Claude B. Erb

First Chicago Investment Management Co. Chicago, IL

Campbell R. Harvey

Duke University, Durham, NC 27708

National Bureau of Economic Research, Cambridge, MA

Tadas E. Viskanta

First Chicago Investment Management Co. Chicago, IL

This paper is an earlier draft of the an article published in the November/December 1996 Financial Analysts Journal

ABSTRACT

How important is an understanding of country risk for investors? Given the increasingly global nature of investment portfolios, we believe it is very important. Our paper measures the economic content of five different measures of country risk: The International Country Risk Guide’s political risk, the financial risk, economic risk and composite risk indices and Institutional Investor’s country credit ratings. First, we explore whether any of these measures contain information about future expected stock returns by conducting trading simulations. Next, we conduct time-series-cross-sectional analysis linking these risk measures to future expected returns. Second, we investigate the relation between these measures and other, more standard, approaches to risk exposures. Finally, we analyze the linkages between fundamental attributes within each economy, such as book-to-price ratios, and the risk measures. Our results suggest that the country risk measures are correlated future equity returns. We find that the country risk measures are correlated with each other, however, financial risk measures contain the most information about future equity returns. Finally, we find that country risk measures are highly correlated with country equity valuation measures. This provides some insight into the reason for higher returns for value-oriented strategies.

1. Introduction

What is country risk and how should it impact global investment strategies? We explore the information in five different measures of country risk. The first four measures are from Political Risk Services’ International Country Risk Guide (ICRG). These measures include political risk, economic risk and financial risk. The ICRG also reports a measure of composite risk which is a simple function of the three base indices. The final measure we examine is Institutional Investor’s country credit ratings. We define the information content in a number of different ways.

We initially explore whether the risk indices contain information about future expected returns. This analysis is conducted in two ways. First, we form a portfolio of countries which experience a decrease in the risk rating (become more risky) and a portfolio of countries which experience an increase in the risk rating (become less risky). We form the portfolio after the risk information is available and rebalance the portfolio every six months. We find that there is, indeed, information about expected equity returns in these measures. We supplement this analysis with a time-series cross-sectional regressions which measure the amount of information contained in each metric. We find that the financial risk measure contains the most information about future expected returns and political risk contains the least.

The next part of our analysis investigates the link between these country risk measures and some more standard measures of risk. We investigate whether there is a correlation between a country’s beta against the MSCI world index. While this is a standard risk measure for integrated capital markets, many have found the world beta model inadequate to characterize risk in emerging markets. As an alternative, we also investigate the relation between the country risk measure and equity volatility.

The final part of our paper explores the interface between country risk analysis and investment strategies based on country fundamental information, such as book-to-price ratios. We find that the risk indices are highly correlated with the fundamental attributes. This provides some economic insight as to why value-oriented strategies earn higher returns — they reflect higher risk exposure.

To ensure the widest possible dissemination of our methodology, we have established a country risk homepage:

This site includes information on 135 different countries which could not be included in this manuscript. For example, the site contains equity return histograms for 48 countries, time-series graphs of the five risk measures for 117 countries, summary statistics, as well as the most recent version of this paper and the associated tables and exhibits.

2. Measuring Country Risk

There are many services that measure country risk. The appendix provides information on the following providers:

• Bank of America World Information Services
• Business Environment Risk Intelligence (BERI) S.A.
• Control Risks Information Services (CRIS)
• Economist Intelligence Unit (EIU)
• Euromoney
• Institutional Investor
• Standard and Poor’s Rating Group
• Political Risk Services: International Country Risk Guide (ICRG)
• Political Risk Services: Coplin-O’Leary Rating System
• Moody’s Investor Services

Each of the index or rating providers must amalgamate a range of qualitative and quantitative information into a single index or rating. In this section, we review in detail the methodologies used by two of the foremost providers of risk ratings: Institutional Investor and International Country Risk Guide (ICRG).

2.1 Institutional Investor

Institutional Investor credit ratings are based on a survey of leading international bankers who are asked to rate each country on a scale from zero to 100 (where 100 represents maximum creditworthiness). Institutional Investor averages these ratings, providing greater weights to respondents with greater worldwide exposure and more sophisticated country analysis systems.

Whenever a survey or expert panel is used to subjectively rate creditworthiness, it is hard to exactly define the parameters taken into account. At any given point in time an expert’s recommendation will be based upon factors the expert feels are relevant.

In order to identify the factors that its survey participants have taken into consideration in the past, Institutional Investor asks them to rank the factors that they take into account in preparing country ratings. The results of this survey are listed in Table 1. Note that the bankers rank factors differently for different groups of countries and that rankings have changed over time within country groups. The ranking of factors affecting OECD country ratings appear to have been the most turbulent over the fifteen-year period.

2.2 International Country Risk Guide

ICRG compiles monthly data on a variety of political, financial and economic risk factors to calculate risk indices in each of these categories as well as a composite risk index. Five financial, thirteen political and six economic factors are used. Each factor is assigned a numerical rating within a specified range. The specified allowable range for each factor reflects the weight attributed to that factor. A higher score indicates lesser risk.

Political risk assessment scores are based on subjective staff analysis of available information. Economic risk assessment scores are based upon objective analysis of quantitative data and financial risk assessment scores are based upon analysis of a mix of quantitative and qualitative information.

Calculation of the three individual indices is simply a matter of summing up the point scores for each factor within each risk category. The composite rating is a linear combination of the three individual indices’ point scores. Note that the political risk measure (100 points) is given twice the weight of financial and economic risk (50 points each). ICRG, as well as many of the other providers, think of country risk as being composed of two primary components: ability to pay and willingness to pay. Political risk is associated with a willingness to pay while financial and economic risk are associated with an ability to pay.

The specific formulas for these calculations are as follows: where PR is political risk, ER is economic risk, FR is financial risk and CRR is the composite risk rating. The specific factors taken into account for each risk index are detailed in Table 2.

ICRG also groups country composite scores into ordinal risk categories to facilitate quick interpretation and comparison of country scores. This categorization scheme is presented in Table 3.

2.3 Index and Rating Provider Comparison

A wide range of groups provide country risk or country credit ratings. Although the factors taken into account by each group and the audience they seek to inform vary, there are significant similarities across the providers of these measures.

Most of the providers transform widely used quantitative economic indicators in roughly the same manner. The important differences are found in the degree of and specific factors included in the qualitative component of the risk index measures. Tables 4 and 5 contrast the primary risk index products available from the indicated groups.

Table 4 identifies the underlying analytical source (i.e. quantitative or qualitative information) for each major index subcomponent. Additionally, the index measure is categorized as ordinal or scalar and the data sources are classified. From this table it can be seen that most firms use a mix of qualitative and quantitative analysis. The extremes are represented by Bank of America World Information Services which is wholly based on quantitative information, and Institutional Investor which is wholly based on a survey of banking professionals. Note that surveys and staff analysis which strive for an overall recommendation are categorized as qualitative analysis on the basis that this form of recommendation takes many non-quantitative factors into account. At the same time, the analysts and experts certainly do take relevant quantitative factors into account, just not in a formulaic manner as is true for a purely quantitative index.

Table 5 looks in more detail at the specific factors comprising each risk index. Moderately broad categories have been created to classify the many specific factors used by each group. This chart helps to identify the mix of quantitative and qualitative factors used by each group as well as the specific similarities and differences in the composition of the indices.

Finally, Table 6 provides a comparison of S&P and Moody’s Ratings with both the II and ICRG ratings. The table reports the ratings in October 1995. There is a close correspondence between the S&P and Moody’s ratings and the II credit risk measure with a rank order correlation of 95%. There is also a strong correlation between these ratings and the ICRG financial rating (rank order correlation of 90%). The correlations are weaker for the other measures. For example, the rank order correlation of the Moody’s rating and the ICRG economic rating is only 68%. This is also seen in Figure 1 which graphs the five different country risk measures for each of the Moody’s and S&P category. There is a sharp correlation evident with the II credit risk measure. While the other risk measures exhibit similar patterns, the relation is not as sharp.

3. Summary Data Analysis

3.1 Variation in risk measures

Our analysis focuses on 117 countries. For these countries, we have all five risk indices. We segment the countries into five groups: All countries, countries with equity markets, developed countries with equity markets, emerging countries with equity markets and countries without equity markets.

Time-series graphs of the risk indices for each country are available through the country risk web site. /../Country_Risk/couindex.htm. Figure 2 presents time-series graphs of the equally weighted risk indices for three groups over the January 1984-July 1995 period: Developed (with equity markets), emerging (with equity markets) and all other countries.

The equally-weighted measures for the developed countries (Panel A) exhibit remarkably little variation through time. The ICRG financial and economic measure remain about the same throughout the sample, as does the II country credit rating. There is a small decline the ICRG political rating.

The analysis for the emerging countries and all other countries (Panels B and C) is similar. Many of the other countries have similar economic characteristics to the emerging country sample. Generally, all of the risk rating measures increase over the sample. This is particularly evident over the 1988 to 1993 period.

The cross-sectional behavior of the risk measures is explored in Figure 3. For the countries with equity returns and all countries (with and without equity markets), we graph the January 1984 risk level against the change in the risk level up to July 1995. The panels analyze all five measures: II credit risk, ICRG political, economic, financial, and composite risk. There appears to be cross-sectional mean reversion in the risk measures. Those countries that began with a very low risk rating tend to improve. Those countries with a high rating have tended to deteriorate.

This mean reversion is particularly evident for the financial and composite risk measures for countries with equity markets. There is also evidence that the political and economic risk exhibit the same type of behavior. The weakest evidence of cross-sectional mean reversion is found for the credit risk.

Table 7 presents some summary statistics on the risk measures for the subset of countries with equity markets. Panel A presents the initial risk measure and the most current rating along with the average equity return, the standard deviation and the beta against the Morgan Stanley Capital International world market portfolio. The equity returns are from MSCI and the International Finance Corporation (IFC).

There are a number of countries which show sharp improvement in the risk ratings. For example, Argentina begins with a ICRG composite rating of 40.0 and rises to 71.5 by July 1995. Similar improvements are evident for many of the other Latin American countries. In contrast, some developed markets have shown deterioration, for example, Switzerland dropped from an ICRG composite rating of 95.0 to 88.5 in July 1995. Similar declines were found for Japan, the U.K. and the U.S.

The second panel details the correlation of the various risk measures. Since the credit rating is only available on a semi-annual basis the correlations are presented based on semi-annual observations. The correlation measure is equally weighted across the different countries. The upper triangle of the matrix reports the correlation based on changes in rating and the lower triangle reports the correlation of the levels.

The correlations are not as high as one might expect. Obviously, the correlation between the composite and the political rating is the highest because, by definition, the political rating is 50% of the composite. However, the highest cross-correlation of the levels of the three ICRG components is 35%. Interestingly, the correlation between the ICRG financial measure and the II credit rating is only 26%.

The third panel documents the degree of predictability in the risk measures. The levels of the measures are very persistent. We report the average autocorrelations of the changes in the risk measures. The change in the II credit rating is the most predictable. The average first order autocorrelation is 24% (20% in developed markets and 26% in emerging markets). The least predictable measure is the change in the political risk rating. The average autocorrelation is -0.01 (0.03 in developed markets and -0.04 in emerging markets).

The final part of Table 7 provides a correlation analysis of the ratings with mean returns, volatility and beta. First examine the beta. The correlation of each of the risk measures and beta is positive (43% for beta and the ICRG composite). This is exactly the opposite of what one would expect. The evidence suggests that the highest rated countries (lowest financial, economic and credit risk) have the highest betas. This is evident in Figure 4 which graphs the betas against the average risk measures. This positive relation is largely due to the fact that the emerging markets have very low betas with respect to the world market portfolio [see Harvey (1995)].

Also notable in this panel is the sharp negative correlation between volatility and the risk measures. This closely squares with intuition. The lowest (highest) rated countries have the highest (lowest) equity return volatility. This volatility is robust across all risk measures except for political risk. In emerging markets, there is a positive relation between the political risk and the volatility.

We also investigate whether the ICRG ratings, given that they are made on a more timely basis, contain advance information regarding the II credit rating. These results are contained in Table 8. We find that the February-August and August-next February ICRG rating change predicts the March-September and September-next March II CCR. The regression results show that the ICRG ratings lead the II CCR. Each of the ICRG components enters regressions predicting next period’s CCR change with coefficients more than four standard errors from zero. The t-ratio on the ICRG composite measure is 7.6. All coefficients are positive. We also assessed the impact of the change in the II credit rating on the next month’s change in the ICRG rating. There is little information here. Only one of the regressions, the financial risk, has a coefficient that is significantly different from zero. However, the coefficients are all negative which makes little sense (increased CCR predicts lower ICRG ratings). We conclude that the ICRG contains information that predicts CCR but the reverse is not true.

3.2 An initial portfolio strategy

The final panel of Table 7 suggests that there is a relation between average return and average rating. One way to analyze this relation is to form a portfolio strategy based on ratings changes. One version of this strategy is analyzed in Diamonte, Liew and Stevens (1996). They form two portfolios: upgrade and downgrade based on the ICRG political risk measure. Importantly, their strategy is ex post — rather than ex ante. That is, their strategy is only investable if you knew in advance what next month’s rating would be.

We form two portfolios: upgrade and downgrades. The portfolios are rebalanced every six months. If the rating does not change, we keep the country in its respective portfolio. This reduces transactions costs and increases the number of countries in the portfolios. Figure 5 details the number of countries in each portfolio for the overall sample.

The first panel of Table 9 complements the Diamonte, Liew and Stevens (1996) results. The upgrade portfolios have higher average returns than the downgrade portfolios. This is true not just for political risk, but also for the other risk measures as well. Indeed, of the four ICRG risk measures, political risk is never the most important measure. In the all country group and the emerging markets, the composite risk measure is more important in the sense that it implies a more profitable portfolio strategy. In the developed countries, the financial risk measure has the most ability to discriminate between high and low expected return portfolios.

The second panel presents the results from an investable strategy. Portfolios are formed based on a previous rating change and held for six months. The results suggest that the composite risk measure has considerable power to identify high and low return portfolios. On a risk adjusted basis, the hedge strategy (buy upgrades and sell downgrades) based on composite risk has an (annualized) alpha of 1070bp per year when all countries are examined. Interestingly, the political risk measure has an alpha of -80bp per year. The portfolios formed on economic risk have an alpha of 760bp. The portfolios based on financial and credit risk fare worse than the political risk measure.

Perhaps the most interesting aspect of the analysis is that this is not just an emerging markets phenomena. If the portfolio strategy is executed on developed countries, the alpha for the composite risk is 950bp. For the economic risk, the alpha is 540bp. Similar to the overall sample, the financial, political and credit measures fare poorly.

When the emerging markets are examined, the alpha on the composite risk measure is 840bp. For the economic risk, the alpha is 820bp. Consistent with the developed market analysis, financial, political, and credit risk are unable to distinguish between high and low returns in a portfolio strategy.

4. Cross-sectional analysis of risk

4.1 Time-series cross-sectional analysis of risk attributes

The analysis in Table 9 suggests that portfolio groupings by certain attributes produce positive risk adjusted profits on an ex ante basis (panel B). This simple analysis was based on two portfolios: increased risk rating and decreased risk rating. Much information is potentially lost with such a coarse aggregation. More information can be obtained by trying to predict both the cross-section and the time-series of expected returns based on the risk attributes.

Table 10 presents attribute regressions of form: where R represents a vector of six-month returns from (earliest) July 1984- June 1995 for all of the countries in our sample (some markets’ returns begin later). A represents the risk attribute which is lagged and matched to the country. With our full sample of all countries, there are 884 observations. When developed countries are examined, there are 441 observations. There are 443 observations in the emerging equity market sample.

The first line of Table 10 reports coefficients and t-statistics for five separate regressions of the returns on the attributes. The R-square from this regression and the R-square from a similar regression with indicator variables for each country is also reported (fixed effect regression or FER2). Regressions are estimated on the lagged level of the attribute as well as lagged changes in the attribute. The number of countries in each cross-sectional regression grows from 28 in March 1984 to 48 in March 1995.

Panel A reports the regressions using the lagged level of the attribute as an explanatory variable for the cross-section of expected returns. With the full sample of 48 countries, each of the five risk attributes has a coefficient significantly different from zero when the attributes are examined separately. In each case, the coefficient is negative implying that lower rating (higher risk) is associated with higher expected returns. In this analysis, the financial risk variable is most important and the composite is second most important. When a multivariate regression is estimated on four attributes, only the financial risk variable enters the regression with a significant coefficient.

When the regressions are estimated separately on developed and emerging countries, it is clear that the developed countries sample is driving the explanatory power of the entire sample. The ICRG financial risk variable enters the developed country regression with a t-ratio close to four in the univariate regression. In both the univariate and multivariate regressions, both the ICRG financial and economic risk variables enter the regression while the political risk and II credit risk do not. In the emerging countries sample, no attribute enters the regression with a significant coefficient. However, the closest variables are the ICRG financial and economic risk variables.

The second panel of Table 10 examines regressions of country returns on the lagged change in the attributes. This panel is closer to the trading strategy results in Table 9 where portfolios were constructed of countries whose rating had changed. While the change regressions have less explanatory power than the lagged levels, there are many similarities between panel A and panel B. In particular, in the developed country sample, the ICRG financial variable is by far the most important with t-ratios close to two. In contrast to Panel A, the political risk variable is important — but only for the emerging market sample.

4.2 The relation between the risk attributes and fundamental variables

Ferson and Harvey (1995) provide evidence fundamental attributes, such as book-to-price, earnings-price, dividend-to-price, and price-to-cash ratios, are linked to the risk exposure of national markets. They propose an asset pricing framework where the fundamental attributes are linked to dynamic country risk (beta) functions. We now explore whether our set of risk variables are also linked to fundamental attributes.

We focus on a set of three attributes that are available for both the developed and emerging markets: book-to-price, earnings-price and dividend-to-price ratios. Table 11 provides cross-sectional regressions that attempt to explain the cross-section of the valuation attributes using our country risk measures. Similar to Table 10, both univariate regressions and multivariate regression are presented for three different samples. We will focus our discussion on the multivariate regressions. The number of countries in the cross-sectional regression varies from 18 in March 1984 to 47 in March 1995.

There are a number of interesting results in Table 11. First, the risk measures have some ability to explain all three valuation ratios. However, the risk measures do the best in explaining the cross-sectional variation in the book-to-price ratios. In the full sample, over 25% of the variation can be explained using all four risk measures. Of all the component risk measures, the ICRG economic risk rating is the most important accounting for 18% of the variation in the full sample of countries. The economic risk measure enters with a negative coefficient. This suggests that higher rating (lower risk) is associated with lower book-to-price ratios.

When the data are segmented by developed and emerging countries, the results are similar to the pooled regression. For developed countries, 18% of the cross-sectional variance of the book-to-price ratios can be explained of which the economic risk variable contributes 10%. For the emerging equity markets, 29% of the variation can be explained in the multivariate model and the economic risk variable contributes 24%.

The risk variable also shows some ability to explain the cross-section of dividend yields. However, with the dividend yield, the results are different across the developed and emerging markets. For the developed markets, a total of 16% of the cross-sectional variation can be explained. The ICRG economic risk measure contributes 13% and the financial risk variable contributes 5% (the contributions need not sum to the total because of correlation between the risk measures). In emerging markets, a similar amount of explanatory power is found, 17%. However, in this case, almost all of the explanatory power is coming from the financial risk which is contributing 16%. In both the developed and emerging market regressions, the two risk variables enter with negative coefficients suggesting that higher rating (lower risk) is associated with lower dividend yields.

The incremental contribution of the risk ratings relative to the book-to-price valuation measure is presented in Table 12. Univariate regressions are estimated in the form

Table 12 reports the coefficients and t-ratios on the risk attribute. The results indicate that there is little or no incremental information in the II credit rating or the ICRG political risk measure. The ICRG financial risk measure is the most important variable followed by the ICRG economic risk. While both the economic and financial risk measures add important incremental explanatory power to the developed country regressions, they have no ability to add to the emerging equity market regressions. For the emerging markets, the book-to-price valuation attribute fully characterizes the information in the risk ratings.

4.3 Trading strategies based on risk attributes

There are two disadvantages of the time-series cross-sectional methodology. First, in stacking the time-series of returns together, important information regarding the cross-sectional correlation of the returns is eliminated. This could cause the standard errors to be understated. Second, the time-series cross-sectional methodology imposes the same slope coefficient for all time-periods. It is possible that the slope coefficient could change through time. Ferson and Harvey (1991, 1993) find that the variation in the slope coefficients is to some degree predictable.

Panel A of Table 13 reports the results of estimating a cross-sectional regression at each six month interval. The slope coefficients are averaged and the standard error of the average is also presented. The results are largely consistent with those reported in Table 10 for the univariate regressions. For each of the risk measures, the average slope coefficient is negative. In all cases, the slope coefficient is greater than 1.5 standard errors below zero.

There is variation in the estimated cross-sectional slope coefficients through time as is evident in Figure 6 which presents the time-series coefficients for the five country risk measures along with the time-series of R-squares. In all cases, the explanatory power of the risk variables increases through time.

Panel B of Table 13 assesses the incremental contribution of the risk attributes when book value-to-price is included in the cross-sectional regression. While the coefficients are negative for each of the risk attributes, only II credit risk and the ICRG composite and political ratings have coefficients which are more than one standard error below zero.

Panel C of Table 13 implements the Ferson and Harvey (1995) attribute adjustment methodology. Ferson and Harvey make the beta risk (with respect to the MSCI world portfolio) a function of the attribute. We estimate a time-series regression for each country: With the results of this regression, an adjusted attribute is formed:

Panel C reports regressions of the cross-section of returns on the cross-section of adjusted attributes. Here the coefficient is positive suggesting a positive relation between beta risk and expected returns. While the many of the coefficients are not significantly different from zero at conventional levels, the explanatory power of the regressions uniformly improve over the panel A regressions which consider only the raw attribute.

The final panel in Table 13 considers both the adjusted and unadjusted attribute. Note that there is problem with collinearity here because the adjustment factors are country specific. Consistent with the results in Panel C, the beta risk factor enters each regression with a positive coefficient. Each of the II and ICRG risk measures enters with a negative coefficient. Notable in this table is the large jump in explanatory power. The average cross-sectional R-square for the II credit risk measure is now 26% (regressing on credit risk alone produces on a 8% R-square). Both the economic and financial risk measures show similar explanatory power.

Conclusions

The goal of this research is to explore the economic content of five country risk measures: Institutional Investor’s country credit rating and the International Country Risk Guide’s political, financial, economic and composite risk ratings. Our analysis suggests that there is considerable information contained in the ICRG composite, financial and economic ratings, in particular. For example, when we form portfolios based on changes in the risk ratings, we find risk-adjusted abnormal returns in the range of 1000bp per year. Our exercise shows that trading on the political risk measure has no ability to produce abnormal returns which is in contrast to the message in Diamonte, Liew and Stevens (1996).

The cross-sectional regressions confirm the results of the portfolio analysis. We find evidence that some of the ICRG risk measures, in particular, economic and financial risk, can predict the cross-section of expected returns. This is most strongly evidenced in the developed markets in our sample. We also find that the change in the political rating has some marginal explanatory power in emerging equity markets — but not in developed markets.

We find that the country risk ratings are correlated with fundamental valuation attributes. For example, 25% of the cross-sectional variation in book-to-price can be explained by the risk ratings. This explanatory power is largely driven by the ICRG economic risk which alone can explain 18% of the cross-sectional variation. These results shed light on the information that determines the fundamental valuation measures. We provide insights on why global value-oriented strategies work.

Our final contribution is to bridge attribute-oriented investment strategies with asset pricing. Ferson and Harvey (1995) argue that popular valuation attributes should enter each country’s dynamic risk function. We follow their suggestion and find a relation between dynamic risk with respect to a world benchmark and expected returns. In addition, similar to Ferson and Harvey’s results, we find that the attributes are still important for the alpha. That is, even after allowing for the attributes to influence the beta risk, they still have marginal cross-sectional explanatory power when included in the prediction exercise.

Acknowledgements

We appreciate the help of Chris Rath at Political Risk Services in supplying us with the data. Rob Feldman provided valuable research assistance.

References

Bekaert, Geert, 1995, Market integration and investment barriers in emerging equity markets, World Bank Economic Review 9, 75—107.

Bekaert, Geert, and Campbell R. Harvey, 1995a, Emerging equity market volatility, Working paper, Duke University and Stanford University.

Bekaert, Geert, and Campbell R. Harvey, 1995b, The cost of capital in emerging markets, Working paper, Duke University and Stanford University.

Bekaert, Geert, and Campbell R. Harvey, 1995c, Emerging capital markets and economic development, Working paper, Duke University and Stanford University.

Black, Fischer, 1972, Capital market equilibrium with restricted borrowing, Journal of Business 45, 444-455.

Diamonte, Robin, John M. Liew and Ross L. Stevens, 1996, Political risk in emerging and developed markets, Financial Analysts Journal, forthcoming.

Erb, Claude B. Campbell R. Harvey and Tadas E. Viskanta, 1994, National risk in global fixed income allocation, Journal of Fixed Income September, 17-26.

Erb, Claude B. Campbell R. Harvey and Tadas E. Viskanta, 1995, Country risk and global equity selection, Journal of Portfolio Management 9, Winter, 74-83.

Erb, Claude B. Campbell R. Harvey and Tadas E. Viskanta, 1996, Expected returns and volatility in 135 countries, Journal of Portfolio Management forthcoming.

Errunza, Vihang R. and Etienne Losq, 1985, International asset pricing under mild segmentation: Theory and test, Journal of Finance 40, 105—124.

Fama, Eugene F. and James D. MacBeth, 1973, Risk, return and equilibrium: Empirical tests, Journal of Political Economy 81, 607-636.

Ferson, Wayne E. and Campbell R. Harvey, 1991, The variation of economic risk premiums, Journal of Political Economy 99, 285-315.

Ferson, Wayne E. and Campbell R. Harvey, 1993, The risk and predictability of international equity returns, Review of Financial Studies 6, 527—566.

Ferson, Wayne E. and Campbell R. Harvey, 1994, An exploratory investigation of the fundamental determinants of national equity market returns, in Jeffrey Frankel, ed. The Internationalization of Equity Markets, (University of Chicago Press, Chicago, IL), 59—138.

Ferson, Wayne E. and Campbell R. Harvey, 1995, Country risk in asset pricing, Working paper, Duke University.

Gibbons, Michael R. 1982, Multivariate tests of financial models: A new approach, Journal of Financial Economics 10, 3-27.

Harvey, Campbell R. 1991, The world price of covariance risk, Journal of Finance 46, 111—157.

Harvey, Campbell R. 1993, Portfolio enhancement with emerging markets and conditioning information, in Stijn Claessens, and Sudarshan Gooptu, eds. Portfolio investment in developing countries, (World Bank, Washington), 110—144.

Harvey, Campbell R. 1995, Predictable risk and returns in emerging markets, Review of Financial Studies 8, 773-816.

Harvey, Campbell R. Bruno H. Solnik, and Guofu Zhou, 1994, What determines expected international asset returns. Working paper, Duke University, Durham, NC.

Harvey, Campbell R. and Guofu Zhou, 1993, International asset pricing with alternative distributional assumptions, Journal of Empirical Finance 1, 107-131.

Liew, John M. Stock returns, inflation, and the volatility of growth in the money supply: Evidence from emerging markets, Working paper, University of Chicago, Chicago, IL.

Lintner, John, 1965, The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets, Review of Economics and Statistics 47, 13—37.

Ross, Stephen A. 1976, The arbitrage theory of capital asset pricing, Journal of Economic Theory 13, 341-360.

Shanken, Jay, 1990, Intertemporal asset pricing: An empirical investigation, Journal of Econometrics 45, 99—120.

Sharpe, William, 1964, Capital asset prices: A theory of market equilibrium under conditions of risk, Journal of Finance 19, 425—442.

Solnik, Bruno, 1974, An equilibrium model of the international capital market, Journal of Economic Theory 8, 500—524.

Solnik, Bruno, 1974b The international pricing of risk: An empirical investigation of the world capital market structure, Journal of Finance 29, 48—54.

Solnik, Bruno, 1977, Testing international asset pricing: Some pessimistic views, Journal of Finance 32 (1977), 503—511.

Solnik, Bruno, 1983, The relationship between stock prices and inflationary expectations: The international evidence, Journal of Finance 38, 35—48.

Solnik, Bruno, 1983, International arbitrage pricing theory, Journal of Finance 38, 449—457.

Stambaugh, Robert F. 1982, On the exclusion of assets from tests of the two parameter model: A sensitivity analysis, Journal of Financial Economics 10, 237-268.

Stehle, Richard, 1977, An empirical test of the alternative hypotheses of national and international pricing of risky assets, Journal of Finance 32, 493—502.

Stulz, Rene, 1981a, On the effects of barriers to international investment, Journal of Finance 36, 923-934.

Stulz, Rene, 1981b, A model of international asset pricing, Journal of Financial Economics 9, 383-406.

World Bank, 1993, Emerging stock markets factbook, (International Finance Corporation, Washington, D.C.).