International Portfolio Diversification Benefits:
Cross-Country Evidence from a Local Perspective
Joost Driessen and Luc Laeven*
December 2004
Abstract: We investigate how the benefits of international portfolio diversification differ
across countries from the perspective of a local investor. We find that the benefits of
investing abroad are largest for investors in developing countries, including when
controlling for currency effects. Most of the benefits are obtained from investing outside
the region of the home country. These global diversification benefits remain large when
controlling for short-sales constraints in developing stock markets. In addition, the gains
from international portfolio diversification appear to be largest for countries with high
country risk. Other differences across countries, such as size of the stock market, size of
the banking sector, and trade openness do not explain differences in the gains from
international portfolio diversification beyond the first-order effect of the level of country
risk. In addition to this cross-sectional evidence, we also provide evidence that
diversification benefits vary over time as country risk changes. We find that
diversification benefits have decreased for most countries in our sample over the past two
decades. Our results show the importance of further liberalization of international
financial markets, especially from the perspective of investors in developing countries.
JEL classifications: F30, G11, G15, G18
Keywords: Financial Globalization, Stock markets, Financial Liberalization, Portfolio
Diversification.
* Driessen is at the University of Amsterdam. Laeven is at the World Bank and CEPR. We would like to
thank Jerry Caprio, Stijn Claessens, Patrick Honohan, Frank de Jong, Giovanni Majnoni, Franz Palm,
Roberto Rocha, Haluk Unal, Dimitri Vittas, and numerous seminar participants for helpful comments and
suggestions. Contact information for Driessen: University of Amsterdam, Roetersstraat 11, 1018 WB
Amsterdam, The Netherlands. E-mail j.j.a.g.driessen@uva.nl. Contact information for Laeven: World
Bank, 1818 H Street NW, 20433 Washington DC, USA. E-mail llaeven@worldbank.org. The views
expressed in this paper do not necessarily represent those of the World Bank.
International Portfolio Diversification Benefits:
Cross-Country Evidence from a Local Perspective
Abstract: We investigate how the benefits of international portfolio diversification differ
across countries from the perspective of a local investor. We find that the benefits of
investing abroad are largest for investors in developing countries, including when
controlling for currency effects. Most of the benefits are obtained from investing outside
the region of the home country. These global diversification benefits remain large when
controlling for short-sales constraints in developing stock markets. In addition, the gains
from international portfolio diversification appear to be largest for countries with high
country risk. Other differences across countries, such as size of the stock market, size of
the banking sector, and trade openness do not explain differences in the gains from
international portfolio diversification beyond the first-order effect of the level of country
risk. In addition to this cross-sectional evidence, we also provide evidence that
diversification benefits vary over time as country risk changes. We find that
diversification benefits have decreased for most countries in our sample over the past two
decades. Our results show the importance of further liberalization of international
financial markets, especially from the perspective of investors in developing countries.
2
1. Introduction
Institutional investors in small countries are often restricted by their government to invest
only in assets traded in their home country (see, for example, IMF (2001)). This is
potentially costly for these investors, since these restrictions reduce the diversification
possibilities.1 Over the past decades several countries have lifted a large number of their
investment restrictions2, which has fostered trading in assets abroad. However, most
investors still invest largely at home. In fact, even when there are no official restrictions,
investors typically still invest large amounts of their money in domestic stocks. This socalled ‘home-bias’ of financial assets has been well documented in the literature3 and has
raised a puzzle, because it is generally thought that the gains from international
diversification are large.4
In this paper, we (i) measure the potential international diversification benefits for
a large number of countries, and (ii) analyze the cross-country differences in these
international diversification benefits. Most of the literature on international portfolio
diversification takes a U.S. perspective. We investigate for a large cross-section of
countries with both large and small stock markets, if adding international stock
investment opportunities leads to diversification benefits for a domestic investor that
invests in local stocks only. We also measure the economic size of these diversification
benefits. Subsequently, we try to explain both the cross-country and time series variation
in these international diversification benefits using financial and macro-economic
variables.
1
Investment restrictions impose a tax to invest in or hold foreign assets (see Black (1974) or Stulz (1981)
for a general model of barriers to investment). In most countries with foreign investment restrictions, these
restrictions are imposed by law; in others, such as Switzerland, they are partly chosen by individual
companies (see Stulz and Wasserfallen (1995)).
2
Investment restrictions are not only related to investments abroad, but can also restrict investment of
residents at home.
3
See, for instance, French and Poterba (1991), Lewis (1996), Baxter and Jermann (1997), and Li (2004).
Coval and Moskowitz (1999) show that even at home there often is a home bias in the sense that local
investors tend to invest mostly in companies that are located close by.
4
For instance, Harvey (1995) shows that from a US perspective, large benefits can be attained from
investing in emerging markets, where stock returns are driven to a larger extent by local factors. See also
Obstfeld (1994). Others argue that the gains from international diversification can largely be achieved
indirectly at home through investments in stocks of multinational firms (see Rowland and Tesar, 1998) or
country funds and depositary receipts (see Errunza, Hogan, and Hung 1999).
3
This paper thus extends Huberman and Kandel (1987), Bekaert and Urias (1996),
DeRoon, Nijman, and Werker (2001), and Li, Sarkar, and Wang (2003), since these
articles all analyze international diversification benefits for a U.S. investor. We compare
two types of diversification for an investor that currently invests in the local equity index.
First, we estimate the regional diversification benefits by allowing the investor to invest
in a regional equity index for the region in which the country is located. Second, we
estimate the global diversification benefits when allowing the investor to invest in equity
indices for Europe, the US, and Far East. The reason for distinguishing between these
two possibilities is that investors often prefer to invest in familiar investment
opportunities as opposed to foreign or unfamiliar investments (see Huberman (2001) and
Grinblatt and Keloharju (2001)). Therefore, it is interesting to see how much of the
international diversification benefits can be achieved by investing regionally only.
Measuring the potential benefits of regional diversification is also of interest to policy
makers, because many countries are currently considering the setup of regional stock
exchanges.
For this empirical analysis, we use monthly data on stock market index returns
from 52 countries for the period 1985 to 2002. We apply the regression framework
developed by Huberman and Kandel (1987) and DeRoon, Nijman, and Werker (2001) to
statistically test whether there are diversification benefits for a domestic investor with
mean-variance utility, who currently only invests in the local stock index (a ‘spanning
test’). We also calculate the economic size of the diversification benefits on the basis of
improvements in Sharpe ratios or expected returns.5 We both investigate the case where
there are no market frictions and the case in which there are short sales constraints (either
on assets in developing countries only or on all assets). Also, we calculate diversification
benefits for investors who are interested in the local currency returns, and investors that
care about U.S. dollar returns.
Our results are as follows. In case of no short sales constraints, the empirical
results reveal economically large and statistically significant diversification benefits for
investors in almost all countries, both when returns are measured in local currency and in
5
Li, Sarkar, and Wang (2003) use the increase in expected return, that is obtained when adding assets to the
investment set, as a measure for the size of diversification benefits.
4
U.S. dollars. Global diversification benefits tend to be much larger than regional
diversification benefits. Nevertheless, a fair amount of diversification can be obtained by
investing only in a regional index. This within-region diversification is often neglected in
the literature.
The results related to the global diversification benefits are not much affected
when assuming short sales constraints in developing countries, because for almost all
countries it is optimal to allocate a positive share of the total asset portfolio to the local
stock market index. In line with DeRoon, Nijman, and Werker (2001), including short
sales constraints for all countries (i.e., including developed countries and global indices)
leads to a decrease in the global diversification benefits for most countries, especially
when diversification benefits are measured in terms of improvements in expected returns.
Focusing on the increase in the Sharpe ratio, we find that the gains from
international portfolio diversification are larger for developing countries, in particular for
countries with high country risk (as measured by the ICRG country risk measure). This
result appears consistent with the finding that these countries have been less integrated in
world financial markets over the last two decades (as shown by, for example, Bekaert,
Harvey and Lumsdaine (2002)). A regression analysis indicates that other differences
across countries, such as size of the stock market, size of the banking sector, and trade
openness do not explain differences in the gains from international portfolio
diversification beyond this first-order effect of the level of country risk.
Finally, we use the country risk measure to analyze how global diversification
benefits change over time. For most countries, we find that the diversification benefits
have decreased over the sample period. This is primarily due to an increase in
correlations of local returns with global indices, and a decrease in the variances of local
indices. These results appear consistent with increased financial integration over the
period 1985-2002.
The paper continues as follows. Section 2 will explain how we measure the
benefits of international diversification for domestic investors. In Section 3 we report our
estimates of these potential gains from international diversification. Section 4 presents a
cross-country comparison of diversification benefits. Section 5 analyzes the time-series
variation in diversification benefits. Section 6 concludes.
5
2. Measuring International Diversification Benefits
In this section, we describe how we measure the benefits of international diversification
for domestic investors. We use the standard mean-variance framework of Markowitz
(1952), thereby assuming either a mean-variance utility function for the investor, or
normally distributed asset returns. We analyze the diversification benefits for an investor
that currently only invests in the local equity index and a risk-free asset. We consider for
each country two types of diversification opportunities:
A. Regional diversification: A regional equity index for the country’s region.
B. Global diversification: MSCI indices for the US, Europe, and Far East.
In case of asset set B, investors are allowed to optimize the portfolio weights on the three
MSCI indices. For each type of diversification, we then measure the diversification
benefits by analyzing whether the risk-return tradeoff of a domestic index portfolio can
be improved by adding one of the two sets of assets A and B.
The existing empirical work on diversification benefits (see Huberman and
Kandel (1987), Bekaert and Urias (1996), and DeRoon, Nijman, and Werker (2001))
typically looks at global diversification benefits for a U.S. investor. We take the
viewpoint of domestic investors in many different countries. This is an important
distinction, because the benefits of investing abroad for investors in countries with less
well-diversified and developed stock markets as those in the United States are likely to
benefit more from investing abroad than U.S. investors do. Since the United States has
one of the most diversified economies and one of the most developed stock markets in
the world by any standard, the U.S. perspective taken in most of the existing work is
unlikely to lead to results that are representative for investors in most other countries. We
see this as an important contribution of our paper to the literature.
We consider both the case of frictionless markets, in which investors can short
assets without any costs, and the case where there are short selling constraints. In reality,
6
short selling constraints are often present for investors, and they can be especially
important for investors in developing countries. While it seems realistic to assume short
selling constraints in developing countries, for many investors in developed markets short
selling is not an issue anymore (especially for institutional investors). We therefore also
examine the diversification benefits when only developing countries are subject to short
selling constraints.
In case we assume short selling constraints for all countries, we also impose short
selling constraints on the regional and global stock market indices. In case we assume
short selling constraints for developing countries only, we also impose short selling
constraints on the regional stock market indices that are based on a set of developing
countries.
In addition to allowing for the presence of market frictions for an investor, we
also vary the currency that the investor is interested in. First, we consider a local investor
that uses the local currency. In this case, we use currency rate data to obtain returns
denominated in the currency of the country of interest. We also calculate results in terms
of U.S. dollar returns for all countries and indices, since, in practice, investors in many
countries use U.S. dollars as their base currency.
We both analyze the statistical significance of the diversification possibilities, as
well as the economic significance of these possibilities. To calculate the statistical
significance, we use the regression tests for mean-variance spanning developed by
Huberman and Kandel (1987), DeRoon, Nijman, and Werker (2001), and DeRoon and
Nijman (2002). In this case, it is examined whether adding N new assets to a given set of
K benchmark assets leads to a significant shift in the mean-variance frontier. In other
words, it is tested whether the K benchmark assets span the market of all K+N assets. In
our case, the K benchmark assets are given by domestic index only, so that K equals 1,
and the N additional assets are given by asset sets A or B, so that N equals 1 or 3
respectively. In case of frictionless markets, this test can be performed using the
following regression
rt +1 = α + βRt +1 + ε t +1
(1)
7
with the usual distributional assumptions. These assumptions include that returns are
i.i.d.. In equation (1), rt+1 is a N-dimensional column vector with the N returns on the
additional assets, Rt+1 is a K-dimensional return vector for the K benchmark assets, α is a
N-dimensional constant term, β is a N by K matrix with slope coefficients, and εt+1 is a Ndimensional vector with zero-expectation error terms. Then, the null hypothesis that the K
benchmark assets span the entire market of all K+N assets is equivalent to the restrictions
α = 0 and βι K = ι N
(2)
where ιN is a N-dimensional vector with ones. If (2) holds, the return on each additional
asset can be decomposed into the return on a portfolio of benchmark assets plus a zeroexpectation error term that is uncorrelated with the benchmark portfolio return. Thus, in
the case of mean-variance spanning, such an additional asset can only add to the variance
of the portfolio return and not to the expected return, and the investor will not include the
additional asset in his portfolio. This implies that, if the spanning hypothesis holds, the
optimal mean-variance portfolio only consists of the K benchmark assets. Using data on
stock returns, equation (1) can be estimated using OLS and the 2N restrictions in equation
(2) can easily be tested using, for example, a Wald test. In our case, the spanning test
tests whether the local stock market index spans a portfolio that also includes asset set A
(a regional stock market index) or asset set B (the stock market indices of the US, Europe
and Far East). Note that for these spanning tests no information on a risk-free rate is
needed.
In case of short selling constraints, similar regression techniques as in equations
(1) and (2) can be used, although now inequality constraints have to be tested. We follow
the approach of DeRoon, Nijman, and Werker (2001). In this case of short selling
constraints, the power of the spanning test may be low in small samples. We will use 17
years of monthly data in order to minimize these small sample problems. Also, in most
cases we will introduce short sales constraints on only one of the assets, namely the
country’s local index. Since the simulation results in DeRoon, Nijman, and Werker
(2001) show that the small sample properties of the spanning test depend on the number
8
of short-sale restrictions, we therefore minimize the adverse effects of adding short
selling constraints.
We measure the economic significance of the diversification benefits in two ways.
First, we calculate the increase in the Sharpe ratio when adding the new N assets to the K
benchmark assets. More precisely, we calculate the Sharpe ratio for the mean-variance
efficient portfolio based on the K benchmark assets (and a risk-free asset) and the Sharpe
ratio for the mean-variance efficient portfolio based on all K+N assets (and a risk-free
asset), both in the case of frictionless markets and in the case of short selling constraints.
A difference between these two Sharpe ratios indicates that investors can increase their
risk-return tradeoff by investing in the additional N assets.
As an alternative measure of the economic significance of the diversification
benefits, we use the increase in expected return when adding the new N assets to the K
benchmark assets given the same (or lower) variance as for the optimal portfolio of K
benchmark assets. In this case we assume that there is no risk-free asset. This measure is
thus more conservative, in the sense that investors are not allowed to use a risk-free asset
to improve the risk-return tradeoff.
Finally, we note that our analysis has so far assumed that returns are i.i.d.. In
Section 5 we relax this assumption and incorporate time-varying diversification
opportunities.
3. Empirical Results on International Diversification Benefits
For the empirical analysis, we use data from several sources. For developed countries, we
use the stock market index developed by Morgan Stanley Capital International (MSCI).
For developing countries, we use the S&P/IFC Global Index from Standard & Poor’s
(S&P). S&P acquired this data in January 2000 from the International Finance
Corporation (IFC). The Global index is developed for developing countries with
relatively well-developed stock markets.6 The total number of countries in the sample is
6
We do not include data from developing countries covered by the S&P/IFC Frontier index because of the
limited sample period for which there is data on stock market index returns for these countries.
9
52, and includes 23 developed markets and 29 developing markets. We collect threemonth U.S. Treasury bill rates from the International Financial Statistics database
maintained by the International Monetary Fund (IMF). Data on GDP per capita (in
constant 1995 U.S. dollar prices), stock market capitalization of listed companies, private
credit to GDP, and trade to GDP (defined as the ratio of exports plus imports to GDP)
come from the World Development Indicators of the World Bank. We also collect
country risk ratings for each country as reported by the International Country Risk Guide
(ICRG) maintained by the Political Risk Services group. We use the ICRG composite
index as a general measure of country risk. The rating ranges from 0 to 100, with a higher
rating indicating less country risk.
For the stock index return data, we use monthly data for the period 1985 to 2002,
when available. A longer history of reliable stock return data is difficult to obtain for
stock markets in developing countries. In addition, we also believe that, since the
characteristics of developing countries typically change structurally over time, using a
longer data period makes the stationarity assumptions that are needed for the spanning
tests less reasonable. For some countries the data start after 1985. We use all available
data for each country to calculate the diversification benefits and to conduct the spanning
tests. For all countries, we download monthly currency data from Datastream to be able
to calculate returns in U.S. $ terms and in terms of the local currency.
As proxy for the risk-free rate in each country, we use the U.S. three-month
Treasury bill rate reported by the IMF.7 Note that this assumption only influences the
Sharpe ratios in case of local currency returns, and not the Sharpe ratio when using U.S. $
returns, the spanning test, and the expected return measure. Data on other country-level
variables are collected for the period 1985 to 2002, when available.
Table 1 presents the regional setup. We assign developing countries to their
regional S&P/IFC stock market index and developed countries to their regional MSCI
stock market index. The table also shows the period for which stock return data are
available for each country in the sample.
We start by estimating the regional and global diversification possibilities for
local investors, considering both returns measured in local currency and in U.S. dollars,
7
We do not have information on local risk-free rates for all countries.
10
and allowing for either no short selling constraints, short selling constraints in developing
countries only, or short selling constraints in all countries. We estimate both increases in
Sharpe ratios and increases in expected returns. Table 2 reports the results summarized by
averaging country-by-country results within each region. Figures 1 and 2 present a
graphical representation of the individual country results for the global diversification
benefits in U.S. dollars and measured in terms of either improvements in Sharpe ratios or
improvements in expected returns.
The regional diversification benefits appear largest on average for countries in
Eastern Europe and can be substantial, even when allowing for short selling constraints.
For instance, the increase in expected return for Eastern European countries of investing
in the region is 0.3 percent per month on average as expressed in U.S. dollar terms, even
when short selling constraints are present. However, for other regions no improvements
in expected returns can be obtained from investing within the region when short selling
constraints are present.
For most countries, the benefits of global diversification significantly outweigh
the benefits of regional diversification. The increase in expected return as a result of
global diversification under the assumption of no market frictions and expressed in local
currency returns ranges from a low of 0.0 percent per month for the Netherlands and
Switzerland to a high of 3.3 percent per month for Argentina (not reported). The large
figure for Argentina can be explained by the fact that the variance of the Argentinean
stock index return is very high, so that an Argentinean investor can choose a globally
diversified portfolio with high variance and high expected return.
We also find strong benefits from global diversification when we use increases in
Sharpe ratios as measure of diversification benefits. The increase in Sharpe ratio without
market frictions ranges from a low of 1.3 percent for Belgium and Switzerland to a high
of 46.3 percent for Turkey when expressed in local currency returns (not reported), and
from a low of 1.5 percent for the Netherlands to a high of 27.1 percent for Brazil when
expressed in U.S. dollars (see Figure 1). Global diversification benefits appear largest on
average for the countries in Eastern Europe. These figures suggest that the economic size
of global diversification benefits is large for most countries. Averaged over all countries,
the increase in the Sharpe ratio without market frictions is about 11 percent on a monthly
11
basis in U.S. dollar terms. Without global diversification, the average Sharpe ratio is
around 10 percent, while for optimal global diversification the average Sharpe ratio
across countries is 21 percent. Thus, the Sharpe ratio more than doubles on average if a
local investor is allowed to diversify globally. For the same case, the increase in expected
returns is 0.98 percent per month, which is also substantial.
In many developing countries, short-selling is not possible. Therefore, using
short-selling constraints is expected to give more realistic results. However, the results
related to the global diversification benefits are not much affected when assuming short
sales constraints in developing countries, because for almost all countries it is optimal to
allocate a positive share of the total asset portfolio to the local stock market index.
However, when we include short sales constraints for all countries (i.e., including
developed countries and the global indices) – arguably a less realistic assumption than
short sales constraints in developing countries only – we find a substantial decrease in the
benefits of global diversification for most countries, especially when diversification
benefits are measured in terms of improvements in expected returns. Still, the results
show that substantial and statistically significant global diversification benefits can be
obtained for most countries when assuming short sales restrictions for all countries. The
increase in Sharpe ratio with market frictions in all countries and expressed in local
currency returns ranges from a low of 0.0 percent for the Netherlands and Switzerland to
a high of 46.2 percent for Turkey, and the increase in expected return with market
frictions in all countries and expressed in local currency returns ranges from a low of 0.0
percent for several countries to a high of 1.4 percent for Argentina (not reported).
For most countries, the results are not altered much when we estimate
diversification benefits on the basis of U.S. dollar returns rather than local currency
returns. However, for some countries with a limited sample period (such as Russia) there
can be quite a substantial difference. In these cases, part of the international
diversification benefits for local currency investors is due to currency effects, instead of
stock returns in the foreign countries. In the empirical analysis, we confirm that our
results are not affected by countries for which we have shorter sample.
We now turn to the results on the statistical tests of mean-variance spanning.
Table 3 summarizes the results of the spanning tests of each country on a regional basis
12
by presenting the percentage of cases in which the spanning hypothesis is rejected. In the
case of global diversification possibilities, the spanning hypothesis is rejected for all
countries at the 1 percent confidence level, both when assuming that the investor does not
face any short sales constraints and when assuming that the invest faces short selling
constraints in developing countries only. This shows that there are statistically significant
global diversification benefits for domestic investors around the world. In most cases, the
p-values associated with the Wald test are so small that one would reject the spanning
hypothesis for all reasonable confidence levels.
When short-selling constraints are added for all countries, the spanning
hypothesis is not rejected for 18 out of the 52 countries in the sample (6 of which in
Western Europe), suggesting that short sales constraints on all assets eliminate the
diversification benefits for a substantial number of countries. Figures 1 and 2 graphically
show the large impact of short sales constraints. Still, even in case of short sales
constraints on all assets, which is admittedly an extreme assumption, there are substantial
diversification benefits for many countries.
Summarizing, in this section we have shown that there are statistically and
economically significant regional and global diversification benefits for domestic
investors that only invest in the local index, provided that these investors can short the
foreign stock indices. If this is not possible, international diversification benefits are
much smaller and in some cases statistically insignificant.
4. Explaining the Cross-Country Variation in Diversification Benefits
The increases in Sharpe ratios reported in Figures 1 and 2 indicate that there is
large cross-country variation in the benefits of international portfolio diversification, or in
other words, in the degree of risk diversification that can be attained by investing in home
country equities. In this section, we explore what country-specific factors can explain
these differences. In particular, we are interested whether the benefits of international
diversification are larger for less developed countries. Previous research has shown that
many developing countries are not (well) integrated with world capital markets (see, for
13
example, Harvey 1995). This would suggest that there are substantial benefits to
international portfolio diversification for investors in less developed countries. To test for
this hypothesis we use a regression framework that controls for other country-specific
factors. We only use the increase in the Sharpe ratio as a measure for diversification
benefits for this cross-country analysis. This is because the increase in the expected return
measure will be influenced by the level of the local stock market variance, so that crosscountry variation in this measure may be partially caused by variation in the local stock
market variance.8
We collect the following country-specific data. First, we collect data on the
market capitalization of the stock market. This variable proxies for the size of the stock
market. It is expected that the benefits of international diversification are smaller for
countries with sizeable local stock markets as such markets allow for substantial
diversification at home. Stock market capitalization may also proxy for the degree of
stock market integration (Bekaert and Harvey 1997). Two markets are fully integrated if
assets with the same risk command the same expected returns in both markets. Bekaert
and Harvey (1995) show that the stock markets of many developing countries are not
well integrated into world capital markets. We also collect data on trade to GDP to
measure trade openness and data on private sector credit to GDP to measure the effect of
the size of a country’s banking system on diversification benefits. Trade openness serves
as a proxy for the country’s integration in world goods markets. It is expected that
countries that are better integrated in world goods markets are in a better position to
achieve integration in world financial markets, and therefore that such countries would
benefit less from international diversification. Trade openness as a measure of economic
and financial integration has been used before in related work by Bekaert and Harvey
(1997). Private credit to GDP serves as a proxy for the level of financial development of
the country and has been used as such in a large number of papers (see, for example,
Beck, Levine and Loayza 2000). We consider this variable because it is expected that
economies with more developed financial systems have better diversification
opportunities at home and would therefore benefit less from international diversification.
8
We get very similar results when using the expected return measure of diversification benefits, but only
report the Sharpe ratio results.
14
More generally, these measures proxy for the level of institutional development in the
country. It is expected that more developed countries benefit less from international
portfolio diversification. Since different measures of institutional development appear to
be highly correlated (see, for example, Beck, Levine and Loayza 2000), we limit our
exercise to a small number of variables.
Table 4 provides summary statistics on the differences of these variables and our
Sharpe ratio measure of diversification benefits between developing and developed
countries. Our classification of developing countries follows the criterion of the World
Bank, which is based on GDP per capita. The average of GDP per capita ranges from a
low of 249.5 U.S. dollars in Nigeria to a high of 44,309 U.S. dollars in Switzerland. We
find that developing countries have on average smaller capital markets. Still, some
developing countries such as Brazil and the Republic of Korea have sizeable stock
markets that are broader than those found in many developed countries. The average of
market capitalization ranges from a low of 1.14 billion U.S. dollars in the Slovak
Republic to a high of 8,140.0 billion U.S. dollars in the United States. We also find that
developed countries have more developed banking sectors, as measured by private credit
to GDP, and that the country risk of developing countries tends to be significantly higher.
The average of private credit to GDP ranges from a low of 11.2 percent in Russia to a
high of 188.7 in Japan, and the average of ICRG ranges from a low of 51.9 in Pakistan to
a high of 90.6 in Switzerland. Trade openness does not differ significantly between
developed and developing countries. The average of trade to GDP ranges from a low of
18.3 percent in Argentina to a high of 349.7 percent in Singapore.
Table 4 also compares the increase in Sharpe ratios between developing countries
and developed countries. Our data set includes 52 countries, of which 29 are developing
and 23 are developed countries. When comparing the increase in Sharpe ratios due to
international diversification for the different groups of countries, we find that the global
diversification benefits tend to be larger for developing countries than for developed
countries, and the difference is statistically significant in all cases. For example, for the
case of U.S. dollar returns and no short sales restrictions, the average Sharpe ratio
increase due to global diversification equals 13.5 percent for developing countries. For
developed countries, the average increase is smaller and equals 7.8 percent.
15
The benefits of international diversification are likely to be larger for developing
countries because these countries are not fully integrated into world markets (Harvey
1995). Factors such as investment restrictions, political risk, investor protection, and
foreign exchange regulations all contribute to the degree of stock market integration. As a
proxy for these factors we use the ICRG country risk rating. This rating measures not
only political risk (including such factors as corruption and law and order tradition) but
also financial and economic risk. The ICRG variable is a widely used proxy for country
risk (see, for example, Bekaert and Harvey (1997), who show that that ICRG explains
cross-sectional variation in equity market volatilities).
We are interested in the effect of differences in country risk and the other country
characteristics on the gains from international diversification. We use as dependent
variable the difference between the Sharpe ratio of the global portfolio (including the
local index) and the Sharpe ratio of the local index. This variable measures the economic
benefits for a mean-variance investor of investing not only in the local index but also in
the global portfolio. A higher figure for this variable indicates that the benefits of foreign
diversification are larger.
Table 5 presents the correlation matrix of the regression variables. We find a high
correlation between the measure of global diversification benefits and our explanatory
variables. Our measures of diversification benefits appear most correlated with the ICRG
rating. For example, the correlation between the ICRG rating and the increase in Sharpe
ratio is about minus 60 percent, under the assumption of short selling constraints in
developing countries only and local currency returns.
We regress the differenced Sharpe ratio variable on a constant, the ICRG country
risk rating, and a number of other country-level regressors. These regressors are the stock
market capitalization in U.S. dollars, trade to GDP, and private credit to GDP. We
include the stock market capitalization variable, because we expect the potential of
diversification to depend on the size of the stock market, and on the range of equities
available to a local investor in particular. We add trade to GDP to control for openness to
trade and private credit to GDP to control for the effect of a country’s banking sector on
diversification benefits. We do not include industry-specific variables, assuming that
16
country index returns are mainly driven by country factors.9 The dependent variables are
calculated using stock market return and interest rate data for the years 1985 to 2002. The
regressors are based on data averaged over the period 1985-2002. The main regression
results are presented in Table 6, and Annex 1 reports the data of the main variables used
in these cross-sectional regressions. We report Huber-White robust estimates of the
standard errors in Table 6. The table only reports the results that assume short selling
constraints in developing countries, as this seems the most realistic assumption. We have
however also checked for robustness of the results by assuming either no frictions or
short selling constraints for all asset markets.
Of all the country-specific variables that we investigate, we find only the ICRG
country risk rating to be a statistically significant explanatory variable for the increase in
the Sharpe ratio variable. In particular, we find that the benefits of foreign diversification
are larger for countries with higher country risk, consistent with the summary statistics in
Table 4. The results do not differ qualitatively when we use U.S. dollar returns rather
than local currency returns to calculate the diversification benefits, or when we assume
no short selling constraints. When we assume short selling constraints on all assets, the
results are also qualitatively similar (not reported).
We also explored other country-level variables as explanatory variables, such as
stock market turnover (proxy for the liquidity of the stock market), inflation rate (proxy
for macroeconomic volatility), the number of listed companies (proxy for the depth of the
stock market), and the law and order tradition in the country (proxy for the quality of the
legal system), but our results do not alter. In all cases, we find that the ICRG country risk
rating has most explanatory power (not reported).
For robustness, we have checked whether the results alter if we exclude the East
Asian countries and/or Russia because the sample period covers both the Asian financial
crisis and the Russian debt crisis of 1997. We find that the results do not qualitatively
alter when these countries are excluded from the sample (not reported).
9
There is a debate going on in the international finance literature whether industrial structure of markets
affects international diversification strategies. For instance, Roll (1992) finds that a large part of the
variation in country index returns can be explained by their industrial composition. On the other hand, more
recent research has found that country index returns are mainly driven by country factors, and only for a
very small part by industry factors (see, for instance, Griffin and Karolyi (1998) and Serra (2000)).
17
As additional robustness test, we exclude countries without complete data for the
sample period 1985 to 2002. Most of these countries are transition countries where stock
markets did not develop until the 1990s. We find that excluding these countries without
full coverage does not qualitatively alter our main results (not reported).
We are also worried that outliers may drive our results. We therefore re-run the
regressions after excluding the 5 percent largest observations and 5 percent smallest
observations of each dependent variable. This outlier rule excludes 5 observations of each
dependent variable, bringing down the sample to 47 or 46 countries depending on the
regression specification. When applying these outlier rules, we obtain similar results.
We have also concerns about the direction of the causality, in particular because
the regressors and the regressand cover the same period. We correct in two ways for this
possible endogeneity problem. First, we find that the results do not alter qualitatively if
we use data at the beginning of the sample period for the regressors, i.e., for the year
1985 (not reported). Second, we use the legal origin of the country as an instrument for
country risk. Legal origin has been used before as instrument for the quality of the
institutional framework of a country (see La Porta et al. 1998). Both the first-stage and
second-stage instrumental variables regression results are reported in Table 7. The firststage results indicate that legal origin is a valid instrument for country risk. For our
sample of countries, country risk is lowest in countries with Scandinavian or German
legal origin. Our main results are not qualitatively altered when using legal origin as
instrument for country risk. We again find that country risk is a major determinant of
global diversification benefits, and that diversification benefits are substantially larger for
countries with higher country risk.
5. Time-Varying Diversification Benefits
In the previous section, we have found evidence that diversification benefits differ across
countries, and that the ICRG country risk measure captures part of this cross-country
variation. Translating these results to a time-series dimension, they imply that, if a
country develops over time and its country risk decreases, the diversification benefits
18
decrease. In this section, we analyze whether this is the case. In this way, this section
extends Section 3 by incorporating time-varying diversification benefits.
To measure the time-varying diversification benefits, we model the expected
returns, variances and correlations for each stock market as a function of the country risk
of the country. As proxy for country risk we again use the ICRG composite risk index
developed by the Political Risk Services Group, because the empirical analysis thus far
has suggested that country risk is a good determinant of the potential global
diversification benefits of a country. For the sake of completeness, we will perform the
analysis of time-varying diversification benefits for both developed and developing
countries. However, we are primarily interested in the results for developing countries,
since for these countries the largest variation in country risk is observed. For developed
countries, other factors besides country risk may be more important to capture timevariation in diversification benefits.
We model expected returns as a linear function of country risk. For the variance
we choose an exponential function to guarantee that the estimated variances are larger
than zero. The correlations are modeled such that they take values smaller or equal to one
in absolute terms. For each country, the correlations are calculated with respect to the
regional or global stock market indices. The exact model is as follows:
E t ( Ri ,t +1 ) = α i + β i ICRGi ,t
(3)
Vt ( Ri ,t +1 ) = exp(γ i + δ i ICRGi ,t )
(4)
ρ t ( Ri ,t +1 , R j ,t +1 ) = 2Φ(ζ ij + η ij ICRGi ,t ) − 1
(5)
where i indicates country i, j indicates regional or global index j, t is the time index (in
months), E(.) is the expected value operator, V(.) is the variance operator, ρ (.) is the
correlation operator, Φ(.) is the cumulative normal distribution function, and ICRG is the
ICRG composite risk index. All operators in equations (3) through (5) are conditional
upon the information at time t, which is summarized by the ICRG measure. We have
19
monthly values for ICRG. For expected returns and variances of regional and global
indices, we use the equally-weighted average ICRG for the specific region as explanatory
variable. Equations (3) and (4) are thus also estimated for the regional and global index
returns.
Equations (3) through (5) are translated into the following regression equations
Ri ,t +1 = α i + β i ICRGi ,t + ε i ,t +1
(6)
ε i2,t +1 = exp(γ i + δ i ICRGi ,t ) + ε i ,t +1
(7)
ε i ,t +1ε j ,t +1
exp((γ i + δ i ICRGi ,t )(γ j + δ j ICRG j ,t ))
= 2Φ(ζ ij + ηij ICRGi ,t ) − 1 + ε~ij ,t +1 (8)
For each country, we estimate equations (6) through (8) step by step using (nonlinear)
regression. Equation (7) and (8) use the error term from the regression results of equation
(6), and in equation (8) the estimates obtained from regression (7) are used. Thus, the
dependent variables in (7) and (8) are replaced by consistent estimates. Given the usual
assumptions on the error terms in regressions (6)-(8), this gives consistent estimates of all
parameters in the model, see Greene (1993, Ch. 14). Standard errors are calculated using
Newey-West (1987), to correct for autocorrelation and heteroskedasticity that is
generated by the step-by-step regression procedure.
Table 8 reports summary statistics on the regression results. The results show that
for the median country, an increase in the ICRG measure (i.e., a decrease in country risk)
causes an increase in the correlation with global indices, and a decrease in variance and
expected return of the local index. Actually, for 42 out of the 52 countries, these patterns
are observed. Since lower country risk is likely to correspond to higher financial
integration, these results are in line with Bekaert and Harvey (1995) and De Jong and De
Roon (2001) who find that most countries have become more financially integrated in
world capital markets. In addition, Bekaert, Harvey, and Lumsdaine (2002) report
20
increases in correlations and decreases in expected returns after liberalization of financial
markets.
The explanatory power of the regression models and the statistical significance of
the slope parameters are not very large, which is not surprising given the large variability
of monthly stock returns and the low-frequency changes in the ICRG measure.10
However, the ICRG measure does seem to pick up economically important swings in
expected returns, variances, and correlations. This follows from Table 9, where we
tabulate the values for the conditional moments in (3) through (5) for two months:
January 1985 and August 2002. The table shows that for most countries the country risk
has decreased over the period 1985 to 2002, which has caused economically important
movements in the first and second conditional moments of the returns. For example, our
regression results imply that for the ‘average’ Asian country, expected monthly U.S.
dollar returns have decreased from 2.2 percent at the beginning of 1985 to 0.9 percent in
2002, while the monthly standard deviation has decreased from 15.2 percent to 9.0
percent. Furthermore, for the ‘average’ Asian country, the correlation with the U.S. index
has increased from 0.107 to 0.262 (similar results are obtained for the correlation with the
other indices). For countries in other regions, qualitatively similar results are found.
Furthermore, the largest changes over the period 1985 to 2002 are found for developing
countries, in line with our interpretation that the ICRG measure captures long-term
swings in financial liberalization and development of these countries.
Next, we turn to the implications of the model in (3) through (5) for the
diversification benefits. Given the results presented above, the effect of the country risk
measure on the diversification benefits is not clear a priori. The increase in correlations
and the decrease in the country’s return variance have a negative effect on the
diversification benefits, while the decrease in the country’s expected return has a positive
effect on the diversification benefits for a local investor. In addition, it turns out that there
is also a (smaller) downward trend in the variance and expected return of the global
indices, for which we also estimated the time-varying behavior in the expected return and
10
The situation of a highly volatile dependent variable and a slowly moving explanatory variable also
shows up in regressions of stock returns on the dividend/price ratio (see Cochrane (2001, Ch. 20) for an
overview). Typically, the R2 in these regression increases if lower-frequency returns are used. For the
21
variance using equations (3) and (4). Therefore, to assess what the net effect is of all
these factors, we calculate implications of the model in (3) through (5) for the
diversification benefits. More precisely, for the first and last month of the sample
(January 1985 versus August 2002) we use the conditional moments in (3) through (5) to
calculate the increase in the Sharpe ratio caused by the opportunity of global
diversification. For the sake of brevity, we focus on the case of frictionless markets. As
reported earlier, short sales restrictions on the local index of the developing countries
only do not have a large impact on these results.
The results are reported in Table 9. For most countries, we find that the
diversification benefits have decreased over this period, so that the effect of the change in
correlations and variance dominates the effect of the change in expected returns. For
example, for the average Asian country the increase in the Sharpe ratio has decreased
from 18.2 percent in January 1985 to 5.9 percent in August 2002. As a final example, we
show in Figure 3 the time-varying diversification benefits for Chile, a country that has
successfully liberalized its financial markets over the sample period (see Gallego and
Loayza (2001)). The official liberalization date for Chile is January 1990. Figure 3 is a
good example of the patterns observed for many other countries. Especially in the
beginning of the period 1985 to 2002, there are large decreases in the country risk and the
diversification benefits. This can be compared to the results in Table 3 of Bekaert,
Harvey, and Lumsdaine (2002), who estimate the liberalization date for Chile, and find
dates that range from February 1983 to January 1993.
5. Conclusions
We have investigated for a sample of 52 countries how the benefits of international
portfolio diversification differ across countries from the perspective of a local investor.
Our main findings are as follows. First, we show that there are substantial regional and
global diversification benefits for domestic investors in both developed and developing
countries, even under the realistic assumption that investors cannot short sell stocks in
expected return equation in (6) we find a similar effect: the R2 increases to around 10% if we use annual
instead of monthly returns.
22
developing countries. Second, we find that the benefits of international portfolio
diversification are larger for developing countries relative to developed countries.
Country risk appears to be a good determinant of diversification benefits, with countries
with higher country risk having a greater potential benefit of global diversification. Third,
we also find that diversification benefits have decreased over our sample period 1985 to
2002. This decrease in diversification benefits corresponds to improvements in country
risk over time.
We add to the existing literature by offering a local rather than a U.S. perspective.
While previous literature has shown that international diversification benefits for U.S.
investors are limited – partly because U.S. investors can largely achieve these gains at
home through investments in stocks of multinational firms (Rowland and Tesar 1998) or
country funds and depositary receipts (Errunza, Hogan, and Hung 1999), but possibly
also because the U.S. has a well-diversified economy and a well-developed stock market
– we show that for investors in many other countries, the benefits from investing abroad
are still substantial.
Our results are consistent with the finding that developing countries on average
are much less integrated in world financial markets (Harvey 1995). We find that meanvariance investors are substantially worse off in less developed countries when not
investing abroad. Unfortunately, investors in developing countries are often restricted to
invest abroad, which shows the importance of a further liberalization of international
financial markets.
23
References
Baxter, Marianne and Urban J. Jermann (1997), “The International Diversification Puzzle
is Worse Than You Think”, American Economic Review 87, 170-180.
Beck, Thorsten, Ross Levine, and Norman Loayza (2000), “Finance and the Sources of
Growth”, Journal of Financial Economics 58, 261-300.
Bekaert, Geert and Campbell Harvey (1995), “Time-Varying World Integration”, Journal
of Finance 50, 403-444.
Bekaert, Geert and Campbell Harvey (1997), “Emerging Equity Market Volatility”,
Journal of Financial Economics 43, 29-77.
Bekaert, Geert, Campbell Harvey, and Robin Lumsdaine (2002), “Dating the Integration
of World Equity Markets”, Journal of Financial Economics, 65(2), 203-249.
Bekaert, Geert and Michael S. Urias (1996), “Diversification, Integration and Emerging
Market Closed-End Funds”, Journal of Finance 51(3), 835-869.
Black, Fischer (1974), “International Capital Market Equilibrium with Investment
Barriers”, Journal of Financial Economics 1, 337-352.
Coval, Joshua D. and Tobias J. Moskowitz (1999), “Home Bias at Home: Local Equity
Preference in Domestic Portfolios”, Journal of Finance 54(6), 2045-2073.
De Jong, Frank and Frans A. De Roon (2001), “Time-Varying Market Integration and
Expected Returns in Emerging Markets”, Working Paper, University of
Amsterdam.
De Roon, Frans A. and Theo E. Nijman (2001), “Testing for Mean-Variance Spanning: A
Survey”, Journal of Empirical Finance 8(2), 111-156.
De Roon, Frans A., Theo E. Nijman, and Bas J.M. Werker (2001), “Testing for MeanVariance Spanning with Short Sales Constraints and Transaction Costs: The Case
of Emerging Markets”, Journal of Finance 56(2), 721-742.
Errunza, Vihang, Ked Hogan, and Mao-Wei Hung (1999), “Can the Gains from
International Diversification be Achieved Without Trading Abroad?”, Journal of
Finance 54(6), 2075-2107.
French, Kenneth R. and James M. Poterba (1991), “Investor Diversification and
International Equity Markets”, American Economic Review 81(2), 222-226.
Gallego, Francisco and Norman Loayza (2001), “Financial Structure in Chile:
Macroeconomic Developments and Microeconomic Effects”, In: Asli Demirgüç-
24
Kunt and Ross Levine, Eds., Financial Structure and Economic Growth,
Cambridge, MA, MIT Press, 299-346.
Greene, William, (1993), Econometric Analysis, 2nd edition, Macmillan, New York.
Griffin, John M. and G. Andrew Karolyi (1998), “Another Look at the Role of the
Industrial Structure of Markets for International Diversification Strategies”,
Journal of Financial Economics 50, 351-373.
Grinblatt, Mark and Matti Keloharju (2001), “Distance, Language, and Culture Bias: The
Role of Investor Sophistication”, Journal of Finance 56 (3), 1053-1073.
Harvey, Campbell R. (1995), “Predictable Risk and Returns in Emerging Markets”,
Review of Financial Studies 8(3), 773-816.
Huberman, Gur (2001), “Familiarity Breeds Investment”, Review of Financial Studies
14(3), 659-680.
Huberman, Gur and Shmuel Kandel (1987), “Mean-Variance Spanning”, Journal of
Finance 42(4), 873-888.
International Monetary Fund (2001), Annual Report on Exchange Restrictions and
Exchange Arrangements, International Monetary Fund, Washington, DC, 2001.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer and Robert W. Vishny
(1998), 1998, Law and finance, Journal of Political Economy 106, 1113-1155.
Lewis, Karin K. (1996), “What Can Explain the Apparent Lack of International
Consumption Risk Sharing?”, Journal of Political Economy 104(2), 267-297.
Li, K. (2004), “Confidence in the Familiar: An International Perspective”, Journal of
Financial and Quantitative Analysis 39, 47-68.
Li, K., A. Sarkar, and Z. Wang (2003), “Diversification Benefits of Emerging Markets
Subject to Portfolio Constraints,” Journal of Empirical Finance, 10 (1-2) 57-80.
Markowitz, Harry (1952), “Portfolio Selection”, Journal of Finance 7(1), 77-91.
Newey, Whitney K. and Kenneth D. West (1987), “A Simple, Positive Semi-Definite,
Heteroskedasticity and Autocorrelation Consistent Covariance Matrix”,
Econometrica 55, 703-708.
Obstfeld, Maurice (1994), “Risk-Taking, Global Diversification, and Growth”, American
Economic Review 84(5), 1310-1329.
25
Roll, Richard (1992), “Industrial Structure and The Comparative Behavior of
International Stock Market Indices”, Journal of Finance, 47(1), 3-41.
Rowland, Patrick F. and Linda L. Tesar (1998), “Multinationals and the Gains from
International Diversification”, NBER Working Paper No. 6733.
Serra, Ana P. (2000), “Country and Industry Factors in Returns: Evidence from
Emerging Markets’ Stocks”, Emerging Markets Review 1(2), 127-151.
Stulz, René M. (1981), “On the Effects of Barriers to International Investment”, Journal
of Finance 36(4), 923-934.
Stulz, René M. and Walter Wasserfallen (1995), “Foreign Equity Investment Restrictions
and Shareholder Wealth Maximization: Theory and Evidence”, Review of
Financial Studies 8(4), 1019-1057.
26
Regional setup
S&P/IFC
Asia
China (1993)
India
Indonesia (1990)
Korea, Rep. Of
Malaysia
Pakistan
Philippines
Sri Lanka (1993)
Taiwan
Thailand
S&P/IFC
Latin America
Argentina
Brazil
Chile
Colombia
Mexico
Peru (1993)
Venezuela
Czech Rep. (1994)
Hungary (1993)
Poland (1993)
Russia (1996)
Slovakia (1996)
Turkey (1987)
S&P/IFC
Eastern Europe
27
Egypt (1996)
Jordan
Morocco (1996)
Nigeria
South Africa (1993)
Zimbabwe
Austria
Belgium
Denmark
Finland
France
Germany
Greece
Ireland
Italy
Netherlands
Norway
Portugal (1986)
Spain
Sweden
Switzerland
UK
S&P/IFC
MSCI
Middle East & Africa Europe 16
Australia
Hong Kong
Japan
New Zealand
Singapore
MSCI
Pacific
Canada
US
MSCI
North America
The regional division for the developing countries follows the definitions of S&P/IFC. For the developed countries we use regional MSCI indices. For the U.S. and Canada we use
an equally-weighted index of the two countries. Sample period is 1985 to 2002, unless otherwise noted. For countries with a different sample period, we report the first year of the
sample period between brackets (in all cases the last year of the sample is 2002).
Table 1
Increase in Sharpe Ratio and Expected Return by Region
0.039
0.038
0.038
0.133
0.030
0.030
0.042
0.021
0.014
0.223
0.000
0.000
0.182
0.181
0.141
1.116
1.037
0.274
0.129
0.127
0.089
1.476
1.442
0.274
SR: no frictions
SR: frictions, developing
SR: frictions, all
ER: no frictions
ER: frictions, developing
ER: frictions, all
SR: no frictions
SR: frictions, developing
SR: frictions, all
ER: no frictions
ER: frictions, developing
ER: frictions, all
SR: no frictions
SR: frictions, developing
SR: frictions, all
ER: no frictions
ER: frictions, developing
ER: frictions, all
SR: no frictions
SR: frictions, developing
SR: frictions, all
ER: no frictions
ER: frictions, developing
ER: frictions, all
0.129
0.126
0.106
1.190
1.178
0.204
0.137
0.137
0.115
0.981
0.852
0.204
0.019
0.003
0.003
0.000
0.000
0.000
0.014
0.010
0.010
0.000
0.000
0.000
0.129
0.122
0.071
1.855
1.840
0.366
28
0.137
0.134
0.091
0.991
0.964
0.271
0.220
0.139
0.217
0.135
0.195
0.107
1.703
0.749
1.623
0.715
0.366
0.262
Global Diversification (U.S. dollars)
0.096
0.082
0.069
0.005
0.041
0.001
0.291
0.099
0.291
0.000
0.291
0.000
Global Diversification (Local Currency)
0.095
0.047
0.084
0.025
0.070
0.017
0.224
0.089
0.224
0.000
0.224
0.000
Regional Diversification (U.S. dollars)
0.070
0.070
0.054
0.423
0.423
0.076
0.057
0.057
0.037
0.340
0.340
0.076
0.036
0.036
0.034
0.092
0.092
0.000
0.035
0.035
0.033
0.084
0.084
0.000
0.124
0.122
0.105
0.780
0.765
0.307
0.101
0.100
0.087
0.609
0.503
0.307
0.017
0.003
0.003
0.027
0.000
0.000
0.015
0.001
0.001
0.017
0.000
0.000
0.086
0.086
0.059
0.423
0.423
0.226
0.079
0.079
0.059
0.349
0.349
0.225
0.057
0.057
0.023
0.259
0.259
0.000
0.050
0.050
0.025
0.198
0.198
0.000
Using the regional setup in Table 1, this table reports the regional average increase in Sharpe Ratio (SR) and increase in expected return (ER) given variance (in percentages) when
a regional index or three global MSCI indices are added to a local stock index portfolio. The Sharpe ratios are calculated for each country both with and without short-selling
constraints and in terms of local currency and U.S. dollar returns. Short-selling constraints are either applied to all countries or only to developing countries.
Asia
Eastern Europe
Middle East & Africa
Europe 16
Pacific
North America
Latin America
Regional Diversification (Local Currency)
Table 2
Rejection of Spanning Tests by Region
7
100%
29%
29%
7
100%
14%
0%
7
100%
100%
100%
7
100%
100%
57%
Total number of countries
No frictions
Frictions, developing
Frictions, all
Total number of countries
No frictions
Frictions, developing
Frictions, all
Total number of countries
No frictions
Frictions, developing
Frictions, all
Total number of countries
No frictions
Frictions, developing
Frictions, all
S&P Global
Latin America
10
100%
100%
90%
10
100%
100%
90%
10
100%
0%
0%
10
100%
0%
0%
S&P Global
Asia
16
100%
100%
31%
16
100%
100%
63%
Global Diversification (Local Currency)
6
6
100%
100%
100%
100%
100%
83%
Global Diversification (U.S. $)
6
6
100%
100%
100%
100%
33%
67%
29
16
100%
100%
19%
Regional Diversification (U.S. $)
6
6
83%
100%
50%
0%
17%
0%
MSCI
Europe 16
16
100%
100%
19%
S&P Global
Middle East & Africa
Regional Diversification (Local Currency)
6
6
83%
100%
50%
50%
33%
17%
S&P Global
Eastern Europe
5
100%
100%
80%
5
100%
100%
80%
5
100%
0%
0%
5
100%
0%
0%
MSCI
Pacific
2
100%
100%
50%
2
100%
100%
50%
2
50%
50%
50%
2
50%
50%
50%
MSCI
North America
We test for each country whether the local stock index (mean-variance) spans a regional index (asset set A) or three global MSCI indices (asset set B). Using the regional setup in
Table 1, the table reports the percentage of countries per region for which this test leads to a rejection at the 5 percent level, both for the case with and without short-selling
constraints and for local currency and U.S. dollar returns. Short-selling constraints are either applied to all countries or only to developing countries.
Table 3
T-test of mean differences
Variable
Increase SR: No frictions, local currency
Increase SR: Market frictions for developing, local currency
Increase SR: Market frictions for all countries, local currency
Increase SR: No frictions, U.S. dollars
Increase SR: Market frictions for developing, U.S. dollars
Increase SR: Market frictions for all countries, U.S. dollars
GDP per capita (US$)
Market cap (US$ bn)
Trade to GDP (%)
Private credit to GDP (in %)
Country risk rating
30
Developed
Developing
t-test of equal means
0.122
0.066
0.167
***-5.730
0.121
0.066
0.165
***-5.683
0.098
0.049
0.137
***-4.983
0.110
0.078
0.135
***-3.993
0.108
0.078
0.131
***-3.858
0.080
0.060
0.096
***-3.001
12,438.56
24,259.83
3,063.07
***11.55
372.40
771.99
55.48
*1.95
72.63
85.06
62.77
1.30
64.64
91.14
42.87
***5.08
72.05
82.40
63.84
***10.54
All countries
Figures are averages for the period 1985-2002, when available. We use the year 1995 classification for developing countries of the World Bank. The column “Developing” reports
averages across all developing countries in the sample. The column “Developed” reports averages across all developed countries in the sample. The column “t-test of equal means”
reports the t-statistic of the test of mean differences in the variables in the columns “Developing” and “Developed” assuming unequal variances. “Increase SR” is the difference
between the Sharpe ratio of the local index plus the global index portfolio and the Sharpe ratio of the local index. GDP per capita is calculated in U.S. dollars. Market cap is stock
market capitalization is in billions of U.S. dollar. Trade to GDP is the sum of exports and imports divided by GDP. Private credit to GDP is credit to the private sector divided by
GDP. Country risk rating is the ICRG country risk index. The number of developed countries in the sample is 23 and the number of developing countries is 29. The total number of
countries is 52. Heteroskedasticity consistent t-statistics between brackets. * indicates significance at the 10% level; ** indicates significance at the 5% level; * indicates
significance at the 1% level.
Table 4
Correlation matrix
Measures of diversification benefits
Increase SR: Increase SR: Increase SR: Increase SR:
LC
Frictions, LC
$
Frictions, $
*-0.232
-0.225
-0.198
-0.180
-0.186
-0.116
-0.073
0.049
***-0.536
***-0.487
**-0.307
-0.184
***-0.645
***-0.601
***-0.406
**-0.277
Country characteristics and diversification benefits
Market capitalization
Trade to GDP
Private credit to GDP
Country risk rating
Panel B
Increase SR: Increase SR: Increase SR: Increase SR:
LC
Frictions, LC
$
Frictions, $
Increase SR: LC
1.000
Increase SR: Frictions, LC ***0.939
1.000
Increase SR: $
***0.574
***0.358
1.000
Increase SR: Frictions, $
***0.467
***0.404
***0.777
1.000
Panel A
31
Figures are based on data for the period 1985-2002, or when available. Increase SR indicates increase in Sharpe Ratio (SR). LC indicates in local currency terms and $ indicates in
U.S. dollars terms. Frictions indicates market frictions for developing countries only. Country risk rating is the ICRG country risk index. Private credit to GDP is credit to the
private sector divided by GDP. Trade to GDP is the sum of exports and imports divided by GDP. Market cap is stock market capitalization is in trillions of U.S. dollar.
Correlations are based on 52 country-observations. * indicates significance at a 10% level; ** indicates significance at a 5% level; *** indicates significance at a 1% level.
Table 5
Cross-country differences in the benefits of foreign diversification
—
—
—
0.42
52
Trade to GDP
Market cap
R-squared
Number of observations
-0.468***
(0.100)
Private credit to GDP
Country risk rating
0.43
51
-0.124
(0.333)
-0.001
(1.446)
-3.007
(2.578)
-0.390***
(0.115)
Panel A: Local currency returns
Increase in
Increase in
Sharpe ratios
Sharpe ratios
0.16
52
—
—
—
-0.196***
(0.060)
32
0.17
51
-0.392
(0.333)
0.153
(1.427)
0.145
(2.865)
-0.191**
(0.083)
Panel B: U.S. dollar returns
Increase in
Increase in
Sharpe ratios
Sharpe ratios
Dependent variable is the increase in Sharpe ratio when adding the world portfolio (consisting of US, Europe and Far East) to the local stock market index under the assumption of
market frictions in developing countries only. Data on the Sharpe ratios and expected returns for 52 countries for the period 1985 to 2002, when available. Panel A has stock
returns in local currency. Panel B has stock returns in U.S. dollars. Independent variables are averages for the period 1985-2001, when available, unless otherwise noted. Country
risk rating is the ICRG index. Private credit to GDP is credit to the private sector divided by GDP. Trade to GDP is the sum of exports and imports divided by GDP. Market cap is
stock market capitalization is in trillions of U.S. dollar. A constant was added, but is not reported. Huber-White robust standard errors between brackets. * indicates significance at
a 10% level; ** indicates significance at a 5% level; *** indicates significance at a 1% level.
Table 6
Cross-country differences in the benefits of foreign diversification: Instrumental variables
2.074
(4.678)
-1.470
(4.548)
15.827***
(5.641)
16.266***
(6.307)
0.32
52
French legal origin
German legal origin
Scandinavian legal origin
R-squared
Number of observations
Country risk rating
English legal origin
First stage
R-squared
Number of observations
Country risk rating
Second stage
33
0.42
52
-0.466***
(0.096)
Panel A: Local currency returns
Increase in Sharpe ratios
0.16
52
-0.169*
(0.091)
Panel B: U.S. dollar returns
Increase in Sharpe ratios
Dependent variable is the increase in Sharpe ratio or the increase in expected return when adding the world portfolio (consisting of US, Europe and Far East) to the local stock
market index under the assumption of market frictions in developing countries only. Data on the Sharpe ratios and expected returns for 52 countries for the period 1985 to 2002,
when available. Panel A has stock returns in local currency. Panel B has stock returns in U.S. dollars. Independent variables are averages for the period 1985 to 2001, when
available, unless otherwise noted. Country risk rating is the ICRG index. We use legal origin as instrument for the ICRG index. GDP per capita is the GDP per capita is in tens of
thousands of U.S. dollar. Stock market capitalization is in trillions of U.S. dollar. Number of listed companies is in thousands of companies. Turnover ratio is the total stock value
traded divided by the stock market capitalization. For a glossary of the variables, see Table 1. A constant was added, but is not reported. Standard errors between brackets. *
indicates significance at a 10% level; ** indicates significance at a 5% level; *** indicates significance at a 1% level.
Table 7
ICRG Regression Results
Local returns:
Median slope estimate
Local returns:
Median t-ratio
Local returns:
Median R2
U.S. dollar returns:
Median slope estimate
U.S. dollar returns:
Median t-ratio
U.S. dollar returns:
Median R2
-0.07
1.02
1.87%
-0.09
1.53
2.18%
1.23
0.93%
0.12
1.38
1.28%
ICRG:
Variance Equation
0.09
ICRG:
Return Equation
34
4.57%
1.72
0.05
3.61%
1.48
0.03
ICRG:
Correlation U.S. Equation
2.98%
1.59
0.03
2.52%
1.27
0.02
3.29%
1.39
0.02
2.19%
1.52
0.01
ICRG:
ICRG:
Correlation Europe Equation Correlation Far-East Equation
This table reports the results of the regressions in equations (6) through (8). For each country, five regression equations are estimated, for respectively the expected return of the
local index, the return standard deviation, and the correlations with the three global indices. In all regressions, the explanatory variable is the ICRG country risk rating of the
country of interest. Data period is January 1985 until August 2002. Results are provided both for local currency returns and U.S. dollar returns. Returns are measured in monthly
percentages. Standard errors are calculated using Newey-West (1987), correcting for fourth-order autocorrelation and heteroskedasticity.
Table 8
Time-varying diversification benefits
(85) St.Dev.
0.096
0.108
0.175
0.062
0.076
0.087
0.058
(02) Corr US (85) Corr US (02)
0.080
0.192
0.244
0.092
0.054
0.155
0.115
0.201
0.200
0.067
-0.139
0.193
0.065
0.279
0.483
0.056
0.221
0.463
0.055
0.570
0.764
35
Region
Change ICRG (85-02) Increase SR (85) Increase SR (02) Exp return (85) Exp return (02) Variance (85) Variance (02) Corr US (85) Corr US (02)
S&P/IFC Latin America
6.014
0.093
0.056
0.022
0.009
0.152
0.090
0.107
0.262
S&P/IFC Asia
10.090
0.182
0.059
0.019
0.000
0.108
0.102
0.060
0.282
S&P/IFC Eastern Europe
2.100
0.161
0.164
0.004
0.007
0.174
0.119
0.170
0.246
S&P/IFC Middle East & Africa
5.500
0.124
0.101
0.019
0.012
0.080
0.074
-0.055
0.244
MSCI Europe 16
3.763
0.089
0.067
0.012
0.009
0.079
0.066
0.310
0.457
MSCI Pacific
1.460
0.262
0.144
0.011
0.001
0.094
0.067
0.264
0.606
MSCI North America
-2.700
0.121
0.085
0.019
0.001
0.060
0.060
0.607
0.815
Panel B: Results by region in US dollars
Region
Change ICRG (85-02) Increase SR (85) Increase SR (02) Exp return (85) Exp return (02) St.Dev.
S&P/IFC Latin America
6.014
0.224
0.175
0.036
0.012
S&P/IFC Asia
10.090
0.175
0.073
0.020
0.003
S&P/IFC Eastern Europe
2.100
0.246
0.289
0.011
0.013
S&P/IFC Middle East & Africa
5.500
0.134
0.110
0.025
0.019
MSCI Europe 16
3.763
0.071
0.053
0.010
0.009
MSCI Pacific
1.460
0.255
0.194
0.012
0.002
MSCI North America
-2.700
0.114
0.086
0.019
0.002
Panel A: Results by region in local currency
This table reports the diversification benefits based on our ICRG-based time-varying model for two dates at the beginning and end of our sample period: January
1985 and August 2002. 85 indicates January 1985. 02 indicates August 2002. Change ICRG is the change in the ICRG rating over the period. Increase SR in the
estimated increase in Sharpe ratio at a specific date. Exp return is the expected stock return at a specific date. Variance is the variance of the stock return at a
specific date. Corr US is the correlation with the return on the U.S. stock market index at a specific date. Panel A summarizes the country-level local currency
results by region. Panel B summarizes the country-level U.S. dollar results by region.
Table 9
Increase in Sharpe ratio
Global diversification in U.S. dollars: Increase in Sharpe ratios
0.00
0.05
0.10
0.15
0.20
0.25
0.30
Without frictions
Short selling constraints
Global Diversification Benefits
Increase in Sharpe Ratio as Measured in US Dollar Returns
This figure shows the difference between the Sharpe ratio of the portfolio that includes the local stock market index and the world portfolio (consisting of the local portfolio plus
the stock market indices of US, Europe and Asia-Pacific as measured by MSCI) and the Sharpe ratio of the local stock market index for each country in our sample. The Sharpe
ratios are estimated using U.S. dollar returns for the period 1985 to 2002. Results are presented for frictionless markets and for markets with short-sales restrictions on all assets.
Figure 1
36
Brazil
Argentina
New Zealand
Portugal
Slovakia
Morocco
Egypt
Czech Republic
Indonesia
Sri Lanka
Malaysia
Japan
Russia
Ireland
China
Jordan
Singapore
Canada
South Africa
India
Peru
Pakistan
Venezuela
Thailand
Finland
Nigeria
Korea
Hungary
Norway
Taiwan
Poland
Philippines
Australia
Austria
Zimbabwe
Colombia
Greece
Italy
Turkey
Germany
Hong Kong
Mexico
Denmark
US
UK
Sweden
Spain
France
Chile
Belgium
Switzerland
Netherlands
Increase in Expected Returns
Global diversification in U.S. dollars: Increase in expected returns
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
4.0%
4.5%
5.0%
Without frictions
Short selling constraints
Global Diversification Benefits
Increase in monthly expected return given fixed variance (measured in US dollars)
This figure shows the difference between the Expected Return of the portfolio that includes the local stock market index and the world portfolio (consisting of the local portfolio
plus the stock market indices of US, Europe and Asia-Pacific as measured by MSCI) and the Expected Return of the local stock market index given variance for each country in
our sample. The expected returns are estimated using U.S. dollar returns for the period 1985 to 2002. Results are presented for frictionless markets and for markets with short-sales
restrictions on all assets.
Figure 2
37
Argentina
Brazil
Indonesia
Czech Republic
Egypt
Slovakia
Russia
China
Sri Lanka
Portugal
New Zealand
Poland
Turkey
Hungary
Malaysia
Finland
South Africa
Peru
Japan
Thailand
Venezuela
Singapore
Ireland
Korea
India
Pakistan
Morocco
Nigeria
Canada
Taiwan
Jordan
Norway
Philippines
Australia
Austria
Zimbabwe
Italy
Greece
Germany
Mexico
Colombia
Hong Kong
UK
Denmark
Sweden
Spain
US
France
Chile
Belgium
Switzerland
Netherlands
Example of time-varying diversification benefits: the case of Chile
0
1985
0.01
0.02
0.03
0.04
0.05
0.06
1990
1995
1985-2002
2000
38
Time Varying Diversification Benefits: Chile
2005
This figure shows the time-varying diversification benefits for Chile, as measured by the increase in the Sharpe ratio when allowing for global diversification. For each month in
the 1985 to 2002 sample, this increase in the Sharpe ratio is calculated using the time-varying model for the expected return, variance, and correlations. The graph present the
results for the case of U.S. dollar returns and frictionless markets.
Figure 3
Increase in Sharpe Ratio
Data of the Variables in the Cross-Sectional Regressions
0.28
0.06
0.05
0.01
0.21
0.11
0.08
0.12
0.24
0.16
0.03
0.21
0.10
0.02
0.05
0.10
0.05
0.16
0.18
0.17
0.11
0.05
0.10
Argentina
Australia
Austria
Belgium
Brazil
Canada
Chile
China
Colombia
Czech Republic
Denmark
Egypt
Finland
France
Germany
Greece
Hong Kong
Hungary
India
Indonesia
Ireland
Italy
Japan
0.15
0.14
0.07
0.13
0.17
0.06
0.10
0.07
0.08
0.12
0.03
0.05
0.17
0.08
0.18
0.03
0.13
0.26
0.13
0.08
0.03
0.08
0.22
0.6
0.6
0.3
0.4
1.7
0.3
1.1
0.3
0.3
0.9
0.1
0.2
2.1
0.3
1.8
0.1
1.6
2.0
0.5
0.4
0.1
0.3
2.8
39
0.9
0.8
0.4
0.8
3.0
0.3
1.1
0.4
0.4
1.1
0.2
0.2
2.3
0.3
2.5
0.2
1.8
3.3
0.7
0.4
0.1
0.4
4.1
86.8
82.2
78.4
59.3
58.4
76.4
70.5
85.6
68.3
84.3
80.4
84.9
60.2
61.1
73.7
69.4
69.0
61.5
83.9
85.6
81.6
80.3
61.7
188.7
64.1
59.8
26.5
38.0
155.9
36.0
99.0
38.0
69.1
87.6
41.6
40.0
28.8
54.7
57.9
92.5
46.2
76.9
92.8
57.5
64.8
20.8
18.6
125.9
43.6
21.4
55.4
262.8
80.0
53.6
45.6
58.8
44.8
67.5
48.2
34.2
112.6
59.6
36.9
18.3
63.2
78.0
140.1
36.9
18.3
332.5
5.8
32.9
9.7
3.1
31.7
0.7
68.5
4.5
9.6
72.0
6.4
1.2
1.0
1.2
4.6
19.4
13.5
43.9
2.5
23.7
24.7
5.4
Increase SR is the increase in Sharpe Ratio (SR) and Increase ER is the increase in expected return (ER) given variance (in percentages) when three global MSCI indices are added
to a local stock index portfolio. The increases in SR and ER are calculated for each country with short-selling constraints for developing markets only and in terms of local
currency and U.S. dollar returns. Country risk rating is the ICRG country risk index. Private credit to GDP is credit to the private sector divided by GDP. Trade to GDP is the sum
of exports and imports divided by GDP. Market cap is stock market capitalization is in billions of U.S. dollar. Figures are averages for the period 1985-2002, when available. Data
on stock market capitalization of listed companies, private credit to GDP, and trade to GDP come from the World Development Indicators of the World Bank. We do not have data
on private credit to GDP for Taiwan. Data on the ICRG country risk ratings are from the International Country Risk Guide (ICRG) maintained by the Political Risk Services group.
The ICRG rating ranges from 0 to 100, with a higher rating indicating less country risk.
Increase SR:
Increase SR:
Increase ER:
Increase ER:
Private credit to
Local currency
U.S. dollars
Local currency
U.S. dollars
GDP
Country
ICRG
Trade to GDP
Market cap
Annex 1
0.13
0.11
0.12
0.10
0.14
0.14
0.14
0.02
0.18
0.03
0.08
0.21
0.17
0.09
0.20
0.15
0.16
0.11
0.21
0.11
0.03
0.20
0.04
0.01
0.08
0.11
0.46
0.02
0.05
0.16
0.20
Malaysia
Mexico
Morocco
Netherlands
New Zealand
Nigeria
Norway
Pakistan
Peru
Philippines
Poland
Portugal
Russia
Singapore
Slovakia
South Africa
Spain
Sri Lanka
Sweden
Switzerland
Taiwan
Thailand
Turkey
United Kingdom
United States
Venezuela
Zimbabwe
0.08
0.05
0.12
0.07
0.05
0.10
0.12
0.04
0.02
0.04
0.17
0.16
0.13
0.12
0.13
0.20
0.09
0.13
0.08
0.10
0.12
0.21
0.11
0.19
0.02
0.15
0.05
Increase SR:
U.S. dollars
Increase SR:
Local currency
Country
Jordan
Korea
0.4
0.2
0.7
1.1
0.1
0.6
0.8
0.2
0.0
0.2
1.0
2.1
0.7
2.3
0.8
1.4
1.2
1.0
0.4
0.5
0.7
1.1
0.1
0.5
0.0
0.5
0.3
0.5
0.7
Increase ER:
Local currency
40
0.4
0.2
0.8
1.1
0.2
0.7
0.9
0.2
0.1
0.2
1.6
2.4
1.0
2.4
0.8
1.4
1.5
1.0
0.5
0.6
0.8
1.5
0.7
0.7
0.1
1.1
0.3
0.6
0.8
Increase ER:
U.S. dollars
54.2
83.5
65.2
54.6
82.3
83.5
70.1
83.6
90.6
76.8
53.1
72.3
65.8
61.9
84.7
65.9
78.4
54.2
58.7
87.8
51.9
81.5
52.7
62.7
87.4
73.6
66.8
ICRG
62.8
76.2
26.0
108.2
25.7
19.7
110.2
n.a.
103.9
102.4
162.1
79.9
21.5
32.3
100.4
11.2
104.8
11.9
77.6
17.7
34.2
77.0
28.4
83.1
12.2
37.9
94.9
112.8
19.5
71.7
72.0
Private credit to
GDP
61.1
21.2
48.7
41.6
53.1
93.1
83.6
68.3
71.2
43.3
73.2
111.3
47.1
56.2
349.7
51.5
68.0
30.9
76.6
72.3
37.0
57.8
68.0
58.1
112.1
165.2
45.8
Trade to GDP
119.0
67.6
0.2
814.0
0.7
3.5
156.4
21.8
6.0
18.9
39.3
25.1
0.2
0.1
18.2
3.5
10.2
1.3
2.9
0.8
3.4
4.2
0.7
2.2
0.2
0.6
43.7
12.8
10.6
Market cap
0.4
16.0