The Behavior of Emerging Market Returns

Geert Bekaert,
Stanford University, Stanford CA, 94305
National Bureau of Economic Research, Cambridge, MA 02138

Claude B. Erb
First Chicago Investment Management Co., Chicago, IL 60670

Campbell R. Harvey
Duke University, Durham, NC 27708
National Bureau of Economic Research, Cambridge, MA 02138

Tadas E. Viskanta
First Chicago Investment Management Co., Chicago, IL 60670

1. Introduction

Currency devaluations, failed economic plans, regulatory changes, coups and other national financial "shocks" are notoriously difficult to predict and may have disasterous consequences for global portfolios. Indeed, these characteristics often define the difference in investment in the capital markets of developed and emerging economies.

Research on emerging markets has suggested three market features: high average returns, high volatility and low correlations both across the emerging markets and with developed markets. Indeed, the lesson of volatility was learned the hard way by many investors in December 1994 when the Mexican stock market began a fall that would reduce equity value in U.S. dollars by 80% over the next three months.

But, we have learned far more about these fledgling markets. First, we need to be careful in interpreting the average performance of these markets. Harvey (1995) points out that the International Finance Corporation (IFC) backfilled some of the index data resulting in a survivorship bias in the average returns. Second, the countries that are currently chosen by the IFC are the ones that have a proven track record. This selection of winners induces another type of selection bias. Third, Goetzmann and Jorion (1996) detail a re-emerging market bias. Some markets, like Argentina, have a long history beginning in the last half of the 19th century. At one point in the 1920's, Argentina's market capitalization exceeded that of the U.K. However, this market submerged. To sample returns from 1976 (as the IFC does), only measures the "re-emergence" period. A longer horizon mean, in this case, would be lower than the one calculated from 1976. This insight is consistent with the out-of-sample portfolio simulations carried out by Harvey (1993) indicating that the performance of the dynamic strategy was affect by the initial five years. Fourth, exposure as measured by the IFC is not necessarily attainable for world investor's [see Bekart and Urias (1996)].

Second, we have learned that the emerging market returns are more predictable than developed market returns. Harvey (1995) details much higher explanatory power for emerging equity markets than developed market returns. The sources of this predictability could be time-varying risk exposures and/or time-varying risk premiums, such as in Ferson and Harvey's (1991, 1993) study of U.S. and international markets. The predictability could also be induced by fundamental inefficiencies.

In many countries, the predictability is of a remarkably simple form: autocorrelation. For example, Harvey (1995) details 0.25 autocorrelation coefficient for Mexico in a sample that ends in June 1992. An investor who followed a strategy based on autocorrelation in this country would have lost 35% like everyone else in December 1994. However, the investor would have been completely out of the market in the next three months (or short if possible). Momentum appears to be important for many of these markets.

Third, we have learned that the structure of the returns distribution is potentially unstable. Ghysels and Garcia (1996) reject the structural stability of the prediction regressions presented in Harvey (1995). These regressions allow for the influence of both local and world information. Bekaert and Harvey (1995, 1996) present a model which explains the results of Ghysels and Garcia. The Bekaert and Harvey model allows for the relative influence of local and world information to change through time. They hypothesize that as a market becomes more "integrated" into world capital markets, the world information becomes relatively more important. Bekaert and Harvey (1996) find that the changing relative importance of world information also influences volatility.

Fourth, the Bekaert and Harvey (1996) framework suggests the increasing influence of world factors on emerging expected returns will manifest itself in increased correlation with developed market benchmarks.

The goal of this paper is to explore three aspects of the emerging markets data. First, we examine the behaviorial characteristics beyond the volatility - the skewness and kurtosis. Second, the paper explores the relation between risk variables and expected returns. Harvey (1995) and Bekaert (1995) find that higher betas (from a capital asset pricing framework) are associated with lower expected returns. This is the opposite from what we would expect from theory, however, it is consistent with these markets being segmented. That is, the countries with the higher betas are the ones that are more likely integrated, hence have lower expected returns relative to the segmented countries. Third, we examine the time-varying correlation of these markets with developed markets. Solnik and Longin (1994) and Erb, Harvey and Viskanta (1995) detail how correlations change through time in developed markets. Harvey (1995) and Bekaert and Harvey (1996) show some evidence that correlations are changing in emerging markets. Finally, we examine what is important for explaining both the cross-section of expected returns and volatility in emerging markets. Following Erb, Harvey and Viskanta (1996), we try to link political, economic, and financial risk, as well as, a number of fundamental attributes to explain the cross-sectional behavior of emerging market returns.

2. Distribution of Emerging Market Returns

2.1 Which emerging market benchmarks should be used?

The two main sources of emerging market benchmarks are the International Finance Corporation (IFC) and Morgan Stanley Capital International (MSCI). Both provide country benchmark indices which are based on a value weighted portfolio of a subset of stocks which account for a substantial amount of the market capitalization within each emerging market.

The IFC produces two types of indices: Global (IFCG) and Investible (IFCI). For nine countries, data exists back to 1976. Currently, the IFC provides data on 27 countries. MSCI also produces both Emerging Markets Global (EMG) and Emerging Markets Free indices (EMF) which resembles the IFCI. Our paper focusses on the global indices. Part of the interest in studying emerging markets is the impact capital market liberalizations have on the returns. Hence, we study markets before and after they are accessible by international investors.

IFC and MSCI use a different hierarchical process in the company selection for the country indices. MSCI follows the same technique that it uses in its popular developed country indices. First, the market is analyzed from the perspective of capitalization and industry categories. Next, a target of 60% coverage of the total capitalization of each market, with industry weightings approximating the total market's weightings is established. Finally, companies are selected based on liquidity, float, and cross-ownership to fulfil these goals.

In contrast, the IFC's order of preference is: size, liquidity and industry. The IFC primarily targets the largest and most actively traded stocks in each market, with a goal of 60% of total market capitalization at the end of each year. As a second objective, the index targets 60% of the trading volume during the year. Industry is of tertiary priority.

Although there is some hierarchical differences in the structure of construction, there is little difference in the behavior of the IFCG and the EMG. Table 1 details the difference between the IFCG and the EMG returns over identical samples for each index. Of the 22 countries where there is MSCI and IFC data, the returns indices have greater than 94% correlation. The volatility differences are quite small - as is the tracking error of the two indices.

The only country where substantial deviations occur is Argentina. The IFC index produced a 10.6% lower average return and an 21.1% lower volatility. For this country, the correlation between the IFC index and the MSCI index is only 76%. However, much of the tracking error is due to 1988-1989 data. When we redo the comparison for Argentina beginning in January 1990, the tracking error drops from 61.9% to 10.6%. The correlation increases from 76% to 99%. There is no difference in the mean returns and little difference in the volatilities. Hence, even for Argentina, there is does not appear to be a substantive difference between the MSCI and IFC indices.

The IFC family of indices presents the longest history and, as a result, we choose to focus on the IFC. In addition, we study total market returns measured in U.S. dollars. The local currency returns are not, in general, available to international investors. Furthermore, hedged returns are not available either. Table 2 presents the total sample of emerging markets followed by the IFC and some summary measures of capitalization (in U.S. dollars) along witht the number of countries in each index and the weight in the IFC Composite.

2.2 Time-varying mean performance

Some summary statistics for the emerging market returns are presented in Table 3 for the sample of 27 countries followed by the IFC Global indices. We report summary statistics for the common period of April 1991 to March 1996. We concentrate our analysis on those countries with at least five years of data.

The summary performance measures detailed by Harvey (1995) with data through mid-1992 are largely replicated in the last five years, with some exceptions. For example, Harvey (1995) found that some the largest average returns across the emerging markets were found in Argentina and Brazil. This is also the case over the last five years with arithmetic average annual returns of 35.5% and 44.3%. Interestingly, these are two of the "re-emerged" markets studied by Goetzmann and Jorion. Both of these markets began in the late 19^th century and effectively disappeared. They are among the original members of the IFC database which begins in 1976. It is not clear what the role of the re-emergence bias is in these markets.

One other difference in performance is Nigeria. The overall average return is 37.5% over the past five years. However, this return should be contrasted to the geometric average (which better represents a buy and hold strategy) of 12.7%. This market suffered a 70% drop in value in March 1995 as a result of the substantial devaluation of the Nigerian naira (from $0.045 to $0.015 per naira).

Of course, the major difference in performance comes from Mexico and Venezuela. The average return in Mexico over the period is only 15.7% which is sharply lower than the average returns detailed in Harvey (1995) reflecting the 80% drop in the Mexican market over the period December 1994-February 1995. Over the past five years, the average return in Venezuela was negative.

Over this period, the average return of the IFC Composite was similar to the MSCI World and the MSCI World-All Countries. The difference between the emerging markets is in the volatility. The IFC Composite had a volatility of 16.4% compared to the MSCI World-AC volatility of 10.5%. Hence, over the past five years, the contribution of a diversified emerging markets investment to a diversified global portfolio must have come from the correlation properties.

Figure 1 presents rolling five year average returns for the 20 emerging markets and the 3 benchmark indices. The "re-emergence" effect of Goetzmann and Jorion (1996) appears important for six countries in particular: Argentina, Chile, the Philippines, Portugal, Taiwan and Turkey. For these countries, the average returns in the five years after emergence in the IFC database are much higher than the subsequent five years. However, there are a number of exceptions. There is no such pattern in Brazil, Greece, Colombia, Mexico, Nigeria, Pakistan, South Korea, Thailand, Venezuela, and Zimbabwe.

These figures suggest that the mean returns are time-varying. The evidence presented in Harvey (1993, 1995), Bekaert (1995) and Bekaert and Harvey (1995) suggest that emerging market returns are more predictable than developed market returns. While rolling five year mean returns are useful descriptors of the data, the evidence on predictability suggests that time-varying means are best captured by regression models.

Bekaert and Harvey (1995, 1996) suggest that care must be taken in specifying the prediction model. In particular, if a market experiences increased (or decreased) integration into world capital markets, it is likely that the parameters of the prediction model change through time. Bekaert and Harvey propose models where the influence of world versus local information changes with the degree of integration. That is, as a market become more integrated into world capital markets, it is more likely that world information will have a greater impact on the time-varying mean returns.

2.3 Changing volatility

Figure 2 presents five-year unconditional volatilities for the 20 emerging markets and the 3 benchmark indices. We will focus on the patterns in the last five years. For many countries, there has been a sharp decrease in volatility. Most notably, the volatility levels in Argentina and Brazil have been cut in half over the past five years. Other countries which have experienced large decreases in volatility are: Greece, Jordan, Portugal, Taiwan, and Turkey. While there are a number of countries that have seen increased volatility, the overall pattern in emerging market volatility is downward. The IFC composite volatility was 28% in 1991 and only 16% in the five years ending March 1996. This decreased volatility mirrors the decreased volatility in the MSCI World index - which dropped from 18% in 1991 to 10.7% in the five years ending March 1996.

There are a number of hurdles that one faces when trying to understand volatility in emerging equity markets. First, given the evidence of nonnormalities in the market returns [see Harvey (1995) and Bekaert and Harvey (1996)], it is unlikely that the standard implementation of autoregressive conditional heteroskedasticity (ARCH) models is fruitful. As a result, only models which explicitly account for leptokurtosis and skewness are likely to be useful. Second, given the existing evidence on return predictablity [see Bekaert and Harvey (1995)], variance specifications allow for time-varying conditional means.

Importantly, volatility process should allow the relative importance of local and world information shift through time as emerging equity markets become more or less integrated into world capital markets. As with the mean process, the increasing impact of world factors on volatility may be consistent with increased market integration.

2.4 Changing skewness and kurtosis

It is well known that emerging market returns depart from the normality. Tests presented in Harvey (1995) and Bekaert and Harvey (1996) show substantial deviations from normality. Part of the goal of this paper is to examine which countries show large deviations from normality - and how those deviations change through time.

There are a number of reasons why we observe non-normality in the equity market returns. First, the presence of limited liability in all equity investments may induce option-like asymmetries in returns [see Black (1976), Christie (1982) and Nelson (1991)]. Second, the agency problem may induce asymmetries in index returns [see Brennan (1993)]. That is, a manager has a call option with respect to the outcome of the firm's investment decisions. Managers may prefer high positive skewness. The behavior of conditional skewness is studied in Harvey and Siddique (1995) for a sample of developed countries.

Emerging market returns have more positive skewness than developed market returns. The coefficient of skewness is greater than zero in 16 of 20 emerging equity markets. The highest skewness is found in Argentina and Taiwan's equity returns. Similarly, emerging markets present more excess kurtosis than the world benchmark. There are only two countries, Chile and Jordan, where the excess kurtosis is lower than the world benchmark.

Tests of normality based on Hansen's generalized method of moments which are detailed in Harvey (1995) are executed. Normality is rejected in 11 of the 20 countries based on the last five years of data. These results are consistent with those in Harvey (1995) and Bekaert and Harvey (1996). Bekaert and Harvey (1996) extend these tests by carrying out Monte Carlo simulations. The Monte Carlo analysis confirms that a number of countries deviate from the assumption of normality.

Figures 3 and 4 presents the rolling five-year skewness and excess kurtosis measures for the 20 emerging markets and the benchmark returns. If the data are normally distributed, then both of these measures should be zero. Over the past five years, skewness has shown large increases (become more positive) in Argentina, Brazil, Colombia, Nigeria, Philippines, South Korea, and Taiwan. In Mexico, skewness sharply increased in 1992 and 1993. However, much of that increase was reversed with the returns in December 1994-February 1995. In the IFC composite index, five-year trailing skewness increased from 1991 through mid-1995. Since then, the skewness has dropped back close to zero.

There are two countries with huge excess kurtosis: Argentina and Taiwan. In both of these countries, there is a progressive increase in kurtosis in the last five years. In addition to these two countries, eight other countries experienced increases in excess kurtosis. Seven countries experienced decreased kurtosis. In the remaining three countries, there was little change in kurtosis. Overall, the IFC composite showed a sharpe increase in the excess kurtosis from 0.8 in 1991 to 3.5 in March 1996. In the MSCI World-AC index the excess kurtosis was only 0.4 in March 1996.

The departures from normality are important to portfolio managers in a number of respects. First, the usual mean-variance framework is no longer adequate to characterize investment decisions. It is reasonable to assume that investors have a preference over skewness (prefer more positive skewness). It is also reasonable to expect that there are preferences over kurtosis. The second implication is that these higher moments are evolving through time. Consistent with the evidence on the means and volatilities, dynamic models of these higher moments are necessary.

2.5 Conditional correlations

Figure 5 presents rolling five-year correlations with the MSCI World-AC for the 20 emerging markets and the IFC Composite index return. As detailed by Harvey (1995), these correlations are generally small. By March of 1996, the five-year trailing correlation with the world is less than 40% in all countries except for Portugal. Interestingly, there has been no clear trend in the correlations across the emerging markets over the past five years.

Bekaert and Harvey (1996) present a model of conditional correlation where the means, volatilities and covariances are influenced by both local and world information. Their model predicts that as the a market becomes more integrated, the relative influence of world and local information changes. This change will affect the conditional correlation between the emerging market and the world benchmark. Bekaert and Harvey offer evidence that conditional correlations increase after capital market liberalizations.

Many of the major liberalizations occurred before 1992 [see Bekaert (1995)]. It is also clear in the data that correlations have increased over the longer horizon. For example, the correlation of Argentina and Brazil with the world was zero or slightly negative in the five years ending in 1981. By March 1996, the correlations were above 30%. Long horizon increases are also evident in South Korea and Thailand. However, in the last five years, there has been little overall change. In the five years ending in 1991, the correlation of the IFC composite and the world was 30%. In March 1996, the correlation was 35%. Slightly higher correlations are found comparing the IFC Composite to the MSCI World-AC, which has a higher emerging market composition.

3. Measuring Risk in Emerging Market Returns

3 .1 Asset pricing theory and emerging market returns

Risk is notoriously difficult to measure in emerging market returns. A simple implementation of the Capital Asset Pricing Model (CAPM) of Sharpe (1964) and Lintner (1965) is problematic for a number of reasons. In these markets, there is a no relation between the risk measured by the CAPM and expected returns.

The risk-return relation is detailed in Figure 6 which plots the average returns over the past five years and the beta against the World-AC index. These betas are also presented in Table 3. While there is a positive relation between beta and average returns, the t-statistic on the beta coefficient is 1.1 which is not significant at conventional levels.

The failure of the CAPM to explain emerging market returns could be interpreted in a number of ways. First, following Roll and Ross (1995) and Kandel and Stambaugh (1995), the benchmark world portfolio may not be mean-variance efficient. Second, perhaps a multifactor representation, following Merton (1973), Ross (1976) and Chen, Roll and Ross (1986) is more appropriate for emerging markets. Third, following Ferson and Harvey (1991), an examination of average returns and average risk could be misleading if the risk and expected returns change through time. Finally, the CAPM is not the appropriate framework if these markets are not integrated into world capital markets. In integrated capital markets, the same risk project commands identical expected returns, irrespective of domicile [see Stulz (1981), Bonsar-Neal, Brauer, Neal and Wheatley (1986), Campbell and Hamao (1992), Chan, Karolyi, and Stulz (1994), Heston, Rouwenhorst and Wessels (1995), Bekaert (1995), Harvey (1991, 1995), and Bekaert and Harvey (1995).]

It is most likely that many of these markets are not integrated into world capital markets. As a result, the beta suggested by the CAPM may not be that useful in explaining the cross-section of average returns. Indeed, in completely segmented capital markets, the volatility is the correct measure of risk. The relation between average returns and volatility is detailed in Figure 7. Similar to the beta graph, there is a positive relation which is now significant at conventional levels of confidence (R-square is 21%). However, even among the segmented markets, there might only be a weak relation between volatility and expected returns because the premium accorded to volatility could vary across countries.

3.2 Alternative risk attributes

Following Ferson and Harvey (1994), Erb, Harvey and Viskanta (1994) and others, we examine the relation between some country-specific risk attributes and distribution of returns. We group these attributes into the following categories:

3.2.1 Survey-based measures

The first of these measures is Institutional Investor's Country Credit Rating (IICCR). Institutional Investor country credit ratings are based on a survey of leading international banks who are asked to rate each country on a scale from zero to 100 (where 100 represents the maximum creditworthiness). Institutional Investor averages these ratings, providing greater weights to respondents with higher worldwide exposure and more sophisticated country analysis systems. These ratings have appeared in the March and September issues of Institutional Investor since 1979 and now covers over 135 countries, for additional details see Erb, Harvey and Viskanta (1995, 1996a).

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 those factors the expert feels are relevant. In a recent survey of participants, the most important factors for assessing emerging markets' credit rating were (i)debt service, (ii) political outlook, (iii) economic outlook, (iv) financial reserves/current account and (v) trade balance/foreign direct investment.

The next four measures are from Political Risk Services' International Country Risk Guide. They include the Political Risk index (ICRGP), Economic Risk index (ICRGE), Financial Risk index (ICRGF) and the Composite Risk index (ICRGC). The political index is studied in Harlow (1993) and Diamonte, Liew and Stevens (1996). Erb, Harvey and Viskanta (1996b,c) examine the information in all four of the ICRG risk indices.

On a monthly basis ICRG uses a blend of quantitative and qualitative measures to calculate risk indices for political, financial and economic risk, as well as a composite index. Five financial factors, thirteen political and six economic factors are used. Each factor is assigned a numerical rating within a specified range. A higher score represents lower risk, for additional details see Erb, Harvey and Viskanta (1996b).

The composite index is simply a linear combination of the three subindices. The political risk is weighted twice that of either financial or economic risk. ICRG, as well as many of the other providers, think of country risk as being composed of two primary components: ability to pay and willingness to pay. Political risk is associated with a willingness to pay, while financial and economic risk are associated with an ability to pay.

We also include Euromoney's Country Credit Risk (EMCCR). Euromoney's rating system is based on both qualitative and quantitative methods. The political component is a qualitative survey of experts. The economic component is quantitative and based on Euromoney's global economic projections. The financial component is also quantitative and based on (i)debt indicators, (ii)debt in default or rescheduled, (iii) credit rating (Moody's or Standard and Poors), (iv) access to bank finance, (v) access to short-term financing and (vi) access to international bond and syndicated loan markets.

3.2.2 Macroeconomy

In summarizing either the future economic health or the investment risk, investors often refer to the inflationary environment . Ferson and Harvey (1993, 1994) argue that asset exposure versus world inflation helps explain both the cross-section and time-series of expected returns in 18 developed markets. Erb, Harvey and Viskanta (1995) examine the interaction of inflation and asset returns in emerging markets. We use a trailing 6 month measure of inflation represented by the consumer price index reported in the International Financial Statistics database of the International Monetary Fund. In the case of Taiwan, who is not a member of the IMF, we use inflation reported in their national accounts.

3.2.3 Demographics

Bakshi and Chen (1994) propose a life-cycle investment hypothesis. Younger investors demand housing more than equities. As age increases, more investment is allocated to the stock market. As a result, a rise in average age should be accompanied by a rise in the stock market. Bakshi and Chen (1994) find support for this hypothesis using U.S. data. Erb, Harvey and Viskanta (1996c) find that average age growth explains the risk premiums in a number of developed countries. We choose to examine three variables: population growth, average age and average age growth. All of these data are based on annual statistics compiled by the United Nations.

3.2.4 Market integration

Bekaert and Harvey (1996) argue that the size of the trade sector to the total economy is a reasonable proxy for the openness of both the economy and the investment sector. They use exports plus imports divided by GDP as an instrument for market integration. This variable, along with other proxies for market integration, is used in a function which assigns time-varying weights to world versus local information. Bekaert and Harvey find that increases in this ratio are associated with the increased importance of world relative to local information for both the mean and the volatility of the country's stock returns.

3.2.5 Persistence

A number of researchers have pointed to momentum as an important firm specific attribute [see Jegadeesh and Titman (1993), Conrad and Kaul (1996), Asness, Liew and Stevens (1996), Ferson and Harvey (1996)]. We examine two measures of momentum: the lagged monthly return and the lagged quarterly return from four months ago to one month ago, i.e. the quarterly return lagged by an extra month

3.2.6 Size

We follow a number of papers beginning with Banz (1981) that document a relation between firm size and expected returns. This attribute has recently been studied on a country level basis by Asness, Liew and Stevens (1996) who find that size helps explain the cross-section of expected returns in a sample of developed markets.

3.2.7 Fundamental valuation measures

Following a number of papers that link "fundamental attributes" to asset valuation [see, for example Chan, Hamao and Lakonishok (1991), Fama and French (1992) and Ferson and Harvey (1994).] We use three valuation ratios: price to book value, price to earnings and dividend yield. Value-weighted indices of company level data are produced by the IFC. Ferson and Harvey (1996) show that some of these ratios, most notably price to book, appear to capture information regarding changing risk in a sample of 21 developed countries.

3.2.7 Summary statistics

Some summary measures for many of these attributes are included in Table 4. The March 1996 value of the attribute is reported. In the lower panel, the rank-order correlation of all of the attributes is reported. Most of the correlations follow from intuition. Consider the ICRG indices. These indices are highly correlated with the Euromoney and Institutional Investor Country Credit Risk measures. All of the survey measure are negatively correlated with inflation (high inflation means low rating). The most negative correlation with inflation is found for the ICRG Economic risk index. Average age is positively correlated with the survey risk indices, indicating that low average age is associated with low rating. Size is positively related to the ICRG ratings (smaller markets are appear more risky). There is also a positive relation between the size of the trade sector and the ICRG ratings. The lowest correlations are found for the ICRG indices and the fundamental attributes. For example, the correlation between the ICRG political rating and the price to book ratio is -36%, suggesting that high price to book countries on average have low political ratings.

4. What matters in choosing an emerging market for portfolio investment

4.1 Portfolio approach

A commonly used technique in examining the cross-sectional importance of a fundamental variable is to form unique portfolios based on their ranking. We will examine the country risk variables by forming portfolios based on the risk level itself. These portfolios are investible with respect to the attribute. That is, lagged attribute information is used to determine which countries are in the portfolios. Given the small number of emerging markets, we examine only two portfolios: high attribute and low attribute. In each case, we track the returns to portfolios that are equally weighted by country, and those that are weighted by each country's equity market capitalization. To reduce potential transactions costs, we only consider quarterly rebalancing.

Table 5 presents the results of the portfolio strategies. Of the ICRG indices, the Composite index (ICRGC) produces the greatest separation of expected returns. With the equally weighted investment scheme, the high attribute portfolio presents 29.7% average return with a 25.7% volatility. The low attribute portfolio delivers a 36.6% average annual return with a slightly lower volatility, 23.5%. Interestingly, the beta against the World-AC is much lower for the low attribute portfolio. Hence, the alpha of this strategy is quite large.

Of the family of ICRG indices, the financial and economic risk appear to be the most important and in the equally weighted portfolio strategies. The political risk measure is only important in the capitalization weighted investment strategies. These results are consistent with those presented in Erb, Harvey and Viskanta (1996b).

Both the Institutional Investor and the Euromoney credit ratings also are able to significantly discriminate between high and low expected return securities. As with the ICRG Composite, the low attribute portfolios have higher means and lower volatilities than the high attribute portfolios.

Inflation also appears to be an important instrument in portfolio selection. In this case, the high attribute portfolio has much higher expected returns than the low attribute portfolio. However, in contrast to the ICRG, EMCCR and IICCR, the high attribute portfolio has much higher volatility than the low attribute portfolio.

Trade to GDP has only marginal ability to distinguish between high and low expected returns. The low attribute portfolio has higher expected returns than the high attribute portfolio. However, the volatility of the low attribute portfolio is greater. Nevertheless, the beta of the low attribute portfolio is close to zero leading to a very high "alpha". But caution needs to be exercised here. The beta of the low attribute portfolio may be low because the market is not intregrated. The idea of Bekaert and Harvey (1996) is that trade to GDP is a proxy for integration. Indeed, the low beta of the low trade to GDP portfolio is consistent with their results.

The three demographic variables: population growth, average age growth and average age offer no ability to discriminate between high and low expected return countries. The demographic asset pricing theory presented in Chen and Bakshi (1994) is most appropriate for time-series analysis of developed countries. That is, holding other factors constant, an increasing average age will be associated with higher demand for equities. It is difficult, if not impossible, to hold other factors constant in emerging markets. For example, a changing degree of market integration could confound the relation between demographics and returns.

Size appears to be an important instrument in discriminating between high and low expected returns. This is consistent with the analysis of developed markets presented in Asness, Liew and Stevens (1996). For example, the low market capitalization portfolio produces an average annual return of 41.0% with a volatility of 25.9%. The high capitalization portfolio delivers 24.8% average annual return with a 26.9% volatility. The beta of the low capitalization portfolio is much lower than the high capitalization portfolio. As with the trade to GDP measure, some caution needs to be interpreted with reference to the betas. Bekaert and Harvey (1996) consider market capitalization to GDP as another proxy for market integration. It is likely that the low market capitalization is indicative of market segmentation.

The evidence for the momentum variables is mixed. The high momentum portfolio (high lagged returns) has a 38.9% average annual return whereas the low momentum (low lagged return) has a 26.8% average annual return. While this appears impressive, there is also a large difference in volatility. The high momentum portfolio has 26.8% volatility compared to 21.8% volatility for the low momentum portfolio. Interestingly, there is little difference between the momentum specifications. In addition, the betas of the low momentum portfolio is 50% smaller than the high attribute portfolio, effectively eliminating the "alpha".

The final set of attributes involves the traditional accounting ratios. While dividend yields (DP) are available on all the indices on year after the market enters the IFC data, the price to book (PB) and price to earnings (PE) ratios are only available from January 1986. Hence, the evaluation of the PB and PE ratios is over a different sample than all of the other portfolio simulations.

There are sharp differences between the high and low attribute portfolios for these attributes. For example, the high PB portfolio earns 20.8% average return with a 27.7% annual volatility. The low PB portfolio delivers 42.3% average return and a 20.7% volatility. The beta of the high PB portfolio is 0.82 compared to 0.26 for the low PB portfolio. Similar, but less dramatic results, are found for the PE ratio.

The dividend yield also discriminates between high and low expected returns over a longer sample than PB and PE. The high DP portfolio has 29.4% average annual return with 29.2% volatility compared to the low DP portfolio which has 39.6% average return and 22.2% volatility.

These results suggest that there are a number of useful attributes in discriminating between those countries which will experience high and low expected returns. It is likely, as argued in Ferson and Harvey (1994, 1996), that these attributes are related to risk. Unfortunately, determinnig the appropriate measure of risk is difficult in emerging markets.

4.2 Trading emerging market portfolios

The cost of trading is high in emerging markets. Table 6 presents estimates of transactions costs from Barings Securities. The percent spread calculation is the difference between the offer and bid price divided by the average of the offer and bid price. Barings uses the midpoint in the divisor in order to avoid the problems caused by large fluctuations in the current price.

The percent spreads in Table 5 are based on snapshots of individual stocks during the weeks of July 17 and July 24, 1995. The country spreads are calculated by capitalization weighting the percentage spreads of the individual firms within each country.

The percentage spreads are, in many countries, much larger than one would expect in developed markets. The spread in Chile is close to 400bp. In both Argentina and Turkey, the percentage spread is more than 150bp. These high transactions costs suggest reinforce the need to minimize trading. Indeed, many investment managers do not practice active stock selection strategies in emerging markets because of the massive transactions costs. "Active" management in emerging markets is often interpreted in the context of country selection rather than stock selection.

While the portfolio analysis in Table 4 does not explicitly account for transactions costs, we do include a measure of average turnover. The highest turnover is found with the momentum strategies. The turnover is so high that it is unlikely that these strategies could be sucessfully implemented in the form specified here. The lowest turnover is found with the demographic variables. This is not unexpected given that the data is only available annually and there is little variation over the years.

The most impressive ratios of low-high portfolio returns to turnover are found for the survey risk attributes - in particular the ICRGC, EMCCR and IICCR. For example, in the low attribute of EMCCR portfolio, the turnover is 46%. With 10 countries in that portfolio, 4 or 5 would change over the entire year (four rebalancing opportunities).

4.3 Risk attributes and the behavior of emerging market returns

The portfolio exercise suggests that many of these risk attributes can discriminate between high an low expected returns environments. What about the other moments? Erb, Harvey and Viskanta (1995), Bekaert and Harvey (1996) argue that attributes like trade to GDP and credit rating impact conditional volatility. Ferson and Harvey (1994, 1996) establish a link between country attributes and time-varying risk exposures. It makes sense to examine the relation between these some of these attributes and the other moments of expected returns.

Figure 8 focuses on three attributes: ICRGC, trade to GDP, and PB. The average values of these attributes over the past five years is plotted against the moments of the country returns: mean, standard deviation, skewness, excess kurtosis and correlation with the World-AC.

For each of the attributes, there is a negative relation between the value of the attribute and the average returns which is consistent with the portfolio exercise. For the ICRGC and the Trade/GDP, there is a sharp negative relation between the rating and volatility. There is no significant association between PB and volatility. None of these three measure does well in explaining the cross-section of skewness. There is a negative relation between the Trade/GDP variable and skewness - but it is not significant at conventional levels. All three measures show a negative association with kurtosis. Low composite rating, low Trade/GDP and low PB are associated with higher kurtosis. PB has considerable success in explaining the cross-section of betas which is consistent with Ferson and Harvey's (1996) evidence that PB is an instrument for time-varying risk exposure. Low PB is associated with higher risk. Finally, there is a positive relation between the ICRG composite risk and correlation with the world market. Consistent with Bekaert and Harvey (1996), there is a positive relation between Trade/GDP and correlation with the world.

5. Conclusions

This paper has further explored the behavior of emerging market returns. The first goal of the paper was to go beyond the mean and variance and explore skewness and kurtosis. Indeed, there is considerable research interest in asymmetric variances, semi-moment analysis and down-side optimization. These advances make a lot of sense in emerging markets where the returns distributions depart from the usual normal assumption. We show that the deviations from normality are persistent in the emerging market returns, show no evidence of disappearing in the near future, and are time-varying in nature.

The next contribution is to explore what matters for emerging market investment. The traditional beta-risk paradigm is problematic in emerging markets because a number of the markets are unlikely fully integrated into world capital markets. Indeed, in a completely segmented market, country variance (which is usually considered idiosyncratic) is the appropriate measure of risk. We explore a number of risk attributes that have been successfully applied in developed markets. We find that a number of these attributes such as the International Country Risk Guide's Composite Risk, trade to GDP and price to book are useful in identifying high and low expected return environments.

Finally, we try to link our attribute analysis to the behavior of the emerging market returns. The risk attributes, not only discriminate among the mean returns, they also offer information about of the moments. For example, we find that low ICRG Composite ratings are strongly associated with high volatility, high excess kurtosis and low correlations with the world benchmark.

References (preliminary)

Adler, Michael, and Bernard Dumas, 1983, International portfolio selection and corporation finance: A synthesis, Journal of Finance 38, 925984.

Asness, Cliff S., John M. Liew and Ross L. Stevens, 1996, Parallels between the cross-sectional predictability of stock and country returns, Working paper, Goldman Sachs Asset Management, New York, 1996.

Bailey, Warren, and Julapa Jagtiani, 1994, Foreign ownership restrictions and premiums for international investment: Some evidence from the Thai capital market, Journal of Financial Economics 36, 5788.

Bansal, Ravi, David Hsieh, and S. Viswanathan, 1993, A new approach to international arbitrage pricing, Journal of Finance 48, 17191747.

Banz, Rolf W., 1981, The relationship between return and market value of common stocks, Journal of Financial Economics 9, 3-18.

Black, Fischer, 1972, Capital market equilibrium with restricted borrowing, Journal of Business 45, 444455.

BonserNeal, Catherine, Gregory Brauer, Robert Neal, and Simon Wheatley, 1990, International investment restrictions and closedend country fund prices, Journal of Finance 45, 523548.

Brown, Stephen J., William Goetzmann, Roger G. Ibbotson, and Stephen A. Ross, 1992, Survivorship bias in performance studies, Review of Financial Studies 5, 553580.

Bekaert, Geert, 1995, Market integration and investment barriers in emerging equity markets, World Bank Economic Review 9, 75107.

Bekaert, Geert and Campbell R. Harvey, 1995a, Timevarying world market integration, Journal of Finance 50, 403444.

Bekaert, Geert and Campbell R. Harvey, 1995b, "Capital markets: An engine for economic growth," Working paper, Duke University and Stanford University.

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

Bekaert, Geert and Campbell R. Harvey, 1996, "Emerging Equity Market Volatility,"

Working paper, Duke University and Stanford University.

Bekaert, Geert, and Robert Hodrick, 1992, Characterizing predictable components in excess returns on equity and foreign exchange markets, Journal of Finance 47, 467509.

Bekaert, Geert and Michael Urias, 1996, Diversification, integration and emerging market closed end funds, Journal of Finance, forthcoming.

Brennan, M. J., 1993, Agency and asset pricing, Unpublished manuscript,UCLA and London Business School.

Campbell, John Y., 1987, Stock returns and the term structure, Journal of Financial Economics 18, 373400.

Campbell, John Y., and Yasushi Hamao, 1992, Predictable bond and stock returns in the United States and Japan: A study of longterm capital market integration, Journal of Finance 47, 4370.

Chan, K. C., G. Andrew Karolyi, and Rene Stulz, 1992, Global financial markets and the risk premium on U.S. equity, Journal of Financial Economics 32, 137168.

Chan, K. C., Yasushi Hamao, and Josef Lakonishok, 1991, Fundamentals and stock returns in Japan, Journal of Finance.

Chen, Naifu, Richard R. Roll, and Stephen A. Ross, 1986, Economic forces and the stock market, Journal of Business 59, 383403.

Cho, Chinhyung D., Cheol S. Eun, and Lemma W. Senbet, 1986, International arbitrage pricing theory: An empirical investigation, Journal of Finance 41, 313330.

Christie, A. A., 1982, The stochastic behavior of common stock variances: value, leverage, and interest rate effects, Journal of Financial Economics, 23, 407-432.

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

Divecha, Arjun, Jaime Drach and Dan Stefek, 1992, Emerging markets: A quantitative perspective, Journal of Portfolio Management 19, 4156.

Dumas, Bernard, 1994, A test of the international CAPM using business cycles indicators as instrumental variables, in Jeffrey Frankel, ed.: The Internationalization of Equity Markets, (University of Chicago Press, Chicago, IL), 2350.

Dumas, Bernard and Bruno Solnik, 1995, The world price of foreign exchange rate risk, Journal of Finance.

Erb, Claude, Campbell R. Harvey and Tadas Viskanta, 1994, Forecasting international equity correlations, Financial Analysts Journal, NovemberDecember, 3245.

Erb, Claude, Campbell R. Harvey and Tadas Viskanta, 1995a, Country risk and global equity selection, Journal of Portfolio Management, 21, Winter, 7483.

Erb, Claude, Campbell R. Harvey and Tadas Viskanta, 1995b, Inflation and world equity selection, Financial Analysts Journal, NovemberDecember, 28-42.

Erb, Claude, Campbell R. Harvey and Tadas Viskanta, 1996a, Expected returns and volatility in 135 countries, Journal of Portfolio Management, 32—48. Also in ../../Country_risk/couindex.htm.

Erb, Claude, Campbell R. Harvey and Tadas Viskanta, 1996b, Politial risk, economic risk and financial risk, Financial Analysts Journal, forthcoming . Also in ../../Country_risk/couindex.htm.

Erb, Claude, Campbell R. Harvey and Tadas Viskanta, 1996c, Demographics and international investment, working paper, Duke University. Also in ../../Country_risk/couindex.htm.

Errunza, Vihang R., 1983, Emerging markets: New opportunity for improving global portfolio performance, Financial Analysts Journal 39, 5158.

Errunza, Vihang R., and Etienne Losq, 1985, International asset pricing under mild segmentation: Theory and test, Journal of Finance 40, 105124.

Errunza, Vihang R. and Etienne Losq, 1989, Capital controls, international asset pricing and investors' welfare: A multicountry framework, Journal of Finance 44, 10251037.

Errunza, Vihang R., and P. Padmanabhan, 1988, Further evidence on the benefits of portfolio investments in emerging markets, Financial Analysts Journal 44, 7678.

Errunza, Vihang R., Etienne Losq, and Prasad Padmanabhan, 1992, Tests of integration, mild segmentation and segmentation hypotheses, Journal of Banking and Finance 16, 949972.

Eun, Cheol S., and S. Janakiramanan, 1986, A model of international asset pricing with a constraint on the foreign equity ownership, Journal of Finance 41, 897914.

Fama, Eugene F. and Kenneth French, 1992, The cross-section of expected stock returns, Journal of Finance 59, 427-465 .

Ferson, Wayne E. and Campbell R. Harvey, 1991, The variation of economic risk premiums, Journal of Political Economy 99, 285315.

Ferson, Wayne E., and Campbell R. Harvey, 1993, The risk and predictability of international equity returns, Review of Financial Studies 6, 527566.

Ferson, Wayne E., and Campbell R. Harvey, 1994a, Sources of risk and expected returns in global equity markets, Journal of Banking and Finance 18, 775803.

Ferson, Wayne E., and Campbell R. Harvey, 1994b, 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), 59138.

Ferson, Wayne E., and Campbell R. Harvey, 1996, Modelling the fundamental determinants of national equity market returns, Working paper, Duke University, Durham, NC.

Glosten, L. R., R. Jagannathan, and D. E. Runkle, 1993, On the relation between expected value and the volatility of the nominal excess return on stocks, Journal of Finance 48, 1779-1801.

Goetzmann, William N. and Philippe Jorion, 1996, Re-emerging markets, Working paper, Yale University.

Gooptu, Sudarshan, 1993, Portfolio investment flows to emerging markets, in Stijn Claessens and Sudarshan Gooptu, eds., Portfolio investment in developing countries, (Washington: World Bank).

Grauer, Fredrick L. A., Robert H. Litzenberger and Richard Stehle, 1976, Sharing rules and equilibrium in a international capital market under uncertainty, Journal of Financial Economics 3, 233256.

Grauer, Robert R. and Nils H. Hakansson, 1987, Gains from international diversification: 19681985 returns on portfolios of stocks and bonds, Journal of Finance 42, 721739.

Gultekin, N. Bulent, Mustafa N. Gultekin, and Alessandro Penati, 1989, Capital controls and international capital market segmentation: The evidence from the Japanese and American stock markets, Journal of Finance 44, 849869.

Hansen, Lars P., 1982, Large sample properties of generalized method of moments estimators, Econometrica 50, 10291054.

Harlow, W. Van, 1993, Political Risk and Asset Markets, Fidelity Investments Monograph, Boston: MA.

Harvey, Campbell R., 1991a, The world price of covariance risk, Journal of Finance 46, 111157.

Harvey, Campbell R., 1991b, The term structure and world economic growth, Journal of Fixed Income 1, 417.

Harvey, Campbell R., 1993a, Portfolio enhancement using emerging markets and conditioning information, in Stijn Claessens and Shan Gooptu, Eds., Portfolio investment in developing countries (Washington: The World Bank Discussion Series, 1993, pp. 110144).

Harvey, Campbell R., 1995a, Predictable risk and returns in emerging markets, Review of Financial Studies 8, 773816.

Harvey, Campbell R., 1995b, The risk exposure of emerging markets, World Bank Economic Review, 9, 1950.

Harvey, Campbell R., 1995c, The crosssection of volatility and autocorrelation in emerging equity markets, Finanzmarkt und Portfolio Management 9, 1234.

Harvey, C. R., and A. Siddique, 1995, Conditional skewness in asset pricing tests, Working paper, Fuqua School of Business, Duke University.

Heston, Steven L, K. Geert Rouwenhorst and Roberto E. Wessels, 1995, Capital market integration and the international cost of funds, Journal of Empirical Finance.

Heston, Steven, and K. Geert Rouwenhorst, 1994, Does industrial structure explain the benefits of international diversification, Journal of Financial Economics.

International Finance Corporation, 1993, IFC index methodology, (World Bank, Washington).

Jorion, Philippe, 1991, The pricing of exchange rate risk in the stock market, Journal of Financial and Quantitative Analysis 26, 363376.

Keim, Donald B. and Robert F. Stambaugh, 1986, Predicting returns in the bond and stock market, Journal of Financial Economics 17, 357390.

King, Robert G., and Ross Levine, 1993, Finance, entrepreneurship, and growth: Theory and evidence, Journal of Monetary Economics 32, 513542.

Korajczyk, Robert and Claude Viallet, 1989, An empirical investigation of international asset pricing, Review of Financial Studies 2, 553585.

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, 1337.

Mariscal, Jorge O. And Rafaelina M. Lee, 1993, The valuation of Mexican stocks: An extension of the capital asset pricing model, Goldman Sachs, New York.

Merton, Robert C., 1973, An intertemporal capital asset pricing model, Econometrica 41, 867887.

Mullin, John, 1993, Emerging equity markets in the global economy, Federal Reserve Bank of New York Quarterly Review 18, 5483.

Nelson, D. B., 1991, Conditional heteroskedasticity in asset returns, Econometrica 59, 347-370.

Roll, Richard, 1988, The international crash of October 1987, Financial Analysts Journal 44, 1935.

Roll, Richard, 1992, Industrial structure and the comparative behavior of international stock market indexes, Journal of Finance 47, 342.

Roll, Richard and Stephen A. Ross, 1994, On the crosssectional relation between expected returns and betas, Journal of Finance.

Schmidt, Dana, 1990, Morgan Stanley Capital International (MSCI) indices, (Morgan Stanley, New York, NY).

Schwert, G. William, 1989a, Why does stock market volatility change over time? Journal of Finance 44, 11151154.

Schwert, G. William, 1989b, Business cycles, financial crises and stock volatility, CarnegieRochester Conference Series on Public Policy 31, 83126.

Senbet, Lemma W., Ishac Diwan and Vihang Errunza, 1992, The pricing of country funds and their role in capital mobilization for emerging economies, Working paper, University of Maryland.

Shanken, Jay, 1990, Intertemporal asset pricing: An empirical investigation, Journal of Econometrics 45, 99120.

Sharpe, William, 1964, Capital asset prices: A theory of market equilibrium under conditions of risk, Journal of Finance 19, 425442.

Solnik, Bruno, 1974a, An equilibrium model of the international capital market, Journal of Economic Theory 8, 500524.

Solnik, Bruno, 1974b The international pricing of risk: An empirical investigation of the world capital market structure, Journal of Finance 29, 4854.

Solnik, Bruno, 1977, Testing international asset pricing: Some pessimistic views, Journal of Finance 32 (1977), 503511.

Solnik, Bruno, 1983, The relationship between stock prices and inflationary expectations: The international evidence, Journal of Finance 38, 3548.

Solnik, Bruno, 1983, International arbitrage pricing theory, Journal of Finance 38, 449457.

Solnik, Bruno, 1993, The unconditional performance of international asset allocation strategies using conditioning information, Journal of Empirical Finance.

Speidell, Lawrence S. and Ross Sappenfield, 1992, Global diversification in a shrinking world, Journal of Portfolio Management 19, 5767.

Stehle, Richard, 1977, An empirical test of the alternative hypotheses of national and international pricing of risky assets, Journal of Finance 32, 493502.

Stone, Douglas, 1990, The emerging markets and strategic asset allocation, Working paper, Frank Russell Co., Tacoma, WA.

Stulz, Rene, 1981a, On the effects of barriers to international investment, Journal of Finance 36, 923934.

Stulz, Rene, 1981b, A model of international asset pricing, Journal of Financial Economics 9, 383406.

Stulz, Rene, 1984, Pricing capital assets in an international setting: An introduction, Journal of International Business Studies, Winter, 5573.

Stulz, Rene, 1993, International portfolio choice and asset pricing: An integrative survey, Working paper, Ohio State University, Columbus, OH.