Drivers of Expected Returns

in International Markets
CAMPBELL R. HARVEY

FALL 2000

kets, and all markets. The idea is to determine

if the same risk factors explain expected returns
in developed and emerging markets.

Dr

is the J. Paul Sticht

professor of international business at
Duke University, and
a research associate
with the National
Bureau of Economic
Research in Cambridge, Massachusetts.

hat are the drivers of average

returns in international markets? Are the sources of risk
that impact developed market returns the same as the risk factors that
affect emerging market returns? These are the
questions explored in this study.
The work is closely related to Harvey
[1995a, 1995b] and to Estrada [2000], who
examines a number of risk measures and relates
these measures to emerging market returns.
The difference is that we expand the list of risk
metrics, and examine both developed and
emerging market returns.
There are many reasons to believe that
the sources of risk in emerging and developed markets may differ. Harvey [1995a]
demonstrates that the pricing relationships
that show some positive success in developed
markets fail in emerging markets. Bekaert and
Harvey [1995] argue that emerging and developed markets cannot be treated similarly. They
propose an asset pricing framework that allows
for time-varying integration. Indeed, the presence of barriers to international investment as
documented in Bekaert [1995] immediately
suggests that risk factors will differ across
developed and emerging markets.
I examine a comprehensive list of 18 risk
factors, and analyze whether these risk factors
explain expected returns in 47 different markets. In much of the analysis, I present three sets
of results: developed markets, emerging mar-

I. DATA AND RISK METRICS

SUMMARY ANALYSIS

af

CAMPBELL R.
HARVEY

Sample

I examine average returns and risk in a

sample of 47 countries using data from Morgan Stanley Capital International (MSCI). The
sample consists of 28 emerging markets and 19
developed markets. Emerging market data are
also available from the International Finance
Corporation (IFC). When I replicate a number of the emerging markets analyses with the
IFC data, I find little difference between the
IFC and MSCI samples.
The data are sampled between January
1988 and December 1999. There are a number of emerging markets whose data do not
begin until later, but I include only markets
that have a history at least from January 1995.
Risk Metrics

A variety of risk metrics are used to

explain the average returns. The notation is
detailed here. For example, the raw materials
I am working with are the average total returns
in U.S. dollars for country i, which I express
as E[Ri].

EMERGING MARKETS QUARTERLY

One-Factor Market Model. Using a world version

of a single-factor model, where Rmt denotes the return on
the MSCI world index, I estimate the regression:
Rit rft = ai + bi[Rmt rft] + eit

(1)

where rft is the U.S. 30-day Treasury bill rate, and eit is the
residual. Also, note that emt = Rmt Avg(Rmt) is used later.
SR (systematic risk) is the beta, bi in equation
(1). TR (total risk) is the standard deviation of country
return i. IR (idiosyncratic risk) is the standard deviation of the residual eit. The one-factor model is studied
in Harvey [1995a, 1995b] and Bekaert and Harvey
[1995].
Size. For size (market capitalization), we take the natural log of average market capitalization over the relevant
period for each country. This variable is studied in Bekaert,
Erb, Harvey, and Viskanta [1997] and Estrada [2000].
Semistandard Deviations. The formula for semistandard deviation is:
Semi - B =

(1/ T) Tt=1(R t - B)2

for all R t < B (2)

Semimean is the semistandard deviation with B =

average returns for the market. Semi-rf is the semistandard
deviation with B = U.S. risk-free rate. Semi-0 is the
semistandard deviation with B = 0.
These variables are studied in Estrada [2000].
Downside Beta Measures. Down-iw is the  coefficient from the market model using observations when
country returns and world returns are simultaneously
negative. Down-w is the  coefficient from the market
model using observations when world returns are negative. The first variable is studied in Estrada [2000]. The
second beta has not been studied before.
Value at risk. VaR is a value at risk measure. It is the
simple average of returns below the 5th percentile level.
This variable is examined in Estrada [2000].
Skewness. Skew is the unconditional skewness of
returns. It is calculated by taking the Mean(ei3) divided by
the [Standard Deviation of (ei)]3. Skew 5%: [(Return at
the 95th Percentile level mean return) (Return at 5th
Percentile Level Mean Return)] - 1.
Coskew1 represents coskewness definition 1. It is
calculated by (sum up ei  em2)/T and divide by [square
root of (sum of (ei2)/T)]  [(sum of (em2)/T)].Coskew2
represents coskewness definition 2. It is calculated by
2

DRIVERS OF EXPECTED RETURNS IN INTERNATIONAL MARKETS

(Sum up ei  em2)/T and divide by [standard deviation of

(em)]3.
These variables are examined in Harvey and Siddique [1999, 2000a, 2000b].
Spread. Kurt is the kurtosis of the return distribution.
Kurtosis is documented in Bekaert and Harvey [1997] as
well as Bekaert, Erb, Harvey, and Viskanta [1998].
Political and Country Risk. ICRGC is the log of the
average monthly International Country Risk Guides
(ICRG) country risk composite. CCR is the log of the
average semiannual country risk rating published by Institutional Investor. ICRGP is the log of the average monthly
ICRG political risk ratings.
These variables are studied in Erb, Harvey, and
Viskanta [1996a, 1996b] as well as Bekaert, Erb, Harvey,
and Viskanta [1997].
II. SUMMARY STATISTICS

Exhibit 1A (emerging markets) and Exhibit 1B

(developed markets) present some summary statistics for
the risk measures over the complete sample for each
country. We report the mean monthly U.S. dollar returns
as well as the geometric mean return and the unconditional correlation with the MSCI world in addition to the
average values of the risk measures.
The last row of each exhibit reports the average across
all markets. We see the usual characteristics. Emerging
markets are more volatile and generally have more risk on
the downside (as measured by the semivariance, VaR,
downside betas). In addition, the coskewness measures for
emerging markets are more negative, suggesting that these
markets are contributing to the negative skewness of a
diversified portfolio.
Exhibits 2A-2C present the correlation matrices of
the risk measures in developed countries, emerging
markets, and all countries. There are some interesting
findings.
Total risk and idiosyncratic risk are 99% correlated
across emerging markets and 93% correlated in developed markets. The three semivariance measures all have
more than 90% correlation with total risk in both emerging and developed markets. VaR is at least -86% correlated with total risk in emerging and developed markets.
The skewness measures are relatively little correlated
with the other risk measures.

FALL 2000

FALL 2000

EMERGING MARKETS QUARTERLY

3.60
3.54
2.14
0.29
0.35
0.12
2.36
2.34
2.84
0.95
2.06
0.89
0.16
1.15
1.12
2.68
1.58
0.22
1.30
1.21
3.62
4.46
1.37
0.09
1.71
1.09
3.06
1.63
1.71

E[R]

2.18
1.84
1.84
-0.53
-0.08
-0.26
2.02
1.77
2.08
0.59
0.72
0.65
0.05
0.44
0.60
2.09
1.48
-0.48
0.79
0.72
2.09
1.40
1.02
-0.36
0.96
0.36
1.51
0.32
0.92

mG
r

0.11
0.30
0.27
0.08
0.17
0.27
0.18
0.22
0.54
0.17
0.23
0.40
0.12
0.37
0.48
0.41
-0.31
0.12
0.41
0.46
0.38
0.50
0.54
0.30
0.30
0.46
0.13
0.25
0.28

0.51
1.40
0.52
0.31
0.43
0.60
0.40
0.63
1.74
0.41
1.01
0.74
0.15
1.16
1.26
1.11
-0.37
0.42
1.15
1.17
2.07
3.19
1.24
0.79
0.95
1.42
0.61
1.09
0.93

SR
18.50
18.32
7.74
13.43
9.36
8.63
8.66
11.41
12.39
8.59
17.62
6.85
4.58
12.47
10.35
10.74
4.54
11.84
10.11
10.05
19.37
24.61
8.30
9.49
12.50
12.12
18.54
15.89
12.04

TR
18.43
17.52
7.47
13.47
9.30
8.39
8.60
11.15
10.54
8.52
17.21
6.33
4.58
11.64
9.10
9.82
4.35
11.83
9.26
8.96
17.97
21.55
7.00
9.11
11.97
10.77
18.44
15.47
11.38

IR
17.60
59.22
23.15
1.07
5.97
8.71
5.20
14.04
6.08
50.31
15.38
16.38
1.04
68.76
71.31
62.36
4.64
5.07
6.94
14.35
4.15
19.61
87.89
0.75
106.48
29.73
9.82
5.54
25.77

Mcap
2.87
4.08
3.14
0.07
1.79
2.16
1.65
2.64
1.81
3.92
2.73
2.80
0.04
4.23
4.27
4.13
1.54
1.62
1.94
2.66
1.42
2.98
4.48
-0.29
4.67
3.39
2.28
1.71
2.53

Size
10.36
12.23
5.39
7.80
6.36
6.52
4.78
6.43
8.55
5.71
10.01
5.02
3.20
7.49
7.06
7.98
3.01
8.06
6.98
6.72
10.88
16.98
6.12
6.63
8.29
8.41
11.30
11.21
7.84

Semimean
8.67
10.77
4.47
7.85
6.38
6.66
3.63
5.38
7.41
5.37
9.15
4.75
3.36
7.09
6.71
6.88
2.34
8.15
6.52
6.32
9.16
14.75
5.67
6.79
7.60
8.06
9.73
10.57
7.15

Market

mG

0.01

0.87
0.48
1.28
0.93
1.29
1.83
1.34
1.25
1.08
0.77
0.18
1.46
0.85
1.10
1.09
1.69
1.34
1 . 11
1.51
1.13

0.54
0.39
0.60
0.68
0.54
0.56
0.66
0.62
0.67
0.51
0.77
0.74
0.56
0.64
0.70
0.71
0.65
0.76
0.76
0.63

SR
0.75
0.69
0.70
0.80
0.71
1.24
0.93
0.91
0.96
0.92
1.41
0.79
0.98
1.19
1.12
1.19
0.84
0.89
0.72
0.93

TR

0.04

5.48
7.16
4.68
4.66
5.27
8.74
5.58
5.85
5.73
7.19
7.24
4.24
6.94
7.38
6.36
6.62
5.12
4.68
3.77
5.93

IR
4.63
6.62
3.76
3.42
4.45
7.27
4.18
4.61
4.29
6.22
4.61
2.87
5.76
5.70
4.56
4.67
3.89
3.08
2.46
4.58

Mcap

9032.25

131.24
19.72
62.37
211.34
40.58
40.39
356.85
396.55
16.78
154.80
1810.75
213.95
22.45
53.13
113.32
105.27
277.93
923.20
3979.84
470.02

Size

9 . 11

2.30
2.40
2.48
2.56
2.64
2.71
2.77
2.83
2.89
2.94
3.00
3.04
3.09
3.14
3.18
3.22
3.26
3.30
3.33
2.90

3.84
4.98
3.08
3.45
3.70
5.77
3.93
4.29
3.95
4.85
4.89
3.17
5.07
5.12
4.48
4.78
3.59
3.20
2.80
4.15

3.52
4.82
2.60
3.14
3.16
4.77
3.38
3.82
3.54
4.50
4.90
2.62
4.71
4.68
4.03
4.01
3.08
2.77
2.26
3.70

Semimean Semi-rf
3.29
4.61
2.38
2.95
2.95
4.53
3.17
3.63
3.32
4.27
4.64
2.43
4.49
4.49
3.83
3.81
2.88
2.55
2.07
3.49

Semi-0

1.93
3.19
1.09
-0.06
1.24
1.51
0.64
1.55
2.57
0.68
1.18
0.80
0.26
0.79
1.91
1.74
-0.32
0.56
2.34
2.01
1.28
4.76
1.89
2.28
1.46
2.02
1.27
3.27
1.57

0.45
1.29
0.56
0.70
0.38
1.06
0.79
1.14
0.69
0.68
0.84
0.65
1.04
1.14
1.22
0.95
0.91
0.41
0.67
0.82

0.60
1.54
0.62
0.94
0.79
1.66
1.00
1.21
0.94
0.88
1.15
0.84
1.23
1.41
1.51
1.35
1.20
0.66
0.84
1.07

VaR

-10.58
-15.81
-7.84
-9.27
-9.18
-14.95
-10.74
-13.36
-11.21
-12.92
-13.92
-8.05
-13.88
-16.80
-13.23
-13.09
-10.62
-7.96
-6.69
-11.58

2.10
0.57
0.07
1.18
0.20
0.08
1.49
1.92
1.43
0.36
2.16
-0.24
-0.20
1.53
0.56
-0.28
0.56
0.03
0.58
0.22
2.95
0.33
-0.51
0.32
0.43
0.15
1.02
-0.05
0.68

0.08
0.73
1.24
-0.29
0.31
0.29
-0.13
0.26
0.27
0.50
0.35
-0.31
-0.14
0.01
0.38
-0.10
0.27
0.49
-0.15
0.21

-0.29
-0.30
-0.27
0.02
-0.28
-0.42
-0.25
-0.19
-0.43
-0.09
0.01
-0.04
-0.07
0.08
-0.17
-0.14
-0.15
0.12
-0.43
-0.18
-0.05
0.04
-0.36
-0.26
-0.21
-0.12
-0.16
-0.42
-0.18

-1.33
-1.31
-0.51
0.07
-0.55
-0.80
-0.49
-0.54
-1.05
-0.16
0.04
-0.05
-0.08
0.22
-0.38
-0.35
-0.15
0.30
-0.84
-0.41
-0.19
0.22
-0.53
-0.49
-0.64
-0.32
-0.72
-1.37
-0.44

0.07
0.01
0.07
0 . 11
-0.04
0.41
0.09
-0.09
0 . 11
0.15
0.25
-0.09
0.03
0.08
0.15
-0.05
-0.04
-0.07
-0.09
0.06

-0.13
-0.23
0.06
-0.22
0.00
-0.14
0.05
-0.23
0.09
0.05
0.29
-0.07
-0.22
-0.06
-0.25
-0.14
-0.05
0.35
-0.23
-0.06

-0.15
-0.38
0.05
-0.19
0.00
-0.25
0.05
-0.26
0.10
0.07
0.34
-0.05
-0.31
-0.08
-0.28
-0.16
-0.05
0.27
-0.14
-0.08

skew5% coskew1 coskew2

0.34
0.20
0.19
0.53
-0.06
-0.07
0.45
0.16
0.11
0.27
-0.01
-0.08
0.27
0.49
-0.10
0.00
-0.10
0.37
0.01
0.06
0.16
0.02
0.11
0.09
0.21
-0.24
0.42
0.14
0.14

skew skew5% coskew1 coskew2

skew

-28.41
-37.25
-13.39
-19.42
-20.90
-20.28
-10.41
-15.91
-25.46
-13.71
-30.20
-14.07
-10.29
-23.40
-21.36
-23.89
-6.79
-26.80
-22.71
-20.63
-29.42
-46.30
-20.13
-19.64
-25.11
-26.16
-27.82
-34.80
-22.67

VaR

Down-bw

Down-biw Down-bw

1.41
1.80
1.26
0.20
1.18
1.50
0.57
0.86
1.79
0.40
1.15
0.23
0.04
0.15
1.46
1.11
1.11
-0.30
1.71
1.68
1.45
2.68
1.60
1.50
1.20
1.65
0.93
1.87
1.15

Down-biw

3.33
4.58
6.78
6.07
2.84
3.34
3.75
4.63
4.17
3.08
3.51
3.33
3.93
5.15
4.31
3.49
4.06
3.24
4.32
4.10

kurt

11.03
6.18
4.02
5.11
4.50
4.07
6.02
8.74
6.12
2.60
11.61
2.95
4.56
9.55
7.02
4.49
4.37
4.60
5.48
5.43
15.62
2.86
5.52
3.37
3.97
4.39
4.98
3.83
5.82

kurt

80.78
86.29
82.08
83.92
85.05
83.36
8 1. 2 9
8 5 . 16
82.18
78.08
86.69
86.79
87.56
84.97
76.56
8 3 . 17
90.05
8 1. 7 8
83.35
83.64

ICRGC

64.57
62.98
73.17
72.83
63.02
80.11
69.93
69.28
76.52
65.92
61.31
70.41
63.87
77.77
75.05
68.05
71.39
58.02
61.34
59.87
77.81
59.87
72.25
63.04
84.29
71.24
55.74
65.94
68.41

ICRGC

70.58
85.42
79.84
82.98
7 7 . 15
75.56
87.67
91. 6 3
70.98
76.42
9 1. 8 6
88.74
8 1. 0 7
79.95
75.60
77.54
92.85
8 7 . 14
9 0 . 14
82.27

CCR

31.26
31.76
48.29
57.49
41.26
56.83
30.52
49.78
46.34
44.42
46.88
43.11
29.26
66.61
61.73
39.90
33.68
28.33
20.62
32.16
32.77
35.26
39.88
27.57
77.38
58.81
41.26
35.05
42.44

8 1. 6 9
85.89
79.47
83.34
85.56
86.32
8 0 . 10
83.07
81.33
74.92
82.39
86.93
84.70
81 .1 7
73.24
84.27
88.65
81. 5 1
81. 8 6
82.44

ICRGP

69.17
65.89
69.56
67.63
55.43
82.28
62.75
70.51
82.09
61.20
53.79
64.59
60.71
72.34
70.05
68.74
68.87
53.93
56.45
55.86
80.16
57.91
71.26
57.09
77.86
64.72
54.16
64.13
65.68

CCR ICRGP

Jan-88

Jan-888
Jan-88
Jan-88
Jan-88
Jan-88
Jan-88
Jan-88
Jan-88
Jan-88
Jan-882
Jan-88
Jan-88
Jan-88
Jan-88
Jan-88
Jan-88
Jan-88
Jan-88
Jan-88

Start

Jan-88
Jan-88
Jan-88
Jan-93
Jan-93
Jan-95
Jan-95
Jan-88
Jan-95
Jan-93
Jan-88
Jan-93
Jan-88
Jan-88
Jan-88
Jan-88
Jan-93
Jan-93
Jan-93
Jan-88
Jan-93
Jan-95
Jan-93
Jan-93
Jan-88
Jan-88
Jan-88
Jan-93

Start

MSCI Data. E[R]: Mean returns; G: Geometric mean returns; : Correlation with world returns; SR: Systematic risk (Beta); TR: Total risk; IR: Idiosyncratic risk; Mcap: Average Market capitalization; Size: Log of average market cap; Semimean: Semideviation with respect to mean; Semi-rf: Semideviation with respect to risk-free rate; Semi-0: Semideviation with respect to 0; Down-iw: Downside beta calculated using observations when country index and world index fall simultaneously; Down-w: Downside beta calculated
using observations when world index falls; VaR: Value at risk; skew: Skewness; skew5%: Skewness calculated using formula 1; Coskew1: Coskewness with world index using formula 2, Coskew-2: Coskew with world index using formula 3; kurt: Kurtosis; ICRGC: Average ICRG composite country risk; CCR: Average CCR country risk; ICRGP: ICRGP
political risk; Start: Data start date.

0.01

World

E[R]

1.02
0.73
1.39
1.04
1.42
2.19
1.49
1.42
1.24
1.02
0.43
1.55
1.10
1.36
1.29
1.91
1.47
1.22
1.58
1.31

Australia
Austria
Belgium
Canada
Denmark
inland
rance
Germany
reland
taly
apan
Netherlands
Norway
ingapore
pain
weden
witzerland
U.K.
U.S.
Average

Developed Markets

8.45
10.56
4.27
7.61
6.16
6.46
3.41
5.17
7.24
5.14
8.93
4.57
3.11
6.86
6.51
6.70
2.13
7.94
6.35
6.09
8.97
14.54
5.50
6.58
7.35
7.85
9.50
10.39
6.94

Semi - rf Semi-0

Summary Statistics (monthly dollar returns)Developed Markets

EXHIBIT 1B

Emerging Markets
Argentina
Brazil
Chile
China
Colombia
Czech Republic
Egypt
Greece
Hungary
India
Indonesia
Israel
Jordan
Korea
Malaysia
Mexico
Morocco
Pakistan
Peru
Philippines
Poland
Russia
South Africa
Srilanka
Taiwan
Thailand
Turkey
Venezuela
Average

Market

Summary Statistics (monthly dollar returns)Emerging Markets

EXHIBIT 1A

DRIVERS OF EXPECTED RETURNS IN INTERNATIONAL MARKETS

1.000
0.350
1.000
-0.823
-0.064
0.372
-0.438
-0.549
-0.004
0.169
0.021
0.127
1.000

0.929
-0.801
-0.023
0.215
-0.523
-0.614
0.193
0.167
0.140
0.018
0.729
-0.956
0.074
0.562
-0.115
-0.254
-0.172
0.038
-0.221
-0.074
0.722
-0.952
0.083
0.570
-0.102
-0.242
-0.178
0.034
-0.224
-0.076
0.787
-0.946
0.040
0.594
-0.166
-0.312
-0.223
0.009
-0.282
-0.027
-0.318
0.440
-0.218
-0.220
0.273
0.418
-0.113
-0.010
0.652
-0.108
0.656
-0.859
0.243
0.580
-0.204
-0.376
-0.166
-0.036
-0.392
-0.011
0.745
-0.919
0.151
0.673
-0.063
-0.218
-0.221
-0.038
-0.313
-0.061
0.584
-0.624
-0.128
0.584
0.224
0.179
-0.214
-0.048
-0.069
-0.149
VaR
skew
skew5%
coskew1
coskew2
kurt
ICRGC
CCR
ICRGP

0.190
0.136
-0.221
-0.047
-0.275
-0.264
-0.101
-0.050
-0.134
0.223
Down-bw

Semi-0
Down-biw

1.000
0.060
-0.025
-0.023
-0.093
-0.023
-0.266
-0.262
0.044

1.000
0.691
0.389
0.052
0.685
0.651
0.648
0.451

1.000
0.933
-0.442
0.987
0.954
0.953
0.625

1.000
-0.607
0.919
0.885
0.885
0.567

1.000
-0.447
-0.402
-0.405
-0.264

1.000
0.969
0.970
0.687

1.000
1.000
0.668

1.000
0.678

Semi-0
Semi - rf
Semimean
Size
IR
TR
SR
E[R]

E[R]
SR
TR
IR
Size
Semimean
Semi - rf

Cross-Section Analysis Correlation MatrixDeveloped Countries

EXHIBIT 2A

Down-biw Down-bw

VaR

1.000
-0.061
-0.458
0.218
0.358
0.066
-0.078
0.180
0.036

skew

1.000
0.184
0.330
0.209
0.264
-0.210
-0.126
-0.273

1.000
0.123
0.069
-0.006
-0.289
-0.397
-0.252

1.000
0.956
-0.258
-0.052
0.112
-0.119

skew5% coskew1

1.000
-0.209
-0.075
0.155
-0.174

coskew2

1.000
0.039
0.096
-0.131

kurt

1.000
0.566
0.876

ICRGC

Bivariate Regressions

The main regression results are presented in Exhibit

3. The regressions are illustrated 18 sets of graphs in Exhibit
4. These regressions examine the bivariate relation between
the average returns and the average risk measures.
I should emphasize that we are comparing averages
to averages over the same time period. This is consistent
with some of the early tests of asset pricing models, such
as Black, Jensen, and Scholes [1972]. There is a substantial literature, beginning with Harvey [1991], however,
that suggests that in an international context time variation in the risk and returns measures is very important.
Nevertheless, this unconditional analysis should prove
interesting.
The first set of regressions is the classic world
CAPM. The analysis shows a significant relation when all
47 countries are included with an R2 of 27%. Interestingly,
the intercept is not significantly different from zero. This
evidence would seem to support the CAPM.
A closer inspection reveals that all of the explanatory
power is coming from emerging markets. The regression
suggests that there is a 0% R2 for developed markets.
But here is where the graphic analysis is especially
insightful. The poor fit for developed markets is driven by
one point; Japan has a beta of 1.41 and an average return
close to zero. If this single observation is excluded, the fit
of the developed countries dramatically improves and
shows an R2 of 27%. Hence, the CAPM fares reasonably
well in this analysis.
The fit for the emerging markets might seem surprising, given the results in Bekaert and Harvey [1995] and
Harvey [1995a], but remember that the sample is for
1988-1999. During the early part of this period, many of
the emerging markets liberalized, and the evidence in
Bekaert and Harvey [2000] would suggest that these markets have become more correlated with the world.
The second risk measure is total risk. Asset pricing
theory says that only systematic risk, or the part of variance that contributes to a well-diversified portfolios variance, should be important. The regressions suggest that
total variance can account for 52% of the variation in the
emerging market returns. Variance explains practically
none of the developed market returns. A combined analysis is heavily influenced by emerging markets returns.
The results for the third risk measure, idiosyncratic
risk, are similar to total risk.

1.000

ICRGP
CCR

III. REGRESSION ANALYSIS

FALL 2000

FALL 2000

EMERGING MARKETS QUARTERLY

0.507
-0.556
0.469
0.092
-0.054
-0.286
0.401
-0.094
-0.132
0.169

E[R]
1.000
0.562
0.716
0.683
0.309
0.678
0.530
0.531
0.499

SR

0.794
-0.761
0.126
-0.241
0.048
0.008
0.194
0.017
0.022
0.119

1.000
0.660
0.555
0.303
0.746
0.724
0.727
0.661
0.593
-0.901
0.480
0.185
0.184
-0.137
0.414
-0.265
-0.021
-0.081

1.000
0.990
0.134
0.966
0.932
0.931
0.439

TR

0.518
-0.865
0.509
0.255
0.190
-0.159
0.429
-0.295
-0.029
-0.113

1.000
0.092
0.937
0.906
0.904
0.361

IR

0.246
-0.227
-0.129
-0.168
-0.020
-0.086
0.031
0.253
0.471
0.259

1.000
0.187
0.145
0.145
0.201

Size

0.732
-0.965
0.265
0.062
0.114
-0.166
0.221
-0.281
-0.029
-0.093

1.000
0.982
0.983
0.544

Semimean

0.724
-0.977
0.182
0.042
0.136
-0.129
0.155
-0.295
0.000
-0.141

1.000
1.000
0.508

Semi - rf

0.515
-0.516
0.442
0.117
-0.149
-0.339
0.406
-0.213
-0.240
-0.039

E[R]
1.000
0.519
0.642
0.595
0.018
0.616
0.488
0.489
0.491

0.751
-0.586
0.099
-0.137
0.067
0.019
0.157
0.009
0.010
0.067

1.000
0.514
0.395
0.171
0.582
0.549
0.551
0.616

SR

0.624
-0.940
0.530
0.319
-0.128
-0.376
0.488
-0.608
-0.532
-0.511

1.000
0.990
-0.394
0.980
0.959
0.959
0.515

TR

0.557
-0.920
0.549
0.359
-0.168
-0.417
0.501
-0.660
-0.586
-0.567

1.000
-0.467
0.960
0.945
0.944
0.457

IR

-0.091
0.362
-0.305
-0.284
0.308
0.310
-0.221
0.592
0.728
0.573

1.000
-0.368
-0.405
-0.405
-0.127

Size

0.726
-0.977
0.376
0.234
-0.174
-0.397
0.353
-0.608
-0.528
-0.510

1.000
0.988
0.988
0.593

Semimean

0.710
-0.985
0.326
0.229
-0.168
-0.380
0.313
-0.628
-0.536
-0.551

1.000
1.000
0.566

Semi - rf

0.712
-0.986
0.325
0.225
-0.172
-0.382
0.314
-0.629
-0.537
-0.551

1.000
0.569

Semi-0

0.727
-0.978
0.181
0.038
0.132
-0.132
0.155
-0.294
-0.002
-0.139

1.000
0.511

Semi-0

0.854
-0.616
0.067
-0.314
-0.540
-0.568
0.119
-0.225
-0.287
-0.142

1.000

Down-biw

0.837
-0.564
-0.024
-0.524
-0.536
-0.502
0.009
-0.024
-0.113
0.124

1.000

Down-biw

1.000
-0.756
-0.009
-0.143
-0.390
-0.503
0.022
-0.333
-0.315
-0.227

Down-b w

1.000
-0.775
-0.108
-0.307
-0.354
-0.434
-0.085
-0.221
-0.169
-0.041

VaR

1.000
-0.153
0.034
-0.056
0.197
-0.197
0.273
0.042
0.104

VaR

1.000
-0.302
-0.167
0.217
0.424
-0.342
0.604
0.537
0.519

Down-bw

1.000
0.359
0.059
-0.135
0.822
-0.179
-0.236
-0.129

skew

1.000
0.337
0.162
-0.023
0.858
0.135
0.050
0.214

skew

1.000
0.103
-0.054
0.218
-0.243
-0.235
-0.210

skew5%

1.000
0.248
0.055
0.179
-0.074
-0.042
-0.024

1.000
0.846
-0.061
0.198
0.360
0.121

coskew1

1.000
0.875
0.155
-0.152
0.165
-0.283

1.000
-0.011
-0.020
0.161
-0.244

1.000
-0.173
0.326
0.444
0.189

kurt

1.000
0.161
-0.001
0.259

1.000
-0.170
-0.276
-0.111

kurt

coskew2

coskew2

skew5% coskew1

1.000
0.867
0.957

ICRGC

1.000
0.651
0.902

ICRGC

1.000
0.800

CCR

1.000
0.499

CCR

1.000

ICRGP

1.000

ICRGP

MSCI Data. E[R]: Mean returns; SR: Systematic risk (Beta); TR: Total risk; IR: Idiosyncratic risk; Size: Log of average market cap; Semimean: Semideviation with respect to
mean; Semi-rf: Semideviation with respect to risk-free rate; Semi-0: Semideviation with respect to 0; Down-iw: Downside beta calculated using observations when country index
and world index fall simultaneously; Down-w: Downside beta calculated using observations when world index falls; VaR: Value at risk; skew: Skewness, skew5%: Skewness calculated using formula 1; Coskew1: Co-skewness with world index using formula 2, Coskew-2: Coskew with world index using formula 3; kurt: Kurtosis; ICRGC: Log of average ICRG: composite country risk; CCR: Log of average CCR country risk, ICRGP: Log of ICRP political risk.

VaR
skew
skew5%
coskew1
coskew2
kurt
ICRGC
CCR
ICRGP

Down-bw

Semi-0
Down-biw

E[R]
SR
TR
IR
Size
Semimean
Semi - rf

Cross-Section Analysis Correlation MatrixAll Countries

EXHIBIT 2C

VaR
skew
skew5%
coskew1
coskew2
kurt
ICRGC
CCR
ICRGP

Down-bw

Semi-0
Down-biw

E[R]
SR
TR
IR
Size
Semimean
Semi - rf

Cross-Section Analysis Correlation MatrixEmerging Market Countries

EXHIBIT 2B

EXHIBIT 3A
Bivariate RegressionsDeveloped Countries
Risk Variable
SR
TR
IR
Size
Semimean
Semi - rf

c0
0.7777
0.9277
0.9155
1.0041
0.9274
1.3345
1.3063
0.8317

p-value
0.091
0.052
0.019
0.007
0.070
0.004
0.004
0.012

c1
0.1126
-0.0076
-0.0071
-0.0244
-0.0107
-0.1220
-0.1214
0.0624

p-value
0.807
0.919
0.925
0.704
0.925
0.271
0.279
0.857

R2
0.004
0.001
0.001
0.009
0.001
0.071
0.069
0.002

Adj-R2
-0.055
-0.058
-0.058
-0.050
-0.058
0.016
0.014
-0.057

0.6351
1.0930
0.9309
0.8909
0.8474
0.8419
1.0414
3.0763
3.5383
-7.2269

0.066
0.011
0.000
0.000
0.000
0.000
0.016
0.777
0.469
0.413

0.2310
0.0182
-0.2259
-0.1423
-0.6217
-0.5410
-0.0387
-0.4956
-0.6027
1.8384

0.436
0.580
0.364
0.850
0.255
0.274
0.681
0.840
0.586
0.359

0.036
0.018
0.049
0.002
0.075
0.070
0.010
0.003
0.018
0.050

-0.021
-0.039
-0.007
-0.057
0.021
0.015
-0.048
-0.056
-0.040
-0.006

Semi-0
Down-biw

c0
0.3699
-0.8754
-0.7778
0.6242
-0.8811
-0.4945
-0.4459
0.2849

p-value
0.272
0.059
0.107
0.202
0.084
0.412
0.447
0.484

c1
1.0050
0.1813
0.1831
0.2704
0.2792
0.2520
0.2526
0.8893

p-value
0.002
0.000
0.000
0.118
0.000
0.003
0.003
0.007

R2
0.319
0.517
0.470
0.092
0.464
0.285
0.286
0.250

Adj-R2
0.293
0.499
0.450
0.057
0.444
0.258
0.259
0.221

Down-bw
VaR
skew
skew5%
coskew1
coskew2
kurt
ICRGC
CCR
ICRGP

0.4131
-0.4617
0.8635
1.2288
1.2324
0.9742
0.3654
5.9795
3.3286
-5.2259

0.258
0.409
0.003
0.000
0.002
0.004
0.447
0.532
0.249
0.495

0.5713
-0.0780
0.6561
0.5588
-0.4184
-0.7513
0.1618
-1.1071
-0.5460
1.5644

0.006
0.002
0.012
0.639
0.781
0.140
0.035
0.625
0.478
0.395

0.259
0.312
0.220
0.009
0.003
0.082
0.160
0.009
0.020
0.028

0.230
0.286
0.190
-0.030
-0.035
0.047
0.128
-0.029
-0.018
-0.009

Semi-0
Down-biw
Down-bw
VaR
skew
skew5%
coskew1
coskew2
kurt
ICRGC
CCR
ICRGP

EXHIBIT 3B
Bivariate RegressionsEmerging Countries
Risk Variable
SR
TR
IR
Size
Semimean
Semi - rf

DRIVERS OF EXPECTED RETURNS IN INTERNATIONAL MARKETS

FALL 2000

EXHIBIT 3C
Bivariate RegressionsAll Countries
Risk Variable
SR
TR
IR
Size
Semimean
Semi - rf
Semi-0
Down-biw

c0
0.2480
-0.1256
0.0934
1.1247
-0.1784
0.0813
0.1170
0.2769

p-value
0.325
0.612
0.693
0.001
0.512
0.789
0.691
0.289

c1
0.9514
0.1318
0.1207
0.0031
0.2070
0.1832
0.1836
0.8451

p-value
0.000
0.000
0.000
0.968
0.000
0.000
0.000
0.000

R2
0.271
0.422
0.365
0.000
0.389
0.247
0.248
0.244

Adj-R2
0.249
0.409
0.351
0.022
0.376
0.230
0.232
0.227

Down-bw
VaR
skew
skew5%
coskew1
coskew2
kurt
ICRGC
CCR
ICRGP

0.3493
0.0753
0.8478
1.0650
1.0229
0.8964
0.3226
8.2620
3.4683
2.4870

0.135
0.793
0.000
0.000
0.000
0.000
0.289
0.081
0.012
0.538

0.5758
-0.0583
0.5886
0.6623
-0.8703
-0.8121
0.1586
-1.6560
-0.5851
-0.3163

0.000
0.000
0.002
0.421
0.300
0.018
0.004
0.130
0.084
0.737

0.269
0.275
0.198
0.015
0.024
0.118
0.167
0.050
0.065
0.003

0.253
0.259
0.180
-0.007
0.002
0.098
0.149
0.029
0.044
-0.020

E[Ri]: Mean return; Riski = Risk Variable; SR: Systematic risk (Beta); TR: Total risk; IR: Idiosyncratic risk; Size: Log of average market cap;
Semimean: Semideviation with respect to mean; Semi-rf: Semideviation with respect to risk-free rate; Semi-0: Semideviation with respect to 0;
Down-iw: Downside beta calculated using observations when country index and world index fall simultaneously; Down-w: Downside beta calculated using observations when world index falls; VaR: Value at risk; skew: Skewness; skew5%: Skewness calculated using formula 1; Coskew1:
Co-skewness with world index using formula 2; Coskew-2: Coskew with world index using formula 3; kurt: Kurtosis; ICRGC: Log of average
ICRG composite country risk; CCR: Log of average CCR country risk, ICRGP: Log of ICRP political risk.

The fourth risk measure is size. Size could be related

to liquidity and the amount of information available in the
market, which are legitimate risk factors. We find that
there is little relation between the average international
returns and size.
The fifth through seventh measures are semivariance
measures. All are similar, and from Exhibit 2 we know
they are all highly correlated with total variance. We find
that the semivariance measures account for a substantial
part of the variation in the emerging market returns but
not the developed market returns.
The eighth and ninth measures are downside betas.
There is some relation between the downside betas and
the emerging market returnsbut little relation with the
developed market returns.
The tenth measure is a value at risk measure. The
VaR measures the mean return in the 5% tail of the distribution. If this type of risk is priced, one would expect
a highly negative VaR (high risk) to command a high average return. This is exactly what we see for emerging
marketsbut not developed markets. Notice that the
slope of the line in the Exhibit 4J Emerging Markets
FALL 2000

graph is negative, implying that a more negative VaR

implies a higher expected return in emerging markets.
Harvey and Siddique [2000a] posit that investors care
about the skewness of their portfolio. Their model is
based on the work of Rubinstein [1973] and Kraus and
Litzenberger [1976]. In this world, investors value the contribution to the skewness of the portfolio, or the coskewness. That is, if an asset contributes positive skewness to
a diversified portfolio (high coskewness), that asset will be
valuable and will have a high price (low expected return).
If the asset contributes negative skewness (negative
coskewness), to get investors to purchase this asset the
price must drop ( higher expected return).
So, in the world of the augmented CAPM, coskewness should count, and skewness itself should not. This is
analogous to beta and variance. In the CAPM, it is only
the beta that is rewarded, not the total volatility. To be
consistent, only the systematic part of skewness (the
coskewness) should command a reward.
When we examine the two total skewness measures, we
see a positive relation for emerging markets but not developed markets. This is similar to what we see for total volatilEMERGING MARKETS QUARTERLY

10.00

0
-10.00

15.00
Total Risk

R2 = 0.4221

y = 0.1318x - 0.1256

5.00

5.00

15.00

Idiosyncratic Risk

10.00

R2 = 0.365

y = 0.1207x + 0.0934

20.00

2.00

Systematic Risk

1.00

EXHIBIT 4C -Idiosyncratic Risk

All Countries

0
-10.00

EXHIBIT 4B - Total Risk

All Countries

-1.00

R = 0.3101

y = 1.0114x + 0.2271

0
0.00
-1

EXHIBIT 4A - Systematic Risk

All Countries

Mean Returns

Mean Returns

Mean Returns

20.00

25.00

3.00

25.00

30.00

4.00

0
-10.00

15.00
Total Risk

5.00

15.00
Idiasyncratic Risk

10.00

R2 = 0.4699

y = 0.1832x - 0.7778

10.00

R2 = 0.5172

20.00

2.00

Systematic Risk

1.00

R = 0.3193

y = 1.005x + 0.3699

y = 0.1813x - 0.8754

5.00

Emerging Markets

0
-10.00

Emerging Markets

-1.00

0
0.00
-1

Emerging Markets

20.00

25.00

3.00

25.00

30.00

4.00

0
0.00

0.5

1.5

Systematic Risk

2.00

4.00

Idiosyncratic Risk

R2 = 0.0005

y = -0.0071x + 0.9155

4.00Total Risk6.00

R2 = 0.0006

y = -0.0076x + 0.9277

2.00

Developed Markets

0
0.00

0.5

1.5

0.50

R2 = 0.2656

y = 0.9764x + 0.046

Developed Markets

0
0.00

0.5

1.5

Developed Markets

Mean Returns
Mean Returns

Mean Returns
Mean Returns
Mean Returns

DRIVERS OF EXPECTED RETURNS IN INTERNATIONAL MARKETS

Mean Returns

FALL 2000

1.00

6.00

8.00

8.00

10.00

1.50

0
-10.00

1.00
Size

2.00

5.00

R2 = 0.3893

Semi - mean

10.00

y = 0.207x - 0.1784

3.00

15.00

4.00

0
0.00
-1

5.00

R2 = 0.2469

y = 0.1832x + 0.0813

Semi-rf

10.00

15.00

20.00

20.00

5.00

R = 0.0624

y = 0.2289x + 0.523

EXHIBIT 4F - Semivariance to Risk -Free Interest Rate

All Countries

0
-10.00

EXHIBIT 4E - Semivariance to Mean

All Countries

-1.00

EXHIBIT 4D - Size
All Countries

Mean Returns

Mean Returns

Mean Returns

0
0.00
-1

0
-10.00

1.00

5.00

R2 = 0.2851

10.00

Size

2.00

Semi-rf

10.00

Semi-mean

y = 0.252x - 0.4945

5.00

R2 = 0.4644

y = 0.2792x - 0.8811

Emerging Markets

0
0.00
-1

Emerging Markets

-1.00

Emerging Markets

Mean Returns
Mean Returns
Mean Returns

3.00

15.00

15.00

4.00

20.00

20.00

5.00

R2 = 0.0915

y = 0.2704x + 0.6242

0
0.00

0.5

1.5

1.00

R2 = 0.056

Size

2.00

y = 0.2939x + 0.0306

1.00

2.00

R2 = 0.0709

Semi-rf

3.00

4.00

Semi-mean

y = -0.122x + 1.3345

2.00

R2 = 0.0005

y = -0.0107x + 0.9274

Developed Markets

0
0.00

0.5

1.5

Developed Markets

0
0.00

0.5

1.5

Developed Markets

Mean Returns
Mean Returns
Mean Returns

FALL 2000

EMERGING MARKETS QUARTERLY

4.00

6.00

3.00

5.00

6.00

8.00

4.00

5.00

R2 = 0.2482

y = 0.1837x + 0.117

Semi-0

10.00

15.00

20.00

0
-10.00

Down Beta - iw

1.00

R2 = 0.2439

y = 0.8451x + 0.2769

-1.00

0
0.00
-1

1.00

3.00

Down Beta -w

2.00

R2 = 0.2691

y = 0.5758x + 0.3493

EXHIBIT 4I - Downside Beta, World Down

All Countries

-1.00

4.00

2.00

5.00

6.00

3.00

EXHIBIT 4H - Downside Beta, World and Local Market Down

All Countries

0
-10.00

EXHIBIT 4G - Semivariance to Zero

All Countries

Mean Returns

Mean Returns

Mean Returns

-1.00

0
0.00
-1

Emerging Markets

-1.00
Down Beta - iw

1.00

R2 = 0.2502

1.00

3.00
Down Beta - w

2.00

R2 = 0.2589

2.00

15.00

4.00

y = 0.8893x + 0.2849

Semi-0

10.00

y = 0.5713x + 0.4131

0
0.00
-1

5.00

R2 = 0.2863

y = 0.2526x - 0.4459

Emerging Markets

0
0.00
-1

Emerging Markets

5.00

6.00

3.00

20.00

0
0.00

0.5

1.5

2.00

0.50

R2 = 0.0361

1.00

Down Beta - w

1.00

3.00

Down Beta - iw

Semi-0

y = 0.231x + 0.6351

0.50

R2 = 0.002

y = 0.0624x + 0.8317

Developed Markets

0
0.00

0.5

1.5

1.00

R2 = 0.0685

y = -0.1214x + 1.3063

Developed Markets

0
0.00

0.5

1.5

Developed Markets

Mean Returns
Mean Returns

Mean Returns
Mean Returns
Mean Returns

DRIVERS OF EXPECTED RETURNS IN INTERNATIONAL MARKETS

Mean Returns

10

FALL 2000

1.50

4.00

2.00

1.50

5.00

-40.00

-20.00

Value At Risk

-30.00

0
0.00
-1

-0.40

-0.20

0
-10.00

Skew 5%

Skew

3.00

-10.00

0.20

0.40

0.60

4.00

0
0.00
-1

y = 0.6623x + 1.0651
R2 = 0.0145

2.00

R2 = 0.1978

y = 0.5887x + 0.8478

1.00

EXHIBIT 4L - Skew 5% Tail

All Countries

-1.00

5
y = -0.0583x + 0.0753 4
R2 = 0.2746
3

EXHIBIT 4K - Unconditional Skewness

All Countries

-50.00

EXHIBIT 4J - Value at Risk

All Countries

Mean Returns

Mean Returns

Mean Returns

-40.00

0
-10.00

-0.40

-0.20

Emerging Markets

-1.00

Emerging Markets

-50.00

Emerging Markets

Mean Returns
Mean Returns
Mean Returns

0
-10.00

Skew 5%

Skew

3.00

-10.00

0.20

0.40

0.60

4.00

0
0.00
-1

y = 0.5588x + 1.2288
R2 = 0.0086

2.00

y = 0.6561x + 0.8635
R2 = 0.2202

1.00

-20.00

Value At Risk

-30.00

R2 = 0.3122

y = -0.078x - 0.4618

-0.20

-0.10

0
0.00

0.5

1.5

0
0.00

0.5

Developed Markets

-0.50

-15.00

1.5

Developed Markets

-20.00

Developed Markets

Mean Returns
Mean Returns
Mean Returns

FALL 2000

EMERGING MARKETS QUARTERLY

11

-5.00

1.00

0.20
Skew 5%

0.10

0.30

y = -0.1423x + 0.8909
R2 = 0.0022

Skew

0.50

0.40

y = -0.2259x + 0.9309
R2 = 0.0486

Value At Risk

-10.00

y = 0.0182x + 1.0932
R2 = 0.0184

0.50

1.50

0
0.00

0.5

1.5

-0.40

-1.00

0
-10.00

Coskew2

-0.50

5.00
Kurtosis

10.00

y = 0.1586x + 0.3226
R2 = 0.1673

EXHIBIT 4O - Kurtosis
All Countries

-1.50

y = -0.8121x + 0.8964
R2 = 0.118

0
-10.00

Coskew1

-0.20

EXHIBIT 4N - Coskewness Measure 2

All Countries

-0.60

y = -0.8703x + 1.0229
R2 = 0.0238

EXHIBIT 4M-Coskewness Measure 1

All Countries

Mean Returns

Mean Returns

Mean Returns

15.00

0
-10.00

0.20

20.00

0.50

0.40

0
-10.00

Emerging Markets

-1.50

Emerging Markets

-0.60
Coskew1

-0.20

Coskew2

-0.50

5.00

Kurtosis

10.00

y = 0.1618x + 0.3654
R2 = 0.1604

-1.00

y = -0.7513x + 0.9742
R2 = 0.0818

-0.40

y = -0.4184x + 1.2324
R2 = 0.003

Emerging Markets

15.00

0
-10.00

0
-10.00

20.00

0.50

0.20

0
0.00

0.5

1.5

-0.20

2.00

-0.40

Developed Markets

-0.60

Developed Markets

-0.40

Developed Markets

Mean Returns
Mean Returns

Mean Returns
Mean Returns
Mean Returns

DRIVERS OF EXPECTED RETURNS IN INTERNATIONAL MARKETS

Mean Returns

12

FALL 2000

4.00

Kurtosis

0.20

0.40

y = -0.541x + 0.8419
R2 = 0.0698

0.40

6.00

8.00

y = -0.0387x + 1.0414
R2 = 0.0102

0
0.00
Coskew2

-0.20

0.5

1.5

0.20

y = -0.6217x + 0.8474
R2 = 0.0754

Coskew1

0
0.00

0.5

1.5

0
-13.90

4.00

4.20

4.30

4.40

ICRG Country Risk Composite

4.10

4.50

0
-13.90

4.00

4.10

3.00

4.30

ICRG Political Risk

4.20

4.00

5.00

4.40

4.50

4.60

y = -0.3163x + 2.487
R2 = 0.0025

CCR Country Rating

2.00

EXHIBIT 4R - ICRG Political Risk

All Countries

1.00

y = -0.5851x + 3.4683
R2 = 0.0647

0
-10.00

4.60

y = -1.656x + 8.262
R2 = 0.0501

EXHIBIT 4Q - Institutional Investor Country Credit Rating

All Countries

Mean Returns

EXHIBIT 4P - ICRGC Composite Country Risk

All Countries

4.30

ICRG Country Risk Composite

4.20

0
4.30

0
-13.90

1.00

4.00

Emerging Markets

0
-10.00

2.00

3.00

4.20

4.30
ICRG Political Risk

4.10

4.40

y = 1.5644x - 5.2269
R2 = 0.0279

CCR Country Rating

4.50

0
4.25

0.5

1.5

4.35

4.40

4.40

4.30

ICRG Political Risk

4.35

4.40

CCR Country Rating

y = 1.8385x - 7.2269
R2 = 0.0497

4.30

4.45

4.45

4.50

4.50

4.50

4.60

4.55

y = -0.4956x + 3.0763
R2 = 0.0025

ICRG Country Risk Composite

y = -0.6027x + 3.5384
R2 = 0.0178

Developed Markets

0
4.20

0.5

1.5

5.00

4.50

4.00

4.40

0.5

1.5

Developed Markets

y = -0.546x + 3.3286
R2 = 0.0195

4.10

y = -1.1071x + 5.9795
R2 = 0.0093

Developed Markets

Emerging Markets

0
-14.00

Emerging Markets

Mean Returns
Mean Returns
Mean Returns

Mean Returns

Mean Returns

Mean Returns
Mean Returns
Mean Returns

FALL 2000

EMERGING MARKETS QUARTERLY

13

EXHIBIT 5A
Multiple RegressionsDeveloped Markets
c0
0.8522
0.8177
0.8946
0.8671
1.0727
1.0284
0.7699

p-value
0.1000
0.1130
0.1150
0.1100
0.0270
0.0310
0.1130

c1
0.2774
0.1526
0.1220
0.2681
0.7604
0.7421
0.0943

p-value
0.6730
0.7670
0.7970
0.6810
0.1970
0.2070
0.8600

c2
-0.0385
-0.0169
-0.0253
-0.0565
-0.2431
-0.2403
0.0304

p-value
0.7160
0.8420
0.7020
0.7280
0.0970
0.1040
0.9390

R2
0.012
0.006
0.013
0.011
0.166
0.159
0.004

0.7115
0.9074
0.8732
0.6725
0.6200
0.6446
0.9574
2.8540
3.3791
-7.9100
0.9280
1.1992
0.9171
0.8827
0.6921
0.9245

0.1310
0.0650
0.0700
0.2100
0.1830
0.1660
0.1700
0.7990
0.5070
0.3930
0.0790
0.1230
0.1000
0.0050
0.0300
0.0600

-0.1446
0.4448
0.0607
0.2483
0.2397
0.2081
0.0755
0.1084
0.0959
0.1787
-0.0079
-0.0249
-0.0264
0.7752
0.7319
-0.0262

0.8010
0.4570
0.8970
0.6710
0.6080
0.6550
0.8760
0.2000
0.8390
0.7030
0.9700
0.7700
0.9560
0.0000
0.0000
0.7900

0.2857
0.0380
-0.2217
-0.3788
-0.6864
-0.5818
-0.0354
-0.4683
-0.5868
1.9555
0.0003
-0.0339
0.0295
-1.2421
-0.1819
0.1385

0.4470
0.3760
0.3920
0.6910
0.2330
0.2610
0.7220
0.8530
0.6070
0.3480
0.9990
0.6450
0.9680
0.0000
0.0000
0.7620

0.040
0.053
0.050
0.014
0.091
0.082
0.012
0.006
0.020
0.059
0.001
0.014
0.001
0.658
0.611
0.007

0.9436
TR / VaR
0.8558
TR / skew
0.9160
TR / skew5%
0.8716
TR / coskew1
0.9235
TR / coskew2
0.9947
TR / kurt
1.1478
TR / ICRGC
3.1693
TR / CCR
4.1312
TR / ICRGP
-7.1810
*Extreme multicollinearity.

0.0460
0.0720
0.0580
0.1500
0.0520
0.0400
0.1100
0.7780
0.4510
0.4330

-0.1133
0.1935
0.0026
0.0035
-0.0129
-0.0263
-0.0150
-0.0081
-0.0224
-0.0035

0.3080
0.3080
0.9730
0.9730
0.8620
0.7290
0.8480
0.9160
0.7800
0.0498

0.5699
0.0967
-0.2272
-0.1659
-0.6278
-0.5798
-0.0429
-0.5057
-0.7070
1.8327

0.2060
0.2510
0.3820
0.8740
0.2660
0.2670
0.6660
0.8450
0.5560
0.3760

0.099
0.082
0.049
0.002
0.072
0.077
0.013
0.003
0.023
0.050

Risk1 / Risk2
SR / TR
SR / IR
SR / Size
SR / Semimean
SR / Semi - rf
SR / Semi-0
SR / Down-biw
SR / Down-bw
SR / VaR
SR / skew
SR / skew5%
SR / coskew1
SR / coskew2
SR / kurt
SR / ICRGC
SR / CCR
SR / ICRGP
TR / IR
TR / Size
TR / Semimean
TR / Semi - rf *
TR / Semi-0*
TR / Down-biw
TR / Down-bw

ity, and is consistent with the notion that the emerging markets are not fully integrated into world capital markets.
For the coskewness measures, the results are similar
across the two methods. In both developed and emerging markets, there is a negative relation, suggesting that
more negative coskewness gets a higher expected return
which is what the theory would suggest.
From an asset pricing perspective again, it is the contribution to kurtosis of a well-diversified portfolio (the
cokurtosis) that should be priced. We do not attempt to
measure the cokurtosis. We find a positive relation
between kurtosis and returns in emerging markets but not
in developed markets.
14

DRIVERS OF EXPECTED RETURNS IN INTERNATIONAL MARKETS

The last measures are country risk ratings. Erb, Harvey, and Viskanta [1996a, 1996b] show that there is a relation between country risk ratings and subsequent returns.
The analysis in Exhibit 3 is different. We are comparing
average country risk to average returns during the same
period. Nevertheless, most of the results are consistent with
those of Erb, Harvey, and Viskanta.
We find for the ICRG Composite and the Institutional Investor Country Credit Rating that there is a negative relation in both developed and emerging markets.
The relation is weak, however. The political risk measure
in exhibit 4R has a positive slope, suggesting that higher
political risk is associated with lower expected returns
which is counter-intuitive.
FALL 2000

EXHIBIT 5B
Multiple RegressionsEmerging Markets

SR / Semi-0
SR / Down-biw

-0.8203
-0.7703
0.1156
-0.7869
-0.2003
-0.1763
0.1567

p-value
0.0800
0.1010
0.7980
0.1450
0.7430
0.7660
0.6890

c1
0.2851
0.4749
0.9271
0.2272
0.6672
0.6641
0.7405

p-value
0.3870
0.1220
0.0050
0.5640
0.1210
0.1240
0.0640

c2
0.1546
0.1436
0.1295
0.2401
0.1238
0.1245
0.4000

p-value
0.0020
0.0030
0.4040
0.0130
0.2720
0.2710
0.3060

R2
0.532
0.519
0.338
0.472
0.352
0.352
0.348

SR / Down-bw
SR / VaR
SR / skew
SR / skew5%
SR / coskew1
SR / coskew2
SR / kurt
SR / ICRGC
SR / CCR
SR / ICRGP
TR / IR
TR / Size
TR / Semimean
TR / Semi - rf
TR / Semi-0
TR / Down-biw

0.2985
-0.2127
0.0722
0.0677
0.2511
0.0283
-0.2439
5.5166
2.5628
-3.5800
-0.8258
-1.2669
-0.8183
-0.3113
-0.4206
-1.0380

0.4070
0.7130
0.8190
0.8600
0.5460
0.9390
0.5900
0.4930
0.2890
0.5820
0.0700
0.0180
0.0980
0.4650
0.3150
0.0280

0.7753
0.5912
0.9142
1.1087
1.0120
1.0088
0.9008
1.0083
1.0108
0.9835
0.5232
0.1743
0.2314
0.4239
0.4182
0.1560

0.1190
0.1900
0.0020
0.0010
0.0020
0.0010
0.0030
0.0020
0.0020
0.0030
0.0400
0.0000
0.1010
0.0000
0.0000
0.0000

0.1825
-0.0427
0.5653
1.4635
-0.6276
-0.7625
0.1222
-1.2202
-0.5938
0.9505
-0.3656
0.1879
-0.0842
-0.4872
-0.4763
0.4062

0.5520
0.2260
0.0100
0.1490
0.6200
0.0720
0.0660
0.5220
0.3580
0.5440
0.1650
0.1290
0.7060
0.0040
0.0040
0.1350

0.329
0.359
0.480
0.375
0.326
0.404
0.407
0.331
0.342
0.330
0.554
0.561
0.520
0.658
0.653
0.559

TR / Down-b w

-0.8699
-0.6711
-0.7931
-0.8635
-1.2476
-1.0189
-1.0111
-5.8392
0.9363
-9.7990

0.0630
0.1500
0.0900
0.0690
0.0210
0.0310
0.0440
0.4140
0.6500
0.0710

0.1624
0.2888
0.1618
0.1833
0.1903
0.1748
0.1684
0.1881
0.1807
0.1860

0.0010
0.0010
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000

0.1424
0.0661
0.2254
-0.2512
-1.4755
-0.5026
0.0501
1.1570
-0.4873
2.1230

0.4640
0.1350
0.3080
0.7700
0.1660
0.1690
0.4180
0.4860
0.3710
0.0980

0.528
0.559
0.537
0.519
0.554
0.553
0.530
0.527
0.533
0.568

Risk1 / Risk2
SR / TR
SR / IR
SR / Size
SR / Semimean
SR / Semi - rf

TR / VaR
TR / skew
TR / skew5%
TR / coskew1
TR / coskew2
TR / kurt
TR / ICRGC
TR / CCR
TR / ICRGP

c0

Multivariate Regressions

The multivariate regressions are presented in Exhibit

5. These regressions use two risk factors. We concentrate
on regressions with beta and one other risk factor and
regressions with total risk and one other risk factor.
If the world behaves according to the augmented
CAPM with skewness, we should have both beta and
coskewness in the regression. It turns out that this type of
regression is reasonably successful. In developed markets,
the market beta has a positive coefficient, and the coskewness has a negative coefficient. Neither coefficient is significant at conventional levels in the regression with only
19 observations.
FALL 2000

In the emerging markets sample, this regression is

more successful. The coefficient on the market beta is significant, and the coefficient on the coskewness measure
is negative and significant, at the 10% level using the second definition of coskewness. This regression is able to
explain 40% of the emerging market returns. There are
28 observations in the regression.
In the full sample, the regression with coskewness
definition 2 is quite successful. Both coefficients are significantly different from zero and are the right sign. The regression explains 40% of the variation in the average returns.
The alternative model is one with total variance and
total skewness. This model does not fare well in developed
markets; it does much better in emerging markets. The
EMERGING MARKETS QUARTERLY

15

EXHIBIT 5C
Multiple RegressionsAll Markets

SR / Semi-0
SR / Down-biw

c0
-0.3051
-0.2491
0.3817
-0.2923
-0.1101
-0.0878
0.0555

p-value
0.2370
0.3280
0.2400
0.2900
0.7130
0.7640
0.8360

c1
0.4638
0.6110
0.9780
0.4350
0.6483
0.6460
0.6375

p-value
0.0540
0.0080
0.0000
0.0970
0.0200
0.0210
0.0320

c2
0.1054
0.0944
-0.0451
0.1610
0.1114
0.1119
0.4776

p-value
0.0000
0.0000
0.5150
0.0010
0.0460
0.0450
0.0840

R2
0.470
0.459
0.278
0.427
0.335
0.335
0.320

SR / Down-bw
SR / VaR
SR / skew
SR / skew5%
SR / coskew1
SR / coskew2
SR / kurt
SR / ICRGC
SR / CCR
SR / ICRGP
TR / IR
TR / Size
TR / Semimean
TR / Semi - rf
TR / Semi-0
TR / Down-biw

0.1791
-0.0937
0.0580
0.0881
0.0873
-0.0099
-0.3296
7.5150
2.6242
2.5330
-0.4942
-0.9385
-0.0596
0.2858
0.1582
-0.2859

0.4770
0.7440
0.8020
0.7430
0.7460
0.9680
0.2950
0.0640
0.0270
0.4660
0.0790
0.0180
0.8290
0.1970
0.4580
0.2760

0.5493
0.5945
0.8795
1.0000
0.9747
0.9638
0.8553
0.9550
0.9561
0.9618
0.5095
0.1566
0.1963
0.4402
0.4334
0.1092

0.1210
0.0360
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0020
0.0000
0.0990
0.0000
0.0000
0.0000

0.3252
-0.0371
0.5256
1.0737
-1.0720
-0.8360
0.1302
-1.6895
-0.5971
-0.5372
-0.3758
0.1638
-0.1075
-0.5842
-0.5715
0.3710

0.1300
0.0320
0.0010
0.1310
0.1380
0.0040
0.0080
0.0720
0.0400
0.5100
0.0150
0.0100
0.5750
0.0000
0.0000
0.1020

0.309
0.344
0.427
0.308
0.307
0.396
0.381
0.323
0.339
0.278
0.495
0.503
0.426
0.622
0.616
0.457

TR / Down-bw
TR / VAR
TR / skew
TR / skew5%
TR / coskew1
TR / coskew2
TR / kurt
TR / ICRGC
TR / CCR
TR / ICRGP

-0.1824
0.0218
-0.0723
-0.1286
-0.1604
0.1218
-0.2510
-9.1055
-1.4252
-10.7762

0.4660
0.9290
0.7740
0.6050
0.5300
0.6230
0.3760
0.0530
0.3100
0.0030

0.1083
0.2729
0.1168
0.1381
0.1299
0.1230
0.1199
0.1654
0.1456
0.1714

0.0010
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000

0.2064
0.0823
0.1853
-0.5315
-0.4071
-0.2730
0.0467
2.0120
0.2930
2.4047

0.2030
0.0240
0.3000
0.4260
0.5340
0.3510
0.3590
0.0560
0.3470
0.0030

0.443
0.486
0.436
0.431
0.427
0.434
0.433
0.469
0.434
0.530

Risk1 / Risk2
SR / TR
SR / IR
SR / Size
SR / Semimean
SR / Semi - rf

loading on total variance is positive, and the loading on

skewness is negative which is what one would expect if
the markets are completely segmented. The two factors
can explain about 52% of the variation in the average
returns. In the model with all countries, the total variance
remains important but the skewness measures are no
longer significant.
Judging by R2, the best-fitting regression is the one
with total risk and a measure of semivariance, although
this is misleading because of the extreme collinearity
between the semivariance measures and the total risk.
The message of this analysis is as follows. A world
CAPM with coskewness appears to have some ability to
explain average returns in both developed and emerging
16

DRIVERS OF EXPECTED RETURNS IN INTERNATIONAL MARKETS

markets. In emerging markets, the systematic risks (beta

and coskewness) are not complete measures of risk. That
is, extra variation can be explained by adding total risks
to the regression. Even so, the results of Bekaert and
Harvey [2000] suggest that these total risks may be becoming less important with capital market liberalizations.
IV. SUMMARY

I have examined 18 measures of risk in 47 international markets. Particular attention is paid to a measure of
downside implied by asset pricing theory: coskewness.
This risk measure captures the contribution that an asset
makes to a well-diversified portfolios total skewness. A
FALL 2000

negative coskewness would imply that adding an asset to

the portfolio will decrease the skewness of the portfolio.
Given that investors like positive not negative skewness,
this asset would have to have a high expected return to get
investors to purchase it.
Risk measures implied by asset pricing theory, in
particular world beta and coskewness, work reasonably
well in capturing the cross-section of average returns in
world markets. Yet, consistent with the evidence in
Bekaert and Harvey [1995], many emerging markets
appear to be impacted by total risk measures like variance
and skewness. This is consistent with their less-thancomplete integration with world capital markets.
ENDNOTES

Erb, Claude B., Campbell R. Harvey, and Tadas E. Viskanta.

Expected Returns and Volatility in 135 Countries. The Journal of Portfolio Management, Vol. 22, No. 3 (1996a), pp. 46-58.
. Political Risk, Financial Risk and Economic Risk.
Financial Analysts Journal Vol. 52, No. 6 (November/December 1996b), pp. 28-46.
Estrada, Javier. The Cost of Equity in Emerging Markets: A
Downside Risk Approach. Emerging Markets Quarterly, Vol. 4,
No. 3 (Fall 2000), pp PAGES TO FOLLOW ONCE
JOURNAL IS PAGINATED
Harvey, Campbell R. Predictable Risk and Returns in Emerging Markets. Review of Financial Studies, Vol. 8. No. 3 (1995a),
pp. 773-816.

I wish to thank Lakshman Easwaran for his excellent

research assistance and comments.

Harvey, Campbell R. The Risk Exposure of Emerging Equity

Markets. World Bank Economic Review, Vol. 9, No. 1 (1995),
pp. 19-50.

REFERENCES

. The World Price of Covariance Risk. Journal of Finance,

46 (1991), pp. 111-157.

Bekaert, Geert. Market Integration and Investment Barriers in

Emerging Equity Markets. World Bank Economic Review, 9
(1995), pp. 75-107.
Bekaert, Geert, Claude B. Erb, Campbell R. Harvey, and
Tadas E. Viskanta. The Cross-Sectional Determinants of
Emerging Equity Market Returns. in Peter Carman, ed.,
Quantitative Investing of the Global Markets: Strategies, Tactics,
and Advanced Analytical Techniques. Chicago: Glenlake Publishing, 1997, pp. 221-272.

Harvey, Campbell R. and Akhtar Siddique. Autoregressive

Conditional Skewness. Journal of Financial and Quantitative
Analysis, Vol. 34, No. 4 (1999), pp. 465-488.
. Conditional Skewness in Asset Pricing Tests. Journal
of Finance, 55 (2000a), pp. 1263-1295.
. The Cross-Section of Expected Risk Exposure. Working paper, Duke University, 2000b.

. Distributional Characteristics of Emerging Market

Returns and Asset Allocation. The Journal of Portfolio Management, Vol. 24, No. 2 (Winter 1998), pp. 102-116.

Kraus, Alan, and Robert Litzenberger. Skewness Preference and

the Valuation of Risk Assets. Journal of Finance, 31 (1976), pp.
1085-1100.

Bekaert, Geert and Campbell R. Harvey. Emerging Equity

Market Volatility. Journal of Financial Economics, Vol. 43, No.
1 (1997), pp. 27-77.

Rubinstein, Mark. The Fundamental Theory of Parameterpreference Security Valuation. Journal of Financial and Quantitative Analysis, 8 (1973), pp. 61-69.

. Foreign Speculators and Emerging Equity Markets.

Journal of Finance, 55 (2000), pp. 565-613.
. Time-Varying World Market Integration. Journal of
Finance, Vol. 50, No. 2 (1995), pp. 403-444.
Black, Fischer, Michael Jensen, and Myron Scholes. The
Capital Asset Pricing Model: Some Empirical Tests. Michael
Jensen, editor. Studies in the Theory of Capital Markets. New
York: Praeger, 1972.

FALL 2000

EMERGING MARKETS QUARTERLY

17