THE WORLD BANK ECONOMIC REVIEW Volume 9 January 1995 Number 1 The Emergence of Equity Investment in Developing Countries: Overview Stiin Claessens The Risk Exposure of Emerging Equity Markets Campbell R. Harvey Emerging Stock Markets and International Asset Pricing Elaine Buckberg Market Integration and Investment Barriers in Emerging Equity Markets Geert Bekaert U.S. Equity Investment in Emerging Stock Markets Linda L. Tesar and Ingrid M. Werner Return Behavior in Emerging Stock Markets Stijn Claessens, Susmita Dasgupta, and Jack Glen Portfolio Capital Flows: Hot or Cold? Stiin Claessens, Michael P. Dooley, and Andrew Warner THE WORLD BANK ECONOMIC REVIEW EDITOR Moshe Svrquin CONSULTING EDITOR Sandra Gain EDITORIAI. BOARD Kaushik Basu, Cornell University and University of Delhi Sebastian Edwards Guillermo Calvo, University of Maryland Gregory K. Ingram Jonathan Eaton, Boston University Mieko Nishimizu Alberto Giovannini, Columbia University John Page Mark R. Rosenzweig, University of Pennsylvania Jacques van der Gaag The Xborld Bank Economic Review is a professional journal for the dissemination of World Bank- sponsored research that informs policy analyses and choices. It is directed to an international readership among economists and social scientists in government, business, and international agencies, as well as in universities and development research institutions. The Reviieu emphasizes policy relevance and opera- tional aspects of economics, rather than primarily theoretical and methodological issues. It is intended for readers familiar with economic theory and analvsis but not necessarily proficient in advanced mathemati- cal or econometric techniques. Articles will illustrate how professional research can shed light on policy choices. Inconsistency with Bank policy will not be grounds for rejection of an article. Articles will be drawn primarily from work conducted bv World Bank staff and consultants. Before being accepted for publication by the Editorial Board, all articles are reviewed by two referees who are not members of the Bank's staff; articles must also be recommended by at least one external member of the Editorial Board. The Review may on occasion publish articles on specified topics by non-Bank contributors. Any reader interested in preparing such an article is invited to submit a proposal of not more than two pages in length to the Editor. The views and interpretations in articles published are those of the authors and do not necessarily represent the views and policies of the World Bank or of its Executive Directors or the coUntries thes represent. The World Bank does not guarantee the accuracyi of the data included in this publication and accepts no responsibility whatsoever for any consequences of their use. When maps are used, the bound- aries, denominations, and other information do not imply on the part of the World Bank Group anx judgment on the legal status of anv territorv or the endorsenient or acceptance of such boundaries. Comments or brief notes responding to Revieu articles are welcome and will be considered for publication to the extent that space permits. Please direct all editorial correspondence to the Editor, Th7e World Bank Economnic Review, The World Bank, Washington, D.C. 20433, U.S.A. The WXorld Bank Eco,tomtic Reuieu' is published three times a sear (Janu.iry, Mas. and September) bn the World Bank. Single copies niay be purch.ased at S 1 0.95. Subscription rates are as follosvs: Indipi'dti,als Institutions 1-year subscription US525 USS45 2-sear subscription lJSS46 USS86 3-year subscription US565 USS125 Orders should be sent to: World Bank Publications, Box 7247-7956, Philadelphia. PA 19170(-7956 U.S.A. Subscriptions are available without charge to readers svith mailing addresses in developing coun- tries and in socialist economies in transition. Written requLest is required every two sears to renewy such subscriptions. >- 1994 The International Bank for Reconstruction and Development THL WaiORI D BANK All rights reserved Manufactured in the United States of America ISSN 0258-6770 Material in this journal is copyrighted. The World Bank encourages dissemination of its wvork and w ill normally give permission promptly and, when the intended reproduction is for noncommercial purposes, without asking a fee. Permission tI) make photocopies is granted through the Copyright Clearance Center, 27 Congress Street, Salem, MA 0)1970 U.S.A. This journal is indexed regularlI in Current Contents Sozal er Bebu'ior.al .Scienccs, lIzdex to Intern,i- tional Statistics, Journal oJ Economic Literature, Public -Altlirs Inorination Service, and Social S ien es Citation Index3. It is available in microform through University Microfilms, Inc., 30)) North Zeeb Road, Ann Arbor, Michigan 4810)6, U.S.A. THE WORLD BANK ECONOMIC REVIEW Volume 9 January 1995 Number 1 The Emergence of Equity Investment in Developing 1 Countries: Overview Stijn Claessens The Risk Exposure of Emerging Equity Markets 19 Campbell R. Harvey Emerging Stock Markets and International Asset Pricing 51 Elaine Buckberg Market Integration and Investment Barriers in Emerging 75 Equity Markets Geert Bekaert U.S. Equity Investment in Emerging Stock Markets 109 Linda L. Tesar and Ingrid M. Werner Retuirn Behavior in Emerging Stock Markets 131 Stijn Claessens, Susmita Dasgupta, and Jack Glen Portfolio Capital Flows: Hot or Cold? 153 Stijn Claessens, Michael P. Dooley, and Andrew Warner THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1: 1-17 The Emergence of Equity Investment in Developing Countries: Overview Stijn Claessens Equity flows to developing countries have increased sharply in recent years. Foreign equity investment can be beneficial to developing countries because of its risk-sharing characteristics and effects on resource mobilization and allocation. Empirical evidence shows that the stock markets of developing countries have become more, although not fully, integrated with world financial markets, and this increased integration implies a lower risk-adjusted cost of capital. Constraints to further increasing the flows and expanding the benefits are macroinstability, poorly functioning stock markets, and insufficiently open financial markets. Empirical evidence does not support the view that equityflows are more volatile than other types of capitalflows or that equityflows have a negative impact on the volatility of stock prices. World financial markets in recent years have been characterized by trends to- ward increased integration, securitization, and liberalization. Financial markets today show a much higher degree of integration, with large amounts of capital flowing across borders to take advantage of the slightest perceived financial or diversification benefit. Gross capital flows among industrial countries are also much larger now than a decade ago. Gross capital outflows from the main industrial countries came to about $850 billion in 1993, compared with an average of about $500 billion during 1985-93 and about $100 billion in the first half of the 1980s (Bank for International Settlements 1993/1994, p. 148). Much of this increased cross-border flow has been in the form of easily tradable securities-bonds, equities, and other negotiable instruments. All of this has happened against a background of increased liberalization of domestic financial markets and an opening up of markets to foreigners as capital controls and other barriers have been removed, especially in developing countries. Buttressed by improved domestic policies and increased economic growth, developing countries have shared in these trends. Total net capital flows (in real Stiin Claessens is with the Technical Department of the Europe and Central Asia and Middle East and North Africa Regions at the World Bank. He thanks Geert Bekaert, Elaine Buckberg, Ash Demirguc- Kunt, Campbell Harvey, Leonardo Hernandez, and Andrew Warner for useful comments. The article draws partly on joint work with Sudarshan Gooptu. Part of this research was funded by World Bank research grants RPO 678-01 and 679-94. ( 1995 The International Bank for Reconstruction and Development/THE WORLD BANK 1 2 THE WORLD BANK ECONOMIC REVIEW, VOL.9. NO. I terms) to all developing countries have in recent years reached their highest level since the debt crisis of the early 1980s (World Bank 1993). In particular, private capital flows have sharply increased in the 1990s and are now about 50 percent higher than they were at their peak in the early 1980s. In comparison with the late 1970s, when there were also large private flows, there has been a shift among private flows from bank to nonbank sources through increased direct and portfolio investment.' Direct investment flows to developing countries have been increasing at very high rates since the mid-1980s. More recently, portfolio flows (bonds, certificates of deposit, commercial paper, and equity) to develop- ing countries have increased sharply in magnitude, especially to the so-called emerging markets.2 Equity portfolio flows have been an important component of these portfolio flows. The large amount of portfolio flows to several developing countries has raised a number of policy and research questions. What benefits does an investor in an industrial country gain from investing in these markets? How well are these developing countries financially integrated with the markets of industrial coun- tries? How has this changed over time? To what extent is financial. integration a function of changes in barriers to free capital movements? What exactly are these barriers (in developing countries as well as industrial countries)? Other questions have also been raised: What is the volume of financing that can be expected from portfolio flows in the coming years, to which countries will the flows go, and from which investors will the flows come? To what extent are these flows a function of factors in the developing countries and factors in the industrial countries? Are these flows volatile ("hot money"), and do they perhaps call for some form of public action? Do the stock markets in these countries price securities efficiently? What is the broader role of stock markets in resource allocation and managerial control? What is the relation between stock market development and domestic economic performance? These are the types of questions addressed in this issue, which includes six articles that were presented at a World Bank Conference on Portfolio Investment in Developing Countries in September 1993 (for the conference proceedings, see Claessens and Gooptu 1993). This overview presents the context for and sum- marizes the main results of the six articles. Where useful, the overview provides a short summary of relevant literature and the results of related papers. It should be kept in mind that this research presents the first wave of serious research on this topic. Many questions are not answered exclusively; more experience and 1. As a percentage of exports or of the gross domestic product (GDP) of developing countries, total aggregate resource flows in the 1990s, although much above their level in the mid-1980s, are still below their level in the mid- to late 1 970s. Compared with that in the 1 970s, the recent increase in flows reflects more a change in composition than an increase in overall net flows. In the 1970s, most financing was from commercial banks (to governments), and equity investment (both portfolio and direct) was a much smaller percentage than today. 2. Emerging markets are commonly considered to be those markets represented in tlle IFC's Emerging Markets Data Base (EMDB), currently twenty-five markets. Claessens 3 research on this important aspect of international capital markets will be neces- sary in the coming decade. Section I provides a short overview of the amounts of and motivation for equity portfolio flows. Section II provides an overview of different approaches to testing for benefits of equity investment in developing countries from an investor point of view and summarizes the most important empirical findings. Section III discusses the role of stock markets and the benefits from equity investment for developing countries. Section IV discusses the various policy issues for developing countries that can arise from increased equity (and other portfolio) flows. Section V summarizes the findings. I. THE AMOUNTS OF AND MOTIVATION FOR EQUITY PORTFOLIO FLOWS From 1989 to 1993 total portfolio flows (bonds, certificates of deposit, com- mercial paper, and equity) increased more than sevenfold to a level of $55.8 billion (see table 1 and Gooptu 1993). Portfolio flows now account for about a third of overall net resource flows to developing countries. Equity flows have been an important component of these portfolio flows. Total equity flows to developing countries were $13.2 billion in 1993, quadruple that of three years earlier. Even though small for developing countries on aggregate (about 7 per- cent of the aggregate net resource flows they received in 1993), equity flows are an important source of external finance for some developing countries (for example, equity flows represented about a quarter of the total net external financing of Mexico in 1991 and 1992). Equity flows have taken several forms: direct equity purchases by investors in the host stock markets; investments through country funds; issues of rights on equities held by depository institutions in the form of American Depository Receipts (ADRS) and Global Depository Receipts (GDRs);3 and direct foreign equity offerings. In the early 1990s equity flows have largely taken place through depository receipts. For 1989-93 the volume of ADRS and GDRS issued for equity claims on developing countries is estimated to have been about $18.2 billion (table 1), including direct offerings on foreign capital markets by corpo- rations in developing countries outside the ADR or GDR structure (under Rule 144A in the United States). Until 1990 most important were (closed-end) coun- try funds: during 1989-93 new country funds were created for developing coun- tries with an aggregated size of $10.3 billion. The highest relative increase in the early 1990s has been in direct purchases of equities: these are estimated to have been about $3.2 billion in 1993, up from $0.8 billion in 1990, and were second in importance during 1989-93. 3. ADRs and GDRS are receipts issued by financial intermediaries in industrial countries against shares held in custody by these intermediaries in the developing countries. 4 THE WORID BANK ECONOMIC REVIEW, VOL. 9, NO. I Table 1. Portfolio Flows to Developing Countries, 1989-93 (estimates in billions of U.S. dollars) Total Type offlow 1989 1990 1991 1992 1993a 1989-93,, Bonds, commercial paper, and certificatesof deposit 4.0 5.5 12.7 23.7 42.6 88.5 Equity 3.5 3.8 7.6 13.0 13.2 41.1 New country funds 2.2 2.9 1.2 1.3 2.7 10.3 American and global deposit receipts 0.0 0.1 4.9 5.9 7'.3 18.2 Direct equity 1.3 0.8 1.5 5.8 3.2 12.6 Total 7.5 9.3 20.3 36.7 55.8 129.6 a. Estimated for 1993. Source: World Bank (1993). Various factors in both developing and industrial countries have played a role in the increased importance of equity flows in recent years. An important factor has likely been the decline in global interest rates in the early 1990s, which was an important "push" for equity flows (Calvo, Leiderman, and Reinhart 1993; Dooley, Fernandez-Arias, and Kletzer 1994; Fernandez-Arias 1994). But there is no doubt that the increase in these flows is partly also motivated by improved domestic policies and better growth performance in the recipient countries, reflected in higher rates of return on equity (Chuhan, Claessens, and Mamingi 1993). Another contributing factor has been the removal of barriers by developing and industrial countries on foreign participation in developing countries' stock markets (Claessens and Rhee 1994). Many developing countries have removed restrictions on foreign ownership, liberalized capital account transactions, im- proved their accounting and information standards, and in general made it easier for foreigners to access their markets. Particularly in Europe and Latin America, many countries now have very few or no restrictions on access by foreigners to their markets and treat foreign and domestic investors almost identically (see IFC various years). Similarly, several industrial countries have removed or relaxed their restrictions on investments in developing countries (Chuhan 1994). 11. BENEFITS FOR INVESTORS The benefits for an investor of equity investment in emerging markets ulti- mately depend on a tradeoff between the expected rate of return and its associ- ated risk. To assess this tradeoff a number of factors are important: the underly- ing factors driving the rate of return and its variability; the efficiency of the domestic stock market; the regulatory, accounting, and enforcement standards in the host country; the ability to invest in the country; the diffejrent forms of transfer risk (for example, the possibility that capital controls will be imposed, affecting the ability to repatriate capital out of the host country); taxes and other Claessens S transaction costs; and restrictions and regulatory and accounting standards im- posed on investors in the home country (for example, restrictions on the share of foreign assets held by pension funds). Returns on emerging stock markets have been high. For example, the annual U.S. dollar rate of return on the IFC composite index for Latin America was 38.88 percent during 1988-93, compared with 14.40 percent for the U.S. Stan- dards and Poor 500. The volatility of U.S. dollar rates of return has, however, also been high. For example, it was as high as 100 percent on an annual basis for Argentina during the same period. The high ex post rates of return and high volatility already suggest a tradeoff from the investor's point of view. The risk-return tradeoff should, however, be investigated from the point of view of an internationally well-diversified investor who is considering investing in emerging markets. Because correlations between equity returns from different countries are lower than those between equity returns in the same country, the benefits of diversification-a lower risk for equivalent return or a higher return for equivalent risk-are stronger across international financial markets than within domestic markets. This is especially true for investments in developing countries, because their stock returns tend to have (even) lower correlations with those of industrial countries. Participation in developing countries is thus likely to lower overall unconditional portfolio risk. This fact is by now well estab- lished. Harvey (1993) and Divecha, Drach, and Stefak (1992), for example, find that by investing up to about 20 percent of an international portfolio in develop- ing countries, the risk-return tradeoff can be Pareto-improved, in which case the unconditional mean-variance frontier shifts dramatically upwards. Harvey (1994) and De Santis (1993) show that this upward shift is not only dramatic but also statistically significant. The important caveat concerning these diversification benefits is whether these gains are attainable. Barriers may restrict investment in these markets. (Another caveat is, of course, transaction costs.) The benefits of diversification relate then also to the degree of capital-market integration. Without barriers, capital markets tend to be fully integrated; financial assets traded in different markets but with identical risk characteristics will yield identical expected re- turns. Alternatively, with barriers, markets tend to be segmented, and assets in different markets may have different expected returns even when their risk characteristics are the same. Ways to Testfor and Measure Stock Market Segmentation There are several ways to test for and measure the degree of stock market segmentation. One way is to explicitly model the barriers, derive the effect on equilibrium asset prices and stock holdings, and then test the model. (See Stulz 1994, for example, for a review of the equilibrium pricing given ownership restrictions, taxes, or other barriers.) This approach is difficult, however, be- cause there are many barriers to consider, and few of them can be easily quan- tified. In any case, using a specific asset-pricing model entails the risk of mis- 6 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I specification, because it is unclear whether rejections should be attributed to the model or to lack of integration. Another way to test for and measure stock market integration is to assume that market integration exists and that a particular asset-pricing model holds. This method also risks misspecification and is hampered by the lack of a well- established international asset-pricing model. The single-factor Capital Asset Pricing Model (CAPM) with one source of risk is often used in a domestic context. This model cannot be applied to international stock market returns because it can describe international stock market returns only if all investors have identi- cal preferences (see Solnik 1974 and the review by Adler and Dumas 1983). Unless purchasing power parity holds-which it does not in the short run- exchange risk will have to enter the tests. Few modern empirical tests have done so, however (recent exceptions are Dumas and Solnik 1994 and Ferson and Harvey 1993). In general, empirical results have been ambiguous in tests of the international CAPM (see Stulz 1994 and Dumas 1994 for reviews). Although a fully satisfying asset-pricing model is lacking, the general direction in empirical tests has been toward more complex, multifactor mociels. Here, risk is measured with respect to the covariances, that is "betas,' of equity returns with various risk factors. Because risk factors need to be prespecified, this approach also runs the risk of misspecification. A related further complication is that recent research has found significant time variation in expected returns, both domestic and international, but no consensus has emerged on what drives this apparent predictability. (A simple, static asset-pricing model would imply no changes over time in the predictability of returns.) It is unclear whether the predictability of returns itself is evidence of market inefficiency, time-varying risk premiums, or infrequent trading of stocks (see Harvey 1995). Without an explicit, dynamic asset-pricing theory, it is im- possible to distinguish between highly variable risk premiums, peso problems, regime switches, knowledge of policy changes, or other inefficiencies. Neverthe- less, it has been found that returns in different industrial countries can be pre- dicted by using a common set of instruments (for example, Harvey 1991). This commonality suggests that industrial countries are relatively well integrated. Thus, a third way to measure integration is to test for any commonality in the factors driving the predictability of returns across countries. A fourth way to measure integration is to look at actual investrnent patterns. Actual portfolios of investors in industrial countries exhibit significant "home bias"; that is, the shares of domestic securities in the portfolios are much higher than one would expect on the basis of risk-return tradeoffs and a reasonable level of risk aversion, and this suggests a lack of integration (French and Poterba 1991; Tesar and Werner forthcoming).4 4. Admittedly, the asset-allocation model used to derive the optimal shares could be misspecified. Possibly, differences in consumption baskets, deviations from purchasing power parity, uninsurable (nontradable) income risk, and borrowing constraints could explain this home bias. French and Poterba Claessens 7 Evidence of Integration The difficulties associated with testing for and measuring stock market inte- gration have motivated some of the authors in this issue (Bekaert, Buckberg, Harvey, and Tesar and Werner) to pursue various approaches (for related work see De Santis 1993 and Bekaert and Harvey 1994). Bekaert and Harvey use the world portfolio as a benchmark for measuring risk. They report that an unconditional, single-factor CAPM is unable to charac- terize returns in emerging markets. This result confirms-actually as a corollary-the unconditional mean-variance diversification benefits already mentioned. Tests for diversification benefits and lack of integration are identical if one (global) source of risk is assumed. Deviations from integration amount, then, to unexploited (mean-variance) diversification benefits and diversification benefits imply lack of integration. Bekaert observes, however, that the slope coefficient of the country return on the world portfolio return (beta) has increased for most emerging markets in recent years. He interprets these higher betas as signs of increased integration, because most industrial countries have high betas. Buckberg goes a step further by testing a conditional, single-factor CAPM in which expected returns are al- lowed to change over time. In contrast to Bekaert and Harvey, Buckberg cannot reject that emerging equity markets were integrated in more recent years, whereas she could reject it for the earlier period, and this suggests that the benefits of further diversifying into emerging markets have been reduced. How- ever, Harvey questions the power of the test Buckberg uses. Harvey (1994) allows both expected returns as well as covariance risk (beta) to change over time and finds that this more general model is rejected. Harvey tests multifactor models and finds significant evidence that global risk factors are not powerful in explaining returns in emerging markets, especially compared with explaining returns in industrial countries. His evidence is con- sisterit with emerging markets' being segmented from industrial countries. He finds, however, that compared with earlier periods, the importance of global factors for many emerging markets has increased, suggesting greater, but still imperfect, integration. Bekaert investigates the degree of predictability of rates of return in emerging markets using both global and domestic variables (see also Buckberg 1993 and Harvey 1994). He finds that rates of return in the emerging markets are more predictable than in industrial countries, mostly because of the higher autocor- relation of returns in emerging countries. Bekaert finds that global instruments (1991) and Tesar and Werner (forthcoming) provide some further explanations. Cooper and Kaplanis (1991) provide some evidence, however, that the magnitude of the home bias cannot be explained by differences in consumption baskets or deviations from purchasing power parity alone. Purchasing power parity deviations combined with deadweight costs of a few percent a year can, however, generate the observed home bias. Tesar and Werner (forthcoming) document that turnover ratios of foreign equities appear to be higher than those of domestic stocks, contradicting an explanation of home bias based on transaction costs. 8 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I are less important for developing countries than for industrial countries. (Claessens, Dasgupta, and Glen; Buckberg; and Harvey also report this higher autocorrelation.) Bekaert also observes, however, that global predictability has declined (com- paring 1985-92 with the pre-1985 period). This complicates the interpretation of predictability through common factors as an indicator of market integration, and no clear pattern emerges on how predictability for individual emerging markets changes over time. Bekaert then also concludes that predictability per se does not yield much useful information about market segmentation. Instead, he calculates for each country the correlation between the fitted values of his expected-return equation for the United States with those for the other coun- tries. This way he can control for the apparent decline in global predictability. He interprets this correlation as a measure of the degree of market integration across countries. (In a one-factor model, and if markets were perfectly inte- grated, expected returns would be perfectly correlated across markets.) Com- paring the correlations for the two periods, he finds evidence of increased inte- gration for most industrial countries and many emerging markets after 1985. Tesar and Werner show that there is a significant "home bias" for developing countries. The cumulative inflow of foreign equity investments represents only a small fraction of stock market capitalization in the emerging markets and only a minute fraction of stock market capitalization in all industrial countries, much below any "optimal" share. Again, turnover ratios for foreign investment in emerging markets appear to be roughly of the same magnitude as, or somewhat lower than, those in the United States. Tesar and Werner document, however, that the share of recent U.S. outward equity investment going to emerging markets in total U.S. outward equity investment is in line with the share of the market capitalization of emerging markets in global capitalization. This suggests that at the margin the home bias has more recently disappeared. In summary, most evidence based on asset prices suggests that emerging mar- kets until the mid-1980s were not integrated with world financial markets but are now increasingly becoming integrated. Actual investment also provides evi- dence that emerging markets are becoming de facto integrated. III. THE ROLE OF STOCK MARKETS AND BENEFITS FROM EQUITY INVESTMENT FOR DEVELOPING COUNTRIES Efforts to improve the functioning of financial markets of developing coun- tries, that is, to allocate capital more efficiently, have often focused on core financial themes such as interest liberalization, smaller government role in credit allocation, and improvement in the role of banks as financial intermediaries. (For a survey on the role of capital markets and the relation between stock market development and the functioning of financial intermediaries, see Demirguc-Kunt and Levine 1993.) Recently, capital markets in general and stock markets in particular have received increased attention from policy- Claessens 9 makers. Here I review the principal roles that stock markets can perform and relate some of the evidence the symposium articles provide of these roles. The Role of Stock Markets First, stock markets can be a vehicle for raising capital for firms. Although this is true for any other form of financial intermediation, stock markets may take on a larger role in developing countries where privatization and a greater role for the private sector imply a large demand for equity finance. Second, capital markets in general, and equity markets in particular, can enable investors to diversify their wealth across a variety of assets, usually more easily than in most other financial markets. Thus, capital markets reduce the risk that investors must bear, thereby reducing the risk premium demanded and the cost of capital. The benefits of a lower risk premium can be particularly large in the case of foreign equity investment, because foreign investors are more diversified. Third, stock markets can perform a screening and monitoring role. Relying on the information and judgment of numerous participants, stock prices quickly reflect changes in underlying values and indicate profitable investment oppor- tunities, thus providing a screening and selecting function. Stock markets- through continuous adjustment of stock prices-can also assist in monitoring managers of publicly traded corporations, thereby possibly improving corporate governance. Foreign investment may be particularly useful because it introduces international practices and cross-country experiences. This monitoring role will take on increased importance as the number of publicly traded companies in developing countries increases. Fourth, a financial system that functions well requires that the whole financial sector function efficiently. There can be important relations among the various financial institutions and complementarities from the point of view of capital suppliers and demanders. For example, the absence of a well-functioning stock market may limit the ability of firms to achieve an efficient mix of debt and equity, in spite of a well-functioning debt market. In this sense, stock markets and other financial intermediaries may function as complements, rather than substitutes, and a stock market that functions well may have positive exter- nalities for the rest of the financial system. These benefits of stock markets have to be balanced, however, against certain costs. Do stock markets perform these roles more efficiently than other financial intermediaries? This question becomes more important when there are some specific costs and concerns associated with stock markets that are not associated with other financial markets. Critics often claim that stock markets are nothing but casinos and that stock prices are not related to fundamentals. In general, stock markets in developing countries are often associated with volatility. Many policymakers are concerned that foreign flows may lead to excessive price move- ments and general macroeconomic instability (for example, real exchange rate 10 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I variability). Thus some argue that stock markets contribute little to economic efficiency and may even be welfare-decreasing (for example, Stiglitz 1991). Evidence of Benefits The large foreign inflows in the past few years (table 1) document convin- cingly the ability of stock markets to raise foreign savings. Whether stock mar- kets increase the overall mobilization of domestic resources-or are simply an- other vehicle for channeling the same amount of resources-is as of yet unclear. The experience of industrial countries suggests that in mature markets the net contribution of stock markets to firms' financing needs is small (see, for exam- ple, Mayer 1989). To date, firms in developing countries have relied to a greater extent on stock markets for their external finance than firms in industrial coun- tries currently do (see Singh and others 1992 and Stiglitz 1993, table 1, p. 22). Whether this higher mobilization of resources will be sustained remains to be seen. Concerning the risk-adjusted cost of capital, the benefits for developing coun- tries are the mirror image of the benefits for an industrial-country investor. The test results reported by Bekaert, Buckberg, and Harvey show increased integra- tion and thus indicate that the risk-adjusted cost of capital in emerging markets has moved more in line with that in industrial markets. This lower cost of capital associated with increased integration is also backed up by studies of individual securities' offerings. Tandon (1994) shows, for example, that offering bonds on the international markets leads to a reduction in the required rate of return of the same firm's equity. He finds a similar effect for the introduction of a country fund. In general, the increased foreign equity flows are part of a process that allocates global savings to the most productive investment, and this process leads not only to direct cost savings but also to higher investment and growth. The static benefits of increased portfolio equity flows-the inflow of foreign resources and the lower risk-adjusted cost of capital-can thus be documented. The dynamic benefits of improved screening and monitoring and of externalities are more difficult to document. To some extent, there is an analogy here with the benefits of increased integration in trade. As has now been well documented, the static gains of increased trade are generally thought to be relatively small. Yet, much evidence shows that increased openness to trade and liberalization of prices is associated with increased overall economic growth, which suggests large dynamic gains. The exact channels through which these gains are realized are unclear, however. A similar situation may exist with respect to foreign equity flows. Foreign investment-therewith bringing domestic asset prices in line with foreign prices-can lead to an improvement in the overall functioning of domestic finan- cial markets and as a result lead to indirect economic efficiency and welfare gains. Although there is anecdotal evidence of these dynamic gains, evidence specific to foreign equity flows is scarce. Claessens 1 1 Several articles in this issue provide evidence that emerging markets are not yet efficient, at least when compared with the (more thoroughly studied) mar- kets of industrial countries. Return behavior, both on a cross-section and a time- series basis, displays some patterns that raise questions about the effectiveness of the asset-allocation role played by the stock markets in the countries examined. For example, Claessens, Dasgupta, and Glen; Harvey; Buckberg; and Bekaert report that rates of return in many emerging markets have significant positive first-order autocorrelation, which indicates return predictability and possible inefficiencies.5 In industrial countries, by contrast, autocorrelations are gener- ally insignificant. At the same time, Claessens, Dasgupta, and Glen document that emerging markets display few of the seasonal and cross-sectional anomalies found for industrial countries. Whether this means that the often institution- based explanations used for the industrial countries do not transpose themselves to emerging markets or that the emerging markets are less or more efficient is still unclear at this point, however. This evidence suggests that the emerging stock markets can still improve their functioning. Foreign capital can play a useful role in this respect as it may speed up domestic competition and stimulate innovation, thus leading to dynamic gains. Bekaert, for example, finds some evidence that opening up markets im- proves market efficiency. Diwan, Errunza, and Senbet (1993) show that country funds, despite their small size, can contribute to pricing efficiency and domestic resource mobilization. Just as with trade in goods, countries may also find it more efficient to import some financial services rather than produce them domestically. Relying on for- eign financial intermediaries and institutions may allow the country to avoid the cost of setting up expensive domestic financial infrastructure and to benefit from economies of scope and scale in international financial markets. The use of ADRS and direct listings on foreign stock exchanges are good examples of this "impor- tation" of foreign financial skills, which can lead to cost savings. Eun, Claessens, and Jun (forthcoming) show that for a sample of Australian firms, for example, issuing an ADR reduced the cost of capital for the firm and possibly also for other firms in the same market. Bekaert (see next section) finds that a higher number of foreign-listed securities is associated with a higher degree of financial integration. IV. POLICY ISSUES FOR DEVELOPING COUNTRIES: BARRIERS AND VOLATILITY Two major policy issues for developing countries with emerging stock markets are barriers to integration and the volatility of equity (and other portfolio) flows. 5. Motivation for various types of time-series tests and results of actual tests for industrial countries can be found in Fama (1970, 1991). Harvey and Claessens, Dasgupta, and Glen also report that many of the emerging-market returns depart from normality. 12 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Barriers The tests showing increased but still incomplete integration of emerging stock markets over time indicate that for both investors and firms in developing coun- tries there still remain some unexploited gains from increased foreign invest- ment. Whether these gains are attainable depends on whether barriers can be overcome or removed. Some barriers may be legal and may therefore be re- moved, if so chosen, easily; others are de facto and may limit foreigners' access to these markets for some time to come. Which are the most important barriers, and how can they be overcome or removed? Bekaert identifies several barriers to equity flows in recipient countries and examines which barriers are effective by relating them to several measures of market integration. First, he finds that (macroeconomic) instability in the recip- ient country can be an important detriment: poor credit ratings, high and vari- able inflation, and exchange rate controls are barriers that have a high rank- correlation with measures of lack of market integration. Second, the degree to which the domestic stock market is developed is an important faciLor: lack of a high-quality regulatory and accounting framework and the limited size of some stock markets are associated with a lower degree of integration. Claessens, Dasgupta, and Glen make the related point that emerging stock markets that function inefficiently may arise from exclusive access to information by certain firms or individuals-"insider trading." This would imply that the market will be stacked against "outsiders" and will be less likely to attract new investors, either foreign or domestic. Consequently, improvement in efficiency may in itself re- duce a barrier. Finally, Bekaert finds that the de facto openness of the stock market matters: a lack of sufficient country funds or cross-listed securities or both is associated with lower integration. Surprisingly, Bekaert finds that formal barriers in the form of ownership restrictions seem to matter little, which sug- gests that these are not binding or are circumvented. Hence, to further benefit, countries need to lower barriers. Some of these barriers obviously cannot be removed overnight. For example, a poor credit rating will remain a constraint on portfolio flows for some developing countries in the near future. But some other barriers may be easily removed. One conclu- sion from Bekaert's work is that countries wishing to attract portfolio flows would do best to list country funds and other securities internationally. Improv- ing the way the stock market functions-investors often mention a malfunction- ing stock market as a barrier-requires adopting improved investor protection safeguards, including disclosure requirements, accounting standards, and cus- tody and settlement procedures. These safeguards can be adopted relatively easily, especially when relying on self-enforcing regulators, such as exchange houses and investment boards. Another example of a barrier that is relatively easy to remove is explored by Demirguc-Kunt and Huizinga (forthcoming). They argue that by harmonizing the taxation of capital gains and dividends with the taxation of capital gains and dividends in industrial countries, emerging Claessens 13 markets can enhance the effective returns to foreigners and lower their cost of capital. Ultimately, investors' perceptions and attitudes matter, too, and this makes it difficult to predict how any change in a given barrier will affect inflows. It is clear, however, that removing many of these barriers would be beneficial. Volatility For many developing countries that receive large portfolio flows, an impor- tant policy issue has been whether portfolio flows in general and equity flows in particular are volatile and potentially destabilizing to financial markets and the economy. Conventional wisdom is that short-term flows, and portfolio flows to developing countries in particular, are inherently unstable (see, for example, Reisen 1993). Many developing countries have actively tried to discourage short-term flows (through quantity constraints, taxes, or other instruments) or triecl to encourage longer-term flows (for example, through subsidized foreign direct investment). Claessens, Dooley, and Warner take issue with the conventional view that short-term flows are inherently more unstable and thus may require some policy action. Using data for five developing countries and five industrial countries, they show that there are no significant differences between the times-series properties of short-term flows and long-term flows. Put differently, if only time- series (statistics) are used, it is not possible to tell the label of the flow. Claessens, Dooley, and Warner also show that because there is much substitu- tion between the various flows, only an analysis of the aggregate capital account is meaningful. This implies that any capital control program or other policy (including subsidies and taxes) aimed at discouraging a particular type of flow because of its (alleged) volatile behavior may be misguided or ineffectual. Ap- propriate aggregate macropolicies aimed at achieving the desired overall capital account behavior are likely more effective in dealing with volatile flows. This is confirmed by Corbo and Hernandez (1994), who review the experience with capital inflows for nine countries and evaluate the various mechanisms used to manage these inflows. They conclude that aggregate demand measures (fiscal contractions) have been the most effective. Tesar and Werner corroborate these findings. They investigate the time-series behavior of U.S. equity flows to emerging markets. They find little indication that countries with high U.S. investment activity have high rates of turnover (the volume of equity traded in relation to the local market capitalization). They also find no relation between U.S. flows and the volatility of stock returns. Kim and Singal (1993) and De Santis and Imrohoroglu (1994) investigate this issue as well. They both study the behavior of stock prices following the opening of a market to foreigners or large foreign inflows. They find that there is no systema- tic effect of liberalization on stock market volatility. These findings corroborate Bekaert's finding that volatility in emerging markets is unrelated to his measure of openness. Therefore the fear that foreign-market access will lead to more volatile domestic markets might be ill-founded. 14 THE WORLD BANK ECONO-MIC REVIEW, VOL. 9, NO. I V. CONCLUSION Equity portfolio flows can benefit developing countries by diversifying the sources of external finance, increasing the risk-bearing by investors, reducing the cost of capital, improving incentives for managing the investrnent process, assisting in the development of domestic capital markets, and enhancing the mobilization of domestic resources. Empirical evidence to date conEirms some of these benefits. Several studies have found that emerging markets were not well integrated until the early 1980s. The corollary to the lack of integration has been that these markets have provided attractive investment and diversification op- portunities for investors in industrial countries, which has likely motivated the larger inflows of foreign equity during the past few years. Now, as a result of increased equity flows and opening up, it is found that emerging markets are increasingly integrated with world financial markets, and this has brought their costs of capital more in line with those in world markets. To further reap these benefits, developing countries should (continue to) lower barriers to foreign equity flows. The most important barriers appear to be instability, under- developed stock markets, and a de facto lack of openness. Policy measures that can help remove these barriers are-in addition to proper fiscal and monetary policies-a solid regulatory and accounting framework, investor protection, and less restrictions on foreign ownership. There is no empirical evidence to support the view that portfolio flows are more volatile than other types of capital flows or that they have a negative impact on stock price volatility. Therefore, the conventional wisdom that short- term flows, and portfolio flows to developing countries in particular, are inher- ently unstable or destabilizing may be unfounded. The research task ahead is still large, nevertheless. Important questions re- main on the degree of capital-market integration, especially as it relates to different kinds of financial instruments (for example, are country funds better or worse than ADRS in this respect, and how about direct equity purchases?). Also, little is known on the specific benefits of foreign equity flows for improving corporate governance, and on the links between stock market development and the functioning of the whole financial system. REFEREN CES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Adler, Michael, and Bernard Dumas. 1983. "International Portfolio Choice and Corpo- ration Finance: A Synthesis. Journal of Finance 38(3):925-84. Bank for International Settlements. 1993 /1994. Annual Report. Basel, Switzerland. Bekaert, Geert. 1995. "Market Integration and Investment Barriers in Emerging Equity Markets." The World Bank Economic Review 9(1):75-107. Bekaert, Geert, and Campbell Harvey. 1994. "Time-varying World Market Integration." NBER Working Paper 4843. National Bureau of Economic Research, Cambridge, Mass. Claessens 15 Buckberg, Elaine. 1993. "Emerging Stock Markets and International Asset Pricing'" In Stijn Claessens and Sudarshan Gooptu, eds., Portfolio Investment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. . 1995. "Emerging Stock Markets and International Asset Pricing." The World Bank Economic Review 9(1):51-74. Calvo, Guillermo, Leonardo Leiderman, and Carmen Reinhart. 1993. "Capital Inflows and the Real Exchange Rate Appreciation in Latin America: The Role of External Factors?' IMF Staff Papers 40(1):108-51. Chuhan, Punam. 1994. "Are Institutional Investors an Important Source of Portfolio Investment in Emerging Markets?" wps 1243. World Bank, International Economics Department, Washington, D.C. Processed. Chuhan, Punam, Stijn Claessens, and Nlandu Mamingi. 1993. "Equity and Bond Flows to Latin America and Asia: The Role of Global and Country Factors." wps 1160. World Bank, International Economics Department, Washington, D.C. Processed. Claessens, Stijn, and Sudarshan Gooptu, eds. 1993. Portfolio Investment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C.: World Bank. Claessens, Stijn, and Moon-Whoan Rhee. 1994. "The Effect of Equity Barriers on Foreign Investment in Developing Countries." In Jeffrey Frankel, ed., Internationaliza- tion of Equity Markets. Chicago, Ill.: University of Chicago Press. Claessens, Stijn, Susmita Dasgupta, and Jack Glen. 1995. "Return Behavior in Emerging Stock Markets." The World Bank Economic Review 9(1):131-52. Claessens, Stijn, Michael Dooley, and Andrew Warner. 1995. "Portfolio Flows: Hot or Cold:' The World Bank Economic Review 9(1):153-74. Cooper, Ian, and Evi Kaplanis. 1991. "What Explains the Home Bias in Portfolio Investment?" Discussion Paper IFA-136-91. London Business School, Institute of Fi- nance and Accounting, London. Processed. Corbo, Vittorio, and Leonardo Hernandez. 1994. "Macroeconomic Adjustment to Cap- ital Inflows: Latin American Style versus East Asian Style." wps 1377. World Bank, International Economics Department, Washington, D.C. Processed. Demirgiu-Kunt, Asli, and Ross Levine. 1993. "Stock Market Development and Finan- cial Intermediary Growth.' wps 1159. World Bank, Policy Research Department, Washington, D.C. Processed. Demirgic-Kunt, Asli, and Harry Huizinga. Forthcoming. "Barriers to Portfolio Invest- ments in Emerging Stock Markets." Journal of Development Economics. De Santis, Giorgio. 1993. "Asset Pricing and Portfolio Diversification: Evidence from Emerging Financial Markets." In Stijn Claessens and Sudarshan Gooptu, eds., Portfo- lio Investment in Developing Countries. World Bank Discussion Paper 228. Washing- ton, D.C. De Santis, Giorgio, and Selahattin Imrohoroglu. 1994. "Stock Returns and Volatility in Emerging Financial Markets." University of Southern California, Department of Fi- nance and Business Economics, Los Angeles, Calif. Processed. Divecha, Arjun, Jaime Drach, and Dan Stefak. 1992. "Emerging Markets: A Quantita- tive Perspective." Journal of Portfolio Management 19(Fall):41-50. Diwan, Ishac, Vlhang Errunza, and Lemma Senbet. 1993. "Country Funds for Emerging Economies'" In Stijn Claessens and Sudarshan Gooptu, eds., Portfolio Investment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. 16 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Dooley, Michael, Eduardo Fernandez-Arias, and Kenneth Kletzer. 1994. "Is the Debt Crisis History? Recent Private Capital Inflows to Developing Countries." wps 1327. World Bank, International Economics Department, Washington, D.C. Processed. Dumas, Bernard. 1994. "Partial-Equilibrium versus General-Equilibrium Models of the International Capital Market." In Frederick van der Ploeg, ed., Handbook of Interna- tional Economics. Oxford, U.K.: Basil Blackwell. Dumas, Bernard, and Bruno Solnik. 1993. "The World Price of Exchange Rate Risk." NBER Working Paper 4459. National Bureau of Economic Research, Cambridge, Mass. Eun, Cheol, Stijn Claessens, and Kwang Jun. Forthcoming. "International Trade of Assets, Pricing Externalities, and the Cost of Capital." Pacific-Basin Finance Journal 3(1). Fama, Eugene. 1970. "Efficient Capital Markets: A Review of Theory and Empirical Work." Journal of Finance 25(2):383-417. . 1991. "Efficient Capital Markets: II." Journal of Finance 46(5):1575-1617. Fernandez-Arias, Eduardo. 1994. "The New Wave of Private Capital Inflows: Push or Pull?" wps 1312. World Bank, International Economics Department, Washington, D.C. Processed. Ferson, Wayne, and Campbell Harvey. 1993. "The Risk and Predictability of Interna- tional Equity Returns'" Review of Financial Studies 6(3):527-66. French, Kenneth R., and James M. Poterba. 1991. "Investor Diversification and Interna- tional Equity Markets." American Economic Review 81(2, May):222-26. Gooptu, Sudarshan. 1993. "Portfolio Investment Flows to Emerging Markets." In Stijn Claessens and Sudarshan Gooptu, eds., Portfolio Investment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. Harvey, Campbell. 1991. "The World Price of Covariance Risks." Journal of Finance 46(4):111-57. . 1993. "Portfolio Enhancement Using Emerging Markets and Conditioning In- formation." In Stiin Claessens and Sudarshan Gooptu, eds., Portfolio Investment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. 1994. "Predictable Risk and Returns in Emerging Markets." Working Paper. Duke University, Fuqua School of Business, Durham, N.C. Processed. . 1995. "The Risk Exposure of Emerging Equity Markets." The World Bank Economic Review 9(1):19-50. IFC (International Finance Corporation). Various years. Emerging Stock Markets Fact- book. Washington, D.C. Kim, E. Han, and Vijay Singal. 1993. "Opening Up of Stock Markets by Emerging Economies: Effect on Portfolio Flows and Volatility of Stock Prices." In Stijn Claessens and Sudarshan Gooptu, eds., Portfolio Investment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. Mayer, Colin. 1989. "Myths of the West: Lesson from Developed Countries for Develop- ment Finance." wps 301. World Bank, Office of the Vice President, Development Economics, Washington, D.C. Processed. Reisen, H. 1993. "The Case for Sterilized Intervention in Latin America." Paper pre- sented at the 6th Annual Inter-American Seminar on Economics, May 28-29, Cara- cas, Venezuela. Claessens 1 7 Singh, Ajit, Javed Hamid, Babram Salimi, and Yoicki Nakano. 1992. "Corporate Finan- cial Structures in Developing Countries." Technical Paper No. 1. International Finance Corporation, Washington, D.C. Processed. Solnik, Bruno. 1974. "An Equilibrium Model of the International Capital Markets." Journal of Economic Theory 8(4):500-24. Stiglitz, Joseph. 1991. "Government, Financial Markets, and Economic Development." NBER Working Paper 3669. National Bureau of Economic Research, Cambridge, Mass. .1993. "The Role of the State in Financial Markets." In Proceedings of the World Bank Annual Conference on Development Economics 1993. Washington, D.C.: World Bank. Stulz, Rene M. 1994. "International Portfolio Choice and Asset Pricing: An Integrative Survey." NBER Working Paper 4645. National Bureau of Economic Research. Cam- bridge, Mass. Tandon, Kishore. 1994. "External Financing in Emerging Economies: An Analysis of Market Responses." World Bank, International Economics Department, Washington, D.C. Processed. Tesar, Linda, and Ingrid Werner. 1995. "U.S. Equity Investment in Emerging Stock Markets." The World Bank Economic Review 9(1):109-30. . Forthcoming. "Home Bias and High Turnover." Journal of International Money and Finance. World Bank. 1993. World Debt Tables 1993-94. Washington, D.C. THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1: 19-50 The Risk Exposure of Emerging Equity Markets Campbell R. Harvey The low correlation between returns in emerging equity markets and industrial equity markets implies that the global investor would benefit from diversification in emerg- ing markets. This article explores the sensitivity of the emerging-market returns to measures of global economic risk. When these traditional measures of risk are used, the emerging markets have little or no sensitivity. This finding is consistent with these markets' being segmented from world capital markets. However, the correlation between the emerging-market returns and the risk factors appears to be changing over time. New interest in international investing has been partly caused by the emerging equity markets, which are attractive because of their high average returns and low correlations with industrial markets. Little is known, however, about how to measure the risk of investment in emerging markets. The goal of this article is to advance the understanding of the investment risk in emerging markets by measuring each market's exposure to a number of global economic forces. To measure risk in a meaningful way, an asset-pricing model is needed. In the international version of the Capital Asset Pricing Model (CAPM) of Sharpe (1964) and Lintner (1965) investors are presumed to hold a diversified portfo- lio of equities from all national markets, that is, a world market portfolio. In this type of model, the portfolio risk is the variance of this well-diversified portfolio. The risk of an individual security is measured by its contribution to (or covariance with) the world market portfolio. Usually, the covariances are scaled by the variance of the world market portfolio and are called betas. The CAPM predicts that equities with higher covariances (higher risk) will command higher expected returns in equilibrium. Roll (1977) emphasizes that testing the CAPM is equivalent to testing the mean-variance efficiency of the market portfolio. That is, any test that tries to Campbell R. Harvey is with the Fuqua School of Business at Duke University and the National Bureau of Economic Research. This work is based on his presentation at the World Bank Conference on Portfolio Investment in Developing Countries, Washington, D.C., September 9-10, 1993. This research is funded by the World Bank. The author appreciates the support of the Batterymarch Fellowship and a grant from the Center for International Business Education and Research (CIBER) at the Fuqua School of Business, Duke University. Yoram Ehrlich, Chris Kirby, and Akhtar Siddique provided excellent research assis- tance, and two anonymous referees provided useful comments. © 1995 The International Bank for Reconstruction and Development/ THE WORLD BANK 19 20 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I investigate whether higher-risk securities receive higher expected returns is contingent on the market portfolio's being the portfolio with the highest possi- ble expected return for any given level of variance (efficiency). Encouragingly, a number of tests fail to reject the null hypothesis of mean-variance efficiency of the world market portfolio (see, for example, Harvey 1991). However, an important caveat is that these tests consider equities only in industrial markets. The focus of this article is on emerging-market equities. The efficiency issue is important in the following sense. Suppose that a world CAPM is a reasonable approximation of equilibrium. Suppose, however, that a world portfolio is an inefficient benchmark; that is, there exists another portfolio with a higher expected return and the same variance. Roll and Ross (1994) point out that there may be little or no relation between risk and expected return in this case. As a result, although estimating risk exposure (or covariance) is possible, this risk exposure may not be that meaningful in distin- guishing between high and low expected returns. An alternative approximation of equilibrium is a multifactor world CAPM. In this case, the risk of an equity investment is measured by its contribution to the variance of a portfolio of the factors (assuming the factors are traded-asset returns). These factors are often specified to represent broad economic forces such as average world interest rates, world inflation growth, and world busi- ness cycle movements. The risk of each asset or domestic market can be char- acterized by a number of betas that represent the sensitivity to changes in these factors. Both the single-factor CAPM and the multiple-factor CAPM present measures of risk. These measures are contingent on the asset-pricing model's being well specified. There are many possible sources of statistical rejection of these models. First, the fundamental assumptions that provide the building blocks for these models, such as utility specification, information environment, or distributional assumptions, could be violated. Second, as mentioned earlier, the benchmark portfolio that is used to measure risk could be improperly specified. Third, there could be problems with the returns data caused by infrequent trading of the component stocks. Fourth, capital markets may not be integrated. In examining emerging markets, this last factor may be crucial. If domestic investors encounter barriers in accessing foreign markets and foreign investors in accessing the domestic market, then risk is even more problematic to define. Indeed, Bekaert (1995) describes numerous barriers to investing in emerging markets. The notion that risk can be defined as the sensitivity to the changes in world factor returns is contingent on the assumption of complete market inte- gration. As the amount of segmentation. increases, risk takes on a new defini- tion as a security's sensitivity to local-market factors. The intuition is as fol- lows. In integrated world capital markets, the sensitivity to many local events can be hedged by a diversified portfolio. That is, a negative event in one country may be offset by positive news in another country. However, if capital Harvey 21 markets are segmented, the sensitivity to local events can have dramatic effects on the required returns for the securities that trade in the local market. The issue of integration and segmentation is a complicated one. The idea here is to start with the world asset-pricing paradigm and to explore the emerging markets' sensitivities to world factors. This is a logical place to start, and the exercise yields important insights that can be used for future studies of market integration. Another important issue concerns how information is incorporated into the analysis. Indeed, the traditional analyses of returns employ static models. For example, risk exposures and hence expected returns are often assumed to be constant. In the context of mature, industrial economies, this might be an innocuous assumption. In the arena of emerging economies, it is unlikely that risk exposure remains fixed over time. Emerging economies are often charac- terized by a shifting industrial structure that will induce changes in risk sensi- tivities. My research provides a first-step examination of time-varying correla- tions between the emerging-market returns and a number of world factors that are designed to capture common sources of risk in the world. These factors include risks related to equity and leverage, foreign exchange, commodity prices, world business cycles, and inflation. The article is organized as follows. Section I describes the data sources and summary statistics. Section II analyzes the risk exposure to a single factor. Section III discusses exchange risk. Section IV broadens the focus to a multi- factor asset-pricing model. Section V offers conclusions. I. DATA SOURCES AND SUMMARY STATISTICS Data on twenty-one industrial markets are from Capital International Per- spective, S.A. and Morgan Stanley & Co. (various issues), from now on re- ferred to as MSCI, and data on twenty emerging markets are from the Interna- tional Finance Corporation (IFC). Each market's return is based on a value- weighted portfolio of securities that trade in that market. The number of stocks included in the market indexes ranges from 17 to 300 (see Harvey 1991 for the MSCI markets and Harvey 1994c and Claessens, Dasgupta, and Glen 1995 for the emerging markets). Stocks are selected for inclusion on the basis of liquidity (how often they trade and the volume of trading) and size (market value). The industrial composition of the index is also important. That is, if two securities have approximately the same size and liquidity, the security that enables the index to better reflect the industrial composition of the local mar- ket may be chosen. All of the indexes reflect total returns, that is, dividends and capital gains. Details of the MSCI indexes are presented in Harvey (1991). The IFC indexes are described in Harvey (1994c). MSCI has also introduced a set of emerging-market indexes. These indexes suffer from a relatively short sample period (the earliest data begin in 1987), whereas IFC indexes for nine markets are available back to December 1975. 22 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I The early IFC data are problematic, however. When the IFC began publishing its indexes in 1981, the portfolio of stocks the IFC formed for each market was "backtracked" to December 1975, and this induced a "look-back" bias. Had the IFC selected the portfolio of stocks in December 1975, the portfolio might have been different from the one selected in January 1981 and backtracked to Decem- ber 1975. Indeed, if the stocks had been selected in December 1975, some might not have made it through the next five years because of bankruptcy. But the stocks selected in 1981, by construction, survived. The look-back bias is an example of survivorship bias. Given the look-back bias, the average returns should have been higher. To deal with the look-back bias problem, I calculated results based on the full sample (beginning in December 1975) and on a more re- cent subsample that bypasses the survivorship problem. 1 Some summary statistics are presented in table 1 for the full sample period, 1976-92. Only U.S. dollar returns are displayed. The statistics include the average (annualized) arithmetic and geometric returns, the standard deviation, and autocorrelations for lag 1, 2, and 12 periods. The industrial-market sum- mary statistics are presented over different samples by other authors and are included in table 1 for comparison with the emerging-market returns. Arithmetic and geometric average returns have an important difference. The arithmetic average is the return to a strategy that requires equal investment in each period. That is, the gains are not reinvested in the market. The geometric average has a more appealing portfolio interpretation. The geometric average represents the average return to a buy-and-hold strategy. In this strategy, a fixed amount is invested in the first period, and the portfolio is held until the end of the sample. The mean U.S. dollar returns in the emerging markets range from 72 percent (in Argentina) to -6 percent (in Indonesia, the sample for which begins only in January 1990). This range sharply contrasts with the range of average returns in the industrial markets. In the MSCI sample, no market has an average arith- metic return that exceeds 25 percent. In the IFC sample, nine markets (Argen- tina, Chile, Colombia, Mexico, the Philippines, Portugal, Taiwan (China), Turkey, and Venezuela) have returns that average above 25 percent. The emerging-market returns are characterized by high volatility, which induces large differences between the arithmetic and geometric mean returns. These differences are especially evident in the sample of emerging markets. The most dramatic example is in Argentina, where the arithmetic average return is 72 percent, and the geometric average return is 27 percent. Volatility (the standard deviation) ranges from 18 percent (in Jordan) to 105 percent (in Argentina). In contrast, volatilities in the MSCI markets range be- tween 15 and 33 percent. Thirteen emerging markets have a volatility higher than 33 percent (Argentina, Brazil, Chile, Greece, Indonesia, Mexico, Nigeria, 1. The full-sample (1976-92) results are reported in the arricle; the subsample (1985-92) results, which are not reported, are available from the author. Harvey 23 Table 1. Means, Standard Deviations, and Autocorrelations of International Equity Returns, 1976-92 (percent) Starting year and Arithmetic Geometric Standard Autocorrelation Market month mean mean deviation Pi P2 P12 Industrial markets Australia 1976.01 15.95 12.17 26.34 0.02 -0.13 -0.10 Austria 1976.01 15.20 12.31 24.21 0.14 0.02 0.01 Belgium 1976.01 18.03 15.80 20.97 0.07 0.07 -0.01 Canada 1976.01 12.44 10.39 19.93 -0.02 -0.07 -0.11 Denmark 1976.01 14.98 13.13 19.08 -0.07 0.06 -0.18 Finland 1988.01 -9.66 -12.17 22.15 0.09 -0.33 0.03 France 1976.01 17.78 14.51 25.26 0.02 -0.02 -0.10 Germany 1976.01 15.17 12.73 21.81 -0.04 -0.01 -0.08 Hong Kong 1976.01 25.45 19.25 33.88 0.02 -0.05 -0.06 Ireland 1988.01 12.61 9.72 24.28 -0.19 -0.11 -0.25 Italy 1976.01 14.68 11.11 26.84 0.18 -0.03 0.07 Japan 1976.01 17.97 15.20 23.38 0.01 -0.03 0.12 Netherlands 1976.01 18.95 17.30 17.53 -0.06 -0.09 0.01 New Zealand 1988.01 -1.98 -5.18 26.12 -0.04 -0.09 -0.10 Norway 1976.01 16.60 12.49 28.41 0.12 -0.04 -0.02 Singapore and Malaysia 1976.01 16.72 13.05 26.21 0.03 0.02 -0.05 Spain 1976.01 10.32 7.32 24.47 0.11 0.00 -0.03 Sweden 1976.01 18.65 15.87 23.24 0.08 0.00 0.01 Switzerland 1976.01 14.18 12.37 18.74 0.05 0.00 -0.03 United Kingdom 1976.01 19.20 16.50 22.90 -0.01 -0.09 -0.14 United States 1976.01 14.27 13.00 15.46 -0.01 -0.06 -0.02 Emerging markets Argentina 1976.01 71.66 27.02 105.06 0.05 0.06 -0.10 Brazil 1976.01 22.69 4.71 60.83 0.03 -0.04 0.03 Chile 1976.01 38.65 30.90 39.84 0.17 0.26 0.09 Colombia 1985.01 45.60 40.27 32.57 0.49 0.16 0.03 Greece 1976.01 9.75 3.82 36.27 0.12 0.18 -0.05 India 1976.01 21.45 17.88 26.87 0.09 -0.10 -0.09 Indonesia 1990.01 -6.29 -12.35 34.95 0.30 0.24 0.19 Jordan 1979.01 10.14 8.53 18.04 0.00 0.02 -0.02 Korea, Rep. of 1976.01 20.02 15.15 31.97 0.01 0.07 0.12 Malaysia 1985.01 13.56 9.81 26.90 0.05 0.08 -0.10 Mexico 1976.01 30.44 19.02 45.00 0.25 -0.08 -0.01 Nigeria 1985.01 2.18 -6.36 37.20 0.09 -0.13 -0.08 Pakistan 1985.01 25.65 23.21 22.38 0.27 -0.24 0.13 Philippines 1985.01 51.16 43.23 38.79 0.33 0.02 0.06 Portugal 1986.02 40.85 29.00 51.43 0.27 0.03 0.03 Taiwan (China) 1985.01 39.93 25.37 54.06 0.06 0.04 0.13 Thailand 1976.01 21.55 18.11 25.69 0.12 0.16 0.05 Turkey 1987.01 47.89 22.04 76.71 0.24 0.10 -0.16 Venezuela 1985.01 37.92 26.23 47.52 0.27 0.18 -0.06 Zimbabwe 1976.01 10.16 4.33 34.30 0.13 0.15 -0.04 Note: Values are based on U.S. dollar returns from monthly data from January 1976 to June 1992. pj denotes the jth-order autocorrelation coefficient. Source: The monthly returns for emerging markets are from the International Finance Corporation (IFC) Emerging Markets Data Base (EMDB). The industrial-market returns are from Capital Interna- tional Perspective, S.A. and Morgan Stanley & Co. (various issues), thereafter referred to as MSCI. 24 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I the Philippines, Portugal, Taiwan (China), Turkey, Venezuela, and Zimbabwe). The autocorrelations in table 1 measure the persistence (or predictability) of the market returns on the basis of past market returns. This persistence could be driven by market imperfections, such as infrequent trading of the compo- nent securities, or by some fundamental forces, such as predictable changes in sensitivities to world risk factors. Among the MSCI markets, only five have first-order autocorrelations that exceed 10 percent. Among the emerging mar- kets, twelve have autocorrelations greater than 10 percent. Indeed, eight have first-order autocorrelations above 20 percent (Colombia, Indonesia, Mexico, Pakistan, the Philippines, Portugal, Turkey, and Venezuela). Although the sample period is shorter for some of these markets, and the standard errors of the autocorrelations are higher, the evidence suggests that returns in many of the emerging markets can be predicted on the basis of past information. There is evidence that many of the emerging-market returns depart from normality. Harvey (1994c) presents a test of normality based on Hansen's (1982) generalized method of moments and rejects normality in fourteen of twenty emerging markets. Claessens, Dasgupta, and Glen (1995) use an alter- native test and find results that are consistent with Harvey's. Although nor- mality is not required for any of the measurements presented here, it may be that the distributional characteristics of the emerging-market returns induce a nonlinear relation between returns and global risk factors. This is a subject for further research. In my examination of the most recent subperiod, 1985-92 (not reported), similar patterns to those in the summary statistics have emerged. For example, the extraordinary arithmetic average return of 72 percent for Argentina is not a function of look-back bias; in the most recent subperiod the average return in Argentina is 88 percent. Indeed, in the most recent subperiod, ten emerging markets have returns exceeding 33 percent. Predictability is also retained, with ten markets exhibiting serial correlation above 20 percent. I also calculated the summary statistics for returns measured in local cur- rency terms, although they are not reported in the tables. The wild inflation in Argentina and Brazil is evident in the 228 and 156 percent average returns over the full sample. Other economies that have experienced severe inflation, such as Colombia, Chile, and Venezuela, also have much higher local returns. Cal- culating the returns in U.S. dollars eliminates the local inflation but retains the U.S. inflation. The correlations within emerging markets and the correlations between emerging markets and the MSCI markets are presented in table 2. Panel A shows the U.S. dollar return correlations within the emerging markets. These correlations are remarkably small. For example, the correlation between Ar- gentina and Brazil is only -3 percent. The correlation between Pakistan and India is -5 percent. The correlation between Colombia and Chile is 0 percent. The correlations in the most recent subperiod (not reported) show the same Harvey 25 characteristics. The correlation between Argentina and Brazil is still -4 per- cent, which is somewhat surprising, given that Argentina and Brazil have recently become important trading partners. The correlations between the emerging and industrial markets are presented in panel B. The average correlations are very small. Over the full sample period, Malaysia has the highest correlation with industrial markets and Mex- ico has the second highest. For the other markets, the correlations are often less than 10 percent. For example, Argentina has correlations of less than 10 percent with eighteen of the twenty-one industrial markets. The Republic of Korea has correlations of less than 10 percent with eight of the twenty-one industrial markets. The same holds true in the most recent subperiod (not reported). The correlations for Argentina and Venezuela with each of the twenty-one MSCI markets are less than 10 percent, and many emerging markets have negative correlations with several industrial markets. Mullin (1993) argues that the low average monthly correlations between emerging markets and Msci markets as well as the cross-correlations within the emerging markets could be caused by market imperfections such as infrequent trading. Mullin shows that the annual correlations are higher than the monthly correlations. However, it is not clear that the annual correlations are statis- tically higher. In my sample (excluding Indonesia), there are 171 cross- correlation coefficients for emerging-market returns. When monthly data are used, twenty-six correlations are significantly different from zero; when an- nual data are used, only five are significantly different from zero. This evi- dence supports the view that the low correlations are real rather than an artifact of infrequent trading. In addition, when monthly data are used, five emerging markets have significant correlation with the U.S. return; whereas when annual data are used, only one market has significant correlation (at the 5 percent level of significance). The low correlations imply that significant benefits are possible in diversify- ing into the emerging markets. Even though the volatility of the individual emerging markets is high, the low correlations should reduce portfolio vol- atility. This reduction in volatility is evident in the work of Divecha, Drach, and Stefek (1992), Harvey (1994a, 1994c), Stone (1990), and Wilcox (1992). These analyses measure the effect of adding emerging markets to portfolios of industrial-market securities. As mentioned above, the studies show that including emerging markets in a well-diversified portfolio reduces overall vol- atility even though the emerging-market equities, held alone, are much more volatile than the industrial-market equities. However, investors usually require information in addition to the mean and variance of the portfolio before making their portfolio decisions. An important control in real-world portfolio management is the level of risk exposure that the portfolio bears. That is, a quadratic program can select the portfolio with the highest expected return for a given level of variance. How- ever, this portfolio might have an unacceptable exposure to, for example, 26 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Table 2. Correlation of International Equity Returns, 1976-92 A. Correlation of Emerging-Market Returns Argen- Co- Indo- Korea, Market tina Brazil Chile lombia Greece India nesia Jordan Rep. of Argentina 1.00 Brazil -0.03 1.00 Chile 0.08 0.00 1.00 Colombia -0.10 0.06 0.00 1.00 Greece 0.06 -0.03 0.14 0.23 1.00 India 0.12 -0.01 0.06 -0.11 0.07 1.00 Indonesia -0.29 0.06 0.09 0.25 0.36 0.07 1.00 Jordan -0.02 0.00 0.01 0.03 0.10 0.16 0.20 1.00 Korea, Rep. of -0.11 0.02 0.03 -0.01 -0.05 0.00 0.00 -0.18 1.00 Malaysia -0.08 0.12 0.24 0.02 0.06 -0.01 0.46 0.07 0.07 Mexico 0.14 -0.02 0.13 0.02 0.05 0.03 0.04 -0.07 0.10 Nigeria 0.11 0.01 -0.03 0.14 0.11 -0.13 -0.10 0.00 0.04 Pakistan -0.03 -0.03 -0.10 0.43 -0.09 -0.10 0.05 0.11 -0.01 Philippines -0.10 0.12 0.20 0.13 0.12 -0.11 0.50 0.09 0.18 Portugal -0.02 0.10 0.21 0.14 0.41 -0.11 0.24 -0.03 0.10 Taiwan (China) -0.04 0.07 0.31 0.11 0.09 -0.11 0.30 0.10 0.04 Thailand -0.04 -0.01 0.10 0.14 0.26 0.10 0.42 0.06 0.01 Turkey 0.15 0.07 0.02 0.13 0.28 0.09 0.28 -0.12 0.02 Venezuela 0.04 -0.15 -0.23 0.09 -0.04 0.00 0.01 -0.01 -0.15 Zimbabwe -0.08 -0.04 0.13 -0.18 0.12 0.06 0.04 0.01 -0.08 B. Correlation of Emerging-Market Returns with Industrial-Market Returns Aus- Aus- Bel- Den- Fin- Ger- Hong Market tralia tria gium Canada mark land France many Kong Ireland Argentina 0.14 0.01 -0.09 0.10 0.00 -0.15 0.03 -0.01 -0.08 -0.20 Brazil 0.06 0.04 0.04 0.00 0.02 0.38 0.04 0.07 0.13 0.27 Chile 0.13 0.15 0.13 0.07 0.03 0.03 0.04 0.08 0.12 0.02 Colombia -0.03 0.03 0.05 0.08 0.01 0.08 0.00 0.02 0.09 0.07 Greece 0.08 0.27 0.18 0.10 0.06 0.05 0.17 0.14 0.07 0.12 India 0.13 0.19 0.07 0.05 0.13 0.02 0.13 0.10 0.04 -0.03 Indonesia -0.05 0.58 0.32 0.20 0.29 0.45 0.17 0.41 0.40 0.27 Jordan 0.15 0.18 0.18 0.15 0.17 0.16 0.19 0.18 0.08 0.33 Korea, Rep. of 0.03 0.02 0.12 0.17 0.07 0.33 0.04 0.09 0.09 0.53 Malaysia 0.40 0.18 0.25 0.50 0.21 0.48 0.15 0.22 0.59 0.55 Mexico 0.19 0.03 0.22 0.19 -0.02 0.23 0.13 0.11 0.15 0.14 Nigeria -0.12 0.12 0.07 0.11 0.21 0.04 0.08 0.13 -0.09 0.09 Pakistan 0.01 0.08 0.11 -0.05 0.10 -0.05 0.04 0.06 0.10 0.13 Philippines 0.11 0.14 0.33 0.31 0.17 0.43 0.20 0.20 0.30 0.31 Portugal 0.26 0.14 0.20 0.21 0.12 0.14 0.18 0.13 0.29 0.54 Taiwan (China) 0.21 0.19 0.18 0.10 -0.02 0.29 0.11 0.12 0.15 0.17 Thailand 0.24 0.21 0.28 0.10 0.11 0.24 0.14 0.23 0.17 0.38 Turkey 0.11 0.31 0.12 0.02 0.13 -0.01 -0.04 0.09 0.05 0.21 Venezuela -0.02 -0.17 -0.03 0.04 -0.11 -0.20 -0.13 -0.25 -0.04 -0.27 Zimbabwe -0.02 0.17 -0.01 0.11 0.11 0.12 0.00 0.04 0.00 0.00 Note: Correlations are based on monthly data in U.S. dollars from January 1976 to June 1992. Harvey 27 Malay- Paki- Philip- Por- Taiwan Thai- Vene- Zim- sia Mexico Nigeria stan pines tugal (China) land Turkey zuela babwe 1.00 0.40 1.00 -0.17 -0.10 1.00 -0.07 -0.4 0.02 1.00 0.31 0.06 0.08 0.00 1.00 0.23 0.35 -0.20 0.03 0.03 1.00 0.23 0.35 -0.14 -0.06 0.06 0.39 1.00 0.51 0.24 -0.11 0.03 0.25 0.35 0.40 1.00 0.26 0.17 0.08 0.04 0.12 0.27 0.17 0.29 1.00 -0.02 -0.05 0.12 0.03 -0.17 -0.07 -0.22 -0.11 -0.10 1.00 0.01 0.03 0.04 -0.07 0.02 0.12 -0.04 -0.05 0.01 0.10 1.00 Singa- Nether- New pore and Switzer- United United Italy Japan lands Zealand Norway Malaysia Spain Sweden land Kingdom States 0.07 -0.04 -0.03 0.03 0.00 -0.01 0.01 0.01 0.04 -0.08 0.01 0.11 0.06 0.04 0.24 0.15 0.07 0.11 0.13 0.09 0.09 0.07 0.05 0.05 0.05 -0.25 0.05 0.08 0.10 0.06 0.02 0.03 0.02 0.07 0.00 0.07 0.01 -0.07 0.08 0.13 0.02 0.07 0.06 0.10 0.15 0.09 0.12 0.06 0.11 0.08 0.17 0.06 0.20 0.13 0.10 0.02 -0.07 0.04 0.07 0.06 0.09 0.08 0.09 0.10 0.05 -0.01 0.51 -0.10 0.32 0.14 0.38 0.43 0.05 0.22 0.28 0.12 0.16 0.03 0.07 0.14 -0.06 0.09 0.10 0.02 0.09 0.26 0.24 0.07 0.08 0.25 0.14 0.16 0.09 0.16 0.10 0.20 0.17 0.19 0.18 0.13 0.20 0.37 0.29 0.52 0.92 0.21 0.38 0.27 0.44 0.53 0.08 0.10 0.17 -0.10 0.23 0.28 0.15 0.21 0.16 0.20 0.28 0.04 0.10 0.19 0.03 -0.04 -0.17 0.25 -0.16 0.15 0.08 0.08 0.03 -0.03 0.13 0.05 0.00 0.01 0.00 0.00 0.06 0.13 0.02 0.30 0.25 0.33 0.24 0.12 0.34 0.36 0.26 0.20 0.18 0.28 0.21 0.39 0.25 0.38 0.35 0.25 0.32 0.31 0.26 0.34 0.22 0.08 0.20 0.01 -0.12 0.19 0.30 0.13 0.14 0.02 0.09 0.14 0.13 0.14 0.18 0.17 0.09 0.38 0.12 0.20 0.25 0.20 0.15 0.05 0.04 0.09 0.03 0.16 0.29 0.22 0.13 0.09 0.01 0.01 -0.22 -0.12 -0.08 -0.07 -0.08 0.02 -0.13 -0.20 -0.21 0.06 -0.07 0.02 0.08 -0.01 -0.01 0.14 0.04 0.11 -0.04 0.03 0.10 0.03 Source: Author's calculations. 28 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 shocks in the price of oil. As a result, it is important to try to measure the risk exposure of the emerging-market equity returns. Il. SINGLE-FACTOR MODELS Implicit in the mean-variance analysis are the assumptions that investors prefer higher expected returns and that a portfolio's risk (which investors dislike) is captured by the overall portfolio variance. It is useful to characterize the risk of the individual markets. As mentioned earlier, in implementing port- folio optimization, constraints are often added to limit exposure to certain types of risk. The problem, in particular, is how to characterize the risk of the emerging markets. If an efficient benchmark portfolio exists, then the risk of the individual market can be measured by the covariance with the efficient benchmark. The expected return on that market will be exactly linear in the efficient benchmark (Roll 1977; Ross 1977). If the benchmark is not efficient (and even if it is very close to being efficient), there may be no relation between the covariance and the expected returns. One potential benchmark is the MSCI world market portfolio in excess of the thirty-day Eurodollar deposit rate. Cumby and Glen (1990), Ferson and Harvey (1994), Harvey (1991), and Harvey and Zhou (1993) fail to reject the mean-variance efficiency of this portfolio within the set of MSCI industrial- market portfolios. Although this is the most widely used world benchmark, one problem with the portfolio is its lack of investment in emerging markets. Currently, emerging markets represent less than 2 percent of the investments of the MSCI world portfolio, whereas emerging markets represent about 7 percent of world equity capitalization (see IFC 1993). Table 3 provides estimates of the one-factor model for both industrial and emerging markets. The loading on the MSCI world market portfolio is signifi- cantly different from zero in each of the industrial markets.2 However, among the emerging markets, only seven (Greece, Korea, Malaysia, Mexico, the Phi- lippines, Portugal, and Thailand) have significant betas. In addition, only one of the markets has a beta greater than unity (Portugal, with a beta of 1.168); therefore, a strong relation between expected returns and this risk exposure is unlikely. A possible explanation of the low betas is that the stocks in the local index trade infrequently. That is, suppose the world market goes up one month and down the next, but that the stocks in the local portfolio do not trade in the first month. Even though the value of the local stocks might rise with the world market in terms of their unobserved market value, the covariance of the re- turns of the local and world markets over the two months may be close to zero. 2. Note that the table presents regressions of returns, not excess returns, on the MSCI world market portfolio. As a result, many of the intercepts are significantly different from zero. Harvey 29 One solution to the problem of infrequent trading was suggested by Scholes and Williams (1977). In this correction the local-market return is regressed on the lagged world return, the contemporaneous world return, and the lead of the world-market return. Then the three betas are averaged and divided by one plus twice the first-order autocorrelation in the world-market return. These calculations provide an adjustment for possible infrequent trading. Although not reported in table 3, the Scholes-Williams (1977) betas are broadly similar to the usual betas, with two exceptions. The beta for Mexico for 1976-92 increases from 0.76 to 1.59. The beta for the Philippines for the same period increases from 0.77 to 1.49. For the other markets, there is little change. For example, the beta for Portugal increases from 1.17 to 1.24. In the more recent subperiod (1985-92), the results (not reported) are similar. Only six emerging markets have betas that are significantly different from zero. Only a single market has a beta greater than one, and two markets have negative betas. The R2s of these regressions range from 0 percent (in thirteen markets) to 20 percent (in Malaysia). The use of the Scholes-Williams (1977) beta has little effect in the most recent subperiod, with the exception of Mexico. For this market, the beta increases from 0.81 to 1.91 using the Scholes-Williams methodology. The inability of the single-factor model to characterize the emerging-market re- turns is a result of the MSCI portfolio's being inefficient in relation to the set of as- sets examined. Indeed, the low or negative betas are expected from the low and negative correlations that many of the emerging markets have with the industrial market. The MSCI world market portfolio is really an industrial world market portfolio. The betas estimated in table 3 assume that the risk is constant throughout the pe- riod examined. Annex figure A-1 shows five-year rolling correlation measures of the local-market returns and the MSCI excess returns. The graphs depict some inter- esting changes in the correlations. In Brazil, correlations increase from 0 percent in the early 1980s to 25 percent by 1992. There is no significant pattern in any of the other South American markets. However, the Mexican correlations increase from 0 percent in 1986 to 30 percent by 1991. In the East Asian markets, the correlations increase progressively reaching 40 percent in Korea, 60 percent in Malaysia, 40 per- cent in the Philippines, 15 percent in Taiwan (China), and 40 percent in Thailand. In India the correlation uniformly decreases over time. In Greece and Portugal the correlations are 25 percent and 50 percent, respectively, by 1992. The time variation in the correlations suggests that the sensitivity of many emerging markets to the MSCI world portfolios is increasing. Although the betas have only limited ability to explain the expected-return variation across different markets (the cross-sectional adjusted R2 is only 4 percent in the overall period), it appears as if the cross-sectional R2 may increase from the beginning of the sample to the end.3 3. In contrast, for the twenty-one industrial markets, the adjusted R2 of the regression of the betas on twenty-one industrial-market average returns is 30 percent. 30 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Table 3. One-Factor Model Loadingsfor Forty-One Equity Markets, 1976-92 Starting year World re- Market and month Intercept turn' beta R2 Industrial markets Australia 1976.01 0.009 0.889 0.239 (1.912) (4.796) Austria 1976.01 0.011 0.488 0.082 (2.171) (3.726) Belgium 1976.01 0.011 0.886 0.377 (3.334) /9.502) Canada 1976.01 0.006 0.932 0.464 (2.093) (10.328) Denmark 1976.01 0.010 0.675 0.263 (2.869) (8.295) Finland 1988.01 -0.008 0.667 0.219 (-1.109) (3.912) France 1976.01 0.010 1.081 0.387 (2.475) (11.681) Germany 1976.01 0.009 0.812 0.292 (2.442) (7.629) HongKong 1976.01 0.017 0.998 0.181 (2.627) (5.015) Ireland 1988.01 0.010 1.047 0.470 (1.465) (6.546) Italy 1976.01 0.008 0.857 0.213 (1.758) (7.689) Japan 1976.01 0.010 1.159 0.521 (2.954) (11.045) Netherlands 1976.01 0.012 0.874 0.527 (4.877) (13.545) NewZealand 1988.01 -0.002 0.452 0.059 (-0.192) (2.247) Norway 1976.01 0.009 1.029 0.276 (1.859) (7.511) Singapore and Malaysia 1976.01 0.010 0.941 0.271 (2.073) (5.475) Spain 1976.01 0.005 0.821 0.236 (1.162) (7.220) Sweden 1976.01 0.012 0.858 0.286 (2.981) (8.061) Switzerland 1976.01 0.008 0.874 0.460 (2.867) (12.096) United Kingdom 1976.01 0.011 1.099 0.488 (3.362) (15.002) United States 1976.01 0.008 0.840 0.626 (4.176) (13.898) Emerging markets Argentina 1976.01 0.061 -0.180 -0.004 (2.752) (-0.430) Brazil 1976.01 0.017 0.407 0.005 (1.397) (1.287) Chile 1976.01 0.032 0.120 -0.003 (3.776) (0.571) Colombia 1985.01 0.037 0.145 -0.006 (3.599) (0.763) Greece 1976.01 0.006 0.381 0.019 (0.883) (2.117) India 1976.01 0.018 -0.024 -0.005 (3.242) (-0.175) Harvey 31 Starting year World re- Market and month Intercept turn, beta R2 Indonesia 1990.01 -0.004 0.126 -0.031 (-0.249) (0.311) Jordan 1979.01 0.008 0.159 0.012 (1.902) (1.548) Korea, Rep. of 1976.01 0.014 0.549 0.058 (2.286) (3.686) Malaysia 1985.01 0.005 0.738 0.199 (0.700) (3.542) Mexico 1976.01 0.022 0.764 0.057 (2.416) (3.107) Nigeria 1985.01 0.000 0.222 -0.001 (0.004) (1.031) Pakistan 1985.01 0.021 0.052 -0.010 (3.022) (0.355) Philippines 1985.01 0.036 0.770 0.099 (3.188) (2.827) Portugal 1986.02 0.027 1.168 0.148 (1.780) (4.807) Taiwan (China) 1985.01 0.028 0.687 0.034 (1.644) (1.629) Thailand 1976.01 0.016 0.379 0.041 (2.989) (1.940) Turkey 1987.01 0.039 0.216 -0.013 (1.459) (0.524) Venezuela 1985.01 0.035 -0.382 0.007 (2.271) (-1.119) Zimbabwe 1976.01 0.008 0.214 0.003 (1.053) (1.151) Note: All returns are calculated in U.S. dollars and are in excess of the thirty-day Eurodeposit rate. Results are reported for a linear regression of the excess market return on the world return. The intercept and slope (beta) are reported with heteroskedasticity-consistent t-ratios (in parentheses). a. The MSci value-weighted world-market portfolio in excess of the thirty-day Eurodollar deposit rate. Source: The monthly returns for emerging markets are from IFC EMDB. The industrial-market returns are from MSCI. It is clear that a single-factor model is not enough to provide a meaningful defi- nition of risk. Harvey (1994c) provides statistical tests of the single-factor model (testing eight markets during the period from February 1976 to June 1992 and eighteen markets from March 1986 to June 1992) and finds strong rejections of the model's implications. Of course, these rejections could be caused by the assumption that the betas and expected returns are constant over time. However, Harvey (1994c) also allows for both the beta and the expected returns to change over time. He finds that this more general model is also rejected. In a way, his evidence contrasts with that of Buckberg (1995), who uses a world CAPM and fails to reject the model for fourteen emerging markets. The difference between these two sets of results can be reconciled. Buckberg estimates her model under the null hypoth- esis that the excess local-market returns are proportional to the excess world- market return. For many markets, her tests cannot reject the null hypothesis. 32 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Harvey (1994c) estimates the model under the alternative hypothesis that excess returns are linear in the excess world-market return. The intercept, under the null (to make linearity equal to proportionality), should be equal to zero. Harvey finds sharp evidence against the null with this formulation. It is reason- able to conclude that Buckberg's test lacks power. It is unlikely that the single- factor model is sufficient to characterize expected returns in emerging markets. III. FOREIGN EXCHANGE EXPOSURE The international asset-pricing models of Adler and Dumas (1983), Sercu (1980), Solnik (1974), and Stulz (1981) all provide a role for exchange risk. In the Adler and Dumas model (their equation 14), with N countries, expected returns in a numeraire currency are generated by the covariance with the world portfolio and by the covariances of the asset returns and inflation rates in all the countries. The weights on these inflation covariances depend on the wealth- weighted risk aversion in each country. The usual way to implement this model is to follow Solnik's (1974) assumption that the asset covariance with the numer- aire country's inflation is zero. Expected returns can then be written in terms of their covariance with the world portfolio and their N - 1 covariances with exchange rate changes (see the discussion in Dumas 1994). Unfortunately, the Adler and Dumas (1983) model is intractable unless a very small number of countries are examined. For example, Dumas and Solnik (1994) are able to estimate the model for only four countries. One possible simplification pursued in a number of papers is to aggregate the exchange rate factor (see Bailey and Jagtiani 1994; Bodurtha 1990; Brown and Otsuki 1993; Ferson and Harvey 1993, 1994; Harvey, Solnik, and Zhou 1994, and Jorion 1991). Given that it is impossible to observe the wealth-weighted risk aversions of the N - 1 markets, trade weights (exports plus imports) are used as an aggregation method. The aggregation of the exchange risk factor departs from the asset-pricing theory but provides tractability. One may also view this as the prespecification of factors in some general multifactor model of asset pricing, following Merton (1973), Ross (1976), and Sharpe (1984). Empirically, Ferson and Harvey (1993, 1994) and Harvey, Solnik, and Zhou (1994) have found the aggregated exchange risk factor to be significant in both conditional and unconditional asset-pricing tests. Harvey, Solnik, and Zhou show that the loadings from these first two factors are able to explain 35 percent of the cross-section of expected bond and stock returns in industrial markets. Both the Organization for Economic Cooperation and Development (OECD) and the Federal Reserve Bank publish indexes of the value of the U.S. dollar. The percentage change in these indexes represents the changes in the exchange rate. The index changes are not "true" returns, however, because investors are usually assumed not to hold cash: an investor purchasing deutsche mark would immediately deposit the deutsche mark in a Euromark account. Hence, to con- Harvey 33 struct a currency return index, local interest rates need to be included in the calculation. Harvey (1994b) describes the construction of a trade-weighted index of cur- rency returns. Because I use this index to measure the global currency-risk exposure of the emerging markets, a brief review of the construction of the index is in order. The index of currency returns is similar to the Federal Reserve index in that it uses trade weights to aggregate each market component. A trade weight is the value of exports plus the value of imports divided by the sum of both for ten markets (Group of Ten plus Switzerland minus the United States). The two indexes are dissimilar, however, in that the Federal Reserve index uses the trade weights that existed during 1972-76 and keeps these weights fixed, whereas the index of currency returns allows the trade weights to change over time. The fixed trade weights for the ten markets in the Federal Reserve index are as follows: Germany, 20.8 percent; Japan, 13.6 percent; France, 13.1 percent; the United Kingdom, 11.9 percent; Canada, 9.1 percent; Italy, 9.0 percent; the Netherlands, 8.3 percent; Belgium, 6.4 percent; Sweden, 4.2 percent; and Switzerland, 3.6 percent. The current value of the Federal Reserve index is calculated by dividing the U.S. dollar per local currency rate in the base period of March 1973 by the U.S. dollar per local currency rate in the current period. Hence, as the U.S. dollar depreciates, the index decreases because the denomi- nator gets larger. Harvey (1994b) allows the trade weights to change through time. Using the same general approach as the Federal Reserve, he lets the weights reflect a five- year moving average of trade. Shifts in trade weights are important. Belgium's trade sector dropped from 7.2 percent in December 1969 to 6.7 percent in November 1992. The drop for Canada is one of the largest, from 10.3 percent to 7.5 percent. France's trade gained from 11.8 percent to 12.7 percent. Ger- many's grew from 19.9 percent to 21.7 percent. Italy's showed an increase from 8.8 percent to 9.9 percent. The most dramatic increase was Japan's, which jumped from 10.7 percent to 15.6 percent. The Netherlands' trade sector lost a small amount of ground, dropping from 7.7 percent to 7.3 percent. Sweden's fell from 4.5 percent to 3.1 percent. Switzerland's was stable at 3.7 percent. Finally, the United Kingdom's trade share plummeted from 15.4 percent to 11.5 percent. To allow for reasonable publication delays, the trade weights are lagged by one year when the index returns are calculated. That is, the average trade weights for the period January 1975 to December 1979 are applied to the currency return for January 1981. Harvey then calculates the exchange rate-investment return for each market by converting 100 U.S. dollars into local currency on the last day of the month and investing in a thirty-day Eurodeposit. One month later the deposit comes due and is converted back to dollars. Notice how this approach is different from that of the Federal Reserve index. As the U.S. dollar depreciates, the investor holding a foreign currency will gain. 34 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I In summary, Harvey's (1994b) global exchange rate index has two features that distinguish it from the traditional exchange rate indexes: currency returns rather than rate changes are used, and trade weights are allowed to change over time. Although this index does not include emerging markets, including emerg- ing markets would probably not affect the index very much, because the trade weight on these markets would be very small. Even though the trade weight is small, swings in the exchange rate are large; however, the index is calculated with returns, not rate changes. Presumably, a large depreciation in currency, for example in Brazil, would be offset by a high interest rate on local deposits. For industrial-market returns, the betas on the exchange rate-investment in- dex presented in table 4 are significantly different from zero for twelve of twenty-one markets within the overall period. The betas range from -0.50 for the U.S. portfolio to 0.94 for Austria. In general, the non-Scandinavian markets in Europe exhibit significant positive betas. The risk for Canada and the United States is negative. The exchange rate factor has marginal explanatory power in eight of the twenty emerging markets: Greece, India, Jordan, Malaysia, Mexico, Pakistan, Taiwan (China), and Zimbabwe. However, in eight other markets the R2 of the two-factor regressions is zero. In the most recent subperiod (not reported), the marginal explanatory power of the foreign exchange risk factor is not substantially altered. Eight emerging markets have t-statistics greater than 1.5 on the exchange portfolio. This portfo- lio has some ability to explain returns in Argentina, Chile, and Thailand. Plots of the five-year rolling correlations between the emerging-market returns and the foreign exchange portfolio are presented in figure A-2. These correlation measures are not the same as betas because there is no control for the correlation with the world market portfolio. However, the plots reveal interesting sim- ilarities to the ones presented in figure A-1. There is a tendency for the correla- tions to increase in absolute magnitude during the last seven years of the sample in some of the markets. This is the case in the South American markets and Mexico. The correlations are 0 percent in the East Asian markets, with the exception of Thailand. The correlation of the foreign exchange index and the Greek equity market is about 30 percent in 1992 and rises to more than 50 percent for Portugal. Again, although foreign exchange exposure does not explain the average returns (measured over the entire sample, the cross-sectional R2 is 7 percent), the graphs indicate that the cross-sectional relation may be strengthening over time.4 IV. MULTIFACTOR MODEL International asset-pricing models that include multiple factors are described in Bansal, Hsieh, and Viswanathan (1993); Ferson and Harvey (1993, 1994); 4. For the twenty-one industrial markets, the cross-sectional adjusted R2 is 37 percent. Harvey 35 Hodrick (1981); Ross and Walsh (1983); Solnik (1983); and Stulz (1981, 1993) find that a number of global risk factors are important in capturing the variation in both the cross-section of expected returns and the time series of expected returns. The three additional factors examined here are similar to theirs. The factors are designed to capture three broad economic forces: commodity prices, inflation, and the world business cycle. A number of researchers have found that shocks in crude oil prices have important effects on stock returns in industrial markets. I specify the factor as the change in the U.S. dollar price per barrel of crude oil at the wellhead less the Eurodollar deposit rate. The world business cycle is proxied by the growth rate in OECD industrial production. Finally, world inflation is proxied by the OECD inflation rate. The risk exposures for the five-factor model are presented in table 5. In the overall sample of twenty-one industrial markets, eight have significant exposure to oil, two have exposure to industrial production growth, and five have significant exposure to inflation. The adjusted R2s of these regressions range from 3 percent to 71 percent. The inclusion of these additional factors does not help to explain the emerging-market returns. Of the twenty emerging markets, five have signifi- cant oil exposure. In four of these markets (Colombia, Jordan, the Philippines, and Taiwan, China), the exposure is negative, which indicates decreasing returns when oil prices rise. In Venezuela, the exposure is positive, as it is (albeit insignificantly) in Mexico and Nigeria. Only three of the emerging markets have significant exposure to world industrial production, and only four markets have significant loadings on world inflation. The adjusted R2s of the five-factor regressions range from 0 percent (in six markets) to 25 percent in Malaysia. Plots of the five-year rolling correlations with the final three factors were calculated but are not presented. In most industrial markets, the oil exposure is negative. This means that an increase in the price of oil is viewed as bad news, on average, by market participants. Even producers such as the United Kingdom and Canada have 0 percent or negative exposure. In the emerging markets, there are a number of different patterns. For example, the Mexican exposure, al- though positive in the early 1980s, is now negative. It appears that the Mexican economy's dependence on the strength of the U.S. economy is more important than Mexico's oil holdings. India's market has an unexpected positive correla- tion with oil, increasing from 0 percent in 1987 to about 35 percent by 1992. Thailand's correlation shows a dramatic change, from 35 percent in the early 1980s to -35 percent by 1992. The graphical analysis also suggests time-varying exposures to growth both in industrial production and in the inflation rate. In seven of the emerging markets, the correlation with OECD industrial production shows an increase over time. In the other thirteen emerging markets, there are no detectable patterns over time. There are no obvious trends in the correlation with OECD inflation across all the emerging markets. 36 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Table 4. Two-Factor Model Loadings for Forty-One Equity Markets, 1976-92 World Exchange rate Starting year returna investment Market and month Intercept beta indexb beta R2 Industrial markets Australia 1976.01 0.010 0.918 -0.128 0.237 (1.970) (4.279) (-0.523) Austria 1976.01 0.010 0.271 0.941 0.213 (2.162) (1.684) (5.156) Belgium 1976.01 0.011 0.740 0.635 0.457 (3.393) (6.890) (5.075) Canada 1976.01 0.007 1.001 -0.299 0.481 (2.195) (11.537) (-2.737) Denmark 1976.01 0.009 0.551 0.540 0.331 (2.880) (6.893) (4.552) Finland 1988.01 -0.008 0.710 -0.195 0.212 (-1.057) (4.161) (-0.649) France 1976.01 0.010 0.920 0.697 0.452 (2.461) (9.536) (4.940) Germany 1976.01 0.008 0.628 0.800 0.410 (2.476) (4.723) (5.479) HongKong 1976.01 0.017 1.013 -0.064 0.177 (2.634) (4.352) (-0.212) Ireland 1988.01 0.010 1.036 0.052 0.460 (1.445) (5.776) (0.235) Italy 1976.01 0.008 0.799 0.253 0.217 (1.718) (7.122) (1.374) Japan 1976.01 0.010 1.063 0.418 0.548 (2.962) (9.987) (3.316) Netherlands 1976.01 0.012 0.805 0.300 0.551 (4.914) (11.123) (3.413) New Zealand 1988.01 -0.002 0.478 -0.120 0.043 (-0.163) (2.578) (-0.343) Norway 1976.01 0.009 0.999 0.129 0.274 (1.841) (6.722) (0.667) Singapore and 1976.01 0.010 1.011 -0.303 0.279 Malaysia (2.167) (5.187) (-1.232) Spain 1976.01 0.005 0.741 0.350 0.251 (1.110) (5.472) (2.074) Sweden 1976.01 0.012 0.848 0.043 0.283 (2.972) (7.510) (0.329) Switzerland 1976.01 0.008 0.729 0.629 0.559 (2.971) (8.170) (5.967) United Kingdom 1976.01 0.011 1.042 0.249 0.496 (3.342) (12.187) (2.108) United States 1976.01 0.009 0.955 -0.499 0.718 (5.157) (17.179) (-7.692) Emerging markets Argentina 1976.01 0.061 -0.036 -0.621 -0.006 (2.783) (-0.089) (-0.995) Brazil 1976.01 0.018 0.561 -0.667 0.010 (1.451) (1.711) (-1.501) Chile 1976.01 0.031 0.065 0.240 -0.005 (3.752) (0.265) (0.782) Colombia 1985.01 0.035 0.103 0.238 -0.011 (3.432) (0.562) (0.754) Greece 1976.01 0.006 0.230 0.655 0.043 (0.824) (1.107) (2.391) Harvey 37 World Exchange rate Starting year returna investment Market and month Intercept beta indexb beta R2 India 1976.01 0.018 -0.136 0.489 0.020 (3.230) (-0.956) (2.444) Indonesia 1990.01 -0.002 0.180 -0.351 -0.059 (-0.089) (0.434) (-0.483) Jordan 1979.01 0.008 0.075 0.356 0.044 (1.972) (0.702) (2.296) Korea, Rep. of 1976.01 0.015 0.627 -0.339 0.063 (2.326) (3.995) (-1.407) Malaysia 1985.01 0.010 0.865 -0.726 0.272 (1.510) (4.643) (-3.012) Mexico 1976.01 0.023 1.003 -1.036 0.100 (2.603) (3.862) (-2.692) Nigeria 1985.01 -0.002 0.159 0.360 -0.002 (-0.187) (0.808) (1.239) Pakistan 1985.01 0.017 -0.040 0.524 0.041 (2.728) (-0.276) (2.295) Philippines 1985.01 0.038 0.819 -0.282 0.094 (3.398) (2.934) (-0.916) Portugal 1986.02 0.028 1.185 -0.108 0.137 (1.754) (5.071) (-0.252) Taiwan (China) 1985.01 0.037 0.937 -1.426 0.102 (2.298) (2.426) (-3.344) Thailand 1976.01 0.016 0.409 -0.132 0.039 (3.060) (1.753) (-0.501) Turkey 1987.01 0.037 0.155 0.445 -0.026 (1.387) (0.345) (0.475) Venezuela 1985.01 0.032 -0.461 0.451 0.005 (2.115) (-1.303) (1.301) Zimbabwe 1976.01 0.007 0.072 0.619 0.028 (1.002) (0.368) (2.508) Note: All returns are calculated in U.S. dollars and are in excess of the thirty-day Eurodeposit rate. Results are reported for a linear regression of the excess market return on the world return and the exchange-investment index. The intercept and slopes (betas) are reported with heteroskedasticity- consistent t-ratios (in parentheses). a. The MSCI value-weighted world-market portfolio in excess of the thirty-day Eurodollar deposit rate. b. The U.S. dollar return to holding a trade-weighted portfolio of Eurocurrency deposits in ten countries (Belgium, Canada, France, Germany, Italy, Japan, Netherlands, Sweden, Switzerland, and the United Kingdom) in excess of the thirty-day Eurodollar deposit rate (see Harvey 1994b for details of the construction of the index). Source: The monthly returns for emerging markets are from the lFC EMDB. The industrial-market returns are from MSCI. Although the addition of the three factors increases the ability to explain the cross-section of expected returns (adjusted R2 rises to 10 percent), much is left unexplained.5 There are two ways to interpret these results. In one sense, the combination of the five prespecified factors can be considered a portfolio. The inability of the factor loadings as a group to explain the cross-section of average returns suggests that this portfolio is inefficient. 5. For the twenty-one industrial markets, the adjusted R2 is 29 percent. Table 5. Five-Factor-Model Loadings for Forty-One Equity Markets, 1976-92 Starting Exchange rate year and World returna investment Commodity World business Inflatione Market month Intercept beta indexb beta pricesc beta cycled beta beta R2 Industrial markets Australia 1976.01 0.002 0.960 -0.140 0.072 0.269 1.225 0.241 (0.188) (4.504) (-0.594) (1.851) (0.389) (0.739) Austria 1976.01 0.019 0.255 0.914 -0.047 0.998 -2.009 0.226 (2.099) (1.680) (5.015) (-0.912) (1.694) (-1.740) Belgium 1976.01 0.020 0.719 0.620 -0.021 0.112 -1.781 0.459 (3.068) (6.665) (4.755) (-0.894) (0.222) (-1.812) Canada 1976.01 0.003 1.030 -0.313 0.053 0.275 0.516 0.487 (0.454) (12.167) (-2.947) (1.975) (0.645) (0.423) Denmark 1976.01 0.010 0.545 0.544 -0.001 -0.277 -0.089 0.322 (1.390) (6.655) (4.526) (-0.021) (-0.541) (-0.086) Finland 1988.01 0.000 0.777 -0.258 0.062 -0.118 -2.073 0.184 (0.025) (4.715) (-0.882) (1.183) (-0.082) (-0.494) x France 1976.01 0.003 0.925 0.710 -0.043 0.605 1.034 0.454 (0.336) (9.267) (5.109) (-1.731) (1.196) (0.792) Germany 1976.01 0.(15 0.609 0.793 -0.024 0.012 -1.213 0.406 (1.965) (4.630) (5.387) (-0.532) (0.020) (-1.159) HongKong 1976.01 0.010 1.018 -0.043 -0.012 -0.147 1.395 0.166 (0.738) (4.439) (-0.142) (-0.335) (-0.137) (0.709) Ireland 1988.01 --0.005 1.061 0.039 -0.013 -0.361 3.630 0.438 (-0.301) (6.190) (0.173) (-0.223) (-0.304) (1.029) Italy 1976.01 0.018 0.754 0.258 -0.088 0.093 -1.703 0.230 (1.851) (6.982) (1.438) (-1.881) (0.124) (-1.093) japan 1976.01 (.005 1.046 0.452 -0.047 -0.495 1.025 0.551 (0.687) (9.949) (3.724) (-1.947) (-1.028) (0.929) Netherlands 1976.01 0.015 0.820 0.277 0.057 0.018 -0.633 0.563 (3.269) (11.368) (3.107) (3.577) (0.050) (-0.869) New Zealand 1988.01 0.001 0.541 -0.176 0.059 2.501 -1.462 0.035 (0.047) (2.751) (-0.550) (1.103) (2.288) (-0.240) Norway 1976.01 -0.01( 1.103 0.097 0.150 1.261 2.971 0.332 (-1.021) (7.765) (0.529) (4.229) (1.715) (1.688) Singaporc and Malaysia 1976.01 -0.006 1.038 -(.270 0.004 -0.039 2.968 0.282 (-0.598) (5.563) (-1.134) (0.110) (-(.051) (1.95.5) Spain 1976.01 0.011 0.695 0.367 -0.100 -0.026 -1.071 0.271 (1.117) (5.162) (2.222) (-2.222) (-0.039) (-0.754) Sweden 1976.01 0.014 0.829 0.044 -0.082 0.900 -0.642 0.304 (1.910) (7.860) (0.332) (-2.835) (1.472) (-0.574) Switzerland 1976.01 0.017 0.712 0.613 0.011 -0.316 -1.642 0.563 (3.094) (7.811) (5.691) (0.454) (-0.695) (-1.893) United Kingdom 1976.01 0.002 1.060 0.262 0.005 0.111 1.569 0.494 (0.328) (12.329) (2.235) (0.223) (0.206) (1.496) United States 1976.01 0.011 0.953 -0.505 0.016 -0.188 -0.411 0.717 (2.896) (17.393) (-7.812) (1.006) (-0.631) (-0.757) Emerging markets Argentina 1976.01 0.076 0.010 -0.714 0.175 0.553 -2.997 -0.017 (1.541) (0.024) (-1.152) (1.234) (0.155) (-0.424) Brazil 1976.01 0.023 0.489 -0.613 -0.116 -1.428 -0.429 0.003 (0.836) (1.577) (-1.347) (-1.123) (-0.748) (-0.112) Chile 1976.01 0.028 0.028 0.297 -0.053 -1.578 1.238 -0.009 (1.804) (0.115) (0.978) (-1.045) (-1.276) (0.476) Colombia 1985.01 0.074 -0.117 0.378 -0.097 -3.129 -8.701 0.049 LV (3.812) (-0.681) (1.097) (-2.109) (-1.892) (-2.390) Greece 1976.01 0.022 0.161 0.680 -0.024 -2.327 -2.047 0.053 (1.539) (0.773) (2.469) (-0.411) (-2.069) (-1.130) India 1976.01 0.007 -0.085 0.474 0.058 0.862 1.572 0.021 (0.543) (-0.606) (2.465) (1.049) (1.177) (0.978) Indonesia 1990.01 0.053 -0.205 -0.107 -0.099 1.541 -13.933 -0.064 (1.511) (-0.429) (-0.150) (-1.051) (0.466) (-1.834) Jordan 1979.01 -0.009 0.090 0.412 -0.063 0.267 3.215 0.081 (-1.088) (0.934) (2.627) (-2.012) (0.421) (2.107) Korea, Rep. of 1976.01 0.032 0.600 -0.379 -0.022 0.553 -3.285 0.064 (2.507) (3.739) (-1.574) (-0.410) (0.579) (-1.394) Malaysia 1985.01 0.006 0.900 -0.759 0.010 0.793 0.845 0.251 (0.336) (4.882) (-3.138) (0.254) (0.681) (0.240) Mexico 1976.01 0.039 1.031 -1.136 0.043 2.556 -3.845 0.111 (1.860) (3.805) (-2.863) (0.752) (1.874) (-1.356) Nigeria 1985.01 -0.005 0.167 0.373 0.023 -0.484 0.820 -0.035 (-0.173) (0.707) (1.233) (0.599) (-0.274) (0.155) Pakistan 1985.01 0.033 -0.106 0.558 -0.010 -0.936 -3.717 0.027 (Table continues on the following page.) Table 5. (Continued) Starting Exchange rate year and World return' investment Commodity World business Inflatione Market month Intercept beta indexb beta pricesc beta cycled beta beta R2 (2.292) (-0.670) (2.107) (-0.237) (-0.642) (-1.473) Philippines 1985.01 0.089 0.607 -0.248 -0.148 -0.176 -13.076 0.173 (3.663) (2.572) (-0.860) (-2.972) (-0.119) (-2.173) Portugal 1986.02 -0.000 1.301 -0.146 0.030 0.808 7.162 0.113 (-0.009) (4.959) (-0.326) (0.622) (0.475) (1.051) Taiwan (China) 1985.01 0,018 0.859 -1.300 -0.151 -1.765 5.977 0.098 (0.609) (2.231) (-2.934) (-1.718) (-0.715) (0.776) Thailand 1976.01 0.031 0.369 -0.151 -0.061 0.322 -2.749 0.049 (2.769) (1.582) (-0.574) (-1.222) (0.450) (-1.671) 411 Turkey 1987.01 0.025 0.185 0.416 0.040 -1.476 3.675 -0.073 o (0.367) (0.377) (0.443) (0.263) (-0.349) (0.252) Venezuela 1985.01 0.021 -0.270 0.281 0.169 3.018 1.399 0.020 (0.646) (-0.869) (0.825) (1.689) (1.166) (0.162) Zimbabwe 1976.01 -0.007 0.139 0.590 0.041 2.065 1.677 0.035 (-0.499) (0.714) (2.341) (0.976) (1.876) (0.788) Note: All returns are calculated in U.S. dollars and are in excess of the thirty-day Eurodeposit rate. Results are reported for a linear regression of the excess market return on the world return, the exchange-investment index, and proxies for commodity prices, the world business cycle, and inflation. The intercept and slopes (betas) are reported with heteroskedasticity-consistent t-ratios (in parentheses). a. The MSCI value-weighted world-market portfolio in excess of the thirty-day Eurodollar deposit rate. b. The U.S. dollar return to holding a trade-weighted portfolio of Eurocurrency deposits in ten countries (Belgium, Canada, France, Germany, Italy, Japan, Netherlands, Sweden, Switzerland, and the United Kingdom) in excess of the thirty-day Eurodollar deposit rate (see Harvery 1994b for details of the construction of the indcx). c. The change in the U.S. dollar price of crudc oil at the wellhead in excess of the thirty-day Eurodollar deposit rate. d. The growth in OECD industrial production. e. IThe growth in OECI inflation. Source: The monthly returns for emerging markets are from the IFC EMDB. The industrial-market returns are from MSCI. Harvey 41 In another sense, in the context of an integrated global market, identical exposure to a source of risk in two different markets commands the same reward. The lack of a cross-sectional relation between the risk loadings and return performance could be symptomatic of market segmentation. As markets become more integrated, the cross-sectional correlation of risk exposures and expected returns should be higher. Lack of integration opens up the possibility that equities are inefficiently priced in some emerging markets. Interestingly, the global investment manager may not care. The manager may prefer to have the opportunity to purchase securities at a price lower than the implied value in an integrated world economy. The notions of underpricing and overpricing are vague without explicit refer- ence to an asset-pricing model. In a globally integrated economy, covariance- not variance-is priced. That is, in integrated capital markets, investors can diversify away much of the idiosyncratic or local market variance by holding stocks from many markets. As a result, increases in the country variance (which could be driven by local events) do not necessarily command increases in ex- pected returns. But in many of the emerging markets, there is a clear relation between average returns and volatility. Indeed, Harvey (1994c) shows that vari- ance in emerging markets explains more of the cross-section of expected returns than covariance. This suggests that many of the markets are not integrated. Global investors may not care if the market is integrated or segmented as long as they can access the market for investment. Indeed, global investors can en- hance their portfolio performance by holding emerging-market assets with high variance and high expected returns. The enhancement results from the ex- tremely high contribution to portfolio expected return per unit of covariance (not variance). Presumably, these opportunities would diminish as emerging markets become more integrated into the world economy. Although the ex- pected return-covariance ratio may drop as a result of integration and the cross- border equity arbitrage may also decrease, integration may also imply that the cost of capital decreases. That is, in a segmented capital market, the cost of capital is high because investors demand a premium for bearing the local, or idiosyncratic, risk. In integrated capital markets, the cost of capital may de- crease because compensation for idiosyncratic risk is not required. A lower cost of capital usually leads to additional foreign direct investment. V. CONCLUSIONS Recently a number of researchers have documented the low correlations be- tween emerging-equity-market returns and industrial-market returns. However, little is known about the risk exposure of equity investments in emerging markets. 42 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Indeed, in real world portfolio selection, investment portfolio weights are chosen subject to a number of constraints. These constraints usually involve the prohibition of short-selling any market, maximum position limits for each mar- ket or groups of markets, and limits on the portfolio exposures to certain sources of risk. For the last constraint, estimates of each market's risk exposures are needed. These risk constraints eliminate the possibility of choosing a global portfolio with a higher expected return than, for example, the Standard and Poor's 500 and with the same volatility-but with an oil beta of -3.00 (com- pared with the Standard and Poor's 500 oil beta of -0.30). In other words, on average, a 3 percent loss on the Standard and Poor's 500 portfolio would result if oil prices increased by 10 percent. In a portfolio with an oil beta of -3.00, the same increase in the price of oil would lead to, on average, a 30 percent loss in portfolio value. To many investors, this type of exposure is unacceptable. Hence, it is important not only to measure the global risk exposures of interna- tional markets, but also to use the estimated risk exposures in portfolio formation. This article has examined five global risk factors: the world-market equity return, the return on a foreign currency index, a change in the price of oil, growth in world industrial production, and the world inflation rate. Only a handful of emerging markets have been found to have significant exposures to these factors. For example, only one of twenty emerging markets was found to have a beta against the world market portfolio that exceeded unity. One implication of the risk analysis is that many of the emerging markets are not well integrated into the global economy. However, the time-series evidence suggests that a number of markets may be becoming increasingly integrated. Models that allow for time-varying integration of world capital markets are explored in Bekaert and Harvey (1994). Figure A-1. Correlation of Emerging-Market Returns with Returns from the MSCI World Market Portfolio Argentina Brazil 0.9- 0.9 - 0.7- 0.7 - 0.5- 0.5 - 0.3 - 0.3 - 0.1- 0.1 -0.3 -0.3 -0.5 8 I I I I I I I I I I I 92 1980 82 84 86 88 90 92 1980 82 84 86 [ 8 910 92 Harvey 43 Chile Colombia 0.9 - 0.9 - 0.7 - 0.7 - 0.5 - 0.5 - 0.3 - 0.3 - 0.1- 0.1 -0.1 -0.1 -0.~3 -0.3 1980 82 84 86 88 90 92 1989 90 91 92 Greece India 0.9 0.9 - 0.7 0.7 - 0.5 0.5 - 0.3 0.3- Oil 0.1 -01 -0.1 -0.3 -0.3 -0.5 . . . . . . -0.5 - , l2 1980 82 84 86 88 90 92 1980 82 84 86 88 90 92 Jordan Republic of Korea 0.9 -- 0.9 - 0.7- 0.7 - 0.5- 0.5- 0.3 0.3 - 0.1 0.1 -0.1 - -0.1 -0.3 -0.3 -o.5 -0.5 - 1983 84 85 86 87 88 89 90 91 92 1980 82 84 86 88 90 92 Malaysia Mexico 0.9- 0.9 l 0.7- 0.7 - 0.5- 0.5- 0.3- 0.3- 0.1- 0.1 -0.1 -0.1 - -0.3- -0.3 -0.5 -0.5 l l l l l l ll 1989 90 91 92 1980 82 84 86 88 90 92 (Figure continues on the following page.) 44 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Figure A-1 (continued) Nigeria Pakistan 0.9 - 0.9 0.7 - 0.7 - 0.5 - 0.5 - 0.3 - 0.3 - 0.1 - 0.1 - -0.1 -0.1 -0.3 -0.3 -0.5 -05 1989 90 91 92 1989 90 91 92 Philippines Portugal 0.9- 0.9 0.7 - 0.7 - 0.5 - 0.5 - 0.3 - 0.3- 0.1 - 0.1 - -0.1 -0.1 -0.3 -0.3 -0.5 -05 1989 90 91 92 1991 92 93 Taiwan (China) Thailand 0.9- 0.9 0.7 - 0.7 - 0.5 - 0.5 - 0.3 - 0.3 - 0.1- 0.1- -0.1 -0.1 -0.3 -0.3 -o.5 --7 \ -0.5-I , i;,lE 1989 90 91 92 1980 82 84 86 88 90 92 Turkey Venezuela 0.9 -0.9 0.7 - 0.7 - 0.5 -0.5- 0.3 0.3 - 0.1 - 01 - -0.1 - -0.1 - -0.3 -0.3 -0.5 -0.5- 1991 92 1989 90 91 92 Harvey 45 Zimbabwe 0.9 - 0.7 - 0.5 - 0.3 - 0.1 -0.1 -0.3 -0.5- 1980 82 84 86 88 90 92 Note: Values are five-year moving correlations with the MSCI value-weighted world market portfolio in excess of the thirty-day Eurodollar deposit rate. The correlations are based on monthly returns calculated in U.S. dollars. Source. Author's calculations. Figure A-2. Correlation of Emerging-Market Returns with Currency Returns from Ten Industrial Markets Argentina Brazil 0.9 - 0.9 0.7 - 0.7 0.5 - 0.5 0.3 - 0.3 0.1 0.1 -0.1- -0.1 -0.3 - -0.3 -0.5 --0.5 1980 82 84 86 88 90 92 1980 82 84 86 88 90 92 Chile Colombia 0.9- 0.9 0.7 - 0.7 - 0.5 - 0.5 - 0.3- 0.3- 0.1- 0.1 m -0.1 -0.1 - -0.3 -0.3 -0.5- , , , R, , -0.5- 1980 82 84 86 88 90 92 1989 90 91 92 (Figure continues on the following page.) 46 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Figure A-2 (continued) Greece India 0.9 - 0.9 - 0.7 - 07 - 0.5 -0.5- 0.3- 0.3- 0.1 - 0.1 - -0.1 --. -0.3 - -0.3 -0 .5 1 , I I , I I I I- i I I I l I 1980 82 84 86 88 90 92 1980 82 84 86 88 90 92 Jordan Republic of Korea 0.9- 0.9 0.7 - 0.7 - 0.5- 0.5 0.3- 0.3- 0.1 0.1 -0.1 - 0.1 -0.3 -0.3 - -0.5 - l l l l l l l l l-0.5 - l l l ll l l l 1983 84 85 86 87 88 89 90 91 92 1980 82 84 86 88 90 92 Malaysia Mexico 0.9 0.9 - 0.7 0.7 - 0.5 0.5 - 0.3 0.3 - 0.1 0.1 -0.1 - -0.1 ' -0.3 -0.3 - -0.5 ----- -0.5 - I I I i l l 1989 90 91 92 1980 82 84 86 88 90 92 Nigeria Pakistan 0.9 - 0.9 0.7 - 0.7 0.5 - 0.5 0.3 - 0.3 0.1- 0.1 -0.1 -0.1 -0.3 -0.3 -o.S- l l l -0.5 1989 90 91 92 1989 90 91 92 Harvey 47 Philippines Portugal 0.9 - 0.9 - 0.7 - 0.7 - 0.5 - 0.5 - 0.3 - 0.3 - 0.1 - 0.1- -0.1 -0.1 -0.3 -0.3 -0.5 - -0.5- 1989 90 91 92 1991 92 93 Taiwan (China) Thailand 0.9 - 0.9 - 0.7 - 0.7 - 0.5 - 0.5 - 0.3 - 0.3 - 0.1 - 0.1 - -0.1 _ -0.1 - -0.3 -0.3 1989 90 91 92 1980 82 84 86 88 90 92 Turkey Venezuela 0.9 - 0.9 - 0.7 - 0.7 - 0.5 - 0.5 - 0.3 - 0.3 - 0.1- 0.1- -0.1 -0.1 - -0.3 -0.3 - -0.5 - -0.5 - I I 1991 92 1989 90 91 92 Zimbabwe 0.9 - 0.7 - 0.5 - 0.3- 0.1 -0.1 - -0.3 - -0.5- 1980 82 84 86 88 90 92 Note: Values are five-year moving correlations with the U.S. dollar return to holding a trade-weighted portfolio of Eurocurrency deposits in ten markets (Belgium, Canada, France, Germany, Italy, Japan, Netherlands, Sweden, Switzerland, and the United Kingdom) in excess of the thirty-day Eurodollar deposit rate: see Harvey (1994b) for details of the construction of the index. The correlations are based on monthly returns calculated in U.S. dollars. Source: Author's calculations. 48 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Adler, Michael, and Bernard Dumas. 1983. "International Portfolio Selection and Cor- poration Finance: A Synthesis." Journal of Finance 38(3, September):925-84. Bailey, Warren, and Julapa jagtiani. 1994. "Foreign Ownership Restrictions and Stock Prices in the Thai Capital Market." Journal of Financial Economics 36(1, August):57- 87. Bansal, Ravi, David Hsieh, and S. Viswanathan. 1993. "A New Approach to Interna- tional Asset Pricing." Journal of Finance 48(5, December):1719-47. Bekaert, Geert. 1995. "Market Integration and Investment Barriers in Emerging Equity Markets." The World Bank Economic Review 9(1):75-108. Bekaert, Geert, and Campbell R. Harvey. 1994. "Time-varying World Market Integra- tion." NBER Working Paper 4843. Cambridge, Mass.: National Bureau of Economic Research. Bodurtha, James N., Jr. 1990. "International Factors and U.S. Equity Excess Returns." Working paper. University of Michigan, School of Business Administration, Ann Ar- bor, Mich. Processed. Brown, Stephen J., and Toshiyuki Otsuki. 1993. "Risk Premia in Pacific-Basin Capital Markets." Pacific Basin Finance Journal 1(3, September):235-62. Buckberg, Elaine. 1995. "Emerging Stock Markets and International Asset Pricing." The World Bank Economic Review 9(1 ):S1-74. Capital International Perspective, S.A., and Morgan Stanley & Co. Various issues. Morgan Stanley Capital International Perspective. New York, N.Y.: Morgan Stanley. Claessens, Stijn, Susmita Dasgupta, and Jack Glen. 1995. "Return Behavior in Emerging Stock Markets." The World Bank Economic Review 9(1):131-52. Cumby, Robert E., and Jack D. Glen. 1990. "Evaluating the Performance of Interna- tional Mutual Funds." Journal of Finance 45(2, June):497-521. Divecha, Arjun, Jaime Drach, and Dan Stefek. 1992. "Emerging Markets: A Quantita- tive Perspective." Journal of Portfolio Management 19(1, Fall):41 -50. Dumas, Bernard. 1994. "Partial-Equilibrium vs. General Equilibrium Models of Inter- national Capital Market Equilibrium." In Frederick van der Ploeg, ed., The Handbook of International Macroeconomics. London: Basil Blackwell. Dumas, Bernard, and Bruno Solnik. 1994. "The World Price of Foreign Exchange Risk." NBER Working Paper 4459. Cambridge, Mass.: National Bureau of Economic Research. Ferson, Wayne E., and Campbell R. Harvey. 1993. "The Risk and Predictability of International Equity Returns." Review of Financial Studies 6(3):527-66. . 1994. "Sources of Risk and Expected Returns in Global Equity Markets." NBER Working Paper 4622. Cambridge, Mass.: National Bureau of Economic Research. Hansen, Lars P. 1982. "Large Sample Properties of Generalized Method of Moments Estimators." Econometrica 50(4, July):1029-54. Harvey, Campbell R. 1991. "The World Price of Covariance Risk." Journal of Finance 46(4, March):111-57. -. 1994a. "Conditional Asset Allocation in Emerging Markets." NBER Working Paper 4623. Cambridge, Mass.: National Bureau of Economic Research. Harvey 49 - 1994b. "Global Risk Exposure to a Trade-weighted Foreign Currency Index." Working paper. Duke University, Fuqua School of Business, Durham, N.C. Processed. . 1994c. "Predictable Risk and Returns in Emerging Markets." NBER Working Paper 4621. Cambridge, Mass.: National Bureau of Economic Research. Harvey, Campbell R., and Guofu Zhou. 1993. "Intemational Asset Pricing with Alterna- tive Distributional Assumptions." Journal of Empirical Finance 1(1):107-31. Harvey, Campbell R., Bruno H. Solnik, and Guofu Zhou. 1994. "What Determines Expected International Asset Returns?" NBER Working Paper 4660. Cambridge, Mass.: National Bureau of Economic Research. Hodrick, Robert. 1981. "International Asset Pricing with Time-varying Risk Premia." Journal of International Economics 11(4, November):573-87. IFC (International Finance Corporation). 1993. Emerging Stock Markets Factbook. Washington, D.C. Jorion, Philippe. 1991. "The Pricing of Exchange Rate Risk in the Stock Market," Journal of Financial and Quantitative Analysis 26(September):363-76. Lintner, John. 1965. "The Valuation of Risk Assets and the Selection of Risky Invest- ments in Stock Portfolios and Capital Budgets." Review of Economics and Statistics 47(1, February):13-37. Merton, Robert C. 1973. "An Intertemporal Capital Asset Pricing Model." Economet- rica 41(5, September):867-87. Mullin, John. 1993. "Emerging Equity Markets in the Global Economy." Federal Re- serve Bank of New York Quarterly Review 18(2, Summer):54-83. Roll, Richard. 1977. "A Critique of the Asset Pricing Theory's Tests. Part I: On Past and Potential Testability of the Theory." Journal of Financial Economics 4(2, March):129- 76. Roll, Richard, and Stephen A. Ross. 1994. "On the Cross-sectional Relation between Expected Returns and Betas." Journal of Finance 49(1, March):101-21. Ross, Stephen A. 1976. "The Arbitrage Theory of Capital Asset Pricing." Journal of Economic Theory 13(3, December):341-60. . 1977. "The Capital Asset Pricing Model (CAPM), Short Sale Restrictions and Related Issues.' Journal of Finance 32(1, March): 177-83. Ross, Stephen A., and Michael Walsh. 1983. "A Simple Approach to the Pricing of Risky Assets with Uncertain Exchange Rates." In Robert G. Hawkins, Richard M. Levich, and Clas Wihlborg, eds., The Internationalization of Financial Equity Markets and National Economic Policy. Greenwich, Conn.: )AI Press. Scholes, Myron, and Joseph Williams. 1977. "Estimating Betas from Nonsynchronous Data." Journal of Financial Economics 5(3, December):309-27. Sercu, Piet. 1980. "A Generalization of the International Asset Pricing Model:' Finance 1(1):91-135. Sharpe, William F. 1964. "Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk: Introduction." Journal of Finance 19(3, September):425-42. . 1984. "Factor Models, CAPM'S and the APT." Journal of Portfolio Management ll(Fall):21-25. Solnik, Bruno H. 1974. "An Equilibrium Model of the International Capital Market?' Journal of Economic Theory 8(4, August):500-24. . 1983. "International Arbitrage Pricing Theory." Journal of Finance 38(2, May):449-57. 50 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO, I Stone, Douglas. 1990. "The Emerging Markets and Strategic Asset Allocation." Working paper. Frank Russell Co., Tacoma, Wash. Processed. Stulz, Rene M. 1981. "A Model of International Asset Pricing." Journal of Financial Economics 9(4, December):383-406. 1993. "International Portfolio of Choice and Asset Pricing: An Integrative Sur- vey." Working paper. Ohio State University, Max M. Fisher College of Business, Col- umbus, Ohio. Processed. Wilcox, Jarrod W. 1992. "Taming Frontier Markets." Journal of Portfolio Management 19(1, Fall):51-55. THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1: 51-74 Emerging Stock Markets and International Asset Pricing Elaine Buckberg This article investigates whether emerging stock markets are now part of the global financial market and characterizes return behavior in these markets. Tests of the condi- tional International Capital Asset Pricing Model (ICAPM) reveal that eighteen of the twenty largest emerging markets were integrated with the world market between De- cember 1984 and December 1991, but that many of the same markets reject the model when data for 1977-84 are used. These results suggest that large capital inflows from industrial economies, beginning in the late 1980s, caused prices in emerging markets to reflect covariance risk with the world portfolio, thus inducing their consistency with the ICAPM. The year 1981 saw the formation of the first emerging-market country funds to attract investors to purchase stocks in developing economies; today, big Wall Street brokerages trade issues from developing economies, and the Financial Times tracks fourteen emerging markets daily. Stock markets offer a promising channel through which developing economies can attract foreign capital to fund investment and growth. So-called emerging markets are large and expanding rapidly, yet they continue to exhibit very different risk and return characteristics fromn comparably sized industrial markets. The combination of supranormal yields, highly autocorrelated returns, and volatile prices suggest that these mar- kets may be inefficient, that excess returns may exist, and that emerging markets may not be fully integrated into global capital markets. More important, con- sistently high rates of return in these economies translate into a high cost of capital, which limits the stock market's role as a source of private financing. Elaine Buckberg is with the Western Hemisphere Department at the International Monetary Fund. She thanks J. Joseph Beaulieu, Andrew Bernard, Tasneem Chipty, Stijn Claessens, Rudiger Dornbusch, Raul Livas Elizando, Stanley Fischer, Campbell Harvey, John Heaton, Ruth Judson, Charles Kramer, Paul Krugman, Mark Stone, Aaron Tomell, and the M.I.T. International Breakfast and Money Lunch groups for helpful discussions. Several anonymous referees and seminar participants at Brandeis Univer- sity, the Brookings Institution, the Congressional Budget Office, the Federal Reserve Bank of New York, the Federal Reserve Board, the International Monetary Fund, the Massachusetts Institute of Technology, RAND Corporation, and the World Bank Conference on Portfolio Investment in Developing Countries, Washington, D.C., September 9-10, 1993, provided useful suggestions. Peter Tropper, of the IFC'5 Emerging Markets Group, and James Poterba generously provided data, and a National Science Founda- tion Graduate Fellowship provided financial support for this research. © 1995 The International Bank for Reconstruction and Development/THE WORLD BANK 51 52 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 This article presents strong evidence that emerging markets now adhere to the International Capital Asset Pricing Model (ICAPM) in that expected returns re- flect expected covariance with the world portfolio. In tests of the conditional ICAPM using monthly data from December 1984 to December 1991, the model cannot be rejected for eighteen of the twenty emerging markets. Tests on data for 1977-84, however, indicate that few emerging markets were integrated before the late 1980s. These results suggest that large capital inflows from industrial economies, catalyzed by Southeast Asian stock market booms around 1987, induced ICAPM relationships to evolve in many previously segmented markets. Section I establishes characteristics of emerging markets that may affect whether standard models can accurately price emerging-market stocks. Section II covers the econometric methodology for estimating the ICAPM, and section III documents the data. Section IV presents results of the ICAPM estimates for 1985- 91; section V presents estimates using data for 1977-84. Section VI offers conclusions. I. CHARACTERIZING EMERGING MARKETS Although emerging stock markets have risen swiftly both in value and as a source of financial flows, they remain different in many ways from the mature markets in industrial economies. This section characterizes return patterns and policy attributes that distinguish emerging markets from their industrial counterparts. Price and return patterns reveal some of the most crucial differences between emerging and mature markets. Many markets in developing economies offer yields far in excess of industrial-market returns and low-to-negative correlation with the world market. Both of these facts suggest that unexploited profit op- portunities may exist. High autocorrelation in returns, characteristic of specula- tive inefficiency, indicates that lagged prices may contain information about future returns; volatile stock prices also suggest inefficiency. In addition, rapidly rising price-earnings ratios (see table 1) signal ongoing transition in emerging markets. From 1988 to 1991, price-earnings ratios more than doubled in seven of twenty emerging markets tracked by the International Finance Corporation's Emerging Markets Data Base (IFC EMDB), and more than tripled in four (Argen- tina, Chile, Pakistan, and Turkey). During the same period, the price-earnings ratio of the Morgan Stanley Capital International Perspective (MscI) world index rose only 22 percent (Capital International Perspective, S.A. and Morgan Stan- ley & Co., various issues). High price volatility in emerging markets may stem from small-market effects and informational imperfections. With few trades occurring, information about stock value-and therefore stock prices-tends to be noisy. Moreover, limited reporting requirements in many markets mean that investors typically have less information about firms and receive less-frequent updates than do investors in Buckberg 53 Table 1. Price-Earnings Ratios in Selected Markets, 1986-91 Market 1986 1987 1988 1989 1990 1991 1991/1986 Emerging markets Argentina 16.0 3.8 11.3 22.1 3.1 38.9 2.4 Brazil 4.2 15.4 8.0 8.3 5.3 7.7 1.8 Chile 5.3 5.0 4.4 5.8 8.9 17.4 3.3 Colombia 8.3 11.6 8.8 7.0 10.7 26.1 3.2 Greece - 30.5 10.6 24.3 26.2 10.4 0.3a India 18.0 22.1 21.5 18.3 20.6 13.9 0.8 Indonesia - - - - 30.8 26.7 - Jordan 12.9 12.8 17.3 14.9 8.2 10.6 0.8 Korea, Rep. of 25.7 21.7 39.5 38.6 21.5 17.6 0.7 Malaysia 32.7 30.7 24.1 30.8 23.0 14.6 0.4 Mexico 10.5 6.2 5.0 10.7 13.2 24.4 2.3 Nigeria 5.8 4.9 6.1 7.0 7.0 9.7 1.7 Pakistan 8.2 6.9 9.4 8.4 8.5 23.9 2.9 Philippines 4.4 8.9 9.9 18.5 24.5 16.2 3.7 Portugal 24.8 27.2 26.5 21.4 15.5 18.9 0.8 Taiwan (China) 12.0 13.0 40.2 51.2 44.4 14.5 1.2 Thailand 12.5 10.5 12.6 23.1 10.9 17.2 1.4 Turkey - 19.8 2.6 17.6 22.5 21.6 1.1a Venezuela 9.4 16.9 11.5 6.4 29.3 30.5 3.2 Zimbabwe 4.2 7.0 4.2 7.0 12.0 8.4 2.0 Industrial markets New York 16.8 15.4 12.2 14.7 15.2 21.9 1.3 Tokyo 47.3 58.3 58.4 70.6 39.8 37.8 0.8 - Not available. Note: The price-earnings ratio is the ratio of end-of-month price to trailing twelve-month earnings. a. 1991/1987. Source: Author's calculations based on data from the IFC EMDB. industrial markets. Uncertainty about the financial condition of firms may intro- duce high variance in expected returns. In a small securities-market, trades that are small by New York standards may adversely affect prices; limitations on the size of transactions may prevent investors from fully exploiting all available information and may explain why emerging-market returns contain a large forecastable component. Many developing economies impose capital controls that insulate the local stock exchange from global markets. Markets in which neither foreign nor local capital can freely cross borders lack the capital flows necessary to induce an ICAPM rela- tionship (assuming that capital controls prove effective in preventing financial flows). All but five of the twenty markets considered here bar nationals from hold- ing foreign securities; as a result, emerging markets are populated by investors who cannot diversify internationally. All but two of the markets banned or severely re- stricted foreign investment during part of the 1985-91 sample period. Because bar- riers to foreign investment limit the number of market participants, they also re- strict the capital supply and thus the market's capital formation potential capital. A look at actual openings to foreign investment shows that market capitalization rose permanently within a few years of many of the earlier openings, as hap- pened in Chile, Mexico, the Philippines, Turkey, and Venezuela (table 2). I Table 2. Year-End Price-Earnings Ratios and Turnover Ratios for the Year before and the Year of Selected Market Openings Price-earnings ratio Market capitalization Turnover ratio Before After Before After Before After Market Opening date opening opening opening opening opening opening Nature of opening Argentina October 1991 3.11 38.89 3.27 18.5 33.6 45.3 Full opening Brazil September 1987 4.24 15.38 42.1 42.7 74.4 41.5 Country fund admitted May 1991 5.34 7.65 16.3 42.7 23.6 22.0 Full opening Chile October 1989 4.4 5.82 6.8 28.0 6.3 8.8 Country fund admitted Colombia October 1991 10.66 26.08 1.4 4.0 5.6 7.1 Full opening Indonesia March 1989 - - 0.2 6.8 2.5 38.6 Minor restrictions on entry and exit; previously, wholly closed Mexico May 1989 5.04 10.66 13.8 101.2 51.7 3.3.3 Restrictionsreduced Pakistan June 1991 8.53 23.87 2.9 7.3 8.7 12.6 Full opening Philippines October 1989 9.92 18.5 4.3 10.2 24.4 29.1 C ountry fund admitted Portugal January 1986 - - 0.2 9.6 4.0 7.1 Full opening Turkey December 1989 2.62 17.64 1.1 15.7 5.5 19.0 Country fund admitted Venezuela December 1988 16.91 11.45 2.3 11.2 8.1 10.9 Minorrestrictions on entry and repatriation; previously, special share classes only - Not available. Source: Author's calculations based on data from the trc EMOFm and IMF (various years). Buckberg 55 Because many of the openings occurred in 1991, it is too early to judge whether market capitalization will sustain a rise. After some lag, the number of listings may also expand as more firms go public, able to meet the now-reduced costs of raising capital in the stock markets. Turkey's market had 50 listings in the three years up to and including its 1989 opening; by 1992, 145 firms were traded. Indonesia had 24 listings one year prior to its 1989 opening, 57 by the end of 1989, and 155 by the end of 1992. Turnover ratios also rose after 9 of 12 openings, with the exceptions of Mexico's marginal 1989 opening and Brazil's 1987 and 1991 openings (see table 2). Those markets that have liberalized restrictions on foreign investment have consistently experienced huge price increases. In 1991 four countries eliminated closed-market regimes in favor of policies of free entry and exit. In the period from December before the opening to December after the opening each country witnessed increases in the price-earnings ratio of between 40 and 1,000 percent; Argentina witnessed an increase of 1,150 percent; Brazil, 43 percent; Colombia, 145 percent; and Pakistan, 180 percent (see table 2). Argentina previously ad- mitted foreign investors but prohibited repatriation of capital before three years had passed from the date of initial investment. Brazil admitted investment only into special share classes or through a country fund and also restricted repatriation. Because broader economic reforms typically accompany financial openings, price-earnings ratios may in some cases rise in the short run in expectation of future gains in earnings and then return to their previous levels once earnings increase. However, in a number of cases price-earnings ratios have risen continu- ously for several years after the opening. The continuous rise suggests that the markets were historically undervalued and that their opening reduced the cost of capital permanently; the 1989 openings in Chile, Mexico, and the Philippines all provide examples. Although a trade opening should reduce rents in monopo- lized local industries, earnings may actually rise in competitive industries. The entry of foreign firms increases competition in the local market, forcing domes- tic firms to become more productive, efficient, and competitive to survive. As a result, the government can no longer appropriate rents from local producers as high as those before the opening, whether through taxation, fees, or kickbacks. The average return retained by local firms may therefore actually rise, and earnings may also rise as the firms become more streamlined and efficient. Returns in emerging markets may also be affected by the fact that the policy environment tends to be less stable in developing than in industrial economies. Uncertainty about policies governing future earnings may introduce "peso prob- lems" if investors fear some low-probability policy shift that will dramatically reduce returns: a devaluation, closing of the stock market, expropriation, or imposition of capital controls that bar repatriation of capital or profits. The potential for any of these events to occur reduces the expected liquidity of holding emerging-market stocks and may cause what appear to be excess returns in a period in which the bad state fails to materialize. 56 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I II. SPECIFICATION OF THE CONDITIONAL ICAPM The International Capital Asset Pricing Model (ICAPM) states that if emerging markets are part of a global market, then each market's expected returns should be proportional to that market's covariance with a capitalization-weighted world portfolio. If emerging markets are not integrated into the world market- that is, if returns do not reflect covariance with the world portfolio-then by adding issues from emerging markets to their portfolios, investors can both reduce overall risk and increase expected returns. As investors in industrial economies increasingly participate in emerging markets (and investors in devel- oping economies increasingly participate in industrial markets), the investors should exploit and subsequently eliminate any excess returns relative to the ICAPM. This article takes as given the ICAPM'S validity as a description of mature markets because of its success at describing return behavior on the exchanges of industrial economies (see Harvey 1991 and Dumas and Solnik 1992). Treating each economy's market index as a portfolio, the article tests whether market returns in developing economies are consistent with the model's predictions. The study examines data for twenty markets during 1985-91 and also for a smaller, longer sample during 1977-91 to investigate whether emerging markets became more integrated after inflows of equity capital from industrial economies took off in the late 1980s. The article's objective is to determine to what extent emerging markets behave like industrial markets in relation to the world portfo- lio and to examine how the relation between emerging markets and the world portfolio has changed over time. Numerous studies have used international data to test the Sharpe-Lintner asset-pricing model (Sharpe 1964 and Lintner 1965). The model assumes that investors divide their wealth between a riskless asset and risky assets or stocks in proportions that depend on each investor's risk aversion. Extending the model internationally allows investors to choose among stocks from many countries rather than from a single stock market; the market portfolio now includes all the assets in the world. In choosing a portfolio of risky assets, investors seek a high expected-return-to-variance ratio. The ICAPM states that the expected return on any given risky asset in excess of the safe rate is proportional to the expected return on the market in excess of the safe rate: E(Rj,) - Rfr = fj [E(Rwt) - Rf1] (1) - cov(R(R,, Rjt) [E(Rw.t) - Rf1] var(R,,) where Rjt is the total return on some asset j, Rf, is the rate of return on the risk- free asset, ,j is the proportionality factor [{j = cov(R,t, Rj,)/var(Rw,)], and R_, is the total return on the world, or market, portfolio. Under the model, optimiz- ing behavior leads investors to care only about covariance risk with the market Buckberg 57 portfolio and about no other sources of risk; the ICAPM relation should evolve out of investors' efforts to diversify risk. A stock or portfolio is integrated with the defined market in an ICAPM sense if its returns are consistent with the model. Many empirical studies reduce the model's complexity by creating portfolios of stocks and allowing the investors to divide their assets between the riskless asset and the risky portfolios. This study takes that approach, using a representative portfolio for each emerging market. A weakness of the ICAPM is that the model assumes independent and identi- cally distributed returns and ignores the presence of serial correlation. Given serially correlated returns, most investors would likely prefer a more compli- cated, intertemporal strategy to the ICAPM. In fact, the presence of serial correla- tion in industrial markets has been established by Cutler, Poterba, and Summers (1991) and emerging-market returns display strong serial correlation at short intervals. Because the objective here is to evaluate whether emerging markets behave like industrial markets in terms of the fit of the ICAPM, the estimates will remain faithful to the standard formulation of the model and will not correct for serial correlation. Several recent studies test asset-pricing models on data from emerging mar- kets. De Santis (1993) finds that adding assets from emerging markets to a benchmark portfolio of U.S. assets induces a statistically significant change in the volatility bounds. This result suggests that standard asset-pricing models that perform well on data on assets from industrial economies may fail to price assets traded on exchanges in emerging markets. Claessens, Dasgupta, and Glen (1993) compare the fit of a single-factor ICAPM and a multifactor model for eighteen markets during 1988-92. They conclude that the additional factors generate a better fit; the improvement is not, however, shown to be statistically significant. Moreover, even the multifactor model is rejected for every economy except Brazil. The poor fit of either model may be caused by the short sample period. Errunza, Losq, and Padmanabhan (1992) attempt to classify eight emerging markets during 1976-87 as integrated, mildly segmented, or seg- mented and find that the hypothesis of mild segmentation is rejected least fre- quently. However, the fact that India rejects all three models raises questions about Errunza, Losq, and Padmanabhan's classification scheme; and their use of the U.S. market to represent the world portfolio, during a period when the U.S. market represented two-fifths of world capitalization, may have affected their results. Following recent work by Harvey (1989, 1991) and Dumas and Solnik (1992), this study uses conditional or expectational asset pricing to test the Sharpe-Lintner model. "Conditional" refers to the use of conditioning information-some information set Z,-1-to calculate expected moments and to test properly the ICAPM as a relation between expected returns and ex ante risk. Earlier tests of the ICAPM used realized or ex post return and covariance data and thereby failed to capture the ex ante relationship described by the theoretical model. S8 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 The conditional formulation restricts the conditionally expected return on an asset (based Zt-1) to be proportional to the asset's covariance with the market portfolio, yet allows expected returns to vary over time:' (2) E[rj,l Z,-.] = OE[rmt, Z,-l] where Oj -I= cov(ri>, r~, I Zt-1) h ,B co~va(r,t, Iw Z, - 1 ) where rj, and rw, represent the excess returns on an asset j and the world portfolio, respectively. The proportionality factor fj, which is also calculated using conditional moments, represents the price of covariance risk or the ex- pected return compensation investors demand for taking on a unit of covariance risk. The modeling and specification borrow from Harvey ( 1991 )). The conditional ICAPM successfully explains both time-series variation and cross-sectional differences in Organization for Economic Cooperation and De- velopment (OECD) market returns.2 Harvey (1991) finds returns consistent with the ICAPM in fourteen of seventeen industrial country markets-an 82 percent success rate-during 1970-89; only Austria, Denmark, and Japan reject the model. Moreover, the betas he obtains for each market correspond in ranking to the ranking of mean returns across these markets, with the exception of under- estimating the Japanese return. The fact that the model fails for Austria and Denmark, the two smallest OECD markets, suggests that illiquidity or other small-market effects may impede the evolution of ICAPM relationships. Harvey also runs a multimarket test over all Group of Seven nations and is unable to reject the model. Using a slightly different formulation, Dumas and Solnik (1992) analyze U.S., German, and Japanese returns simultaneously and are unable to reject the model as a description of return patterns in all three markets. The estimates in this article impose the restriction that a is constant over time and test whether expected returns in the local market are proportional to the expected return on a benchmark portfolio, in this case the world portfolio. Hansen's (1982) generalized method of moments (GMM) is used to estimate a constant proportionality factor j, (3) ejt= -rj -rt 0, which is equivalent to equation 2. In the orthogonality condition (4) f (r'tX, O) = e", Zt-l rj, and r.,, denote the local and world excess returns in U.S. dollars, and e1t is the vector of errors from estimating equation 3. Zt-, contains I information vari- ables (instruments) and is a subset of the t - 1 information set. The rise of international capital flows to these markets suggests that ( may have changed 1. Conditional variances and covariances may also be allowed to vary over time, with sufficient data. The short sample of emerging-markets data does not permit estimation of time-varying variances or covariances. See Harvey (1991) for an example. 2. The unconditional formulation of the model has done poorly. Buckberg 59 over time. However, the brevity of the data series currently available prohibits estimation of time-varying betas (because of inadequate degrees of freedom). Estimation of time-varying betas will be possible and advisable in future work, when longer time series become available. Single-market tests determine whether the time-series behavior of local returns accords with equation 4 and identify which emerging markets do not reflect covariance with the world market. However, the single-market tests do not impose one of the ICAPM's restrictions: that the price of covariance risk must be the same for each market. The price of covariance risk is the conditionally expected world-market return divided by the conditional variance of the world market return: E[r.,IZ,- f = K and var[r,,,IZ,-i] ov = cov[rjt, rt IZt-11 *K for allj. A stricter test of the model, including the cross-asset restriction, is obtained by estimating equation 3 on multiple markets in a system. For details of GMM estimation see the attached technical appendix or Ogaki (1992). III. DATA AND SUMMARY STATISTICS The data for stock exchanges in developing economies come from the Emerg- ing Markets Data Base (EMDB) compiled by the International Finance Corpora- tion (IFC, various years). The IFC has constructed its own indexes for every market. The indexes typically include 10 to 20 percent of listed stocks, selected on the basis of high trading volume or large capitalization or to give the index an industry composition representative of the market overall. Like many industrial- market indexes, the IFC indexes are biased toward local blue-chip stocks, and this somewhat diminishes their representativeness. All empirical work in this article uses the IFC'S representative indexes and treats each as a stock portfolio. I choose to work with the IFC indexes instead of the locally calculated market indexes because the former offer greater comparability across markets and are, I believe, more carefully calculated. Monthly index (and stock) data are available for seventeen markets from December 1984 to December 1991; data for Portu- gal, Turkey, and Indonesia start in December of 1986, 1987, and 1988, respec- tively. The data base includes series dating back to December 1976 for ten markets; prior to 1985, however, fewer stocks were sampled to create the mar- ket indexes. The ten markets with longer data series are those of Argentina, Brazil, Chile, Greece, India, Jordan (beginning in 1978), the Republic of Korea, Mexico, Thailand, and Zimbabwe. Market data for industrial nations come from MSCI. MSCJ'S indexes represent European, Japanese, U.S., and world market portfolios. The MSCI'S capitalization-weighted thirteen-market index for Europe includes Austria, Bel- 60 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I gium, Denmark, Finland, France, Germany, Italy, the Netherlands, Norway, Spain, Sweden, Switzerland, and the United Kingdom. Ideally, the world index would include all the emerging-market stocks used in this analysis as well as stocks from the industrial markets, but neither the MSCI index nor any other standard world index does so. Like the EMDB indexes, MSCI indexes contain only selected stocks and are weighted toward large capitalization issues; studies have found, however, that MSCI indexes are highly correlated with commonly quoted national indexes such as the New York Stock Exchange index (99.1 percent correlation) or the Nikkei 255 (93.8 percent; Harvey 1991). The MSCI indexes differ from EMDB indexes in that investment companies and foreign-domiciled companies are excluded to avoid double counting. However, other forms of double counting still pose problems. According to McDonald (1989) and French and Poterba (1989), MSCI's world index weights Japanese stocks too heavily because it fails to correct for the extensive corporate cross-ownership. Unfortunately, no world index corrected for cross-holding exists (see Harvey 1991). All returns in industrial and emerging markets are calculated as the excess return in dollars over the holding yield on the U.S. Treasury bill closest to thirty days to maturity on the last trading day of the month. Government bond data come from the Center for Research on Securities Prices (CRSP) Government Bond File. The U.S. inflation component in dollar stock returns cancels out against the inflation component in the Treasury bill return. Ideally, the set of instrumental variables should replicate the information investors use to predict prices. For each period, actual rates of return set during the previous period serve as conditioning information. This choice allows ex- pected returns to vary over time. The common-instrument set, identical for all emerging markets, contains information about industrial-market returns only; local-instrument sets also include the lagged rate of return in the local market that investors would likely consider in their investment decision. Preliminary tests also considered the lagged local-dividend yield and the lagged return on the dollar-local currency exchange rate as instruments. In regressions of returns on lagged instruments (not reported), the addition of the lagged local return sub- stantially improves return prediction. The selection of common instrumental variables draws on studies of U.S. stock returns and returns in other industrial economies. Following Harvey (1991), the information set includes the lagged world excess stock return, a dummy variable for the month of January, the dividend yield on the MSCI world index, a U.S. term structure premium, and the U.S. default risk-yield spread. The lagged return on the world market portfolio is included; Fama (1970) and many studies since have found autocorrelation in returns. The January dummy is included because Keim (1983) and Gultekin and Gultekin (1983) find system- atically higher January returns in the United States and other industrial econ- omies; Claessens, Dasgupta, and Glen (1993) find statistically significant Janu- ary effects in seven of eighteen emerging markets. The U.S. term structure Buckberg 61 premium is calculated as the one-month return to holding a three-month Trea- sury bill less the return on a bill thirty days to maturity. Campbell and Hamao (1989) show that measures of the term structure statistically explain returns in Japan and the United States. The default risk spread, measured here as the difference between the yield on a Moody's Baa bond and a Moody's Aaa bond, is included because of the findings of Keim and Stambaugh (1986) and Fama and French (1989) that the junk bond spread helps in predicting stock market returns. The final instrumental variable is the dividend yield on the world port- folio. Fama and French (1988, 1989) show the importance of this term in predicting U.S. returns; Cutler, Poterba, and Summers (1989) find that lagged dividend yields also influence international returns. Again, all returns are calcu- lated in excess of the return on a Treasury bill thirty days to maturity. Table 3 presents summary statistics on the twenty emerging markets in the EMDB and the MSCI European, Japanese, U.S., and world indexes, calculated from January 1985 to December 1991. All calculations use monthly total (not excess) returns. Fourteen emerging markets have mean returns exceeding those of all the industrial-market indexes, whereas five have mean returns below the industrial-market range. However, only five emerging markets (Chile, Col- ombia, Pakistan, the Philippines, and Zimbabwe) have higher reward-to-risk ratios (mean/standard deviation) than the industrial markets, and all five of these opened during the sample period. Figure 1 plots mean returns against return variance for all twenty-four markets. All four industrial-market indexes are closely clustered, offering very low variances but also lower-than-average returns. Three emerging markets (India, Korea, and Pakistan) also join this tight cluster, and another four (Chile, the Philippines, Thailand, and Zimbabwe) offer higher means for only a small increase in variance. Note that Argentina, Brazil, and Turkey have very large return variances. More striking is the high return autocorrelation in many emerging markets, shown in table 3. Among industrial markets, the world index exhibits the high- est one-month autocorrelation at 0.0713, and only Japan exhibits positive auto- correlation at two months. In contrast, fifteen emerging markets exhibit one- month autocorrelation exceeding 0.1000, and the predictable component rises as high as 0.5509 for Indonesia and 0.4999 for Colombia; for Chile, Colombia, Indonesia, and the Philippines the autocorrelations are significant at 95 percent confidence. Six markets also demonstrate autocorrelation over 0.1000 at two- month intervals, although most emerging markets exhibit mean reversion after two to three months. The high predictable component in returns in emerging markets far exceeds the autocorrelation in industrial markets. High serial cor- relation suggests that returns contain a predictable component and, according to efficient-markets models that assume risk-neutral investors, that the market is inefficient (see LeRoy 1989 for a survey of the efficient-markets literature). The correlation matrices of total equity returns in table 4 show that emerging markets are generally less correlated with the world portfolio and with industrial markets than industrial markets are with each other. During 1985-91, the U.S., Table 3. Summary Statistics on Monthly Total Equity Returns, Including Reinvested Dividends, January 1985 to December 1991 Standard Reward/ Autocorrelation Market Mean deviation risk ratio Pi P2 P3 P4 Ps Europe 0.0178 0.0574 0.3103 -0.0176 -0.0646 0.0141 0.0566 -0.1214 Japan 0.0164 0.0786 0.2080 0.0192 0.0012 0.0025 0.0388 0.1383 United States 0.0138 0.0513 0.2682 0.0545 -0.0903 -0.0921 -0.1850 -0.0481 World 0.0146 0.0488 0.2991 0.0713 -0.0458 -0.0329 -0.1073 0.0274 Argentina 0.0358 0.2797 0.1279 -0.1045 -0.1087 0.0949 -0.2005 -0.1158 Brazil 0.0069 0.2239 0.0308 -0.0483 0.0768 -0.0717 -0.0921 -0.0234 Chile 0.0402 0.0779 0.5159 0.3147' -0.0404 -0.2566 -0.1809 0.1005 Colombia 0,0334 0.0779 0.4287 0.4999"' 0.1961 0.0374 0.0271 0.0378 Greece 0.0244 0.1292 0.1891 0.1282 0.1591 -0.0060 -0.1650 -0.1327 India 0.0146 0.0852 0.1712 0.0500 -0.0291 0.0538 0.0249 -0.1425 Indonesia, -0.0215 0.1006 -0.2138 0.5509** 0.2867 0.1835 0.2753 0.6103 Jordan 0.0038 0.0531 0.0707 0.0751 0.0526 0.2962" 0.0243 0.2791 Korea, Rep. of 0.0194 0.0833 0.2323 -0.0268 0.2075: -0.0186 0.2030 0.1256 Malaysia 0.0079 0.0812 0.0974 (.1109 0.0808 -0.0438 0.0423 -0.0893 Mexico 0.0361 0.1593 0.2268 0.3494 -0.1669 -0.2875 -0.0560 -0.0889 Nigeria -0.0001 0.1179 -0.0007 0.1241 -0.0488 -0.1300 0.1124 -0.0006 Pakistan 0.0206 0.0547 0.3759 0.3789" 0.0365 0.0647 0.1375 0.0264 Philippines 0.0348 0.1101 0.3162 0.3652" 0.0261 0.0520 0.1669 0.0864 Portugal 0.0303 0.1390 0.2177 0.2846 -0.0098 0.0271 0.2686 0.0646 Taiwan (China) 0.0234 0.1591 0.1468 0.0391 0.0730 -0.0411 0.0410 0.1679 Thailand 0.0261 0.0906 0.2883 0.1034 0.0360 -0.0534 -0.2401 -0.1184 Turkey 0.0579 0.2294 0.2524 0.1804 0.0298 0.2486 0.2723 -0.0977 Venezuela 0.0267 0.1393 0.1913 0.2124 0.2012"" 0.0558 -0.0167 -0.1320 Zimbabwe 0.0270 0.0833 0.3234 0.2749* 0.2544 0.2134 0.1678 -0.2273 Significant at 90 percent. -' Significant at 95 percent. a. The calculations are based on only twenty-four observations becausc the EMDB began tracking the Jakarta market only in December 1989. Source: Author's calculations based on data from the IFC EMDB. Buckberg 63 Figure 1. Mean Returns and Return Variances in Selected Emerging and Industnial Markets, 1985-91 Mean of monthly returns o.06 - & Turkey 0.05 - Chile 0.04 - Philippines XP Mexico Argentina a 0.03 -Zimbabwe a Portugal % Thailand o Venezuela Pakistan - Greece S Nigeria 0.02 -@ J Korea EuW. l *Japan 0.01 -U.s. Malaysia * Brazil 0 -JoraanC Colombia *t Nigeria -0.01 - -0.02 - Indonesia -0.03 - I l l l I 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 Return variance Source: Author's calculations based on the Emerging Markets Data Base (EMDB). Japanese, and European portfolios all exhibit correlations of 0.75 or above with the world market. However, correlations with the world portfolio that exceed 0.25 are exhibited in only seven emerging markets: Malaysia (with the highest correlation, at 0.51), Thailand, Portugal, Korea, the Philippines, Mexico, and Taiwan (China). Meanwhile, Argentina, Indonesia, India, Turkey, Venezuela, and Zimbabwe all offer consistent hedges against the world markets. From December 1976 to December 1984, emerging markets exhibited on average even less correlation with the industrial markets. Two facts-that the markets best known to foreign investors are those most correlated with the industrial markets and that correlations between industrial and emerging markets have risen over time-suggest that once foreign investors "discover" an emerging market, it quickly becomes part of the global market. Bekaert (1993) further discusses the link between market correlation and market integration. IV. TESTS OF A CONDITIONAL ICAPM WITH TIME-VARYING EXPECTED RETURNS, I985-91 Tests of the conditional ICAPM on two different time periods reveal that many emerging markets were integrated into the global capital market during 64 THE WORLD BANK ECONOMIC REVIEW, VOL- 9, NO. 1 Table 4. Total Return Correlations including Reinvested Gross Dividends, in Selected Markets, 1985-91 and 1976-84 A. 1985-91 Colom- Indo- Korea, Mex- Malav- Nige Market Brazil Chile bia Greece nesia India Jordan Rep. of ico sia roa Argentina -0.06 -0.02 -0.07 0.04 -0.23 0.22 -0.13 -0.14 0.11 -0.04 0.1 Brazil 0.13 0.03 -0.01 -0.12 -0.03 -0.04 0.01 -0.01 0.12 O.C Chile 0.10 0.17 0.03 -0.06 -0.06 0.12 0.33 0.27 O.C Colombia 0.22 0.16 -0.08 0.05 -0.09 0.07 0.03 O.C Greece 0.34 0.04 0.07 -0.14 0.15 0.08 O.C Indonesia -0.03 0.21 -0.11 0.23 0.43 -0.2 India 0.14 -0.03 -0.00 -0.05 O.X Jordan -0.17 -0.04 0.06 -O.C Korea, Rep. of 0.15 0.10 OC Mexico 0.44 -0.C Malaysia -0.2 Nigeria Taiwan (China) Pakistan Philippines Portugal Thailand Turkey Venezuela Zimbabwe United States Japan Europe B. 1976-84 Market Brazil Chile Greece India Jordan Korea, Rep. of Argentina -0.07 0.08 0.05 0.09 -0.03 0.06 Brazil -0.10 -0.17 -0.22 -0.01 -0.02 Chile 0.23 0.11 0.17 0.01 Greece 0.25 0.32 -0.04 India 0.35 -0.02 Jordan -0.16 Korea, Rep. of Mexico Thailand Zimbabwe United States Japan Europe Source: Author's calculations. Buckberg 65 Taiwan Philip- Portu- Thai- Vene- Zim- United Eu- (China) Pakistan pines gal land Turkey zuela babwe States Japan rope World -0.02 -0.03 -0.11 -0.03 0.10 0.10 0.08 -0.26 0.04 -0.13 -0.08 -0.09 0.10 0.01 0.13 0.11 0.07 0.13 -0.18 -0.02 0.12 0.16 0.14 0.17 0.35 -0.05 0.23 0.24 0.27 0.01 -0.20 -0.02 0.33 0.07 0.19 0.22 0.07 0.51 0.08 0.17 0.11 0.06 0.02 -0.19 0.14 0.00 0.08 0.09 0.05 -0.10 0.12 0.42 0.32 0.20 -0.08 0.01 0.15 0.06 0.18 0.16 0.18 0.20 0.48 0.09 0.48 0.17 0.04 0.28 0.20 -0.20 0.18 -0.01 -0.12 0.08 -0.09 -0.06 0.01 0.23 0.05 0.07 -0.05 -0.12 0.08 -0.06 0.12 0.08 0.11 -0.02 0.15 -0.16 0.01 0.06 0.02 0.10 0.23 0.13 0.05 -0.02 0.20 0.10 -0.02 -0.09 -0.23 -0.12 0.26 0.35 0.23 0.36 0.34 0.03 0.10 0.43 0.39 0.12 -0.05 -0.09 0.45 0.09 0.23 0.30 0.30 -0.03 0.34 0.28 0.57 0.20 0.01 0.01 0.57 0.24 0.43 0.51 -0.18 0.01 -0.00 -0.22 -0.12 0.08 0.03 -0.06 0.02 0.08 0.06 0.05 --0.02 0.12 0.35 0.43 0.12 -0.25 -0.08 0.20 0.21 0.17 0.26 -0.01 0.00 0.12 -0.00 0.05 -0.14 -0.01 -0.05 0.08 0.00 0.07 0.27 -0.01 -0.19 0.01 0.30 0.29 0.29 0.36 0.39 0.22 -0.06 0.12 0.23 0.40 0.34 0.42 0.19 -0.09 -0.04 0.42 0.26 0.42 0.46 -0.16 0.05 -0.05 -0.00 0.05 -0.00 0.07 -0.09 -0.16 -0.13 -0.16 -0.09 0.01 -0.06 -0.05 0.25 0.64 0.75 0.52 0.79 0.84 Zim- United Mexico Thailand babwe States Japan Europe World 0.16 -0.16 0.17 0.07 0.08 0.09 0.11 -0.13 -0.17 -0.10 -0.06 -0.13 -0.02 -0.07 0.04 -0.03 0.23 -0.15 0.08 0.04 -0.06 -0.20 0.13 0.31 0.03 0.19 0.25 0.15 -0.00 0.14 0.27 -0.01 0.28 0.28 0.17 -0.04 -0.04 -0.01 0.15 0.10 0.24 0.20 0.05 0.07 -0.09 0.04 0.15 0.12 0.09 0.06 0.15 0.12 0.03 0.13 O.1S -0.06 -0.13 0.05 0.13 -0.04 0.05 0.06 0.15 0.11 0.22 0.46 0.86 0.53 0.59 0.79 66 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I 1985-91 but that many of the same markets reject the model when data for 1977-84 are used. Indeed, the late 1980s marked a change in the relation between emerging markets and world financial markets as the emerging markets began receiving large capital inflows from industrial economies. The greater success of integration tests in the more recent sample indicates that many emerg- ing markets were segmented from international capital markets until the late 1980s and suggests that rising cross-border capital flows served as the means of integration. Table 5 presents tests of a conditional ICAPM with heteroskedasticity-corrected errors on individual markets from January 1985 to December 1991. As modeled in section II, the original Sharpe-Lintner formulation assumes that asset returns are proportional to the return on the world market portfolio, with coefficient of proportionality 3. The third and fourth columns of table 5 provide chi-squared statistics for each test; high chi-squared statistics indicate poor fit. The p-value (in parentheses in the third and fourth columns) indicates the probability that if the model's errors were indeed distributed chi-squared with the given degrees of freedom, a value exceeding the calculated chi-squared value would be obtained. The discussion will focus on whether or not the model is rejected with 90 percent or more confidence (denoted by a p-value smaller than 0.10). For eighteen of twenty markets, the ICAPM cannot be rejected using either instrument set; only Mexico and Zimbabwe reject the model, both with over 95 percent confidence and using both the local- and common-instrument sets.3 That the ICAPM can be rejected in only two markets gives strong evidence that emerging markets were integrated in the late 1980s. That the model rejects for Zimbabwe is unsurprising because Zimbabwe is the smallest market in the sample (with a capitalization of $1.3 billion in December 1991) and likely the least liquid; this last feature would affect pricing patterns. Although Mexico's market was large by 1991, with a capitalization of $100 billion, it stood at a mere $2.2 billion in 1985 and the model's rejection may relate to changes in return patterns during its dramatic growth. The estimated betas are statistically distinct from zero in ten of twenty mar- kets, of which six are more than two standard deviations from zero and four are at least one standard deviation from zero; in the remaining ten markets the standard error exceeds the estimated coefficient. In contrast, Harvey's study of industrial markets yields betas that are two or more standard errors from zero for every market except Austria, for which he rejects the model. The failure to obtain insignificant betas for emerging markets is unsurprising for two reasons. One reason is that because of high idiosyncratic risk, the markets are very noisy and volatile, leading to large standard errors and making 3. Without the heteroskedasticity correction, the ICAPM can also be rejected with at least 90 percent confidence for Colombia, Pakistan, Portugal (with local instruments only), and Turkey (with common instruments only). In each of these cases, the chi-squared statistics are much higher in the tests with uncorrected errors than in the heteroskedasticity-corrected tests. The betas differ between the corrected and uncorrected tests, but not in any consistent manner. Buckberg 67 Table 5. Heteroskedasticity-consistent Estimates of a Conditional ICAPM with Time-varying Expected Returns and Constant Conditional Proportionality Factors (ej = rjt - r,3), 1985-91 Mean Local Common Coefficient of pro- excess instruments, instruments, Market portionality, f3 return xi xi Argentina 1.204 0.030 5.48 2.71 (1.413) (0.484) (0.745) Brazil 1.173 0.002 2.66 2.11 (1.346) (0.851) (0.834) Chile 2.127 0.035 8.99 7.11 (0.685) (0.174) (0.213) Colombia 0.919 0.028 6.98 4.92 (0.420) (0.323) (0.425) Greece 0.081 0.019 7.79 6.93 (0.731) (0.254) (0.226) India 0.052 0.009 1.90 1.90 (0.574) (0.928) (0.863) Indonesia 0.276 -0.027 7.81 3.60 (0.583) (0.252) (0.609) Jordan 0.280 -0.002 1.57 1.57 (0.367) (0.955) (0.905) Korea, Rep. of 1.293 0.014 5.93 1.06 (0.516) (0.431) (0.958) Malaysia 0.405 0.003 3.08 1.97 (0.463) (0.799) (0.853) Mexico 1.397 0.031 13.63* 13.82* (0.797) (0.034) (0.017) Nigeria -1.368 -0.005 3.37 3.40 (0.843) (0.760) (0.639) Pakistan 0.395 0.015 6.06 5.98 (0.233) (0.416) (0.308) Philippines 2.812 0.029 2.15 2.14 (0.814) (0.906) (0.830) Portugal 2.271 0.025 7.58 4.89 (0.638) (0.270) (0.429) Taiwan (China) 1.720 0.018 2.93 2.53 (0.642) (0.818) (0.771) Thailand 0.624 0.021 7.14 7.05 (0.468) (0.308) (0.217) Turkey -1.859 0.053 8.07 6.19 (2.574) (0.232) (0.288) Venezuela 1.156 0.021 6.98 4.14 (0.703) (0.323) (0.528) Zimbabwe -0.375 0.022 15.52* 15.46* (0.450) (0.017) (0.009) * Null hypothesis of the ICAPM can be rejected with 95 percent confidence. Note: All estimates use Hansen's (1982) generalized method of moments (GMM) for the model e, = rj - r_,, ,. The GMM procedure iterates three to five times over the weighting matrix because Monte Carlo simulations have found that repeated iteration improves the small-sample properties of the esti- mates. Common instruments include a constant, the lagged return on the world index less the return on a thirty-day Treasury bill, a dummy for the month of January, the lagged differential between the return to holding a ninety-day Treasury bill for one month and the return on a Treasury bill thirty days to maturity, the lagged differential between the yield on a Moody's Baa bond and the yield on a Moody's Aaa bond, and the lagged dividend yield on the world portfolio less the return on a thirty-day Treasury bill. The local instrument set includes all common instruments plus the lagged return on the local index less the return on a thirty-day Treasury bill. Values in parentheses in column 1 are standard errors; in columns 3 and 4 they are p-values. Source: Author's calculations. 68 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I the beta unstable. Allowing conditional moments to vary across time might alleviate this problem; unfortunately, however, adequate data are not yet avail- able to estimate the additional parameters. Even with a strong covariance, a high variance could lead to an insignificant coefficient. As shown in section III, the correlation between returns on the world portfolio and returns in a number of emerging markets is still negligible. A second reason is that although correlations have risen on average since the late 1970s, covariance risk may remain too small or too idiosyncratic for local returns to reflect covariance with the world market. In that case, obtaining significant betas is unlikely. Indeed, the strongest ,Bs are obtained in the markets most correlated with the world portfolio (as shown in table 4): Chile, Korea, Mexico, the Philippines, Portugal, Taiwan (China), and Thailand. Only Malay- sia has a similarly high correlation with the world market yet fails to yield a strong beta. Bekaert (1993) also finds larger betas (in relation to the U.S. mar- ket) for larger emerging markets and associates high betas with a higher degree of market integration. These markets are also among the most developed in the sample and the most popular with foreign investors. The lack of significant betas in no way invalidates the ICAPM tests. The ulti- mate test of the model's fit lies in the chi-squared statistic, not the estimated beta. Strong rejections of the model for Mexico and Zimbabwe (with 97 percent to over 99 percent confidence) confirm that the test indeed has the power neces- sary to discriminate between markets that do and do not reject the model. Small betas may be consistent with either the null hypothesis of ICAPM integration or the alternative hypothesis of nonintegration. The model should not be rejected as long as the conditionally expected rate of return E(r) falls within the range predicted by the model-that is, between the minimum and maximum vertical intercepts of the security-market line between (1 - a) and (1 + a) as shown in figure 2. If E(r) is not predicted by any beta in the standard error range of the estimate, then the model should be rejected. The model's success at describing emerging markets in the late 1980s slightly exceeds that of similar studies of industrial markets. This can likely be explained by the fact that ICAPM studies on industrial markets (such as Harvey 1991) typically use time series that include data from the 1970s, when international capital flows were more restricted, even among industrial markets, than in the late 1980s. Moreover, the results in this article may make emerging markets look more efficient and integrated than they in fact are because the IFC indexes include only each market's largest and most heavily traded stocks and are calcu- lated with capitalization-based weights. Emerging markets are very concentrated by international standards; for ex- ample, the Buenos Aires stock market lists 175 companies, but the 10 largest companies accounted for more than 68 percent of both market capitalization and value traded at the end of 1992. A valuable extension would be to examine the correlation of capitalization-based portfolios with the world market and perform ICAPM tests to determine whether increased integration applies only to Buckberg 69 Figure 2. 7he Secuities-Market Line and Model Rejection and Nonrejection Ranges Conditionally expected rate If EQ) is here, no i Market of retum, EQ) in standard enror range can explain it; line therefore, reject ICAPM/ Effilcient portffolio variance / / ' If E() is here, 10 Olifol ' do not reject ICAPM rf, ---___________ 3-oa 0 Pi \ 3+o Standard error a(i) Standard error range the largest stocks or is reflected throughout markets. Claessens, Dasgupta, and Glen (1993) examine the behavior of capitalization-based portfolios in eighteen emerging markets during 1988-92. A stricter test of the ICAPM is achieved by imposing the cross-asset restriction that a common price of covariance risk holds across markets, in addition to requiring that the time-series behavior in each market accords with the model. Estimating all seventeen markets with complete time series (Indonesian, Por- tuguese, and Turkish data start after December 1984) as a system provides a stricter test of the model on the most complete sample of markets possible. This test is expected to reject the model, because the ICAPM cannot explain the time- series behavior of returns for two of the markets in the system. The instrument set is reduced to include only the lagged returns on the world- and local-market portfolios so that the number of unique elements of the weighting matrix will be smaller than the number of data points (including market data and instruments). The results are presented for completeness but should be interpreted cautiously because their numerical properties are very weak. As table 6 shows, the ICAPM iS not rejected. The nonrejection of the model despite the presence of two markets whose time-series behavior the model cannot explain suggests that the test may have inadequate power. Harvey (1991) also found multimarket tests unable to reject the ICAPM even when component markets rejected the model in time-series tests. 70 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Table 6. Multimarket Estimates of a Conditional ICAPM Time period and sample Estimate 17-market sample,a 1985-91 24.43 (0.108) 9-market sample,b 1977-84 3.07 (0.962) Note: The estimates are from a conditional ICAPM with time-varying expected returns and constant conditional proportionality factors, 4, and are heteroskedasticity-consistent. The lagged return on the world market portfolio, r_,-,, and the lagged return on the local portfolio, r1,,1, are used as instruments in the model ej, = rj, - r_,3. The seventeen-market test has seventeen degrees of freedom-the number of orthogonality conditions (17 times 2), less the number of parameters (17); the nine-market test has nine degrees of freedom. p-values are in parentheses. a. Argentina, Brazil, Chile, Colombia, Greece, India, Jordan, Korea, Malaysia, Mexico, Nigeria, Pakistan, the Philippines, Taiwan (China), Thailand, Venezuela, and Zimbabwe. b. Argentina, Brazil, Chile, Greece, India, Korea, Mexico, Thailand, and Zimbabwe. Source: Author's calculations. V. TESTS OF THE CONDITIONAL ICAPM FOR I977-84 To investigate whether emerging markets have become more integrated over time, this section presents ICAPM tests on a ten-market sample from January 1977 to December 1984. The earlier data cover a period when emerging mar- kets were effectively isolated from world capital markets and were much smaller in relation to industrial markets than they are today. Emerging markets proba- bly became more integrated beginning in the late 1980s, when investors from industrial economies became active participants in emerging stock markets and many markets lifted restrictions on foreign investment. The critical year may have been 1987 or 1988, when booming Southeast Asian markets first attracted capital from industrial economies to emerging markets. Table 7 shows ICAPM estimates with heteroskedasticity-corrected errors for ten markets. For 1977-84, five markets reject the ICAPM under both the local- and common-instrument sets. These are Greece, India, Korea, Mexico, and Zim- babwe. Chile rejects the model in the common-instrument test and comes close to the 90 percent rejection level in the local-instrument test. All rejections with the local-instrument set are in the 95 percent confidence range except Korea, which rejects with 90 percent confidence; all rejections with the common- instrument tests are in the 90 percent confidence range except Mexico, which rejects with 95 percent confidence.4 The fact that the ICAPM successfully de- scribes 1985-91 return behavior (that is, that the model is not rejected) in four of these markets-Chile, Greece, India, and Korea-indicates that they became integrated sometime in the mid-1980s. Six of the estimated betas are one or more standard deviations from zero; the betas for Brazil, Greece, Mexico, and Thailand are all two or more standard deviations from zero. The betas do not, 4. In tests without the errors correction for heteroskedasticity, the only major difference is that the ICAPM is not rejected for Korea. Buckberg 71 Table 7. Heteroskedasticity-consistent Estimates of a Conditional ICAPM with Time-varying Expected Returns and Constant Conditional Betas (Ejt= rj, -r,3), 1977-84 Local Local Common proportionality instruments, instruments, Market factor, 0I'-e Xi Xi Argentina -0.654 7.96 7.93 (1.222) (0.241) (0.160) Brazil -1.645 2.56 2.24 (0.683) (0.861) (0.816) Chile -1.607 10.62 10.70* (0.969) (0.101) (0.058) Greece 1.167 13.43* * 10.83* (0.556) (0.037) (0.056) India 0.113 14.91** 10.76* (0.374) (0.021) (0.056) Jordan -0.256 3.023 3.234 (0.341) (0.806) (0.664) Korea, Rep. of 0.839 11.00* 10.72* (0.647) (0.088) (0.057) Mexico 2.057 15.11** 15.07** (0.889) (0.019) (0.010) Thailand 1.229 2.08 1.06 (0.569) (0.912) (0.957) Zimbabwe 0.218 13.63** 10.83* (0.821) (0.034) (0.055) * Null hypothesis of the ICAPM is rejected with 90 percent confidence. * Rejection with 95 percent confidence. Note: All estimates use Hansen's (1982) generalized method of moments (GMM) on the model fi, r, - r_, 3. The GMM procedure iterates three to five times over the weighting matrix, as Monte Carlo simulations have found that repeated iteration improves the small-sample properties of the estimates. The common-instrument set includes a constant, the lagged return on the world index less the return on a thirty-day Treasury bill, a dummy for the month of January, the lagged differential between the return to holding a ninety-day Treasury bill for one month and the return on a Treasury bill thirty days to maturity, the lagged differential between the yield on a Moody's Baa bond and the yield on a Moody's Aaa bond, and the lagged dividend yield on the world portfolio less the return on a thirty-day Treasury bill. The local-instrument set includes all common instruments plus the lagged return on the local index less the return on a thirty-day Treasury bill. Standard errors are in parentheses. Source: Author's calculations. however, rank closely with the markets' correlation with the world portfolio (as shown in table 4). For completeness, cross-country tests on nine markets (Jordanian data starn later) appear in table 6. Although six markets reject the model in time-series tests, the system estimates fail to reject the model. These results should be viewed as evidence of power problems, not evidence of integration. As in the tests of the seventeen-market system, only the lagged world- and local-market returns are used as instruments, and the numerical properties of the estimates are weak. VI. CONCLUSIONS This article finds that emerging markets became increasingly integrated into global financial markets in the late 1980s, with rising capital flows from indus- 72 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I trial economies evidently the mechanism of integration. For 1977-84, six of ten markets reject the ICAPM in tests of the conditional formulation of the Sharpe- Lintner ICAPM. However, evidence of integration is strong for 1985-91: eighteen of twenty markets do not reject the model in heteroskedasticity-consistent esti- mates, although high rate-of-return variance impedes estimation of market betas. Stylized evidence from recent market openings suggests that opening expands markets in terms of participants, firms listed, and value, and promotes the capital flows necessary for integration. TECHNICAL APPENDIX: THE GENERALIZED METHOD OF MOMENTS The article uses the generalized method of moments (GMM) to obtain a con- sistent and efficient estimate of beta. The weighting matrix WT is chosen opti- mally as in Hansen (1982).5 The orthogonality condition of equation 4 implies that the moment restrictions (A-1) E(,E) = E[f(rj,, 1)] = 0 hold, where rj, denotes excess returns in U.S. dollars, ,B is the proportionality factor for expected returns in the local market, and f(rj, ,B) is referred to as the wing. GMM estimation mimics this moment restriction by minimizing a quadratic form of the sample means, (A-2) JT(1) = | b(rt, )' jWTL+ if(rjt, 3)J with respect to 13. WT, the variance-covariance matrix of wing f (rj ), is a positive definite matrix satisfying (A-3) limtt WT= WO with probability approaching one for positive definite matrix WT; both WT and W, are referred to as the weighting matrix. Hansen (1982) shows that the optimal weighting matrix, that is, the one that minimizes the variance- covariance matrix of 1, is the inverse variance-covariance matrix of the wing. Generally, two iterations over WT should yield an asymptotically consistent estimate for 1, flT. However, Monte Carlo simulations have shown that repeated iteration over the weighting matrix improves the small-sample properties of the estimates (Tauchen 1986). All single-market estimates in this article are obtained after three to five iterations over WT, with the procedure ceasing before five iterations if WT has converged; multimarket estimates use as many as sixteen iterations to obtain convergence. The minimized value of the quadratic form in equation A-2 is distributed x2 under the null hypothesis, with degrees of freedom equal to the number of 5. To perform my GMM estimation, I use the Gauss procedures described in "GMM Programs for Gauss" by Lars Peter Hansen, John Heaton, and Masao Ogaki (1992). This work was sponsored by the National Science Foundation and is available from Lars Peter Hansen, Department of Economics, University of Chicago, Chicago, Ill. Buckberg 73 orthogonality conditions less the number of parameters, or equal to the number of instruments less the number of parameters. The x2 statistic measures how close the errors are to zero (Ho) after repeated iteration, and can be interpreted as an indicator of the model's goodness of fit as it measures whether the quadra- tic maximand evaluated at the optimal parameter estimates is statistically differ- ent from zero. A high value of the x2 statistic signals that the disturbances are correlated with the instrumental variables and that the model may be misspecified. REFERENCES Bekaert, Geert. 1993. "Market Integration and Investment Barriers in Emerging Equity Markets." In Stijn Claessens and Sudarshan Gooptu, eds., Portfolio Investment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. Campbell, John W., and Yasushi Hamao. 1989. Predictable Stock Returns in the United States and Japan: A Study of Long-Term Capital Market Integration. NBER Working Paper 3191. Cambridge, Mass.: National Bureau of Economic Research. Capital International Perspective, S.A., and Morgan Stanley & Co. Various issues. Morgan Stanley Capital International Perspective. New York, N.Y: Morgan Stanley. Claessens, Stijn, Susmita Dasgupta, and Jack Glen. 1993. "Stock Price Behavior in Emerging Markets." In Stijn Claessens and Sudarshan Gooptu, eds., Portfolio Invest- ment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. Cutler, David M., James M. Poterba, and Lawrence H. Summers. 1989. "What Moves Stock Prices?" Journal of Portfolio Management 15:4-12. 1991. "Speculative dynamics." Review of Economic Studies 58(3, May):529- 46. Dumas, Bernard, and Bruno Solnik. 1992. "The World Price of Exchange Rate Risk." The University of Pennsylvania, The Wharton School, Department of Finance, Phila- delphia, Pa. Processed. De Santis, Giorgio. 1993. "Asset Pricing and Portfolio Diversification: Evidence from Emerging Financial Markets." In Stijn Claessens and Sudarshan Gooptu, eds., Portfo- lio Investment in Developing Countries. World Bank Discussion Paper 228. Washing- ton, D.C. Errunza, Vihang, Etienne Losq, and Prasad Padmanabhan. 1992. "Tests of Integration, Mild Segmentation and Segmentation Hypotheses." Journal of Banking and Finance 16(5, September):949-72. Fama, Eugene E 1970. "Efficient Capital Markets: A Review of Theory and Empirical Work." Journal of Finance 25(2, May):383-417. Fama, Eugene F., and Kenneth R. French. 1988. "Permanent and Temporary Compo- nents of Stock Prices." Journal of Political Economy 96(2, April):246-73. . 1989. "Business Conditions and Expected Returns on Stocks and Bonds." Jour. nal of Financial Economics 25(1, November):23-50. French, Kenneth R., and James M. Poterba. 1989. "Are Japanese Stock Prices toc, High?" Working Paper. University of Chicago, Graduate School of Business, Chicago. Ill. Gultekin, Mustafa N., and N. Bulent Gultekin. 1983. "Stock Market Seasonality: Inter national Evidence." Journal of Financial Economics 12(4, December):469-81. 74 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Hansen, Lars Peter. 1982. "Large Sample Properties of Generalized Method of Moments Estimators." Econometrica 50:1029-54. Harvey, Campbell R. 1989. "Time-varying Conditional Covariances in Tests of Asset Pricing Models." Journal of Financial Economics 24:289- 317. . 1991. "The World Price of Covariance Risk." Journal of Finance 46(1, March): 111-57. IFC (International Finance Corporation). Various years. Emerging Stock Markets Fact- book. Washington, D.C. IMF (International Monetary Fund). Various years. Annual Report on Exchange Ar- rangements and Exchange Restrictions. Washington, D.C. Keim, Donald B. 1983. "Size-related Anomalies and Stock Return Seasonality: Further Empirical Evidence." Journal of Financial Economics 12(1, June):13-32. Keim, Donald B., and Robert F. Stambaugh. 1986. "Predicting Returns in the Stock and Bond Markets." Journal of Financial Economics 17(2, December):357-90. LeRoy, Stephen F. 1989. "Efficient Capital Markets and Martingales." Journal of Eco- nomic Literature 27(4, December):1583-1621. Lintner, John. 1965. "The Valuation of Risk Assets and the Selection of Risky Invest- ments in Stock Portfolios and Capital Budgets." Review of Economics and Statistics 47:13-37. McDonald, Jack. 1989. "The Mochiai Effect: Japanese Corporate Cross-Holdings." Journal of Portfolio Management 16(1, Fall):90-94. Ogaki, Masao. 1992. "Generalized Method of Moments: Econometric Applications." University of Rochester, Department of Economics, Rochester, N.Y. Processed. Sharpe, William F 1964. "Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk." Journal of Finance 19:425-42. Tauchen, George. 1986. "Statistical Properties of Generalized Method-of-Moments Esti- mators of Structural Parameters Obtained from Financial Market Data." Journal of Business and Economic Statistics 4(4, October):397-416. THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1: 75-107 Market Integration and Investment Barriers in Emerging Equity Markets Geert Bekaert This article develops a return-based measure of market integration for nineteen emerg- ing equity markets. It then examines the relation between that measure, other return characteristics, and broadly defined investment barriers. Although the analysis is ex- ploratory, some clear conclusions emerge. First, global factors account for a small fraction of the time variation in expected returns in most markets, and global predic- tability has declined over time. Second, the emerging markets exhibit differing degrees of market integration with the U.S. market, and the differences are not necessarily associated with direct barriers to investment. Third, the most important de facto bar- riers to global equity-market integration are poor credit ratings, high and variable inflation, exchange rate controls, the lack of a high-quality regulatory and accounting framework, the lack of sufficient country funds or cross-listed securities, and the lim- ited size of some stock markets. Equity portfolio flows to developing economies, especially to the so-called emerging markets, have sharply increased in magnitude in recent years. The increase in financial flows to emerging markets raises three important questions: • What are the expected return and diversification benefits of investing in these markets? • How well are these markets integrated with the markets of industrial econ- omies and to what extent is integration a function of identifiable barriers to investment? * What are the opportunity costs, in terms of higher cost of capital, associated with these barriers? These questions are closely related. The return properties and potential diver- sification benefits from investing in emerging markets have been investigated by Geert Bekaert is with the Graduate School of Business at Stanford University. This article was commis- sioned by the Debt and International Division of the World Bank for its Conference on Portfolio Invest- ment in Developing Countries, Washington, D.C., September 9-10, 1993. The author would like to thank Michael Urias for excellent research assistance and many useful comments; Stijn Claessens, Steve Grenadier, Bob Hodrick, Ingrid Werner, the discussant Cheol Eun, and three anonymous referees for suggestions and comments; Steve Gray and Rohit Kumar for their assistance with some of the computa- tions; and Bob Korajzyek for providing part of the data. (D 1995 The International Bank for Reconstruction and Development/THE WORLD BANK 75 76 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I a number of authors, including Divecha, Drach, and Stefek (1992); Harvey (1993); Speidell and Sappenfield (1992); and Wilcox (1992). However, barriers to investment can make potential diversification benefits unattainable for for- eign investors. As a consequence, capital flows from the industrial world, which might reduce domestic capital costs and increase economic welfare through more efficient resource mobilization, might not be forthcoming. This article will try to shed some light on the last two questions, with primary emphasis on market segmentation. The analysis is restricted to nineteen equity markets con- tained in the Emerging Markets Data Base (EMDB) of the International Finance Corporation (IFC). There are two major approaches to testing and measuring the degree of mar- ket segmentation. The first approach assumes that markets are integrated and that a particular asset-pricing model holds (for example, Campbell and Hamao 1992). The second approach models the restrictions to integration explicitly and derives their effects on equilibrium returns (for example, Cooper and Kaplanis 1986, 1994; Errunza and Losq 1985; Eun and Janakiramanan 1986; Hietala 1989; Stulz 1981; and Wheatley 1988). The second approach is unsatisfactory because I do not want to restrict the analysis to the effects of one particular barrier to investment and there are too many different barriers to consider. The first approach is hampered by the lack of a universally accepted international asset-pricing model. Recent research on international equity and foreign exchange markets, for instance, has uncovered considerable time variation in expected excess returns, but no consensus has emerged on what drives this apparent predictability. Some empirical papers show that common risk factors explain a large fraction of the time and cross- sectional variation in returns (for example, Harvey 1991). This suggests that markets in industrial economies, at least from 1980 onward, are relatively well integrated. In any case, the use of a formal asset-pricing model requires further research on capital market integration in general and is left for future work. My approach consists of two steps. First, in section I, I examine whether pre- dictable components in the excess returns from investing in emerging markets are similar to those observed in industrial equity markets. If the predictable compo- nents track time-varying risk premiums, examining these components can inform on market integration as well. I include both local factors (the lagged return and the dividend yield) and global factors (the lagged return on the U.S. market, the U.S. dividend yield, and the U.S. interest rate) in regression analysis to investigate the relative importance of global, compared with local, components in the predic- tability of excess returns in emerging markets. I interpret the predictive power of global factors as indicative of some degree of integration. Similarly, I interpret the lack of predictive power by the local instruments as indicative of integration, al- though some international asset-pricing models imply that economy-specific fac- tors are priced (Adler and Dumas 1983). Second, I use the regressions to compute correlations of expected returns in emerging markets with expected returns in equity markets in industrial econ- Bekaert 77 omies. If there were only one source of risk and markets were perfectly inte- grated, expected returns would be perfectly correlated (see Cumby and Huizinga 1992). Bekaert (forthcoming), for instance, uses a vector autoregressive frame- work to compute correlations between expected returns on foreign exchange and finds that they are highly correlated. Although it seems unlikely that one risk factor explains all of the cross-sectional and time variation in equity returns, it is equally unlikely that expected returns in perfectly integrated markets would show low correlation. In fact, as shown in section II, the expected equity returns in the major industrial markets are highly correlated. This correlation is a mea- sure of the common component in expected stock returns and hence, indirectly, of market integration (see also Campbell and Hamao 1992). However imper- fect, the correlation of expected returns is the measure of market integration used in this article. To check for robustness, I have provided an alternative measure of market integration, based on the change in predictable variation in returns when an observable proxy for the world factor (the world market port.. folio return) is added to the forecasting equations. In section III, I discuss various other return characteristics and examine how they relate to the measure of market integration. The remainder of the article links the degree of market integration, as measured by the expected return correlation with the U.S. market, to various barriers to investment. I distinguish between three kinds of barriers. First are legal barriers arising from the different legal status of foreign and domestic investors, for example, ownership restrictions and taxes. Second are indirect barriers arising from dif- ferences in available information, accounting standards, and investor protec- tion. Third are barriers arising from emerging-market-specific risks (EMSRS) that discourage foreign investment and lead to de facto segmentation. EMSRS include liquidity risk, political risk, economic policy risk, macro- economic instability, and, perhaps, currency risk. Some might argue that these risks are in fact diversifiable and are not priced. However, such an argument seems inconsistent with the amount of resources spent on, for example, measur- ing political risk throughout the world. Chuhan (1992), for instance, on the basis of a survey of market participants in Canada, Germany, Japan, the United Kingdom, and the United States, reports liquidity problems as a major impedi- ment to investing in emerging markets. But the survey yielded the surprisin,g result that restrictions in host economies are not a crucial factor. The other EMSRS are related to the notion of country risk. For example, credit ratings not only reflect assessments of political stability but also incorporate factors related to the economic environment. Unstable macroeconomic policies, for instance, appear to have detrimental effects on stock market performance. Barriers to investment are a direct function of the domestic policies pursued in the various economies. This article is intended as a preliminary empirical investi- gation into the association between a set of broadly defined barriers to invest- ment and measures of market integration and other return characteristics. Be- cause quantitative measures of these barriers to investment are necessarily crude, 78 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I the association is simply measured through rank correlations. This approach has the obvious disadvantage of precluding strong quantitative policy implications, but it allows a broader analysis that can provide useful insights for further research. In section IV, I investigate the association between market integration and direct and indirect barriers to investment. I also examine whether any of the described return characteristics are related to measures of "openness" of the emerging markets, for example, the existence of country funds and cross-listed securities or the extent of ownership restrictions. Section V focuses on EMSRS. Because I do not specify a formal asset-pricing model, I cannot make an explicit link between market integration and the cost of equity capital. The analysis here takes as a starting point the belief that a higher degree of market integration is necessarily accompanied by lower costs of capital and increased capital flows. Some of the return characteristics reported in section III are corre- lated with the cost of capital, but without a generally accepted asset-pricing model, estimating the cost of capital precisely is extremely difficult and is not formally attempted. A related disadvantage of the approach here is that the rankings are typically taken at a point in time or are based on averages. No dynamic relation between changes in barriers to investment and return proper- ties is described. Some further implications for future research are discussed in section VI, which offers conclusions. Finally, cost-of-capital issues cannot be fully analyzed without incorporating the configuration of the entire financial market in the developing economy, including bond, money, and informal markets, all of which are ignored in this analysis. Eventually, it would be fruitful to take the viewpoint of the developing economy, rather than a global asset-pricing perspective, as the basis of the analysis. A model of a developing economy with rudimentary financial markets could explicitly address how opening up the equity market to foreign investors would affect returns, the cost of capital, and ultimately social welfare. I. THE PREDICTABILITY OF RETURNS IN EMERGING EQUITY MARKETS To assess the predictability of excess returns earned on investments in emerg- ing markets, I regressed the dollar index return in excess of the U.S. interest rate onto five instrumental variables (see also Bekaert and Harvey 1994, Buckberg 1995, and Harvey 1993). I used two local instruments, the local dividend yield and the lagged excess return, and three global instruments, the U.S lagged excess return, the U.S. dividend yield, and the U.S. interest rate relative to a one-year backward-moving average. These instruments were shown to predict excess returns on equities and foreign exchange in Germany, Japan, the United King- dom, and the United States in Bekaert and Hodrick (1992). Because no reliable interest rate data are available for most emerging markets, I could not emulate Bekaert and Hodrick's specification, which uses the local excess return as the dependent variable and the forward premium as a predictor. Bekaert 79 Table 1 reports the regression results for 1985-92, using data sampled at the end of each month for nineteen emerging markets.' The emerging-market in- dexes used are those compiled by the IFC as their so-called global indexes (IFC various issues). (Indonesia was excluded from the sample because of insufficienit data.) The 1980s were a decade of increasing globalization and deregulation of financial markets. These developments, and the fact that the large financial flows to emerging equity markets only occurred near the end of the sample, motivated the choice of the sample period. Moreover, for some markets, data are only available since 1986. Several test statistics are reported. The x2(5) statistic is a Wald test of the joint predictive power of the five instruments, and the X2(2) and x2(3) statistics test the predictive power of the local and global instruments, respectively. The 1-statistic is a test developed by Cumby and Huizinga (1993) for the remaining serial correlation in the residuals. It is robust to conditional heteroskedasticity and to the fact that the residuals are estimated. The adjusted R2 is greater than 10 percent in Chile, Colombia, Mexico, the Philippines, Portugal, Turkey, Venezuela, and Zimbabwe but is negative in Ar- gentina, India, Nigeria, and Thailand. The joint predictability test for all five instruments rejects the null of no predictability at the 1 percent level for six economies: Chile, Colombia, the Philippines, Portugal, Venezuela, and Zim- babwe. Except for Portugal, this rejection appears to derive from the local instruments. For Malaysia the test for no predictability of the local instruments also rejects at the 1 percent level; for Brazil, the Republic of Korea, and Turkey, it rejects at the 5 percent level. Although this result could be construed as evidence of market inefficiency, it is important to point out that the predictive power of the dividend yield, not the lagged return, drives some of the rejections (see, for example, Brazil, Portugal, and Zimbabwe). The dividend yield predicts excess returns in the industrial equity markets as well (see, for example, Bekaert and Hodrick 1992). Campbell and Ammer (1993) use a log-linear decomposi- tion of stock returns to show that the dividend yield should perform well as a proxy for the long-horizon expected excess return. The predictive power of the global instruments is generally weak. The Wald test only rejects at the 1 percent level for Portugal, at the 5 percent level for Turkey, and at the 10 percent level for Chile. For Malaysia, the predictability is primarily caused by the local instruments. This does not necessarily mean that the Malaysian market is segmented, because the local instruments might par- tially track the common component in expected returns. This possibility will be examined in section II. Note that the return for the emerging-markets composil:e index is significantly predictable at the 1 percent level using all five instrumenits and at the 10 percent level using the global instruments. The same type of analysis was done for four industrial economies: Germany, Japan, the United Kingdom, and the United States (not reported).2 Surprisingly, 1. See the appendix for more details on all data used in the article. 2. Here and throughout the article, results that are not reported are available from the author on request. Table 1. Predictable Components in Emerging Equity Markets, December 1985 to December 1992 Coefficient estimates Predictability statisticsb U.S. ex- Local U.S. Allfive Local Global cess dollar Local ex- U.S. dividend interest instru- instru- instru- Residual returns, cess dollar dividend yield, rate, Adjusted ments, ments, ments, autocorrela- Market r1, returns, ri, yield, dy1, dyit ilt R2 x2(5) x2(2) X2(3) tion,c 1(5) 00 Argentina 0.39 -(.19 -3.55 -0.65 8.18 -0.022 1.95 1.18 1.33 2.72 (0.37) (0.18) (88.79) (13.74) (36.34) [0.86] [0.55] [0.72] [0.74] Brazil 0.34 -0.13 -37.3 29.2 1.96 0.025 9.66 7.60 1.44 6.67 (0.40) (0.10) (90.7) (10.74) (39.7) [0.085] [0.022] [0.70] [0.25] Chile 0.38 0.25 45.3 0.67 -7.38 0.111 25.8 5.67 6.68 7.39 (0.15) (0.11) (44.6) (4.71) (14.82) [0.0001] [0.06] [0.08] to.I9] Colombia -0.005 0.40 -3.73 2.94 -11.04 0.166 20.13 14.94 1.36 2.95 (0.18) (0.18) (37.3) (2.80) (11.33) [0.001] [0.0006] [0.71] [0.71] Greece 0.42 0.09 -50.8 2.14 8.74 0.003 9.12 1.98 5.66 5.12 (0.27) (0.08) (63.0) (4.75) (17.8) [0.10] [0.37] [0.13] [0.40] India -0.06 0.077 -34.3 41.5 -37.46 -0.011 4.78 4.03 4.23 2.98 (0.18) (0.14) (41.5) (22.2) (19.6) [0.44] [0.13] [0.24] [0.70] Jordan 0.0017 -0.20 -3.73 -1.84 -11.75 0.003 5.43 3.02 3.04 8.22 (0.09) (0.14) (14.2) (2.13) (7.55) [0.36) [0.22] [0.39] [0.14] Korea, Rep. of 0.26 -0.24 35.2 13.03 17.06 0.090 12.41 8.56 4.81 4.28 (0.16) (0.12) (33.64) (5.15) (10.81) [0.029] [0.014] [0.19] [0.51] Malaysia 0.10 0.01 86.66 68.53 -24.27 0.088 10.95 9.93 2.85 3.77 (0.20) (0.13) (52.22) (22.15) (15.8) [0.05] [0.007] [0.41] [0.58] Mexico 1.34 0.19 90.2 3.43 -24.5 0.259 8.04 2.40 5.22 7.99 (0.60) (0.15) (85.14) (5.58) (24.50) [0.15] [0.30] [(0.16] [0.16] Nigeria 0.26 0.08 -33.1 -0.83 16.1 -0.035 0.71 0.30 0.59 6.67 (0.35) (0.14) (71.1) (4.79) (25.43) [0.98] [0.86] [0.90] [0.25] Pakistan -0.13 0.21 -1.72 -0.92 8.39 0.015 3.83 0.88 1.79 10.37 (0.15) (0.22) (23.9) (6.15) (8.86) [0.57] [0.64] [0.62] [0.07] Philippines 0.19 0.27 30.61 14.9 2.33 0.183 27.42 19.1 2.37 5.53 (0.19) (0.10) (56.26) (6.22) (16.97) [0.00005] [0.00007] [0.50] [0.35] Portugal 0.72 -0.12 -279.2 -75.5 -4.04 0.299 29.8 10.7 29.00 8.77 (0.34) (0.17) (69.3) (23.2) (21.42) [0.00002] [0.005] [0.000002] [0.12] Taiwan (China) 0.64 -0.04 9.24 37.03 7.58 0.008 7.80 2.57 2.93 2.88 (0.40) (0.14) (87.6) (23.6) (26.24) [0.17] [0.28] [0.40] [0.72] Thailand 0.33 0.02 14.9 2.87 -3.07 -0.018 3.41 0.45 1.63 11.4 (0.28) (0.16) (65.14) (4.26) (19.03) [0.64] [0.80] [0.65] [0.04] Turkey 1.05 -0.01 -171.3 20.9 46.22 0.162 14.6 7.78 9.05 2.89 (0.46) (0.11) (109.0) (8.21) (30.96) [0.01] [0.02] [0.03] (0.72) Venezuela -0.37 0.21 -52.5 47.66 -7.98 0.116 14.6 9.83 2.81 1.59 (0.24) (0.08) (45.7) (18.5) (19.6) [0.012] [0.007] [0.42] [0.90] Zimbabwe -0.06 -0.008 24.72 12.5 12.44 0.207 25.3 17.1 3.08 2.53 (0.13) (0.10) (31.5) (3.65) (11.36) [0.0001] [0.0002] [0.38] [0.77] Emerging-markets 0.41 -0.0026 11.13 14.93 2.60 17.38 2.24 6.40 7.16 composite (0.19) (0.13) (43.5) (10.0) (13.76) 0.057 [0.004] [0.33] [0.09] [0.21] Note: Figures in parentheses are heteroskedasticity-consistent standard errors; those in brackets are p-values. a. In relation to a one-year backward-moving average. b. Tests on the joint explanatory power of all five instruments, x2(5); the two local instruments, x2(2); and the three U.S. instruments, x2(3). c. Tests for residual serial correlation using the first five autocorrelations of the residuals and is distributed x2(5) (Cumby and Huizinga 1993). Source: Author's calculations. Table 2. Predictable Components in Emerging Equity Markets, December 1976 to September 1985 Coefficient estimates U. S. Predictability statisticsb excess Local Local U.S. Allfive Local Global Chow- dollar excess U.S. dividend interest instru- instru- instru- Residual type test returns, dollar dividend yield, rate,a Adjusted ments, ments, ments, autocorrela- for Market rtt returns, r,, yield, dylt dyi, ill R2 x2(S) x2(2) X2(3) tion,c I(S) stabilityd Argentina 0.08 0.07 -56.7 30.6 3.036 -0.021 5.05 2.25 1.82 1.54 7.39 (0.52) (0.08) (45.7) (29.1) (11.7) [0.41] [0.32] [0.61] [0.91] [0.29] Brazil -0.37 0.13 -31.2 1.37 -2.40 -0.003 4.73 1.64 2.73 9.9 14.3 (0.30) (0.11) (24.8) (3.49) (7.04) [0.45] [0.44] [0.43] [0.08] [0.027] Chile -0.04 0.04 -12.9 20.3 7.53 0.079 13.5 9.26 2.32 6.41 10.0 (0.30) (0.09) (19.1) (7.11) (5.70) [0.02] [0.01] [0.51] [0.27] [0.13] Greece -0.12 0.02 -9.6 -3.62 -7.71 0.009 6.57 3.72 5.09 5.73 10.8 (0.12) (0.12) (12.2) (1.95) (3.75) [0.25] [0.15] [0.16] [0.331 [0.094] ,s India 0.40 -0.06 -15.3 9.30 -0.66 0.092 23.7 1.34 19.1 9.16 14.4 (0.10) (0.12) (9.54) (8.34) (2.84) [0.0002] [0.51] [0.0002] [0.10] [0.026] Jordan 0.08 0.07 15.80 31.26 -3.53 -0.012 4.46 3.46 1.73 7.29 7.72 (0.17) (0.12) (18.24) (20.48) (3.94) [0.48] [0.18] [0.63] [0.20] [0.259] Korea, -0.34 0.04 -26.1 3.47 3.28 0.005 9.00 0.37 6.76 3.93 10.1 Rep. of (0.19) (0.09) (13.9) (9.68) (4.71) [0.11] [0.83] [0.08] [0.56] [0.12] Mexico 0.62 -0.002 -34.7 0.59 12.45 0.034 12.4 0.01 11.77 5.95 4.33 (0.23) (0.14) (23.2) (6.62) (4.97) [0.03] [1.00] [0.008] [0.31] [0.63] Thailand 0.038 0.06 -26.3 --0.53 -4.14 0.039 18.7 0.40 16.8 4.58 2.76 (0.16) (0.10) (11.1) (2.92) (2.84) [0.002] [0.82] [0.0008] [0.47] [0.84] Zimbabwe 0.34 0.12 -10.0 -0.40 -2.97 -0.006 3.51 1.19 2.33 7.37 10.9 (0.30) (0.11) (20.2) (2.84) (6.81) [0.62] [0.55] [0.51] [0.19] [0.091] Note: Figures in parentheses are heteroskedasticity-consistent standard errors; those in brackets are p-values. a. In relation to a one-year backward-moving average. h. Tests on the joint explanatory power of all five instruments, x2(5); the two local instruments, x2(2); and the three IJ.S. instruments, x 2(3)- c. Tests for residual serial correlation using the first five autocorrelations of the residuals and is distributed X2(5) (Cumby and Huizinga 1993). d. Robust to heteroskedasticity on thc six coefficients in the regressions, including the constant (see, for example, Hodrick and Srivastava 1984). Source: Author's calculations. Bekaert 83 there is only marginal evidence of predictability; for the excess returns on Ger- man, Japanese, and U.K. equity, all R2s are negative and the Wald statistics never reject at the 5 percent level. This result is in sharp contrast to the large body of empirical literature on international predictability of equity returns (Bekaert and Hodrick 1992; Ferson and Harvey 1993; Harvey 1991). It is therefore of independent interest and deserves further scrutiny. Because similar instruments were used in previous studies, it is probable that the lack of significant predictability is specific to the more recent sample period. The differences between the 1985-92 sample and the 1976-85 sample are strik- ing. For the earlier period there is evidence of strong predictability that primarily derives from global instruments. The decrease in predictability for the later period complicates the interpretation of the predictable variation through global factors as an indicator of global-market integration. One possible explanation would be that the predictability is merely an indication of market inefficiency that was eliminated with increasing globalization at the end of the 1980s. Alter- natively, the nature of time-varying risk premiums may have changed, making them more difficult to track with the instruments typically used in empirical studies. For Japan, for instance, including the local interest rate or the forward premium as an instrument improves predictability marginally, whereas in Ger- many changes that have occurred in the exchange rate help to predict future returns. However, formal tests for stability fail to reject the hypothesis that the coefficients have not changed for Germany, Japan, and the United Kingdom, but the tests reject at the 1 percent level for the United States. And, based on the i-tests, the forecasting variables used here suffice to eliminate all serial correla- tion in the residuals. Table 2 reports the results for 1976-85 for ten emerging markets where data were available. The test for stability rejects for Greece and Zimbabwe at the 10 percent level and for India and Brazil at the 5 percent level. There is no clear pattern in how the predictability patterns move over time. For example, it is striking how the predictability arising from global factors was actually stronger in the early period for Greece, India, Korea, Mexico, and Thailand. The appar- ent decline of global predictability is not necessarily inconsistent with the fact that most markets became more open to foreign investment during the 198Cls (see below). In sum, expected returns generally vary through time, although predictability is stronger for both industrial and emerging markets before 1986 than it is in the late 1980s and early 1990s. I conclude that predictability tests do not yield much useful information on market segmentation. II. A MEASURE OF MARKET INTEGRATION I interpret the fitted values of the regressions of excess returns on five prede- termined variables to be estimates of expected returns. There are several asset- 84 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 pricing models that justify this procedure. For example, suppose the returns satisfy a multifactor model with expected returns depending on the risk loadings (Os) with respect to risk factors and on the prices of these risks (their expected returns). In a K-factor model, the conditional expected value of an excess return, ri,,,, satisfies K (1) Et(rit+1) E Z (ik Xk k=1 where fikt is the factor loading of asset i for the kth factor at time t and Xkt is the market price of risk for the kth factor at time t. To yield a projection equation on a number of forecasting variables as the reduced-form model, several auxiliary assumptions are needed. One sufficient set of assumptions is constant Os and time-varying prices of risk, with the time variation assumed to be a linear function of the information set (see, for example, Campbell and Hamao 1992). To see this, let Zt be a vector of forecasting variables, Zt = (r1t, ,rt, dy1t, dyit, it)', where r1l is the U.S. lagged excess return, ri, is the local lagged excess return, dy1l is the U.S. dividend yield, dyit is the local dividend yield, and it is the U.S. interest rate relative to a one-year backward-moving average. Let L (2) Xkt =X Ik Zlt ikt = Oik 1=1 with a/lk the sensitivity of the kth price of risk to the Ith variable in Zt. Combining equations 1 and 2, K L L (3) Et(riti+) = E Oik EX alk Zlt E ilzl k=1 1=1 1=1 The 6i, coefficients can be recovered from a linear projection of rit+ 1 onto Zt. Alternatively, the Os could be assumed to be linear functions of the informa- tion set (see, for example, Ferson and Harvey 1993), which also would imply an equation such as equation 3. In either case, the coefficients on the forecasting variables are a function of coefficients that determine the O3s or prices of risk in a multifactor model. The advantage of this reduced-form approach is that it is model free, and the factors do not have to be specified or measured. Allowing for time variation in expected returns is important, given the rapidly changing nature of the economies and stock markets. The evidence for predictability detected in the previous section confirms the presence of time variation in ex- pected returns. My integration measure is the correlation of the regression esti- mates of the expected returns in the United States and the emerging markets. This correlation is an indicator of the common component in expected returns and hence, indirectly, of market integration. A couple of caveats must be noted. First, because I compute the unconditional correlation coefficient, no changes in the degree of market integration are al- lowed over the sample period. This is another motivation for using the relatively short sample (1986-92) as opposed to the full sample available, which is longer Bekaert 85 for many economies. I examine whether integration changed over time by com- puting the correlations for an earlier sample as well. Harvey (1995) computes five-year rolling correlations between emerging-market returns and the world market. His results suggest that these correlations are increasing for many emerging markets. Second, I want to stress that the measure is only a perfect measure of market integration in a one-factor world with constant risk exposures. Suppose the world equity market is fully integrated and assets are priced according to a multifactor model. Emerging markets might display dramatic cross-sectional differences in their risk exposures. These differences, in turn, might affect the correlation of expected returns with the U.S. market, without reflecting actual barriers to investment (broadly defined). For instance, the various emerging markets have different industrial structures, which might result in different ex- posures to "industry factors" (see Divecha, Drach, and Stefek 1992 on emerging markets, and Heston and Rouwenhorst 1994 and Roll 1992 on industrial mar- kets). Moreover, some economies are dependent on a limited number of natural resources (for example, Nigeria on oil), which might give rise to different "com- modity exposures." In section VI, I briefly assess the importance of these indus- try and commodity factors in the measurement of market segmentation. In table 3, I report three different correlations. The regression decomposes the return into an expected and unexpected part. The reported correlations are then the correlation of the return, of the expected return, and of the unexpected return in economy i with its counterparts in the United States. The methodology borrows from Bekaert and Hodrick (1992) and Bekaert (forthcoming). Assume that Zj, which includes the U.S. excess return, the emerging-market excess return, the two dividend yields, and the relative U.S. interest rate, follows a first-order vector autoregression: (4) Zt+= i + A Zit + uit+,1 If the vector autoregressive framework is correctly specified, E,(uit+1) = 0. Let the variance-covariance matrix of the innovations ui, be V. Let E be the variance-covariance matrix of Zj,t As it is found from (5) vec (E) = (I - A' ® A' )-1 vec(V), V and E are sufficient to compute the correlation of returns and unexpected returns. To compute the correlation of expected returns, the covariance matrix of EI(Zil+1), SE, is derived to be (6) ZE = A EA' . Standard errors are obtained by estimating A and V using the general method of moments and applying the Mean Value Theorem. Note that this technique assumes that the vector autoregressive framework generates the expected returns correctly. If there is measurement error in the resulting expected-return estimate s that is uncorrelated across the United States and emerging markets, the estl'- Table 3. Return Correlations with the U.S. market, 1976-85 and 1985-92 1985:12-1992:12 1976:12-1985:09 Rank based Rank on expected- based on Expected Unexpected return a Expected Unexpected Return, return, return, correlation variance Return, return, return, Market p (r1, ri,) p (rpi, rpid p (ul, ui,) estimatea ratiol, p (rl, rit) p (rp1,, rpit) p (uit, Uit) Argentina 0.10 -0.14 0.12 12 12 0.03 -0.575 0.07 (0.09) (0.56) (0.11) (0.08) (0.45) (0.075) Brazil 0.13 -0.06 0.15 14 14 -0.07 -0.41 -0.04 (0.08) (0.38) (0.09) (0.10) (0.33) (0.09) °o Chile 0.32 0.485 0.30 5 11 -0.11 -0.47 -0.07 (0.12) (0.48) (0.17) (0.08) (0.31) (0.08) Colombia 0.11 -0.16 0.16 15 19 - - - (0.12) (0.46) (0.08) Greece 0.145 -0.38 0.19 18 20 0.04 0.57 0.005 (0.11) (0.45) (0.11) (0.08) (0.32) (0.08) India -0.13 -0.57 -0.07 21 8 0.03 -0.14 0.06 (0.075) (0.35) (0.08) (0.08) (0.32) (0.08) Jordan 0.06 -0.44 0.10 19.5 22 0.05 0.13 0.04 (0.13) (0.43) (0.14) (0.1() (0.53) (0.10) Korea, Rep. of 0.21 0.135 0.21 11 7 0.04 -0.67 (.(95 (0.08) (0.39) (0.08) (0.10) (0.34) (0.10) Malaysia 0.66 0.80 0.64 3 9 - - - (0.07) (0.24) (0.10) Mexico 0.49 0.33 0.54 7 10 0.125 -0.77 0.21 (0.09) (0.52) (0.13) (0.085) (0.21) (0.09) Nigeria 0.04 -0.33 0.06 13 21 - - - (0.08) (0.78) (0.06) Pakistan -0.02 0.25 -0.05 10 16 - - (0.10) (0.37) (0.10) Philippines 0.29 0.74 0.20 2 6 - - - (0.16) (0.34) (0.14) Portugal 0.26 -0.265 0.43 17 13 (0.10) (0.45) (0.11) Taiwan (China) 0.195 0.12 0.20 9 5 (0.12) (0.67) (0.13) Thailand 0.43 0.30 0.44 6 4 -0.09 -0.02 -0.10 (0.14) (0.68) (0.16) (0.11) (0.46) (0.11) Turkey -0.16 -0.71 -0.08 22 15 (0.15) (0.34) (0.11) Venezuela -0.06 -0.35 -0.02 19.5 17 - - - (0.07) (0.34) (0.10) Zimbabwe -0.14 -0.01 -0.18 16 18 0.13 0.28 0.12 (0.09) (0.28) (0.11) (0.10) (0.415) (0.105) Emerging-markets composite 0.40 0.19 0.42 - - - - - (0.12) (0.77) (0.15) Germany 0.42 0.73 0.43 4 1 0.25 0.31 0.24 (0.10) (0.345) (0.12) (0.10) (0.27) (0.12) 00 Japan 0.23 0.34 0.23 8 3 0.21 -0.21 0.25 (0.09) (0.49) (0.09) (0.09) (0.445) (0.10) United Kingdom 0.67 0.96 0.67 1 2 0.39 0.56 0.37 (0.06) (0.13) (0.07) (0.07) (0.17) (0.08) -Not available. Note: The correlations are computed using the dynamic structure of a vector autoregressive framework on the U.S. excess return, the emerging-market excess return, the two dividend yields, and the relative U.S. interest rate. Standard errors (in parentheses) are computed as in Bekaert (forthcoming) using three Newey-West lags. a. The ranking is based on the sum of two ranks: one according to the point estimate of the correlation of expected returns estimated for the most recent sample, one based on the number of standard errors away from perfect correlation computed for the same sample. b. Returns are regressed on the instruments plus the world market portfolio return. The statistic used for the rankings is the ratio of the predictable variation caused by the instruments in the model with the world market portfolio as an observable factor relative to the predictable variation in the regressions without the world market portfolio reported in tables 1 and 2. See Campbell and Hamao (1992). Source: Author's calculations. 88 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I mated correlations will overestimate the true degree of expected-return correlation. By far the highest expected-return correlation in table 3 is observed for the United Kingdom (0.96), as would be expected given the high degree of integra- tion and the extent of cross-listing of securities between the London and New York markets. Germany, Malaysia, and the Philippines exhibit correlations of over 0.60. Japan has an expected return correlation of about 0.34, which is similar to the expected-return correlations of Chile, Mexico, and Thailand, which are 0.49, 0.33, and 0.30, respectively. Korea, Pakistan, and Taiwan (China) have expected-return correlations of 0.14, 0.25, and 0.12, respectively. All the other economies display negative expected-return correlations. Most markets show fairly large correlations for their unexpected returns. Hence, there must exist global news factors affecting many markets simultaneously, including the emerging markets. The results for the 1976-85 sample conform to the trend toward increasing integration of equity markets. According to this measure, the industrial markets all became more integrated with the U.S. market during the last half of the 1980s, the change being most dramatic for Japan. Chile, Korea, and Mexico show negative correlation with the U.S. market in the early sample. In fact, before 1984, when the first Korean country fund was introduced, the Korean market was virtually closed to foreign investment. Some conundrums, however, do exist. For example, markets in Greece and Zimbabwe show high, albeit imprecisely measured, expected-return correlations with the U.S. market in the early sample. In table 3 the rank based on the expected-correlation estimate is the sum of a ranking on the point estimate and a ranking on the size of the deviation from perfect correlation in number of standard errors. The expected-return correla- tions might not give an adequate picture of the common component in expected returns because the evidence for predictability is weak for some markets. To check the robustness of the results, I also provide an alternative measure of market integration based on the analysis in Campbell and Hamao (1992). Sup- pose that the emerging equity markets obey a multifactor model, where the first factor is international and the other factors are domestic; suppose that the international factor is well proxied by the world-stock-index return; and con- sider a regression of the excess equity returns on that world-market return and the forecasting variables. The variance of the predictable variation caused by the forecasting variables in that regression, in relation to the variance of the fitted values of the regressions reported in table 1, is a measure of the variation in risk prices of domestic factors relative to the variation in the risk prices of all factors. I interpret low ratios as indicative of more integration. In table 3, the column "Rank based on a variance ratio" ranks the markets on the basis of this ratio. For lack of space, further results are not reported. There are some notable differences in the rankings based on the variance ratio compared with the earlier rankings (India and Nigeria are examples), but the Bekaert 89 rank correlation between the measures is 0.693, which is more than three stan- dard errors from zero. The ratio is lower than 0.7 for only two emerging markets, Taiwan (China) and Thailand. I also checked whether inclusion of the world-market return changed the predictability tests. Significant rejections of the null of no predictability only disappeared for Greece (at the 10 percent level) and for Malaysia (at the 5 percent level). I also substituted a regional index for the world-market index to test whether there was any evidence of regional integration. (From Morgan Stanley Capital International I used the Pacific index for the Asian markets, the Europe index for the European and African markets, and the North America index for the Latin American markets.) The only mar- kets for which the variance ratio dropped relative to the world-market regres- sion were Chile and Korea. The regional Os were substantially higher only for Argentina, Chile, Mexico, and Venezuela. Hence there is weak evidence that regional integration is stronger than global integration in Latin America. In what follows, I will occasionally refer to results that use the ratio when they differ from the results that use the expected-return correlation measure. III. MARKET INTEGRATION AND RETURN CHARACTERISTICS In this section I provide a fuller picture of the properties of emerging-equity- market returns in order to relate them to various measures of barriers to invest- ment. Some of the return properties might be correlated with popular cost-of- capital measures. Tables 4 and 5 summarize some return properties for the two sample periods for nineteen emerging markets, an emerging-markets composite, and four major industrial markets. The first three columns report the mean, standard deviation, and Sharpe ratio. The Sharpe ratio is a measure of the risk-return tradeoff, computed as the excess return divided by the standard deviation of the excess return. Emerging markets offer higher but more variable returns compared with the industrial markets, although there are some notable exceptions to this rule (for example, in Jordan and Zimbabwe). The risk-return tradeoff during 1986- 92 is most favorable in a number of Latin American markets (in Chile, Col- ombia, and Mexico), and it is generally better in emerging markets than in the industrial world. The composite index has a slightly higher mean return than the U.K stock market and a slightly higher risk. Its diversification potential stems from the relatively low correlation with the industrial markets. This is further illustrated in figure 1. Tables 4 and 5 report the constant (a) and slope coefficient (fi) of a regression of the excess return onto a constant and the world-market return. The Capital Asset Pricing Model (CAPM) would predict that the a coefficient equals zero. The major markets display very high Os with respect to the world-market portfolioc and relatively small as (pricing errors). In the emerging markets, high Os are found for Brazil, Korea, Malaysia, Mexico, the Philippines, Taiwan (China), and Thailand. Chile, Colombia, Mexico, the Philippines, and Thailand alsoa 90 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Table 4. Properties of Emerging-Market Equity Returns, 1986-92 World risk First-order Dollar excess returns World loading autocor- Standard Sharpe pricing coeffi- relation Market Mean deviation ratioa error, a cient, 3 coefficient Argentina 62.048 104.396 0.594 64.735 -0.422 -0.082 (40.616) (0.638) Brazil 22.299 77.824 0.287 17.368 0.773 -0.030 (28.419) (0.504) Chile 41.186 28.508 1.445 39.593 0.250 0.310* (10.788) (0.243) Colombia 43.069 33.524 1.285 41.899 0.184 0.479* (12.868) (0.193) Greece 26.684 50.089 0.533 23.627 0.479 0.120 (18.720) (0.285) India 7.476 35.458 0.211 9.638 -0.339 0.103 (13.473) (0.212) Jordan -2.369 17.139 -0.138 -3.113 0.117 -0.160 (6.539) (0.130) Korea, Rep. of 16.836 32.028 0.526 13.115 0.584 -0.099 (11.388) (0.178) Malaysia 11.782 26.520 0.444 6.789 0.783 0.031 (9.031) (0.212) Mexico 49.925 48.832 1.022 44.419 0.864 0.355 (18.239) (0.374) Nigeria -5.263 39.075 -0.135 -6.720 0.229 0.086 (15.132) (0.217) Pakistan 15.455 24.739 0.625 15.288 0.026 0.255* (9.408) (0.151) Philippines 38.654 40.482 0.955 33.559 0.799 0.345 * (14.575) (0.278) Portugal 27.666 50.224 0.551 20.087 1.189 0.287* (17.364) (0.242) Taiwan (China) 29.377 55.994 0.525 24.810 0.716 0.058 (21.190) (0.429) Thailand 29.567 30.964 0.955 25.247 0.678 0.114 (11.566) (0.301) Turkeyb 30.815 74.146 0.416 30.212 0.247 0.114 (30.036) (0.402) Venezuela 32.684 46.452 0.704 34.454 -0.278 0.312* (18.465) (0.337) Zimbabwe 1.245 26.669 0.047 1.428 -0.029 0.280* (10.338) (0.195) Emerging-markets 10.453 25.249 0.414 7.215 0.508 0.130 composite (9.370) (0.221) Germany 2.909 25.113 0.116 -2.603 0.865 -0.083 (7.678) (0.152) Japan 8.156 29.528 0.276 -0.790 1.403 0.008 (6.959) (0.174) United Kingdom 10.372 23.173 0.448 3.374 1.098 -0.049 (5.218) (0.078) United States 8.210 17.091 0.480 3.563 0.729 -0.006 (4.668) (0.109) ' Significant at the 5 percent level. Note: All returns are annualized percentages. The reported mean is arithmetic. Standard errors are in parentheses. a. The excess mean return scaled by the standard deviation. b. Data begin January 1987. Source: Author's calculations. Bekaert 91 Table 5. Properties of Emerging-Market Equity Returns, 1976-85 First- order World risk autocor- Dollar excess returns World loading relation Standard Sharpe pricing coeffi- coeffi- Market Mean deviation ratioa error, a cient, ,S cient Argentina 61.236 105.525 0.580 60.265 0.321 0.151 (33.981) (0.505) Brazil 7.612 44.323 0.172 7.848 -0.078 0.116 (14.115) (0.283) Chile 16.977 46.527 0.365 17.223 -0.081 0.130 (15.150) (0.353) Greece -21.814 20.119 -1.084 -22.769 0.315 0.077 (6.314) (0.158) India 11.502 19.721 0.583 10.258 0.411 -0.003 (6.013) (0.135) Jordanb 4.932 18.651 0.264 4.360 0.264 0.115 (6.996) (0.170) Korea, Rep. of 5.490 32.437 0.169 4.280 0.400 0.036 (10.158) (0.265) Mexico -1.291 40.754 -0.032 -3.003 0.565 0.136 (12.873) (0.310) Thailand 3.547 21.519 0.165 3.571 -0.008 0.101 (6.982) (0.122) Zimbabwe -3.601 39.089 -0.092 -5.420 0.601 0.099 (12.383) (0.335) Germany 2.307 17.995 0.128 0.315 0.750 0.022 (0.138) (0.123) Japan 9.150 18.807 0.487 0.309 0.910 -0.016 (0.129) (0.129) United Kingdom 7.583 23.053 0.329 0.607 1.171 0.038 (0.158) (0.143) United States 2.138 14.071 0.152 -0.849 0.987 -0.023 (2.221) (0.060) Note: All returns are annualized percentages. The reported mean is arithmetic. Standard errors are in parentheses. a. The excess return scaled by the standard deviation. b. Data begin January 1979. Source: Author's calculations. display significantly positive as. Clearly, an unconditional world CAPM model does not explain much of the cross-sectional variation in emerging-equity- market returns. Hence, it would be incorrect to conclude that higher Os increase the cost of capital. On the contrary, high Os seem to indicate a higher degree of integration with the industrial world. Compared with the earlier sample, the Os and the Sharpe ratio have increased for most but not all emerging markets. The exceptions are Argentina, India, and Zimbabwe.3 Finally, tables 4 and 5 report the first-order autocorrelation coefficient for the various markets. This coefficient is clearly insignificantly different from zero for 3. It will be interesting to see whether the recent capital-market liberalization in India will have an effect on these statistics. 92 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO_ I Figure 1. Mean-Standard Deviation Frontiers of Monthly Dollar Total Returns of Selected Portfolios, January 1986-December 1992 (percent) Mean of total monthly retums 80 - / , _, _ 70 - - / _, _._- l Argentina 60 - o o Mexico - - / Colomb~ia -',~ 50 - Chile Philippines ....- a~~~~~~~ / o o~- '--' Vezul 40 - Thailand -ez a , ..-... a Taiwan 30 - Pakistan '- Greece Portugal X Brazil EB ndex , <' a Korea 20 - u s7'- 5a Malaysia W;orldf(t a U.K. a India 10 NTVP Zi,Japan aZimbabwe O - ordanrdanGermny aNigeria 0 10 20 30 40 50 60 70 80 90 10( 110 120 Standard deviation Emerging Markets, Germany, Japan, United Kingdom, and United States. - - - - Emerging Markets Index, Germany, Japan, United Kingdom, and United States. - -- -- Germany, Japan, United Kingdom, and United States. Aote. MVP is the minimum-variance portfolio. Source: IFC EMDB. the major markets, but it is significantly positive for some emerging markets, potentially signaling market inefficiencies. However, not a single emerging mar- ket displays significant positive serial correlation in the early sample. Table 6 contains information on dividend yields and price-earnings (P/E) ratios. Both variables exhibit large cross-sectional and time variation in the emerging markets. As Buckberg (1995) points out, P/E ratios typically increase substantially when a market is opened to foreign investment. By the same token, openness would result in lower dividend yields. Significant increases in P/E ratios coupled with significant decreases in dividend yields over the sample period are observed for Colombia, Mexico, and Pakistan. Both variables are factors in simple cost-of-capital computations and are likely to be affected by the degree of market segmentation. However, given the large differences in corpo- rate and accounting practices, the absolute magnitude of dividend yields or P/E ratios may not be very informative on market segmentation. Table 7 provides a matrix containing the rank correlations between all the return characteristics discussed above, including the measure of market integra- tion. The ranking for all measures is such that the three industrial markets rank high. For example, markets with high values for mean return, volatility, Sharpe ratio, the ar-pricing error from the world-market model, and dividend yield get a Table 6. Dividend Yields and Price-Earnings Ratios in Emerging Markets, 1986-92 (percent, annual averages) Dividend yields- Price-earnings ratio' Rankb Rankb Market 1986-92 1986 1989 1992 1992 1986-92 1986 1989 1992 1992 Argentina 3.71 0.20 11.82 0.73 22 2.81 6.25 106.24 -25.19 23 Brazil 5.85 3.63 5.87 3.30 10 26.30 5.73 5.66 132.48 1 Chile 7.89 8.53 10.93 3.61 8 7.61 3.59 4.34 14.57 12 Colombia 8.19 15.39 7.62 3.50 9 6.71 -27.71 5.24 34.55 4 Greece 7.38 9.95 10.13 6.91 1 15.05 7.14 13.04 8.68 20 India 2.37 2.27 2.93 1.13 17 19.25 12.81 15.40 38.58 3 Jordan 4.82 3.00 3.63 4.23 5 11.56 12.13 13.66 12.22 17 Korea, Rep. of 2.04 6.20 1.08 0.02 23 21.71 16.43 31.18 19.17 11 Malaysia 2.42 2.26 2.64 2.62 13 28.05 21.40 29.65 20.78 8 Mexico 4.23 10.74 3.85 1.38 16 9.41 4.89 5.73 12.72 16 Nigeria 8.52 9.81 12.21 6.84 2 6.53 4.95 5.61 10.19 18 Pakistan 6.56 7.03 6.94 3.19 11 10.78 6.23 7.96 24.87 5 w Philippines 2.75 7.75 1.82 0.91 20 12.62 - 13.13 13.99 14 Portugald 2.27 1.59 1.79 3.76 7 15.11 10.85 16.75 9.82 19 Taiwan (China) 1.12 2.42 0.97 1.08 18 29.23 13.44 57.94 19.58 10 Thailand 4.71 9.77 3.87 2.44 14 12.47 8.47 14.01 13.73 15 Turkey 7.07 - 11.58 4.74 4 11.29 - 6.71 7.39 21 Venezuela 1.64 3.13 1.88 0.79 21 14.43 11.81 4.76 22.10 7 Zimbabwe 8.86 10.10 11.31 5.03 3 5.05 3.80 3.69 3.47 22 Germany 2.22 1.82 2.17 2.34 15 15.60 17.80 14.70 14.30 13 Japan 0.67 0.79 0.46 1.01 19 45.50 51.90 45.70 38.90 2 United Kingdom 3.60 4.03 3.41 3.80 6 14.93 11.70 13.40 19.70 9 United States 3.42 3.57 3.46 3.02 12 16.97 14.10 14.10 22.70 6 - Not available. a. The dividend yield can be interpreted as a twelve-month reinvested average yield. b. From high to low. c. Twelve-month average computed by the IFC for IFC index stocks. d. 1987 values substitute for 1986 values. Source: Author's calculations. The earnings and dividend measures for the major markets are taken from Morgan Stanley and Company (various issues). Table 7. Rank Correlations between Market Integration and Other Return Characteristics, 1986-92 Pricing error, a, Expected- Price- Mean Sharpe Risk from the world- return Dividend earnings Characteristic returns Volatility ratio loading, ,B market model correlation yield ratio Market integration -0.092 0.336 -0.268 0.610 -0.023 0.234 0.269 0.207 Mean returns 0.563 0.872 0.035 0.942 0.493 0.007 0.269 Volatility 0.209 0.054 0.472 0.051 -0.055 -0.010 Sharpe ratio -0.027 (.905 0.612 0.028 0.251 Risk loading, / 0.175 0.228 0.403 0.557 Pricing error, ae 0.584 0.156 0.366 SD) First-order autocorrelation 0.357 0.503 Dividend yield 0.738 Note: The mean returns, volatility, Sharpe ratio, risk loading, pricing error, and autocorrelation coefficients are the return characteristics for the 1986-92 sample depicted in table 4. The 1986-92 sample averages for dividend yield and price-earnings ratio are given in table 6. The rank correlation coefficient is the Spearman rank correlation computed for nineteen emerging markets and Germany, Japan, and the United Kingdom. The standard error for each statistic is 0.22. The ranking for all measures is such that the three industrial countries rank high, that is, from low to high for the mean, volatility, Sharpe ratio, pricing error from the world- market model, and dividend yield and from high to low for the world-market risk loading and the price-earnings ratio. For the pricing error, the ranking is based on the sum of two ranks, one according to the absolute magnitude of the pricing error, another according to distance in the number of standard deviations away from zero. For the expected return correlation coefficients, countries with insignificantly small autocorrelations (smaller than 0.100 in absolute magnitude) are given the same rank. Source: Author's calculations. Bekaert 95 low rank for that particular statistic. The ranking is from low to high because the industrial markets typically display low values for these statistics. Likewise, the ranking is from high to low for expected-return correlation, world market 3, and P/E ratio because the industrial markets typically have high values for these statistics. To help interpret the numbers, consider two examples. First, a positive rank correlation between the market-integration measure and volatility indicates that low volatility is associated with a high degree of market integration because industrial markets display low volatility. Second, a positive rank correlation between market integration and the P/E ratio indicates that higher P/E ratios typically imply higher degrees of market integration. The only significant relation between the market-integration measure and other return characteristics is with the world 3. As conjectured above, higher fs are associated with higher degrees of market integration and do not necessarily translate into higher expected returns. Although not significant, the rankings also reveal positive associations between market integration and P/E ratios and negative associations between market integration and dividend yields. This con- firms the intuition that the capital flows associated with opening up markets tend to increase P/E ratios and decrease dividend yields. Similarly, it would be expected that market integration contributes to domes- tic market efficiency. Because the autocorrelation ranking is from low to high, table 7 reveals the association between market-integration and the first-order autocorrelation coefficient to be negative but not significantly different from zero. The table also shows that the market-integration measure and volatility co-vary negatively. This supports my conjecture that concerns about excess price volatility in newly opened emerging stock markets might be unnecessary. The alternative market-integration measure yields similar results (not reported) with the exception that the associations with dividend yields and P/E ratios are significantly different from zero in this case. There is no relation between mean returns and either the market-integration measures or the Ols. Harvey (1995) examines the sensitivity of the emerging- market returns to measures of global risk, including the world-market portfolio. He finds that emerging markets have little or no sensitivity, which confirms the results of table 7. There is a strong positive rank correlation between average returns and volatility, the pricing error, and the autocorrelation coefficients. Consequently, high mean returns cannot be explained by the world-market model, but they might partially reflect inefficiencies in domestic markets. IV. MARKET INTEGRATION AND BARRIERS TO INVESTMENT Foreign investors face many barriers when investing in emerging markets. I distinguish two groups of direct barriers to investment and one group of indirect barriers. In the first group are direct restrictions on foreign ownership. For example, certain sectors may be closed to foreign investment, or limits may be imposed on direct ownership of equity. Table 8. Foreign Ownership Restrictions and Dates of Recent Liberalizations Affecting Foreign Investors Percent Investable Exchange rate investable, index/global Market regime, 1991 1992 index (ratio) Liberalization Date Argentina Free float 100 88.2 All limits on foreign capital abolished December 1989 Brazil Free float 49a 60.4 Group of foreign investment trusts March 1987 approved Interbank foreign exchange market March 1990 allowed Foreign ownership levels increased May 1991 Foreign portfolios without local custody July 1991 allowed Chile Pegged to basket 25 20.9 Non-Central Bank foreign exchange April 1990 market authorization Colombia Central Bank control 100 76.0 Made 100 percent investable February 1991 Greece Managed float 100 80.8 n.a. n.a. India Free float 24 19,1 All shares made investable November 1992 Managed exchange rate abolished March 1992 Jordan Pegged to basket 49 29.0 n.a. n.a. Korea, Rep. of Pegged to dollar 10a 9.6 Government-announced sweeping December 1988 liberalization Investment preapproval rules softened January 1990 Market average exchange rate system March 1990 C1\ ~~~~~~~~~~~~~~~~~~~~~~introduced Foreign ownership levels increased January 1992 Malaysia Free float 30 67.4 n. a. n.a. Mexico Free float 100k' 87.7 Made 100 percent investable May 1989 Dual exchange rate system unified November 1991 Nigeria Pegged to French franc 0. 0.0 n. a. n.a. Pakistan Pegged to dollar 100 29.3 Made 1 00 percent investable February 1991 Philippinies Free float 40- 47.3 All shares made investablc November 1991 Portugal Europeani Monietary Systemi 100' 54.1 n.a. n.a. I-aiwan (China) Central Banik control 10', 3.0 Equity market broadly openied, $5 January 1991 billioni maximnum foreign holdings Maximum foreign security holdings March 1993 limit increased to $l()0 billion Thailand Pegged to basket 100' 27.0 n.a. n . a. Turkey Free float 100 97.3 n.a. n.a. Venezuela Free float 100 36.3 Foreign ownership allowed with limits December 1988 All restrictions lifted January 1990 Zimbabwe Pegged to basket -. 0.0 n.a. n.a. nia. Not applicable. a. Industry exceptions. Bekaert 97 In the second group are exchange and capital controls that affect investment in emerging markets and the repatriation of dividends and capital from emerging markets. For example, some economies have direct restrictions, such as a mini- mum investment period, on the remittance of profits. Taxes on dividends and capital gains are considered direct barriers in this second group (see Demirgiu- Kunt and Huizinga 1992). Some economies, such as Nigeria and Zimbabwe, are still completely closed to foreign investment. Overall, however, restrictions have been gradually relaxed, and this process has accelerated in the 1990s. Examples of economies in which restrictions have been recently lifted include Brazil, Col- ombia, India, Korea, and Taiwan (China). Table 8 gives some information on both groups of direct barriers. A more detailed survey of the existing restrictions on foreign investors at the end of 1992 is given in IFC (various issues). In the third group are indirect barriers having to do with the regulatory and accounting environment. Investors might not have adequate information on these markets and on the financial health of the companies, the settlement systems might be inefficient and slow, accounting standards might be poor, and investor protection might be minimal. These factors might play a large role in the investment decisions of international investors. In her survey of market participants, Chuhan (1992) lists limited information on emerging markets as one of the key impediments to investing in emerging markets. I considered several measures of "openness" and their relation to the market- integration measure. The difficulty was to quantify the extent of the restrictions in the various economies in order to make the computation of rank correlations possible. The IFC has recently launched indexes that take direct foreign owner- ship restrictions into account. The investable market capitalization of each stock is used for its weight in the index instead of the stock's total market capitaliza- tion, as in the IFC's regular global indexes.4 Consequently, one measure of the extent of foreign ownership restrictions is the ratio of the IFC investable index to the IFC global index. That ratio is reported in table 8 and is the basis of my openness measure, Open I. Unfortunately, the scarcity of available information prevented me from rank- ing the economies directly according to the severity of other capital and ex- change restrictions. To gauge the effects of these restrictions indirectly, I com- puted a ranking based on the mean black-market premium and a ranking based on the sums of the ranks according to the mean and volatility of the premiums during 1988-92. The data used for these calculations are described in Chuhan, Claessens, and Mamingi (1993). Because some economies (Greece, Jordan, Nigeria, Portugal, Turkey, and Zimbabwe) are missing from the data set, the rank correlations have a standard error of 0.25 8. However, the rank correlation between black-market premiums (ranked from low to high) and the market- 4. For details on how a variety of restrictions on foreign ownership change the weights used to construct the index, see IFC (various issues). 98 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I integration measure is 0.711 if the ranking is based on the means and 0.697 if the ranking is based on the means and variances. For indirect barriers, I used the EMDB table on the availability of market and company information, and the quality of accounting standards and investor protection, as reported by Harvey (1993). From this information, I computed a summary measure (unreported), which is the basis for my "Open II" ranking. Despite the persistence of various restrictions on foreign investors, several emerging markets have been open to some form of foreign investment for a surprisingly long time. One of the first vehicles for foreign investment in emerg- ing markets was country funds. Four Asian economies-Korea, Malaysia, Tai- wan (China), and Thailand-individually have more than ten country funds listed abroad. There is of course potentially very useful information on market integration in the premiums that some of these closed-end funds command when traded in industrial markets (see, for instance, Bekaert and Urias 1994 and Diwan, Senbet, and Errunza 1992). More recently, some companies in emerging markets have begun to list their stock on the exchanges of industrial markets. No less than thirty Mexican companies are listed on American exchanges. I used the number of country funds and cross-listed securities to construct a third measure of openness, "Open III." The measure is imperfect because the lack of data has prevented me from weighting the funds and companies by market capitalization, and the cross-listings are restricted to the United States. I calculated the rank correlations for the Open I, Open II, and Open III measures for the emerging markets (not reported). The rank correlations be- tween market integration and the three measures are 0.214 for Open I, 0.601 for Open II, and 0.794 for Open III. The market-integration measure is most significantly positively associated with the Open III measure. This result indi- cates that the best way to effectively open up a market may be to mobilize foreign resources through country funds or cross-listed securities. Such an ap- proach confirms the theoretical analysis of Diwan, Senbet, and Errunza (1992). They show that country funds, despite their small size, contribute significantly to capital mobilization and pricing efficiency in the originating capital markets. These results are robust to the use of the alternative measure of market segmen- tation; in fact, they are even stronger with the alternative measure than without it. Somewhat surprisingly, the relation between the market-integration measure and the Open I measure is not significantly positive: either the ownership restric- tions are circumvented or they are not binding. The Open I measure does correlate significantly with the world-market O3s (not reported). Markets with less severe ownership restrictions tend to have high Os. In the theoretical analysis of Eun and Janakiramanan (1986) and Stulz and Wasserfallen (1992), the pres- ence of ownership restrictions in a world CAPM leads to higher expected returns for foreign investors and to "home bias" in their portfolio holdings. How this super-risk premium affects the empirical estimates of ce and j3, however, is unclear. Again, the data reveal that openness goes hand in hand with higher Os. Bekaert 99 The rank correlation between the world O3s and the Open II and Open III measures is even higher than that between the world Os and the Open I measure. (The correlations with the world Os are 0.447 for Open I, 0.721 for Open II, and 0.700 for Open III.) The Open II measure correlates significantly with the market-integration mea- sure. The results suggest that providing more and better information on the markets and companies and improving accounting standards and investor pro- tection should contribute to making emerging markets better integrated in the global equity market. In fact, such simple policy actions might be more impor- tant than fully abolishing ownership restrictions. The Open II and Open III measures also correlate positively with P/E ratios and negatively with dividend yields. These results confirm that open markets tend to have lower dividend yields and higher P/E ratios. There are few other significant associations between return characteristics and the openness mea- sures. In particular, there is no significant relation between the openness of a market and stock return volatility. Therefore, the fear that foreign-market access leads to more volatile markets might be ill-founded. In fact, the relatively high correlations with the autocorrelation measure, although not statistically signifi- cant, suggest that opening up markets is likely to improve domestic market efficiency. V. MARKET INTEGRATION AND EMERGING-MARKET-SPECIFIC RISKS The first emerging-market-specific risk (EMSR) I investigated is political or, more broadly, country risk. Political instability and economic mismanagement might add substantial risk premiums to returns and deter some foreign investors. A crude and indirect measure of political risk is the secondary-market price of bank debt. Unfortunately, this is only available for a limited number of econ- omies (unreported). Nevertheless, it is remarkable how the prices of Mexican and Chilean debt increased recently in conjunction with investors' renewed in- terest in these markets. Between 1989 and 1992, the price increased from 60.6 cents to 89.4 cents to the dollar for Chilean debt and from 39.7 cents to 64.3 cents to the dollar for Mexican debt. For all markets, a more direct measure of political risk is the Institutional Investor country credit rating. The credit rating did not change noticeably from 1986 to 1992 for most markets. However, the credit rating improved considerably for Chile (from 25.1 to 45.9) and for Mex- ico (from 30.8 to 42.6). By far, the credit rating for Taiwan (China) (77.5 in 1992) is the highest for the emerging markets. The Economist regularly ranks the industrial countries according to three macroeconomic indicators: inflation, real gross domestic product (GDP) growth, and current account balance as a percentage of GDP. I computed this ranking for the emerging markets in the sample (not reported). The top performers are Korea and Malaysia. The United Kingdom's macroeconomic performance is only average in relation to that of the set of emerging markets; it ranks 10.5 100 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I among the sample markets. Because a current account deficit does not neces- sarily signal instability but could be the healthy mirror image of large capital inflows, I also computed a ranking based only on GDP growth and inflation. Korea, Malaysia, and Thailand are the best performers. The variability, rather than the level of inflation, might be a better indicator of the soundness of economic policies. Inflation is even less variable in Malaysia and Thailand than in the United Kingdom. I also computed a ranking of the economies based on their performance with respect to both level and variability of inflation. Not surprisingly, Latin American economies perform worst whereas Asian econ- omies perform best. Currency movements can have a dramatic impact on equity returns for foreign investors. Many developing economies manage to keep exchange rate volatility lower than that in the industrial economies. This is not surprising as many devel- oping economies try to peg their exchange rates to the U.S. dollar or to a basket of currencies (see table 8). Dramatic exceptions are Argentina, Brazil, and Nigeria. The second EMSR I investigated is liquidity risk. Because liquidity might be cor- related with the size of the stock market, I also investigated the relative size of the market using locally compiled (not the IFC'S) indexes. Most of the emerging mar- kets are relatively small compared with the major industrial markets.5 Mexico's market, the largest emerging market, has a market capitalization value of $127.1 billion, about one-third the size of Germany's (market-capitalization value $393.5 billion). Zimbabwe's market, the smallest (market-capitalization value $869.2 million), is about 150 times smaller than Mexico's. Recently, some mar- kets have grown tremendously. Between 1989 and 1992 the stock markets of Chile, Colombia, Greece, India, Malaysia, Mexico, Pakistan, Thailand, Turkey, and Venezuela all more than doubled in size in dollar terms. In percentage of GDP, the markets of Chile and Malaysia, with market capitalization values of 82 per- cent of GDP in 1991 and 81 percent of GDP in 1989, respectively, have surpassed the level of the U.S. market. The size of other markets, such as those of Colombia, Nigeria, Turkey, and Venezuela, having market capitalization values of between 3 and 5 percent of GDP, is still tiny compared with the size of their economies. I calculated a turnover measure (value traded as a percentage of market cap- italization; not reported), which could serve as a liquidity indicator. Sur- prisingly, turnover is larger in many emerging markets than in the United King- dom and Japan. Markets with particularly large turnover are those of Korea, Taiwan (China), and Thailand. Table 9 reports rank correlations between the return characteristics and the EMSR measures. Macroeconomic performance, inflation volatility, size, and rela- tive size all have an impact on the extent of market integration. Political risk is positively associated with market segmentation, but the correlation is not signif- 5. Data on liquidity and size of markets in this and the next paragraph are from the Emerging Markets Data Base for the emerging markets and from Allen and O'Connor (1992) for the industrial markets. GDP data are from IMF (various issues). Bekaert 101 icantly different from zero. Not surprisingly, these variables also correlate signif- icantly with the Os and, except for the macroeconomic performance variable, with dividend yields and P/E ratios. The only marginally significant relation between mean returns and EMSRS involves the inflation variable. High and vari- able inflation also contributes to volatility in the stock markets.6 Economies with relatively small stock market capitalization as a percentage of GDP and bad macroeconomic policies tend to have more-volatile stock markets. Exchange rate volatility is not significantly related to return characteristics. The turnover measure is positively correlated with volatility and mean returns, although the correlations are not significantly different from zero. Either the result is caused by some of the Asian markets, where trading is "excessive,' or turnover is a bad proxy for liquidity and does not capture the liquidity problems mentioned by foreign investors in the Chuhan (1992) survey. When the alternative measure of market integration is used, the rank correla- tion between liquidity and market integration is higher (0.390). The other re- sults are robust; in particular, the relations between market integration and macroeconomic performance, inflation volatility, size, and relative size are in- variably stronger. The alternative measure also exhibits a 0.698 rank correlation with the political-risk measure. VI. CONCLUSIONS In this article I have attempted to identify significant relations between a number of barriers to investment, broadly defined, and a return-based measure of market integration, as well as other return characteristics. The policy pre- scription is that an economy should try to eliminate or lessen the impact of those barriers that are most likely to effectively segment the local market from the global capital market. I have identified the following effective barriers to global equity-market integration: poor credit ratings, high and variable inflation, ex- change rate controls, the lack of a high-quality regulatory and accounting frame- work, the lack of sufficient country funds or cross-listed securities, and the limited size of some stock markets. I have not found a significant link between return characteristics and ownership restrictions or a turnover measure. My analysis has some major drawbacks. Foremost, I have simply assumed that my measure of capital-market integration is positively related to capital flows and negatively related to domestic capital costs. As to the former, I could not detect highly significant correlations between my market-integration mea- sures and cumulative capital flows (as a percentage of market capitalization). The correlation between my market-integration measure and the capital-flow data (cumulated real net U.S. purchases of foreign equity studied in Chuhan, Claessens, and Mamingi 1993) is 0.542 with a standard error of 0.258, but it is 6. Although returns are measured in nominal terms, they are measured in dollars, so that high inflation should not necessarily lead to higher stock returns. Table 9. Rank Correlations between Return Characteristics and Emerging-Market-Specific Risks z Level and Exchange volatility z° 0 Political Macroeconomic rate Inflation of Relative Return characteristic risks performanceb volatility volatility inflation Sizec sized Liquidityc ¢ Market integration 0.379 0.462 0.137 0.514 0.519 0.584 0.615 0.012 m Mean returns 0.013 0.228 -0.078 0.238 0.410 -0.037 0.257 -0.258 , Volatility 0.145 0.633 0.327 0.578 0.686 0.042 0.436 -0.263 Sharpe ratio -0.069 -0.073 -0.298 -0.044 0.108 -0.016 0.173 -0.098 Risk loading, a 0.632 0.439 -0.021 0.535 0.504 0.632 0.557 0.199 r Pricingerror, a 0.138 0.215 -0.186 0.278 0.431 0.108 0.377 -0.153 Expected-return correlation 0.384 0.212 -0.090 0.173 0.271 0.516 0.437 0.332 o Dividend yield 0.593 0.295 0.028 0.454 0.486 0.645 0.466 0.441 Price-earnings ratio 0.634 0.272 0.186 0.495 0.511 0.641 0.567 0.317 Note: Economies are ranked from high to low for size, relative size, and liquidity and from low to high for exchange rate volatility, inflation volatility, and level and volatility of inflation. The standard errors for the statistics are 0.22, except for relative size, for which it is 0.23; Argentina and Taiwan (China) are not included. a. The ranking based on the Institutional Investor Credit Ratings for 1992. b. Based on inflation and real GDP growth performance. c. Based on 1992 data. d. Average over 1986, 1989, and 1992. Source: Author's calculations. Bekaert 103 close to zero using the data reported in Tesar and Werner (1995). Note that both data sets involve somewhat different markets and time periods. As for the cost of capital, I have demonstrated the difficulties associated with trying to measure the level of expected equity returns, and hence domestic capital costs. First, the lack of a relation between mean returns and any of the barriers to investment that I have considered does not bode well for approaches based on a history of returns. The market-integration measures correlate signifi- cantly with P/E ratios and dividend yields, which feature in some capital-cost calculations. Second, most efforts to measure expected returns use some version of the CAPM (see, for example, Demirguic-Kunt and Huizinga 1992). However, my results show that high world market ,Bs do not necessarily reflect higher expected returns but rather seem to reflect a higher degree of global capital- market integration. There are a number of possible interpretations for this outcome. The CAPM could be a reasonable description of the returns but should be modified to allow for time-varying degrees of market segmentation. Bekaert and Harvey (1994), for instance, allow conditionally expected returns in emerging markets to de- pend on their covariance with a world benchmark portfolio and on the variance of the country return. The integration measure is a time-varying weight applied. to these two moments, which arises from a conditional regime-switching model. It is also possible that the effect of the world-market factor is confounded by other factors in a multifactor world, where the risk exposures vary through time (Harvey 1993 makes a similar point). Additional factors that come to mind are industry factors and commodity factors. The fact that my integration measure does not correct for different industry exposures and the general lack of diversification within indexes for emerging markets is another potential drawback. Divecha, Drach, and Stefek (1992) consider four "concentration measures": the proportion of capitalization in the top ten companies, an asset concentration factor (which is valued at one if the entire market capitalization is concentrated in one market), a sector concentra- tion measure, and the average correlation between stocks in the index. To exam- ine whether the industry and sectoral patterns affect my market-integration measure, I computed rank correlations between the four concentration measures as reported in Divecha, Drach, and Stefek (1992: 46) and my market-integration measure. If all markets were perfectly integrated and the correlation of expected returns only reflected different concentration or industry effects, one would expect to find positive correlations between the market integration measure and the concentration measures (ranked from low to high). I found the rank correla- tions to be 0.393 for the concentration measures based on the top ten com- panies, 0.399 for asset concentration, 0.448 for sector concentration, and -0.055 for the average correlation between stocks. Because data on Germany are not reported in Divecha, Drach, and Stefek, I used twenty-one countries, and the standard error was 0.224. When the alternative measure was used, the first three correlation coefficients were significantly different from zero. These 104 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 results seem to contradict the conclusion of Divecha, Drach, and Stefek, who state that sector concentration is not important for explaining emerging-market returns. On the contrary, the results suggest that industry factors should be an important part of future analyses. Finally, in this article I have ignored dynamic interactions between changes in barriers to investment and market returns. Future work should explore panel- data approaches that incorporate global and domestic risk factors jointly with quantitative indicators of barriers to investment. In such an analysis, the risk exposures should be made a function of the degree of market segmentation. This article has implications for some other interesting policy questions. Gen- uine concern exists among policymakers about the impact of international in- vestment on local-market turnover and the volatility of equity returns. Tesar and Werner (1995) find no evidence that U.S. investment activity contributes to either volatility in equity returns or to higher local turnover in emerging mar- kets. This result is confirmed in the present article. Section IV has shown that volatility is unrelated to any measure of openness. In fact, volatility is actually negatively, although not significantly, correlated with the market-integration measure. Furthermore there is no association between turnover and the market- integration measures. Policymakers might be concerned that increasing integration between the cap- ital market and the economy will lead to lower diversification benefits. Lower diversification benefits, in turn, might reduce the appetite of the international investment community for stocks in emerging equity markets. Table 3 reports a correlation of the composite index with the United States that is 0.40, which is not unlike correlations noted between industrial countries. The more relevant correlation of expected returns is still only 0.19, which is fairly low. I would argue that these concerns are ill-founded for two reasons. First, as shown in this article, I have not detected any relation between the risk-return tradeoff of individual markets (as measured by the Sharpe ratio) and market integration or the openness measures. Second, capital-market integration might help secure long-lasting portfolio flows from institutional investors. The trend toward inter- national diversification has caused an increasing number of money managers and institutional investors to practice global-asset-allocation strategies. Typ- ically, asset-allocation models start from a neutral benchmark that is close to the world-market portfolio as, for instance, defined by Morgan Stanley Capital International. Emerging markets should eventually strive to become part of the global world-market portfolio, used as a benchmark by investors worldwide. APPENDIX. DATA SOURCES The stock return data for the emerging markets are from the IFc Emerging Markets Data Base. Annualized dividend yields are constructed as the sum of twelve monthly dividend yields. P/E ratios, market capitalizations, and turn- over ratios are also taken from that data set. The stock return data for the Bekaert 1 05 industrial countries are from Morgan Stanley Capital International. The U.S. interest rate used in the article is the one-month Eurorate obtained from Data Resources Incorporated (DRI) until mid-1988, from Citicorp Data Services be- tween mid-1988 and July 1991, and from the Financial Times for the remainder of the sample. Equity returns for Germany, Japan, and the United Kingdom were computed using exchange rate data from Citicorp Data Services which were updated from mid-1991 onward with data from the Financial Times. Macroeconomic data were taken from IMF (various issues). REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Adler, Michael, and Bernard Dumas. 1983. "International Portfolio Choice and Corpo- ration Finance: A Synthesis." Journal of Finance 38(3, June):925-84. Allen, S., and S. O'Connor, eds. 1992. The GT Guide to World Equity Markets. Lon- don, U.K.: Euromoney Publications. Bekaert, Geert. Forthcoming. "The Time-Variation of Expected Returns and Volatility in Foreign Exchange Markets." Journal of Business and Economic Statistics. Bekaert, Geert, and Robert Hodrick. 1992. "Characterizing Predictable Components in Excess Returns on Equity and Foreign Exchange Markets." Journal of Finance 47(2):467-509. Bekaert, Geert, and Campbell R. Harvey. 1994. "Time-varying World Integration.' NBER Working Paper 4843. Cambridge, Mass.: National Bureau of Economic Research. Bekaert, Geert, and Michael Urias. 1994. "Diversification, Integration, and Emerging Market Closed-End Funds." Stanford University, Graduate School of Business, De- partment of Finance, Stanford, Calif. Processed. Buckberg, Elaine. 1995. "Emerging Stock Markets and International Asset Pricing." The World Bank Economic Review 9(1):51-74. Campbell, John, and John Ammer. 1993. "What Moves the Stock and Bond Markets? A Variance Decomposition for Long-Term Asset Returns." Journal of Finance 48(1): 3-38. Campbell, John Y., and Yasushi Hamao. 1992. "Predictable Stock Returns in the United States and Japan: A Study of Long-Term Capital Market Integration." Journal of Finance 47(1, March):43-69. Capital International Perspective, S.A., and Morgan Stanley & Co. Various issues. Morgan Stanley Capital International Perspective. New York: Morgan Stanley. Chuhan, Punam. 1992. "Sources of Portfolio Investment in Emerging Markets." Work- ing Paper, World Bank, International Economics Department, Washington, D.C. Processed. Chuhan, Punam, Stijn Claessens, and Nlandu Mamingi. 1993. "Equity and Bond Flows to Latin America and Asia: The Role of Global and Country Factors.' World Bank, International Economics Department, Washington, D.C. Processed. Cooper, Ian, and Evi Kaplanis. 1986. "Costs to Crossborder Investment and Interna- tional Equity Market Equilibrium?' In Jeremy Edwards, Julian Franks, Colin Mayer, 106 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 and Stephen Schacter, Recent Developments in Corporate Finance. Cambridge, U.K.: Cambridge University Press. . 1994. "Home Bias in Equity Portfolios, Inflation Hedging, and International Capital Market Equilibrium." Review of Financial Studies 7(1, Spring):45-60. Cumby, Robert E., and John Huizinga. 1993. "Investigating the Correlation of Unob- served Expectations: Expected Returns in Equity and Foreign Exchange Markets and Other Examples." Journal of Monetary Economics 30(2, November):217-53. . 1992. "Testing the Autocorrelation Structure of Disturbances in Ordinary Least Squares and Instrumental Variables Regressions." Econometrica 60(1, January): 185- 95. Demirgui-Kunt, Asli, and Harry Huizinga. 1992. "Barriers to Portfolio Investments in Emerging Stock Markets." Policy Research Working Paper wps 984. World Bank, Washington, D.C. Processed. Divecha, Arjun B., Jaime Drach, and Dan Stefek. 1992. "Emerging Markets: A Quan- titative Perspective." Journal of Portfolio Management 19(1, Fall):41-56. Diwan, Ishac, Lemma Senbet, and Vlhang Errunza. 1992. "Pricing of Country Funds and Their Role in Capital Mobilization for Emerging Economies." University of Maryland, Center for International Business Education and Research, College Park, Md. Processed. Errunza, Vihang, and Etienne Losq. 1985. "International Asset Pricing under Mild Segmentation: Theory and Test." Journal of Finance 40(1, March):105-24. 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(4, September): 897-914. Ferson, Wayne E., and Campbell R. Harvey. 1993. "The Risk and Predictability of International Equity Returns." Review of Financial Studies. 6(3):527-66. Harvey, Campbell R. 1991. "The World Price of Covariance Risk." Journal of Finance 46(1, March):111-57. 1. 1993. "Predictable Risk and Returns in Emerging Markets." Duke University, Fuqua School of Business, Raleigh, N.C. Processed. . 1995. "The Risk Exposure of Emerging Equity Markets." The World Bank Economic Review 9(1):19-50. Heston, Steve, and K. Geert Rouwenhorst. 1994. "Does Industrial Structure Explain the Benefits of International Diversification?" Journal of Financial Economics 36(1):3-27. Hietala, Pekka. 1989. "Asset Pricing in Partially Segmented Markets: Evidence from the Finnish Market." Journal of Finance 44(3, July):697-718. Hodrick, Robert, and Sanjay Srivastava. 1984. "An Investigation of Risk and Return in Forward Foreign Exchange." Journal of International Money and Finance 3:5-29. IFC (International Finance Corporation). Various issues. IFC Index Methodology. Wash- ington, D.C. IMF (International Monetary Fund). 1992. 1992 Annual Report. Washington, D.C. . Various issues. International Financial Statistics. Washington, D.C. Park, Keith K. H., and Antoine W. van Agtmael, eds. 1993. The World's Emerging Stock Markets: Structure, Developments, Regulations, and Opportunities. Chicago, Ill.: Probus Publications. Roll, Richard. 1992. "Industrial Structure and the Comparative Behavior of Interna- tional Stock Market Indices." Journal of Finance 47(1, March):3-41. Bekaert 107 Speidell, Lawrence S., and Ross Sappenfield. 1992. "Global Diversification in a Shrink- ing World." Journal of Portfolio Management 19(1, Fall):57-67. Stulz, Rene M. 1981. "On the Effects of Barriers to International Investment."Journal of Finance 36(4, September):923-34. Stulz, Rene, and Walter Wasserfallen. 1992. "Foreign Equity Investment Restrictions and Shareholder Wealth Maximization: Theory and Evidence." Ohio State University, De- partment of Finance, Columbus, Ohio. Processed. Tesar, Linda, and Ingrid Werner. 1995. "U.S. Portfolio Investment and Emerging Stock Markets." The World Bank Economic Review 9(1):109-30. Wheatley, Simon. 1988. "Some Tests of International Equity Integration." Journal of Financial Economics 21(2, September) :177-212. Wilcox, Jarrod W. 1992. "Taming Frontier Markets." Journal of Portfolio Management 19(1, Fall):51-56. THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1: 109-129 U.S. Equity Investment in Emerging Stock Markets Linda L. Tesar and Ingrid M. Werner This article examines U.S. equityflows to emerging stock markets from 1978 to 1991 and draws three main conclusions. First, despite the recent increase in U.S. equity investment in emerging stock markets, the U.S. portfolio remains strongly biased to- ward domestic equities. Second, of the fraction of the U.S. portfolio that is allocated to foreign equity investment, the share invested in emerging stock markets is roughly proportional to the share of the emerging stock markets in the global market capitaliza- tion value. Third, the volatility of U.S. transactions in emerging-market equities is higher than in other foreign equities. The normalized volatility of U.S. transactions appears to befalling over time, however, and we find no relation between the volume of U.S. transactions in foreign equity and local turnover rates or volatility of stock returns. In the past several years the opportunities for equity investment in developing' economies have increased remarkably. The expansion of local equity markets. and the development of instruments for issuing equity directly on international. markets have given firms in these economies increased access to the world supply of capital. Indeed, developing economies as a group increased their bor- rowing through new equity issues from virtually zero in 1987 to $5 billion in 1991.1 This borrowing accounted for more than 10 percent of the total capital raised by these economies in 1991 (Gooptu 1993). From the perspective of international investors, these rapidly growing markets offer potentially high rates of return and an important means of diversifying portfolio risk (Divecha, Drach, and Stefak 1992; Harvey 1993, 1995; Wilcox 1992). As the share of foreign investment in emerging markets has risen, however, policymakers have become increasingly concerned about the factors determining international in- vestment, the permanence of foreign capital investments, and the impact of foreign investment on local turnover and on the volatility of stock prices. 1. A billion is 1,000 million. Linda L. Tesar is with the Department of Economics at the University of California at Santa Barbara and the National Bureau of Economic Research. Ingrid M. Werner is with the Graduate School cf Business at Stanford University, the Institute for International Economic Studies in Stockholm, and the National Bureau of Economic Research. This article was commissioned by the Debt and International Finance Division of the World Bank for its Conference on Portfolio Investment in Developing Countries, Washington, D.C., September 9-10, 1993. The authors thank Geert Bekaert and Stiin Claessens for their helpful comments, and Tasos Mastroyiannis, Jon Riddle, Patrick Rowland, and Michael Urias for their excellent research assistance. © 1995 The International Bank for Reconstruction and Development / THE WORLD BANK 109 110 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. ] To better understand the behavior of foreign investors in emerging stock markets (ESMS), we examine U.S. equity investment in nineteen economies: the industrial markets of Canada, Germany, Japan, and the United Kingdom, and fifteen emerging markets in Asia, Latin America, and Europe. This narrows the focus of the study to the portfolio allocation decisions of U.S. investors; however, U.S. investment currently accounts for a substantial fraction of port- folio flows to developing economies (see Chuhan 1992). The richness of data on U.S. equity transactions allows us to examine portfolio investment behavior at a level not possible (at least to our knowledge) with other data sources. In particular, the availability of data on U.S. purchases and sales of equity on a bilateral basis makes it possible to identify market-specific factors that affect U.S. investment. We focus on purchases and sales in foreign equity; in future work we plan to extend our analysis to investment in foreign corporate bonds. We have chosen to include industrial markets in the sample in order to place portfolio investment in emerging markets in the context of changes in global portfolio investment behavior. In general, we find that the investment patterns observed in the emerging markets seem to be mirrored in U.S. investment patterns in the industrial markets. The principal difference is that the increased investment activity in emerging markets began slightly later than in the indus- trial markets. Our findings suggest that the recent increase in equity investment flows in emerging markets should be viewed as part of a global trend toward increased cross-border investment and not as a separate phenomenon. Section I describes the data sources. Section II discusses the changes in the U.S. investment portfolio since 1976. Section III provides some descriptive sta- tistics on U.S. net purchases of foreign equity. Section IV turns to gross pur- chases of equity and the relation between U.S. equity investment and local market turnover as well as stock price volatility. Section V draws some general conclusions that may be of relevance for policymakers concerned about the rise of investment in ESMS. 1. DATA SOURCES The data on equity flows were collected from U.S. Department of the Trea- sury (various issues). By law, all banks, brokers, dealers, and individuals are required to report the value of any security transaction involving a foreign resident. The data record the value of U.S. purchases and sales of domestic and foreign securities (bonds and equity) on a bilateral basis in sixty-four economies. See Tesar and Werner (forthcoming) and (1994) for more details about these data. The international investment position data used in section II are reported by the U.S. Department of Commerce. These series are based on a cumulation of the transactions series and are adjusted by the Department of Commerce for changes in equity prices and exchange rates. There are some shortcomings with the data. First, although investors are required to report their transactions with foreigners, they incur no penalty for Tesar and Werner 111 failing to do so. There are obvious reasons for underreporting equity transac- tions when capital controls are present. Second, the rapid expansion of markets, the emergence of new markets, and the development of new types of financial instruments make it difficult for the reporting agencies to keep pace with the volume of flows. (See Steckler and Truman 1992 for a complete description of the problems involved in collecting data on portfolio flows.) Finally, estimates of stock positions are plagued by the usual problems of incorporating capital gains on the portfolio and changes in exchange rates. Despite these problems, the data remain the sole public source of information on bilateral equity flows between the United States and the rest of the world. Data on equity returns and market capitalization values for emerging markets are drawn from the International Finance Corporation (IFC) Emerging Markets Data Base (EMDB). Equity returns for the industrial economies and market cap- italization values are from Capital International Perspective, S.A. and Morgan Stanley & Co. (various issues), hereafter referred to as MSCI. U.S. Treasury bill rates are from the Center for Research in Securities Prices. II. THE INTERNATIONAL PORTFOLIO OF U.S. INVESTORS The increasing role of emerging markets in the global equity market can be seen in table 1. The first two columns show each economy's equity market capitalization values as a fraction of global market capitalization in the first quarter of 1986 and again in the first quarter of 1991. The third column shows the percentage change in the market shares during the 1986-91 period. The share of the United States in the world market has fallen from roughly 45 to 35 percent, whereas Japan's share has risen from 26 to 32 percent. Taken as a group, the share of the five industrial economies has decreased slightly, from 87 percent of the global total in 1986 to 83 percent in 1991. Within the set of emerging markets, there appears to be some heterogeneity in the growth of market shares across regions. Market capitalization shares have generally increased in Asia and Europe. The Republic of Korea and Taiwan (China) each accounted for about 1 percent of the global total in 1991. The market shares of Greece and Portugal have also increased dramatically, although their absolute market sizes remain small. In Latin America, the increases in relative market sizes are less pronounced, and there has been an actual decline iin market share in Brazil, from 1.6 to 0.3 percent, during this period. One simple strategy for achieving an internationally diversified portfolio is to invest in foreign equities in proportion to each economy's share of the world market. This strategy abstracts from the effect of currency denomination on portfolio choice. In 1991 such a strategy would have required investors to hold roughly one-third of their portfolio in U.S. equities, one-third in Japanese equi- ties, about 10 percent in British equities, 4 percent in German equities, 3 percent 112 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Table 1. Equity Market Shares, 1986-91 (percent) Percentage Market 1986.1 1991.1 change Industrial markets Canada 3.13 2.53 -19.22 Germany 4.15 3.52 -15.21 Japan 26.02 32.36 24.38 United Kingdom 8.13 9.86 21.26 United States 45.26 35.04 -22.57 Total 86.68 83.31 -3.89 Emerging stock markets Asia India - 0.42 Indonesia - 0.09 Korea, Rep. of 0.19 1.15 507.53 Malaysia 0.27 0.64 141.57 Philippines 0.01 0.10 553.44 Taiwan (China) 0.26 1.35 430.84 Thailand 0.04 0.38 937.78 Total 0.76 4.13 443.17 Europe Greece 0.02 0.19 1020.23 Portugal 0.02 0.11 505.59 Turkey - 0.21 Total 0.03 0.50 1360.42 Latin America Argentina 0.04 0.06 59.71 Brazil 1.59 0.27 -82.80 Chile 0.05 0.21 304.16 Colombia 0.01 0.02 47.77 Mexico 0.08 0.43 415.36 Total 1.78 0.99 -44.43 Percentage of world market capitalization covered by these countries 87.48 87.94 - Not available. Note: Values are for each economy's equity market capitalization as a share of the global equity market capitalization value, based on data for the first quarters of 1986 and 1991. Source: For industrial economies and world total, MscI; for emerging stock markets, the IFC EMDB. in Canadian equities, and the balance distributed across other European and emerging markets.2 Figure 1 shows the equity portfolio chosen by U.S. residents. For each econ- omy or region for which data are available, investment positions are expressed as a fraction of the value of U.S. equity market capitalization.3 The U.S. 2. If investors are identical, this is the equilibrium portfolio held in all economies. Tesar and Werner (forthcoming) demonstrate that for investors in the five industrial markets in our sample, such a portfolio generally dominates a portfolio of domestic securities in terms of the tradeoff between risk and return. Harvey (1993, 1995) finds further gains from international diversification by combining a portfolio of eighteen industrial markets (those covered by the MSCI index) with emerging-market securities. 3. For the purposes of this paper, "region" denotes both a geographical region and a group of markets. Tesar and Werner 113 Figure 1. 7he International Equity Portfolio of US. Residenzts, 1976-90 (percentage of U.S. equity market capitalization value) Percent 2.4 - 2.2 - 2.0 - 1.8 - 1.6 - 1.4 - 1.2- 8. - > z 0.8 0.4 - . 0.2 - 0 - -- -- - - - - - 1976 78 80 82 84 86 88 90 - Canada -- International organizations Latin America .. . Other markets Westem Europe Source: U.S. Department of the Treasury (various issues). investment share in Western European equities increased from 0.3 percent in 1976 to about 2.2 percent in 1990. The share invested in Canada remained fairly constant, at less than 1 percent, and there has been a very slight increase in invest- ment in other markets. Data on the U.S. international investment position in Ja- pan are not reported by the Department of Commerce. The remaining 96 percent is invested in domestic equities. Note that if U.S. investors were to hold the mar- ket portfolio, 6 to 7 percent of their portfolio would be allocated to the emerging markets and the smaller European markets listed in table 1. This figure exceed.s the current total of U.S. equity investment in all markets. This strong bias toward domestic securities, despite the apparent gains from international diversification, has been documented for other industrial markets as well (French and Poterba 1991; Tesar and Werner, forthcoming). The reallocation of international equity investments across regions over time is illustrated in figure 2. Here the investment positions are plotted as shares of the total international portfolio held by U.S. residents. The share of the interna- tional portfolio allocated to Western Europe has risen from 30 percent in 1976 to nearly 70 percent in 1989. This was accomplished by pulling investment out of Canadian equities. There has been very little reallocation involving equity from the market groups "Latin America" or "Other markets." Thus, although 114 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Figure 2. The Share of Total Intemnational Equity Investment Helcd by US. Residents, 1976-89 Percent 70 - 60 - 50- 30- 20- 10 1976 78 80 82 84 86 89 - Canada --- International organizations Latin America .. - .. Other miarkets - Western Europe Source: U.S. Department of the Treasury (various issues). the increased investment in emerging-market equities appears large from the perspective of the recipient economies, it remains a small fraction of the U.S. portfolio and is in fact much smaller than the investment position suggested by a portfolio that is weighted by market capitalization. The small U.S. investment position in emerging markets is consistent with the figures reported by pension fund managers in the United States and Europe. Fund Research, a group- studying institutional investors, reports that investment in emerging markets "probably represents only around 0.5 percent of institutional investors' portfolio holdings" (Financial Times, October 26, 1992). Chuhan (1992) finds that investment in emerging markets was roughly 2 to 3.5 percent of the international portfolio held by U.S. pension funds from 1988 to 1991. III. U.S. NET PURCHASES OF FOREIGN EQUITY Data on international investment positions allow us to examine portfolio allocations directly. Unfortunately, these data are limited to a small set of econ- omies or regions and are rarely available on a bilateral basis. We therefore turn to the data on U.S. transactions in foreign equity. Unlike the data on investment positions, these figures do not reveal the overall portfolio allocation; however, Tesar and Werner 115 Figure 3. Quarterly Net Purchases of Equity in Foreign Industrial Economies by U.S. Residents, 1978-91 Millions of dollars 6,ooo - 5,000- 4,000- 3,000- 2,000 -'l 0 -~~~~~~~~~~~~~~ -1,000 . i - 2,000 - 1978 79 80 81 82 83 84 85 86 87 88 89 90 91 - Canada Germany - Japan - -- United Kingdom Source: U.S. Department of the Treasury (various issues). the transactions do reflect the marginal adjustments in the portfolios of those individuals and institutions actively participating in these markets.4 In this sec- tion we examine patterns in U.S. net portfolio flows to four industrial and fifteen developing economies. U.S. investment flows cover a substantial fraction of total foreign equity investment in developing economies. In 1990 U.S. net equity flows to develop- ing economies reached $1.4 billion, compared with net flows of $200 million from Canada and $60 million from Germany (see Chuhan 1992, table 2). In 1991 net flows from these three economies to developing economies approxi- mately doubled. In terms of gross flows, U.S. equity investment was about $10 billion in 1991, compared with gross equity flows of $310 million from Ger- many. Data on gross equity flows from other industrial economies are unfor- tunately not available. Figures 3 through 6 show the quarterly net purchases by U.S. residents of foreign equity in various markets from the first quarter of 1978 to the third quarter of 1991. All flows are in millions of dollars. Net purchases are defined as U.S. purchases of foreign equity from foreign residents less U.S. sales of 4. Tesar and Werner (1994) examine the cross-border portfolio flows between the United States arid Canada, Germany, Japan, and the United Kingdom, exploiting both U.S. and foreign data sources. Despite the data collection problems discussed above, we find a high correlation between U.S. net purchases of Canadian and German equities across the official statistics of the three economies. 116 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Figure 4. Quarterly Net Purchases of Equity in Emerging Stock Markets in Latin America by U.S. Residents, 1978-91 Millions of dollars 2,000 - 1,800 - 1,600 - 1,400 - 1,200 - 1,000 i.1 800 -i 600 - 200 -.--'- ;- -200 - -400 -6oo~~~~~~~~~~~~~~~~~~~~~~~~~~~i' 1978 79 80 81 82 83 84 85 86 87 88 89 90 91 Argentina --- Brazil c Chile Mexico Source. U.S. Department of the Treasury (various issues). foreign equity to foreign residents during the quarter. From figure 3 it is apparent that U.S. transactions in foreign equity began to pick up in the mid-1980s. Inter- estingly, a relatively large net sale of roughly $2.5 billion of foreign equity to Japa- nese residents occurred in 1987. By the early 1990s this flow had reversed to a net purchase of equity from Japanese residents of $5 billion. There was also a large net purchase of equity from British residents in 1989 of approximately $3 billion. This volatility in net purchases is a characteristic of equity flows to both the indus- trial and emerging-market economies. It is evident in figures 3 through 6 that this volatility has increased in both sets of economies over the sample period. Figure 4 shows that net purchases of foreign equities from Latin America are on a smaller scale than those from the industrial economies. In Latin American economies, the increased investment activity was delayed until the late 1980s. The variance of these flows is especially large for Mexico. U.S. residents pur- chased on net approximately $1.5 billion worth of Mexican equities in the second quarter of 1991. This large net outflow from the United States to Mexico coincided with the American Depository Receipt (ADR) issue of Telmex securities in the United States in May 1991. Issues of ADRS, Global Depository Receipts (GDRS), and securities issued in the United States under Rule 144A tend to produce lumpiness in portfolio flows between the United States and the ESMS Tesar and Werner 117 Figure 5. Quarterly Net Purchases of Equity in Emerging Stock Markets in Asia by U.S. Residents, 1978-91 A. India, Indonesia, Republic of Korea, and Malaysia Millions of dollars 80 - 70 - 60 - 50 - 40 - 30 - I' 20 - 10 - 60- 50 - -30 -40 -50 , I III, I 1978 79 80 81 82 83 84 85 86 87 88 89 90 91 India - Indonesia --. Korea -- - Malaysia B. Pakistan, the Philippines, Taiwan (China), and Thailand Millions of dollars 80 - __________________________________ 70- 60- 50- 40- 30- 20- 10 1 0-5 1978 79 80 81 82 83 84 85 86 87 88 89 90 91 Paitn-Philippines - - -Taiwan (China). Thailand Source: U.S. Departmnent of the Treasury (various issues). 118 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Figure 6. Quarterly Net Purchases of Equity in Emerging Stock Mfarkets in Europe by US. Residents, 1978-91 Millions of dollars 50 - 40 i 30- 20- 10- 0 -10 -20 - IIli,II' I iiII 1978 79 80 81 82 83 84 85 86 87 88 89 90 91 Greece ..- Portugal __ Turkev Source: U.S. Department of the Treasury (v-arious issues). because these securities are first issued in large blocks to the sponsor and are then reissued on the U.S. market (see Gooptu 1993 for a more complete discussion). Figure 5 shows net purchases of equity from ESMS in Asia. The flows are on a yet smaller scale-roughly one-tenth of the scale of figure 3. Excluding the net purchase of Filipino equity in late 1979, the increase in U.S. purchases of Asian equity began at about the same time as that of equities in the industrial econ- omies. The flows to the emerging markets in Europe plotted in figure 6 are of the same order of magnitude as those in Asia, with a notable spike in net purchases of Turkish and Portuguese equity in the third quarter of 1989. Table 2 provides descriptive statistics of net purchases of foreign equity during the same time period. All data are now deflated by the U.S. consumer price index. The United Kingdom tends to be the largest recipient of U.S. equity investment, with a mean level of $293 million. Investment in Canada follows with a mean of $128 million, and in Japan the mean is $84 million. As seen in figures 4-6, mean flows to emerging markets are on a smaller scale, the largest recipients being Mexico, with a mean of $53 million; Brazil, $12 million; Tesarand Werner 119 Malaysia, $3 million; and Portugal, $2 million. Three economies-Taiwan (China), Argentina, and Colombia-experienced mean outflows of equity invest- ment in this period. The coefficient of variation, defined as the standard deviation divided by the mean, gives some indication of the volatility of net purchases in re- lation to the mean level of equity flows. With the exception of the very high coeffi- cient for Argentina, the variability of net purchases in the emerging markets is of the same order of magnitude as that in the industrial markets in our sample. The theory of portfolio choice has no direct implications for net purchases of foreign equities. Given the roughly random-walk behavior of equity returns, however, it follows that changes in portfolio allocations due to changes in ex- pected returns would exhibit little serial correlation. If, however, U.S. investors are moving gradually toward an internationally diversified portfolio, or if they re- spond slowly to changes in financial markets, net purchases may exhibit some persistence. Columns 4 through 7 of table 2 indicate that net purchases of equity from both sets of markets tend to exhibit high first-order autocorrelation. There are some notable exceptions, however. Argentina, Brazil, and Colombia have p-values of 0.85 or higher, indicating a high probability that the autocorrelation coefficients are equal to zero, and the first-order autocorrelation in India is signifi- cantly negative. To get a measure of the evolution of the U.S. portfolio over time, table 3 shows the cumulated value of net purchases of foreign equity for each of the markets in our sample. Note that these figures do not correspond to a true measure of the stock value of the equity held in each market, because we do not take into account capital gains or changes in exchange rates. However, these figures provide a rough estimate of the U.S. investment position. The first column of table 3 shows the value accumulated over the entire period, from the first quarter of 1978 to the third quarter of 1991. The second column shows the value accumulated since 1986, when activity on these markets began to increase. The final column shows the percentage of the equity position accumulated in the latter part of the sample. For almost all markets, we see a positive accumulation of foreign equity since 1978 (column 1). The exceptions are the emerging markets of Taiwan (China), Argentina, and Colombia. These data suggest that U.S. investors are, albeit gradually, moving from having zero foreign assets to having positive claims. Most of this accumulation has taken place at the end of the period, particularly in Indonesia, Chile, and the European emerging markets. Note that equity from ESMS is roughly 12 percent of the total amount of foreign equity accumulated by U.S. residents since 1978. The figures in table 1 suggest that if U.S. residents were to allocate their portfolio of foreign equities in proportion to market size, about 6 percent of their portfolio would be in the form of ESM equities. Thus, the recent increase in investment in the emerging markets appears to be slightly larger than the expansion of their market size. The correlations between net purchases of equity in different markets provide some additional information about the factors affecting U.S. portfolio choice. If Table 2. Descriptive Statistics of U.S. Net Purchases of Foreign Equity by Market, 1978-91 (millions of 1985 dollars) Standard Coefficient Autocorrelations Market Mean deviation of variationa AR(1) AR(2) AR(3) AR(4) p-value(4)b Industrial markets Canada 127.79 283.05 2.22 0.54' 0.20 -0.06 -0.15 0.00 Germany 46.23 195.18 4.22 0.39* -0.04 -0.06 0.14 0.04 Japan 84.05 1023.43 12.18 0.32* 0.32' 0.25 0.04 0.00 United Kingdom 292.75 640.85 2.19 0.27 -0.09 0.00 0.13 0.24 Emerging stock markets Asia India 0.05 0.56 10.58 -0.42' 0.05 0.00 -0.01 0.03 Indonesia 1.76 6.51 3.70 0.42* 0.07 0.21 0.47" 0.00 Korea, Rep. of 0.73 8.14 11.15 0.21 0.01 -0.21 -0.09 0.21 Malaysia 3.05 9.75 3.19 0.56* 0.09 -0.11 -0.05 0.00 Philippines 1.06 7.92 7.45 0.10 0.28* 0.00 0.01 0.28 Taiwan(China) -0.65 6.65 -10.20 -0.05 0.35 0.02 -0.04 0.12 Thailand 3.26 7.28 2.23 0.45' 0.11 0.02 0.10 0.01 Europe Greece 0.44 2.80 6.42 0.19 0.19 0.22 0.13 0.09 Portugal 1.97 5.19 2.64 0.48' 0.27 0.33'- 0.22 0.00 Turkey 0.93 6.12 6.58 -0.16 0.23 0.11 0.00 0.24 Iatin America Argentina -0.43 7.07 -16.33 0.09 0.00 0.01 -0.01 0.98 Brazil 11.75 44.61 3.80 0.11 -0.06 0.05 0.07 0.85 Chile 1.32 10.59 8.03 0.34* 0.28* 0.23 -0.25 0.00 Colombia -0.25 1.26 ---5.06 -0.11 -0.02 0.04 -0.06 0.90 Mexico 52.65 190.76 3.62 0.38* 0.31* 0.13 0.15 0.00 Significant at the 5 percent level. Note: Net purchases are U.S. purchases of foreign equity from foreign residents lcss U.S. sales of foreign equity to foreign residents. Values are based on quarterly data from the first quarter of 1978 to the third quarter of 1 991, deflated by the U.S. consumer price index (1985.3 = 100). a. The ratio of the standard deviation divided by the mean. b. The probability that all four autocorrelation coefficients are zero. Source: For net purchases, U.S. Department of the Treasury (various issues); for the U.S. consunier price index, IMF (various issues). Tesar and Werner 121 Table 3. Cumulated Real U. S. Net Purchases of Foreign Equity by Market, 1978-91 (millions of 1985 dollars) Percentage of the equity position Cumulated value of net purchases accumulated during Market 1978.1-1991.3 1986.1-1991.3 1986.1-1991.3 Industrial markets Canada 7,028.3 3,833.2 54.54 Germany 2,542.4 2,137.8 84.09 Japan 4,622.7 2,616.7 56.61 United Kingdom 16,101.4 13,680.2 84.96 Total 30,294.8 22,267.9 73.50 Emerging stock markets Asia India 3.0 1.8 60.00 Indonesia 96.6 97.6 101.04 Korea, Rep. of 40.1 -20.8 -151.87 Malaysia 167.9 164.6 98.03 Philippines 58.4 -15.0 -125.68 Taiwan (China) -35.8 -35.2 98.32 Thailand 179.3 177.3 98.88 Total 509.5 370.3 72.68 Europe Greece 24.0 39.9 166.25 Portugal 108.3 108.3 100.00 Turkey 51.2 61.3 119.73 Total 183.5 209.5 114.17 Latin America Argentina -23.8 -3.6 15.13 Brazil 646.4 639.9 98.99 Chile 72.5 79.0 108.97 Colombia -13.7 -7.4 54.01 Mexico 2,895.9 2,670.8 92.23 Total 3,577.3 3,378.7 94.45 Note: Values are based on quarterly data deflated by the U.S. consumer price index (1985.3 = 100). Source: For net purchases, U.S. Department of the Treasury (various issues); for the U.S. consumer price index, IMF (various issues). U.S. investors turn to foreign equity investments when the returns to domestic investments are low, or if U.S. investors are moving generally toward a more broadly diversified portfolio, we would expect to see a positive correlation between net purchases across markets. It has been suggested, for example, that low U.S. interest rates have prompted the recent increase in U.S. foreign invest- ment in emerging markets. A tendency to reallocate the portfolio between equity investments in two markets, perhaps in response to changes in expected returns, would be reflected in a negative correlation between net equity purchases. The amount that can be learned from these simple correlations is limited, of course, by the fact that investors are likely to care about both the return on the invest- ment and the minimization of risk for the overall portfolio. 122 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Table 4. Correlation Coefficients between U.S. Net Equity Purchases, 1978-91 Ger- United Indo- Rep. of Malay- Philip- Taiwan Thai- Market many Japan Kingdom India nesia Korea sia pines (China) land Canada 0.00 -0.19 0.19 -0 16 -0.16 0.19 -0.17 0.09 0.02 -0.06 Germany 0.35* 0.04 0.12 0.18 0.08 0.07 0.03 0.25 0.18 Japan 0.23 0.11 0.43' -0.11 -0.11 0.35- 0.38* 0.07 United Kingdom 0.01 0.48' 0.07 -(0.23 -0.02 0.25 0.22 India -0.01 -0.23 -0.08 0.08 0.25 0.(0 Indonesia -0.06 0.05 0.09 0.22 0.35' Korea, Rep. of -0.16 -0.17 0.17 -0.50* Malaysia -0.04 -0.12 0.04 Philippines 0.02 -0.06 Taiwan (China) 0.10 Thailand Argentina Brazil Chile Colombia Mexico Greece Portugal Turkey Significant at the 5 percent level. Note: The correlations are based on quarterly data from the first quarter of 1978 to the third quarter of 1991. Net purchases are in millions of dollars, deflated by the U.S. consumer price index (1985.3 = The figures shown in table 4 indicate that net equity purchases are generally not significantly correlated across markets. Of the 171 correlation coefficients shown, only 36 are significant at the 5 percent level. Of these, 29 are positive, providing weak evidence that U.S. investors tend to seek more than one foreign channel for investment at the same time. The boxed sections of the table high- light the correlations within the four regions. A positive correlation across mar- kets within a region would be observed if U.S. investors were to make invest- ment decisions based on region-specific as opposed to market-specific factors. There is no evidence of such a region-specific determinant for investment in the industrial markets or the emerging markets of Asia and Latin America. There is, however, a significant positive correlation between U.S. investment in Greece and in Portugal and between U.S. investment in Portugal and in Turkey. There is also a significantly positive correlation between U.S. equity investment in Japan and Indonesia, the Philippines, and Taiwan (China). IV. U.S. EQUITY TRANSACTIONS IN FOREIGN MARKETS We define transactions as the sum of U.S. purchases of foreign equity plus U.S. sales. Table 5 shows the means, standard deviations, and coefficients of variation of U.S. transactions in two periods: from the first quarter of 1978 to the fourth quarter of 1985 and from the first quarter of 1986 to the third quarter Tesar and Werner 123 Argentina Brazil Chile Colombia Mexico Greece Portugal Turkey -0.03 0.12 -0.01 0.06 -0.13 -0.13 0.18 0.07 -0.17 0.35* 0.11 -0.12 0.28 0.33* 0.35* 0.20 0.18 0.33k -0.22 -0.13 0.43* 0.15 0.24 0.30* 0.28 0.20 -0.20 -0.09 0.46* 0.23 0.16 0.18 0.04 -0.02 -0.05 -0.13 -0.03 -0.09 -0.05 0.02 0.21 0.17 0.02 -0.10 0.81 0.54* 0.24 0.24 -0.07 -0.13 -0.05 0.13 -0.03 0.03 -0.02 -0.03 -0.30* -0.18 0.47* 0.10 -0.13 0.10 0.09 0.05 0.01 0.04 0.00 0.05 0.06 -0.02 0.04 0.02 -0.03 0.16 0.10 -0.02 -0.18 0.10 0.17 0.28 0.04 0.493 -0.05 -0.1.5 0.38* 0.37* 0.36' 0.32* -0.33 -0.53 i -0.16 0.32' -0.01 -0.38" -0.46" 0.26 -0.08 0.26 0.31 0.85* 0.83* 0.14 -0.29* 0.04 0.54* 0.40* -0.16 -0.01 -0.01 -0.02 0.62" 0.18 0.14 0.31 0.21 0.76 100). The boxed sections of the table highlight the correlations within the four regions: industrial markets and emerging stock markets in Asia, Europe, and Latin America. Source: For purchases, U.S. Department of the Treasury (various issues); for the U.S. consumer price index, IMF (various issues). of 1991. The means and standard deviations of the volume of U.S. transactions are substantially higher in the industrial markets in both periods. However, the coefficient of variation, which normalizes the volatility of the transactions rate by the mean level of transactions, is higher in emerging markets than in indus- trial markets in every case and across the two time periods. Thus it appears that the transaction rate of U.S. investors in equity from emerging markets is some- what higher than the U.S. transaction rate in equity from more established markets. A second conclusion that can be drawn from table 5 is that the volatility of U.S. transactions in relation to the mean appears to be falling over time. The coefficient of variation is smaller in the second half of the sample for all markets but one. The single exception is Mexico, whose coefficient of variation doubled over the time period. The general decline in the variability of U.S. transactions may be exaggerated for markets with very low levels of U.S. equity investment because the data are reported in millions of dollars. Transactions that tend to fall below 1 million dollars will be reported as 0 or 1, inflating the variance in relation to the mean. However, the nearly universal decline in the coefficient of variation over time seems to suggest that as U.S. investors increase their hold- ings of foreign equity, they tend to transact less frequently. It also suggests that the concern in emerging markets about the high volatility of foreign investment in their markets may be relevant only in the short run. 124 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Table 5. Volatility of U.S. Transactions in Foreign Equity (millions of 1985 dollars) 1978-85 1986-91 Coeffi- Coeffi- cient of cient of Standard varia- Standard varia- Market Mean deviation tiona Mean deviation tiona Industrial markets Canada 1,383.31 542.s7 0.39 2,609.26 844.89 0.32 Germany 195.10 150.88 0.77 1,775.36 759.58 0.43 Japan 1,561.85 872.55 0.56 11,209.40 2,947.49 0.26 United Kingdom 1,359.14 933.32 0.69 14,912.77 4,412.64 0.30 Emerging stock markets Asia India 0.17 0.50 2.91 0.95 1.73 1.83 Indonesia 0.10 0.56 5.66 6.15 10.82 1.76 Korea, Rep. of 2.42 4.51 1.86 28.17 14.69 0.52 Malaysia 0.85 2.22 2.63 44.68 56.67 1.27 Philippines 5.05 8.92 1.77 15.63 15.97 1.02 Taiwan (China) 0.33 0.78 2.38 47.17 39.85 0.84 Thailand 0.06 0.35 5.66 33.47 27.09 0.81 Europe Greece 2.01 2.81 1.40 6.06 4.40 0.73 Portugal 0.22 0.67 3.04 7.42 9.20 1.24 Turkey 0.78 3.69 4.73 13.65 26.76 1.96 Latin America Argentina 2.05 5.33 2.60 9.98 12.08 1.21 Brazil 1.58 4.59 2.91 59.38 101.33 1.71 Chile 0.71 1.88 2.65 17.41 22.40 1.29 Colombia 0.63 1.41 2.24 2.37 2.50 1.06 Mexico 19.32 18.69 0.97 277.33 530.15 1.91 Note: U.S. transactions are U.S. purchases of foreign equity from foreign residents plus U.S. sales of foreign equity to foreign residents. Values are based on quarterly data deflated by the U.S. consumer price index (1985.3 = 100) for two periods: 1978.1-1985.4 and 1986.1-1991.3. a. The ratio of the standard deviation divided by the mean. Source: For transactions, U.S. Department of the Treasury (various issues); for the U.S. consumer price index, IMF (various issues). Table 6 provides information on four indicators. The first column gives the turnover rate in each market. This is the volume of equity traded over the quarter divided by the local market capitalization value. We use the average of the market capitalization value in the current quarter and the previous quarter. This is to avoid underestimating the turnover rate in economies with rapidly rising nominal market capitalization values caused by high inflation. The data reported are the four-quarter averages for 1990. The second column gives the U.S. transaction rate in each market, which is the U.S. volume of transactions rate divided by the volume traded on the local market. The U.S. volume of transactions in foreign equity is defined as the average of U.S. purchases and sales of foreign equity. The data reported are the four-quarter averages for 1990. The third column gives the standard deviation of excess returns in each market. Tesar and Werner 125 Table 6. Turnover and U.S. Transactions in Foreign Equity Markets Correlation be- U.S. trans- Local volatility, tween U.S. transac- actions share in 1986.1- tions share and Local turnover, local market, 1990.12' local turnover, Market 1 990a 1990b (percent) (dollars) 1986.1-1991.3d United States 0.119 n.a. 0.0299 n.a. Emerging Stock Markets Asia India 0.236 0.017 0.0890 -0.383 Indonesia 0.159 1.328 - -0.294 Korea, Rep. of 0.128 0.262 0.0895 -0.318 Malaysia 0.046 5.049 0.0863 0.110 Philippines, the 0.027 7.067 0.1271 0.116 Taiwan (China) 0.806 0.053 0.1789 -0.236 Thailand 0.133 2.010 0.0936 -0.079 Average 0.209 2.255 0.1107 n.a. Europe Greece 0.071 1.341 0.1526 -0.538* Portugal 0.028 5.149 0.1650e -0.144 Turkey 0.105 1.635 0.2361f -0.248 Average 0.068 8.125 0.1846 n.a. Latin America Argentina 0.072 3.654 0.3185 -0.325 Brazil 0.060 3.562 0.2299 -0.268 Chile - - 0.0824 Colombia 0.009 29.048 0.0651 -0.329 Mexico 0.111 12.537 0.1585 -0.438* Average 0.047 12.200 0.1709 n.a. * Significant at the 5 percent level. -Not available. n.a. Not applicable. a. Quarterly volume of equity traded (defined as total purchases) divided by the average of the current and last quarter's local market capitalization value. Figures shown are the average of the four quarters in 1990. b. Average of U.S. purchases plus sales of foreign equity during the quarter divided by the sum of total purchases of equity in the local market during the quarter. Figures shown are the average of the four quarters in 1990. C. Standard deviation of monthly excess returns over the U.S. Treasury bill rate. Values are local monthly returns in dollars less the returns on the U.S. Treasury bill. d. Correlation between the first and second columns for the longest time period for which data are available (typically 1986.1-1991.3). e. Calculated for 1986.2-1990.12. f. Calculated for 1987.1-1990.12. Source: IFC EMDB and Msci. Excess returns are local monthly returns in dollars less the return on the U.S. Trea- sury bill. The final column shows the correlation between the U.S. transaction share of total transactions on the local market and the local turnover rate. The turnover rate and the standard deviation of excess returns of U.S. equity are pro- vided as benchmarks. The first column of table 6 provides a comparison of turnover rates across markets.5 We find that Taiwan (China) stands out with a quarterly rate of 0.806. 5. See also Mullin (1993) for an extensive discussion of turnover, volatility, and market breadth in emerging markets. 126 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 With that exception, the remaining ESMs exhibit turnover rates of roughly the same magnitude or somewhat smaller rates than the United States.6 This seems to suggest that liquidity in ESMS is fairly high. The standard deviations of excess returns, however, are higher in all emerging markets than in the United States. Regionally, returns to equity investments in Latin America and Europe tend to exhibit higher volatility than rates in Asian markets. The figures in the second column of table 6 indicate that U.S. transactions are concentrated in a few of the emerging markets. U.S. residents are involved in roughly 29 percent of all transactions in Colombia, 12 percent in Mexico, 7 percent in the Philippines, 5 percent in Malaysia, and 5 percent in Portugal. When comparing regional averages, we see that U.S. investors tend to transact most frequently (as a share of all local transactions) in Latin America. This suggests that in the United States, at least, investors tend to turn to foreign markets with the closest proximity. This accords with the evidence in Tesar and Werner (forthcoming) that the United States invests disproportionately in Ca- nadian equity and that Canada invests disproportionately in U.S. equity. Policymakers have become increasingly concerned about the impact of inter- national investment on local market turnover and the volatility of equity re- turns. To address this question, we examine whether increased transactions by U.S. residents in foreign equity markets contribute to higher rates of turnover or higher standard deviations of excess returns. The fourth column of table 6 reports the correlation between the U.S. transactions share in the local market and the local turnover rate. Of the fourteen markets for which data are avail- able, only two of the correlations are significant at the 5 percent level. Both are strongly negative, suggesting that U.S. investment activity apparently has no effect, or a slightly dampening effect, on the local turnover rate. As further evidence, figure 7 plots a cross-section of turnover rates against the U.S. transactions share for 1990. There is little indication that markets with high U.S. investment activity experience higher rates of turnover than other mar- kets. Finally, figure 8 plots U.S. transactions shares against the standard devia- tions of local excess returns. Again, there is no evidence that U.S. investment activity contributes to volatility in equity returns. This corroborates the find- ings of Bekaert (1995), who reports that volatility in ESMS is unrelated to a number of measures of openness, such as foreign ownership restrictions and market integration through cross-listing of market funds. V. CONCLUSION We draw three main conclusions from our analysis of data on U.S. equity investment in 1978-91. First, despite the recent increase in U.S. equity invest- 6. Note that the U.S. transactions data reflect trading on the New York Stock Exchange and the American Stock Exchange. Thus, to the extent equity trading has migrated away from these markets in recent years, we underestimate the rate of turnover. Tesarand Werner 127 Figure 7. U.S. Transactions and the Local Turnover Rate Local turnover (quarterly rate) 1.0 * 0.9 0.8 Taiwan (China) 0.7- 0.6 0.5 0.4 0.3 India 0.2 0.1 El 3 aMexico 0.1 t3 E Colombia 0 4 8 12 16 20 24 28 U.S. transactions share in local market (percent) Source: Authors calculations (see table 6). Figure 8. U.S. Transactions and Local Volatility Local volatility (dollars) 0.36 - 0.34 Argentina 0.32 a 0.30 0.28 0.26 0.24 - 0.22 0.20 0.18 a Mexico 0.16 - 0.14 0.12 - 0.10 0.08 0 Colombia o.o6 - 0.04 l l 0 4 8 12 16 20 24 28 U.S. transactions share in local market (percent) Source: Authors' calculations (see table 6). 128 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I ment in foreign equities, including investment in emerging stock markets, the U.S. portfolio remains strongly biased toward domestic equities. Second, the increase in U.S. equity investment in ESMS is roughly in propor- tion to the share of ESMS in the global market capitalization value. That is, of the fraction of the U.S. equity portfolio that is held in the form of international equities, roughly 12 percent is allocated to ESMS. Although the volume of flows to these markets may appear large from the perspective of the recipient markets, these flows are substantially smaller than what would be observed if U.S. inves- tors were to follow an overall strategy of holding the market portfolio. Third, the volatility of U.S transactions is higher in emerging-market equities than in other foreign equities. However, the volatility of U.S. transactions ap- pears to be falling over time when normalized by the mean level of transactions. We find no evidence of a relation between the volume of U.S. transactions in foreign equity and local turnover rates or volatility of stock returns. These findings suggest that the activity of U.S. investors is not the source of excess volatility or high turnover on local equity markets. Thus, these data provide little support for limiting U.S. access to these markets. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Bekaert, Geert. 1995. "Market Integration and Investment Barriers in Emerging Equity Markets." The World Bank Economic Review 9(1):75-108. Calvo, Guillermo A., Leonardo Leiderman, and Carmen M. Reinhart. 1992. "Capital Inflows and Exchange Rate Appreciation in Latin America: The Role of External Factors." IMF Working Paper. International Monetary Fund, Washington, D.C. Capital International Perspective, S.A., and Morgan Stanley & Co.. Various issues. Morgan Stanley Capital International Perspective. New York, N.Y.: Morgan Stanley. Chuhan, Punam. 1992. "Sources of Portfolio Investment in Emerging Markets." Work- ing Paper. Processed. Divecha, Arjun B., Jaime Drach, and Dan Stefak. 1992. "Emerging Markets: A Quan- titative Perspective." Journal of Portfolio Management 19(Fall):41-50. French, Kenneth R., and J. Poterba. 1991. "Investor Diversification and International Equity Markets." The American Economic Review 37:222-26. Gooptu, Sudarshan. 1993. "Portfolio Investment Flows to Emerging Markets." Paper presented at a symposium sponsored by the World Bank, Debt and International Finance Division, Washington, D.C., September 9-10. Processed. Harvey, C.R. 1993. "Predictable Risk and Returns in Emerging Markets." Duke Univer- sity Working Paper. Duke University, Fuqua School of Business, Durham, N.C. Processed. . 1995. "The Risk Exposure of Emerging Equity Markets." The World Bank Economic Review 9(1):19-50. IMF (International Monetary Fund). Various issues. International Financial Statistics. Washington, D.C. Tesar and Werner 129 Mullin, John. 1993. "Emerging Equity Markets in the Global Economy." Federal Re- serve Bank of New York Quarterly Review 18(2):54-83. Steckler, L. E., and E. M. Truman. 1992. "The Adequacy of the Data on U.S. Interna- tional Financial Transactions: A Federal Reserve Perspective." International Finance Discussion Papers, Board of Governors No. 430. Processed. Tesar, Linda L., and Ingrid M. Werner. 1994. "International Securities Transactions and U.S. Portfolio Choice." In NBER, The Internationalization of Equity Markets. Chi- cago, Ill.: The University of Chicago Press. --- . Forthcoming. "Home Bias and High Turnover." Journal of International Money and Finance. U.S. Department of the Treasury. Various issues. Quarterly Bulletin of the U.S. Treasury. Washington, D.C. Wilcox, Jarrod W. 1992. "Taming Frontier Markets?' Journal of Portfolio Management 19(Fall):51-56. THE WORLD BANK ECONOMIC REVIEW, VOL. 9. NO. 1: 131-151 Return Behavior in Emerging Stock Markets Stijn Claessens, Susmita Dasgupta, and Jack Glen This article investigates the behavior of stock returns in the twenty stock markets represented in the International Finance Corporation's Emerging Markets Data Base. The aim is to test for return anomalies and predictability. Using statistical meth- odologies that have identified seasonal and size-based return differences, as well as general return predictability in industrial markets, we find that these emerging markets display few of the same anomalies. In particular, we find limited evidence of turn-of- the-tax-year effects and small-firm effects. We do find, however, evidence of return predictability. Emerging markets have seen a sharp increase in equity inflows in recent years (see Gooptu 1993). Despite apparent strong interest, however, little is known about the price determination process in these markets and how it compares with that of the more thoroughly studied markets in industrial economies. This article provides descriptive statistics on emerging stock markets and investigates the presence of some of the return anomalies that have been documented in other markets. It also presents what are, by now, commonly used measures of the time-series behavior of rates of return. There are several motivations specific to developing economies for research of stock market behavior. First, stock markets are believed to be very efficient at allocating capital to its highest-value users. Improved capital allocation in- creases overall economic efficiency. Second, stock markets play an important role in encouraging savings and investment, which are essential in economic development. Third, by allowing diversification across a variety of assets, stock markets reduce the risk that investors must bear, thus reducing the risk premium demanded by suppliers of capital and, through the risk premium, the cost of capital. The result should be increased investment levels and enhanced develop- ment. This is particularly relevant for foreign investment in which foreigners Stijn Claessens is with the Technical Department of the Europe and Central Asia and Middle East and North Africa Regions at the World Bank; Susmita Dasgupta is with the Policy Research Department at the World Bank; and Jack Glen is with the Economics Department at the International Finance Corporation. This article was funded in part through the World Bank research grant RPO 678-01. The authors thank the staff in the Emerging Markets Data Base division at the International Finance Corporation, especially Peter Tropper and Peter Wall, for assisting with the data, and Michael Adler, Geert Bekaert, Campbell Harvey, and the three referees for very useful comments. ©) 1995 The International Bank for Reconstruction and Development / THE WORLD BANK 131 132 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I share the risks. Fourth, the public corporation places ultimate decisionmaking power in the hands of shareholders. Managers of public corporations with traded equity claims are not free to make decisions in a vacuum but must consider shareholder and social responsibilities. The improved corporate gover- nance that results should contribute to economic development. The tests reported in this article provide information relevant to these issues. Return behavior, both in cross-sectional and in time-series analyses, provides information on the effectiveness of the asset-allocation role played by the stock markets in the economies examined. But even if capital allocation is efficient, foreign and domestic investors may be discouraged from investing in emerging stock markets if return behavior differs from expectations. To the extent that these deviations arise from exclusive access to information by certain firms or individuals-"insider trading"-the market will be stacked against "outsiders" and may be less likely to attract new investors, either foreign or domestic. This may limit the market's ability to generate savings and investment for the econ- omy. Consequently, measures of return behavior may be useful in research on the determinants and behavior of flows into the stock markets. And a fuller understanding of the general pattern of return behavior sheds important light on the risk diversification benefits of the stock market and the cost of capital. Section I describes the theory behind the tests. Section II describes the data sources. Section III presents the test results for economy indexes, and section IV the results for size-based portfolios of individual stocks. Section V offers conclusions. I. THE THEORY UNDERLYING THE TESTS Empirical tests of asset-pricing behavior are plentiful for industrial econ- omies. The terminology, however, is not always uniform. We make a distinction here between anomaly tests and time-series tests, both of which involve, implic- itly or explicitly, a test of an asset-pricing model. The distinction we make is that the anomaly tests focus on seasonal or cross-sectional patterns in rates of return whereas time-series tests focus on the predictability of rates of return over time. 1 For both types of tests we use a very parsimonious model for rates-of-return behavior-that is, rates of return are independently and identically distributed (i.i.d.). In this way we try to minimize the problem of jointly testing a particular asset-pricing model and market efficiency. For industrial economies, anomaly tests center on seasonal effects; for exam- ple, effects related to the beginning or end of the tax year or to the beginning or end of the calendar year and small-firm effects. See Keim (1987) and Fama 1. Instead of the terms "anomaly tests" and "time-series tests," we could have used the more general term "efficiency tests." As extensively discussed by Fama (1991), however, anomalies and unexpected time-series behavior may reflect risk factors that are priced in the market and do not necessarily imply inefficiencies. In principle, returns could be corrected for these factors by specifying an asset-pricing model, but the test will then jointly test market efficiency and the specific model. Claessens, Dasgupta, and Glen 133 (1991) for reviews of the literature on anomalies. The search for market anoma- lies is motivated by the concern that certain institutional features may induce return behavior that deviates from expected behavior (such as a random walk or other form of martingale). Whether policy changes aimed at reducing these anomalies are desirable cannot be answered in isolation. Anomalies do not necessarily indicate market inefficiencies but may simply reflect certain institu- tional features of markets (or certain risk factors for which the asset-pricing model used does not correct). For example, tax structures may cause turn-of- the-tax-year effects. Embracing or rejecting institutional causes for observed anomalous behavior on the basis of results from one or a few markets may be risky. A search for anomalies across a large group of markets, however, may lead to the acceptance or rejection of certain institutional-based explanations of mar- ket anomalies when there is sufficient variation in institutional features. The inclusion of the additional twenty markets in this study could potentially be more powerful than a marginally more sophisticated test on an already studied market. Time-series tests are a subset of market-efficiency tests. Efficiency tests center on the extent to which available information is absorbed in individual stock prices. As distinguished by Fama (1991), efficiency tests can be of several forms: tests for return predictability (including trading rules and autocorrelations), event studies (the adjustment of prices to public announcements), or tests for private information (whether private investors have specific information that is not included in market prices). We restrict our attention here exclusively to various forms of tests for return predictability. Motivation for the various types of time-series tests and the results of actual tests for industrial economies can be found in Fama (1991).2 Seasonality The literature on anomalies is both broad and deep. The search for sea- sonality in returns has been particularly rewarding in the markets in industrial economies. Keim (1988) reviews this literature and argues that although sea- sonality is an anomaly in the sense that it cannot be explained by existing asset- pricing models, seasonality does not necessarily represent deviation from market efficiency. Moreover, there is some evidence that institutional features of the market may play a role in return seasonality (see, for example, Roll 1983). Emerging markets have a variety of institutional features that differentiate them from one another and from the markets in industrial economies. The search for seasonality or other anomalies in the returns of the emerging markets can pro- vide important information on the role of institutional features on return behav- ior. This information may help stock exchange and regulatory authorities when they make policy decisions. 2. Fama (1991) also explains why he prefers the term "tests for time-series predictability" over the term "weak-form efficiency,' which he had previously introduced (Fama 1970). 134 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Our tests are strongly influenced by the findings of others. In particular, we examine returns to identify extraordinary returns in January and at the turn of the tax year. To test for seasonality, we employ three different but related tests. We test for equal-mean rates of return using the Kruskal-Wallis test as described in Gultekin and Gultekin (1983). This is a nonparametric test of the hypothesis that the mean returns in all months are equal. Two tests of abnormal returns for a particular month, for example January, are also performed. The first, as described in Corhay, Hawawini, and Michel (1987), is a parametric test that uses regressions of returns on dummy variables for each of the months being tested for abnormal behavior. The second test is a variant of the Kruskal-Wallis test and is also described in more detail in Gultekin and Gultekin (1983). That statistic tests whether the mean rate of return for a specific month is significantly greater than the mean return for any other month of the year. Predictability The predictability of returns was originally viewed as a form of market ineffi- ciency. Views on this have changed somewhat for two reasons. First, even though the existence of some forms of return predictability is accepted, invest- ment decisions made on the basis of that evidence are inevitably risky because of the low level of predictability that has been identified. Consequently, any returns earned by predicting returns may simply reward the investor for the extra risk borne. Second, predictability may arise from the basic factors that drive stock returns. The fact that the models of returns are unable to capture those factors may indicate model failure rather than market inefficiency. Our tests of return predictability follow the literature (as reviewed by Fama 1991). We first examine the time-series behavior of returns by estimating auto- correlation coefficients. These are tested for differences from zero both individu- ally (using t-tests) and jointly (using the Ljung-Box Q-statistic). The hypothesis that returns follow a random walk provides a complementary test of return predictability. Much of the interest in this hypothesis has centered on the approach taken by Lo and MacKinlay (1988). They develop a statistic that employs the variance of returns using different periods for which stocks are held and over which returns are calculated. Under the random walk hypothesis, which assumes that returns are i.i.d. across time, return variance increases linearly with the length of the holding period. Consequently, the ratio of the return variances calculated over different holding periods should reflect the difference in the holding period in a known manner. Deviations indicate rejec- tion of the random walk hypothesis and provide evidence that returns are predictable. We employ the Lo-MacKinlay statistic (which is consistent under hetero- skedasticity) for holding periods of two and four months. It would be interesting to examine longer holding periods as well, but the relatively short sample pe- riods available preclude doing so. Claessens, Dasgupta, and Glen 135 Size Banz (1981) was the first to document a relation between market capitalization (or firm size) and mean return. He found, and others have confirmed, that small firms have, on average, higher returns than large firms. Subsequent work has also found that this small-firm effect can be seasonal; for the United States, it is typ- ically found that small-size stocks display a positive January effect; that is, returns in January are higher than in other months (see Keim 1988 and Fama 1991). The relation between size and mean return has been typically analyzed by grouping individual stocks into portfolios on the basis of market capitalization, with periodic regrouping to account for changes in size over time. Often, ten size-based portfolios are formed and then the performance of these portfolios is examined. Statistically significant differences in mean returns between large and small firms have been documented. Specifically, we rank for each market all stocks by market capitalization at the beginning of each year. We then assign the individual stocks-ranked by market capitalization-to one of four portfolios, each of which has the same number of stocks. The first portfolio has the smallest market capitalization stocks and the fourth portfolio the largest, with the other two portfolios in between. We then calculate the return on these four portfolios for the following twelve months, using the market capitalization of the individual stock relative to the market capitalization of the whole portfolio to weigh the individual stock returns. At the end of the year, we redo the portfolio-creating scheme. We thus create a timc series of returns for the four size-based portfolios in each market that are ranked from small market-capitalization to large market-capitalization, where the as- signment of stocks to the portfolios is updated annually. Once the size-based portfolios have been formed, we examine their returrL behavior using several of the methods employed for examining the market in- dexes. The examination includes testing for size effects on returns by comparing the mean return on the smallest market-capitalization of the size-based portfo.- lios with those of the largest market-capitalization (with a t-test) and by compar.- ing the mean returns on all four portfolios jointly (using an F-test). We also examine the portfolio returns for seasonality and for differences in predictability using the methods previously mentioned. II. DATA The data for developing economies are from the International Finance Corpo- ration (IFC) Emerging Markets Data Base (EMDB). We use monthly data oIn returns for the EMDB indexes as well as for individual stocks.3 The sample period varies by market; all time series end in December 1992, but there are several 3. The criteria used by the IFC for inclusion of stocks and the methodology for creating the indexes are described in IFC (1993b). Note that the EMDB indexes are value weighted. As pointed out by Harvey (1993), the IFC indexes before 1981 have a look-back survivorship bias because the sample of stocks that was included to create the pre-1981 index was based on a sample of stocks traded as of 1981. Table 1. Market Coverage and Market Concentration, December 1992 Percentage of value IFC EMDB Market Percentage of market traded held by the Number Capitalization Number Capitalization capitalization held by ten most-active Economy of stocks (billions of dollars) of stocks (billions of dollars) the ten largest stocks stocks Emerging markets Argentina 29 14.29 175 18.63 68.8 72.5 Brazil 69 23.20 565 45.26 29.3 51.2 Chile 35 21.93 80 29.64 78.5 62.9 Colombia 20 5.10 80 5.68 78.5 62.9 Greece 32 5.38 129 9.49 43.6 50.4 India 62 25.36 2,781 65.12 22.6 32.2 Indonesia 63 8.66 155 12.04 39.2 61.4 Jordan 27 1.99 103 3.37 49.4 31.6 Korea, Rep. of 91 66.46 688 107.45 30.5 22.4 Malaysia 62 47.94 366 94.00 30.9 14.0 Mexico 62 66.11 195 139.06 31.7 39.4 Nigeria 24 0.80 153 1.22 48.5 53.6 Pakistan 58 3.77 628 8.03 23.0 19.1 Philippines 30 8.17 170 13.79 52.2 30.6 Portugal 30 4.87 191 9.21 31.4 22.1 Taiwan (China) 70 60.45 256 1,101.12 30.2 15.4 Thailand 51 28.37 305 58.26 28.5 36.3 Turkey 25 2.87 145 9.93 29.9 11.4 Venezuela 17 4.99 66 7.60 59.6 80.0 Zimbabwe 17 0.27 62 0.63 36.5 47.7 Industrial markets Canada n.a. n.a. 1,119 243.02 30.8 n.a. France n.a. n.a. 786 350.86 28.7 n.a. Germany n.a. n.a. 665 348.14 40.4 n.a. Japan n.a. n.a. 2,118 2,399.00 16.5 n.a. Switzerland n.a. n.a. 180 195.29 53.5 n.a. United Kingdom n.a. n.a. 1,874 838.58 24.2 n.a. United States n.a. n.a. 7,014 4,757.88 14.9 n.a. n.a. Not applicable. Note: Market refers to the local index as reported by the national stock exchanige or local authority. Market capitalization is the number of shares outstanding per stock, times the end-of-period price for each stock, summed over stocks. In all tables, amounts refer to U.S. dollars, unless otherwisc noted. Source: IFC EMDB, IFC (1993a, b), and MSCI. Claessens, Dasgupta, and Glen 137 starting dates. Thus, the sample period is quite short for some economies (most notably Indonesia, Portugal, and Turkey), and consequently the results for those economies should be interpreted cautiously. The data for industrial economies are from Capital International Perspectives, S.A. and Morgan Stanley & Co. (various issues), hereafter referred to as MSCI. The EMDB indexes cover only a subset of all stocks listed on the various exchanges, varying between 5 percent (Taiwan, China) and 90 percent (Col- ombia) in terms of market capitalization and 2 percent (India) and 44 percent (Chile) in terms of number of stocks. Table 1 presents data on the relative coverage of the EMDB indexes. Typically, because of its selection criteria, the EMDB index will be weighted toward the larger market-capitalization stocks. Returns are calculated as the percentage change in price adjusted for divi- dends, rights issues, and stock splits. For each market, returns can be calculated in the local currency or in a foreign currency, such as the U.S. dollar. For ease of comparison across economies, we use dollar rates of return. We begin by performing tests on the EMDB indexes. Those tests are followed by tests using the size-based portfolios of individual EMDB stocks. III. TEST RESULTS FOR EMDB ECONOMY INDEXES Table 2 presents the mean, standard deviation, Sharpe ratio (mean return over standard deviation), skewness, kurtosis, range, and Jarque-Bera normality test statistic for monthly returns from the EMDB economy indexes. The statistics are for emerging stock markets in twenty economies; for composite indexes for all twenty economies, for Latin America, and for Asia; and for several industrial economy indexes. The mean rates of return for emerging markets in table 2 are in general high but so, too, are the standard deviations. The highest mean rate of return is for Argentina, almost 6 percent on a monthly basis (68 percent on an annual basis). However, Argentina also has the highest standard deviation, almost 30 percent (103 percent on an annual basis), and the highest return range, 243 percent. Indonesia has the lowest mean return (-1.019 percent), and Jordan has the lowest standard deviation (5.165 percent). In general, the returns and their standard deviations for the individual emerging markets, as well as the regional and composite EMDB indexes, are higher than those for the industrial economies. Figure 1 plots the mean rate of return against the standard deviation. For developing economies, the tradeoff between the rate of return and the standard deviation is high, especially when compared with the tradeoffs of industrial economies. There is some risk diversification among the markets, as indicated by the standard deviation for the EMDB composite index (of all emerging mar- kets), which is less than the index for almost all of the individual markets (and significantly less than the average of those markets). The emerging markets, however, are still risky. The Sharpe ratio (mean return over standard deviation) in table 2 is an indicator of the relative risk-return tradeoff in each market. Table 2. Summary Statistics of Monthly Percentage Changes in Total Return Indexes Starting date Jarque-Bera (year and Standard Sharpe normality Economy month) Mean deviation ratioa Skewness Kurtosis Range test statistic Emerging markets Argentina 1976:01 5.656 29.996 0.189 1.98 7.44 243.1 567.7 Brazil 1976:01 1.841 17.390 0.106 0.52 1.07 114.4 17.6 Chile 1976:01 3.055 11.425 0.267 0.94 3.23 90.9 110.8 Colombia 1985:01 3.637 9.279 0.392 1.74 4.33 54.8 109.5 Greece 1976:01 0.622 10.456 0.060 1.85 7.54 89.4 563.8 India 1976:01 1.684 7.860 0.214 0.67 2.33 59.7 56.6 Indonesia 1990:01 -1.019 9.397 -0.108 0.13 -0.07 39.6 0.2* Jordan 1979:01 0.896 5.165 0.173 0.46 0.98 29.2 11.3 Korea, Rep. of 1976:01 1.772 9.335 0.190 1.00 2.04 64.1 65.1 Malaysia 1985:01 1.154 7.606 0.152 -0.65 2.33 51.4 24.0 Mexico 1976:01 2.533 12.864 0.197 -0.83 3.82 98.9 137.0 Nigeria 1985:01 0.225 10.538 0.021 -1.80 11.90 94.9 540.6 Pakistan 1985:01 1.794 6.698 0.268 2.13 10.17 51.1 426.9 co Philippines 1985:01 3.775 11.023 0.343 0.46 2.30 71.7 20.3 Portugal 1986:01 2.881 14.498 0.199 1.61 5.48 100.1 119.9 Taiwan (China) 1985:01 2.835 15.271 0.186 0.25 1.08 88.9 4.3k Thailand 1976:01 1.861 7.435 0.250 -0.10 3.38 61.5 89.4 rurkey 1987:01 3.145 21.436 0.147 1.07 1.08 100.8 15.2 Venezuela 1985:01 2.682 13.657 0.196 0.11 3.33 98.3 36.9 Zimbabwe 1976:01 0.648 9.865 0.066 0.29 1.97 73.9 32.4 Composite 1985:01 1.503 6.984 0.215 -0.64 4.29 40.7 13.2 Asia 1 985:01 1.498 7.421 0.202 -0.70 4.33 40.8 14.9 Latin America 1985:01 2.605 11.210 0.232 -0.05 3.75 66.9 2.3 Industrial markets Japan 1976:01 1.02 5.20 0.196 -0.04 5.06 39.6 38.1 UJnited Kingdom 1975:01 2.04 6.87 0.297 1.84 18.49 80.0 2,281.1 United States 1976:01 1.19 4.39 0.271 -0.44 5.88 34.4 76.7 World 1976:01 1.18 4.07 0.290 -0.47 4.91 29.0 38.9 * Denotes inability to reject normality at the 1 0 pcrcent level. For other markets, normality could be rejected at the 1 percent level. Note: All series are in U.S. dollars and end in December 1992. a. 'rhe Sharpe ratio is the ratio of the mean return (the second column) to the standard deviation (the third column). Source: IFC EMOB, MSCI (various issues), and authors' calculations. Claessens, Dasgupta, and Glen 139 Figure 1. Mean-Variance Frontier in Return Indexes Mean rate of return (percent) 50 - E Philippines ta Colombia 40 - Turkey a Chile 3 Portugal E; m Taiwan (China) Latin a n Venezuela 30 - America EIlMexico United Kingdom La Hong Kong p k ta Thailand El India S Korea C Brazil 20 - France Asa 0 Belgiurnrj Asaomposite 20~Europe \ Italy World Australia E? N herlandS3axalaysia U.S. Caaa ina JMapan m 10 - J 3 Jordan Germany Zimbabwe Switzerland 3 El Greece 0 Nigeria 0 I I I I 10 20 30 40 50 60 70 80 Standard deviation of rate of return (percent) Note: Composite is the IFC Composite Index for Emerging Markets; EAFE represents Europe, America, and the Far East (MscI Index); and Sinmal represents Singapore and Malaysia. Returns have been calculated'as annualized monthly dollar rates of return. Source: Authors' calculations. Colombia has the highest Sharpe ratio, almost 0.4. Except for Indonesia, which has a negative Sharpe ratio (-0.108), Nigeria has the lowest Sharpe ratio (0.021). The Sharpe ratios for many emerging markets are similar to those for industrial economies, and the EMDB composite index has a marginally lower ratio than the MSCI world index, indicating that, by this measure, these econ- omies have a similar risk-return tradeoff as industrial economies do. The skewness and kurtosis measures indicate that the rates of return are not likely drawn from normal distributions. In general, the kurtosis measure is mucl higher than the value of 3 associated with the normal distribution, indicating that the distributions have fat tails. The Jarque-Bera test for normality bears this out; normality is rejected for eighteen of the twenty economies at the 1 percent level. This is also true for industrial economies: normality cannot be accepte(d for any of the sixteen economies or indexes reported by MSCI.4 Similar rejections 4. Not all sixteen industrial economies are reported here; see Claessens, Dasgupta, and Glen (1993). We would like to thank Michael Adler for pointing out that normality is also rejected for industrial economies. 140 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I have been found in the foreign exchange literature. It has become clear that this is primarily because the time series are not i.i.d.; they could have been produced by mixtures of independently but nonidentically distributed normal distribu- tions. The rejections here could be caused by similar processes. An explanation that is more specific to emerging markets is that these devia- tions from normality are the result of "peso" (or regime-switching) effects. It may have been the case that investors thought that certain (adverse) events could happen during the sample period, but ex post the events did not happen (or at least not at the frequency investors had in mind). This would result in rates of return that ex post are not only higher than ex ante rates of return but also display deviations from normality. Similarly, investors could have become con- vinced that the economy's economic prospects and policies were better than they had expected earlier on, making ex post rates of return higher than ex ante returns and displaying the distributions found here. Seasonality As mentioned, a frequently tested market anomaly is seasonality. To test for seasonality in the return series, we use the nonparametric Kruskal-Wallis statis- tic, which tests the equality of returns for all months. The first column of table 3 presents the Kruskal-Wallis test statistic for the EMDB economy indexes. Equality of returns for all months is rejected (at the 5 percent level) for only three markets: Chile, Greece, and Jordan. This result contrasts with the results of Gultekin and Gultekin (1983), who, using the same test, find evidence of sea- sonality (at the 10 percent level) in twelve of the seventeen industrial economies they studied. To determine the existence of a turn-of-the-tax-year or other seasonal effect, we also test whether the return in any given month is different from that of the other months considered jointly. As mentioned above, we do this in two ways: first using a parametric test (Corhay, Hawawini, and Michel 1987) and then using a variant of the Kruskal-Wallis test. The second column of table 3 provides the months for which there is a significant dummy variable in at least one of the tests. There is a significant (at the 10 percent level) month-of-the-year effect for the economy indexes for sixteen markets. For three markets-Chile, Jordan, and the Philippines-three of the twelve months are significantly different from the other months; for six markets-Argentina, India, Indonesia, Malaysia, Nigeria, and Portugal-two of the twelve months are significantly different from the other months. January is significantly different for three markets-the Republic of Korea, Mexico, and Turkey-and the significantly different month coincides with the beginning of the tax year for only two markets-Malaysia (May) and Mexico. Pakistan is the only economy in which the end-of-the-fiscal-year month is significantly different from other months. These results contrast with those of Corhay, Hawawini, and Michel (1987), who find that for each of the four markets they studied (the New York Stock Exchange, the London Stock Ex- Claessens, Dasgupta, and Glen 141 Table 3. Tests of Seasonal Differences in EMDB Monthly Returns Economy index Size-based portfolios Kruskal- Wallis Positive Negative statistics significant significant Economy (p-value) Significant monthsb monthsc monthsc Argentina 11.27 August, October (0.42) Brazil 9.47 April (0.58) Chile 21.02* February, October, September (0.03) November Colombia 11.98 March January (0.37) Greece 21.51 * October October (0.03) India 14.28 October, November July, November (0.22) Indonesia 13.08 February, September June, (0.29) September Jordan 19.76* May, August, December August, October (0.05) November Korea, 10.33 January March, Rep. of (0.50) September, October Malaysia 15.26 May, August (0.17) Mexico 9.16 January June (0.61) Nigeria 10.19 March, June (0.51) Pakistan 7.55 December (0.75) Philippines 11.95 June, August, August, (0.37) September October Portugal 10.25 September, November February (0.51) Taiwan 8.31 February, March (China) (0.69) Thailand 6.56 (0.83) Turkey 10.77 January September (0.46) Venezuela 4.77 February, (0.94) September Zimbabwe 3.53 August, January (0.98) December * Indicates that equality of returns for all months is rejected at the 5 percent level. Note: Blank spaces denote no significant month(s). a. Tests whether the ranks of the rates of return in each month are equal. The statistic is derived as follows: H = 12/[M(M+1)] E_l T, (X,-X)2-X2C11) where X_ is the average rank received by the rates of return in the mth month and X is the average rank of all observations. b. Those months for which the t-statistic for the dummy measuring the difference between the mean return in the particular month and the mean returns during the rest of the year is significant for the index. c. The months for which a t-test between the mean return on the smallest-size portfolio and that on tlae largest-size portfolio is significant at the 5 percent level, either positive or negative. Source: Authors'calculations. 142 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 change, the Paris Stock Exchange, and the Belgian Stock Exchange) there is a significant seasonal effect in January.5 Table 3 also reports (in the third and fourth columns) for which months there is a significant difference between the mean return on the smallest-size portfolio and that on the largest capitalization portfolio. This is one way to check whether the significant seasonal effect identified for the economy indexes (first and sec- ond columns) can be attributed to the fact that the return of the smallest capital- ization stocks is significantly different from the largest capitalization stocks in the same month(s). For five countries (Greece, India, Indonesia, Jordan, and the Philippines) the smallest-size portfolio has a larger return than the largest-size portfolio in the same month in which the economy index has a significant month-of-the-year effect. This suggests that for these economies part of the seasonality effect may be attributable to the behavior of the smallest-size portfo- lios. We explore this further in Section IV. To check further for specific seasonality tests, and again conforming to Gultekin and Gultekin (1983), we test whether the mean rate of return for a spe- cific month is significantly greater than the mean return in any other month. We do this for January for all markets and for the four markets for which the first month of the fiscal year is not January (Greece, India, Malaysia, and Pakistan). We find no January seasonal effect for any market, nor are any of the four other tax-year effects significant (results are not reported). This result contrasts with that of Gultekin and Gultekin, who find that for twelve of the thirteen industrial economies they studied, the mean rank of the rate of return in the first month of the tax year is significantly greater than the mean rank in any other month. Our results indicate that these findings do not carry over to emerging markets. From these results, it appears that the turn-of-the-tax-year effects found for many industrial economies do not extend to the emerging markets studied here. The weak power of some of our tests (given the short time series and the high volatility of the rates of return) could explain why we do not find this effect. Equally possible, however, could be that the tax codes of these economies do not give rise to the selling of stocks at the end of the tax year to generate a loss for tax purposes, the hypothesis often put forward as an explanation for the turn- of-the-tax-year effect in the industrial economies. In addition to tax codes that are designed differently in developing economies (compared with industrial economies), weak enforcement of tax codes could well explain the lack of evi- dence for the tax-loss-selling hypothesis. And, of course, other institutional factors are likely different for emerging-market economies. Predictability Predictability of returns can be evidence of market inefficiency if the asset- pricing model used implies that expected returns are constant through time. In 5. Their seasonal findings are restricted to January only in the United States; there is also an April effect that coincides with the beginning of the tax fiscal year for the U.K. market; for the other two markets, the other month-of-the-year effects do not occur in a turn-of-the-tax-year month. Claessens, Dasgupta, and Glen 14i models that allow for time-varying risk premiums, predictability of returns does not necessarily indicate market inefficiency. For further work along these lines, see Harvey (1995). Here we investigate one simple form of market predictabil- ity-autocorrelation-as a test of predictability. Table 4 presents the first and second autocorrelations and Ljung-Box Q-tests to check for the joint signifi- cance of the first twelve autocorrelations. We find, as others did (for example, Harvey 1993 and Bekaert 1993), signifi- cant predictability in rates of return. First-order autocorrelations are significani: for nine economies: Chile, Colombia, Greece, Mexico, Pakistan, the Philip- pines, Portugal, Turkey, and Venezuela. For Chile and Greece the second-order autocorrelations are also high. For Chile, Colombia, Greece, Mexico, Pakistan, and Zimbabwe, the Ljung-Box Q-test indicates that the first twelve autocorrela. tions are jointly significantly different from zero (at the 5 percent level). For most industrial economies, first-order autocorrelations are generally not higher than 0.2. For the sixteen industrial economies in the MsCI data base, using the same period (1976-92), Belgium, Italy, and Switzerland have the highest first- Table 4. Autocorrelations of EMDB Monthly Returns, 1976-92 Number of size- based portfolios for First-order Second- Ljung-Box which Ljung-Box auto- order auto- Q-statistica Q-statistic is Economy correlation correlation (12 lags) significantb Argentina 0.054 0.066 10.18 Brazil 0.029 -0.038 8.54 0 Chile 0.169* 0.260* 44.64* 4 Colombia 0.489* 0.147 35.35* 2 Greece 0.132* 0.179* 25.18* 4 India 0.079 -0.099 18.09 1 Indonesia 0.284 0.196 22.93 0 Jordan 0.000 0.024 15.27 0 Korea, Rep. of -0.001 0.082 7.81 0 Malaysia 0.052 0.060 9.54 0 Mexico 0.247* -0.074 24.02* 4 Nigeria 0.085 -0.126 14.69 0 Pakistan 0.250* -0.273 45.94* 3 Philippines 0.338* 0.028 17.53 0 Portugal 0.287* 0.040 19.38 1 Taiwan (China) 0.074 0.051 12.50 0 Thailand 0.114 0.149 20.84 3 Turkey 0.232* 0.106 20.03 0 Venezuela 0.267* 0.184 14.31 0 Zimbabwe 0.138 0.154 42.24* 3 - Not available. * Indicates rejection at a 5 percent level of significance. a. Distributed as x2 and with 12 degrees of freedom. b. The number of size-based portfolios for which the Ljung-Box Q-statistic is significant at the 5 percent level. Source: Authors' calculations. 144 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Table 5. Variance Ratio Tests of EMDB Monthly Rates of Return Two-month holding period Four-month holding period Number of Number of IFC market index significant IFC market index significant Variance size-based Variance size-based Economy ratio z(q) portfolios* ratio z(q) portfoliosa Argentina 1.027 0.258 0 .916 -.447 0 (.796) (.655) Brazil 1.007 0.084 0 .976 -.166 0 (.933) (.868) Chile 1.212 2.910* 2 1.553 3.902* 4 (.004) (.000) Colombia 1.497 2.251 2 1.842 2.135' 2 (.024) (.033) Greece 1.155 2.202* 1 1.422 2.607* 2 (.028) (.009) India 1.078 0.747 0 .991 -.047 0 (.455) (.962) Indonesia 1.312 1.828 1 1.709 2.241 ' 0 (.068) (.025) Jordan 1.005 .050 0 1.146 .898 1 (.961) (.369) Korea, Rep. of 1.023 .326 0 1.145 1.047 0 (.744) (.295) Malaysia 1.121 1.026 1 1.222 1.083 2 (.305) (.279) Mexico 1.279 1.889 2 1.320 1.295 2 (.059) (.195) Nigeria 1.117 .960 0 1.017 .082 0 (.337) (.935) Pakistan 1.266 1.080 0 1.001 .003 0 (.280) (.998) Philippines 1.421 3.208* 1 1.736 3.264* 1 (.001) (.001) Portugal 1.333 1.679 1 1.564 1.522 1 (.093) ( .12) Taiwan (China) 1.084 .587 0 1.202 .751 0 (.557) (.453) Thailand 1.123 1.182 0 1.332 1.807 1 (.237) (.071) Turkey 1.255 2.149* 1 1.310 1.422 1 (.032) (.155) Venezuela 1.241 2.405* 2 1.623 3.499 4 (.016) (.001) Zimbabwe 1.158 2.025* 1 1.541 3.818* 2 (.043) (.000) * Indicates that the null hypothesis that the rates of return are i.i.d. is rejected at the 5 percent level. Note: Significance levels of the test statistics are in parentheses. The statistics have been adjusted for heteroskedasticity. a. The number of size-based portfolios for which the z(q) statistic (q denotes the observation interval, two or four months) is significant at the 5 percent level. Source: Authors' calculations. Claessens, Dasgupta, and Glen 145 order autocorrelations (not reported), but these are still much below 0.2.6 We find that seven of twenty emerging-market economies have a first-order autocor- relation coefficient higher than 0.2. Compared with industrial economies, the aggregate indexes for the emerging markets thus display a high degree of auto- correlation in their rates of return. We also examine the variance ratios of returns as a test of the random walk hypothesis described above. The results, adjusted for heteroskedasticity, are reported in table 5. We report the variance ratios (a measure of the autocor- relation) and the z-statistics, using a one-month base observation period and periods of two and four months to form the variance ratios. Using a period of two months to form the variance ratio, we can reject the null hypothesis that: the rates of return are i.i.d for seven economies (Chile, Colombia, Greece, the Philippines, Turkey, Venezuela, and Zimbabwe) at the 5 percent level. Using a period of four months to form the variance ratio, we can reject the i.i.d. assumption again for seven economies (the same seven except for Turkey, which is replaced by Indonesia). The variance ratios are higher for the four month period than for the two-month period, and the rejections are also (mostly) at higher significance levels for the four-month period. The fact thait we can reject with higher significance using a longer comparison period con- trasts with the findings of Lo and MacKinlay (1988); when they compare variances for longer holding periods to monthly variances, the significance of rejection declines. The variance ratio test leads to results that are similar to those of the Ljung- Box Q-test. Only for Mexico and Pakistan, which have a significant Ljung-Box Q-statistic, does the variance ratio test not reject (at the 5 percent level). There is an even closer correspondence between the variance ratio test and the level of first-order autocorrelation. For all seven economies except Zimbabwe, the first- order autocorrelation is also high. The general results for the emerging markets contrast with those of Lo and MacKinlay, who investigate the value-weighted and equally weighted joint New York Stock Exchange-American Stock Ex- change index. Although Lo and MacKinlay are able to reject the random wallk hypothesis for various sample periods when using weekly rates of return, they are not able to reject the hypothesis when using monthly rates of return. IV. TEST RESULTS FOR SIZE-BASED PORTFOLIOS OF INDIVIDUAL EMDB STOCKS In this section we examine the differences in the behavior of portfolios of EMDB stocks. The portfolios are formed by sorting the stocks in each economy by size and then allocating the stocks to four size-based portfolios as previously described. The tests follow the form of those in section III but provide additional evidence that is not available from the economy indexes. To conserve space, we 6. Gultekin and Gultekin (1983) find that only three (Austria, Denmark, and Singapore) of the seventeen economies in their study have first-order autocorrelations higher than 0.2. 146 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 do not present basic statistics on the portfolios but only summary statistics of the tests of predictability. Details are available in Claessens, Dasgupta, and Glen (1993). As with the economy indexes, we calculate statistics on market capitalization, monthly mean, standard deviation, Sharpe ratios, skewness, kurtosis, range, and Jarque-Bera normality tests (not reported). What is surprising is that the difference in the relative importance of each of the four portfolios in terms of total market capitalization is quite large, in spite of the limited number of stocks in each market. For example, in Greece the portfolio of large stocks represents 70 percent of the market capitalization of all EMDB stocks, even though each of the four portfolios contains the same number of shares. In contrast, the portfolio of small stocks represents a mere 3 percent of EMDB stocks. Other statistics reveal that the returns on the four portfolios are, in general, more volatile than the aggregate market indexes. The differences are small, however, and in part reflect the fact that we create only four portfolios and that the number of stocks in the indexes is small (and consequently the portfolios are not very diversified). The Sharpe measures are in the same range as those for the indexes, as are the skewness and kurtosis measures. Also similar to the indexes, the portfolio rates of return do not appear to come from normal distributions. Using the Jarque-Bera test, seventy of the seventy-six portfolios reject normality (at the 5 percent level). There is a strong relation between the size-based portfo- lio tests and the economy index tests of normality, as is evident from the behav- ior of the portfolios in Indonesia and Taiwan (China). For both economies, for two of the four portfolios normality cannot be rejected, just as for the indexes of these economies. In addition, two of the Malaysian portfolios appear to be normally distributed. Seasonality The relation between seasonality of returns and size-based portfolios is re- ported in table 6. The first column presents in summary form the results of Kruskal-Wallis tests of equality of rates of return across all months for each of the four size-based portfolios. From the tests, there appears to be no relation between seasonality and size, because the number of economies in which returns differ significantly is about equally distributed across the different size portfolios. The individual portfolio results show that the significant seasonality found for the indexes of Chile, Greece, and Jordan can be attributed to the behavior of a particular portfolio only for Jordan, in which case the portfolio of largest firms and not that of the smallest firms rejects equality across all months. For Chile and Greece the seasonal effect is significant (at the 5 percent level) in three of the four portfolios. We find additional seasonal effects at the portfolio level for only two other economies; the portfolio of the smallest Colombian stocks and the portfolio of the largest Indonesian stocks display seasonality. Claessens, Dasgupta, and Glen 14 7 Table 6. Test of End-of-Tax-Year or Other Seasonal Effects by Size Portfolios oJ EMDB Stocks (number of economies for which test statistics reject equality of returns) Kruskal- First month Wallis of the tax Other Portfolio statistica Januaryb yearc monthsd Total Small firms 2 3 0 8 13 Next-to- smallest firms 2 4 1 8 15 Next-to- largest firms 2 4 0 11 17 Largest firms 3 3 1 15 22 Total 9 14 2 42 67 a. Tests that all months have equal returns. b. Shows for how many economies January returns differ from other months' returns for that size portfolio. c. Shows for how many economies the return for the first month of the tax year (if different fromn January) differs from other months' returns for that size portfolio. d. Shows for how many economies the return is different for at least one month (other than January or the first month of the tax year) for that size portfolio. Source: Authors' calcuilations. To identify the month in which any seasonality occurs, we again test equality of returns across months for each size portfolio by using the dummy-variable approach of Corhay, Hawawini, and Michel (1987). We focus on the effects in the first month of the tax year and on the months in which the economy indexes display seasonality. Again, we do not find that in these cases seasonality is explained by small market-capitalization stocks. There is no strong, consistent pattern across size portfolios, and of the 912 tests, only 67 (7 percent) are significant. The detailed results are as follows: for two markets-Thailand and Zimr- babwe (both January)-it is the portfolio of smallest stocks that displays sea- sonality; for others-India (April, October, and November), Jordan (August'>, Malaysia (May), and the Philippines (August)-it is the portfolio of the largest stocks; and for the other markets, intermediate-size portfolios or portfolios of various sizes have a significant size effect in the month in which the index shows a significant size effect or in which the (tax) year starts (for example, for Jordan the two intermediate-size portfolios show a significant month-of-the-year effect in May, and for Chile all size portfolios show a significant rate-of-return effect in February, October, and November). If anything, we find that larger-size portfolios are more likely to display seasonality in months other than the first month of the tax year (the fourth column). On balance, we conclude that there is little evidence of a seasonal or turn-of-the-tax-year effect that varies with the market capitalization of the stocks. Size Effects The basic statistics on mean returns suggest that for some markets there may be a size effect but one that is not necessarily restricted to the smallest-size 148 THE WORLD BANK ECONOMIC REVIEW, VOL 9, NO. I stocks. For eight of nineteen markets, the stock portfolio with the smallest market capitalization has the highest rate of return of the four portfolios (but for many of these markets these portfolios also have the highest standard deviation). For only two of these markets (Mexico and Zimbabwe), however, can we reject (through a pairwise t-test) the hypothesis that the average annual rate of return for the smallest-size stocks is significantly (at the 5 percent level) higher than that of the largest-size stocks (not reported). For six other markets, the second- quartile portfolio has the highest rate of return. To investigate further whether there is a significant size effect, we use an F-test for equality of returns across all portfolios. For no economy is the difference between any of the four portfolios significant at the 5 percent level. To test whether there are size effects that differ by month of the year, we perform a pairwise t-test for equality of rates of return between the smallest- and largest-size portfolios for each month of the year. These tests, summarized in the third and fourth columns of table 3, indicate that the positive size effects for the whole year may be attributed to the month of June for Mexico and to August and December for Zimbabwe. We find significant positive size effects for four- teen economies: for six economies in one month, for seven economies in two months, and for one economy in three months. We find negative size effects for Zimbabwe (January) and Jordan (October). Altogether, of the 228 market- month combinations (19 economies times 12 months), we find 23 market- months in which there is a positive size effect and 2 market-months in which there is a negative size effect. These results confirm what was reported above. That is, although we identify significant size effects, they are not limited to the turn of the tax year, as is most often found for industrial economies, nor are they always positive. To further check for size effects that differ by month of the year, we perform a number of other (cross-sectional) regressions of the individual rates of return in every month on their (lagged as well as contemporaneous) market capitaliza- tions and, respectively, each individual month and year, each month of the year, and the whole period. We find very little evidence that rates of return and market capitalization are consistently negatively related. See Claessens, Dasgupta, and Glen (1993). It appears, therefore, that the size effect found in many industrial economies does not prevail as systematically in the emerging markets. One explanation could be that the EMDB data do not include the smallest-size stocks and that we therefore cannot pick up a small-size effect. The relative market capitalizations, however, indicate that there is considerable size difference in the stocks in the sample. A small-size effect would appear, to some extent, in any differences between the rates of return on the EMDB indexes and the local market indexes (which cover more stocks and thus also smaller ones). To investigate this possi- bility, we run pairwise, one-way t-tests over the 1987-92 period to check whether the local market indexes have significantly higher rates of return than the EMDB indexes, both measured in local currency (not reported). We find this Claessens, Dasgupta, and Glen 149 to be the case for none of the twenty emerging markets. To the contrary, for three economies-Chile, Colombia, and Nigeria-we find that the local market: index is significantly (at the 5 percent level) lower than the IFC index. Again, we do not find evidence that the small-firm effect found in industrial economies is prevalent in these markets. Predictability To investigate predictability at the portfolio level, the fourth column of table 4 lists the number of significant Ljung-Box Q-statistics (for up to lag 12 autocor- relations) for the size-based portfolios in each market. We find evidence of significant (at the 5 percent level) autocorrelations for twenty-five of seventy-six portfolios. All four size portfolios are significant for three markets (Chile, Greece, and Mexico). For six other markets (Colombia, India, Pakistan, Portu- gal, Thailand, and Zimbabwe), the autocorrelations for one or more of the size portfolios are significant (at the 5 percent level). For the three markets with significant Ljung-Box Q-tests for all size portfo- lios, there does not appear to be a great difference in the levels of the first autocorrelations across the portfolios. A higher first autocorrelation for smaller stocks is generally found for industrial economies (using daily ancl weekly rates of return). For example, Fama (1991) reports that the autocor- relations for weekly rates of return of the four smallest deciles of the New York Stock Exchange during 1962-85 is 0.3, whereas it is 0.09 for the largest decile. Using monthly data, Lo and MacKinlay (1987) report autocorrelations of 0.23 for the smallest quintile of stocks and 0.06 for the largest quintile for the combined New York Stock Exchange-AMEx sample of stocks. Part of this higher autocorrelation for smaller stocks may be spurious. As Roll (1984) points out, because bid-ask spreads are likely larger for small stocks, spurious autocorrelation may be induced. Consequently, autocorrelations can be ex- pected to differ by the market capitalization of the stocks without necessarily indicating a failure of market efficiency. French and Roll (1986) adjust auto- correlations to offset this, in which case they decline from 0.27 to about 0.05 for the smallest-size stocks. For the developing economies, the first-order autocorrelation for small stocks is the highest of the four size portfolios for only six of nineteen economics (Chile, Jordan, Korea, Malaysia, Nigeria, and Turkey). Autocorrelation is high- est for the largest-size portfolios in five economies (Greece, India, the Philip- pines, Portugal, and Taiwan, China). For Colombia, the first-order autocorrela- tions for the smallest- and largest-size portfolios are equal. As pointed out by Lo and MacKinlay (1988) and others, positive autocorrela- tion for the indexes and portfolios may be an artifact of the portfolio-formation process. For individual securities, autocorrelations for industrial economies are mostly negative. We therefore calculate for each market the (simple) mean of the first-order autocorrelation for all securities (not reported). The means of the first-order autocorrelations for the individual stocks are smaller than the firsr- 150 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I order correlations of the economy indexes (they are actually negative for most markets). This suggests that the portfolio-formation process influenced the higher degree of predictability found for the index and portfolio rates of return. We also repeat the variance ratio tests for the size-sorted portfolios. A sum- mary of the results is presented in the third and sixth columns of table 5. The random walk hypothesis is rejected for fifteen of seventy-six portfolios (at the 5 percent level) using two-month holding-period returns. Using four-month holding-period returns, the hypothesis is rejected for twenty-three portfolios and, as for the market indexes, at generally higher significance levels than for the shorter horizon. The economies for which the portfolio results indicate rejection of the random walk largely overlap with those for which the random walk for the market index is rejected. As with the autocorrelations, there is no pattern across portfolios, and the random walk is equally often rejected for the smallest- and the largest-size portfolios. V. CONCLUSIONS Using standard anomaly and time-series tests, we find that stock price behav- ior in the twenty stock markets represented in the IFC'S EMDB displays few of the anomalies found for industrial economies. In particular, we find limited evidence of turn-of-the-tax-year effects, of small-firm effects, or of a relation between seasonal effects and size effects. We do find, however, a significant predictability of returns that does not appear to be related to size. As we do not explicitly specify an asset-pricing model, we are not able to determine whether this predictability is caused by market inefficiencies, time- varying risk premiums, peso problems, or possible regime switching. The short time series, the large volatility of the rates of return, and the large structural changes these markets have undergone in recent years may well contribute to our findings. Nevertheless, the relatively high degree of predictability could arise because these markets may not be a level playing field. If that were the case, the high predictability could in the long run be harmful: it may prevent uninformed investors from participating in these markets, reducing the markets' depth and liquidity and perhaps negatively affecting the role of the stock market in overall economic development. Facilitating better stock market behavior-for example, through improved disclosure requirements, accounting standards, and settle- ment procedures-may then be called for. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Banz, Rudolf W. 1981. "The Relationship between Return and Market Value of Com- mon Stocks." Journal of Financial Economics 9(1):3-18. Bekaert, Geert. 1993. "Market Integration and Investment Barriers in Emerging Equity Markets." In Stijn Claessens and Sudarshan Gooptu, eds. Portfolio Investment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. Claessens, Dasgupta, and Glen 151 Capital International Perspective, S.A., and Morgan Stanley & Co. Various issues. Morgan Stanley Capital International Perspective. New York,, N.Y.: Morgan Stanley. Claessens, Stijn, Susmita Dasgupta, and Jack Glen. 1993. "Stock Price Behavior in Emerging Markets." In Stijn Claessens and Sudarshan Gooptu, eds. Portfolio Invest- ment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. Corhay, Albert, Gabriel Hawawini, and Pierre Michel. 1987. "Seasonality in the Risk- Return Relationship: Some International Evidence." Journal of Finance 42(1):49-68. Fama, Eugene. 1970. "Efficient Capital Markets: A Review of Theory and Empirical Work." Journal of Finance 25(2):383-417. -.1991. "Efficient Capital Markets: II:' Journal of Finance 46(5):1575-1617. French, Kenneth R., and Richard Roll. 1986. "Stock Return Variances: The Arrival of New Information and the Reaction of Traders." Journal of Financial Economics 17(1):5-26. Gooptu, Sudarshan. 1993. "Portfolio Investment Flows to Developing Countries:' In Stijn Claessens and Sudarshan Gooptu, eds. Portfolio Investment in Developing Coun- tries. World Bank Discussion Paper 228. Washington, D.C. Gultekin, Mustafa N., and N. Bulent Gultekin. 1983. "Stock Market Seasonality: Inter- national Evidence." Journal of Financial Economics 12 (December):469-81. Harvey, Campbell. 1993. "Portfolio Enhancement Using Emerging Markets and Condi- tioning Information:' In Stijn Claessens and Sudarshan Gooptu, eds. Portfolio Invest- ment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. . 1995. "The Risks Exposure of Emerging Markets." The World Bank Economic Review 9(1):19-50. iFC (International Finance Corporation). 1993a. Emerging Markets Factbook. Washing- ton, D.C. . 1993b. "IFC Index Methodology.' IFC, Central Capital Markets Department, Washington, D.C. Processed. Keim, Donald. 1987. "Stock Market Regularities: A Synthesis of the Evidence and Explanations." In Elroy Dimson, ed., Stock Market Anomalies. Cambridge, U.K.: Cambridge University Press. 1988. "Stock Market Regularities: A Synthesis of the Evidence and Explana- tions.' In Elroy Dimson, ed., Stock Market Anomalies. Cambridge, U.K.: Cambridge University Press. Lo, Andrew, and A. Craig MacKinlay. 1987. "Stock Market Prices Do Not Follow Random Walks: Evidence from a Simple Specification Test." Working Paper 5-87. Rodney While Center for Financial Research, The Wharton School, University of Pennsylvania, Philadelphia, Pa. Processed. . 1988. "Stock Market Prices Do Not Follow Random Walks: Evidence from a Simple Specification Test." Review of Financial Studies 1(1)41-66. Roll, Richard. 1983. "Vas is Das? The Turn-of-the-Year Effect and the Return Premia of Small Stocks." Journal of Portfolio Management 9(2): 18-28. . 1984. "A Simple Implicit Measure of the Bid/Ask Spread in an Efficient Mar- ket." Journal of Finance 39(4)1127-39. THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 153-174 Portfolio Capital Flows: Hot or Cold? Stijn Claessens, Michael P. Dooley, and Andrew Warner A distinction is often made between short-term and long-term capitalflows: the former are deemed unstable hot money and the latter are deemed stable cold money. Using time-series analysis of balance of payments data for five industrial and five developing countries, we find that in most cases the labels "short-term" and "long-term" do not provide any information about the time-series properties of the flow. In particular, long-term flows are often as volatile as short-term flows, and the time it takes for an unexpected shock to a flow to die out is similar acrossflows. Long-term flows are also at least as unpredictable as short-term flows, and knowledge of the type of flow does not improve the ability to forecast the aggregate capital account. Several developing countries have received large capital inflows in recent years, reversing a trend of outflows for most of the 1980s (see Gooptu 1993). Much of this new capital inflow has been short-term portfolio investment, including bonds, equities, and short-term instruments such as certificates of deposit and. commercial paper. This surge in short-term flows has raised the question of whether these flows will be sustained or instead be reversed in the near future. Some observers argue that the recent flows are inherently unsustainable be- cause they have short maturities. For example, on the basis of this argument. Reisen (1993:2) concludes that "the majority of flows [to Latin America] are hot rather than cool." Nunnenkamp (1993) employs a similar approach and points out that the composition of inflows varies considerably among developing coun- tries. His conclusion is that hot money transactions have been relatively small in. the Chilean case but significantly large in Brazil. And Turner (1991), in his review of capital flows for industrial countries, ranks short-term bank lending as most volatile and long-term bank flows as least volatile, followed by foreigrn direct investment (FDI) as the next-to-least volatile (Turner 1991, table 35, p. 95). This article focuses on the implicit reasoning that reliable inferences can be drawn about the degree to which a flow tends to sustain itself at its current level, Stijn Claessens is with the Technical Department of the Europe and Central Asia and Middle East and North Africa Regions at the World Bank; Michael P. Dooley is with the Department of Economics at the University of California at Santa Cruz; and Andrew Warner is with Harvard University. At the time this article was written, Michael P. Dooley and Andrew Warner were with the International Economics Department at the World Bank. The authors would like to thank Guillermo Calvo, Maxwell Fry, Camp- bell Harvey, Ricardo Hausmann, participants in a Bank seminar, and the referees for their comments. © 1995 The International Bank for Reconstruction and Development/THE WORLD BANK 153 154 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 that is, the degree of persistence, solely on the basis of data categories given in balance of payments statistics. That is, the conventional view asserts that persis- tence can be inferred from labels. Here we ask whether, given direct statistical measures of persistence, labels can be drawn from persistence. In effect, our way of verifying this proposition is to turn it around and see if it also holds in the opposite direction. When only time-series statistics on persistence are used, can the label of the flow be identified? As we explain in section II, we are skeptical on theoretical grounds that such an inference is reliable; however, we think the issue is worth a systematic examination. Section I provides the motivation for the analysis. Section II discusses the modeling of capital flows. Section III discusses what evidence is relevant. Section IV provides the classifications for the various capital flows and the data sources. Section V analyzes simple univariate statistics for the various individual flows and looks at the predictability of these flows to answer the question: hot or cold? Section VI investigates the interactions between the various flows. It also looks at the predictive power of the overall capital flows and the interactions between the various flows and the overall capital account. Section VII is a summary. I. MOTIVATION The notion that inferences can be made about the characteristics of financial flows by just observing their labels is, of course, not new in economics. The flows-of-funds approach used by many central banks and others to analyze developments in the domestic economy embodies the implicit view that there is information in labels. This view has also long been an important part of the traditional analysis of international finance (see Nurske 1944 for a description of experiences in the 1920s). A distinction is often made between short-term and long-term capital flows: short-term capital movements are deemed speculative and reversible-hot money-and long-term capital flows-cold money-are based on fundamentals and are deemed reversible only when the fundamentals change. The fact that capital control programs in many countries distinguish between short- and long-term flows already points to the importance attached to this distinction. Capital account transactions are often also subject to policy inter- ventions that differ according to the type of flow. Withholding taxes are often levied on one kind but not another kind of capital flow. Subsidies often take the form of a government guarantee of private liabilities, favorable tax treatment on earnings (for example, on FDI), or access to special government facilities (for example, debt-equity swaps). Each of these distortions is designed to encourage or discourage a given type of capital transaction. In fact, the whole structure of balance of payments accounts reflects the implicit view that different types of capital flows have different economic implications. However, it is easy to be skeptical about the information value of the balance of payments labels for any purpose, not just for assessing persistence. Claessens, Dooley, and Warner 155 Increasingly, multiple financial instruments are available to finance any project, and so if a tight link ever existed between the financing method and the underlying nature of the project, it is probably becoming increasingly loose. A Treasury bond with a thirty-year maturity can be sold on the secondary mar- ket, and short-term assets can be continuously rolled over. Many observers seem to base their notion that short-term flows are more volatile on the fact that short-term-maturity inflows need to be repaid more quickly than long- term flows. Although rapid repayment may lead to higher volatility of gross short-term flows, it need not make net flows more volatile. Short-term flows that are rolled over are equivalent to long-term assets, and a disruption of gross FDI inflows, for example, can cause its net flow to be equivalent to a repayment of a short-term flow. In addition, the explicit label given to a flow may not cover its implicit nature. Dooley (forthcoming) argues, for example, that, although the inflows to developing countries in the 1970s were private capital flows in name, the universal government guarantees of both lenders and borrowers considerably subdued the discipline of the market. The flows, which helped generate the debt crisis of 1982, should have been considered official capital flows. The reasoning based on the label of the flow can nevertheless underpin substantive policy measures. Once a flow is identified as hot money, it is often seen to require some policy response. At various times countries (especially developing countries) have responded with exchange rate management, (ster- ilized) intervention, fiscal contraction, borrowing taxes, absolute foreign bor- rowing constraints, and reserve requirements (see, for example, Kiguel and Caprio 1993, Fischer and Reisen 1992, and Corbo and Hernandez 1993 for reviews of developing countries' responses to recent capital inflows). Many of these responses have differed depending on the "label" of the flow, thus presuming that labels are meaningful. II. MODELING CAPITAL FLOWS Research on international capital flows has differed on whether it is more accurate to treat the flows as exogenous (with respect to the country in question) or endogenous. In this study, although we do not think it vital to take a stand on this issue, we do clarify how the interpretation of our findings depends on this issue. If capital flows are exogenous from the point of view of the domestic economy, perhaps because they are driven by changes in international financial variables and market perceptions of the country, then the policymaker's con- cern about the volatility of capital flows can make good sense. Depending on the exchange rate policy being pursued by the country, volatile capital flows can translate into exchange rate volatility (in the case of a flexible exchange rate) or into variations in official reserves (in the case of a fixed or pegged exchange rate). Either consequence can be undesirable because it leads to 156 THE WORLD BANK ECONOMiC REVIEW, VOL. 9, NO. I temporary signals to shift resources between traded and nontraded sectors or because it requires monetary adjustments. If flows are exogenous, it would clearly be useful to know whether the data support the conventional view that certain kinds of flows are inherently more volatile and that certain flows can be predicted better. If capital flows are endogenous, however, an analysis of the behavior of capital flows in isolation makes little sense. Here, everything depends on the nature of the shock that gives rise to changes in the current account. The behavior over time of the flows would reflect the behavior over time of the underlying shocks. In the unlikely event that different flows have different ultimate causes and that the causes have different time-series properties, the flows themselves would have different time-series properties. But this seems to be a remote possibility. If capital flows are predominantly endogenous, there is no deep reason to expect any particularly tight relationship between types of flows and time-series properties. It may be argued that rather than taking an agnostic approach to the causality question, it would be better to present a model and try to identify the important causes, and then to use that framework to assess the question of persistence of financial flows. It has proven difficult, however, to develop such a structural model empirically with underlying sources of shocks. Capital flows in general, and perhaps even more so portfolio flows to developing countries, have been difficult to explain. Recent studies by Calvo, Leiderman, and Rein- hart (1993); Chuhan, Claessens, and Mamingi (1993); and Fernandez-Arias (1994) find low explanatory power, and the authors have difficulty identifying which factors exactly determine capital flows. One finding common to these three papers is that external factors, particularly the lower interest rates in the United States in the early 1990s, may have been important in motivating capital flows to developing countries. III. WHAT EVIDENCE Is RELEVANT? The view that labels convey information about persistence underlies laws about capital controls and is implicit in some academic research, but it is not expressed in a way that makes it obvious how to evaluate it empirically. Broadly speaking, to evaluate claims that certain kinds of flows are more volatile, we look at coefficients of variation; to evaluate claims about persistence, we look at measures of (positive) serial dependence and half-lives from impulse responses; and to evaluate claims about predictability, we look at time-series measures of forecasting performance. This evidence provides part of what we want; we also look at additional evidence on total capital flows and on how the flows interact with each other. Questions about the volatility of certain types of capital flows are (presumably) motivated by concerns about the volatility of the total capital account, not just about the volatility of one particular capital flow. Policy- makers after all wish to assess the likelihood of sudden and destabilizing changes in the total capital account, not just in its components. Claessens, Dooley, and Warner 157 To frame this point in a more concrete setting, suppose that a policymaker observes a rise in FDI during the latest quarter. Does this mean anything for the level of total capital flows? At an extreme, the FDI inflow can be merely a shift from, for example, long-term (bank) inflows to FDI inflows. In this case there would be an exact offset because the two flows would be perfectly negatively correlated, and the rise in FDI would carry no information about the level of total capital inflows and its volatility. An actual example of this phenomenon of substitution between various capi- tal flows was the rapid growth of holdings of deposits in offshore banks by U.S. residents in the 1970s. These deposits were attracted by higher yields offshore that were made possible by the absence of deposit interest rate ceilings, reserve requirements, charges for deposit insurance, and other factors. The capital out- flow in the U.S. balance of payments was matched at first by interbank loans from the branches to the U.S. head office. When these inflows were discour- aged, other forms of capital inflows took their place. Such inflows and outflows were offsetting and had little to do with an analysis of the U.S. balance of payments position. A similar effect occurred in the context of the Voluntary Restraint Program that the United States launched in February 1965, and the effect was particularly strong with respect to FDI (see, for example, Brimmer 1966 and Dooley 1981 and 1990). More generally, if a flow is a close substitute for other flows, it can be quite volatile, but this need not necessarily be a cause for concern, because other flows may be offsetting its volatility. Correspondingly, attributing volatility to a par- ticular flow can be misleading in the presence of substitution or complemen- tarity. The possibility of systematic interactions between components of the capital account needs thus to be addressed before making inferences from the parts to the whole. We examine these questions by first looking at correlations among the various flows. Then we ask a question that we think is central to assessing whether data on the components of capital flows can provide an early warning for future levels of capital flows: to what extent does the knowledge of the composition of the capital account improve the ability to forecast future levels of the capital account? This article analyzes data on components of capital flows in five industrial and five developing countries. It investigates whether volatility and persistence match up with categories of capital flows as expected and whether the data reveal systematic relationships among the flows, as well as the extent to which the available categorization of data provides useful information for forecasting total capital inflows. Because the article relies completely on time-series analysis, data requirements are limited. IV. DATA For our analysis, we classified balance of payments flow data according to the type of instrument within the country studied. We distinguished between 158 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I four categories of flows: FDI, portfolio equity, long-term (official and private), and short-term. The focus of the analysis was on net flows, because our concern was net financing. A list of the categories of data used is in table A-1. (Note that we took the labels as they are given in IMF, various years, and did not make any data corrections.) From the same source, we used data on quarterly changes in claims and liabilities to investigate the short-run time- series behavior of various flows. All flows are in millions of U.S. dollars deflated by the U.S. producer price index to convert them into real (1987) dollars. A second distinction we made is by transactor, that is, foreign direct investors (FDI plus other long-term flows), banks, government, and the private sector. The results of using this distinction are provided in Claessens, Dooley, and Warner (1993).1 In addition, countries could be concerned with the occurrence of rare but large, sudden movements in capital flows, such as those that occurred in 1982 in many developing countries with the start of the debt crisis and in 1992-93 in some industrial countries with the breakdown of the European Exchange Rate Mechanism. This concern, and the possible lumpiness of quarterly data, would suggest that annual data should also be analyzed. Given the rare occurrence of such crises and the limited annual data, however, analysis of time series is of little use. Instead, case studies are more appropriate for investigating such events. We selected five industrial countries (France, Germany, Japan, the United Kingdom, and the United States) and five developing countries (Argentina, Bra- zil, Indonesia, the Republic of Korea, and Mexico). The five industrial countries are the largest economies in the world, and flows to and from these countries represent the majority of capital flows between industrial countries. The five developing countries are among the developing countries that have received the largest share of private capital flows in recent years, and they represent very different country circumstances and institutional backgrounds. Such a choice has given us a broad selection of country circumstances on which to make some generalizations. 1. A third classification we could have made would have been by source (type of creditor). For developing countries, the source can be determined by using the World Bank Debtor Reporting System (DRS). One source distinction could, for example, be between official (bilateral and multilateral) and commercial sources (further distinguished, if desired, by destination, public, publicly guaranteed, and privately nonguaranteed). Adding the IMF balance of payments flows (short-term, FDI, and equity) to the breakdown of the DRS gives a more complete picture of the sources of external financing. But using DRS data (in addition to IMF data), instead of IMF data exclusively, has three drawbacks: (1) DRs data are reported annually, not quarterly; (2) the DRS does not cover industrial countries; and (3) DRS and IMF data can differ (greatly) for a given developing country (because of different data sources, conceptual prob- lems, and capital flows not recorded in IMF data, such as capital flight). Nevertheless, the advantages of being able to distinguish long-term flows by source would likely be considerable, given the different objectives of official and commercial creditors. Claessens, Dooley, and Warner 159 V. HOT OR COLD? Table 1 provides means, standard deviations, and coefficients of variation (cvs) for various kinds of flows, broken down by type.2 To provide an indica- tion of the relative magnitude of these flows compared with the total capital account, the third column of table 1 presents the average for the flows as shares of total financing. Long-term flows are the most important for all countries except the United States and Japan, where, respectively, short-term flows and portfolio equity flows are more important. There does not appear to be a systematic pattern in the volatility (as measured by the cv) of various types of flows across countries. Long-term flows have the highest cv for four countries; FDI for four countries; and portfolio equity flows for two countries. Perhaps surprising to those claiming that short-term flows are hot is the fact that short- term flows have the lowest cv in seven countries. Note also that the volatility of the total capital account is often less than that of a component. High relative volatility is one of the notions that has been associated with hot money. A related notion is that a hot-money inflow is likely to disappear or reverse itself in the near future, whereas a cold-money inflow is more likely to persist. Degree of persistence and level of volatility are two complementary measures: hot flows are associated with low persistence and high volatility. Figure 1 provides data on net capital flows in Japan (in millions of 1987 dollars, positive figures denoting inflows), by type of flow. These data provide the best corroboration we have found for conventional ideas about the persistence of various kinds of flows. Figure 1 shows that the FDI and portfolio equity flows display much less volatility over short periods than do the short-term flows and that the long-term flows are somewhere in between. One efficient way to summarize the idea of persistence that is apparent in figure 1 is to calculate autocorrelations for each type of capital inflow. A persis- tent series will be positively autocorrelated, whereas a transitory series will have a low or negative autocorrelation. In general, the classic case of a cold-money flow would be a flow that is highly positively autocorrelated, whereas a hot- money flow would exhibit zero or even negative autocorrelations. Referring back to figure 1, we would expect the FDI flows for Japan to have large positive autocorrelations and the short-term flows to exhibit far lower or even negative autocorrelations. The autocorrelations for Japan in figure 2 conform to these expectations, given the time-series plots. Note that FDI and portfolio equity have high positive autocorrelations. In contrast, the short-term flows exhibit negative autocorrela- tions, and the signs change from quarter to quarter. The long-term flows are only positively correlated at short horizons. The main finding for the other countries is that, if anything, the conventional pattern exhibited by the Japanese data is the exception rather than the rule. In 2. The detailed results for the categorization of flows by transactor are available in Claessens, Dooley, and Warner (1993). 160 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Table 1. Basic Statistics on Capital Flows, by Country Standard Average Coefficient Mean deviation share in total of variation Country, period, (millions of (millions of financing (CV) and type offlow 1987 dollars) 1987 dollars) (percent) (percent) Argentina, 1976:1-1992:1 Foreign direct investment 21 237 1.8 1,123 Portfolio equity 94 444 -2.5 473 Long-term -896 1,829 100.0 204 Short-term 747 1,158 0.7 155 Short- and long-term -149 1,549 100.7 1,040 Total -34.05 1,456.47 100.0 4,278 Brazil, 1975:1-1991:4 Foreigndirectinvestment -55 309 -l.o 562 Portfolio equity -6 99 0.3 1,611 Long-term -1,056 4,961 78.0 470 Short-term 3,000 3,080 22.7 103 Short- and long-term 1,944 3,052 100.7 157 Total 1,883 3,043 100.0 162 France, 1978:1-1992:3 Foreign direct investment - - - Portfolio equity - - - Long-term -534 5,569 88.1 1,043 Short-term -1,170 2,704 11.9 231 Short- and long-term -1,196 s,020 100.0 420 Total - - - Germany, 1979:1-1992:1 Foreign direct investment -376 755 2.1 201 Portfolio equity 2,140 3,169 -7.9 148 Long-term -2,761 9,642 70.0 349 Short-term -5,866 4,067 35.9 69 Short- and long-term -8,627 9,8s3 105.8 114 Total -6,864 9,752 100.0 142 Indonesia, 1976:1-1992:1 Foreign direct investment 3 100 0.7 3,719 Portfolio equity 15 68 1.3 454 Long-term -560 1,438 70.7 257 Short-term 1,062 342 27.3 32 Short- and long-term 501 1,496 98.0 298 Total 519 1,519 100.0 293 Japan, 1979:1-1992:1 Foreign direct investment -1,261 1,234 5.9 98 Portfolio equity -8,451 10,282 66.0 122 Long-term -125 11,280 21.9 9,007 Short-term -959 4,506 6.1 470 Short-andlong-term 355 10,941 28.1 3,081 Total -10,854 13,407 100.0 124 Korea, 1976:1-1992:1 Foreign direct investment -1 159 2.8 23,366 Portfolio equity 63 189 2.7 300 Long-term -757 2,271 92.4 300 Short-term 669 745 2.0 111 Short- and long-term -88 2,187 94.5 2,490 Total -25 2,304 100.0 9,036 Mexico, 1975:1-1992:1 Foreign direct investment 79 224 -1.3 283 Portfolio equity -2 873 -2.7 53,811 Long-term -358 3,783 92.0 1,057 Short-term 1,348 1,186 11.9 88 Short- and long-term 990 3,599 103.9 363 Total 1,068 3,340 100.0 313 Claessens, Dooley, and Warner 161 Standard Average Coefficient Mean deviation share in total of variation Country, period, (millions of (millions of financing (CV) and type offlow 1987 dollars) 1987 dollars) (percent) (percent) United Kingdom, 1973:1-1992:1 Foreign direct investment -2,041 2,302 2.8 113 Portfolio equity 413 969 -4.6 235 Long-term 1,075 8,927 102.2 830 Short-term -378 1,612 -0.4 426 Short- and long-term 697 8,688 101.8 1,247 Total -931 8,014 100.0 860 United States, 1973:1-1992:1 Foreign direct investment -478 5,308 13.0 1,110 Portfolio equity -2,046 3,045 0.4 149 Long-term 3,635 11,169 40.7 307 Short-term 5,897 11,929 45.9 202 Short- and long-term 9,532 16,886 86.7 177 Total 7,007 20,118 100.0 287 - Not available. Note: The statistics are based on quarterly data deflated by the U.S. producer price index. Source: IMF (various years) and authors' calculations. many other countries the conventional pattern simply breaks down. Figures 3 and 4 show this for Germany and Mexico. For Germany, for example, FDI appears at least as volatile as short-term flows, and long-term flows appear the most stable for Germany. And for Mexico, when broken down by transactor, flows to the banking system appear more stable than flows to the government (as well as to the private sector). The autocorrelations for Germany (figure 5) confirm that FDI flows are the least stable and long-term flows the most stable. For Mexico (figure 6), the government and private sector flows have indeed the lowest (for some lags even negative) autocorrelations. Another way to summarize the evidence on persistence for all countries is to compute half-lives from impulse response functions. To do this, we estimated a univariate fourth-order autoregressive or AR(4) model for a given flow and then examined how a given shock to the error term in the estimated equation propa- gated itself through time.3 If a time series is highly positively autocorrelated, it will take a long time for a shock to die out; if the autocorrelations are low, the shock should vanish quickly. The half-life in this context is simply the number of quarters it takes for the shock to lose half or more of its initial value. The results are reported in table 2. 3. We chose an AR(4) model because we used quarterly flows and we wanted to account for possible seasonal effects. We used four lags because we thought it was important to at least exceed one year in the lag length and because experiments with longer lags did not alter the results importantly. 162 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Figure 1. Net Capital Flows by Type, Japan, 1977-91 (millions of 1987 dollars) A. Foreign direct investment B. Long-term flows Value Value 0 40,000 -800 30,000 -1,600- 20,000 -2,400- 10,000 -3,200- 0 0 V V -4,000- -10,000 -4,800 - -20,000 -5,600- ... .. .... . ......" ' " -30,000 .. .. ..1" .. 1977 79 81 83 85 87 89 91 1977 79 81 83 85 87 89 91 C. Portfolio equity D. Short-term flows Value Value 5,000 0 _ Ao 12,000 -5,000 8,000 -10,000 4,000 -15,000 0 - VAP V jV -20,000 -4,000 -25,000, -8,000 -30,000 -12,000 -35,000 . .. ....... 1977 79 81 83 85 87 89 91 1979 81 83 85 87 89 91 Note: Positive values denote inflows. Source: IMF (various years) and authors' calculations. Claessens, Dooley, and Warner 163 Figure 2. Autocorrelations of Net Capital Flows by Type, Japan A. Foreign direct investment B. Long-term flows Autocorrelation Autocorrelation 1.0 . 0.8 02 0° -0.6 0.6 -0.4 -0.4- -0.8 -0.8 -1.0 1-0 ,, I ,, I I I I . . . -.0 . . . . . ,,'' 1 2 3 4 5 6 7 8 9 10111213141516 1 2 3 4 5 6 78 8 9 10 112131415 16 Lag (quarters) Lag (quarters) C. Portfolio equity D. Short-term flows Autocorrelation Autocorrelation 1.0 - ~~~~~~~~~~~~1.0 0.8 - 0.8 0.6 0.6 040.4 0.202 0 -0.2 -0.2 -0.4 - -0.4 - -0.6 --0.6 -0.8 -0.8 -1.0 - -1.0- ,11111 1 2 3 456 7 8 10111213141516 1 2 3 4 5 6 7 8 9 10111213141516 Lag (quarters) Lag (quarters) Source: Authors' calculations. 164 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Figure 3. Net Capital Flows by Type, Germany, 1973-91 (millions of 1987 dollars) A. Foreign direct investment B. Long-term flows Value Value 2,000 15,000 1,000 10,000 -25,000 0 4 vv 0 A -1,000 - 5,000 q -10,000, -2,000 --15,000 -3,000 --20,000 -25,000 -4,000o ......... I7 I -30,000 . 1973 76 79 82 85 88 91 1973 76 79 82 85 88 91 C. Portfolio equity D. Short-term flows Value Value 16,000 - 4,000 12,000] 0 8,000 --4,000- 4,000 -8,000 0 r ~~~~~~vtl -12,000 -4,000 _ -16,000 - -8,000 ..,'.*..'.............. -20,000 . i... . ..'.I .. -.. 1I. 1973 76 79 82 85 88 91 1973 76 79 82 85 88 91 Note: Positive values denote inflows. Source: IMF (various years) and authors' calculations. Claessens, Dooley, and Warner 165 Figure 4. Net Capital Flows by Transactor, Mexico, 1975-91 (millions of 1987 dollars) A. Foreign direct investment B. Government Value Value 1,400 12,000 1,200. 1,000 8,000 800400 400 R ' 200 4,000 -200 -8,0_ _ 4;0 0_ _ 1975 77 79 81 83 85 87 89 91 1975 77 79 81 83 85 87 89 91 C. Banks D. Private sector Value Value 10,000 6,000- 8,000 4,000 - 6,000 2,000 4,000~~~~~~~~~~~~2,0 2,000 a 200 -2,000 - 2,000 -4,000 -4,000 |6,00 .... .., ...... |l-6 oo-. ...... ..,...,!.."". "" 1975 77 79 81 83 85 87 89 91 1975 77 79 81 83 85 87 89 91 Note: Positive values denote inflows. Source: IMF (various years) and authors' calculations. 166 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Figure 5. Autocorrelations of Net Capital Flows by Type, Germany A. Foreign direct investment B. Long-term flows Autocorrelation Autocorrelation 1.0- 1 0 0.8- o.6' . -0.4 -0.4 -0.2 - -0.2 -0.4 -. -0.6 --. -0.8 -0.8 - 1.0- -1.0 I I I II I I I 1 2 3 4 5 6 7 8 9 10111213141516 1 2 3 4 5 6 7 8 9 10111213141516 Lag (quarters) Lag (quarters) C. Portfolio equity D. Short-term flows Autocorrelation Autocorrelation 1.0 - 1.0 0.8 - 0.8- o.6 - 0.6 -0.4 -0.4 -o.6 -0.6 -0.8 -0.8 -1.0 1, I -1.0- 1 2 3 4 5 6 7 8 9 10111213141516 1 -2 3 4 5 6 7 8 9 101112131415 16 Lag (quarters) Lag (quarters) Source: Authors' calculations. Claessens, Dooley, and Warner 167 Figure 6. Autocorrelations of Net Capital Flows by Transactor, Mexico A. Foreign direct investment B. Government Autocorrelation Autocorrelation 1.0 - 1.0 _ 0.8 0.8 o.6 o.6 0.4 Z7 :047 -0.4 -0.4 -0.6 -0.2 -0.48 -0.48 -0.8 - 1.0 I - 1.0 I 1 2 3 4 5 6 7 8 9 10111213141516 1 2 3 4 5 6 7 8 9 1111213 141516 Lag (quarters) Lag (quarters) C. Banks D. Private sector Autocorrelation Autocorrelation 1.0 1.0 0.8 0.8 0.6 0 o.6 0.4 0.4 0. 0.2~ R 0 0 -0.2 -0.2 -0.4 --0.4 -0.6 -0.6 -0.8 -0.8 - 1.0 I I I I I I I I I I I f I I I -1.0 I I I I I I I I I I I T I I 1 2 3 4 5 6 7 8 910111213141516 1 2 3 4 5 6 7 8 9 10111213141516) Lag (quarters) Lag (quarters) Source: Authors' calculations. 168 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Table 2 provides little support for those who wish to infer persistence from labels. With the exception of Japan, most of the half-lives are 1: that is, more than 50 percent of the shock has dissipated before even one quarter has elapsed. This is true for the breakdown of flows by type as well as by transactor (not reported). It basically reflects the fact that the lag coefficients in the estimated autoregressive equations are small. In addition, there is little evidence, apart from the case of Japan, that the allegedly persistent flows- such as FDI and long-term flows-exhibit more memory than the other flows. So far, we have examined the question of whether persistence-as measured by autocorrelations and half-lives-matches up with the categories, and we have found that often there is not a close correspondence. A different but related question is whether the flows are forecastable solely on the basis of their own past and whether the forecasts match up with the categories in the expected way. The key point here is that a flow can be predictable without necessarily being significantly positively autocorrelated or having a high half- life. As long as a flow follows some pattern, not necessarily that of positive serial correlation, it can be predictable. It is possible that instead of making statements about half-lives or autocorrelations, conventional wisdom may sim- ply be saying that long-term flows are more predictable than short-term flows. The test we employ on predictability is a simple measure of the goodness of predictive power: the residual mean square error (RMSE). We estimated again, starting from 1982, a univariate AR(4) model for all flows and then performed out-of-sample forecasts for the next four quarters on the level of the flows. Using the new data, we updated the AR(4) model each year and performed another out-of-sample forecast-for the next year-repeating this procedure for each year. We then standardized the out-of-sample forecasting RMSE with Table 2. Half-Lives from Impulse Response Functions (in quarters) Foreign direct Portfolio Long-term Short-term Country investment equity flows flows Argentina - 1 - 1 Brazil 1 3 la 3 France - 1 1 Germany la 1 la 1 Indonesia - 1a la Japan 9a 1 3 1 Korea, Rep. of lb la 2 Pa Mexico lb 1 - 2 United Kingdom 2a 1 la 1 United States 1 1 1 1 - Not available. Note: The half-lives measure the number of quarters it takes before the impulse response is half or less of the initial shock. For countries with incomplete data, we analyzed liabilities or claims separately. a. Based on data from liabilities side only. b. Based on data from claims side only. Source: Authors' calculations. Claessens, Dooley, and Warner 169 Table 3. Relative Ability to Predict Individual Categories of Flows (ratios of residual mean square errors) Foreign direct Portfolio Long-term Short-term Country investment equity flows flows Argentina - - 1.11 1.00 Brazil 0.97 1.24 1.02 1.00 France - - 1.35 1.00 Germany 1.40 1.31 1.13 1.00 Indonesia - 1.03 0.77 1.00 Japan 0.76 0.89 1.06 1.00 Korea, Rep. of 0.91 0.95 0.91 1.00 Mexico 1.16 1.23 1.15 1.00 United Kingdom 1.18 1.17 1.05 1.00 United States 0.86 0.98 0.83 1.00 -Not available. Note: We first estimated AR(4) univariate forecasting equations for each of the four categories. Then we computed all possible four-step-ahead dynamic forecast errors from these equations and calculated residual mean square errors (RMSES). These were then divided by the standard deviation of the time series to obtain a unit-free measure of forecasting performance. These normalized RMSES were then divided by the normalized RMSE for the last category, short-term flows, to compare forecasting performance across categories. These ratios are reported above. A value over 1.0 indicates that the category exhibits poorer forecasts than the forecasts for short-term flows. Quarterly data were used for the forecast period frorn the first quarter of 1982 to the fourth quarter of 1991, except for France (from the first quarter of 1985 tto the fourth quarter of 1991) and Brazil (from the first quarter of 1982 to the third quarter of 1988). Source: Authors' calculations. the standard deviation of the respective flow to get a measure of the relative ability to forecast the various flows. Finally, we compared the ability to forecast: the various flows with the ability to forecast short-term flows. Short-term flows are commonly assumed to be the most volatile and least predictable type of flow, and we would thus expect the other flows to be more predictable than short- term flows. The evidence on this issue (table 3) is that, compared with the benchmarkc (short-term flows), other flows cannot systematically be predicted more accu- rately. For about half of the countries, our forecasts for the other flows were actually worse than the forecast for short-term flows. The two countries for which we were able to predict all other flows better were the United States and the Republic of Korea. Altogether, only about half of the other flows were more predictable than short-term flows. Compared with the autocorrelations and half-lives, the evidence on relative predictability using the time-series model showed that a low score on the first two measures generally translates into high unpredictability. VI. How Do THE FLOWS INTERACT? The evidence so far provides some basis for skepticism about gauging vol- atility and persistence by using only the labels given to the flows. As explained 170 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. I Table 4. Marginal Sources of Financing the Current Account (slope coefficients) Foreign direct Portfolio Long-term Short-term Country investment equity flows flows Argentina -0.03 -0.08 1.06*'* 0.06 Brazil -0.00 0.01 1.27** -0.27 France - - 0.98*'* 0.02 Germany -0.02* 0.02 0.83** 0.17** Indonesia 0.01 0.01 0.99** -0.01 Japan -0.03** 0.17 0.86t* 0.00 Korea, Rep. of -0.02 0.00 1.28` -0.26* Mexico -0.00 -0.10** 1.07* 0.03 UnitedKingdom 0.03 -0.02* 0.99** 0.00 United States 0.11 -0.01 0.42 * 0.49 * - Not available. Significant between the 5 and 10 percent level. * Significant at the 5 percent level or better. Note: Slope coefficients for the regressions of the quarter-to-quarter change in the level of a particular flow on the quarter-to-quarter change in the level of the total capital account. Source: Authors' calculations. before, there may be some offsets between various flows. It would thus clearly help to know whether there are systematic correlations in the data along these lines. We started by calculating the simple correlation matrixes between the various categories of flows for all countries (not reported). The correlations showed some degree of substitution (that is, negative correlations) between most flows for almost all countries. For some countries, the substitution was very pronounced, with high negative correlation between short- and long-term flows. We next performed an analysis on the marginal source of financing the current account.4 We ran regressions of the changes in the various types of flows on the change in the total capital account. Slope coefficients provide a measure of the degree to which a particular flow "finances" at the margin the country's overall financing requirements or surplus (under the assumption that the current ac- count movements drive capital flows). Table 4 provides the slope coefficients. Long-term flows appear to be the most sensitive: for all countries except one the slope coefficient for long-term flows is the highest. The lack of consistent rank- ing among the other flows across countries suggests that short-term flows do not differ on this measure from, for example, FDI. Our results differ from those of Turner (1991) and Fry (1993). Two facts may account for this: we used quarterly as opposed to annual data; and we 4. As mentioned before, this approach does not allow us to distinguish between the endogenous and exogenous natures of the various types of capital flows. Doing so would require a structural model that covered, among other things, shifts in investment opportunities. Nevertheless, our approach gives an indication of the empirical relation between the total capital account and its components. Claessens, Dooley, and Warner 171 took the total capital account as the right-hand variable, whereas the other two studies excluded the balance of official monetary movements (that is, official reserve movements) from the total capital account. The classification by transactor (not reported) showed that, except for the United States, flows to and from the government sector have been the most accommodative. Although it may not be surprising that the government sector is the most accommodative, it does raise the question of whether the government is the source of volatility in capital flows or whether it is accommodating to the volatility of other flows. The results so far suggest there is sufficient substitution between and interac- tions among the various flows to make an analysis of the time series of a single flow possibly misleading.5 This possibility can best be further addressed by investigating how well we can forecast the total capital account using past information. Table 5 presents the ratio (ratio 1) of the RMSE of the forecast of the total capital account using a time-series predictor, AR(4) regression, to the RMSE of a naive predictor (one with no change in the capital account). As can be observed, we can generally improve on the naive forecasts (by up to 34 percent), and this indicates that there is some information in the past time series for the aggregate capital account. Does knowledge of the breakdown category of flows convey any information about total capital flows? To analyze this, we forecasted the total capital account by using past information on the total capital account as well as information on the contemporaneous shares of the individual flows. We reasoned that if the total capital account is independent of a particular flow, then adding the con- temporaneous share of the flow should not affect our forecasting ability. Con- versely, if a flow helps determine the total capital account, then adding the contemporaneous share should help the forecast. Ratio 2 in table 5 presents this horse race in relation to the AR(4) time-series predictor. Using the shares as additional information did not greatly improve our forecasting ability. At most, it improved our ability to forecast net flows by 10 percent (Japan), and in most cases the gain is less than 3 percent. This result: provides evidence that, in general, movements in the overall capital account are little influenced by the type of capital flow. Because there is much substitution going on between the various flows, analyzing individual flows may not be very meaningful. It is thus better to focus attention on the determinants of the overall capital account; for example, the impact of the aggregate external shocks the economy is exposed to and the overall (macro-) economic policies the govern- ment pursues. 5. This analysis ignores any impact of the various flows on domestic savings or investment in this or future periods. 172 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Table S. Relative Ability to Predict Overall Flows (ratio) Country Ratio I Ratio 2 Argentina 0.85 0.99 Brazil 0.77 1.00 France - - Germany 0.94 1.00 Indonesia 0.67 0.97 Japan 0.84 0.90 Korea, Rep. of 0.96 1.00 Mexico 0.66 0.95 United Kingdom 0.92 0.92 United States 0.87 0.99 - Not available. Note: Ratio 1 is RMSE(f2)/RMSE(f1), where RMsE(fl) is the residual mean square error (RMSE) of a naive predictor (no change in the capital account) and RMSE(f2) is the RMSE of the forecast of the total capital account using a time-series predictor, AR(4) regression, without share variables. Ratio 2 is RMSE(f3)/RMSE(fl), where PMSE(f3) is the RMSE of the forecast of the total capital account using a time- series predictor, AR(4) regression, with all the contemporaneous share variables plus their two lags. The RMSEs are based on four-step-ahead forecast errors. Flows include claims and liabilities. Source: Authors' calculations. VII. SUMMARY Using time-series analysis of data for five industrial and five developing coun- tries, we find that in most cases the labels "short-term" and "long-term" in relation to capital flows do not provide any information about the time-series properties of the flows. Put differently, if only time-series statistics are available, it is not likely that the label of the flow can be identified. In particular, long-term flows are often as volatile as short-term flows, and the time it takes for an unexpected shock to a flow to die out is similar across flows. And, arguably the most relevant measure from a policymaker's point of view, short-term flows are at least as predictable as long-term flows. There is also little evidence that information about the composition of flows is useful in forecasting the overall level of flows, and this suggests that the overall capital account is independent of the type of flow. The implication is that the emphasis on analyzing specific flows, especially short-term portfolio flows, is overdone. We find virtually no time-series prop- erty that can be regarded as an inherent property of any particular kind of flow. An attempt to reduce capital account volatility by administratively limiting short-term inflows is unlikely to be effective because there is little evidence that these flows really are more volatile than other flows. The evidence here is consistent with the view that capital flows are fungible, highly substitutable, and endogenous with respect to external shocks and internal policies. Claessens, Dooley, and Warner 173 Table A-1. Data Classifications Item Line numbers Flow 1. Change in direct investment claims 45,46 Change in direct investment liabilities 49, 50 2. Change in bond claims 53, 56 Change in bond liabilities 54,55,57,58 3. Change in equities claims S9 Change in equities liabilities 60, 61 4. Change in bank loan claims 69-71,77-79 Change in bank loan liabilities 72-76,80-83 5. Change in bank deposit claims 89 Change in bank deposit liabilities 90-92 6. Change in other short-term claims 48,93,94 Change in other short-term liabilities 52, 95-97 7. Change in other long-term claims 47 Change in other long-term liabilities 51 8. Change in long-term official claims (non reserve) 62-64 Change in long-term official liabilities (non reserve) 65-68 9. Change in short-term official claims 84, 85 Change in short-term official liabilities 86-88 10. Change in official reserves 98-111 11. Errors and omissions 112 Classification Foreign direct investment 1 Long-term flows 2, 4, 7, 8, 10 Short-term flows 5, 6,9 Portfolio equity 3 Note: Flow items and line numbers refer to those in IMP (various years); line numbers for the items under Classification refer to the numbered items under Flow. Source: IMF (various years) and authors' classifications. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Brimmer, Andrew. 1966. "Direct Investment and Corporate Adjustment Techniques under the Voluntary U.S. Balance of Payments Program." Journal of Finance 21(2):266-82. Calvo, Guillermo, Leonardo Leiderman, and Carmen Reinhart. 1993. "Capital Inflows and the Real Exchange Rate Appreciation in Latin America: The Role of External Factors.' IMFStaff Papers 40(1, March):108-51. Chuhan, Punam, Stiin Claessens, and Nlandu Mamingi. 1993. "Equity and Bond Flows to Latin America and Asia: The Role of Global and Country Factors. Working Paper 1160. World Bank, International Economics Department, Washington, D.C. Processed. 174 THE WORLD BANK ECONOMIC REVIEW, VOL. 9, NO. 1 Claessens, Stijn, Michael P. Dooley, and Andrew Warner. 1993. "Portfolio Capital Flows: Hot or Cool." In Stijn Claessens and Sudarshan Gooptu, eds., Portfolio Invest- ment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. Corbo, Vittorio, and Leonardo Hernandez. 1993. "Macroeconomic Adjustment to Port- folio Capital Inflows: Rationale and Recent Experience." In Stijn Claessens and Sud- arshan Gooptu, eds., Portfolio Investment in Developing Countries. World Bank Dis- cussion Paper 228. Washington, D.C. Dooley, Michael P. 1981. "Implications of the Internationalization of Banking for the Definition and Measurement of U.S. Credit and Monetary Aggregates." International Finance Discussion Paper 177. Board of Governors of the Federal Reserve System, Washington, D.C. Processed. . 1990. "Comment." In Assaf Razin and Joel Slemrod, eds., Taxation and the Global Economy. Chicago, Ill.: University of Chicago Press for the National Bureau for Economic Research. . Forthcoming. "A Retrospective on the Debt Crisis." In Peter Kenen, ed., Under- standing Interdependence: The Macroeconomics of an Open Economy. Princeton, N.J.: Princeton University Press. Fernandez-Arias, Eduardo. 1994. "The New Wave of Private Capital Inflows: Push or Pull?" Working Paper 1312. World Bank, International Economics Department, Wash- ington, D.C. Processed. Fischer, Bernard, and Helmut Reisen. 1992. Towards Capital Account Convertibility. OECD Development Centre Policy Brief No. 4. Paris: Organization for Economic Cooperation and Development. Processed. Fry, Maxwell. 1993. "Foreign Direct Investment in a Macroeconomic Framework: Fi- nance, Efficiency, Incentives, and Distortions." Working Paper 1141. World Bank, International Economics Department, Washington, D.C. Processed. Gooptu, Sudarshan. 1993. "Portfolio Investment Flows to Developing Countries." In Stijn Claessens and Sudarshan Gooptu, eds., Portfolio Investment in Developing Countries. World Bank Discussion Paper 228. Washington, D.C. IMF (International Monetary Fund). Various years. Balance of Payments Yearbook. Washington, D.C. Kiguel, Miguel, and Gerard Caprio. 1993. "Capital Inflows-Hot or Cool?" Outreach 8. World Bank, Policy and Research Department, Washington, D.C. Processed. Nurske, Ragnar. 1944. International Currency Experiences. Geneva: League of Nations. Nunnenkamp, Peter. 1993. "The Return of Foreign Capital to Latin America." Kiel Working Paper 574. The Kiel Institute of World Economics, Kiel, Germany. Processed. Reisen, Helmut. 1993. "The Case for Sterilized Intervention in Latin America." Paper presented at the 6th Annual Inter-American Seminar on Economics, May 28-29, Caracas, Venezuela. Paris: Organization for Economic Cooperation and Development (OECD), Development Centre. Processed. Turner, Philip. 1991. "Capital Flows in the 1980s: A Survey of Major Trends." BIS Economic Papers 30. Basel: Bank for International Settlements, Monetary and Eco- nomic Department. Processed. World Bank 'i New Publications j World Debt Tables 1994-95: External Finance for Developing Countries A complete and up-to-date presentation of World Population Projections 1994-9;5: data and a review of development on the Estimates and Projections with Related external debt of and financial flows to develop- Demographic Statistics ing countries. Volume I--Analysis and Summary Contains the most recent demographic Tables. 248 pages. Stock No. 12912. $21.95. statistics on the world's population, with esti- Volume II--Country Tables. 560 pages. Not mates for 206 countries and territories projected available separately. Two-volume set--Stock as far as the year 2150. This edition incorporates No. 12914. $150.00. census and fertility-mortality survey results that were unavailable for the previous edition. It Forthcoming: contains a new subregion, Central Asia, to World Debt Tables 1994-985: Data on categorize the southernmost former Soviet Diskette. Stock No. 12915. $95.00 republics. 532 pages. Stock No. 44947. $34.95. Averting the Old Age Crisis: Policies to Protect the Old and Promote Growth The first comprehensive and global examina- Voting for Reform: Democracy, Political tion of old age security. Systems providing Liberalization, and Economic Adjustment financial security for the old are under growing Drawing on case studies from 8 countries, strain, and changing demographics have led to this volume assesses the interaction between two an increase in the number of old people. global trends: political liberalization and market- Economists, policymakers, and government oriented reform. The study shows that although officials will find valuable evaluations of policy there is no necessary contradiction between the options in this detailed report, along with two trends, the success of efforts to consolidate critical analyses of the issues affecting income both democracy and reform depends on appropri- sucurity for the elderly in both developing and ate political institutions and strategies. 536 pages. industrial nations. 630 pages. Stock No. 60996. Stock No. 60987. $24.95. $19.95. World Bank Atlas 1995 An easy-to-use collection of color maps, charts, and graphs representing key social, economic, and environmental indicators for 209 To ordler, please countries and territories. Organized under three development themes: The Poeple, The fill O(lit the CoIIpoI that follows. Economy, The Environment. 36 pages. Stock No. 12922. $7.95. World Bank Publications Order Coupon CUSTOMERS IN THE UN1TED STATES: CUSTOMERS OUTSIDE THE UNITED STATES Complete this coupon and return to Contact your local World Bank Publications distributc The World Bank for information on prices in local currency and payme Box 7247-8619 terms. (See next page for a complete list of distributoi Philadelphia, PA 19170-8619 If no distributor is listed for your country, use this ord( U.S.A. form and return it to the U.S. address. Orders that ar Charge by credit card by calling (202) 473-1155 or sent to the U.S. address from countries with distrib fax this completed order coupon to (202) 676-0581. tors will be returned to the customer. Quantity Title Stock Number Price Total Subtotal US$ * If a purchase order is used, actual postage will Postage and handling* US$ be charged. If payment is by check or credit card, postage and handling charges are Total US$ US$5.00 per order. For air mail delivery outside the U.S., include US$8.00 for the first item and US$6.00 for each additional item. CHECK METHOD OF PAYMENT U Enclosed is my check payable to The World Bank. O Charge my O VISA O MasterCard O American Express Credit card account number Expiration date Signature O Bill me. (Institutional customers only. Purchase order must be included.) PLEASE PRINT CLEARLY Name Firm Address City State Postal code Country Telephone 1098 Distributors of World Bank Publications Prices and terms vary by country ARGENTINA COLOMBIA GREECE Carlos Hirsch, SRL Infoenlace Ltda. Papasotiriou S.A. Galeria Guemes Apartado Aereo 34270 35, Stournara Str. Florida 165,4th Floor-Ofc. 453/ Bogota D.E. 106 82 Athens 465 1333 Buenos Aires COTE D'VOIRE HONG KONG, MACAO Centre d'Edition et de Asia 2000 Ltd. Oficina del Libro Internacional Diffusion 46-48 Wyndham Street Alberti 40 Africaines (CEDA) Winning Centre 1082 Buenos Aires 04 B.P. 541 7th Floor Abidjan 04 Plateau Central Hong Kong AUSTRALIA, PAPUA NEW GUINEA, CYPRUS HUNGARY FIJI, SOLOMON ISLANDS, Center of Applied Research Foundation for Market VANUATU, AND Cyprus College Economy WESTERN SAMOA 6, Diogenes Street, Engomi Dombovari Ut 17-19 D.A. Information Services P.O. Box 2006 H-1117 Budapest 648 Whitehorse Road Nicosia Mitcham 3132 INDIA Victoria CZECH REPUBLIC Allied Publishers Private Ltd. National Information Center 751 Mount Road AUSTRIA PO Box 668 Madras - 600 002 Gerold and Co. CS-113 57 Prague 1 Graben 31 INDONESIA A-1011 Wien DENMARK Pt. Indira Limited SamftmdsLitteratur Jalan Borobudur 20 BANGLADESH Rosenoerns Alle 11 P.O. Box 181 Micro Industries Development DK-1970 Frederiksberg C Jakarta 10320 Assistance Society (MIDAS) House 5, Road 16 DOMINICAN REPUBLIC IRAN Dhanmondi R/Area Editora Taller, C. por A. Kowkab Publishers Dhaka 1209 Restauraci6n e Isabel la P.O. Box 19575-511 Cat6lica 309 Tehran BELGIUM Apartado de Correos 2190 Z-1 Jean De Lannoy Santo Domingo IRELAND Av. du Roi 202 Government Supplies Agency 1060 Brussels EGYPT, ARAB REPUBLIC OF 4-5 Harcourt Road Al Ahram Dublin 2 BRAZIL Al Galaa Street Publicacoes Tecnical Cairo ISRAEL Intemacionais Ltda. Yozmot Literature Ltd. Rua Pexoto Gomide, 209 The Middle East Observer P.O. Box 56055 01409 Sao Paulo, SP 41, Sherif Street Tel Aviv 61560 Cairo CANADA R.O.Y. International Le Diffuseur FINLAND P.O. Box 56055 151A Boul. de Mortagne Akateeminen Kirjakauppa Tel Aviv 61560 Boucherville, Quebec P.O. Box 128 J4B 5E6 SF-00101 Helsinki 10 ITALY Licosa Commissionaria Renouf Publishing Co. FRANCE Sansoni SPA 1294 Algoma Road World Bank Publications Via Duca Di Calabria, 1/1 Ottawa, Ontario K1B 3W8 66, avenue d'Iena Casella Postale 552 75116 Paris 50125 Firenze CHINA China Financial & Economic GERMANY JAMAICA Publishing House UNO-Verlag Ian Randle Publishers Ltd. 8, Da Fo Si Dong Jie Poppelsdorfer Allee 55 206 Old Hope Road Beijing 53115 Bonn Kingston 6 JAPAN PERU SRI LANKA & THE MALDII Eastem Book Service Editorial Desarrollo SA Lake House Bookshop Hongo 3-Chome, Bunkyo-ku 113 Apartado 3824 P.O. Box 244 Tokyo Lima 1 100, Sir Chittampalam A. Gardiner Mawatha KENYA PHILIPPINES Colombo 2 Africa Book Service (E.A.) Ltd. International Book Center Quaran House, Mfangano St. Suite 1703, Cityland 10 SWEDEN P.O. Box 45245 Condominium Tower 1 Fritzes Fackboksforetaget Nairobi Ayala Avenue, H.V. dela Regeringsgatan 12, Box 16356 Costa Extension S-106 47 Stockholm KOREA, REPUBLIC OF Makati, Metro Manila Pan Korea Book Corporation SWITZERLAND P.O. Box 101, Kwangwhamun POLAND Librairie Payot Seoul International Publishing Service Case postale 3212 Ul. Piekna 31/37 CH 1002 Lausanne Korean Stock Book Centre 00-677 Warszawa P.O. Box 34 Van Dierman Editions Techniq Yeoeida PORTUGAL P.O. Box 465 Seoul Livraria Portugal CH 1211 Geneva 16 Rua Do Carmo 70-74 MALAYSIA 1200 Lisbon TANZANIA University of Malaya Cooperative Oxford University Press Bookshop, Limited SAUDI ARABIA, QATAR Maktaba Street P.O. Box 1127, Jalan PantaiBaru Jarir Book Store P.O. Box 5299 59700 Kuala Lumpur P.O. Box 3196 Dar es Salaam Riyadh 11471 MEXICO THAILAND INFOTEC SINGAPORE, TAIWAN, Central Department Store Apartado Postal 22-860 MYANMAR,BRUNEI 306 Silom Road 14060 Tlalpan, Mexico D.F. Gower Asia Pacific Pte Ltd. Bangkok Golden Wheel Building NETHERLANDS 41, Kallang Pudding, #04-03 TRINIDAD & TOBAGO, De Lindeboom/InOr-Publikaties Singapore 1334 Systematics Studies Unit P.O. Box 202 #9 Watts Street 7480 AE Haaksbergen SLOVAK REPUBLIC Curepe Slovart G.T.G. Ltd. Trinidad, West Indies NEW ZEALAND Krupinska 4 EBSCO NZ Ltd. P.O. Box 152 UGANDA Private Mail Bag 99914 852 99 Bratislava 5 Gustro Ltd. New Market 1st Floor, Room 4, Geogiadis Auckland SOUTH AFRICA, BOTSWANA Chambers For single titles: P.O. Box 9997 Oxford University Press Plot (69) Kampala Road NIGERIA Southern Africa Kampala University Press Limited P.O. Box 1141 Three Crowns Building Jericho Cape Town 8000 UNITED KINGDOM Private Mail Bag 5095 Microinfo Ltd. Ibadan For subscription orders: P.O. Box 3 International Subscription Service Alton, Hampshire GU34 2PG NORWAY P.O. Box 41095 England Narvesen Information Center Craighall Book Department Johannesburg 2024 ZAMBIA P.O. Box 6125 Etterstad University Bookshop N-0602 Oslo 6 SPAIN Great East Road Campus Mundi-Prensa Libros, S.A. P.O. Box 32379 PAKISTAN Castello 37 Lusaka Mirza Book Agency 28001 Madrid 65, Shahrah-e-Quaid-e-Azam ZIMBABWE P.O. Box No. 729 Libreria Intemacional AEDOS Longman Zimbabwe (Pte.) Ltd Lahore 54000 Consell de Cent, 391 P.O. Box ST 125 08009 Barcelona Southerton Harare Coming in the next issue of THE WORLD BANK ECONOMIC REVIEW May 1995 Volume 9, Number 2 * The Role of Infrastructure in Mexican Economic Reform by Andrew Feltenstein and liming Ha * The Current Account in Developing Countries: A Perspective from the Consumption-Smoothing Approach by Atish R. Ghosh and Jonathan D. Ostry * An Eclectic Approach to Estimating the Determinants of Achievement in Jamaican Primary Education by Paul Glewwe, Margaret Grosh, Hanan Jacoby, and Marlaine Lockheed * Does Participation Improve Performance? Establishing Causality with Subjective Data by Jonathan Isham, Deepa Narayan, and Lant Pritchett * Natural Resource Management and Economywide Policies in Costa Rica: A CGE Modeling Approach by Annika Persson and Mohan Munasinghe