Sie sind auf Seite 1von 44

Fixed

Income
Research
October 1991
Global Asset
Allocation With
Equities, Bonds,
and Currencies
Fischer Black
Robert Litterman
Fixed
Income
Research
Acknowledgements Al though we dont have space to thank them each i ndi vi du-
al l y, we woul d l i ke to acknowl edge the thoughtful comments
and suggesti ons that we have recei ved from our many col -
l eagues at Gol dman Sachs and the dozens of portfol i o man-
agers wi th whom we have di scussed our work thi s past year.
I n addi ti on, we want to offer speci al thanks to Al ex Bergi er,
Tom I ben, Pi otr Karasi nski , Scott Pi nkus, Scott Ri chard, and
Ken Si ngl eton for thei r careful readi ng of an earl i er draft of
thi s paper. We are al so grateful to Greg Cumber, Ganesh
Ramchandran, and Suresh Wadhwani for thei r val uabl e
anal yti cal support.
Fischer Black
(212) 902-8859
Robert Litterman
(212) 902-1677
Fi scher Bl ack i s a Partner i n Gol dman Sachs Asset Manage-
ment.
Robert Li tterman i s a Vi ce Presi dent i n the Fi xed I ncome
Research Department at Gol dman, Sachs & Co.
Editor: Ronald A. Krieger
Copyri ght 1991 by Gol dman, Sachs & Co. Si xth Pri nti ng, October 1992
Thi s materi al i s for your pri vate i nformati on, and we are not sol i ci ti ng any acti on
based upon i t. Certai n transacti ons, i ncl udi ng those i nvol vi ng futures, opti ons, and
hi gh yi el d securi ti es, gi ve ri se to substanti al ri sk and are not sui tabl e for al l
i nvestors. Opi ni ons expressed are our present opi ni ons onl y. The materi al i s based
upon i nformati on that we consi der rel i abl e, but we do not represent that i t i s
accurate or compl ete, and i t shoul d not be rel i ed upon as such. We, or persons
i nvol ved i n the preparati on or i ssuance of thi s materi al , may from ti me to ti me
have l ong or short posi ti ons i n, and buy or sel l , securi ti es, futures, or opti ons
i denti cal wi th or rel ated to those menti oned herei n. Further i nformati on on any of
the securi ti es, futures, or opti ons menti oned i n thi s materi al may be obtai ned upon
request. Thi s materi al has been i ssued by Gol dman, Sachs & Co. and has been
approved by Gol dman Sachs I nternati onal Li mi ted, a member of The Securi ti es and
Futures Authori ty, i n connecti on wi th i ts di stri buti on i n the Uni ted Ki ngdom, and
by Gol dman Sachs Canada i n connecti on wi th i ts di stri buti on i n Canada.
Fixed
Income
Research
Global Asset Allocation With
Equities, Bonds, and Currencies
Contents
Executive Summary
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
II. Neutral Views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Nai ve Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Historical Averages . . . . . . . . . . . . . . . . . . . . . . . . 7
Equal Means . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Risk-Adjusted Equal Means . . . . . . . . . . . . . . . . . 10
The Equi l i bri um Approach . . . . . . . . . . . . . . . . . . . . . . 11
III. Expressing Views . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
IV. Combining Investor Views With Market
Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Three-Asset Exampl e . . . . . . . . . . . . . . . . . . . . . . . . . . 16
A Seven-Country Exampl e . . . . . . . . . . . . . . . . . . . . . . 21
V. Controlling the Balance of a Portfolio . . . . . . . . . . 23
VI. Benchmarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
VII. Implied Views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
VIII. Quantifying the Benets of Global
Diversication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
IX. Historical Simulations . . . . . . . . . . . . . . . . . . . . . . . 31
X. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Appendix. A Mathematical Description of the
Black-Litterman Approach . . . . . . . . . . . . . . . . . . . 37
Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Fixed
Income
Research
Executive Summary
A year ago, Gol dman Sachs i ntroduced a quanti tati ve model that
offered an i nnovati ve approach to the management of xed i ncome
portfol i os.* I t provi ded a mechani sm for i nvestors to make gl obal
asset al l ocati on deci si ons by combi ni ng thei r vi ews on expected
returns wi th Fi scher Bl acks uni versal hedgi ng equi l i bri um. Gi v-
en an i nvestor s vi ews about i nterest rates and exchange rates,
thi s i ni ti al versi on of the Bl ack-Li tterman Gl obal Asset Al l ocati on
Model has been used to generate portfol i os wi th opti mal wei ghts
i n bonds i n di fferent countri es and the opti mal degree of currency
exposure.
I n thi s paper, we descri be an updated versi on of the Bl ack-
Li tterman Model that i ncorporates equi ti es as wel l as bonds and
currenci es. The new versi on of the model wi l l be especi al l y useful
to portfol i o managers who make gl obal asset al l ocati on deci si ons
across equi ty and xed i ncome markets, but i t wi l l al so have advan-
tages for pure xed i ncome managers.
The addi ti on of the equi ty asset cl ass to the model al l ows us to use
an equi l i bri um based on both bonds and equi ti es. Thi s equi l i bri um
i s more desi rabl e from a theoreti cal standpoi nt because i t i ncorpo-
rates a l arger fracti on of the uni verse of i nvestment assets. I n our
model (as i n any Capi tal Asset Pri ci ng Model equi l i bri um), the
equi l i bri um expected excess return on an asset i s proporti onal to
the covari ance of the assets return wi th the return of the market
portfol i o. Even for pure xed i ncome managers, i t i s useful to use
as broad a measure of the market portfol i o as i s practi cal .
As we descri bed i n our earl i er paper, the equi l i bri um i s i mportant
i n the model because i t provi des a neutral reference poi nt for
expected returns. Thi s al l ows the i nvestor to express vi ews onl y
for the assets that he desi res; vi ews for the other assets are de-
ri ved from the equi l i bri um. By provi di ng a center of gravi ty for
expected returns, the equi l i bri um makes the model s portfol i os
more bal anced than those from standard quanti tati ve asset al l oca-
ti on model s. Standard model s tend to choose unbal anced portfol i os
unl ess arti ci al constrai nts are i mposed on portfol i o composi ti on.
* See Bl ack and Li tterman (1990). [Note: a compl ete l i sti ng of references ap-
pears at the end of thi s report on page 40.]
Fixed
Income
Research
Global Asset Allocation With
Equities, Bonds and Currencies
I. Introduction nvestors wi th gl obal portfol i os of equi ti es and bonds are
I
general l y aware that thei r asset al l ocati on deci si ons
the proporti ons of funds that they i nvest i n the asset
cl asses of di fferent countri es and the degrees of currency
hedgi ng are the most i mportant i nvestment deci si ons they
make. I n attempti ng to deci de on the appropri ate al l ocati on,
they are usual l y comfortabl e wi th the si mpl i fyi ng assump-
ti on that thei r objecti ve i s to maxi mi ze expected return for
any gi ven l evel of ri sk, subject i n most cases to vari ous types
of constrai nts.
Gi ven the strai ghtforward mathemati cs of thi s opti mi zati on
probl em, the many correl ati ons among gl obal asset cl asses
requi red i n measuri ng ri sk, and the l arge amounts of money
i nvol ved, one mi ght expect that i n todays computeri zed
worl d, quanti tati ve model s woul d pl ay a domi nant rol e i n
thi s gl obal al l ocati on process.
Unfortunatel y, when i nvestors have tri ed to use quanti tati ve
model s to hel p opti mi ze thi s cri ti cal al l ocati on deci si on, the
unreasonabl e nature of the resul ts has often thwarted thei r
efforts.
2
When i nvestors i mpose no constrai nts, asset
wei ghts i n the opti mi zed portfol i os al most al ways ordai n
l arge short posi ti ons i n many assets. When constrai nts rul e
out short posi ti ons, the model s often prescri be corner sol u-
ti ons wi th zero wei ghts i n many assets, as wel l as unreason-
abl y l arge wei ghts i n the assets of markets wi th smal l capi -
tal i zati ons.
These unreasonabl e resul ts have stemmed from two wel l -
recogni zed probl ems. Fi rst, expected returns are very di ffi -
cul t to esti mate. I nvestors typi cal l y have vi ews about abso-
l ute or rel ati ve returns i n onl y a few markets. I n order to
use a standard opti mi zati on model , however, they must state
a set of expected returns for al l assets and currenci es. Thus,
they must augment thei r vi ews wi th a set of auxi l i ary as-
2
For some academi c di scussi ons of thi s i ssue, see Green and Hol l i el d
(1990) and Best and Grauer (1991).
1
Fixed
Income
Research
sumpti ons, and the hi stori cal returns that portfol i o manag-
ers often use for thi s purpose provi de poor gui des to future
returns.
Second, the opti mal portfol i o asset wei ghts and currency
posi ti ons i n standard asset al l ocati on model s are extremel y
sensi ti ve to the expected return assumpti ons. The two prob-
l ems compound each other because the standard model has
no way to di sti ngui sh strongl y hel d vi ews from auxi l i ary
assumpti ons, and gi ven i ts sensi ti vi ty to the expected re-
turns, the opti mal portfol i o i t generates often appears to
have l i ttl e or no rel ati on to the vi ews that the i nvestor wi sh-
es to express.
Confronti ng these probl ems, i nvestors are often di sappoi nted
when they attempt to use a standard asset al l ocati on model .
Our experi ence has been that i n practi ce, despi te the obvi ous
conceptual attracti ons of a quanti tati ve approach, few gl obal
i nvestment managers regul arl y al l ow quanti tati ve model s to
pl ay a major rol e i n thei r asset al l ocati on deci si ons.
I n thi s paper we descri be an approach that provi des an
i ntui ti ve sol uti on to these two probl ems that have pl agued
the use of quanti tati ve asset al l ocati on model s. The key to
our approach i s the combi ni ng of two establ i shed tenets of
modern portfol i o theory: the mean-vari ance opti mi zati on
framework of Markowi tz (1952) and the capi tal asset pri ci ng
model (CAPM) of Sharpe (1964) and Li ntner (1965). We
al l ow the i nvestor to combi ne hi s vi ews about the outl ook for
gl obal equi ti es, bonds, and currenci es wi th the ri sk premi -
ums generated by Bl acks (1989) gl obal versi on of the CAPM
equi l i bri um. These equi l i bri um ri sk premi ums are the excess
returns that equate the suppl y and demand for gl obal assets
and currenci es.
3
As we wi l l i l l ustrate wi th exampl es, the mean-vari ance
opti mi zati on used i n standard asset al l ocati on model s i s
extremel y sensi ti ve to the expected return assumpti ons that
the i nvestor must provi de. I n our model , the equi l i bri um
ri sk premi ums provi de a neutral reference poi nt for expected
returns. Thi s, i n turn, al l ows the model to generate opti mal
portfol i os that are much better behaved than the unreason-
abl e portfol i os that standard model s typi cal l y produce, whi ch
3
For a gui de to terms used i n thi s paper, see the Gl ossary, page 39.
2
Fixed
Income
Research
often i ncl ude l arge l ong and short posi ti ons unl ess otherwi se
constrai ned. I nstead, our model gravi tates toward a bal -
anced i .e., market-capi tal i zati on-wei ghted portfol i o but
ti l ts i n the di recti on of assets favored by the i nvestor s vi ews.
Our model does not assume that the worl d i s al ways at the
CAPM equi l i bri um, but rather that when expected returns
move away from thei r equi l i bri um val ues, i mbal ances i n
markets wi l l tend to push them back. Thus, we thi nk i t i s
reasonabl e to assume that expected returns are not l i kel y to
devi ate too far from equi l i bri um val ues. Thi s i ntui ti ve i dea
suggests that the i nvestor may prot by combi ni ng hi s vi ews
about returns i n di fferent markets wi th the i nformati on
contai ned i n the equi l i bri um.
I n our approach, we di sti ngui sh between vi ews of the i nvestor
and the expected returns that dri ve the opti mi zati on anal ysi s.
The equi l i bri um ri sk premi ums provi de a center of gravi ty for
expected returns. The expected returns that we use i n the
opti mi zati on devi ate from equi l i bri um ri sk premi ums when
the i nvestor expl i ci tl y states vi ews. The extent of the devi a-
ti ons from equi l i bri um depends on the degree of condence
the i nvestor expresses i n each vi ew. Our model makes adjust-
ments i n a manner as consi stent as possi bl e wi th hi stori cal
covari ances of returns of di fferent assets and currenci es.
Our use of equi l i bri um al l ows i nvestors to speci fy vi ews i n a
much more exi bl e and powerful way than i s otherwi se
possi bl e. For exampl e, rather than requi ri ng a vi ew about
the absol ute returns on every asset and currency, our ap-
proach al l ows i nvestors to speci fy as many or as few vi ews
as they wi sh. I n addi ti on, i nvestors can speci fy vi ews about
rel ati ve returns, and they can speci fy a degree of condence
about each of the vi ews.
I n thi s paper, through a set of exampl es, we i l l ustrate how
the i ncorporati on of equi l i bri um i nto the standard asset
al l ocati on model makes i t better behaved and enabl es i t to
generate i nsi ghts for the gl obal i nvestment manager. To that
end, we start wi th a di scussi on of how equi l i bri um can hel p
an i nvestor transl ate hi s vi ews i nto a set of expected returns
for al l assets and currenci es. We then fol l ow wi th a set of
appl i cati ons of the model that i l l ustrate how the equi l i bri um
sol ves the probl ems that have tradi ti onal l y l ed to unreason-
abl e resul ts i n standard mean-vari ance model s.
3
Fixed
Income
Research
II. Neutral Views
Exhibit 1
Historical Excess Returns
(J anuary 1975 through August 1991)
Total Historical Excess Returns
Germany France J apan U.K. U.S. Canada Australia
Currenci es -20.8 3.2 23.3 13.4 12.6 3.0
Bonds CH 15.3 -2.3 42.3 21.4 -4.9 -22.8 -13.1
Equi ti es CH 112.9 117.0 223.0 291.3 130.1 16.7 107.8
Annualized Historical Excess Returns
Germany France J apan U.K. U.S. Canada Australia
Currenci es -1.4 0.2 1.3 0.8 0.7 0.2
Bonds CH 0.9 -0.1 2.1 1.2 -0.3 -1.5 -0.8
Equi ti es CH 4.7 4.8 7.3 8.6 5.2 0.9 4.5
Annualized Volatility of Historical Excess Returns
Germany France J apan U.K. U.S. Canada Australia
Currenci es 12.1 11.7 12.3 11.9 4.7 10.3
Bonds CH 4.5 4.5 6.5 9.9 6.8 7.8 5.5
Equi ti es CH 18.3 22.2 17.8 24.7 16.1 18.3 21.9
________________________________
Note: Bond and equi ty excess returns are i n U.S. dol l ars currency hedged (CH). Excess returns on bonds and
equi ti es are i n excess of the London i nterbank offered rate (LI BOR), and those on currenci es are i n excess of
the one-month forward rates. Vol ati l i ti es are expressed as annual i zed standard devi ati ons.
hy shoul d an i nvestor use a gl obal equi l i bri um
W
model to hel p make hi s gl obal asset al l ocati on
deci si on? A neutral reference i s a cri ti cal l y i mpor-
tant i nput i n maki ng use of a mean-vari ance opti mi zati on
model , and an equi l i bri um provi des the appropri ate neutral
reference. Most of the ti me i nvestors do have vi ews feel -
i ngs that some assets or currenci es are overval ued or under-
val ued at current market pri ces. An asset al l ocati on model
shoul d hel p them to l everage those vi ews to thei r greatest
advantage. But i t i s unreal i sti c to expect an i nvestor to be
abl e to state exact expected excess returns for every asset
and currency. The purpose of the equi l i bri um i s to gi ve the
i nvestor an appropri ate poi nt of reference so that he can
express hi s vi ews i n a real i sti c manner.
4
Fixed
Income
Research
We begi n our di scussi on of how to combi ne vi ews wi th the
equi l i bri um by supposi ng that an i nvestor has no vi ews.
What then i s the opti mal portfol i o? Answeri ng thi s questi on
i s a sensi bl e poi nt of departure because i t demonstrates the
useful ness of the equi l i bri um ri sk premi um as the appropri -
ate poi nt of reference.
We consi der thi s questi on, and exampl es throughout thi s
paper, usi ng hi stori cal data on gl obal equi ti es, bonds, and
currenci es. We use a seven-country model wi th monthl y
returns for the Uni ted States, Japan, Germany, France, the
Uni ted Ki ngdom, Canada, and Austral i a from January 1975
through August 1991.
4
Exhi bi t 1 presents the means and standard devi ati ons of
excess returns. The correl ati ons are i n Exhi bi t 2 (pages 6-7).
Al l resul ts i n thi s paper have a U.S. dol l ar perspecti ve, but
other currency poi nts of vi ew woul d gi ve si mi l ar resul ts.
5
4
I n actual appl i cati ons of the model , we typi cal l y i ncl ude more asset
cl asses and use dai l y data to more accuratel y measure the current
state of the ti me-varyi ng ri sk structure. We i ntend to address i ssues
concerni ng uncertai nty of the covari ances i n another paper. For the
purposes of thi s paper, however, we treat the true covari ances of excess
returns as known.
5
We dene excess return on currency hedged assets to be total return
l ess the short rate and excess return on currency posi ti ons to be total
return l ess the forward premi um (see Gl ossary, page 39). I n Exhi bi t 2,
al l excess returns and vol ati l i ty are i n percent. The currency hedged
excess return on a bond or an equi ty at ti me t i s gi ven by:
E
t

,
P
t 1
/X
t 1
P
t
/X
t
1 100 (1 R
t
)FX
t
R
t
where P
t
i s the pri ce of the asset i n forei gn currency, X
t
i s the ex-
change rate i n uni ts of forei gn currency per U.S. dol l ar, R
t
i s the do-
mesti c short rate, and FX
t
i s the return on a forward contract, al l at
ti me t. The return on a forward contract, or equi val entl y the excess
return on a forei gn currency, i s gi ven by:
FX
t

,
F
t 1
t
X
t 1
X
t 1
100
where i s the one-peri od forward exchange rate at ti me t. F
t 1
t
5
Fixed
Income
Research
Exhibit 2
Historical Correlations of Excess Returns
(J anuary 1975 through August 1991)
Germany France J apan
Equi ti es Bonds Currency Equi ti es Bonds Currency Equi ti es Bonds Currency
CH CH CH CH CH CH
Germany
Equi ti es CH 1.00
Bonds CH 0.28 1.00
Currency 0.02 0.36 1.00
France
Equi ti es CH 0.52 0.17 0.03 1.00
Bonds CH 0.23 0.46 0.15 0.36 1.00
Currency 0.03 0.33 0.92 0.08 0.15 1.00
J apan
Equi ti es CH 0.37 0.15 0.05 0.42 0.23 0.04 1.00
Bonds CH 0.10 0.48 0.27 0.11 0.31 0.21 0.35 1.00
Currency 0.01 0.21 0.62 0.10 0.19 0.62 0.18 0.45 1.00
U.K.
Equi ti es CH 0.42 0.20 0.01 0.50 0.21 0.04 0.37 0.09 0.04
Bonds CH 0.14 0.36 0.09 0.20 0.31 0.09 0.20 0.33 0.19
Currency 0.02 0.22 0.66 0.05 0.05 0.66 0.06 0.24 0.54
U.S.
Equi ti es CH 0.43 0.23 0.03 0.52 0.21 0.06 0.41 0.12 0.02
Bonds CH 0.17 0.50 0.26 0.10 0.33 0.22 0.11 0.28 0.18
Canada
Equi ti es CH 0.33 0.16 0.05 0.48 0.04 0.09 0.33 0.02 0.04
Bonds CH 0.13 0.49 0.24 0.10 0.35 0.21 0.14 0.33 0.22
Currency 0.05 0.14 0.11 0.10 0.04 0.10 0.12 0.05 0.06
Australia
Equi ti es CH 0.34 0.07 0.00 0.39 0.07 0.05 0.25 0.02 0.12
Bonds CH 0.24 0.19 0.09 0.04 0.16 0.08 0.12 0.16 0.09
Currency 0.01 0.05 0.25 0.07 0.03 0.29 0.05 0.10 0.27
Naive Approaches To moti vate the use of the equi l i bri um ri sk premi ums as a
neutral reference poi nt, we rst consi der several other nai ve
approaches i nvestors mi ght use to construct an opti mal
portfol i o when they have no vi ews about assets or curren-
ci es. We wi l l cal l these the hi stori cal average approach, the
equal mean approach, and the ri sk-adjusted equal mean
approach.
6
Fixed
Income
Research
Historical Averages The hi stori cal average approach denes a neutral posi ti on to
Exhibit 2 (Continued)
Historical Correlations of Excess Returns
(J anuary 1975 through August 1991)
United Kingdom United States Canada Australia
Equi ti es Bonds Currency Equi ti es Bonds Equi ti es Bonds Currency Equi ti es Bonds
CH CH CH CH CH CH CH CH
U.K.
Equi ti es CH 1.00
Bonds CH 0.47 1.00
Currency 0.06 0.27 1.00
U.S.
Equi ti es CH 0.58 0.23 0.02 1.00
Bonds CH 0.12 0.28 0.18 0.32 1.00
Canada
Equi ti es CH 0.56 0.27 0.11 0.74 0.18 1.00
Bonds CH 0.18 0.40 0.25 0.31 0.82 0.23 1.00
Currency 0.14 0.13 0.09 0.24 0.15 0.32 0.24 1.00
Australia
Equi ti es CH 0.50 0.20 0.15 0.48 0.05 0.61 0.02 0.18 1.00
Bonds CH 0.17 0.17 0.09 0.24 0.20 0.21 0.18 0.13 0.37 1.00
Currency 0.06 0.05 0.27 0.07 0.00 0.19 0.04 0.28 0.27 0.20
be the assumpti on that excess returns wi l l equal thei r hi s-
tori cal averages. The probl em wi th thi s approach i s that
hi stori cal means are very poor forecasts of future returns.
For exampl e, i n Exhi bi t 1 we see many negati ve val ues.
Lets see what happens when we use these hi stori cal excess
returns as our expected excess return assumpti ons. We may
opti mi ze expected returns for each l evel of ri sk to get a fron-
ti er of opti mal portfol i os. Exhi bi t 3 (page 8) i l l ustrates the
fronti ers wi th the portfol i os that have 10.7% ri sk, wi th and
wi thout shorti ng constrai nts.
6
We may make a number of poi nts about these opti mal
portfol i os. Fi rst, they i l l ustrate what we mean by unreason-
6
We choose to normal i ze on 10.7% ri sk here and throughout thi s paper
because i t happens to be the ri sk of the market-capi tal i zati on-wei ghted,
80% currency hedged portfol i o that wi l l be hel d i n equi l i bri um i n our
model .
7
Fixed
Income
Research
abl e when we cl ai m that standard mean-vari ance opti mi za-
ti on model s often generate unreasonabl e portfol i os. The
portfol i o that does not constrai n agai nst shorti ng has many
l arge l ong and short posi ti ons wi th no obvi ous rel ati onshi p
to the expected excess return assumpti ons. When we con-
strai n shorti ng we have posi ti ve wei ghts i n onl y two of the
14 potenti al assets. These portfol i os are typi cal of those that
the standard model generates.
Exhibit 3
Optimal Portfolios Based on
Historical Average Approach
(percent of portfolio value)
Unconstrained
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 78.7 46.5 15.5 28.6 65.0 5.2
Bonds 30.4 40.7 40.4 1.4 54.5 95.7 52.5
Equi ti es 4.4 4.4 15.5 13.3 44.0 44.2 9.0
With constraints against shorting assets
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 160.0 115.2 18.0 23.7 77.8 13.8
Bonds 7.6 0.0 88.8 0.0 0.0 0.0 0.0
Equi ti es 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Gi ven how we have set up the opti mi zati on probl em, there
i s no reason to expect that we woul d get a bal anced set of
wei ghts. The use of past excess returns to represent a neu-
tral set of vi ews i s equi val ent to assumi ng that the constant
portfol i o wei ghts that woul d have performed best hi stori cal l y
are i n some sense neutral . I n real i ty, of course, they are not
neutral at al l , but rather are a very speci al set of wei ghts
that go short assets that have done poorl y and go l ong assets
that have done wel l i n the parti cul ar hi stori cal peri od.
Equal Means Recogni zi ng the probl em of usi ng past returns, the i nvestor
mi ght hope that assumi ng equal means for returns across al l
8
Fixed
Income
Research
countri es for each asset cl ass woul d be a better representa-
Exhibit 4
Optimal Portfolios Based on Equal Means
(percent of portfolio value)
Unconstrained
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 14.5 12.6 0.9 4.4 18.7 2.1
Bonds 11.6 4.2 1.8 10.8 13.9 18.9 32.7
Equi ti es 21.4 4.8 23.0 4.6 32.2 9.6 10.5
With constraints against shorting assets
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 14.3 11.2 4.5 0.2 25.9 2.0
Bonds 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Equi ti es 17.5 0.0 22.1 0.0 27.0 8.2 7.3
ti on of a neutral reference. We show an exampl e of the opti -
mal portfol i o for thi s type of anal ysi s i n Exhi bi t 4. Agai n, we
get an unreasonabl e portfol i o.
7
Of course, one probl em wi th the equal means approach i s
that equal expected excess returns do not compensate i nves-
tors appropri atel y for the di fferent l evel s of ri sk i n assets of
di fferent countri es. I nvestors di versi fy gl obal l y to reduce
ri sk. Everythi ng el se bei ng equal , they prefer assets whose
returns are l ess vol ati l e and l ess correl ated wi th those of
other assets.
Al though such preferences are obvi ous, i t i s perhaps surpri s-
i ng how unbal anced the opti mal portfol i o wei ghts can be, as
Exhi bi t 4 i l l ustrates, when we take everythi ng el se bei ng
equal to such a l i teral extreme. Wi th no constrai nts, the
l argest posi ti on i s short Austral i an bonds.
7
For the purposes of thi s exerci se, we arbi trari l y assi gned to each coun-
try the average hi stori cal excess return across countri es, as fol l ows: 0.2
for currenci es, 0.4 for bonds, and 5.1 for equi ti es.
9
Fixed
Income
Research
Risk-Adjusted Equal Means Our thi rd nai ve approach to deni ng a neutral reference
poi nt i s to assume that bonds and equi ti es have the same
expected excess return per uni t of ri sk, where the ri sk mea-
sure i s si mpl y the vol ati l i ty of asset returns. Currenci es i n
thi s case are assumed to have no excess return. We show the
opti mal portfol i o for thi s case i n Exhi bi t 5. Now we have
i ncorporated vol ati l i ti es, but the portfol i o behavi or i s no
better. One probl em wi th thi s approach i s that i t hasnt
taken the correl ati ons of the asset returns i nto account. But
there i s another probl em as wel l perhaps more subtl e, but
al so more seri ous.
Exhibit 5
Optimal Portfolios Based on
Equal Risk-Adjusted Means
(percent of portfolio value)
Unconstrained
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 5.6 11.3 28.6 20.3 50.9 4.9
Bonds 23.9 12.6 54.0 20.8 23.1 37.8 15.6
Equi ti es 9.9 8.5 12.4 0.3 14.1 13.2 20.1
With constraints against shorting assets
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 21.7 8.9 14.0 12.2 47.9 6.7
Bonds 0.0 0.0 0.0 7.8 0.0 19.3 0.0
Equi ti es 11.1 9.4 19.2 6.0 0.0 7.6 19.5
So far, the approaches we have used to try to dene the
appropri ate neutral means when the i nvestor has no vi ews
have been based on what mi ght be cal l ed the demand for
assets si de of the equati on that i s, hi stori cal returns and
ri sk measures. The probl em wi th such approaches i s obvi ous
when we bri ng i n the suppl y si de of the market.
Suppose the market portfol i o compri ses 80% of one asset
and 20% of the other. I n a si mpl e worl d, wi th i denti cal i n-
10
Fixed
Income
Research
vestors al l hol di ng the same vi ews and both assets havi ng
equal vol ati l i ti es, everyone cannot hol d equal wei ghts of
each asset. Pri ces and expected excess returns i n such a
worl d woul d have to adjust as the excess demand for one
asset and excess suppl y of the other affect the market.
The EquilibriumApproach To us, the onl y sensi bl e deni ti on of neutral means i s the
set of expected returns that woul d cl ear the market i f al l
i nvestors had i denti cal vi ews. The concept of equi l i bri um i n
the context of a gl obal portfol i o of equi ti es, bonds, and cur-
renci es i s si mi l ar, but currenci es do rai se a compl i cati ng
questi on. How much currency hedgi ng takes pl ace i n equi l i b-
ri um? The answer, as descri bed i n Bl ack (1989), i s that i n a
gl obal equi l i bri um i nvestors worl dwi de wi l l al l want to take
a smal l amount of currency ri sk.
Thi s resul t ari ses because of a curi osi ty known i n the cur-
rency worl d as Si egel s paradox. The basi c i dea i s that
because i nvestors i n di fferent countri es measure returns i n
di fferent uni ts, each wi l l gai n some expected return by tak-
i ng some currency ri sk. I nvestors wi l l accept currency ri sk
up to the poi nt where the addi ti onal ri sk bal ances the ex-
pected return. Under certai n si mpl i fyi ng assumpti ons, the
percentage of forei gn currency ri sk hedged wi l l be the same
for i nvestors of di fferent countri es gi vi ng ri se to the name
uni versal hedgi ng for thi s equi l i bri um.
The equi l i bri um degree of hedgi ng the uni versal hedgi ng
constant depends on three averages: the average across
countri es of the mean return on the market portfol i o of as-
sets, the average across countri es of the vol ati l i ty of the
worl d market portfol i o, and the average across al l pai rs of
countri es of exchange rate vol ati l i ty.
I t i s di ffi cul t to pi n down exactl y the ri ght val ue for the
uni versal hedgi ng constant, pri mari l y because the ri sk pre-
mi um on the market portfol i o i s a di ffi cul t number to esti -
mate. Neverthel ess, we feel that uni versal hedgi ng val ues
between 75% and 85% are reasonabl e. I n our monthl y data
set, the former val ue corresponds to a ri sk premi um of 5.9%
on U.S. equi ti es whi l e the l atter corresponds to a ri sk premi -
um of 9.8%. For the purposes of thi s paper, we wi l l use an
equi l i bri um val ue of 80% currency hedgi ng.
11
Fixed
Income
Research
Exhibit 6
EquilibriumRisk Premiums
(percent annualized excess return)
Germany France J apan U.K. U.S. Canada Australia
Currenci es 1.01 1.10 1.40 0.91 0.60 0.63
Bonds 2.29 2.23 2.88 3.28 1.87 2.54 1.74
Equi ti es 6.27 8.48 8.72 10.27 7.32 7.28 6.45
Exhi bi t 6 shows the equi l i bri um ri sk premi ums for al l as-
sets, gi ven thi s val ue of the uni versal hedgi ng constant.
8
Let us consi der what happens when we adopt these equi l i b-
ri um ri sk premi ums as our neutral means when we have no
vi ews. Exhi bi t 7 shows the opti mal portfol i o. I t i s si mpl y the
market capi tal i zati on portfol i o wi th 80% of the currency ri sk
hedged. Other portfol i os on the fronti er wi th di fferent l evel s
of ri sk woul d correspond to combi nati ons of ri sk-free borrow-
i ng or l endi ng pl us more or l ess of thi s portfol i o.
By i tsel f, the equi l i bri um i s i nteresti ng but not parti cul arl y
useful . The real val ue of the equi l i bri um i s to provi de a
neutral framework to whi ch the i nvestor can add hi s own
perspecti ve i n terms of vi ews, opti mi zati on objecti ves, and
constrai nts. These are the i ssues to whi ch we now turn.
8
The uni versal hedgi ng equi l i bri um i s, of course, based on a set si mpl i -
fyi ng assumpti ons, such as a worl d wi th no taxes, no capi tal con-
strai nts, no i nati on, etc. Exchange rates i n thi s worl d are the rates of
exchange between the di fferent consumpti on bundl es of i ndi vi dual s of
di fferent countri es. Whi l e some may nd the assumpti ons that justi fy
uni versal hedgi ng overl y restri cti ve, thi s equi l i bri um does have the
vi rtue of bei ng si mpl er than other gl obal CAPM equi l i bri ums that have
been descri bed el sewhere, such as i n Sol ni k (1974) or Grauer, Li tzen-
berger, and Stehl e (1976). Whi l e these si mpl i fyi ng assumpti ons are
necessary to justi fy the uni versal hedgi ng equi l i bri um, we coul d easi l y
appl y the basi c i dea of thi s paper to combi ne a gl obal equi l i bri um
wi th i nvestor vi ews to another gl obal equi l i bri um deri ved from a
di fferent, l ess restri cti ve, set of assumpti ons.
12
Fixed
Income
Research
III. Expressing Views
Exhibit 7
EquilibriumOptimal Portfolio
(percent of portfolio value)
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 1.1 0.9 5.9 2.0 0.6 0.3
Bonds 2.9 1.9 6.0 1.8 16.3 1.4 0.3
Equi ti es 2.6 2.4 23.7 8.3 29.7 1.6 1.1
he basi c probl em that confronts i nvestors tryi ng to
T
use quanti tati ve asset al l ocati on model s i s how to
transl ate thei r vi ews i nto a compl ete set of expected
excess returns on assets that can be used as a basi s for
portfol i o opti mi zati on. As we wi l l show here, the probl em i s
that opti mal portfol i o wei ghts from a mean-vari ance model
are i ncredi bl y sensi ti ve to mi nor changes i n expected excess
returns. The advantage of i ncorporati ng a gl obal equi l i bri um
wi l l become apparent when we show how to combi ne i t wi th
an i nvestor s vi ews to generate wel l -behaved portfol i os, wi th-
out requi ri ng the i nvestor to express a compl ete set of ex-
pected excess returns.
We shoul d emphasi ze that the di sti ncti on we are maki ng
between i nvestor vi ews on the one hand and a compl ete set
of expected excess returns for al l assets on the other i s
not usual l y recogni zed. I n our approach, vi ews represent the
subjecti ve feel i ngs of the i nvestor about rel ati ve val ues of-
fered i n di fferent markets.
9
I f an i nvestor does not have a
vi ew about a gi ven market, he shoul d not have to state one.
I f some of hi s vi ews are more strongl y hel d than others, the
i nvestor shoul d be abl e to express that di fference. Most
vi ews are rel ati ve for exampl e, when an i nvestor feel s one
market wi l l outperform another or even when he feel s bul l -
i sh (above neutral ) or beari sh (bel ow neutral ) about a mar-
ket. As we wi l l show, the equi l i bri um i s the key to al l owi ng
the i nvestor to express hi s vi ews thi s way i nstead of as a set
of expected excess returns.
9
As we wi l l show i n Secti on I X, vi ews can al so represent feel i ngs about
the rel ati onshi ps between observabl e condi ti ons and such rel ati ve
val ues.
13
Fixed
Income
Research
To see why thi s i s so i mportant, we start by i l l ustrati ng the
extreme sensi ti vi ty of portfol i o wei ghts to the expected ex-
cess returns and the i nabi l i ty of i nvestors to express vi ews
di rectl y as a compl ete set of expected returns. We have al -
ready seen how di ffi cul t i t can be si mpl y to transl ate no
vi ews i nto a set of expected excess returns that wi l l not l ead
to an unreasonabl e portfol i o i n an asset al l ocati on model .
But l et us suppose that the i nvestor has al ready sol ved that
probl em and happens to start wi th the equi l i bri um ri sk
premi ums as hi s neutral means. He i s comfortabl e wi th a
portfol i o that has the market capi tal i zati on wei ghts, 80%
hedged. Consi der what can happen when thi s i nvestor now
tri es to express one si mpl e, extremel y modest vi ew.
I n thi s exampl e, we suppose the i nvestor s vi ew i s that over
the next three months i t wi l l become i ncreasi ngl y apparent
that the economi c recovery i n the Uni ted States wi l l be weak
and bonds wi l l perform rel ati vel y wel l and equi ti es poorl y.
We suppose that the i nvestor s vi ew i s not very strong,
10
and he quanti es thi s vi ew by assumi ng that over the next
three months the U.S. benchmark bond yi el d wi l l drop 1 basi s
poi nt (bp) rather than ri se 1 bp, as i s consi stent wi th the
equi l i bri um ri sk premi um. Si mi l arl y, the i nvestor expects
U.S. share pri ces to ri se onl y 2.7%over the next three months
rather than to ri se 3.3%, as i s consi stent wi th the equi l i bri -
um.
To i mpl ement the asset al l ocati on opti mi zati on, the i nvestor
starts wi th the expected excess returns equal to the equi l i b-
ri um ri sk premi ums and adjusts them as fol l ows: he moves
the annual i zed expected excess returns on U.S. bonds up by
0.8 percentage poi nts and expected excess returns on U.S.
equi ti es down by 2.5 percentage poi nts. Al l other expected
excess returns remai n unchanged. I n Exhi bi t 8 we show the
opti mal portfol i o gi ven thi s vi ew.
Note the remarkabl e effect of thi s very modest change i n the
expected excess returns. The portfol i o wei ghts change i n
dramati c and l argel y i nexpl i cabl e ways. The opti mal portfo-
l i o wei ghts do shi ft out of U.S. equi ty i nto U.S. bonds, but
the model al so suggests shorti ng Canadi an and German
10
I n thi s paper, we use the term strength of a vi ew to refer to i ts mag-
ni tude. We reserve the term condence to refer to the degree of cer-
tai nty wi th whi ch i t i s hel d.
14
Fixed
Income
Research
bonds. Thi s l ack of apparent connecti on between the vi ews
Exhibit 8
Optimal Portfolios Based on a Moderate View
(percent of portfolio value)
Unconstrained
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 1.3 8.3 3.3 6.4 8.5 1.9
Bonds 13.6 6.4 15.0 3.3 112.9 42.4 0.7
Equi ti es 3.7 6.3 27.2 14.5 30.6 24.8 6.0
With constraints against shorting assets
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 2.3 4.3 5.0 3.0 9.2 0.6
Bonds 0.0 0.0 0.0 0.0 35.7 0.0 0.0
Equi ti es 2.6 5.3 28.3 13.6 0.0 13.1 1.5
the i nvestor attempts to express and the opti mal portfol i o
the model generates i s a pervasi ve probl em wi th standard
mean-vari ance opti mi zati on. I t ari ses because, as we saw i n
tryi ng to generate a portfol i o representi ng no vi ews, i n the
opti mi zati on there i s a compl ex i nteracti on between expected
excess returns and the vol ati l i ti es and correl ati ons used i n
measuri ng ri sk.
IV. Combining
Investor Views
With Market
Equilibrium
ow our approach transl ates a few vi ews i nto expect-
H
ed excess returns for al l assets i s one of i ts more
compl ex features, but al so one of i ts most i nnovati ve.
Here i s the i ntui ti on behi nd our approach:
(1) We bel i eve there are two di sti nct sources of i nforma-
ti on about future excess returns: i nvestor vi ews and
market equi l i bri um.
(2) We assume that both sources of i nformati on are un-
certai n and are best expressed as probabi l i ty di stri -
buti ons.
15
Fixed
Income
Research
(3) We choose expected excess returns that are as consi s-
tent as possi bl e wi th both sources of i nformati on.
The above descri pti on captures the basi c i dea, but the i mpl e-
mentati on of thi s approach can l ead to some novel i nsi ghts.
For exampl e, we wi l l now show how a rel ati ve vi ew about
two assets can i nuence the expected excess return on a
thi rd asset.
11
Three-Asset Example Let us rst work through a very si mpl e exampl e of our ap-
proach. After thi s i l l ustrati on, we wi l l appl y i t i n the context
of our seven-country model . Suppose we know the true struc-
ture of a worl d that has just three assets: A, B, and C. The
excess return for each of these assets i s known to be gener-
ated by an equi l i bri um ri sk premi um pl us four sources of
ri sk: a common factor and i ndependent shocks to each of the
three assets.
We can wri te thi s model as fol l ows:
R
A
=
A
+
A
Z +
A
R
B
=
B
+
B
Z +
B
R
C
=
C
+
C
Z +
C
where:
R
i
i s the excess return on the i
th
asset,

i
i s the equi l i bri um ri sk premi um on the i
th
asset,

i
i s the i mpact on the i
th
asset of Z, the common
factor, and

i
i s the i ndependent shock to the i
th
asset.
I n thi s worl d, the covari ance matri x, , of asset excess re-
turns i s determi ned by the rel ati ve i mpacts of the common
factor and the i ndependent shocks. The expected excess
returns of the assets are a functi on of the equi l i bri um ri sk
premi ums, the expected val ue of the common factor, and the
expected val ues of the i ndependent shocks to each asset.
11
I n thi s secti on, we try to devel op the i ntui ti on behi nd our approach
usi ng some basi c concepts of stati sti cs and matri x al gebra. We provi de
a more formal mathemati cal descri pti on i n the Appendi x.
16
Fixed
Income
Research
For exampl e, the expected excess return of asset A, whi ch
we wri te as E[R
A
], i s gi ven by:
E[R
A
] =
A
+
A
E[Z] + E[
A
]
We are not assumi ng that the worl d i s i n equi l i bri um, i .e.,
that E[Z] and the E[
i
]s are equal to zero. We do assume
that the mean, E[R
A
], i s i tsel f an unobservabl e random vari -
abl e whose di stri buti on i s centered at the equi l i bri um ri sk
premi ums. Our uncertai nty about E[R
A
] i s due to our uncer-
tai nty about E[Z] and the E[
i
]s. Furthermore, we assume
the degree of uncertai nty about E[Z] and the E[
i
]s i s pro-
porti onal to the vol ati l i ti es of Z and the
i
s themsel ves.
Thi s i mpl i es that E[R
A
] i s di stri buted wi th a covari ance
structure proporti onal to . We wi l l refer to thi s covari ance
matri x of the expected excess returns as . Because the
uncertai nty i n the mean i s much smal l er than the uncertai n-
ty i n the return i tsel f, wi l l be cl ose to zero. The equi l i bri um
ri sk premi ums together wi th determi ne the equi l i bri um
di stri buti on for expected excess returns. We assume thi s
i nformati on i s known to al l ; i t i s not a functi on of the ci r-
cumstances of any i ndi vi dual i nvestor.
I n addi ti on, we assume that each i nvestor provi des addi ti on-
al i nformati on about expected excess returns i n the form of
vi ews. For exampl e, one type of vi ew i s a statement of the
form: I expect Asset A to outperform Asset B by Q, where
Q i s a gi ven val ue.
We i nterpret such a vi ew to mean that the i nvestor has
subjecti ve i nformati on about the future returns of A rel ati ve
to B. One way we thi nk about representi ng that i nformati on
i s to act as i f we had a summary stati sti c from a sampl e of
data drawn from the di stri buti on of future returns data i n
whi ch al l we were abl e to observe i s the di fference between
the returns of A and B. Al ternati vel y, we can express thi s
vi ew di rectl y as a probabi l i ty di stri buti on for the di fference
between the means of the excess returns of A and B. I t does-
nt matter whi ch of these approaches we want to use to
thi nk about our vi ews; i n the end we get the same resul t.
I n both approaches, though, we need a measure of the i nves-
tor s condence i n hi s vi ews. We use thi s measure to deter-
mi ne how much wei ght to gi ve to the vi ew when combi ni ng
17
Fixed
Income
Research
i t wi th the equi l i bri um. We can thi nk of thi s degree of con-
dence i n the rst case as determi ni ng the number of obser-
vati ons that we have from the di stri buti on of future returns,
i n the second as determi ni ng the standard devi ati on of the
probabi l i ty di stri buti on.
I n our exampl e, consi der the l i mi ti ng case: the i nvestor i s
100% sure of hi s one vi ew. We mi ght thi nk of that as the
case where we have an unbounded number of observati ons
from the di stri buti on of future returns, and that the average
val ue of R
A
R
B
from these data i s Q. I n thi s speci al case,
we can represent the vi ew as a l i near restri cti on on the
expected excess returns, i .e. E[R
A
] E[R
B
] = Q.
I ndeed, i n thi s speci al , case we can compute the di stri buti on
of E[R] = {E[R
A
], E[R
B
], E[R
C
]}condi ti onal on the equi l i bri -
um and thi s i nformati on. Thi s i s a rel ati vel y strai ghtforward
probl em from mul ti vari ate stati sti cs. To si mpl i fy, l et us
assume a normal di stri buti on for the means of the random
components.
We have the equi l i bri um di stri buti on for E[R], whi ch i s
gi ven by Normal (, ), where = {
A
,
B
,
C
}. We wi sh to
cal cul ate a condi ti onal di stri buti on for the expected returns,
subject to the restri cti on that the expected returns sati sfy
the l i near restri cti on: E[R
A
] E[R
B
] = Q.
Let us wri te thi s restri cti on as a l i near equati on i n the ex-
pected returns:
12
P E[R] = Q where P i s the vector [1, 1, 0].
The condi ti onal normal di stri buti on has the mean
+ P [P P]
-1
[Q P ],
whi ch i s the sol uti on to the probl em of mi ni mi zi ng
(E[R] ) ()
-1
(E[R] ) subject to P E[R] = Q.
For the speci al case of 100% condence i n a vi ew, we use
thi s condi ti onal mean as our vector of expected excess re-
turns. I n the more general case where we are not 100%
12
A pri me symbol (e.g., P) i ndi cates a transposed vector or matri x.
18
Fixed
Income
Research
condent, we can thi nk of a vi ew as representi ng a xed
number of observati ons drawn from the di stri buti on of fu-
ture returns i n whi ch case we fol l ow the mi xed esti ma-
ti on strategy descri bed i n Thei l (1971), or al ternati vel y as
di rectl y reecti ng a subjecti ve di stri buti on for the expected
excess returns i n whi ch case we use the techni que gi ven
i n the Appendi x. The formul a for the expected excess returns
vector i s the same from both perspecti ves.
I n ei ther approach, we assume that the vi ew can be summa-
ri zed by a statement of the form P E[R] = Q + , where P
and Q are gi ven and i s an unobservabl e, normal l y di stri b-
uted random vari abl e wi th mean 0 and vari ance . repre-
sents the uncertai nty i n the vi ew. I n the l i mi t as goes to
zero, the resul ti ng mean converges to the condi ti onal mean
descri bed above.
When there i s more than one vi ew, the vector of vi ews can
be represented by P E[R] = Q + , where we now i nterpret
P as a matri x, and i s a normal l y di stri buted random vector
wi th mean 0 and di agonal covari ance matri x . A di agonal
corresponds to the assumpti on that the vi ews represent
i ndependent draws from the future di stri buti on of returns,
or that the devi ati ons of expected returns from the means of
the di stri buti on representi ng each vi ew are i ndependent,
dependi ng on whi ch approach i s used to thi nk about subjec-
ti ve vi ews. I n the Appendi x, we gi ve the formul a for the
expected excess returns that combi ne vi ews wi th equi l i bri um
i n the general case.
Now consi der our exampl e, i n whi ch correl ati ons among
assets resul t from the i mpact of one common factor. I n gen-
eral , we wi l l not know the i mpacts of the factor on the as-
sets, that i s the val ues of
A
,
B
, and
C
. But suppose the
unknown val ues are [3, 1, 2]. Suppose further that the i nde-
pendent shocks are smal l so that the assets are hi ghl y corre-
l ated wi th vol ati l i ti es approxi matel y i n the rati os 3:1:2.
For exampl e, suppose the covari ance matri x i s as fol l ows:
19
Fixed
Income
Research

1
1
1
]
9.1
3.0
6.0
3.0
1.1
2.0
6.0
2.0
4.1
Al so, for si mpl i ci ty, l et the equi l i bri um ri sk premi ums (i n
percent) be equal for exampl e, [1, 1, 1]. There i s a set of
market capi tal i zati ons for whi ch that i s the case.
Now consi der what happens when we speci fy a vi ew that A
wi l l outperform B by 2. I n thi s exampl e, si nce vi rtual l y al l of
the vol ati l i ty of the assets i s associ ated wi th movements i n
the common factor, we cl earl y ought to i mpute from the
hi gher return of A than of B (rel ati ve to equi l i bri um) that a
shock to the common factor i s the most l i kel y reason A wi l l
outperform B. I f so, C ought to perform better than equi l i bri -
um as wel l .
The condi ti onal mean i n thi s case i s [3.9, 1.9, 2.9], and i n-
deed, the vi ew of A rel ati ve to B has rai sed the expected
returns on C by 1.9.
Now suppose the i ndependent shocks have a much l arger i m-
pact than the common factor. Let the matri x be as fol l ows:
Suppose the equi l i bri um ri sk premi ums are agai n gi ven by

1
1
1
]
19.0
3.0
6.0
3.0
11.0
2.0
6.0
2.0
14.0
[1, 1, 1], and agai n we speci fy a vi ew that A wi l l outperform
B by 2.
Thi s ti me, more than hal f of the vol ati l i ty of A i s associ ated
wi th i ts own i ndependent shock not rel ated to movements i n
the common factor. Now, al though we ought to i mpute some
change i n the factor from the hi gher return of A rel ati ve to
B, the i mpact on C shoul d be l ess than i n the previ ous case.
I n thi s case, the condi ti onal mean i s [2.3, 0.3, 1.3]. Here the
i mpl i ed effect of the common factor shock on asset C i s l ower
than i n the previ ous case. We may attri bute most of the
outperformance of A rel ati ve to B to the i ndependent shocks;
i ndeed, the i mpl i cati on for E[R
B
] i s negati ve rel ati ve to equi -
20
Fixed
Income
Research
l i bri um. The i mpact of the i ndependent shock to B i s expect-
ed to domi nate, even though the contri buti on of the common
factor to asset B i s posi ti ve.
Noti ce that we can i denti fy the i mpact of the common factor
onl y i f we assume that we know the true structure that
generated the covari ance matri x of returns. That i s true
here, but i t wi l l not be true i n general . The computati on of
the condi ti onal mean, however, does not depend on thi s spe-
ci al knowl edge, but onl y on the covari ance matri x of returns.
Fi nal l y, l ets l ook at the case where we have l ess condence
i n our vi ew, so that we mi ght say (E[R
A
] E[R
B
]) has a
mean of 2 and a vari ance of 1.
Consi der the ori gi nal case, where the covari ance matri x of
returns i s:
The di fference i s that i n thi s exampl e the condi ti onal mean

1
1
1
]
9.1
3.0
6.0
3.0
1.1
2.0
6.0
2.0
4.1
i s based on an uncertai n vi ew. Usi ng the formul a gi ven i n
the Appendi x, we nd that the condi ti onal mean i s gi ven by:
[3.3, 1.7, 2.5].
Because we have stated l ess condence i n our vi ew, we have
al l owed the equi l i bri um to pul l us away from our expected
rel ati ve returns of 2 for A B to a val ue of 1.6, whi ch i s
cl oser to the equi l i bri um val ue of 0. We al so nd a smal l er
effect of the common factor on the thi rd asset because of the
uncertai nty expressed i n the vi ew.
A Seven-Country Example Now we appl y thi s approach to our actual data. We wi l l try
to represent the vi ew descri bed previ ousl y on page 14 that
bad news about the U.S. economy wi l l cause U.S. bonds to
outperform U.S. stocks.
The cri ti cal di fference between our approach here and our
earl i er experi ment that generated Exhi bi t 8 i s that here we
say somethi ng about expected returns on U.S. bonds versus
U.S. equi ti es and we al l ow al l other expected excess returns
21
Fixed
Income
Research
to adjust accordi ngl y. Above we adjusted onl y the returns to
Exhibit 9
Expected Excess Returns
Combining Investor Views With Equilibrium
(annualized percent)
Germany France J apan U.K. U.S. Canada Australia
Currenci es 1.32 1.28 1.73 1.22 0.44 0.47
Bonds 2.69 2.39 3.29 3.40 2.39 2.70 1.35
Equi ti es 5.28 6.42 7.71 7.83 4.39 4.58 3.86
U.S. bonds and U.S. equi ti es, hol di ng xed al l other expect-
ed excess returns. Another di fference i s that here we speci fy
a di fferenti al of means, l etti ng the equi l i bri um determi ne
the actual l evel s of means; above we had to speci fy the l evel s
di rectl y.
I n Exhi bi t 9 we show the compl ete set of expected excess
returns when we put 100% condence i n a vi ew that the
di fferenti al of expected excess returns of U.S. equi ti es over
bonds i s 2.0 percentage poi nts, bel ow the equi l i bri um di ffer-
enti al of 5.5 percentage poi nts. Exhi bi t 10 shows the opti mal
portfol i o associ ated wi th thi s vi ew.
Exhibit 10
Optimal Portfolio
Combining Investor Views With Equilibrium
(percent of portfolio value)
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 1.4 1.1 7.4 2.5 0.8 0.3
Bonds 3.6 2.4 7.5 2.3 67.0 1.7 0.3
Equi ti es 3.3 2.9 29.5 10.3 3.3 2.0 1.4
Thi s i s i n contrast to the i nexpl i cabl e resul ts we saw earl i er.
We see here a bal anced portfol i o i n whi ch the wei ghts have
ti l ted away from market capi tal i zati ons toward U.S. bonds
and away from U.S. equi ti es. Gi ven our vi ew, we now obtai n
a portfol i o that we consi der reasonabl e.
22
Fixed
Income
Research
V. Controlling the
Balance of
a Portfolio
n the previ ous secti on, we i l l ustrated how our approach
I
al l ows us to express a vi ew that U.S. bonds wi l l outper-
form U.S. equi ti es, i n a way that l eads to a wel l -behaved
opti mal portfol i o that expresses that vi ew. I n thi s secti on,
we focus more speci cal l y on the concept of a bal anced
portfol i o and show an addi ti onal feature of our approach:
Changes i n the condence i n vi ews can be used to control
the bal ance of the opti mal portfol i o.
Exhibit 11
Goldman Sachs Economists Views
Currencies
Germany France J apan U.K. U.S. Canada Australia
Jul y 31, 1991
Current Spot Rates 1.743 5.928 137.3 1.688 1.151 1.285
Three-Month Hori zon
Expected Future Spot 1.790 6.050 141.0 1.640 1.000 1.156 1.324
Annual i zed Expected
Excess Returns 7.48 4.61 8.85 6.16 0.77 8.14
Interest Rates
Germany France J apan U.K. U.S. Canada Australia
Jul y 31, 1991
Benchmark Bond Yi el ds 8.7 9.3 6.6 10.2 8.2 9.9 11.0
Three-Month Hori zon
Expected Future Yi el ds 8.8 9.5 6.5 10.1 8.4 10.1 10.8
Annual i zed Expected
Excess Returns 3.31 5.31 1.78 1.66 3.03 3.48 5.68
We start by i l l ustrati ng what happens when we put a set of
stronger vi ews, shown i n Exhi bi t 11, i nto our model . These
happen to have been the short-term i nterest rate and ex-
change rate vi ews expressed by our Gol dman Sachs econo-
23
Fixed
Income
Research
mi sts on Jul y 31, 1991.
12
We put 100% condence i n these
Exhibit 12
Optimal Portfolio Based on Economists Views
(percent of portfolio value)
Unconstrained
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 16.3 68.8 35.2 12.7 29.7 51.4
Bonds 34.5 65.4 79.2 16.9 3.3 22.7 108.3
Equi ti es 2.2 0.6 6.6 0.7 3.6 5.2 0.5
vi ews, sol ve for the expected excess returns on al l assets,
and nd the opti mal portfol i o shown i n Exhi bi t 12. Gi ven
such strong vi ews on so many assets, and opti mi zi ng wi th-
out constrai nts, we generate a rather extreme portfol i o.
Anal ysts have tri ed a number of approaches to amel i orate
thi s probl em. Some put constrai nts on many of the asset
wei ghts. However, we resi st usi ng such arti ci al constrai nts.
When asset wei ghts run up agai nst constrai nts, the portfol i o
opti mi zati on no l onger bal ances return and ri sk across al l
assets. Others speci fy a benchmark portfol i o and l i mi t the
ri sk rel ati ve to the benchmark unti l a reasonabl y bal anced
portfol i o i s obtai ned. Thi s makes sense i f the objecti ve of the
opti mi zati on i s to manage the portfol i o rel ati ve to a bench-
mark,
13
but agai n we are uncomfortabl e when i t i s used
si mpl y to make the model better behaved.
An al ternate response when the opti mal portfol i o seems too
extreme i s to consi der reduci ng the condence expressed i n
some or al l of the vi ews. Exhi bi t 13 shows an opti mal portfo-
l i o where we have l owered the condence i n al l of our vi ews.
By putti ng l ess condence i n our vi ews, we have generated
12
For detai l s of these vi ews, see the fol l owi ng Gol dman Sachs publ i ca-
ti ons: The I nternational Fixed I ncome Analyst, August 2, 1991, for
i nterest rates and TheI nternational Economics Analyst, Jul y/August
1991, for exchange rates.
13
We di scuss thi s si tuati on i n Secti on VI .
24
Fixed
Income
Research
a set of expected excess returns that more strongl y reect
the equi l i bri um and have pul l ed the opti mal portfol i o
wei ghts toward a more bal anced posi ti on.
Exhibit 13
Optimal Portfolio With Less Condence
in the Economists Views
(percent of portfolio value)
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 12.9 3.5 10.0 6.9 0.4 17.9
Bonds 3.9 21.0 19.6 2.6 7.3 13.6 42.4
Equi ti es 0.8 2.2 24.7 7.1 26.6 4.2 1.2
Let us now expl ai n preci sel y a property of a portfol i o that we
cal l bal ance. We dene bal ance as a measure of how si mi -
l ar a portfol i o i s to the gl obal equi l i bri um portfol i o that i s,
the market capi tal i zati on portfol i o wi th 80% of the currency
ri sk hedged. The di stance measure that we use i s the vol ati l -
i ty of the di fference i n returns of the two portfol i os.
We nd thi s property of bal ance to be a useful suppl ement
to the standard measures of portfol i o opti mi zati on, expected
return and ri sk. I n our approach, for any gi ven l evel of ri sk
there wi l l a conti nuum of portfol i os that maxi mi ze expected
return dependi ng on the rel ati ve l evel s of condence that
are expressed i n the vi ews. The l ess condence the i nvestor
has, the more bal anced wi l l be hi s portfol i o.
Suppose that an i nvestor does not have equal condence i n
al l hi s vi ews. I f the i nvestor i s wi l l i ng to rank the rel ati ve
condence l evel s of hi s di fferent vi ews, then he can generate
an even more powerful resul t. I n thi s case, the model wi l l
move away from hi s l ess strongl y hel d vi ews more qui ckl y
than from those i n whi ch he has more condence. For exam-
pl e, we have speci ed hi gher condence i n our vi ew of yi el d
decl i nes i n the Uni ted Ki ngdom and yi el d i ncreases i n
France and Germany. These are not the bi ggest yi el d chang-
es that we expect, but they are the forecasts that we most
strongl y want to represent i n our portfol i o. I n parti cul ar, we
put l ess condence i n our vi ews of i nterest rate moves i n the
Uni ted States and Austral i a.
25
Fixed
Income
Research
Before, when we put equal condence i n our vi ews, we ob-
tai ned the opti mal portfol i o shown i n Exhi bi t 13. The vi ew
that domi nated that portfol i o was the i nterest rate decl i ne i n
Austral i a. Now, when we put l ess than 100% condence i n
our vi ews, we keep rel ati vel y more condence i n the vi ews
about some countri es than others. Exhi bi t 14 shows the
opti mal portfol i o for thi s case, i n whi ch the wei ghts repre-
senti ng more strongl y hel d vi ews are l arger.
Exhibit 14
Optimal Portfolio With Less Condence
in Certain Views
(percent of portfolio value)
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 10.0 0.4 4.8 2.8 6.2 7.8
Bonds 10.3 34.3 25.5 1.6 22.9 2.4 28.1
Equi ti es 0.1 2.3 25.9 7.0 26.3 6.0 1.3
VI. Benchmarks ne of the most i mportant, but often overl ooked, i nu-
O
ences on the asset al l ocati on deci si on i s the choi ce of
the benchmark by whi ch to measure ri sk. I n mean-
vari ance opti mi zati on, the objecti ve i s to maxi mi ze return
per uni t of portfol i o ri sk. The i nvestor s benchmark denes
the poi nt of ori gi n for measuri ng thi s ri sk i n other words,
the mi ni mum ri sk portfol i o.
I n many i nvestment probl ems, ri sk i s measured, as we have
done so far i n thi s paper, as the vol ati l i ty of the portfol i os
excess returns. Thi s can be i nterpreted as havi ng no bench-
mark, or as deni ng the benchmark to be a portfol i o 100%
i nvested i n the domesti c short-term i nterest rate. I n many
cases, however, an al ternati ve benchmark i s cal l ed for. For
exampl e, many portfol i o managers are gi ven an expl i ci t
performance benchmark, such as a market-capi tal i zati on-
wei ghted i ndex. I f such an expl i ci t performance benchmark
exi sts, then i nvesti ng i n bi l l s i s cl earl y a ri sky strategy, and
the appropri ate measure of ri sk for the purpose of portfol i o
opti mi zati on i s the vol ati l i ty of the tracki ng error of the
portfol i o vi s--vi s the benchmark.
26
Fixed
Income
Research
Another exampl e of a si tuati on wi th an obvi ous al ternati ve
benchmark i s that of a manager fundi ng a known set of
l i abi l i ti es. I n such a case, the benchmark portfol i o repre-
sents the l i abi l i ti es.
Unfortunatel y, for many portfol i o managers, the objecti ves
fal l i nto a l ess wel l -dened mi ddl e ground, and the asset
al l ocati on deci si on i s made di ffi cul t by the fact that thei r
objecti ve i s i mpl i ci t rather than expl i ci t. For exampl e, a
gl obal equi ty portfol i o manager may feel hi s objecti ve i s to
perform among the top ranki ngs of al l gl obal equi ty manag-
ers. Al though he does not have an expl i ci t performance
benchmark, hi s ri sk i s cl earl y rel ated to the stance of hi s
portfol i o rel ati ve to the portfol i os of hi s competi ti on.
Other exampl es are an overfunded pensi on pl an or a uni ver-
si ty endowment where matchi ng the measurabl e l i abi l i ty i s
onl y a smal l part of the total i nvestment objecti ve. I n these
types of si tuati ons, attempts to use quanti tati ve approaches
are often frustrated by the ambi gui ty of the i nvestment
objecti ve.
When an expl i ci t benchmark does not exi st, two al ternati ve
approaches can be used. The rst i s to use the vol ati l i ty of
excess returns as the measure of ri sk. The second i s to speci -
fy a normal portfol i o one that represents the desi red
al l ocati on of assets i n the absence of vi ews. Such a portfol i o
mi ght, for exampl e, be desi gned wi th a hi gher-than-market
wei ght for domesti c assets, to represent the domesti c nature
of l i abi l i ti es wi thout attempti ng to speci fy an expl i ci t l i abi l i -
ty benchmark.
An equi l i bri um model can hel p i n the desi gn of a normal
portfol i o by quanti fyi ng some of the ri sk and return tradeoffs
between di fferent portfol i o wei ghts i n the absence of vi ews.
For exampl e, the opti mal portfol i o i n equi l i bri um i s market-
capi tal i zati on-wei ghted wi th 80% of the currency hedged. I t
has an expected return (usi ng equi l i bri um ri sk premi ums) of
5.7% wi th an annual i zed vol ati l i ty of 10.7%.
A pensi on fund wi shi ng to i ncrease the domesti c wei ght to
85% from the current market capi tal i zati on of 45%, and not
wi shi ng to hedge the currency ri sk of the remai ni ng 15% i n
i nternati onal markets, mi ght consi der an al ternate portfol i o
27
Fixed
Income
Research
such as the one shown i n Exhi bi t 15. The hi gher domesti c
Exhibit 15
Alternative Domestic Weighted Benchmark Portfolio
(percent of portfolio value)
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 1.5 1.5 7.0 3.0 2.0 0.0
Bonds 0.5 0.5 2.0 1.0 30.0 1.0 0.0
Equi ti es 1.0 1.0 5.0 2.0 55.0 1.0 0.0
wei ghts l ead to 0.4 percentage poi nts hi gher annual i zed vol a-
ti l i ty and an expected excess return 30 bp bel ow that of the
opti mal portfol i o. The pensi on fund may or may not feel that
i ts preference for domesti c concentrati on i s worth those costs.
VII. Implied Views nce an i nvestor has establ i shed hi s objecti ves, an
O
asset al l ocati on model denes a correspondence be-
tween vi ews and opti mal portfol i os. Rather than
treati ng a quanti tati ve model as a bl ack box, successful port-
fol i o managers use a model to i nvesti gate the nature of thi s
rel ati onshi p. I n parti cul ar, i t i s often useful to start an anal y-
si s by usi ng a model to nd the i mpl i ed i nvestor vi ews for
whi ch an exi sti ng portfol i o i s opti mal rel ati ve to a benchmark.
To i l l ustrate thi s type of anal ysi s, we assume that a portfol i o
manager has a portfol i o wi th wei ghts as shown i n Exhi bi t
16. The wei ghts, rel ati ve to those of hi s benchmark, dene
the di recti ons of the i nvestor s vi ews. By assumi ng the i nves-
tor s degree of ri sk aversi on, we can nd the expected excess
returns for whi ch the portfol i o i s opti mal . I n thi s type of
anal ysi s, di fferent benchmarks may i mpl y very di fferent
vi ews for a gi ven portfol i o. I n Exhi bi t 17 (page 30) we show
the i mpl i ed vi ews of the portfol i o shown i n Exhi bi t 16, wi th
the benchmark al ternati vel y (1) the market capi tal i zati on
wei ghts, 80% hedged, or (2) the domesti c wei ghted al terna-
ti ve shown i n Exhi bi t 15. Unl ess a portfol i o manager has
thought careful l y about what hi s benchmark i s and where
hi s al l ocati ons are rel ati ve to i t, and has conducted the type
of anal ysi s shown here, he may not have a cl ear i dea what
vi ews are bei ng represented i n hi s portfol i o.
28
Fixed
Income
Research
Exhibit 16
Current Portfolio Weights for Implied View Analysis
(percent of portfolio value)
Germany France J apan U.K. U.S. Canada Australia
Currency exposure 4.4 3.4 2.0 2.2 2.0 5.5
Bonds 1.0 0.5 4.7 2.5 13.0 0.3 3.5
Equi ti es 3.4 2.9 22.3 10.2 32.0 1.7 2.0
VIII. Quantifying the
Benets of Global
Diversication
hi l e i t has l ong been recogni zed that most i nves-
W
tors demonstrate a substanti al bi as toward domes-
ti c assets, many recent studi es have documented a
rapi d growth i n the i nternati onal component i n portfol i os
worl dwi de. I t i s perhaps not surpri si ng, then, that there has
been a reacti on among many i nvestment advi sers, who have
started to questi on the tradi ti onal arguments that support
gl obal di versi cati on. Thi s has been parti cul arl y true i n the
Uni ted States, where gl obal portfol i os have tended to under-
perform domesti c portfol i os i n recent years.
Of course, what matters for i nvestors i s the prospecti ve
returns from i nternati onal assets, and as noted i n our earl i -
er di scussi on of neutral vi ews, the hi stori cal returns are of
vi rtual l y no val ue i n projecti ng future expected excess re-
turns. Hi stori cal anal yses conti nue to be used i n thi s context
si mpl y because i nvestment advi sors argue there i s nothi ng
better to measure the val ue of gl obal di versi cati on.
We woul d suggest that there i s somethi ng better. A reason-
abl e measure of the val ue of gl obal di versi cati on i s the
degree to whi ch al l owi ng forei gn assets i nto a portfol i o rai s-
es the opti mal portfol i o fronti er. A natural starti ng poi nt for
quanti fyi ng thi s val ue i s to compute i t based on the neutral
vi ews i mpl i ed by a gl obal CAPM equi l i bri um. There are
some l i mi tati ons to usi ng thi s measure. I t assumes that
there are no extra costs to i nternati onal i nvestment; thus,
rel axi ng the constrai nt agai nst i nternati onal i nvestment
cannot make the i nvestor worse off. On the other hand, i n
measuri ng the val ue of gl obal di versi cati on thi s way, we
are al so assumi ng that markets are effi ci ent and therefore
29
Fixed
Income
Research
we are negl ecti ng to capture any val ue that an i nternati onal
portfol i o manager mi ght add through havi ng i nformed vi ews
about these markets. We suspect that an i mportant benet
of i nternati onal i nvestment that we are mi ssi ng here i s the
freedom i t gi ves the portfol i o manager to take advantage of
a l arger number of opportuni ti es to add val ue than are af-
forded i n domesti c markets.
Exhibit 17
Expected Excess Returns Implied by a Given Portfolio
Views Relative to the Market Capitalization Benchmark
(Annualized Expected Excess Returns)
Germany France J apan U.K. U.S. Canada Australia
Currenci es 1.55 1.82 0.27 1.22 0.63 2.45
Bonds 0.30 0.30 0.58 1.03 0.13 0.01 1.22
Equi ti es 2.82 3.97 0.30 6.73 4.15 5.01 5.88
Views Relative to the Domestic Weighted Benchmark
(Annualized Expected Excess Returns)
Germany France J apan U.K. U.S. Canada Australia
Currenci es 0.05 0.20 0.50 0.54 0.01 0.90
Bonds 0.01 0.21 0.72 0.85 1.45 1.01 0.18
Equi ti es 2.24 2.83 5.24 4.83 1.49 0.28 2.38
I n any case, as an i l l ustrati on of the val ue of the equi l i bri um
concept, we use i t here to cal cul ate the val ue of gl obal di ver-
si cati on for a bond portfol i o, an equi ty portfol i o, and a
portfol i o contai ni ng both bonds and equi ti es, i n each case
both wi th and wi thout al l owi ng currency hedgi ng. We nor-
mal i ze the portfol i o vol ati l i ti es at 10.7% the vol ati l i ty of
the market-capi tal i zati on-wei ghted portfol i o, 80%hedged. I n
Exhi bi t 18, we show the addi ti onal return avai l abl e from
i ncl udi ng i nternati onal assets rel ati ve to the opti mal domes-
ti c portfol i o wi th the same degree of ri sk.
What i s cl ear from thi s tabl e i s that gl obal di versi cati on
provi des a substanti al pi ckup i n expected return for the
domesti c bond portfol i o manager, both i n absol ute and per-
centage terms. The gai ns for an equi ty manager, or a portfo-
30
Fixed
Income
Research
l i o manager wi th both bonds and equi ti es, are al so substan-
Exhibit 18
The Value of Global Diversication
Expected Excess Returns in Equilibrium
at a Constant 10.7%Risk
Without Currency Risk Hedging
Basis Point Percent
Domestic Global Difference Gain
Bonds Onl y 2.14 2.63 49 22.9
Equi ti es Onl y 4.72 5.48 76 16.1
Bonds and Equi ti es 4.76 5.50 74 15.5
Allowing Currency Hedging
Basis Point Percent
Domestic Global Difference Gain
Bonds Onl y 2.14 3.20 106 49.5
Equi ti es Onl y 4.72 5.56 84 17.8
Bonds and Equi ti es 4.76 5.61 85 17.9
ti al , though much smal l er as a percentage of the excess
returns of the domesti c portfol i o. These resul ts al so appear
to provi de a justi cati on for the common practi ce of bond
portfol i o managers to currency hedge and of equi ty portfol i o
managers not to hedge. I n the absence of currency vi ews, the
gai ns to currency hedgi ng are cl earl y more i mportant i n both
absol ute and rel ati ve terms for xed i ncome i nvestors.
IX. Historical
Simulations
t i s natural to ask how a model such as ours woul d have
I
performed i n si mul ati ons. However, our approach does
not, i n i tsel f, produce i nvestment strategi es. I t requi res
a set of vi ews, and any si mul ati on i s a test not onl y of the
model but al so of the strategy produci ng the vi ews.
For exampl e, one strategy that i s fai rl y wel l known i n the
i nvestment worl d, and whi ch has performed qui te wel l i n
recent years, i s to i nvest funds i n hi gh yi el di ng currenci es.
31
Fixed
Income
Research
I n thi s secti on we i l l ustrate how a quanti tati ve model such
as ours can be used to opti mi ze such a strategy, and al so to
compare the rel ati ve performances of di fferent i nvestment
strategi es. I n parti cul ar, we wi l l compare the hi stori cal
performance of a strategy of i nvesti ng i n hi gh yi el di ng cur-
renci es versus two other strategi es: (1) i nvesti ng i n the
bonds of countri es wi th hi gh bond yi el ds and (2) i nvesti ng i n
the equi ti es of countri es wi th hi gh rati os of di vi dend yi el d to
bond yi el d. Our purpose i s to i l l ustrate how a quanti tati ve
approach can be used to make a useful compari son between
al ternati ve i nvestment strategi es. We are not tryi ng to pro-
mote or justi fy these parti cul ar strategi es. We have chosen
to focus on these three pri mari l y because they are si mpl e,
rel ati vel y comparabl e, and representati ve of standard i nvest-
ment approaches.
Our si mul ati ons of al l three strategi es use the same basi c
methodol ogy, the same data, and the same underl yi ng secu-
ri ti es. The di fferences i n the three si mul ati ons are i n the
sources of vi ews about excess returns and i n the assets to-
ward whi ch those vi ews are appl i ed. Al l of the si mul ati ons
use our approach of adjusti ng expected excess returns away
from the gl obal equi l i bri um as a functi on of i nvestor vi ews.
I n each of the si mul ati ons, we test a strategy by performi ng
the fol l owi ng steps. Starti ng i n Jul y 1981 and conti nui ng
each month for the next 10 years, we use data up to that
poi nt i n ti me to esti mate a covari ance matri x of returns on
equi ti es, bonds, and currenci es. We compute the equi l i bri um
ri sk premi ums, add vi ews accordi ng to the parti cul ar strate-
gy, and cal cul ate the set of expected excess returns for al l
securi ti es based on combi ni ng vi ews wi th equi l i bri um.
We then opti mi ze the equi ty, bond, and currency wei ghts for
a gi ven l evel of ri sk wi th no constrai nts on the portfol i o
wei ghts. We cal cul ate the excess returns that woul d have
accrued i n that month. At the end of each month we update
the data and repeat the cal cul ati on. At the end of ten years
we compute the cumul ati ve excess returns for each of the
three strategi es and compare them wi th one another and
wi th several passi ve i nvestments.
The vi ews for the three strategi es represent very di fferent
i nformati on but are generated usi ng si mi l ar approaches. I n
si mul ati ons of the hi gh yi el di ng currency strategy, our vi ews
32
Fixed
Income
Research
are based on the assumpti on that the expected excess re-
turns from hol di ng a forei gn currency are above thei r equi -
l i bri um val ue by an amount equal to the forward di scount on
that currency.
For exampl e, i f the equi l i bri um ri sk premi um on yen, from
a U.S. dol l ar perspecti ve, i s 1% and the forward di scount
(whi ch, because of covered i nterest rate pari ty, approxi mate-
l y equal s the di fference between the short rate on yen-de-
nomi nated deposi ts and the short rate on dol l ar-denomi nat-
ed deposi ts) i s 2%, then we assume the expected excess
return on yen currency exposures to be 3%. We compute
expected excess returns on bonds and equi ti es by adjusti ng
thei r returns away from equi l i bri um i n a manner consi stent
wi th 100% condence i n the currency vi ews.
I n si mul ati ons of a strategy of i nvesti ng i n xed i ncome
markets wi th hi gh yi el ds, we generate vi ews by assumi ng
that expected excess returns on bonds are above thei r equi -
l i bri um val ues by an amount equal to the di fference between
the bond-equi val ent yi el d i n that country and the gl obal
market-capi tal i zati on-wei ghted average bond-equi val ent
yi el d.
For exampl e, i f the equi l i bri um ri sk premi um on bonds i n a
gi ven country i s 1.0% and the yi el d on the 10-year bench-
mark bond i s 2.0 percentage poi nts above the worl d average
yi el d, then we assume the expected excess return for bonds
i n that country to be 3.0%. We compute expected excess
returns on currenci es and equi ti es by assumi ng 100% con-
dence i n these vi ews for bonds and adjusti ng them away
from equi l i bri um i n the appropri ate manner.
I n si mul ati ons of a strategy of i nvesti ng i n equi ty markets
wi th hi gh rati os of di vi dend yi el d to bond yi el d, we generate
vi ews by assumi ng that expected excess returns on equi ti es
are above thei r equi l i bri um val ues by an amount equal to 50
ti mes the di fference between the rati o of di vi dend to bond
yi el d i n that country and the gl obal market-capi tal i zati on-
wei ghted average rati o of di vi dend to bond yi el d.
For exampl e, i f the equi l i bri um ri sk premi um on equi ti es i n
a gi ven country i s 6.0% and the di vi dend to bond yi el d rati o
i s 0.5 wi th a worl d average rati o of 0.4, then we assume the
expected excess return for equi ti es i n that country to be
33
Fixed
Income
Research
11.0. We compute expected excess returns on currenci es and
Jun-81 Sep-82 Dec-83 Mar-85 Jun-86 Sep-87 Dec-88 Mar-90 Jun-91
600
500
400
300
200
100
0
Equi l i bri um
Currency Strategy
Bond Strategy
Equi ty Strategy
Currency Strategy
Equi ty Strategy
Equi l i bri um
Bond Strategy
($)
Exhibit 19
Historical Cumulative Monthly Returns
(U.S. Dollar-Based Perspective)
bonds by assumi ng 100% condence i n these vi ews for equi -
ti es and adjusti ng them away from equi l i bri um i n the appro-
pri ate manner.
We show the resul ts graphi cal l y i n Exhi bi ts 19 and 20. I n
Exhi bi t 19, we compare the cumul ati ve val ue of $100 i nvest-
ed i n each of the three strategi es as wel l as i n the equi l i bri -
um portfol i o, whi ch i s a gl obal market portfol i o of equi ti es
and bonds wi th 80% currency hedgi ng. The strategi es shown
i n Exhi bi t 19 were structured to have ri sk equal to that of
the equi l i bri um portfol i o. Whi l e such a graph gi ves a cl ear
pi cture of the rel ati ve performances of the di fferent strate-
gi es, i t cannot easi l y convey the tradeoff between ri sk and
return that can be obtai ned by taki ng a more or l ess aggres-
si ve posi ti on for any gi ven strategy.
34
Fixed
Income
Research
We can make such a compari son i n Exhi bi t 20, whi ch pl ots
the poi nts wi th actual annual i zed excess returns on the
verti cal axi s and the vol ati l i ty of returns from the si mul a-
ti ons of the di fferent strategi es on the hori zontal axi s. Be-
cause we perform our si mul ati ons wi th no constrai nts on
asset wei ghts, the ri sk/return tradeoff that we obtai n by
combi ni ng our si mul ati on portfol i os wi th cash i s l i near and
denes the appropri ate fronti er for each strategy. We show
each of these fronti ers together wi th the ri sk/return posi -
ti ons of several benchmark portfol i os: domesti c bond and
equi ty portfol i os, the equi l i bri um portfol i o, and gl obal mar-
ket-capi tal i zati on-wei ghted bond and equi ty portfol i os wi th
and wi thout currency hedgi ng.
0 2 4 6 8 10 12 14 16
Risk (%)
Excess Return
10
9
8
7
6
5
4
3
2
1
0
U.S.
LI BOR
U.S. Bonds
Gl obal Hedged
Equi l i bri um
Gl obal Undhedged
U.S. Equi ti es
Bond Strategy
Equity Strategy
Currency Strategy
(%)
Exhibit 20
Historical Risk/Return Tradeoffs
(J uly 1981 Through August 1991)
What we nd i s that strategi es of i nvesti ng i n hi gh yi el di ng
currenci es and i n the equi ty markets of countri es wi th hi gh
35
Fixed
Income
Research
rati os of di vi dend yi el ds to bond yi el ds both have performed
remarkabl y wel l over the past 10 years. On the other hand,
a strategy of i nvesti ng i n hi gh yi el di ng bond markets has
not i mproved returns. Al though past performance i s certai n-
l y no guarantee of future performance, we bel i eve that these
resul ts, and those of si mi l ar experi ments wi th other strate-
gi es, suggest some i nteresti ng l i nes of i nqui ry.
X. Conclusion
uanti tati ve asset al l ocati on model s have not pl ayed
Q
the i mportant rol e that they shoul d i n gl obal portfol i o
management. We suspect that a good part of the
probl em has been that users of such model s have found
them di ffi cul t to use and badl y behaved.
We have l earned that the i ncl usi on of a gl obal CAPM equi -
l i bri um wi th equi ti es, bonds, and currenci es can si gni cantl y
i mprove the behavi or of these model s. I n parti cul ar, i t al l ows
us to di sti ngui sh between the vi ews of the i nvestor and the
set of expected excess returns used to dri ve the portfol i o
opti mi zati on. Thi s di sti ncti on i n our approach al l ows us to
generate opti mal portfol i os that start at a set of neutral
wei ghts and then ti l t i n the di recti on of the i nvestor s vi ews.
By adjusti ng the condence i n hi s vi ews, the i nvestor can
control how strongl y the vi ews i nuence the portfol i o
wei ghts. Si mi l arl y, by speci fyi ng a ranki ng of condence i n
di fferent vi ews, the i nvestor can control whi ch vi ews are
expressed most strongl y i n the portfol i o. The i nvestor can
express vi ews about the rel ati ve performance of assets as
wel l as thei r absol ute performance.
We hope that our seri es of exampl es desi gned to i l l ustrate
the i nsi ghts that quanti tati ve model i ng can provi de wi l l
sti mul ate i nvestment managers to consi der, or perhaps to
reconsi der, the appl i cati on of such model i ng to thei r own
portfol i os.
36
Fixed
Income
Research
Appendix.
A Mathematical
Description of the
Black-Litterman
Approach
1. n assets bonds, equi ti es, and currenci es are
i ndexed by i = 1 , . . . , n.
2. For bonds and equi ti es, the market capi tal i zati on i s
gi ven by M
i
.
3. Market wei ghts of the n assets are gi ven by the vec-
tor W = {W
1
, . . . , W
n
}. We dene the W
i
s as fol l ows:
I f asset i i s a bond or equi ty:
W
i
M
i
i
M
i
I f asset i i s a currency of the j
th
country:
W
i
W
c
j
Where i s the country wei ght (the sum of market W
c
j
wei ghts for bonds and equi ti es i n the j
th
country) and
i s the uni versal hedgi ng constant.
4. Assets excess returns are gi ven by a vector R =
{R
1
, . . . , R
n
}.
5. Assets excess returns are normal l y di stri buted wi th
a covari ance matri x .
6. The equi l i bri um ri sk premi ums vector i s gi ven by
= W, where i s a proporti onal i ty constant based
on the formul as i n Bl ack (1989).
7. The expected excess return E[R] i s unobservabl e. I t
i s assumed to have a probabi l i ty di stri buti on that i s
proporti onal to a product of two normal di stri buti ons.
The rst di stri buti on represents equi l i bri um; i t i s
centered at wi th a covari ance matri x , where i s
a constant.
37
Fixed
Income
Research
The second di stri buti on represents our vi ews about k
l i near combi nati ons of the el ements of E[R]. These
vi ews are expressed i n the fol l owi ng form:
PE[R] = Q +
Here P i s a known k n matri x, Q i s a k-di mensi onal
vector, and i s an unobservabl e normal l y di stri buted
random vector wi th zero mean and a di agonal covari -
ance matri x .
8. The resul ti ng di stri buti on for i s normal wi th a E R
mean : E R
E R ()
1
P
1
P
1
()
1
P
1
Q
I n portfol i o opti mi zati on, we use as the vector E R
of expected excess returns.
38
Fixed
Income
Research
Glossary (Note: Deni ti ons on thi s page expl ai n terms as used in this
paper.)
Asset Excess Returns: Returns on assets l ess the domesti c short rate (see formul as
i n footnote 4 on page 5).
Balance: A measure of how cl ose a portfol i o i s to the equi l i bri um
portfol i o.
Benchmark Portfolio: The standard used to dene the ri sk of other portfol i os. I f a
benchmark i s dened, the ri sk of a portfol i o i s measured as
the vol ati l i ty of the tracki ng error the di fference between
the portfol i os returns and those of the benchmark.
Currency Excess Returns: Returns on forward contracts (see formul as i n footnote 4 on
page 5).
Expected Excess Returns: Expected val ues of the di stri buti on of future excess returns.
Equilibrium: The condi ti on i n whi ch means (see bel ow) equi l i brate the
demand for assets wi th the outstandi ng suppl y.
EquilibriumPortfolio: The portfol i o hel d i n equi l i bri um: i n thi s paper, market
capi tal i zati on wei ghts, 80% currency hedged.
Means: Expected excess returns.
Neutral Portfolio: An opti mal portfol i o gi ven neutral vi ews.
Neutral Views: Means when the i nvestor has no vi ews.
Normal Portfolio: The portfol i o that an i nvestor feel s comfortabl e wi th when
he has no vi ews. He can use the normal portfol i o to i nfer a
benchmark when no expl i ci t benchmark exi sts.
Risk Premiums: Means i mpl i ed by the equi l i bri um model .
39
Fixed
Income
Research
References
Best, Mi chael J., and Robert R. Grauer, On the Sensi ti vi ty of
Mean-Vari ance-Effi ci ent Portfol i os to Changes i n Asset Means:
Some Anal yti cal and Computati onal Resul ts, The Review of Fi-
nancial Studies, Vol . 4, No. 2, 1991, pp. 315-342.
Bl ack, Fi scher, Uni versal Hedgi ng: How to Opti mi ze Currency
Ri sk and Reward i n I nternati onal Equi ty Portfol i os, Financial
Analysts J ournal, Jul y/August 1989, Vol . 45, No. 4, pp. 16-22.
Bl ack, Fi scher, and Robert Li tterman, Asset Allocation: Combining
I nvestor Views With Market Equilibrium, Gol dman, Sachs & Co.,
September 1990.
Grauer, Frederi ck L. A., Robert H. Li tzenberger, and Ri chard E.
Stehl e, Shari ng Rul es and Equi l i bri um i n an I nternati onal Capi -
tal Market Under Uncertai nty, J ournal of Financial Economics,
Vol . 3, 1976, pp. 233-256.
Green, Ri chard C., and Burton Hol l i el d, When Wi l l Mean-
Vari ance Effi ci ent Portfol i os Be Wel l Di versi ed? Worki ng Paper,
Graduate School of I ndustri al Admi ni strati on, Carnegi e-Mel l on
Uni versi ty, February 1990 (revi sed).
Li ntner, John, The Val uati on of Ri sk Assets and the Sel ecti on of
Ri sky I nvestments i n Stock Portfol i os and Capi tal Budgets, Re-
view of Economics and Statistics, Vol . 47, No. 1, February 1965,
pp. 13-37.
Markowi tz, Harry, Portfol i o Sel ecti on, J ournal of Finance, Vol . 7,
March 1952, pp. 77-91.
Sharpe, Wi l l i am F., Capi tal Asset Pri ces: A Theory of Market
Equi l i bri um Under Condi ti ons of Ri sk, J ournal of Finance, Vol .
19, No. 3, September 1964, pp. 425-442.
Sol ni k, Bruno H., An Equi l i bri um Model of the I nternati onal
Capi tal Market, J ournal of EconomicTheory, Vol . 8, August 1974,
pp. 500-524.
Thei l , Henri , Principles of Econometrics, Wi l ey and Sons, 1971.
40

Das könnte Ihnen auch gefallen