reflecting the broad spectrum of
factors which affect creditworthiness, but understandably add a 'black box'
dimension to the model's results. Application of such models to the emerging
markets is particularly difficult due to the added volatility of these markets
compounded by higher uncertainty as to information quality. Differences in
accounting and reporting standards alone, makes cross-border comparisons of
credit quality difficult, and exacerbate any efforts to model or score such
differences absent significant interpretation of the information.
LEMMA W. SENBET
University of Maryland
INTRODUCTION
value use, and going concern firm value exceeds liquidation value, are reorgan-
ized and continue to operate, while inefficient firms are liquidated, with proceeds
distributed among claimants, in accordance with the promised debt obligation
and the seniority of claims.
Unfortunately, conflicts of interest between the firm's diverse claimant classes
lead to divergent preferences for resolving financial distress. In general, senior
claimants will favor premature shutdown of the firm (even if there is a loss of
going concern value) in order to preserve the value of their claims. The residual
and junior priority claim holders (e.g. equityholders), on the other hand, will
favor continued operation of the firm. Past attempts to reform the bankruptcy
process have generally lead to one of two undesirable outcomes: either ineffi-
cient firms were reorganized and survived, or viable firms were prematurely
(and inefficiently) liquidated. It should be noted here that the provisions of the
1978 Bankruptcy Act effectively lead to the former outcome, while the previous
bankruptcy laws led to the latter.
There are also problems of managerial entrenchment and free ridership in
workouts and restructuring. The voting process for the approval of a reorgani-
zation plan may be advantageous to the incumbent management and the
shareholders. Restructuring of public debt outside of the bankruptcy process
(i.e. informal or private workouts) is governed by the Trust Indenture Act of
1939. Curiously, the Act mandates that any changes to the terms of outstanding
public debt, such as interest, principal, or maturity, may be made only upon
unanimous approval of the issue's holders. In practice, this virtually precludes
any change in these terms directly. Consequently, informal restructuring of
public debt generally takes the form of an exchange offer. Thus, the Bankruptcy
Code impacts the balance of power among managers, equity holders, and the
firm's remaining stakeholders, with potentially high impact on the allocation
of resources and wealth distribution among the claimholders.
Moreover, it should be recognized that the existing bankruptcy procedure
can have important impact (indirectly) on the initial contractual process and
even private resolution of financial distress outside the court system. Since the
Code specifies the set of rules under which claimants bargain for their entitle-
ments, it also influences the behavior of the various stakeholders outside the
formal bankruptcy process. This observation is important because it suggests
that any reform of the Code or codification of an entirely new bankruptcy
procedure (say in an emerging economy), must also consider its impact on the
behavior of corporate stakeholders outside the formal bankruptcy process. This
is a point that I will return to in a later section in evaluating the proposed
procedure of the current paper under discussion.
In this section I wish to highlight the salient features of the reform proposal.
The next section will provide an evaluation of the proposed bankruptcy pro-
cedure. The core of the reform proposal is designing a bankruptcy procedure
Comment 435
that separates real asset allocation, particularly whether the firm should be
restructured or liquidated, from financial claims structure. In the parlance of
corporate finance, the procedure calls for separation of investment and financ-
ing decisions, while preserving absolute priority rule. The separability is accom-
plished by the expedient of converting a company in bankruptcy essentially
into an all-equity firm, or unifying claims into firm value maximization as an
appropriate objective. This is in the same vein of the private workout proposal
of Haugen and Sen bet (1978, 1988), whereby claims structure of an imminently
bankrupt firm would be unified or restructured into an all-equity firm either
through informal reorganization or through a takeover mechanism. While the
Haugen-Senbet proposal is in the context of private workouts, the current
paper seeks to codify it in the bankruptcy law.
How is the separability property achieved in the bankruptcy procedure?
Reorganization rights (RRs), which are effectively equity ownership rights in the
post-bankruptcy firm, are issued by a court-supervised receiver. Insider claimhold-
ers have preferential treatment and hold options to purchase these rights at
exercise prices determined on the basis of priority structure. Another way to
appreciate the procedure is to reverse the exercise process from the most senior
claims to the lowest in the pecking order. Using the authors' example, the most
senior claimants receive RRs free and sell them to the next at their claim value
($100), who then buy at $100 and sell it to the next group in the pecking order at
$200, obtaining the difference as their full claim. Thus, the most junior claimants
buy at $200 and get the difference between the reorganization value ($250) and
the exercise price ($200). Their total claim is, of course, impaired, because the
total claims outstanding ($300) exceed the reorganization offer of $250. Note also
that RRs are subject to public auction, which will mitigate the liquidity constraint
problem. This is important, which I will return to later.
Thus, at the end ofthe game, you have all claimants holding RRs, or effectively
a new group of shareholders, in an all-equity firm. These claimants vote for an
optimal investment policy, since there are no conflicts of interests. That is, an
optimal reorganization or liquidation of a bankrupt firm will take place.
Presumably, the firm will then be relevered to an optimal level of debt financing
to take advantage of the benefits (e.g. tax subsidy) associated with debt.
inherent in the existing institutional framework and market systems are not
confounded with the actual costs of the bankruptcy procedure. For instance,
there is a prevailing concern that Chapter 11 reorganizations take too long
and cost too much. However, this view may be confounding the economic
costs of reorganizations with the costs of Chapter 11. That is, the costs of
reorganizations might have been incurred even in the absence of Chapter 11.
In this context, Eisenberg (1992) argues that Chapter 11 proceedings did not
take longer than receivership reorganizations that preceded the Chapter in the
earlier days of bankruptcy settlements.
The issue of confounding of economic costs of reorganization and costs of
bankruptcy procedures is important, and it has an analogy to the confounding
of potential liquidation and bankruptcy costs in the determination of optimal
financial structure. Liquidation and bankruptcy can be thought of as indepen-
dent events, but when bankrupt firms also get liquidated, the costs are often
incorrectly attributed to bankruptcy costs. Or alternatively, there is a tendency
to confound economic and financial distress. Fundamentally bankruptcy does
not cause poor performance, although it is tempting to infer from news reports
of financially and economically distressed firms and contend that their economic
distress is caused by financial distress (see Haugen and Senbet, 1978 and Senbet
and Seward, 1995 for further discussion on these issues).
The deviations from absolute priority rule (APR) are also often viewed as
inefficient features of current bankruptcy procedures. Are APR violations
inefficient? There is growing literature that is shedding light on this question.
It is my view, though, that the absolute priority rule (APR) is an inappropriate
benchmark to judge the efficiency of Chapter 11. It is interesting to note that
large reorganizations involved deviations from APR even prior to Chapter 11
(see Eisenberg, 1992). Of course, we have a growing evidence in empirical
finance for frequent violations of APR in Chapter 11 reorganizations. Eberhart
et al. (1990) report an average deviation of 7.6% from APR, and there is
evidence that market participants anticipate the magnitudes of APR violations.
This suggests that, if APR violations are fully internalized and priced out in
the initial contract, there are no efficiency losses from APR deviations.
The authors of the current paper require adherence to the APR in allocating
claims. I agree with them that there ought to be prioritization of claims so as
to preserve the role of debt in disciplining management. However, why should
this priority be absolute? As above, the deviations from the APR can be
internalized in bond pricing; also, the relative priority rules may be beneficial
in mitigating risk incentives of financially distressed firms.1 Reliance on the
APR has the danger of dismissing the other simple alternatives of recapitaliza-
1 There is work suggesting that APR violations have a beneficial effect of mitigating risk shifting
costs of debt financing under financial distress (e.g. Eberhart and Senbet, 1993). Consequently, the
APR violations can be viewed as an implicit feature of an efficient, rather than distortionary, bond
contract, and policy calls for abolition of Chapter lion the ground of APR violations may be
unwarranted.
Comment 437
Proposals for a new bankruptcy procedure, or calls for the reform of an existing
one, should take careful account of two considerations. First, the bankruptcy
code determines not only how financial distress is resolved within a formal,
court-supervised framework, but also exerts influence on how corporate claim-
ants (e.g. bondholders, stockholders, trade creditors, managers) attempt to work
out privately outside the legal court system. Thus, any attempt at legislative
reform must be undertaken by considering its economic impact on various
claimants both inside and outside the formal bankruptcy process.
Second, it is important to recognize that markets play a vital role in the
reorganization of distressed firms. That is, legislative and judicial rules can
exert a significant influence on the efficiency of markets in resolving financial
distress. A ruling in January 1990 by a bankruptcy judge in the LTV case
illustrates clearly the process by which judicial discretion within the current
bankruptcy procedure can substantially change the cost of resolving financial
distress (Senbet and Seward, 1995). A private workout was arranged in which
a segment of bondholders volunteered for bonds with market value below the
original face value, and subsequently LTV filed for Chapter 11 in 1986. Under
the judge's ruling, the participants in the swap gave up their original claim,
while the holdout bondholders ended up with a larger reward per bond. The
ruling can be thought of effectively skewing the resolution of financial distress
towards the costlier court-supervised, formal bankruptcy procedure, and away
from successful private workouts. The basic point here is that markets must
be allowed to operate sufficiently unencumbered so as to achieve efficient
resource allocation, regardless of whether the firm is financially solvent or
distressed. Thus, it is important to assess whether a bankruptcy process, includ-
ing precedents established through judicial rule-making, weakens the influence
438 L. W. Senbet
2 Conversely, it should be clear that the firm that is efficiently run may go bankrupt, or face
financial distress, just on the basis of carrying excessively high debt in its capital structure. See
Haugen and Senbet (1978) and Senbet and Seward (1995) for further elaboration of the dangers
of confounding economic distress and financial distress.
Comment 439
The last issue that I wish to raise, which is particularly acute in emerging
economies characterized by severe informational problems, has to do with the
3 Note that the mechanism of acquisition under financial distress entails more than simply
acquiring ownership of the firm through acquisition of equity. The reason is that, if an outside
arbitrageur were to acquire the firm, and then implement a value-enhancing operating strategy,
the gains would accrue primarily to the debt holders. Thus, in order to retain the benefits of the
value enhancement, the bondholder claims must generally be extinguished at market value at the
time when the equity is purchased. An additional issue to note here is the existence of potential
impediments which are identified and analyzed in the literature, such as the free rider problem
and informational issues, and the solutions thereof. See, for instance, Haugen and Senbet (1988)
and Senbet and Seward (1995). In a nutshell, while the potential impediments often assume that
the form of debt finance is exogenously specified, their solutions view security design and corporate
capital structure decisions as endogenous to redressing the problems that may arise in the bank-
ruptcy process.
440 L. W. Sen bet
CONCLUDING NOTE
REFERENCES
Haugen, Robert A. and Lemma W. Senbet (1978). The insignificance of bankruptcy costs to the
theory of optimal capital structure. Journal of Finance, 33, 383-393.
Haugen, Robert A. and Lemma W. Senbet (1988). Bankruptcy and agency costs: their significance
to the theory of optimal capital structure. Journal of Financial and Quantitative Analysis, 23,
27-38.
Roe, M. (1983). Bankruptcy and debt: a new model for corporate reorganization. Columbial Law
Review, 83, 527-602.
Senbet, Lemma and James Seward (1995). Financial distress, bankruptcy and reorganization. In
R. Jarrow et al., Handbooks in OR and MS (Finance). Amsterdam: Elsevier, 921-961.
Operational problems
Management is burdened by rules and restrictions
Heightened conflicts of interest among management, bondholders, and
equity holders
Public scrutiny (suppliers, customers)
Deadweight costs
Direct costs (lawyers, accountants, etc.)
Indirect costs: Hard to estimate, potentially high
Allocational inefficiency
Too many liquidations/continuations
Managerial entrenchment
Deviations from absolute priority rule
Inefficient capital structure for emerging firm
Agency and free rider problems
Impediments to private workouts
List of contributors
PETER AERNI
EDWARD I. ALTMAN
Edward I. Altman is the Max L. Heine Professor of Finance at the Stern School
of Business, New York University. Since 1990, he has directed the research
effort in Fixed Income and Credit Markets at the NYU Salomon Center and
is currently the Vice Director of the Center. Prior to serving in his present
position, Professor Altman chaired the Stern School's MBA Program for 12
years. He has been visiting Professor at the Hautes Etudes Commerciales and
Universite de Paris-Dauphine in France, at the Pontificia Catolica Universidade
in Rio de Janeiro, at the Austrian Graduate School of Management in Sydney
and Luigi Bocconi University in Milan. Dr. Altman has an international
reputation as an expert on corporate bankruptcy and credit analysis. He was
named Laureate 1984 by the Hautes Etudes Commerciales Foundation in Paris
for his accumulated works on corporate distress prediction models and pro-
cedures for firm financial rehabilitation and awarded the Graham & Dodd
Scroll for 1985 by the Financial Analysts Federation for his work on Default
Rates on High Yield Corporate Debt. He is currently an advisor to the Centrale
dei Bilanci in Italy and a member of its Scientific and Technical Committee.
He was named to the Max L. Heine endowed professorship at Stern in 1988.
He received his MBA and PhD in Finance from the University of California,
Los Angeles.
Professor Altman is Editor of the Journal of Banking and Finance and a
publisher series, the Wiley Frontiers in Finance Series. Professor Altman has
published over a dozen books and numerous articles in scholarly finance,
accounting and economic journals. He is the current editor of the Handbook
of Corporate Finance and the Handbook of Financial Markets and Institutions
and the author of the recently published books Recent Advances in Corporate
Finance; Investing in Junk Bonds; Default Risk, Mortality Rates and the
Performance of Corporate Bonds; Distressed Securities: Analyzing and Evaluating
Market Potential and Investment Risk and his most recent work, Corporate
Financial Distress and Bankruptcy, second edition. His work has appeared in
many languages including French, German, Italian and Japanese.
443
R. Levich (ed.), Emerging Market Capital Flows, 443-460.
~ 1998 Kluwer Academic Publishers.
444 List of contributors
GEERT BEKAERT
ROBERT J. BERNSTEIN
GORDON M. BODNAR
JACOB BOUDOUKH
ROBERT WHITELAW
MATTHEW RICHARDSON
PABLO CABEZAS
RICHARD CANTOR
WILLIAM J. CHAMBERS
STIJN CLAESSENS
Stijn Claessens is at the World Bank and is currently one of the principal
authors of the 1996 World Development Report on Transition Economies. He
holds a PhD in business economics from the Wharton School of the University
of Pennsylvania. He has worked in the World Bank's Finance complex, the
Chief Economist's Office, the Debt and International Finance Division, and
the Finance and Private Sector Development Division for the Europe and
Central Asia regions. His assignments in the Bank have included operational
work on external debt management and financial sector development in various
countries, and policy work on external finance for developing countries.
His research interests are how countries can manage external risks; alterna-
tive forms of external finance, including portfolio flows; and financial sector
and enterprising restructuring in transition economies. He has published extens-
ively in these areas.
ROBERT E. CUMBY
IAN DOMOWITZ
MICHAEL DOOLEY
VIHANG ERRUNZA
MARTIN D: EVANS
RENE GARCIA
JACK GLEN
WILLIAM N. GOETZMANN
LAWRENCE GOODMAN
OLIVER D. HART
Oliver D. Hart has been a Professor of Economics at Harvard since July 1993.
Prior to that he taught at the London School of Economics and Massachusetts
Institute of Technology. At Harvard he is teaching a graduate course in
Contract Theory and also part of the first-year graduate microeconomics
sequence. His research interests include the theory of the firm, corporate gover-
nance, the bankruptcy reform. He is a research associate of the National Bureau
of Economic Research and is a former managing editor of the Review of
Economic Studies. He recently published a book, entitled Firms, Contracts, and
Financial Structure (Oxford University Press, 1995).
JOHN HARTZELL
CAMPBELL R. HARVEY
Empirical Finance, The Journal of Fixed Income, The Pacific Basin Finance
Journal, The Journal of Banking and Finance, The Journal of International
Financial Institutions, Markets and Money, and The European Journal of
Finance.
Professor Harvey received the 1993-1994 Batterymarch Fellowship, an
annual award given to the person most likely to establish a new area of research
in finance. He has also received the R.L. Rosenthal Award for Innovation in
Investment Management as well as three Graham and Dodd Scrolls for excel-
lence in financial writing from the Association for Investment Management
and Research. The American Finance Association awarded him a Smith-
Breeden prize for his publication The World Price of Covariance Risk and he
has received the American Association of Individual Investors' Best Paper in
Investments Award for Predictable Risk and Returns in Emerging Markets.
PHILIPPE JORION
GEORG JUNGE
Georg Junge is Managing Director and Head of the Industry and Country
Risk Analysis Group at Swiss Bank Corporation in Basel. He is specialized in
country risk analysis and designed the bank's system of country risk assessment.
More recently he has worked on models with which to quantify credit risk.
Before joining Swiss Bank Corporation in 1984 he was Research Fellow at the
Kiel Institute of World Economics. In 1983 he completed his studies in
Economics with a PhD at the Graduate Institute of International Studies in
List of contributors 453
Geneva. His dissertation dealt with the transmission of inflation under fixed
and flexible exhange rates.
TIMOTHY J. KEHOE
ANYA KHANTHAVIT
RAPHAEL LA PORTA
RICHARD M. LEVICR
JOHN M. LIEW
John M. Liew joined Goldman Sachs Asset Management in July 1994 and is
a member of the quantitative research team. Prior to joining Goldman Sachs,
he was a global equity trading analyst with Trout Trading Company where he
developed quantitative trading strategies. He received a BA from the University
of Chicago in 1989, an MBA in 1994 and a PhD in 1995 from the University
of Chicago Graduate School of Business.
FLORENCIO LOPEZ-DE-SILANES
ANANTH MADHAVAN
MATTHEW PECK
JOHN A. PENICOOK, JR
KENNETH S. ROGOFF
LEMMA W. SENBET
international finance and has received numerous honors and awards. He has
been an elected member of the Board of Directors of the American Finance
Association and past President of the Western Finance Association. Professor
Sen bet has served on numerous journal editorial boards, including the Journal
of Finance, the Journal of Financial and Quantitative Analysis, the Journal of
Banking and Finance, and Financial Management. He has also produced a string
of doctoral graduates and placed them in major research universities in the
USA. He is a recipient of the 1994 Allen Krowe Award for Teaching Excellence
at the University of Maryland Business School. Professor Sen bet has been
a consultant for the World Bank, the International Monetary Fund, United
Nations Economic Commission for Africa, and various government agencies
in USA, Canada, and Africa. He has also served as a resource person for the
African Economic Research Consortium.
Roy C. SMITH
Roy C. Smith has been on the faculty of the Stern School of Business at New
York University since September 1987 as a professor of finance and inter-
national business. Prior to assuming this appointment he was a General Partner
of Goldman, Sachs & Co. specializing in international investment banking and
corporate finance. At the time of his retirement from the firm to join the faculty,
he was the senior international partner. During his career at Goldman Sachs
he set up and supervised the firm's business in Japan and the Far East, headed
business development activities in Europe and the Middle East and served as
President of Goldman Sachs International Corporation while resident in the
firm's London office from 1980 to 1984.
Mr Smith received his BS degree from the US Naval Academy in 1960, and
his MBA from Harvard University in 1966, after which he joined Goldman,
Sachs & Co. He is a frequent guest lecturer at other business schools in the
USA and in Europe. His principal areas of research include international
banking and finance, global capital market activity, mergers and acquisitions,
leveraged transactions, foreign investments, and finance in emerging markets
and Eastern Europe.
In addition to various articles in professional journals and op-ed pieces, he
is the author of The Global Bankers (E.P. Dutton, 1989), The Money Wars
(E.P. Dutton, 1990) and Comeback: The Restoration of American Banking Power
in the New World Economy (Harvard Business School Press, 1993). He is also
co-author with Ingo Walter of Investment Banking in Europe: Restructuring in
the 1990s (Basil Blackwell, 1989), Global Financial Services (Harper and Row,
1990) and Global Banking (Oxford University Press, 1996).
Mr Smith is currently a Limited Partner of Goldman, Sachs & Co., a former
Director of Harsco Corporation and Tootal pIc, a UK Corporation, and a
founding partner of Large, Smith & Walter, a European financial services
consulting company. He is also a Director of the Atlantic Council of the United
458 List of contributors
States and a member of the Internal Research Council of the Center for
Strategic and International Studies, Washington, DC.
Ross L. STEVENS
PHILIP J. SUTTLE
KISHORE T ANDON
ANDRES VELASCO
T AD AS E. VISKANTA
INGO WALTER
HOLGER C. WOLF
461
462 Index
1. I.T. Vanderhoof and E. Altman (eds.): The Fair Value ofInsurance Liabilities. 1997
ISBN 0-7923-9941-2
2. R. Levich (ed.): Emerging Market Capital Flows. 1997 ISBN 0-7923-9976-5
3. Y. Amihud and G. Miller (eds.): Bank Mergers & Aquisitions. An Introduction and an
Overview. 1997 ISBN 0-7923-9975-7