Victor Silverio Posadas Hernandez
Stock Market Anomalies The Latin American Evidence
Deutscher Universitats-Verlag
Bibliografische Information Der Deutschen Bibliothek Die Deutsche Bibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet uber abrufbar.
Dissertation Universitat Frankfurt am Main, 2004
1.AuflageFebruar2006 Alle Rechte vorbehalten © Deutscher Universitats-Verlag/GWV Fachverlage GmbH, Wiesbaden 2006 Lektorat: Brigitte Siegel / Nicole Schweitzer Der Deutsche Universitats-Verlag ist ein Unternehmen von Springer Science+Business Media. www.duv.de Das Werk einschlieSlich aller seiner Telle ist urheberrechtlich geschiitzt. Jede Verwertung auBerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlags unzulassig und strafbar. Das gilt insbesondere fiir Vervielfaltigungen, Ubersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Die Wiedergabe von Gebrauchsnamen, Handelsnamen, Warenbezeichnungen usw. in diesem Werk berechtigt auch ohne besondere Kennzeichnung nicht zu der Annahme, dass solche Namen im Sinne der Warenzeichen- und Markenschutz-Gesetzgebung alsfrei zu betrachten waren und daher von jedermann benutzt werden diirften. Umschlaggestaltung: Regine Zimmer, Dipl.-Designerin, Frankfurt/Main Druck und Buchbinder: Rosch-Buch, ScheBlitz Gedruckt auf saurefreiem und chlorfrei gebleichtem Papier Printed in Germany ISBN 3-8350-0273-2
Vorwort der Herausgeber Das betriebswirtschaftliche Gebiet Finanzierung hat sich in den vergangenen dreifiig Jahren im Hinblick auf die Abgrenzung von anderen wirtschaftswissenschaftlichen Teildisziplinen, aber auch im Hinblick auf die Forschungsinhalte und die Forschungsmethoden stark gewandelt. Finanzierung wird heute meist, dem amerikanischen Gebrauch des Begriffes „Finance" folgend, als Oberbegriff fiir die Gebiete Untemehmensfinanzierung, Investition und Bankbetriebslehre verwendet. Diesen drei Gebieten ist gemein, da6 die Funktionsweise der relevanten Geld-, Kapital- und Devisenmarkte von zentraler Bedeutung ist. In der Forschung wird ublicherweise mit mehr oder weniger stark formalisierten Modellen in einem ersten Schritt versucht, Hypothesen iiber die betrachteten Sachverbalte abzuleiten, in einem zweiten Schritt werden diese Hypothesen dann empirisch uberpruft, d.h. mit der Realitat konfrontiert. Gemessen am zeitlichen Aufwand der beteiligten Wissenschaftler und am Umfang der vorgelegten Arbeiten hat im Gebiet Finanzierung die empirische Forschung in den vergangenen Jahren betrachtlich an Bedeutung gewonnen. Dabei haben die EDVmaBige Verfugbarkeit von Daten und die verbesserten Moglichkeiten ihrer Verarbeitung eine wichtige Rolle gespielt. In die vorliegende Schriftenreihe sollen Dissertationen und Habilitationen aufgenommen werden, die zur empirischen Finanzmarktforschung (im weitesten Sinne) einen wichtigen Beitrag leisten. Autoren bzw. die sie betreuenden Hochschullehrer werden aufgefordert, sich bei Interesse an der Aufnahme einer Arbeit in die Reihe mit den Herausgebem in Verbindung zu setzen. Eine Liste der bisher erschienenen Schriften ist am Ende dieser Arbeit und auf den Web-Seiten der Herausgeber zu finden. Prof. Dr. J. P. Krahnen Johann Wolfgang Goethe Universitat Fachbereich Wirtschaftswissenschaften Professur fiir Kreditwirtschaft und Finanzierung MertonstraBe 17-21 D-60054 Frankfurt am Main Tel.: (069) 798-22568 Fax:(069)798-28951 E-Mail:
[email protected] http://www.uni-frankfurt.de/professoren/krahnen
Prof R. Stehle, Ph.D. Humboldt-Universitat zu Berlin WirtschaftswissenschaftlicheFakultat Institut fiir Bank-, Borsen- und Versicherungswesen Spandauer StraBe 1 D-10178 Berlin Tel.: (030) 2093-5761 Fax: (030) 2093-5666 E-Mail:
[email protected] http://www.wiwi.hu-berlin.de/finance
Acknowledgments It would have been impossible to undertake the present study without the support of many helpful persons. First of all, I would like to thank Prof. Dr. Jan Pieter Krahnen for directing the present thesis. He has provided me with a wealth of detailed, thoughtful, and constructive conmients. I would also like to extend my thanks to Prof. Dr. Dieter Nautz for taking it upon himself to write the second evaluation of my thesis, as well as to Prof. Dr. Uwe Walz and Prof. Dr. Mark Wahrenburg for their participation in the examination-commission. The team of the Institute for Capital Market Analysis at the Johann Wolfgang Goethe University and the Center for Financial Studies in Frankfurt am Main made important contributions to my work that helped assure the success of the present study. I would like to extend special thanks to Dr. Ralf Elsas, who read earlier versions of parts of this thesis and made me aware of some inconsistencies. In addition, I am deeply grateful to Jesko Meyer, who not only offered me his friendship, but also read the thesis with great care and made many helpful suggestions. My thanks also go to Jesus Diaz, Freddy Moncayo, and Esteban Lx)mbeyda for their econometric and IT-support. Marcus Brainard and Anne helped me with the correction of the language. I greatly appreciate the support provided by the National Council of Science and Technology (CONACYT) of the Mexican Government; without the financing from CONACYT, my studies at the University of Frankfurt am Main would not have been possible. Finally, my greatest thanks is for my wife, Lorena Mondragon. Not only did she leave her work and family in Mexico, but she has been a constant source of support and motivation, and has helped me in the preparation of the final version of this thesis.
Victor Silverio Posadas Hernandez
VII
Contents 1 Introduction 2 Latin American Emerging Markets
1 11
2.1
Introduction
11
2.2
The Latin American Emerging Markets under Study
13
2.3
Previous Investigations of LAEM
14
2.4
Risk Development in Latin American Emerging Countries 2.4.1 Political Risk 2.4.2 Economic Risk 2.4.3 Financial Risk 2.4.4 Composite Risk Index
15 17 19 21 22
2.5
Investment Law Index 2.5.1 Shareholder Rights and Law Enforcement 2.5.2 Insider Trading Index 2.5.3 Barriers to Foreign Equity Investment 2.5.4 Investment Law Index
25 26 28 31 34
2.6
Conclusions
36
3 An Index Methodology for Analyzing and Comparing the Development State and Trading Architecture of Stock Markets 38 3.1
Introduction
38
3.2
Trading Costs across Latin American Stock Exchanges
40
3.3
Development Indicators of the LAEM 3.3.1 Stock Market Size 3.3.2 Liquidity 3.3.3 Stock Market Concentration 3.3.4 Number of Quoted Firms 3.3.5 Composite Index of Stock Market Development 3.3.6 Stock Trading Architecture 3.3.7 Stock Market Intermediaries 3.3.8 Trading Systems 3.3.9 Custody, Clearing, and Settlement Process 3.3.10 Trading Architecture Index
42 43 44 46 50 51 53 54 58 77 89
3.4
Conclusions
90
4 Univariate Portfolio Approach
94
4.1
Introduction
94
4.2
Empirical Evidence 4.2.1 Evidence from Developed Markets
96 97
4.2.2 Evidence of Emerging Markets 4.3
Database
98 100
4.4
Characteristics of the Analyzed Stock Markets
103
4.5
Return Characteristics
104
4.6
Methodology 4.6.1 Portfolio Approach 4.6.2 Preliminary Computations Results
110 110 110 112
4.7
IX
4.8
4.7.1 3-Portfolio Approach
112
4.7.2 6-Portfolio Approach
123
Conclusions
5 Regression Approach
125 129
5.1
Introduction
129
5.2
Methodology 5.2.1 Time Series Regression Approach 5.2.2 Cross-sectional Regression Approach Results 5.3.1 Results of the Time Series Approach 5.3.2 Results of Cross-sectional Regressions
131 133 135 136 137 152
Conclusions
156
5.3
5.4
6 Conclusions
158
References
175
Appendixes
187
Appendix 1: Rating System of the ICRG
187
Appendix 2: DS-Index without Market Concentration
188
Appendix 3: Three-Factor Regression
189
List of Tables Table 2.1: Composite Risk Index Table 2.2: Shareholder Rights Index Table 2.3: Law Enforcement Index Table 2.4: Insider Trading Index Table 2.4: Financial Liberalization Dates Table 2.6: Investment Law Index Table 3.1: Trading Costs Table 3.2: Market Concentration Table 3.3: Development State Index Table 3.4: Market-Makers and Intermediary Index Table 3.5: Trading Mechanisms Table 3.6: Stock Orders Table 3.7: Segments of the Stock Markets Table 3.8. Trading Mechanism for Order Size Table 3.9: Electronic Trading Systems Table 3.10: Transparency of Stock Trading Table 3.11: Use of Call Auctions Table 3.12: TS-Index Table 3.13: Central Securities Depositories-Registries Shareholders Table 3.14: Central Securities Depositories-Registries Table 3.15: Clearing and Settlement Table 3.16: Settlement Assurance Table 3.17: Custody, Clearing, and Settlement Index Table 3.18. Trading Architecture Index Table 3.19: DS- and TA-Indexes vs. Implicit Cost of Trading Table 4.1: Summary Statistics for the LAEM Table 4.2: Characteristics of the Market Monthly Retums Table 4.3: Return Correlation between the Latin America IFCG Indexes Table 4.4: IFCG Price Index Correlation Table 4.5: Summary Statistics for Portfolios Sorted by Size Table 4.6: Summary Statistics for Portfolios Sorted by P/BW Table 4.7: Summary Statistics for Portfolios Sorted by P/E Table 4.8: Summary Statistics for Portfolios Sorted by Turnover Table 4.9: Summary Statistics for Different Periods Table 4.10: Summary Statistics for Portfolios Sorted on a Monthly or Yearly Basis Table 4.11: Summary Statistics for the 6-Portfolio Approach Table 4.12: Summary Results of Chapter 4 Table 5.1: CAPM Regressions Table 5.2: Correlation of the Portfolio Return Differences Table 5.3: Two-Factor Regressions Table 5.4: Summary of the Three-Factor Regressions Table 5.5: Cross-Sectional CAPM Table 5.6: Cross-sectional Regressions Table 6.1: Summary Results of Implicit Trading Costs Table 6.2: Summary Results
25 26 30 31 33 35 41 48 52 57 60 62 63 64 68 72 75 76 79 80 83 87 88 90 93 105 106 107 109 114 116 117 118 121 124 126 128 139 142 144 150 154 155 165 173
XI
List of Figures Figure 2.1 Political Risk Index Figure 2.2: Economic Risk Index Figure 2.3: Financial Risk Index Figure 2.4: Elements of the Composite Risk Index Figure 2.5: Composite Risk Index Figure 2.6: Stock Market Openness to Foreign Investments Figure 2.7: Components of the Investment Law Index Figure 3.1: Stock Market Size (MC/GDP) Figure 3.2: Liquidity Figure 3.3: Number of Quoted Firms Figure 3.4: Components of the Development State Index Figure 3.5: Components of the Trading Architecture Index Figure 4. l:IFC-S&P-Indexes
18 20 22 23 24 34 35 44 45 50 52 91 101
XIII
Table of Abbreviations B/M CAR CATS CCS-Index CRI CSDs DS-Index DvP E/P EMDB ETS FIBV GDP ICAPM lAPT ICRG lEP IFCG IFCI IFC-S&P IID I-Index LAEM MC MSWB NQF 0-Index OLS P/BV P/E S TA-Index TO TS-Index VT/GDP
Book-to-Market Capital Adjusted Rate Computer Assisted Trading System Custody, Clearing, and Settlement Index Composite Risk Index Central Securities Depositories-Registries Development State Index Delivery versus Payment Eamings-to-Price Emerging Market Database Electronic Trading System Federation of Stock Exchanges Gross Domestic Product International CAPM International APT International Country Risk Guide Indicative Equilibrium Price IPC Global Index IPC Investable Index International Finance Corporation and Standard and Poor's Independently and Identically Distributed Intermediary Index Latin America Emerging Markets Market Capitalization Modified P of Scholes and Williams Number of Quoted Firms Openness Index Ordinary Least Squares Price-to-Book-Value Price-to-Eamings Stock Market Capitalization Trading Architecture Index Turnover Trading System Index Value of all Stocks Traded in a year
XV
1 Introduction Interest in the Latin American emerging markets (LAEM) has increased considerably in recent years. However, in their stock markets the price determination process and how it compares with that of developed markets is still an open issue. Thus far, the LAEM and most of the emerging markets may have, as it is often claimed, paid a price for being too different, that is, for having weak institutions, failed macroeconomic programs, political instability, poor corporate governance, and high trading costs. Although they may have indeed suffered for these reasons, this claim ignores the heterogeneity that exists among emerging markets regarding their market development and institutional infrastructure (Yilmaz (2001)). Practitioners still think that the LAEM may lower an international investor's unconditional portfolio risk. In view of this belief concerning emerging markets, the present thesis seeks to answer three sets of questions: (1) What are the investment laws in the LAEM and how do they compare to developed countries? (2) How heterogeneous are the implicit trading costs in the LAEM and which factors are responsible for the heterogeneity? And how different is the implicit trading cost of the LAEM from the developed stock markets? And (3) does the predictability of stock returns in the LAEM differ from those documented for developed markets? A large number of investigations in modem financial economics have been dedicated to quantifying the trade-off between risk and expected returns of financial securities. Financial economists and practitioners argue that this trade-off will depend on the degree to which a security's return covaries with the consumption of a representative investor. However, the implementation of this explanation poses a problem because the representative investor's consumption is not known with certainty. A solution to this problem is offered by the Capital Asset Pricing Model (also called CAPM) developed by Sharpe (1964), Lintner (1965), and Black et al. (1972). The CAPM makes two assumptions: (1) the expected returns and the risk are linearly related and (2) the risk of an investment is determined only by the covariance of its returns with the returns of the portfolio of all assets in the economy, which is to say, ^ is sufficient to describe the expected returns. Since the CAPM was first developed, a large number of studies have examined whether it explains the cross-section of realized average returns. Empirical investiga-
tions have shown that the CAPM is unable to explain the cross-sectional variation in expected returns. For example, in their seminal paper Fama and French (1992) show that stock risks are multidimensional, that is, stock market returns (R^ -Rj)ov
P'm
combination with other variables (such as size, price-to-book-value, eamings-to-price, and turnover, which are usually called market anomalies) better explain the crosssectional average returns. Several theories seek to explain the contradictions of the CAPM. Four explanations are most important: (1) Fama and French (1992, 1993) argue that anomalies must be approximations of risk. (2) De Bondt and Thaler (1987) claim that investors set prices irrationally. (3) Daniel and Titman (1997) argue that firm characteristics rather than the covariance structure of returns appear to explain the cross-sectional variation in average stock returns. And MacKinlay (1995) maintains that the strong relations between the anomalies and the average returns are the result of chance. Giving support to any one of the theories is difficult, since there are enough facts that can be used either to support or to reject them. However, as long as we cannot observe the meanvariance efficient market portfolio, we can not confirm or reject the CAPM. Four anomalies are among the most popular in the academic literature for developed markets: (1) the firm size effect, which was first documented by Banz (1981). He showed that stock market capitalization (5), together with J3, explains the cross-section of average returns and that stocks with a small market capitalization have higher average returns than do large stocks. (2) The book-to-market (B/M) ratio reported by Stattman (1980), who found that average returns on U.S. stocks are positive related to the B/M ratio. (3) The eamings-to-price (E/P) ratio documented by Ball (1978), who argued that the E/P helps to explain expected returns: stocks with higher risks and expected returns are likely to have a higher E/P, regardless of the unnamed sources of risk. And (4) the turnover (TO) or liquidity. A decrease in liquidity may increase expected returns because the investors must be compensated for the higher transaction costs they may face.^ Liquidity continues to gain in importance (Amihud and Mendelson (1986)).
' The size of a security is determined by the market capitalization of its common stock; B/M is the accounting value of the security to its market capitalization; accounting net profits reported by the firm to price per share is the E/P ratio; and the TO is the ratio of shares traded of a security to its outstanding shares.
The joint role of anomalies has also been the subject of numerous papers. Ball (1983) argues that the E/P ratio, together with size and P, explains the cross-sectional average returns on U.S. stocks. Chan, Hamao, and Lakonishok (1991) found that of four variables considered, the B/M ratio and cash flow yield have the most significant positive impact on expected returns on Japanese stocks. For the U.S. stock markets, Fama and French (1992), using the cross-sectional approach of Fama and MacBeth (1973), have found that their test does not support the predictions of the Sharpe, Lintner, and Black model. With data for only nonfinancial firms from the CRSP and COMPUSTAT, they show that for the 1963-1990 period, size and book-to-market capture the cross-sectional variation in average stock returns. Fama and French (1993) expands the asset price tests of (1992) by using the time series regression approach of Black, Jensen, and Scholes (1972). Fama and French (1993) corroborate that portfolios constructed to mimic risk factors related to size and book to market capture the strong variation in returns and remains significant even when other factors are included. They interpret this result as evidence that size and book-to-market do indeed proxy for sensitivity to common risk factors in stock return. In recent years fund managers have substantially increased their participation in emerging stock markets, especially in Asia and Latin America. However, despite the high equity flows to the emerging markets, researchers have only recently become interested in these markets. As a consequence, the amount of research on them is still limited. By contrast, a large number of investigations have been devoted to quantifying the trade-off between risk and expected returns for developed markets. Over the past years countries initially known as "less-developed countries" (such as Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela) eventually became "emerging markets." There are many reasons for this shift in designation. Despite the recent episodes of financial turmoil in the countries just listed,^ their stock market capitalization increased significantly over the past years as a result of the implementation of market-oriented policies, which promote private ownership, including ownership by non-national investors. Furthermore, massive privatization programs, changes in established corporate law, and the use of financial innovations, improvements in
Or thanks to them, because the interest on the sovereign debt of these nations diminished (Levich (2001)).
the regulatory framework, structural and technological changes of the stock markets have characterized the LAEM of late. Before beginning to test whether the CAPM explains the cross-section of realized average returns in the Latin American stock markets, it is first necessary to be conscious of the specific features that differentiate the LAEM from developed markets. To better understand the LAEM, in Chapters 2 and 3 we present and discuss their country risks, investment laws, and stock trading infrastructure. In Chapter 2 we focus on two questions: (1) What are the financial, economic, and political conditions of the LAEM? And (2) what are the investment laws there? To answer the first question, the political, economic, financial, and compound risk indexes of the ICRG are used. In order to evaluate the investment laws directly, information on investor regulation, shareholder rights, their enforcement, insider trading, and barriers to foreign investors is aggregated into an investment law index. This is contrary to La Porta et al. (1998), who analyze shareholder rights, law enforcement, and insider trading separately. Before an investor decides to participate in a stock market, he will compare it with other markets. The literature often mentions three factors that are relevant for their comparison: stock returns, their associated risk, and cost of trading. Chapter 3 concentrates on costs of trading in the LAEM.^ It is important to study the trading costs since it can reduce the return on an investment, especially in illiquid markets. Furthermore, the composition of a global portfolio can change when trading costs are taken into account (Domowitz et al. (2001)). Trading costs can be decomposed into explicit and implicit costs. Due to the difficulties involved in quantifying them. Chapter 3 concentrates on implicit costs. According to the literature, especially two characteristics of a stock market influence implicit costs: state of development and trading architecture. By means of indexes, in Chapter 3 we answer three questions: (1) How heterogeneous are the implicit trading costs in the LAEM? (2) Do trading costs of the LAEM differ significantly from those of developed economies? And of considerable importance: (3) Which factors are responsible for the differences? The chief aim of this chapter is to answer these questions. To do so, two main indexes are constructed here: the Development State Index (DS-Index) and the Trading Architecture Index (TA-Index). The design of indexes has proved to be an efficient way to make comparisons be^ Stock returns and its associated risk are the subjects of Chapters 4 and 5.
tween stock markets (DemirgUc-Kunt and Levine (1996), Erb, Harvey, and Viskanta (1996)). Chapters 2 and 3 are largely descriptive, their principal aim being to help the reader gain an appreciation of the great heterogeneity that exists among the LAEM. This heterogeneity may imply that the risk factors underlying expected stock returns vary among the LAEM and differ from those of developed markets. Using the Emerging Market Database (EMDB) of the International Finance Corporation and Standard and Poor's (IFC-S&P), the stock return determinants in the LAEM are investigated on an aggregate firm level in Chapters 4 and 5. The EMDB contains data of series at stock, index, and market levels for emerging markets. A discussion of the database is undertaken in Chapter 4. In recent years some papers have analyzed the risk and returns in emerging markets. The existing investigations were conducted on country or firm levels. On an aggregate country level, portfolios consist of country indexes (stock markets indexes). The country indexes are assigned to a portfolio according to country or stock market characteristics. On the other hand, on an aggregate firm level, the return premiums are studied by comparing the return on portfolios that are constructed by sorting stocks according to observable firm characteristics. Due to the properties of the LAEM, the analysis in Chapters 4 and 5 is conducted at firm level and by means of a one-way grouping procedure. This approach is supported by several studies. First, Harvey (1995) has shown that emerging market returns are influenced by local rather than by global information variables. The influence of local information might be due to the fact that emerging markets are segmented from world capital markets. Bekaert (1995) has identified a significant number of barriers that might effectively segment emerging markets from global capital markets. He has also shown that emerging markets have a different degree of market integration. Furthermore, Bekaert and Harvey (1995) have found that some emerging markets have become more segmented despite recent waves of liberalization."^ We use the one-way grouping procedure since the number of stocks quoted on each LAEM is not enough to carry out a two- or three-way grouping procedure. Furthermore, the EMDB does In 4.5 it is shown that the return cross-country correlations of the IFCG indexes have increased but that they are still small compared to developed markets.
not cover all stocks. For example, only twelve stocks on the Venezuelan stock exchange are included for one month. An alternative would be to combine the stocks of each country in order to apply a two-way grouping procedure and calculate an International CAPM or APT (ICAPM or IAPT, respectively). However, these models assume integrated capital markets with no differences in the purchasing power parity, legal and trading systems, etc. Thus, using an ICAPM or lAPT might be inappropriate for the LAEM. We pursue two goals in Chapter 4. The first is to investigate whether additional risk factors explaining stock return variation in developed markets are also present in the LAEM. In this chapter we concentrate on the predictability of average returns at an aggregate stock level and test whether, as in the case of the developed markets, S and the ratios price-to-book-value (P/BV), price-to-eamings (P/£), and TO, are related to the stock returns on Latin American stocks. The selection of these variables is motivated by the existing evidence on U.S. and other developed markets and by the practice of security analysts (see Chan, Hamao, and Lakonishok (1991)). Since 5, P/BV, and P/E are all scaled by the price, we test whether TO is related to the return premiums. We use TO because investors commonly monitor it in order to make an investment decision. Furthermore, TO is important for the firms, because less liquid stocks may have to pay an extra premium as compensation. The second objective of Chapter 4 is to reconcile or to find the causes for the contradictory results concerning several anomalies documented in previous studies. Rouwenhorst (1999) found that return factors in emerging markets are similar to those documented for many developed markets. In particular, all emerging market stocks exhibit momentum and small stocks outperform large stocks, and value stocks outperform growth stocks. On the other hand, Claessens et al. (1995) show that there is a size effect for some markets, but this effect is not necessarily related to the smallest size stocks. Furthermore, they do an F-test for the equality of returns across all portfolios in each market. They found that the difference between all portfolios is not significant at the 5% level for any country. In summary, both papers show contradictory results for the anomalies that they both analyze (size and value).^
^ Firms with high ratios of B/M, E/P, or cash-flow-to-price are classified as value stock (see Fama and French (1998)).
The sources of the contradictions in their results are still unclear. However, there might be four important reasons for the differences in results. (1) The analyzed sample period varies in both investigations. (2) Stocks are grouped in different number of portfolios: while Rouwenhorst constructs three portfolios (top 30, middle 40, and bottom 30%), Claessens et al. construct four portfolios (top 25%, top-middle 25%, bottom-middle 25%, and bottom 25%). (3) Rouwenhorst sorts stocks monthly, while Claessens et al. does so on a yearly basis. And (4) to compute portfolio returns, they each weight stocks differently. Therefore, in this chapter we conduct a simple portfolio approach by which stocks are sorted into one of three or six portfolios, the stock allocation takes place each month or year, and the calculated returns are either equally or value-weighted. Finally, with a r-test we check whether the returns of the top and bottom portfolios are significantly different and compare results. Here we also present the empirical evidence on developed and emerging stock markets, introduce the EMDB, and discuss and document the stock return characteristics of the LAEM. Although helpful, the portfolio (or univariate) approach taken in Chapter 4 cannot answer questions such as: Would P be able to absorb the return variation if the fundamental variables were to become insignificant? Or if one of the fundamental variables were to become significant while y^did not? Or if both of them were to explain the return variation? Or if a third factor may even be needed? To answer these questions, a multivariate regression approach is conducted in Chapter 5. The use of a multivariate analysis makes sense since 5, P/BV, and P/E are multiplied by the price. As a consequence, some of the explanatory variables may prove to be redundant in the description of average returns. It is therefore necessary to clarify the relation between the factors and to find the combination that best explains the expected stock returns. The next aim of Chapter 5 is very much in keeping with Chen and Kan (1989).^ In order to verify the robustness of the results obtained by the econometric methodologies, two different multivariate regression approaches are implemented: the time series regression approach developed by Black et al. (1972) at a portfolio level and the crosssectional approach developed by Fama and MacBeth (1973) at a stock level. The time series regression approach is used first. Monthly excess returns of each portfolio are
Davis et al. (1999) have also documented that the long-term return anomalies are sensitive to methodology.
regressed on market portfolio excess return (R^ ~ ^ / ) ^^^ ^n return differences in portfolios that mimic S, P/BV, P/E, and TO. For each stock market, different variants of the equation below are computed in order to define the combination of factors that best captures the common variation of stock returns: R,, -Rf^ =c-^AKt - ^ / J + ' ^ V P S ) . + ^ ^ / ^ V P 3 ) . ^^Pl^in-n),
^ITO^FX-F,), +^P.'
where P = PI, P2, or PJ; r = 1, ... 163; Rpt is the return on portfolio P in month r; R/t is the risk-free rate in r, and Rm,t is the return on the stock market portfolio in t and S(pi.p3),t is the return difference between the sized portfolios PI and P3 in r, P/BV(Pi. P3)j is the return difference of the P/BV portfolios PI and P3 in /, P/E(Pi.p3)j is the return difference of the P/E portfolios PI and P3 in r, and T0(PJ.P3)J is the return difference of the TO portfolios PI and P3 in t. The time series approach has several advantages. (1) It offers a simple and formal return metric by which to choose the best combination of risk factors. (2) With the intercepts it is possible to test how well the different combinations of the variables explain the average returns.^ (3) The high volatility of stock returns in these markets will not lower the power of the asset pricing test. And (4) the number of stocks included in the EMDB is sufficient in order to carry out the necessary calculations. The cross-sectional approach is the second multivariate methodology used here. It is a two-pass regression approach. First, the fi is estimated for each asset based on a time series regression. Then, for each time period the market risk premium At must be estimated based on a regression across assets. This methodology has two advantages: (1) it can aggregate additional risk measures beyond >^ and (2) >^and the coefficients of the explanatory variables are updated periodically. Since we do not have enough stocks to construct portfolios in each market and since analysis at a portfolio level may generate biases in statistical inferences (Lo and MacKinlay (1990)), we evaluate whether the market returns completely explain the realized stock returns without
^ See Fama and French (1993).
forming portfolios. The cross-sectional approach with one-factor (CAPM), two-factor, and five-factor is only applied in the Brazilian and Mexican stock markets.^ Although helpful, the methodology developed by Fama and MacBeth (1973) also has problems. One is the bias it generates against finding a systematic relation between P and returns.^ The reason for this is that tests on the unconditional CAPM are based on expected and not on realized returns (Pettengil et al. (1995)). If one tests the CAPM with realized returns, one should also consider the segmented relation between realized returns and P in order to avoid the bias against the systematic relation between P and returns. Since our results confirm the hypothesis of Pettengil et al. (1995), the conditional CAPM is tested. Considering the conditional relation between P and realized returns. Chapter 5 presents estimates of some combinations of Equation (5.15): /?,,, -R,,, =c, +A,AA, +\,{\-D,)fi,^
+s,S„ +b,P/BV,^ +e,P/E,^ +l,TO,, +s,
(5.15)
where/?,, -R^^is the excess return on stock /, Pit is the modified Scholes and Williams P, Dt is a dummy variable that takes on the value 1 if the market risk premium in t is positive ({R^^ - R^^) > 0) and otherwise 0, and 5/,^ P/BVtt, P/Eit, and TOtt are the market capitalization, price-to-book-value, price-to-eamings and turnover of stock / in t. Adding 5, P/BV, P/E, or TO to the conditional relation between returns and P does not pose a problem. For the additional factors, it is not necessary to test for a conditional relation since their values are always positive. There are four main results of this thesis: (1) The ICRG indexes show that the Latin American countries analyzed here are still riskier than the developed countries considered. However, the ICRG index has been improving faster in the LAEM than in the developed countries. (2) The Investment Law Index in the LAEM is as heterogeneous as in the developed countries (Chile has an IL-Index similar to those of the UK and the U.S., while Brazil, Argentina, and Peru have IL-Indexes similar to those of France, Germany, and Italy). (3) From the Development State Index and the Trading
The other markets are not included in the cross-sectional analysis due to the low number of stocks. ' Other complications with the cross-sectional methodology are the "error-in-variables," which is caused by the use of estimated (from data) ^s, and the unobservability of the true market portfolio (see Roll and Ross (1994)).
Architecture Index can be inferred that trading in Latin America is more expensive than in the developed markets, although there are important differences among the LAEM. (4) The empirical evidence from the developed markets presented by Fama and French (1992 and 1993) regarding market anomalies is not corroborated in the LAEM. The statistical significance of the firm-related variables depends on the country, period, frequency at which stocks are sorted (that is, monthly or yearly) and on the way in which stocks are weighted (that is, equally or value-weighted).
10
2 Latin American Emerging Markets 2.1 Introduction This chapter presents Latin American emerging markets. Here we shall discuss two questions that are important to investors: (1) What are the financial, economic, and political characteristics of Latin America? And (2) what are the investment laws there? To answer the first question, the political, economic, financial, and compound risk indexes of the International Country Risk Guide (ICRG) are used. As regards investor regulation, on the other hand, shareholder rights, their enforcement, insider trading, and the barriers imposed on foreign investors are first discussed. Thus, to answer the second question, it is necessary to construct an Investment Law Index by which the relevant information is aggregated. In the literature it is often said that finance, economic, and political conditions are very important aspects of a stock market since they influence the stock returns and fundamental variables of firms (Erb et al. (1996)). Furthermore, a discussion of these activities is also relevant due to the transformation process in which Latin American countries became involved. To describe each stock market and compare it with the other markets, a direct evaluation of their financial, economic, and political conditions is carried out. The direct evaluation consists in discussing four country risk indexes that have been computed by the ICRG: political, economic, and financial risk indexes, as well as the compounded or total risk index. The ICRG's risk indexes are well regarded by researchers because they include relevant factors for investors and are correlated with ratings of Standard & Poor's, Moody's, and Institutional Investors. Theoretical and empirical studies show that the analysis of market regulation and enforcement is important since both can generate frictions that negatively affects statistical properties of stock returns: If frictions were to exist, agents would not possess symmetric information and prices would not reflect expected values. These might cause prices not to follow processes such as random walk or other forms of martingale. For this reason, in this chapter we discuss laws that incentive investments, that is, laws designed to avoid frictions. Shareholder rights, law enforcement, insider trading, and investment barriers are discussed in this chapter, since they are among the
most studied market characteristics in the literature that might cause non-trivial frictions.^^ The literature on the effects of regulation on stock returns is abundant. La Porta et al. (1998) found that premia on securities in 49 developed and emerging countries depend on the ownership rights and on the legal capacity to enforce these rights, as well as on how well rights are protected. Other researchers point out how insider trading and barriers to foreign investments influence features of stock returns (for instance, Stulz (1981), Meulbroek (1992), Bailey and Jagtiani (1994), Hart (1995), Stulz and Wasserfallen (1995), Domowitz et al. (1997), Bhattacharya and Daouk (2002), and Lombardo and Pagano (2000 and 2002)). In the second part of this chapter, the existence and enforcement of laws is first discussed. Then insider trading and barriers imposed on foreign investors are analyzed. Finally, contrary to La Porta et al. (1998), the discussed variables are aggregated in order to compute an IL-Index. The development of such an index has the advantage that it makes it easier to interpret information and to compare information across stock markets. The discussion of risk sources, shareholder rights, law enforcement, insider trading, and investment barriers is of considerable importance not only for investors but also for regulators. Investors may get a better understanding of stock returns since the financial, economic, and political conditions directly influence the global meanvariance efficient portfolios. Furthermore, laws and their enforcement also affect international investors' returns. For example, due to barriers imposed on them, investors will not get all the gains from international diversification that they would get were those barriers not in place. On the other hand, differences in the investment regulation will give policy makers a clearer understanding of the types of issues they need to address. The present chapter comprises five sections. In the first and second sections, the stock markets under consideration are defined and results of previous studies are presented, some of which prove to be contradictory. The third section contains a discussion of the political, economic, and financial risk characteristics, followed by an analysis of
'° Governments could also pursue appropriate policies in order to the improve regulatory framework of capital markets and bring domestic accounting and supervision standards into line with international standards, actions that eventually boost public confidence in the domestic market (see Pohl et al. (1995)).
12
the total risk index, which compounds the indexes of poUtical, economic, and financial risk. Section 2.5 compares the analyzed markets regarding their respective investment laws and their enforcement, insider trading, and restrictions on foreign investments. To do so, an index is designed. And in the final section, conclusions are drawn.
2.2 The Latin American Emerging Markets under Study To begin with, it is important to define the term "emerging market." The International Finance Corporation defines an "emerging market" as a stock market in transition, increasing in size, activity, or level of sophistication. Stock markets that retain or introduce investment restrictions—such as foreign limits, capital controls, extensive government involvement with listed companies, and other legislative restraints on market activity (particularly on foreign investors)—are usually considered to be emerging markets. In general, the IFC classifies a stock market as "emerging" if it meets at least one of two general criteria: 1. The market is located in a low- or middle-income economy as defined by the World Bank. (According to 1997 data, these are economies with a GNP per capita lower than US$ 9,656.) 2. The investiable market capitalization is low in comparison with its most recent GDP figures. According to this definition, all Latin American stock markets are regarded as emerging markets. However, a great heterogeneity exists among these markets. Areas such as operational efficiency, quality of market regulation, supervision and enforcement, corporate governance practices, minority shareholder rights, transparency, level of accounting standards, and information levels, vary substantially among these stock markets (see Section 2.5, below). While it would have been very interesting to analyze all the Latin American stock markets, unfortunately the requisite information is readily available only for the region's largest stock markets. The stock markets analyzed here are Buenos Aires (Argentina), Sao Paulo (Brazil), Santiago (Chile), Bogota (Colombia), Mexico, Lima
13
(Peru), and Caracas (Venezuela). Hereinafter, these markets are referred to as LAEM.^^ The IPC also mentions Ecuador, Jamaica, and Trinidad as "frontier markets." These three markets are not treated in this study because they tend to be relatively small and illiquid even by the criteria established for emerging markets. In addition, information about them is typically less available than for the main emerging markets in Latin America. According to the IPC, these markets do not even have the breadth (e.g., listings), the depth (e.g., market capitalization and turnover), and infrastructure (e.g., regulatory structure, custody, clearance, and settlement) that would be required in order to calculate indexes for them.
2.3 Previous Investigations of LAEM Results of previous investigations of LAEM are contradictory. On the one hand, it is said that in recent years countries initially regarded as "less-developed countries" such as Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela - became "emerging markets" (see, e.g., Korajczyk (1996), Bekaert (1995), Bekaert and Harvey (1995)). There are many reasons for this shift in classification. Despite the recent episodes of financial turmoil in these countries,^^ stock market capitalization increased significantly over the last years as a result of their market-oriented policies, which promote private ownership, including ownership by non-national investors. The market capitalization of the largest Latin American markets, such as Brazil (US$ 186.2 billion) and Mexico (US$ 126.3 bilUon), is larger than in several developed markets, such as Austria (US$ 25.2 billion) and New Zealand (US$ 17.74 billion). ^^ The total market capitalization of LAEM was US$ 434.4 billion at the end of June 1992. From June 1992 to June 1999 it grew 65.7%. As the importance of equity flows increased, the LAEM have become more integrated in the developed markets. ^"^ ^* These stock markets are the most studied of Latin American markets (see, e.g., La Porta et al. (1997, 1998, 1999, and 2000), Domowitz et al. (1997), Henry (2000), Rouwenhorst (1999), and Fama and French (1998)). '^ Or thanks to the turmoil, because the interest on the sovereign debt of these nations diminished (Levich (2001)). '^ Market capitalization of shares of domestic companies excluding investment funds, rights, warrants, convertibles, foreign companies, and including common and preferred shares and shares without voting rights. ^^ In particular, private capital flows increased and a shift occurred in private flows from bank to nonbank sources to portfolio and direct investment (see Claessens (1995)). The increase in order flow to Latin America has been in the form of ADRs, GDRs, funds, and a small amount of derivatives. Several
14
Furthermore, the LAEM were characterized by massive privatization programs, changes in established corporate law, and financial innovations, improvements in the regulatory framework, and structural and technological changes in the stock markets. The other point of view in the literature from the 1980s and '90s argues that Latin American and all other emerging markets pay a price for being "too different," that is, for having political instability, failed macroeconomic programs, small financial depth, low liquidity, high stock market concentration, a small number of quoted firms, weak institutions, and law enforcement problems (Levich (2001)). This framework either is a disincentive for investors to participate in Latin American stock markets or results in investors demanding higher premiums. As a result of these problems, the weight of the LAEM on the total market capitalization of the IFCG composite index diminished from 35% in 1992 to 22% at the end of 1999. In fact, interest in the LAEM has increased in recent years. Relative to their economies, Latin American stock markets are large in comparison with other emerging markets, such as Poland (19.9% of GDP) or Indonesia (17.5% of GDP), although still small (37% of GDP) compared to developed markets. Furthermore, the market capitalization of Latin American stock markets at the end of 2001 was smaller compared to some developed markets, such as in the United States (US$ 13.8 trillion),^^ Japan (US$ 2.3 trillion), the United Kingdom (US$ 2.16 trillion), and Germany (US$ 1.1 trilUon).^^
2.4 Risk Development in Latin American Emerging Countries The literature frequently mentions that the financial, economic, and political conditions are very important for investors since those conditions also influence the risk associated with stock returns. The analysis of these conditions is also relevant in view of the transformation process of Latin American economies. To assess the impact of the reforms that have been implemented during this process, the outcome of such reforms are evaluated next.
factors have been important in the increase in the capital flows to the LAEM. The decline in international interest rates (Calvo, Leiderman, and Reinhart (1993)) and improved domestic policies give rise to higher growth rates (Chuhan et al. (1993)) and market liberalization (Claessens and Rhee (1994)). '^ This figure represents the sum of Amex, Nasdaq, and NYSE. ^^ Only Frankfurt is included.
15
The ICRG's political, economic, and financial risk indexes are used here to describe the status of the Latin American countries. ^^ There are three reasons for using the ICRG risk indexes. First, they are well regarded among researchers (see Erb, Harvey, and Viskanta (1996)). Second, they include factors relevant to investors. Surveys by Institutional Investors^^ show that the most relevant factors for investing in a country are: economic outlook, debt service, financial reserves-to-current account, fiscal policy, political outlook, access to capital markets, trade balance, portfolio investments inflow, and foreign direct investments. Using these and additional information, the ICRG computes their risk indexes. The ICRG computes their political, economic, and financial risk indexes using twelve political, five economic, and five financial factors. Appendix 1 lists the factors and their weights. ^^ The last reason for using the ICRG risk indexes is their high correlation with ratings of Standard & Poor's, Moody's, and Institutional Investors (see Erb, Harvey, and Viskanta (1996)). The ICRG's rating system consists of assigning numerical values to each risk factor. The values are specified within a range that reflects the weight of the risk factor. To determine the value of an index, the values assigned to each risk factor are summed. The minimum number of points that can be assigned to a risk factor is zero, while the maximum number of points depends on the fixed weight that the factor is assigned from the overall risk assessment (for the maximum points assigned to each factor see Appendix 1).^^ The scale is designed to indicate the highest value with respect to the lowest risk and the lowest value with respect to the highest risk. This applies to all countries. The risk points of the political index components are also assigned each month by the editor for the region in which the country falls, on the basis of the questionnaire for each risk component. On the other side, the risk points for the economic '^ The ICRG is a private provider of risk ratings that have proved to be reliable. The ICRG's risk ratings have been cited by important publications, such as LaPorta et al. (1997) and Erb et al. (1996), as well as in documents published by the International Monetary Fund and World Bank, among other international institutions. Country reports of the ICRG include descriptive assessments and economic data. The ICRG provides ratings for 140 countries on a monthly basis. ^* Institutional Investors is a provider of country risk ratings. The risk measures of Institutional Investors are based on a survey applied to leading international bankers who are asked to rate each country. According to the information of Institutional Investors, to compute the ratings, respondents with greater exposure and more sophisticated country analysis get greater weights. To identify the factors to be considered, survey participants are asked to rank the factors that they consider in preparing country ratings. Two important facts should be considered: (1) bankers rank factors differently for different group of countries and (2) their rankings have changed over time. Erb et al. (1996) provides a detailed description of the way in which Institutional Investors compute its ratings. '^ A full description of each factor can be found at www.icrgonline.com/icrgMethods.asp. ^ Also, the Guide to the ICRG Rating System explains in detail what the rating process is.
16
and financial risk components are automatically assigned each month on the basis of the estimated ratios for the components.^^ The remainder of the present section is divided into four parts. In the first, political risk is analyzed. Economic risk is discussed in the second part, while financial risk is treated in the third. Finally, the CR-Index is discussed and compared across countries. 2.4.1
Political Risk
The importance that investors give to the political, economic, and finance risk indexes depends on the market in question. For developed markets, Chan, Chui, and Kwok (1999) argue that there is a sharp contrast between the nature of political and economic news. Economic news impacts directly on the economy and thus on the stock market. The impact of political news is less clear because: (1) it is not directly related to economic activities, (2) politicians usually obfuscate the informational content of the news, and (3) most of the analysts are well trained in economics and finance, while they are less confident when it comes to analyzing political news. Therefore, economic news has a greater impact on the stock market than does political news.^^ According to the ICRG, the evidence shows the reverse in emerging markets; the ICRG considers political risk to be twice as important as economic and financial risk. Furthermore, to participate in an emerging market, investors have to take different factors into account than they do in a developed market or, in the most optimistic case, the factors are the same but their weights are different (Erb, Harvey, and Viskanta (1996)). Political risk is so important for the ICRG because it reflects a country's willingness to pay its debts while economic and financial risk is associated with the ability to pay. The political difficulties that have arisen in Latin America over the past two decades give support to the ICRG. Most of the countries were transformed from dictatorships into civil governments. In this process, elections used to be violent and fraudulent. Elected governments lacked control over their congress; hence, officials had difficulties in passing bills aimed at instituting structural reforms. Furthermore, there were allegations of corruption in the privatization and sale of the various state-owned companies, from which politicians and their families presumably profited illegally. In ad^^ The available information for the political risk scores is subjectively evaluated by the staff, while the financial and economic risk assessments are made solely on the basis of objective information.
17
dition, the increasing role of drug and guerrilla organizations gave rise to serious political uncertainty. However, there is a heterogeneity among the Latin American countries, and these events have been felt differently in the political and economic landscape of each country (see Figure 2.1). Figure 2.1 depicts the political risk indexes of the ICRG. Over the period analyzed, the behavior of the political index differs among the LAEM. While the political index increased by 71% in Chile and 59% in Peru, it decreased by 25% and 10% in Venezuela and Colombia, respectively. Compared to the other LAEM, the political risk indexes were stable in Brazil and Mexico and belong to the highest over the whole period analyzed. Figure 2.1 also shows the decreases in the index in most of the LAEM from 1995 to 1999 as a consequence of the crises in Mexico and Brazil.
Ly
M
Arg
Bra
ILvy^
Chi
Col
Peru
Ven
Fra
Ger
UK
USA
Figure 2.1 Political Risk Index The Political Risk Index of the ICRG is shown for the time period January 1986 to July 2002. The Political Risk Index scores range from 0 to 100. Zero points mean a very high risk and 100 points indicate the lowest potential risk. The Political Risk Rating includes 12 weighted variables covering both political and social attributes (see Appendix 1).
Compared to developed countries, the political risk ratings of the LAEM are significantly lower, which is to say, the political instability in the LAEM is higher. However, Chile, Mexico, and Peru showed greater improvements than France, Germany, the UK, and the U.S. From 1986 to 2002 the political risk rating of the U.S. dropped by 9%, while it increased in Germany by 3%, in France by 3.2%, and in the UK by ^^ That is, the relation between the quality of information and market activity (Blume et al. (1994)).
18
8%. Interestingly, one can see in Figure 2.4 that when the U.S.-index diminished (from 1989 to 1993), the indexes of most of the other countries also diminished. The converse does not hold, however. For example, the political events in Mexico during 1994 which impacted on the whole of Latin America apparently did not influence either the U.S. or Germany. 2.4.2 Economic Risk In this section the focus is on macroeconomic factors, sunmiarized by risk ratings. The economic risk rating aims to show the economic strengths and weaknesses of Latin American countries. To assess economic development, the ICRG uses five risk components: current account to GDP, real GDP growth, inflation, budget balance to GDP, and per capita GDP. The relationship between stock returns and economic variables has been widely discussed for developed markets. Balvers, Cosimano, and McDonald (1990) derived a model in which stock returns are a function of macroeconomic conditions. Fama (1990) shows for U.S.-stocks that substantial proportions of the variances in stock returns can be attributed to economic variables such as real economic growth, industrial production, and investment. For the UK, Lovatt (1996) explores the relationship between annual real total return on the FTSE All Share and the dividend yield, the inverted yield curve, the expected growth of real GDP, and the expected rate of consumer price inflation. Before we begin to analyze the economic risk index, it is important to review briefly the economic events in Latin America since 1970 in order to gain a better understanding of the current state of its economies. New factors came into play by the early 1970s (Lewis (1995)): Financial and commodity markets were more volatile,^^ world trade grew at rates never before seen, international capital markets allocated resources in the form of bonds to Latin America in order to finance government dissavings and dubious projects. In contrast to some Asian countries, which oriented capital inflows to export industries, to increase international competitiveness, and to build human capital, Latin America invested in welfare projects and import substitution industries. By the early 1980s, Mexico was in the throes of a debt crisis as a result of these poli^^ The breakdown of the Bretton Wood system caused inflation and increases in interest rates. On the other hand, the prices of commodities fell. Especially in 1974 the price of oil dropped drastically.
19
cies, although it was triggered by Mexico's announcement of a moratorium. After Mexico most Latin American countries fell into similar debt crises. The region entered into a period that came to be known as the "Lost Decade of the 1980s." The subcontinent began to change considerably: export-oriented industries were promoted; the market economy replaced heavy government intervention and import substitution; and governments tried to become smaller. Liberalization programs played a role in increasing the GDP growth rates and reducing inflation. All these factors, including the capital market reforms, help to explain to some extent the increase in private savings by local and foreign investors, which in turn influenced the stock market activity. Although the recent crises in Mexico and Brazil were actually brief and their effects on the region's other economies were not as severe as previously, economic risk in the region remains high (see Figure 2.2).
Figure 2.2: Economic Risk Index The Economic Risk Index of the ICRG is shown for the time period January 1986 to July 2002. The Economic Risk Index scores ranges from 0 to 50. Zero points mean a very high risk and 50 points indicate the lowest potential risk. The Economic Risk Rating comprises 5 weighted variables (see Appendix 1).
The trajectories of economic and political risk ratings show some similarities (compare Figures 2.1 and 2.2). Chile, Peru, and Mexico have the lowest economic risk in the region. First, economic risk diminishes over the period for most of the LAEM. Second, the decrease in risk is most evident from 1990 to 1994; thereafter, the risk ratings diminished (i.e. risk increased) as a consequence of the crises in Mexico and Brazil. Third, Chile and Brazil registered the highest increases in economic risk, while
20
in Venezuela it decreased. And fourth, compared to developed markets, the LAEM have significantly lower economic risk ratings, much greater volatility, and show faster improvements. 2.4.3 Financial Risk The Financial Risk Rating of the ICRG aims to provide a measure of the countries' ability to pay. The financial risk rating includes measures related to a country's ability to finance its official, commercial, and trade debt obligations (see Appendix 1 for the financial risk components and their weights).^"^ Latin America's history of debt management has been highly dependent on business cycles (Mexico 1995, Brazil and Colombia 1999, and Argentina 2002).^^ During a debt crisis, governments and international finance institutions develop emergency economic programs that specify objectives to be reached. In most cases, fiscal deficits should diminish when several tax reforms are adopted. Once the economy begins to recover, the country returns to the international capital markets. As a consequence, the composition of external debt usually changes dramatically. The debt profile is reconfigured and the maturity of debt is extended. In the best case, the external debt diminishes slightly (as in Mexico in 1998). Due to both the timely servicing of their international debt obligations and their lower debt burden, countries earned upgrades in their credit ratings (Argentina and Colombia 1996, Brazil and Mexico 2001). After a period of growth with stability, governments begin to relax spending, which increases the fiscal deficit and public sector debt. Financing the larger deficits pushes interest rates higher, with contractionary effects on the private sector.^^ The increasing uncertainty results in capital outflows, leaving the domestic savings and reserves insufficient to meet their term financing needs. Furthermore, the debt ratings are downgraded and the spreads get significantly wider. The difficulties in accessing international capital markets obligates governments and international financial institutions to develop new adjustment programs (Mexico 1995, Brazil 1998, Colombia 1999, and Argentina ^^ A complete description of the financial risk factors and the way in which the ICRG assigns risk points to each component can be found at www.icrgonline.com. ^ During the period, Chile is the exception. Its foreign debt no longer constitutes a major structural problem. As of November 1999, Chile's public and private foreign debt represented 43% of GDP, while in 1985 it was 125%. Public-sector debt has remained low for the past years; it reached 7.4 % of GDP in 1999, reflecting ten years without fiscal deficits. In 1995 the government and the Central Bank prepaid over US$ 1.5 billion in debt to the International Monetary Fund.
21
2001). Hence the time at which one invests in the Latin American emerging countries might well be extremely important.
Figure 2.3: Financial Risk Index The Financial Risk Index of the ICRG is shown for the time period January 1986 to July 2002. The Financial Risk Index scores range from 0 to 50. Zero points mean a very high risk and 50 points indicate the lowest potential risk. The Financial Risk Rating comprises 5 weighted variables (see Appendix 1).
The financial risk indexes of the ICRG are presented in Figure 2.3. Similar to the political and economic indexes, during the analyzed period most of the LAEM showed an increase in the risk rating, which was most evident from 1986 to 1993. Peru and Mexico registered the highest improvements; Colombia and Venezuela, the lowest. In 1994 and 1995 all of the LAEM reverted to a high risk as a consequence of the crises in Mexico and Brazil. Compared to France, Germany, the UK, and the U.S., the LAEM have lower financial risk indexes. The latter indexes increased also at higher rates than in the four developed countries^^ during the whole period. 2.4.4
Composite Risk Index
Two questions are of concern in this section: (1) How risky are the LAEM compared to the most developed stock markets? (2) How different is risk among the LAEM? In order to combine the results of the risk indexes and to answer these questions, the composite risk index (CRI) calculated by the ICRG is used. The CRI-computation is a
^^ Additional domestic and foreign factors, such as political uncertainty, violence, or events in other emerging countries, reinforce the uncertainty. ^^ In the U.S. the financial risk rating dropped from 49 to 35 points, and in Germany from 50 to 40 points.
22
sum of political, financial, and economic risk indexes. The political risk rating contributes 50% and the financial and economic risk ratings each contribute 25%. Risk Index (ICRG)
Polical Risk Index
Economic Risk Index |
Financial Risk Index
Figure 2.4: Elements of the Composite Risk Index The Composite Risk Index is calculated by adding the Political, Economic and Financial Risk Indexes. The political risk rating contributes 50% and the financial and economic risk ratings each contribute 25%.
The CRI is calculated by the ICRG with Formula 2.1: CRL = 0.5 * P/? + 0.25 *ER-\- 0.25 * FR
(2.1)
where CRIi is the composite risk index of country /, PR is the political risk index, FR is the financial risk index, and ER is the economic risk index. The highest rating (100 points) indicates the lowest risk, and the lowest rating (0 points) indicates the highest risk. The index is discussed for all the LAEM and two developed markets.
Table 2.1 reports the mean, standard deviation, growth rates, and average correlation of the CRI of each country for the whole sample and two subperiods. The correlation among the risk indexes for the whole sample period is also shown in Table 2.1. To compute the average correlation, the correlation of country / with each other is first computed and then averaged. The mean of the correlation of this country with the developed and Latin American markets is also calculated. The results of Figure 2.5 and Table 2.1 clearly show that the LAEM are riskier than the four analyzed developed countries and that the LAEM are heterogeneous. Compared to the four developed markets, the CRI of the LAEM have much lower means (riskier) and higher standard deviations over the whole period and also over the two subperiods. However, the CRI in all LAEM, except Venezuela, grew at higher rates than in Germany and the U.S. The LAEM are heterogeneous and can be divided into two groups. The first group consists of the less risky countries: Chile and Mexico. The
23
country with the lowest CRI, thus the riskiest one, is Argentina followed by Venezuela, Peru, Colombia, and Brazil. From the LAEM Chile and Peru had the highest growth rates, while Venezuela had the lowest (its CRI decreased by 13% in the whole period). These results do not change significantly from one subperiod to another. Also interesting, the correlation between the risk indexes is not as high as one might have expected. However, the correlation between the compound and political risk indexes is the highest in most of the countries. This was expected because the political risk index makes up 50% of the CRI. Furthermore, the correlation between the risk indexes is usually lower in the Latin American countries than in the developed countries.^^ Finally, the CRI-correlation between the LAEM is low but higher than the U.S. or Germany. However, the correlation among the LAEM and with the developed markets tends to increase from the first to the second period.
-Vx^
90
^
ao 70
eo 50 40 30
20
f][/ V
f
^ ^
v*
UK
USJV
k
1
1^
v^/
/
\A/^
/
y
10
AB
Bra
ON
Od
Max
Per
v^
Fra
Ger
Figure 2.5: Composite Risk Index The Composite Risk Index of the ICRG is shown for the time period January 1986 to July 2002. Zero points mean a very high risk and 100 points indicate the lowest potential risk
The reasons for this result would be worth analyzing in future research.
24
Table 2.1: Composite Risk Index Arg
Bra
Chi
Col
Mex
Per
Van
Fran
Ger
U.K.
U.S.
Jap
a^ Mean of the Composite Risk Index 1986-2002
31,55 33,52
48,69
41,03
41,51 22,81 35,37
88,06 92,18
87,85
90,92
91,44
1986-1994
24,96
29,18 37,56
38,14
36,17 13,82 36,42
86,13
92,06
86,23
90,33
93,51
1995-2002
39,45
38,73
44,50
47,91 33,59 34,12
90,38
92,32
89,79
91,61
88,96
3,12
62,04
b) Standard Deviation of the Composite Risk Index 1986-2002
9,04
5,62
14,64
4,31 8,78 10,73
2,52
2,84
1,64
2,37
2,14
1986-1994
5,66
2,75
10,39
2,90 7,03 2,64
2,45
1,72
1,93
1,33
2,69
1,96
1995-2002
5,04
3,13
2,99
2,93 5,97 5,14
2,04
2,06
J,24,_.
1:.81
0,83
2,36
c) Growth Rates of the Composite Risk Index 1986-2002
4,85 26,02
160,16
2,60 57,14 133,33 -13,35 12,94
-0,11
5,64
-3,95 -10,05
1986-1994
64,32
-5,02
123,17
15,63 26,65
32,08
-9,32
7,13
-4,56
-1,69
-5,71
-4,71
1995-2002
-38,82 20,36
15,11
13,79 21,96 56,54
3,93
4,94
3,86
6,84
2,32
-6,53
0,439 -0,146 0,349
0,089
0,379
0,277
-0,48
d) Country Correlation of the Composite Risk Index Average Correlation 1986-2002
0,368
0,453
0,304
0,31 0,287
1986-1994
0,241 0,282
-0,116
0,304 -0,027 0,321 0,123 -0,124 -0,046 0,133
0,257 -0,061
1994-2002
0,019 0,395
0,429
0,334 0,282
0,325 -0,413
0,42
0,407
0,355
0,387
0,213
Average Correlation to Developed Countries 1986-2002
-0,025 0,2929
-0,156
0,002 -0,132 0,1304 -0,093 -0,065
0,44
0,162 0,430
-0,169
1986-1994
-0,304 0,6213
-0,762
0,094 -0,693 -0,098 -0,286
0,809
0,750
0,80
1994-2002
-0,125 0,341
-0,7
0,736
Averagie Correlation to Latin American Countries 1986-2002
0,631 0,559
0,611
0,518
0,566
1986-1994
0,604
0,314
0,569
0,417 0,600
1994-2002
0,115 0,432
0,464
-0,359
0,242
0,056
0,645 -0,181 0,585 -0,062 0,472 0,396
0,461 0,464
0,211 -0,613
0,207 -0,413 -0,132 0,052 -0,43 0,363
0,462
0,224
0,351 -0,284
The Composite Risk Index is from the ICRG. The composite scores ranges from 0 to 100. Zero points mean a very high risk and 100 points indicate the lowest potential risk. The CRI comprises the Political, Economic, and Financial Risk Index. In Panel C, the Growth Rates are indicated in percentage. In Panel D, for the average correlation, the correlation of country i with each other is first calculated and then the correlations are averaged.
2.5 Investment Law Index In the literature much discussion has been devoted to how shareholder rights, law enforcement, insider trading, and restrictions on foreign investments affects the defining features of stocks (e.g., Stulz (1981), Meulbroek (1992), Bailey and Jagtiani (1994), Hart (1995), Stulz and Wasserf'allen (1995), Domowitz et al. (1997), La Porta et al.
25
(1998), Bhattacharya and Daouk (2002), and Lombardo and Pagano (2000 and 2002)). In particular. La Porta et al. (1998) discussed shareholder rights, law enforcement, and insider trading differently for 49 countries. Such an analysis, although helpful, is nevertheless difficult for an investor to interpret. For this reason, we aim to develop an index that aggregates information about the existing shareholder rights, law enforcement, insider trading, and barriers to foreign investors and that is easy to interpret and to compare across stock markets. Such an index should correctly reflect how differently investors are protected around the world and the extent to which they can invest in a foreign stock market. Due to the difficulty of getting information, the Investment Law Index (IL-Index) is computed only for the LAEM and five developed countries. In these countries we want to determine the differences regarding (1) shareholder rights, (2) law enforcement, (3) insider trading, and (4) investment barriers to foreigners across the LAEM, and how they compare to some developed markets. In what follows shareholder rights and enforcement of laws are first discussed and then insider trading and barriers imposed on foreign investors are analyzed. In the final section the collected information is grouped in order to compute the IL-Index. 2.5.1 Shareholder Rights and Law Enforcement Recent finance research shows that shareholder rights and the quality of law enforcement also affect the defining features of stocks (Hart (1995)). The purchase of stocks gives investors the right to receive dividends, but also to elect their company's managers. The latter right is important because investors can vote against company directors. For this reason, laws aimed at establishing this right would enable investors to get paid and firms to receive external finance. This section examines how shareholder rights and law enforcement vary across the LAEM and across some developed markets. Based on the results of La Porta et al. (1998), two indexes are computed: a Shareholder Index (SH-Index) and a Law Enforcement Index (LE-Index). The SH-Index includes eight variables of Table 2 from La Porta et al. (1998).^^ To each variable, a value of 1 or 0 is assigned (explanation is given in the lower part of Table 2.2 for the conditions under which a variable receives ^^ They consider that these eight variables represent the voting rights attached to shares, rights that protect the voting mechanism against interference from the insiders, and remedial rights.
26
1 and or 0). Then, averaging the values of the variables, the SH-Index is computed for each stock market. The only difference between the SH-Index and the AntidirectorRights Index of La Porta et al. (1998) is that the SH-Index includes mandatory dividends and the percentage of capital to call an extraordinary shareholder meeting. ^^ These variables are considered since both represent a right for a shareholders and not a restriction. Furthermore, mandatory dividends are more important in countries with a weak shareholder protection (La Porta et al (1998)).^^ In Table 2.2 the notorious differences of SH-Indexes among the LAEM can be observed. Chile, Argentina, and Peru offer the strongest legal protection to investors, while Mexico and Venezuela provide comparatively weak legal protection. Brazil and Colombia fall in the middle. More surprising is that the evidence for developed markets is largely similar. The UK and the U.S. are located on the top, while Germany and Italy offer the weakest legal protection. Compared to developed countries, Chile has a SH-Index as high as the U.S.'s and the UK's, while Mexico's and Venezuela's are as low as Germany's and Italy's. These results are slightly different from the outcomes of the Antidirector-Rights Index of La Porta et al. (1998), since French-civillaw countries come closer to connmon-law countries La Porta et al. (1998) classified countries according to the origin of their commercial laws. All the LAEM are ranked in the French-civil-law group, which proves to give the weakest legal protection to investors. In addition, the authors did not find a relation between per capita income and shareholder rights. This result is confirmed in Table 2.3, which was elaborated using the information taken from Table 5 of the paper just cited. Law enforcement is also a potential determinant of the rights that security holders have. Weak investor protection can be enhanced with a stronger system of legal enforcement since courts can discourage the abuse of shareholders by managers. To address this issue. La Porta et al. (1998) examine proxies for the quality of the enforcement of these rights. They use estimates of law and order compiled by private credit risk agencies. The Law Enforcement Index is calculated using information from Table ^^ For each of the first antidirector rights. La Porta et al. (1998) assign a country a score of 1 if it protects minority shareholders according to this measure, and otherwise 0. They also give each country a 1 if the percentage of share capital needed to call an extraordinary shareholder meeting is at or below the world median of 10%. Finally, they combine these six scores into one score.
27
5 in La Porta et al. (1998). First, to get value ranges between 0 and 1, the values of the variables are divided by 10. Then, by averaging the six variables used, the LE-Index is computed. Contrary to most of the other indexes, the LE-Index does not show a heterogeneity among the LAEM, although Chile continues to be on the top of the list, and Mexico and Venezuela are actually very close to Chile and Brazil (see Table 2.3). On the other hand, Peru, the country with the lowest per capita GDP of the seven LAEM, is now ranked at the bottom of the list. The ranking of the developed countries is the same as the SH-Index, although the differences are now much smaller. It is also interesting to notice that the quality of law enforcement is higher in all developed countries. Unfortunately, it is not possible to say whether a weaker legal framework in one country is offset when its enforcement is of higher quality. 2.5.2 Insider Trading Index Insider trading usually consists in the purchase or sale of a security, where a fiduciary duty or other relationship of trust and confidence is breached, and an individual or firm profits from material or non-public information about the security. Insider trading violations also include, for example, tipping off investors based on such information, trading securities based on such information, and security trading carried out by any persons who abuse such information. How insider trading affects stock returns is still discussed in the literature (Meulbroek (1992)). On the one hand, opponents of insider trading argue that insider trading creates significantly different stock prices, reduces market liquidity, provides incentives for abusive managerial practices, and is unfair to uninformed investors. On the other hand, proponents of insider trading allege that insider trading fosters price discovery and mitigates incentives for many individuals to collect the same information. Regulators across the world might agree with the opponents of insider trading and rules designed to curb such trading. However, regulation against insider trading might be different in each stock market. To compare it across markets, an Insider Treading-Index (IT-Index) is computed next.
^* This is the case of most of the LAEM, as we will see below.
28
i
Tt •*
OJ C\J
h". «3
CVJ CJ
o" o" o" o" o" o" 6"
CO
liH
lO
r».
•^
CM CO
O
O
O
t-
1-
^
O
1-
O
1-
O
O
O
O
1-
1-
^
O
1-
O
O
O
O
•^
T- O
1-
cvj
«^
ill
>
^
C -^3
^ G0.2 S -2 c t C 3 ^ 9 ^
u
^.8-
U
^•
*- o
§2^
T- -r-
O
C/2
X5 c3
C
6 T3
o o o o o o o
T-
O
O
•>-
S u
2
ii
ii^ O
T-
T-
O
O
^
O
o o o o o E o •o
H
llillll
eg
-o
©
II —
u
O ^ ID
5«S
Pill ilfjti 29
Table 2.3: Law Enforcement Index Efficiency Risk of exof judicial Rule of law Corruption propriation system Argentina Brazil Chile Colombia Mexico Peru Venezuela France Germany Italy United Kingdom United States
Risk of con- Rating on LEtract repu- accounting Index diation standards
0,60
0,54
0,60
0,59
0,49
0,45
0,58
0,63
0,63
0,76
0,63
0,54
0,63
0,73
0,70
0,53
0,75
0,68
0,52
0,65
0,73
0,21
0,50
0,70
0,70
0,50
0,56
0,60
0,54
0,48
0,73
0,66
0,60
0,60
0,68
0,25
0,47
0,55
0,47
0,38
0,47
0,65
0,64
0,47
0,69
0,63
0,40
0,58
0,80
0,90
0,91
0,97
0,92
0,69
0,86
0,90
0,92
0,89
0,99
0,98
0,62
0,88
0,68
0,83
0,61
0,94
0,92
0,62
0,77
1,00
0,86
0,91
0,97
0,96
0,78
0,91
1.00
1,00
0,86
1,00
0,90
0,71
0,91
0,54
Description of the variables: (a) efficiency of judicial system assesses the efficiency and integrity of the legal environment as it affects business, especially foreign firms; (b) Rule of Law assesses the law and order tradition in the country; (c) corruption assesses the corruption in government; (d) Risk of Expropriation asses the risk of outright confiscation or forced nationalization; (e) Repudiation of Contracts by Government assess the risk of a modification in a contract taking the form of a repudiation, postponement, or scaling down due to budget cutbacks, indigenization, pressure, a change in government, or a change in government's economic and social priorities; and (f) Accounting Standards are represented by an index that examines and rates companies' 1990 annual reports on their inclusion or omission of 90 items. All variables scale from 0 to 10, where 0 is the worst note and 10 the best. The note is the average of the 6 variables divided by 10. This data is obtained from Table 2 in La Porta et al. (1998).
The IT-Index is based on results from Bhattacharya and Daouk (2002). They analyzed the impact of the existence of the law first and then the impact of the enforcement of the law against insider trading on the cost of equity. They discovered that the establishment of insider trading laws does not cause a reduction in the cost of equity. However, they found that there is a statistically and economically significant drop in the cost of equity after the first insider trading enforcement action. Using this information, the IT-Index is estimated. This Index includes two factors: (1) the existence of laws against insider trading and (2) the existence of prosecutions. The weight of each factor is determined according to the results of Bhattacharya and Daouk (2002), who found that the equity cost does not change after introducing insider trading laws, but decreases after the first prosecution. Twenty percent of the IT-Index is determined by the existence of laws, while the remaining 80% is determined by the existence of prosecution.
30
Results of the IT-Index are presented in Table 2.4. The IT-Indexes show that there exist large differences of law enforcement against insider trading in Latin American countries. Compared to developed countries, insider trading laws were developed sooner in the LAEM than in Germany. However, the first prosecution took place only recently in most of the LAEM (e.g., in Mexico in 2001),^^ while in Colombia and Venezuela none have been registered thus far. Thus, developed markets enforce better insider laws than the LAEM. Compared to other emerging markets, all the LAEM have insider trading laws, while such a law exists in only 80% of other emerging markets. Furthermore, prosecutions have taken place in 71% of the LAEM, but only 25% in other emerging markets."^^ Table 2.4: Insider Trading Index IT-Law
Argentina Brazil Chile Colombia Mexico Peru Venezuela France Germany Italy United Kingdom USA
First Enforcement
IT-lndex
Points IT-Law
II First Enforcement
1991
1995
0,2
0,8
1
1976
1978
0,2
0,8
1
1981
1996
0,2
0,8
1
1990
not enforced
0,2
0
0,2 0,2
1975
not enforced
0,2
0
1991
1994
0,2
0,8
1
1998
not enforced
0,2
0
0,2
1967
1975
0,2
0,8
1
1994
1995
0,2
0,8
1
1991
1996
0,2
0,8
1
1980
1981
0,2
0,8
1
1934
961
0,2
0,8
1
The note is obtained by adding the points of the existence and enforcement of insider law. For the existence of laws again insider trading, a country gets 0.2 points and 0.8 for the existence of the first enforcement. The information in columns 2 and 3 is from Bhattacharya and Daouk (1999).
2.5.3 Barriers to Foreign Equity Investment Interest in emerging markets has grown in recent years as investors try to achieve higher returns and diversify their portfolios. However, to invest in emerging markets, investors have to consider some barriers. In most of the countries, there are restricSince 2001, 239 indictments have been registered in Mexico (Fernandez-Vega, C , 2002. Mexico S.A., La Jornada, March 27, 2002, Mexico.). The CNBV has imposed fines on individuals and intermediaries for insider trading and market manipulation. However, the names of the penalized individuals and intermediaries have not been disclosed, since authorities contend that there have been no sufficient legal grounds to do so. ^^ Information from Bhattacharya and Daouk (2002).
31
tions that are imposed by law or by individual companies (Stulz and Wasserfallen (1995)).^'^ Barriers imposed by law are usually motivated by the interest in preserving the independence of industrial sectors that are strategically important to the particular nation, whereas the management of a company imposes barriers in order to preserve its powers of decision. Returns to foreign investors are affected by the investment barriers they confront in many emerging stock markets. The implementation of these restrictions possibly causes segmentation since prices for identical claims of cash flows and voting rights may vary across investor groups (Domowitz et al. (1997)). Thus investment barriers could render the diversification of benefits and high returns in emerging markets unattainable for international investors. Research shows substantial influence of investment barriers on share prices (see, e.g., Stulz (1981), Bailey and Jagtiani (1994), Domowitz et al. (1997), and Lombardo and Pagano (2000 and 2002)). Most of the empirical studies on ownership restrictions have shown that series open to foreign investors yield higher returns than those restricted to local investors. These results are mainly explained by two theories: differential valuation models and liquidity models. Differential valuation models argue that differences in prices reflect differences in cost of capital, different investor sentiments, taxes, or other factors among local and foreign investors. For example, taxation of capital gains and dividends used to be lower in the LEAM as in developed markets. Therefore, the net effect of the Latin American systems of taxation makes investments in stocks more attractive to local investors. Thus taxes will tend to reduce the premia of non-restricted shares (Domowitz et al. (1997)). On the other hand, liquidity models show that the differences in prices are a consequence of trading costs. Series restricted to foreign investors are traded less frequently than non-restricted series. The Latin American stock markets impose different barriers on foreign investors. The LAEM permit firms to issue several classes of shares that restrict ownership; nevertheless, the LAEM have been "liberalized" (see Table 2.5). The EMDB includes two share classes of Argentina, four of Brazil, two of Chile, three of Colombia, three of
^'* Restrictions formerly placed limits on the percentage of equity, which is permitted to be held by a foreign investor.
32
Mexico, three of Peru, and two of Venezuela. To limit control by foreign investors, non-nationals can buy only some classes of shares or a percentage of them. For example, in Mexico foreigners are restricted to buying only classes B or CPO. Brazil does not permit that any one foreign shareholder possess more than 5% of the voting classes or 20% of aggregate capital. In Colombia a single foreigner is not allowed to possess more than 10% of voting shares, although he can invest in all traded firms. The LAEM also limit foreign ownership in some sectors, such as financial institutions, energy producers, public utilities, and print and broadcast media. Table 2.5: Financial Liberalization Dates Bekaert and Harvey (2000) Buckberg (1995) IFC-Factbook Argentina Oct. 91 Dec. 1989 Oct. 91 Brazil May 91 May 91 May 91 Apr 90/Jan 92 Chile Oct. 89 Dec 88 Colombia Oct. 91 Feb 91 Feb 91 Mexico May 89 May 89 May 89 Peru Venezuela Dec. 88 Jan 90 Jan 90 The liberalization dates are from Bakaert and Harvey (2000), Buckberg (1995), and the IFC-Factbook.
Exchange and capital controls are also two important barriers in most of the LAEM. These barriers affect repatriation of dividends and capitals. Money-convertibility restrictions are imposed especially during periods of crisis. Another way of restricting capital repatriation is found in Chile, where a minimum investment period is still required. To represent the extent to which the Latin American stock markets are open to foreign institutions, the ratio IFCIMCAP / IFCGMCAP is used, where IFCIMCAP is the market capitalization in U.S.-dollars of the investable index computed by the IPC, and IFCGMCAP is the market capitalization in U.S.-dollars of the global index, also computed by the IPC. This ratio is employed because IFCIMCAP includes only the market capitalization of securities and their percentages that are legally and practically available to foreign investors. Stocks included in this index pass the tests for minimum size and liquidity. On the other hand, IFCGMCAP is meant to represent the broadest capitalization's indicator of the markets. In order to ensure that the IFCG indexes capture the properties of the market, the target market capitalization of IFCG
33
index constituents is between 60% and 75% of the total capitalization of all listed shares. ^^ Figure 2.6 shows the ratio IFCIMCAP / IFCGMCAP, which represents the Openness Index (0-Index) from 1988 to 1999. From Figure 2.6 one can see how the LAEM opened their markets to foreign investments. At the end of the '80s, non-local investors were permitted to invest in less than 50% of the traded stocks. However, during the '90s the LAEM gradually relaxed their barriers to foreign investments, and in the last year of the observed period foreigners could invest in 98% of the quoted stocks. Contrary to the other indicators, the openness measure is homogeneous across the markets. Only the Colombian stock market still restricts foreign ownership to no more than 20%.
Figure 2.6: Stock Market Openness to Foreign Investments The ratio IFCI.MCAP/IFCG.MCAP is used to represent the Openness Index (0-Index) from 1988 to 1999. The IFCI.MCAP is market capitalization of the investable index of the IFC; it includes only the market capitalization of securities and their percentage that are legally and practically available to foreign investors. The IFCG.MCAP is the market capitalization of the global index of the IFC-S&P; it is used to represent the broadest capitalization's indicator of the markets.
2.5.4 Investment Law Index The IL-Index should help to answer two questions: (1) What are shareholder rights, law enforcement, insider trading, and barriers to foreign investors in the LAEM? And (2) how do they compare among the LAEM and with developed markets? To answer ^^ For details on how IFCI.MCAP and IFCG.MCAP are computed, see EMDB-International Finance Corporation, July 1999. The IFC Indexes: Methodology, Definitions, and Practices.
34
these questions, the IL-Index is calculated by averaging the values of the indexes on investor protection, law enforcement, insider trading, and barriers to foreign investors (see Figure 2.7 below). The best rating is 1 and the worst is 0. The results for each market are shown in Table 2.6. Investment Law Index (IL-Index)
Law Enforcement Index (LE-lndex)
Sharehol der Index (SH-I ndex)
InskJer Trading Index (IT-lndex)
[—One share, one vote
k-Rule of law
k-IT-Law
^ Proxy by mail allowed
r-Cormption
'—First enorcement
1 Shares not blocked ^ b e f o r e meeting
_Efficiency of judicial system
[—Cummulative voting
—Risk of expropiation
1 Preempty rights to new [shares
—Risk of reputatran
Investment Restrictions
MFCI-MCAP/IFCG.MCAP
Rating of accounting standards
[_% of shar es to call an extraordinary meeting
Figure 2.7: Components of the Investment Law Index The IL-Index is calculated by averaging the values of the indexes on investor protection, law enforcement, insider trading, and barriers to foreign investors.
Table 2.6: Investment Law Index
Argentina Brazil Chile Colombia Mexico Peru Venezuela
Shareholder Rights
Law Enforcement
0.67 0,56 0,78 0,44 0,22 0,67 0,22
0,54 0,63 0,65 0,56 0,60 0,47 0,58
Insider Trad- Investment Reing strictions 1 1 1 0,2 0,2 1 0,2
1 0,94 0,95 0,75 0,99 0,97 0,98
France 1 0,56 0,86 Germany 0,22 1 0,88 Italy 0,77 1 0,22 United Kingdom 0,78 1 0,91 United States 1 0,78 0,91 To calculate the Investment Law Index, the values of the investor protection, trading, and barriers to foreign investors index are averaged. The best rating is
IL-Index
0,80 0,78 0,84 0,49 0,50 0,78 0,50
1 0,85 1 0,78 1 0,75 1 0,92 1 0,92 law enforcement, insider 1 and the worst 0.
The results of Table 2.6 clearly show that laws that stimulate investments vary in Latin America. Chile, Argentina, Brazil, and Peru have the highest IL-Index, and Co-
35
lombia, Mexico, and Venezuela have the lowest.^^ Compared to developed markets, the Latin American group with the highest ratings protects investors as well as they are protected in Italy, Germany, and France. However, in the UK and the U.S. investors are much better protected than in any other analyzed market.
2.6 Conclusions This chapter dealt with the Latin American emerging markets, focusing particularly on two questions: (1) What are the financial, economic, and political conditions in the LAEM? And (2) what investment laws do those markets have? To answer the first question, the political, economic, financial, and composite risk indexes of the ICRG were compared. For the second question an index was developed that aggregates information on the existing shareholder rights, law enforcement, insider trading, and barriers to foreign investors. The trajectory of the risk indexes clearly supports the hypothesis that the LAEM are transforming from "less developed countries" into "emerging countries." The risk indexes showed greater increases for the LAEM over the analyzed period, except in the case of Venezuela. However, the Latin America stock markets are still riskier than the analyzed developed countries. The ICRG indexes also show that the LAEM are heterogeneous. Of the seven LAEM, Chile, Mexico and Colombia are the least risky, while Argentina, Brazil, Peru, and Venezuela are the riskiest. The second question studied in this chapter is how investment laws vary across countries. Empirical research shows that the defining features of securities are influenced not only by shareholder rights and law enforcement, but also by insider trading and barriers imposed on foreign investors. Contrary to La Porta et al. (1998), who discussed shareholder rights, law enforcement, and insider trading each on its own, we aggregate the information on existing shareholder rights, law enforcement, insider trading, and barriers to foreign investors into an IL-Index. The development of such an index makes it easier for investors to interpret the information and compare it
^** The indexes are calculated using information until 1999. If current information were used, the indexes would change. ^^ La Porta et al. (1998) developed an antidirector rights index, which is similar to the SH-Index. The difference between them is that the SH-Index additionally includes mandatory dividends and the per-
36
Empirical research shows that the defining features of securities are influenced not only by the shareholder rights and law enforcement, but also by insider trading and barriers imposed on foreign investors. For this reason the Investment Law-Index was developed in the second part of this chapter. The IL-Index was calculated by averaging the values of the indexes on investor protection, law enforcement, insider trading, and barriers to foreign investors. The IL-Index suggests that the laws that provide incentives for investment vary among the countries. Chile has the highest IL-Index and is followed by Brazil, Argentina, and Peru. Colombia, Mexico, and Venezuela have the lowest. Compared to developed markets, Chile has an IL-Index similar to the UK and the U.S. The index score of Brazil, Argentina, and Peru is comparable to the index score of France, Germany, and Italy. The IL-Indexes of Colombia, Mexico, and Venezuela are much lower than in any developed country. Indeed, this result differs from those obtained by the Antidirector-Rights Index of La Porta et al. (1998), according to which investors in Mexico, Venezuela, Italy, and Germany would have the same rights. The reason for this lies in the differences showed by each of the four indexes included in the ILIndex. In this chapter it was also shown that risk and laws encouraging stock investment in the LAEM differ significantly from developed markets. Furthermore, among the LAEM diversification of risk is still possible and investment incentives and restrictions still exist. In light of these results, it is necessary to investigate whether the factors determining the average stock returns in the LAEM are similar to comparable factors in developed markets. This will be investigated in Chapters 4 and 5. But before addressing this topic, the cost of trading will be analyzed in the following chapter. It is relevant because this cost can dramatically reduce the return on an investment.
centage of capital to call an extraordinary shareholder meeting. However, the results do not have an important impact on the IL-Index.
37
3 An Index Methodology for Analyzing and Comparing the Development State and Trading Architecture of Stock Markets 3.1 Introduction Before participating in a stock market, investors compare it with other stock markets. For the comparison, three factors are often mentioned in the literature: stock returns, their associated risk, and the cost of trading. The present chapter concentrates on the influence of the development state and trading architecture on the trading costs in the LAEM.^^ The design of indexes has proved to be an efficient way to make comparisons between stock markets (Demirguc-Kunt and Levine (1996), Erb, Harvey, and Viskanta (1996)). By means of indexes, it is possible to answer three questions: (1) How heterogeneous are the implicit trading costs in the LAEM? (2) How different are the implicit trading costs of the LAEM from the developed markets? And of considerable importance: (3) Which factors are responsible for the differences? To answer these questions is the chief aim of this chapter. To do so, two main indexes are constructed here: the DS-Index and the TA-Index. Because they determine the implicit costs, the focus in what follows will be on the development state and the trading architecture. The determination of the causes of increased trading costs is important, especially for illiquid markets, where trading costs can dramatically reduce the return on an investment. Domowitz et al. (2001) demonstrate that the composition of global efficient portfolios can change dramatically when trading costs are taken into account. The reason for this is that the perceived gain from international diversification can be impacted if trading cost is included in computations of returns. Comparisons of the elements that give rise to trading costs across markets are relevant due to the competition for international capital flows. Large institutional traders tend to concentrate their holdings in those emerging markets in which the implicit trading cost is low (Domowitz et al. 2001). Trading cost can be broken down into explicit and implicit costs. Explicit costs are broker commissions or fees, taxes, etc. Unlike explicit costs, implicit costs are not represented by visible accounting charges. Implicit costs comprise indirect trading
38
costs, such as the price impact of trading. Of particular interest in the present context are the implicit costs, since according to the literature they are determined by two characteristics of a stock market: development state and trading architecture.^^ Furthermore, implicit costs in the LAEM represent a higher percentage of the total trading costs than in other regions."^^ A large quantity of research analyzes how market development and some elements of trading architecture affect stock returns. For example, Levich (2001) and Aylward and Glen (2000) discuss various issues of emerging markets in relation to the importance of their stock market development and its impact on the international portfolio equity flows. Regarding the trading architecture of an organized exchange market, Demsetz (1968) analyzed the importance of market-makers in the trading process. Freihube, Kehr, and Krahnen (1997) studied how the activities of the Kursmaklers influence the market liquidity in the German exchanges. Friehube, Krahnen, and Theissen (2002) analyzed the interaction between market structure, order size, and liquidity. Hasbrouck and Schwartz (1988) showed that quality of price discovery, opening procedure, and trading systems have a significant impact on investment returns through its effects on transaction costs, which could in turn cause intraday price movements to be excessively volatile. Madhavan (2000) summarized the literature on market structure and implications on metrics of market quality, such as liquidity and volatiUty. In order to represent the development state of each of the Latin American stock markets, we have constructed the DS-Index. Summarizing existing investigations on this topic, we conclude that market development is a multifaceted concept described by four market indicators: market size, liquidity, stock market concentration, and number of quoted firms. These indicators have been computed for the countries under study here and compared across stock markets in order to determine differences in development. Then the DS-Index is calculated by averaging the four development indicators. There is no empirical or theoretical explanation for this way of averaging the indicators; however, as the results show, the DS-Index is robust (see section 3.3.10).
^* Stock returns and their associated risk are the subjects of Chapters 4 and 5, respectively. ^^ The development state and the trading architecture of a stock market might also indirectly influence explicit cost since authorities usually define commissions and fees by taking into account, e.g., liquidity, market concentration, or costs of clearing and settlement. ^ Explicit costs are not considered in the present study since they are determined by fees and taxes, which in many cases are simply imposed by the governments.
39
The consolidation of organized securities exchanges is becoming increasingly frequent. This motivates a continuous implementation of new market participants, new trading mechanisms commonly based on electronic systems, and improvements to their central securities depositories and registries. Therefore, the Latin American trading architectures must be evaluated with the aid of a methodology that is brought up to date periodically. The trading architecture of the Latin American stock markets is analyzed and compared by means of the constructed TA-Index. Three elements play a role in the trading process: market intermediaries, trading systems, and central securities depositories and registries. For each element we compute an index: the Intermediary Index (IIndex), the Trading System Index (TS-Index), and the Custody, Clearing, and Settlement Index (CCS-Index). Then the TA-Index is calculated by averaging these three indexes. The remainder of this chapter is divided into four sections. In 3.2 information on trading costs is presented and compared across the analyzed stock markets. In 3.3 four stock market development indicators are calculated and discussed. In 3.4 trading architecture is presented in two subsections (3.4.1 and 3.4.2). In the first an analysis and comparison of the evolution and current state of the stock trading architectures across the LAEM and developed markets is carried out. This subsection falls into three parts: stock market intermediaries, trading systems, and depository, clearing, and settlement process. In 3.4.2 the TA-Index is calculated by averaging the three indexes computed in 3.4.1 (I-, TS-, and CCS-Indexes). Finally, in 3.5 conclusions are presented.
3.2 Trading Costs across Latin American Stock Exchanges Transaction costs are another important characteristic that distinguish emerging from developed markets. For the period from September 1996 to December 1998, Domowitz et al. (2001) calculate a list of one-way equity trading costs using information from Elkins & McSherry. They compute explicit costs (brokerage commissions, fees, taxes) and implicit costs (impact of trading on stock prices). Domowitz et al. (2001) discover that transaction costs in emerging markets are significantly higher than in developed markets, even after correcting for factors affecting costs, such as market
40
capitalization and volatility."^^ Furthermore, their results show the importance of implicit costs, which represent one-third of the total cost. Table 3.1: Trading Costs Total Trading Cost
Explicit Cost
Implicit Cost
Quarterly Return
Buenos Aires Sao Paulo Santiago Colombia Mexico Lima Caracas Mean Std. Dev Min Max
76,9 58,0 84,3 97,5 61,7 95,8 134,1 86,9 25,8 58,0 134,1
47,3 36,7 45,7 55,3 34,4 60,6 99,4 54,2 22,0 34,4 99,4
29,6 21,4 38,6 42,2 27,3 35,2 34,7 32,7 7,1 21,4 42,2
725 551 423 693 383 301 757 547,6 182,5 301 757
Paris Frankfurt London USA Mean Std. Dev Min Max
29,5 37,7 54,5 38,1 40,0 10,5 29,5 54,5
22,8 24,3 39,3 8,3 23,7 12,7 8,3 39,3
6,7 13.4 15,2 29,8 16,3 9,7 6,7 29,8
294 333 296 374 324,3 37,7 294 374
Source of the information is Domowitz et al (2000). The one-way trading costs are represented in basis points for active managers from September 1996 to December 1998 based on quarterly data from Elkins/McSherry Co., Inc. Explicit costs include commissions and fees. Market impact costs (a component of implicit costs) are computed by comparing the trade price to a benchmark price on the day of the trade. Data for the U.S. represents average costs across Amex, Nasdaq, and NYSE markets. Quarterly return is from 1990 to 1998.
Table 3.1 reports the average one-way implicit, explicit, and total equity transaction costs for the seven Latin American and the four developed stock markets from Table 1 in Domowitz et al. (2001). An enormous variation in trading costs across markets is observed. Trading costs range from 30 basis points (bp) in Paris to 134 bp in Caracas. Across Latin American markets, trading cost varies from 58 bp in Sao Paulo and 61.7 bp in Mexico to 97.5 bp and 134 bp in Colombia and Caracas, respectively. The equally weighted portfolio of the Latin American markets has a one-way trading cost of 86.9 bp compared to 40 bp of the four developed markets. If these portfolios turn over two times per year, the annual average trading costs will be 347.6 bp for the *^ High trading cost in the LAEM motivate firms to issue, e.g., ADRs in the U.S. markets or to list their shares abroad.
41
LAEM and 159.8 bp of the developed markets. Of the average annual portfolio return, trading costs constitute 16% in the LAEM and 12% in the developed markets. The variation in the composition of trading costs across Latin American and developed stock markets is not very different. Explicit costs represent 62.4% in the LAEM and 59.3 in the developed markets, while implicit costs made up 37.7% in the LAEM and 40.7% in the developed markets. Across the LAEM, the cost components are heterogeneous. Explicit cost is low in Mexico (34.4 bp) and high in Caracas (99.4 bp), while imphcit cost is low in Sao Paulo (21.4 bp) and high in Bogota (42.2 bp).
3.3 Development Indicators of the LAEM In this section a development index with five development indicators for the seven largest Latin American stock markets, France, Germany, the UK, and the U.S. is constructed and discussed. The analysis of development indexes is important since, together with the trading architecture, they impact on the implicit trading costs. The goal of this section is threefold: (1) to determine how developed the LAEM are compared to the most developed markets, (2) to investigate the development heterogeneity among the LAEM, and (3) to determine reasons for the differences. To achieve these goals, this analysis follows the recent literature. The existing investigations point out that stock market development is a multifaceted concept (DemirgiicKunt and Levine (1996)). As a consequence, no single indicator will completely describe stock market development. For this reason, four development indicators and one index (compounded from the four indicators) are discussed with regard to and compared among the analyzed markets below. Due to the fast changing conditions of emerging markets, the development indicators are discussed for a period of several years. The sample period varies among the markets. For Argentina, Chile, Colombia, Mexico, and Venezuela the analyzed period runs from 1986 to 1999, while for Brazil and Peru indicators are observed over different periods."^^ For France, Germany, the UK, and the U.S., the sample period runs from 1990 to 1999. The information source for the LAEM is the EMDB; for developed markets it is the International Federation of Stock Exchanges (FIBV). InformaThe analyzed periods for Brazil are from 1986 to 1999 for TO and number of quoted firms, and from 1994 to 1999 for MC/GDP and value traded. For Peru, the time periods are from 1986 to 1999 for MC/GDP, value traded, and number of quoted firms and from 1994 to 1999 for TO.
42
tion on market capitalization, value traded, turnover, and number of quoted firms belong to the most important stock exchange of each country, except the U.S, where information on Amex, Nasdaq, and NYSE is included on account of the significance of the stock exchanges."^^ To determine the stock concentration of stock exchanges, information is used from 1995 to 2001 for only five Latin American"^ markets and for five developed markets. The remainder of this section is divided into five subsections. In the first, for each country the stock market size indicator is computed and discussed, while the liquidity indicator is examined in the second. The same approach is taken regarding the indicators of stock market concentration and number of quoted firms in the third and fourth subsections. Finally, after compounding these indicators, the development index is calculated and compared among the markets. 3.3.1
Stock Market Size
The ratio of market capitalization over Gross Domestic Product (MC/GDP) is the first indicator. The literature calls this ratio "Stock Market Size." The Stock Market Size is important because of its possible correlation with the ability to mobilize capital (DemirgUc-Kunt and Levine (1996)). This ratio has several interpretations. (1) Analysts use it as a measure of stock market size. (2) MC/GDP also represents financial depth. (3) If market development and long-run growth are positively related (Levine and Zervos (1996)), MC/GDP will also be a proxy of the stage of economic development. In that case, as the economy develops, we would expect the market to grow in size and in depth."^^ And, very important, (4) this ratio is also viewed as an inverse indicator for trading costs: an increase in financial depth is expected to cause a decline in transaction cost (Aylward and Glen (2000)).
^^ From 1990 to 2000, 47% of the value traded occurred on Amex and Nasdaq; 71% of firms quoted here represented 19% of market capitalization. ^ The FIBV does not report information on the Colombian and Venezuelan stock markets. For this reason neither was included. ^^ It is also assumed that more mature economies rely more on equity markets (Boyd and Smith (1996)).
43
Arg
Bra
Chi
Col
Mex
Peru
Ven
Fran
Ger
Figure 3.1: Stock Market Size (MC/GDP) The market capitalization is from the EMDB for the LAEM and from FIBV for the developed countries. The GDP is from the IMF.
From Figure 3.1 one can see that since 1986 all LAEM were involved in a transformation process. In the late 1980s and throughout the 1990s, MC/GDP increased significantly in the LAEM. Specifically, the stock market size reached a peak between 1993 and 1994 in most of the LAEM. As a consequence of the crises in Mexico, Asia, Russia, and Brazil, from 1995 to the first part of 1999 stock market size diminished. However, it began to increase in the second part of 1999. Compared to developed markets, this ratio is significantly lower in most of the Latin America markets. Chile is the only stock market in the LAEM that has a MC/GDP comparable to Germany's. Mexico has also a high ratio, but it is significantly lower than Chile's. 3.3.2
Liquidity
Market liquidity is the second indicator of market development. While several definitions of liquidity may be found in the literature, all of them refer to the ability to buy and sell stocks. According to theory, greater liquidity means that stocks can be exchanged easily and cheaply, investments are less risky, more profitable, and longer term. Thus a negative link between liquidity and the cost of trading is expected. Liquidity is an important development indicator because it influences the allocation of capital.
It is difficult to find a measure that represents liquidity since all costs associated with trading must be quantified: time cost, counterpart cost, and settlement cost (Demirguc-Kunt and Levine (1996)). Because data is very hmited, the comparison of stock markets is our principal goal here. To do this, two measures of liquidity are used: the ratio of the value of all stocks traded in a year over the GDP (VT/GDP) and TO, the annual value of total shares traded as a percentage of market capitalization. Both liquidity measures are related to transaction costs. Panel a) VT/GDP
Figure 3.2: Liquidity The volume traded and market capitalization arefromthe EMDB for the LAEM andfromFIBV for the developed countries. The GDP isfromthe IMF.
45
It is important to analyze both liquidity measures because they are complementary and can move in contrary directions. VT/GDP explains trading compared with the size of the economy, and TO captures trading compared to the size of the stock market. For example, it is possible for a large economy with a small market to have a low VT/GDP and a high TO if its market is active. The behavior of VT/GDP and TO substantially varies over time and across the Latin American markets. From Figure 3.2 one can observe that in the late 1980s and at the beginning of the 1990s VT/GDP increased significantly, while TO did not show this tendency. In most of the Latin American stock exchanges, VT/GDP attained a maximum between 1994 and 1995. At that time and during the years following, this ratio diminished, although it began to increase in 1999 in Chile and Mexico. During the analyzed period, Brazil (14.95%), Chile (8.49%), and Mexico (11.7%) have the highest VT/GDP, while in Argentina (3.9%), Colombia (1.6%), Peru (3.4%), and Venezuela (2.9%) the VT/GDP is the lowest. On the other hand, from 1990 to 1999 TO is considerably higher in Argentina, Brazil, and Mexico; these averages are higher than in Japan (30.8), though they are still lower than in the U.S. (81.18%), Germany (121.8%), and the UK (86.9%). Compared to developed markets, the VT/GDP ratio is also much lower in the LAEM. These results suggest that liquidity varies across markets. However, liquidity can also vary within a market, since a small number of highly capitalized firms usually have considerable liquidity (see the following subsection). 3.3.3 Stock Market Concentration Market concentration is another important development indicator. It shows the extent to which a stock market is dominated by only a few companies. A high concentration is not desirable for an exchange since it might have a negative influence on the liquidity of the market and on trading costs. In the literature, two different definitions of market concentration exist: the first is taken from the FIBV and the second is stated by IFC-S&P, mainly for emerging markets. For five of the LAEM,"^^ in the present subsection we show the results of both definitions, each of which has two alternatives. Furthermore, market concentrations of the Latin American exchanges are compared with those of the five most important stock exchanges (Frankfurt, London, Nasdaq, NYSE, and Tokyo).
46
According to the FIBV's definition, market concentration is the percentage of market capitaUzation of the largest 5% of all firms. Taking this measure, market concentration in Latin America is heterogeneous (see Table 2.2). At the beginning of the analyzed period, the stock markets in Buenos Aires, Sao Paulo, and Lima had a concentration of about 70%, while the concentration in Santiago and Mexico was 50%. During the analyzed period, Buenos Aires, Sao Paulo, and Mexico converged to a concentration of around 60%, while Santiago remained at 50%. The FIBV's concentration measure is heterogeneous across the Latin American and the developed markets. Mexico has a concentration tendency similar to that of the NYSE and Nasdaq, while Buenos Aires, Sao Paulo, and Lima has one similar to Frankfurt's. The source of the large difference between the Latin American and the developed markets is the number of firms that represent 5% of the market capitalization. The market with the highest number of firms in Latin America representing 5% of market capitalization is Sao Paulo (25 on average over the period). This number is not much lower than the number for the Frankfurt stock exchange (29 on average over the period). ^'^ The percentage of value traded by the largest 5% of all firms is an alternative measure of market concentration. For Latin America this alternative measure of concentration confirms some results for the previous indicator. (1) At the beginning of the period, Buenos Aires, Sao Paulo, and Lima have a concentration that is much higher than in Santiago and Mexico. (2) However, as time passes, concentration in Buenos Aires, Sao Paulo, and Lima diminishes, while it increases in Santiago and Mexico. In the five Latin American markets, the indicator seems to converge to around 70%. In contrast to the first concentration indicator, this alternative measure of the LAEM is not as heterogeneous as in the developed markets. This variable has a different trajectory in each developed market.
The stock exchanges of Colombia and Venezuela are not included since the FIBV does not report information on them. ^"^ In Frankfurt the average is so low because in 1997 and 1998 only 13 and 12 firms, respectively. made up 5% of the market capitalization.
47
8
" G. c3" S Z
z
^s
r-' z :
in 2 . o>" S^
< o >z
Tr«r-T •^
o 2^ 5 s' "* wI-. " ^ JJ. § -. g ^. P « ° CO ° CO ^ lO —- lO ' - ' U5 >-'
(0 •o (0 (0
3 ^ ^. S J ^ 5 ^ 5 ^ § § *-' U)
T---^CO'-«i-^
CM"
3
2.
<M"
°
in °
Q"
y-
U)
^O
P* of 2 . '"" Si-
Q"
•
C0,_-C0^_T-^_^00^^Q^_^^^,^O,_^
^- 5 ^- 3 ° - "^ •*- «^ - ^ "*- f^ "^- "^ r*»S!ih»£2,co£5,orLO)T"_ in CO
h0)_ CO CO
105_ in CO
cOto^in^f«vCOu2Cvi^oo*Oo^ oofJoTwcxjWodWisr^od^Zi in^^in^^co'^S'^'S^^t^'^
M 0)
h00 d" in
T- 00 CO CO 00 m t^ CO h* N h- N
S £i-D5 2 l S £tS 2^i^ 2t?^ 21?: :
z c o •o c o
0 o oi" 't
CO CO ' CO 0_ ' o in h< CO
CM m
o
s
r^ CO
5
m
'E M (0
o 2 O
£ o o
m N 00 ^ ">* -
1- o r^ i 1^. O O
8 8 :
cvT o " CO*
I
CM CM CM
z
S5 §
< S N S
(Q •o 10 (0
Z
u) co" oo" U) CO CO
zc
oo CO -^ NT- U5 O CO
o c o
•o
^' j:5 5 ^ ••- Q m CD 00 CO CM CO
S S 5 ^"
S
O) CO •^ m CO •*
§ f^ CM < oo" ^" CO* Z
s
s 5 CB
E
co" r«-" ui" o>" m
in lO lo
^Dw h- in TT r^
^. ° ^. o 0)
lO lO S
CO
S ID
!S E o
5 O O) CQ
c «••
Q^
E o § s $^ i§ o Oi co" Q " T-" o T; "««• CM" "f ^ -^ "o5 E o
o
lO ^ CJ) CO -«*• CO T- lO
K S. ° -J'S
Jo" 5 §" 5"
=
(0
o c a»
CMCO00lO
[ T T - T - O O C M
(Q"^C0CMO>
O ^ ' ^ ' ^ f ^
c ^
3 ffi
s s 5
49
IFC-S&P measures concentration for emerging markets differently than does the nVB. Market concentration is the percentage of market capitalization of the largest ten quoted firms. According to this definition, stock market concentration also varies among the LAEM. Once again, Sao Paulo and Santiago have the lowest concentrations, while Buenos Aires, Mexico, and Lima show the highest. In comparison to developed markets regarding this measure, only Sao Paulo and Santiago have concentration levels similar to Frankfurt's, the exchange with the highest market concentration of the developed markets analyzed. The last part of Table 2.2 shows a final alternative to measure concentration: the turnover value of the largest ten quoted firms. Using this alternative measure, it is possible to observe that the market concentration in the LAEM is much higher than in the developed markets. Taking into account these four concentration measures, we may conclude that concentration varies. It is possible to observe that (1) concentration is significantly higher in the LAEM than in developed markets and that (2) the Santiago stock exchange has the lowest concentration and in recent years it was similar to the concentration of the developed markets. To a lesser extent, Sao Paulo also exhibited a diminished concentration. 3.3.4 Number of Quoted Firms
Figure 3.3: Number of Quoted Firms These graphs have been constructed based on information from the IFC (for the LAEM) and FIBV (for the developed markets).
50
Finally, the fourth market development indicator is expressed by the number of quoted firms (NQF). This proxy is included because it represents the prevalence of public financing. Furthermore, NQF is sometimes used as an additional measure of market size. The number of quoted firms shows the different development states among the Latin American stock markets. While it tended to diminish over the period in Argentina, Brazil, and Venezuela, NQF rose in the other four markets and most of the increment took place between 1986 and 1994 (see Figure 3.3). Among the Latin American markets, Brazil has the highest NQF: 486 in 1999, which is half of that of Germany and only 5% of the firms traded in the U.S. The market capitalization per firm is also an interesting aspect. In the LAEM the average capitalization per firm over the sample period is smaller than in the markets of the developed countries under analysis. Among the LAEM, however, considerable variation exists. The average market capitalization of Mexican firms over the period 1990-1999 (US$ 622.6 million) is the highest in Latin America and is even higher than in several developed markets, such as Canada (US$ 312.2 million) and New Zealand (US$ 132.6 miUion), and very close to the UK (US$ 624.1 milHon). 3.3.5
Composite Index of Stock Market Development
Once again, the development of a DS-Index is the main goal of this section. The DSIndex should help to answer three questions: (1) How underdeveloped are the LAEM compared to the most developed stock markets? (2) How heterogeneous are the LAEM? And (3) what are the determinants of the differences? The computation of the index involves two steps. First, the mean-adjusted market capitalization, VT/GDP, TO, NQF, and market concentration are computed for each market. The meanadjusted value of a development indicator X for country / is defined as: Xp =[X. - mean (X)]/mean (X)
(3.2)
where mean(X) is the average value of indicator X across all countries over a period of time. The mean-adjusted value of the market concentration measure is multiplied by (-1) since larger numbers correspond to lower stock market development. The second step is to take a simple average of the mean-adjusted market capitalization. Liquidity {(VT/GDP+T0)/2), market concentration, and NQF (see Figure 3.4). Due to a lack of information, the DS-Index is calculated only from 1995 to 1999.
51
Development State Index (DS-lndex)
Liquidity (VT7GDP + TO)/2
stock Market Size (MC/GDP)
Market Concentration
Nr. Quoted Firms
Figure 3.4: Components of the Development State Index The DS-Index is calculated by averaging the mean-adjusted market capitalization, liquidity, market concentration, and NQF
For market concentration, the percentage of value traded by the largest 5% of all firms is used, since nowadays this is the most common measure of concentration. For Colombia and Venezuela, the FIBV does not include data on market concentration, and so market concentration for these countries is calculated using information of the EMDB. Table 3.3: Development State Index 1995
1996
1997
1998
1999
Buenos Aires Sao Paulo Santiago Colombia Mexico Lima Caracas
-0,66 -0.37 -0,02 -0,65 -0,33 -0,45 -0,71
-0,68 -0,29 -0,19 -0,68 -0,32 -0,52 -0,65
-0,50 -0,24 -0,31 -0,65 -0,38 -0,57 -0,57
-0.60 -0,30 -0,42 -0,66 -0,52 -0,60 -0,74
-0.60 -0.33 -0.42 -0.69 -0.47 -0.59 -0.75
Paris Frankfurt London New York
-0,22 0,22 1,37 1,88
-0,19 0,26 1,33 1,99
-0,15 0,11 1,27 1,99
-0,04 0,32 1,48 2,07
-0.06 0.25 1,36 2.30
The computation of the index involves two steps. First, the mean-removed market capitalization, VT/GDP, TO, NQF, and market concentration are computed for each market. The mean-removed value of the market concentration measure is multiplied by - 1 since larger numbers correspond to lower stock market development. The second step consists in taking a simple average of the meansremoved market capitalization, VT/GDP, TO, and NQF.
Table 3.3 reports the values and rankings of the DS-Index for each country. This index shows that the LAEM are heterogeneous. For 1999 Brazil had the best rating, followed by Chile and Mexico. On the other hand, Venezuela had the worst rating, followed by Colombia, Peru, and Argentina. During the analyzed periods, the markets changed places but not the group, that is to say, during the period Brazil, Chile, and
52
Mexico are always among the first three markets, while the other four always occupy the last four places. Among developed markets, the U.S. has the most developed stock market, followed by the UK, Germany, and France. In comparison with developed markets, all the LAEM have considerably lower ratings than developed markets in each period. To obtain a longer period, a second DS-Index is computed that excludes market concentration. This version of the DS-Index shows the same results, except that the differences between the development indicators become larger."*^ The results of both indexes are similar. Indeed, the developed markets are more developed than the LAEM, and the LAEM are heterogeneous. Three (of four) development indicators support these results: market size, liquidity, and the NQF. Only the market concentration does not differ significantly between Latin American and the developed markets. 3.3.6
Stock Trading Architecture
Trading architecture can be understood as the participating elements that make the trading process possible. In any stock exchange, the trading process can be divided into three phases. First, investors buy and sell securities by means of intermediaries in a given country's stock exchangeCs)."^^ Second, orders are transmitted from the intermediaries' headquarters to the stock exchange's central order book via the trading floor or electronic trading system (ETS) for equities, where they are matched with a similar but opposite order, if there is one. And third, a few business days after the transaction has been completed, the Central Securities Depositories-registries (CSDs), upon receiving instructions from the selling brokerage firm, transfer the securities from the selling brokerage firm's account to the buying brokerage firm's account. The corresponding funds are transferred from the buying brokerage firm's cash account to the selling brokerage firm's cash account. To analyze and compare the trading architectures, three indexes are constructed in this section, one index for each market participant: the I-Index, the TS-Index, and the CCS-Index. Then the TA-Index is calculated by averaging these indexes. The averaging procedure of the elements does not follow a theoretical or empirical foundation, but it is based on the author's own criteria, which are based on the literature of trading ^^ The second version of the DS-Index is presented in Appendix 2. ^^ And if permitted by law, in the stock markets of other countries.
53
architecture. The results of 3.3.10 show that the averaging method is robust. However, if we continue calculating the indexes for the subsequent years, we can test if this way of averaging continues been correct, and if it does not, then the averaging procedure can be easily changed. In the next sections, the importance and the role of the stock market intermediaries in Latin America is first discussed and then their trading systems, as well as their depository, clearing, and settlement processes, are compared. 3.3.7 Stock Market Intermediaries Intermediaries are fundamental to a well-functioning trading system. In all the stock exchanges of Latin America, two types of intermediaries exist: the brokerage firms or brokers and the market-makers.^^ The role of intermediaries depends on the system used for trading. Since all the LAEM use order-driven systems, the problems in the trading system and the functions of the intermediaries should be similar. In this subsection we investigate whether this conclusion bears out. First, some problems of the stock trading systems are briefly presented. Then we investigate whether the functions of intermediaries are the same in each market. Finally, an I-Index is calculated in order to compare the role of the intermediaries among the stock markets. Stock exchanges exist because they facilitate the transactions of stocks between buyers and sellers. To facilitate this saving-investment flow, all Latin American stock exchanges have introduced order-driven systems. However, order-driven systems face problems. Some of the most important are: (1) short-term price fluctuations,^^ (2) increasing dominance of order flow by institutional participants (see Becker and Angstadt (1995)),^^ and (3) the concentration of trading in some stocks.^^ There are mainly two solutions to these problems. First, in most of the developed markets brokerage firms are allowed to buy and sell stocks on their own behalf and
^ It was not possible to get information on market-makers in Buenos Aires. ^^ In their centralized and automated trading systems, the brokers input into the system the investors' bids and offers (see 2.2.2, above). Orders are matched when a similar but opposite order is received. However, if there are no orders on the opposite side, the intermediary has to wait until one arrives. This waiting time may cause short-term price fluctuation. ^^ Institutional order flow is more difficult to handle because it is usually one-sided and the orders are large. Therefore, prices become more volatile and the time needed to execute an order increases. ^^ The great number of stocks listed in exchanges does not allow all stocks to be traded continuously. It makes some stocks less liquid. Evidence from the developed markets shows that most of the stocks report few operations and that the trades are spaced over time.
54
not only for a third party, since this provides a means of partially diminishing price volatility. Second, to improve liquidity and diminish the time needed to trade an order (immediacy), continuous order-driven markets allowing the participation of marketmakers have been introduced. Market-makers play a significant role in the determination of prices because they play the key role of price-setters. Since they are willing to buy or sell securities, they provide liquidity to the market and permit continuous trading. However, the benefits from the incorporation of market-makers are still controversial.^"^ For example, Madhavan (2000) states that market-makers can alter prices in response to considerations of their inventory and information. As mentioned above, in the LAEM there are two types of intermediaries: the brokerage firms or brokers and the market-makers. The name of the brokerage firms may vary across the exchanges of Latin America and developed markets, but not their obligations and privileges. To be registered as a brokerage firm, a number of requirements - such as qualification, solvency, moral and net worth - should be met. As compensation, they are allowed to be members of the stock exchanges and to buy and sell stocks by themselves and for third parties in every market of their country.^^ Since the functions (privileges and obligations) of brokerage firms do not vary among the analyzed markets, they are not included in the I-Index. Since stock trading is also concentrated in the Latin American stock exchanges on a few stocks (see section 3.3.3, above), all stock exchanges have introduced marketmakers. They were introduced in order to increase the liquidity of the less frequently traded stocks. The exception is Santiago, where each stock can register a marketmaker. The bid-ask spread and its determinants are defined in different ways in the LAEM. It is generally believed that the bid-ask spread determined under perfect competition is ^^ The analysis of the importance of market-makers in the trading process is not new. Demsetz (1968) showed that market-makers receive the bid-ask spread as the premium for their predictive immediacy services. The bid-ask spread is important because it is part of intraday price dynamics. Smith (1971) extended the analysis of the market-makers. He argued that the primary activity of the market-makers remains the supply of immediacy, but they are also active in the price-setting process in order to adjust their inventories. It implies that prices may depart from their expected values if the dealer position differs from their target, thereby giving rise to transitory price movements over the course of the day and possibly over longer periods. More recent studies have considered the impact of information on market prices. Informed traders expect to make profits from uninformed traders. Market-makers will on average lose with informed traders and win with uninformed ones, which suggests that an informational component might be contained in the spread (Glosten and Milgrom (1985)). ^^ They can buy or sell stocks, but they need not.
55
the reward of market-makers for the immediacy they provide. For example, Freihube, Krahnen, and Theissen (2002) find by means of experimental studies that the role of market-makers will be beneficial only if they are subject to competition. All Latin American exchanges, except Colombia's, permit more than one market-maker per stock. But this does not guarantee the conditions of perfect competition that would justify considering the bid-ask spread in these markets to be the fair reward to market-makers. In Santiago, Mexico, and Lima, a council estabUshes the bid-ask spread, the smallest lot, the minimum number of stocks to sell and buy, and the time duration of the orders posted by the specialist according to liquidity, free floater, antiquity, and volatility. In Sao Paulo and Colombia the council only establishes the minimum level of demand and supply. The compensations received by the market-makers for their services also vary across the Latin American exchanges (see Table 3.4). Market-makers in Santiago do not pay trading fees, while in Colombia they can trade directly with investors and outside of the exchange. In Mexico and Caracas the only remuneration for the market-makers (called especialistas) are the bid-ask spread. In Lima market-makers (called promoters) do not pay trading fees and can also trade directly with the investors, as in Sao Paulo. Another compensation for market-makers in each LAEM is that they are allowed to specialize in more than one stock. The Intermediary Index Since the role of intermediaries in the continuous order-driven systems is fundamental for a well-functioning market, an I-Index is computed and will be included in the TA-Index. As was seen above, two types of intermediaries exist in all analyzed stock markets: the brokerage firms or brokers and the market-makers. Furthermore, it was shown that while the functions (privileges and obUgations) are similar across the exchanges for brokerage firms, they are not contained in the I-Index. Therefore, only the characteristics of market-makers are included in the I-Index. The I-Index is calculated using information from Table 3.4. Stock exchanges have imposed some obligations on and given some privileges to market-makers. The seven obligations and privileges included in the I-Index were identified as the most important by Demarchi and Foucault (1998). Information for these variables is obtained from the web pages of the Latin American stock exchanges. To compute the I-Index,
56
each obligation or privilege gets one point, since each of them induces competitiveness by augmenting liquidity and diminishing trading costs. Then the points for each market are added up and divided by seven. Table 3.4: Market-Makers and Intermediary Index
ill
>
Ill
l-lndex
Obligations
Privileaes
II
If CO
^
•^
5ii.^»f|i|r »|
• • -^
o B ^ .9-2 §
^ it
r
w
CO
c °'S
ao
0)
i- (D W (D
oo t (0
I Pi III 0) ( 0
O
2.^ 9 Q.w c
li I
W D)
O ^ T3 ^ O c
= 1^
CO O) ( 0
©
« r^ 0? :i E
£ _L -^ _L
III lill II
69
not all systems can compute quotes using different methods or compute indexes automatically and in real time. Finally, the integration of internet brokerage services also varies. The last group of characteristics in Table 3.9 refers to the facilities that are available for actualizing information and tools for evaluating stocks. All electronic trading systems in Latin America update in real time the information that they themselves generate and that they receive from external sources. Furthermore, all of them also provide access to historical data. However, only the ETS in the Santiago Stock Exchange offers a broad array of technical and statistical functions to analyze stocks and other traded instruments. Transparency Transparency is also an important component of the ETS since it also plays a critical role in the price determination (Demarchi and Foucault (1998)). Transparency is the degree to which information on securities quotes, transaction prices, and transaction volume is made publicly available. Transparency is usually divided into pre-trade and post-trade information. Pre-trade information includes information on the orders that are currently standing in the trading system and on prices at which incoming orders can be executed. On the other hand, post-trade information is the information on the details of past transactions, such as transaction price, transaction size, and the identity of parties involved in the transaction. Panel A of Table 3.10 presents the information available during the pre-trade phase in the ETS of the seven Latin American stock markets and four developed markets. As in the trading systems of developed markets, member firms get more pre-trade information than public investors in the ETS in the Latin America stock exchanges. Member firms can obtain the entire limit order book, and in the ETS of Brazil, Colombia, Mexico, and Venezuela they know the identity of the traders placing orders. Pre-trade information made available by the market organizers to investors also varies across markets. As in Xetra, in the ETS of Buenos Aires, Santiago, Mexico, and Peru public investors only know the best bid and ask prices. As in SETS, none of the electronic trading systems of the LAEM completely display all limit orders. Similar to the NSC, in Sao Paulo, Colombia, and Caracas public investors have access to five best bid and
70
ask prices. Furthermore, in the ETS of Argentina, Mexico, and Venezuela traders can submit hidden orders in the same way as in Paris.^^ It is common that order-driven markets open the daily trading operations with a call auction, usually known as a pre-trading session. Organizers of the pre-trading session can provide information on the submitted orders to the market and on the Indicative Equilibrium Price (lEP).^^ As in the trading systems of London and Paris, in the trading systems of Sao Paulo, Santiago, Colombia, and Caracas the order book is open. By contrast, the order book of Buenos Aires, Lima, and Frankfurt is closed. An lEP is published in Sao Paulo, Colombia, Bogota, Paris, and Frankfurt, while in Buenos Aires, Santiago, Lima, and London it is not.^^ Post-trade information of the analyzed ETS is shown in Panel B of Table 3.10. Similar to developed markets, all Latin American trading systems immediately publish the trades of small and medium orders. The information of large trades is published with a delay that varies from 48 hours (in Sao Paulo, Santiago, Mexico, and Caracas or Paris) to 72 hours (in Buenos Aires, Bogota, and Lima). Use of Call Auction Call markets are commonly used in developed countries because of the uncertainty in the valuation of assets generated by the overnight trade interruption. With call auctions the information is aggregated more efficiently. This facilitates price discovery in subsequent continuous trading. Therefore, informational efficiency might increase because overnight uncertainty diminishes (see Demarchi and Foucault (1998)). Call auctions in Latin American stock markets are not as commonly used to open and closed trade sessions of the ETS market as in the developed markets (see Table 3.11). The Chilean, Colombian, and Mexican ETS do not use an auction to open the market.
^^ Harris (1996) finds that hidden orders must be used more frequently in volatile markets, but also that the fraction of hidden orders and the size of the hidden portion increase with the volatility. ^ lEP is the price at which an order would be realized if opening occurred at that instant. ^^ In the Mexican Stock Exchange there is not an opening phase.
71
i? «S 2 o §»cnz
« {2 2 J2
w O
"gil
a) T -
O
I
CM CVJ C M
_^ h. r^ r^ ID
a ®
o
> c ^
S §
3
o o
^o
"(5 c5
o 13 o (0 0) E CD > • 3 S II
^
S E E E •8 E E E
J2
h.
0) •Q g r - T 0)
•o O
1 - - - T-
s
© « c • ^
a
•a
ol ®l (Ol
w.
*"
p
^
"(ol O
; (0 (0 (0
ixxi
T- T-
l
^1
2 K COQ .3? W 0) o .- - 1 -D ^ LU 0) H - ± r oI 8© T3 O O ili
.2 .2 «*-
CO . ^ (D
O . i -S ~ (0 Q. _l 2 I Q. O S
o
(d (0
•S ^ ^
i o 0) £ CD ©
-
w I
00 00 00
I^ o ^
5 o
2
0
0)
•S ^ "§
_
2l
;r T3 T3 "D O ® 0) 0
£ o ui g CD o
E '-
-•
S E E E
^ u.
s:
OT
w
w
•g X X X 0 CVJ CM CM _^ fs. r^. r».
© is z
1 PPP i2 Q
$
| 3 ^
o
0
on
0
o ;? T3 T3 T3 O 0 0 0
E -
2 E E E
0)
"2 o ^^^
o
0) ^ >2 i2 •g X X X
~ z
00 00 QOI " * -^ Tl-
S E E E
g .i .i .i (0
on o E -
O
-
^
T-
S
0)
52
0 •g
0
• >
O
o
o
a.
O
a. Q. CO
CD . ?
3 aj
111
o s
2 2 H
I-
Is I '5 M c
O 3 J« CO .2" 0
0 H_ •D O
zi :2 X
^
s
0)
o E 0 £L ZJ 2
Q- CO g l c c L J Q.
^ O
:2 X
.2 .2 •*o c3 S O £
CO w c CS
' S c 0
o o|
li: ^: P 73
"S X X X O
00 00 00
^ "^ "^ "^ I ppp C 0 0
0
•8 .^ .? .? O 0 0
0
2 E E E
s8
$
O
M m 0 "O c3 T3 O 0
.g D ^
1
1c 1c ."t^l "cl CO
1-
(0
0)
O £
JO fl) h- T3
CO
c
fO
0
These systems use the closing price as the opening price. The Sinac of the Buenos Aires Stock Exchange is the only system that does not have a module to organize an auction in order to close its trading session. In Latin America only the ETS of Sao Paulo and Mexico use intraday call auctions integrated with a continuous order-driven market, such as Xetra. Table 3.11: Use of Call Auctions Market Opening
Sinac Mega Bolsa Telepregon Electronic Sentra Elex Sibe
Market Closing
X X
X X X X
X X
Intraday
X
X X
Points
Note
1 2 1 1 2 2 2
0,33 0,67 0,33 0,33 0,67 0,67 0,67
NSC X 2 0,67 X X X Xetra X 1 3 Sets X 1 0,33 Points are obtained by adding the number of call auctions used by each electronic system. To compute the note, the number of points is divided by 3 (maximum number of points an electronic trading system can get). The information was collected from web-pages of the stock exchanges and from Demarchi and Foucault (1998).
3.3.8.3
Trading System Index
The interest in trading systems grew quickly due to its influence on price formation. Stock trading systems in each of the analyzed exchanges showed some similarities and differences. It is important to identify these because of their impact on the price discovery process, that is, on transaction costs and risk. The best way to grasp the quality of the trading systems and to compare them is to determine a quantitative measure, such as an index. Hence, the TS-Index is computed here for each stock exchange analyzed. The TS-Index is computed by averaging the scores of market segmentation and the scores on the electronic trading system. Before this computation is performed, the scores of market segmentation and electronic trading system are calculated by averaging the components of each (see Figure 3.5). For the score of market segmentation two components (types of stock orders and segments of the stock market) are aver-
75
aged, while for electronic trading system three (functions of the electronic trading systems, the use of call auctions, and information transparency) are averaged. The score of each component is given in a corresponding table, where it is also specified how the rating was computed. The components of the index are the most often cited in the literature or in the trading system descriptions of the stock exchanges. The way the components are aggregated does not have a theoretical or empirical foundation, but the results of the TS-Index are in accordance with the implicit trading costs of Domowitz et al. (2000).^^ Table 3.12: TS-Index Market Segmentation Orders Segmentation Note
Electronic'rrading System TSFuncIndex tions Auctions Transparency Note
Buenos Aires Sao Paulo Santiago Colombia Mexican Lima Caracas
0,70 0,80 0,50 0,50 0,80 0,50 0,70
0,6 0,6 0,5 0,2 0,6 0,3 0,2
0,65 0,70 0,50 0,35 0,70 0,40 0,45
0,67 0,78 0,78 0,50 0,78 0.78 0,89
0,33 0,67 0,33 0,33 0,67 0,67 0,67
0,38 0,85 0,62 0,62 0,69 0,31 0,92
0,46 0,76 0,58 0,48 0,71 0,58 0,83
Paris Frankfurt London New York Nasdaq
0,90 0,80 0,50
0,9 1 1
0,90 0,90 0,75
0,83 0,89
0,67 1 0,33
0,92 0,62 0,69
0,81 0,85 0,83 0,87 0,51 0,63
0,56 0,73 0,54 0,42 0,71 0.49 0,64
The TS-Index is an average of the notes of market segmentation and electronic trading system. Both notes are computed by averaging their components. The note of each component (orders, segments, functions, auctions, and transparency) are from their tables above.
Table 3.12 shows the TS-Index and its components for the analyzed markets. Among the Latin American trading systems, a significant heterogeneity can be detected. Sao Paulo, Mexico, and Caracas have the highest TS-Index, and Colombia and Lima the lowest, while Buenos Aires and Santiago are in the middle. Compared to the trading systems of developed markets, the seven Latin American systems have index ratings below those of the analyzed developed markets, except London. Trading segmentation shows the greatest differences among the Latin American and developed trading systems. The main source of this might be the lower volume and the smaller number of stocks traded in the Latin American exchanges. Furthermore, the electronic trading ^* To my knowledge, no TS-Index was developed in any previous investigation.
76
systems of the Latin American exchanges do not offer certain functions that are common in the developed markets. The reduced alternatives for trading and the absence of some functions of the ETS might increase trading cost, which in turn affects liquidity and the cost of capital in Latin America. If we compare these results with the implicit costs of Table 3.1, the results are meaningful. 3.3.9
Custody, Clearing, and Settlement Process
The clearing and settlement systems of stocks are an important component of the infrastructure of financial markets. This phase of the trading process begins after the brokers execute the orders. Although this phase seems to be administrative, problems can arise if one of the two parties does not fulfill the contract. A financial or operational problem in the clearing and settlement process could give rise to liquidity pressures or credit losses for other participants. Therefore, an additional risk can be generated for which the investors may demand a higher premium. When the clearing and settlement process is riskier in the Latin American markets, the systems used, the legal framework, and the payment systems are weaker. To quantify the risk of the custody, clearing, and settlement process in the LAEM and to compare them to some developed markets, an index is computed. The Custody, Clearing, and Settlement Index (CCS-Index) is made up of information on each of its three components (see Figure 3.4). The first central element analyzed is the Central Securities Depository registries. The CSDs are important because, before a stock can be traded on an exchange and before its ownership can be transmitted from one owner to another, the stock must be stored or registered in a CSDs. The clearing and the settlement are the second and third central elements, respectively. They are two relevant steps that should be followed in order to achieve an efficient and safe transfer of value (stock vs. payment) between the counterparts. In the next three subsections, the three elements of the CCS-Index are discussed, and in the final subsection the index is calculated and discussed. 3.3.9.1 Custody A central element of a clearing and settlement system is its CSDs (Guadamillas and Keppler (2001)). CSDs are important since they enable smoother and more efficient operations of book entry systems. Experience indicates that several features and func-
77
tions of a CSDs influence efficiency. First, the ownership structure is important due to the pressure management can place on parties in order to avoid default on transactions. Second, the additional services provided by the CSDs. And third, the form in which records of ownership are held influences the associated risks related to lost, stolen, or altered documents. We now turn to a discussion of each of these features and at the same time present the CSDs of Latin America and some developed markets. The ownership structure of the Latin American CSDs is similar to those of developed markets. In the LAEM and in most developed countries, CSDs are private firms constituted as self-regulatory organizations, which are overseen by the market regulator. CSDs have a close relation to the stock exchange because it is generally considered unacceptable for a stock exchange to cancel a transaction when, for example, a buyer or a seller could not meet his obligations. Although the direct participation of the Latin American stock exchanges in their CSDs is not as high as in developed markets, exchanges have a considerable influence in an indirect way. Usually the members of the stock exchanges are also shareholders in the CSDs (see Table 3.13). Similar to the services provided by the CSDs in developed markets, CSDs in most Latin American countries provide additional services, such as securities administration (cash dividends, interest payments, redemption, capitalization, exchanges, conversions and splits, and subscriptions) and issuer services (accounting registry of stockholders, corporate actions such as cash dividend payments, capitalization, subscriptions, exchanges, splits, etc., and updating rights). The mechanism for recording ownership of a CSDs influences the quality of the book entry systems. There are primarily two mechanisms for registering ownership. The first makes use of paper-based instruments with the physical delivery of certificates between counterparties, while the second system relies on computer-based transfer mechanisms in which records of ownership are held in so-called book entry form. Nowadays, most of the CSDs use computer-based mechanisms since they are more efficient and safer inasmuch as they eliminate the movement of paper. Thus, associated risks related to lost, stolen, or altered documents are reduced. In turn, computerbased mechanisms can be subdivided in immobilized and dematerialized systems. In an immobilized system, physical certificates are held in secure vaults and provide the essential support for book entry maintained ownership positions of market partici-
78
pants. In a dematerialized system, physical securities do not exist as they are replaced by book entry records. Table 3.13: Central Securities Depositories-Registries Shareholders Stock Ex- Brokerage pj^ancial Central Bank change Firms institutions
Merval (Buenos Aires) CBLC (Sao Paulo) DVC (Santiago) Deceval (Bogota) Indeval (Mexico) Cavali (Lima) C W (Caracas)
Issuers
Others
50,02037*
49.98 23,00
1,00
76,00
1,79 20,00 26,41
57,14 60,42 3,30
39,29 10,41 52,88
1,79 9,17 17,41
Euroclear (Paris) Clearstream (Frankfurt) 100,00 NSCC (NYSE) * 49,98 is held by the Mercado de Valores de Buenos Aires and the rest by the other exchanges of Argentina. The information was collected from web-pages of the stock exchanges.
The current state of ownership stock records varies across Latin American CSDs.^^ First, all CSDs except the Chilean use only computer-based mechanisms. The Chilean CSDs is the only one in Latin America where more than 10% of the physical shares are held in the form of paper. In the other six Latin American CSDs—just as in London, Frankfurt, and New York—most of the stocks are dematerialized, and a small percentage is held as physical stock certificates.^^ For identification of the stocks, all the CSDs use the internationally applied code system IS IN, except for Cajval and DCV. The use of a common code by all CSDs is important since it facilitates the identification of a stock and the reduction of transaction risk.
This information is taken from the Centre for Latin American Monetary Studies, 2000, Yellow Book for Argentina, Chile, and Peru, as well as from the web-pages of the ten stock exchanges and their CSDs. ^® Physical certificates are normally issued for special corporate events. In very few cases, there still are physical stocks that are stored in vaults.
79
CO CO
o
o
r^
r^
r^
r^
r^
1^
r^
O
CO
CO
CO
CO
CO
CO
CO
o"
o"
o"
o
o"
o
o
CO ^ •^ 0 O .N CO
o
III 2|
O
T-
1 -
c CO
"
«2 > c O
DC (Q •C £ Q. (0
2 a
To
(0 o
CO C
•Is "a ^
> O
Q.
c o w
O
O 0 CO
O ~
2 o i2 ©Eg)
^o -c O 2i
•Q -
c
"O
2
k_
0 U
O
CL'CO
D CO
o 0
.C "•- Q. 0>
O
CO 0 • • ^
S S 0-g
D
X3 .2 t !
CO
0
0
CO
Z
CO LL O
0
Z. Q. c ^ /l^ CO 0 Q) CO © CO "P
O co" • K "O CO c
o w
P >
-f„
|5
According to the international trend, six of the Latin American markets (all except Santiago) have an electronic system for deposits and transfers. Each depositor has access to the automated operations via remote terminals. The users can electronically manage their securities, operate and control trades executed on the stock market or in over-the-counter trading. The systems also provide modules for the electronic transfer of securities, so that brokers can record, confirm, look up, and settle trades directly. DvP-Transfers, checking of trade lines, credit and trading accounts, money transfers between the CSDs, and the central bank payment systems are also automated processes. Only CBLC (Brazil) and Indeval (Mexico) have created an electronic system to arrange securities loans.^^ 3.3.9.2 Clearing Process The first step of the third phase is the clearance of trade. Assuring that safety, soundness, certainty, and efficiency are achieved at an acceptable level of cost to all participants, clearing is an important part of the financial infrastructure of a nation.^^ This step includes trade capture, matching, confirmation, comparison and affirmation mechanisms, and the calculation of settlement obligations (Guadamillas and Keppler (2001)). The obligations of the direct market participants (brokers, promoters, specialists, etc.) to deliver securities and funds are determined in this step. There are two methods for computing settlement obligations. The selection of one of them is very important because of their impact on the efficiency and risk exposure of the system. Their weaknesses and strengths are the result of a trade-off between liquidity requirements and risk mitigation (for a discussion see Guadamillas and Keppler (2001)). Both methods are discussed below. The calculation of settlement obligations is the most important task in this step and can be effected on a net or a gross basis. Netting can be done bilaterally or multilaterally. By netting on bilateral basis the credits and debits are offset pair by pair. In multilateral netting mechanisms, an agent's transactions (both buys and sells) are combined at the end of the clearing cycle, and only the net amount is settled. As a result. ^' For complete descriptions of their automated processes, see the web-pages of the relevant CSDs. ^^ However, settlement and the other components of the trading architecture are also important and with the analysis in this and following chapters cannot be answered which element is more important. While this question would worth pursuing in further research, the goal of this chapter is solely to construct indexes with the characteristics identified in the literature as the most important. The development of
81
the settlement amount is reduced. So a smaller amount of intraday liquidity will be needed. However, the interdependence of the different transactions increases thereby, and thus also the domino effects of a failing delivery. So there is a trade-off between costs and systematic risk. Due to its characteristics, multilateral netting is the most common used system. The other alternative for calculating the settlement obligations are gross settlement mechanisms. These systems compute each transaction separately. Such methods have the advantage of reducing the interdependence problem of netting. Thus, if transactions are settled on a gross basis, credit and liquidity risk would diminish. However, the use of these systems gives rise to several costs (Niemeyer (2001)). First, the total cost will increase as a consequence of a higher amount and number of transfers. Second, the costs of capital to which participants are exposed will rise due to the greater amount of capital needed to handle imbalances. Third, the higher costs might limit the intraday trading, causing a lower market liquidity; and fourth, payment problems may become more frequent as a consequence of the disadvantages mentioned above. The computational method used to determine the net volume and value of stocks and cash to be exchanged between counterparties varies among the CSDs. First, the cash credits and debits to be transferred are computed on a multilateral netting basis in all the CSDs. By contrast, the netting variants for stocks vary across markets. Like in Euroclear and NSCC, the CSDs of Buenos Aires, Mexico, and Caracas effect their netting on a multilateral basis, while only in Santiago it is done on a bilateral basis. Similar to Clearstream, the CSDs of Brazil, Colombia, and Peru compute the stock net volume and value of stocks on a gross basis. 3.3.9.3 Settlement Procedure The last step of the third phase is the settlement. It involves the discharge of settlement obligations through the final transfer of funds from the buyer to the seller, and the final transfer of securities from the seller to the buyer. Three aspects, which will be discussed below, are very important in the settlement phase: Delivery versus Payment, settlement assurance, and the time a trade takes to be settled.
indexes is a practical and implementable assessment methodology encompassing key issues that have
82
Table 3.15: Clearing and Settlement Clearing Cash Net- Stock Netting ting Buenos Aires Multilateral Multilateral Gross Santiago Multilateral Bilateral Colombia Multilateral Gross Mexico Multilateral Multilateral Lima Multilateral Gross Caracas Multilateral Multilateral Paris
Multilateral Multilateral
Frankfurt
Multilateral
NYSE
Multilateral Multilateral
Gross
Settlement Transfer Transfer Settlement Foreign of Stocks of Funds1 DvP Model Time Investors BE BE Ph/BE BE BE
NPS Model III NPS Model 11 NPS Model III Model li&l NPS Model III Bank Model II Model Ml
T+3 T+3 T+2 T+3 T+2 T+3 T+3
Model III Beetween Model ll&l Model III
T+3
BE
NPS
BE
NPS
BE
NPS
Custodian Custodian Custodian Direct Custodian
T+2 T+3
Ph stands for Physical, BE for Book Entry, and NPS for National Payment System. The information was recollected from web-pages of the stock exchanges.
Delivery versus Payment Settlement can be carried out using a variety of procedures, but the most common is Delivery versus Payment (DvP). The DvP procedure is important for security transactions because it can help to mitigate counterparty risk, which is one of the most important in the settlement process. A reduction of this risk can be achieved by ensuring that sellers give up their securities if, and only if, they receive full payment and vice versa. According to Guadamillas and Keppler (2001), a DvP transaction contains three essential elements: a good and irrevocable delivery of securities, a final and irrevocable delivery of funds, and a simultaneous exchange. The transfer of stocks can be carried out physically or using the book entry method. The latter is usually employed in developed as well as in Latin American markets. On the other hand, cash can be transferred electronically through accounts within the national payment system or through a bank that assumes the role of cash settler. The first alternative is employed in Argentina, Brazil, Chile, and Mexico, while the second is usual in Colombia, Peru, and Venezuela (see Table 3.15).
an impact on the quality of the treading architecture of a country.
83
To the theoretical aspects of DvP in depth, the Bank for International Settlements (BIS) has identified three different basic models for its application. In Model I, the system settles instructions to transfer stocks against simultaneous payment on a tradeby-trade basis throughout the day. In Model II, stocks are transferred and settled on a trade-for-trade basis throughout the processing cycle, while funds transfers are settled on a net basis at the end of the processing cycle. Finally, in Model III, systems settle instructions to transfer both funds and securities on a net basis, and the transfer of funds and securities occur at the end of the processing cycles. Given the market conditions of the LAEM, the selection of one of the models is not straightforward. Model I, on the one hand, reduces the potential for settlement failure by reducing the progressive build-up of credit exposure between participants and on the other hand, it requires a critical mass of marketable securities and system-wide liquidity for its efficient operation (Guadamillas and Keppler (2001)). The liquidity needs may represent a restriction to the implementation of this models in the LAEM. Model in reduces the volume and value of the final transfers; however, it can increase the systemic risk.^^ Under these considerations, Model III with periodically settlement during the day would be an interesting alternative, which could be analyzed in further research. The DvP-systems applied in the Latin American stock exchanges must be distinguished. The stock exchanges of Buenos Aires, Santiago, Mexico, and Caracas settle funds and stocks using Model III, as in the Paris Bourse and the NYSE. Sao Paulo and Lima use Model II, and only Bogota and Frankfurt apply a mechanism intermediate between Models I and II. Settlement Time Another important aspect of settlement is the time necessary for a trade to be settled. It is widely recognized that the longer it takes to settle a securities trade, the higher the risk that the settlement may not be realized. Consequently, a chief goal in the design of settlement systems is to shorten the time between trade date and settlement date.^"^ The duration of the settlement cycle in developed and Latin American exchanges var^^ There are some mechanisms such a novation (see below) to reduce the systemic risk, but they require a legal environment that support them. ''^ However, shortening the settlement time generates some disadvantages. For example, the number of trades that fail to settle may increase (see Guadamillas and Keppler (2001)).
84
ies between two and three days after the trade date. As in Frankfurt, the settlement of domestic trades in Santiago and Mexico is effected within 48 hours (T+2), while it takes 72 hours (T+3) in other exchanges. Therefore, the G30's suggestion that trade settlement should occur by T+3 or less is fulfilled by the LAEM. Settlement Assurance Due to the risks that arise in this step, settlement assurance is an important concept and can be understood as the arrangements by which a system tries to remove counterparty risk (principal, replacement cost, and liquidity risk) from its participants (Guadamillas and Keppler (2001)). There are different ways to achieve settlement assurance. In case of default, fraud, or negligence on the part of a member, the DvP mechanism reduces only principal risk. Therefore, a CSDs has to implement additional preventive measures. To do so, CSDs usually take the stocks from the member who defaults and sells them inmiediately on the open market. But the counterpart or the CSDs (by novation, see below) will thereby be exposed to market risk. To mitigate the risk, CSDs have several alternatives or a combination of them: (1) the CSDs may liquidate the securities that the defaulter may have on deposit at the Custody Institution. (2) If the generated revenues of the liquidation are insufficient, CSDs in most countries maintain a guarantee fund, which can be financed with a percentage of CSDs-profits and/or with member contributions.^^ (3) CSDs also demand collateral to reduce market risk. All members are required to buy shares of the CSDs and pledge them, for example, in favor of an insurance company and/or of deposit provisions. If a member fails, the CSDs would try to sell the collateral to cover any shortfall. (4) Most of the CSDs operate stock loans facilities that are used mainly to reduce the impact of shortfalls in the settlement process. And (5) some CSDs stipulate operational limits, which are based on an analysis of the member's financial and economic standing. In addition to the mechanisms mentioned above to reduce risk, novation is a very important mechanism. In systems that include novation, the original trade between two counterparties is split in two separate trades (Bemanke (1990)). The clearinghouse ^^ There are only two funds in Argentina: The first, the Guarantee Fund, is financed with 50% of the CSDs profits and is used if a member does not fulfill his obligations. The second, the Special Guarantee Fund, is financed with the contributions of members and is used to reimburse investors in case of gross stockbroker negligence or fraud.
85
becomes an official party to every trade, substituting itself as a seller to every buyer and a buyer to every seller. Therefore, counterparties' exposures to one another are extinguished and replaced by exposures to a central counterparty. This facilitates multilateral netting and provides the basis for counterparty risk guarantees. However, novation causes a default risk to the CSDs. Therefore, novation should be implemented with a combination of the preventive measures explained before. Table 3.16 shows a list of different settlement assurance procedures used in the selected stock exchanges.^^ According to Table 2.12, CSDs can be subdivided into three groups. The first group is Merval with Euroclear, Clearstream, and NSCC, where settlement is guaranteed (i.e., novation is implemented) and the mechanisms for reducing market risk are similar. The second group is integrated by CBLC and Indeval. In order to diminish settlement risk, these CSDs have implemented a guarantee fund, a collateral plan, and a money and stock lending mechanism. Finally, the third group is composed of DVC and Cavali, where buy-ins and sell-outs have been implemented, but a guarantee fund is still under construction and securities lending is generally not regulated, even though there are specific sets of rules for short sales and for the lending of shares. In these CSDs, if one of the counterparties defaults, that counterparty can be punished, which may involve the cancellation of his right to trade in the exchange. For resolutions, such penalty can be maintained until the affected member receives the agreed stocks or cash plus interests.^^ 3.3.9.4 Custody, Clearing, and Settlement Index As discussed above, the systems of clearing and settlement are important components of the financial market's infrastructure because they could give rise to liquidity pressures or credit losses for market participants. Therefore, it is necessary to have an index that compares the clearing and settlement systems used in the LAEM with those of certain developed countries. Table 3.17 presents the CCS-Index of the LAEM and three developed markets. It includes the three components discussed in the previous section: (1) the CSDs, (2) clearance of stock trade, and (3) settlement. For each of these components a rating is
^^ The stock exchanges of Colombia and Caracas have not been included here because it was not possible to find enough information on their settlement assurance procedures. ^^ See the Yellow book of Chile and Peru.
86
calculated based on information on some variables7^ In Table 3.17 the variables use for each component can be seen. Also in the lower portion of Table 3.17, an explanation is given of how the ratings of the three components are calculated.^^ Then the CCS-Index for each stock market is computed by averaging the notes of each element. The results are conclusive: the custody, clearing, and settlement process varies across the CSDs of developed and Latin American countries. According to the CCS-Index, the custody, clearing, and settlement process is safest in New York and Mexico.^^ Then follow Paris, Frankfurt, Buenos Aires and Sao Paulo. However, the reasons for the lower CCS-Index in these financial centers are different. Buenos Aires should enforce the functions of its CSDs, Sao Paulo should improve the clearing and settlement process, in Paris the score on settlement is lower than in the other developed markets. Table 3.16: Settlement Assurance Novation Q " ' - * - Co..a.era. ^^^^^
Merval (Buenos Aires) CBLC (Sao Paulo) DVC (Santiago) Deceval (Bogota) Indeval (Mexico) Cavali(Lima) CW (Caracas)
X
Euroclear (Paris) Clearstream (Frankfurt) NSCC(NYSE)
X X X
X
X X
X X X
X X X X
X X X*
2 1 0
X**
1 0
X X X
2 2 2
X
X X X
Note
The Note is determined by assigning 2 if novation exists; 1 if at least two elements (Guarantee Fund, Collateral or Stock/Money lending) exist to guarantee settlement, and 0 otherwise. The information Was collected from web-pages of the stock exchanges. * A Negotiation System does not exist and securities lending is generally not regulated. There is a specific set of rules for short sales and for lending of shares. ** A Negotiation System does not exist.
^^ The rating of a CSDs does not include variables of ownership and services since they are quite similar in all the markets. ^^ When assigning points to the variable "DvP Model," multilateral netting gets 2 points, bilateral netting 1 point, and gross settlement 0 points. These point assignments were decided upon because the LAEM have considerable problems with liquidity. In section 2.3.2 above, the advantages and disadvantages of the different methods of computing settlement obligations were discussed. As was noted, multilateral netting might be the best alternative for markets with low liquidity.
87
0) (/) •o o ^
CNJ CM 0> 00 CO O CO r^ r^ CO m 00 l o 00 o o o o o o o
X
*
CO O O 00 i n m o o o
O
z
0 0 CM CO 1 ^ r^ 0 0
o o o
g o «»
ill I
r^ CO CO k o 00 00
CO CO 00 CO o o
"2 00 S
o o o
Q) O
c 4^
3
c o
(/}
CVJ 1 - O
T-
O
E 0)
e0>
1-
1-
T - OJ T - CJ "•- 1 -
1^ r^ o
*©
"o z
CM
o
L_ (O CO U) ^
o" o o"
T - CM 1 -
lO T-
o"
r-
> 60 O b
' 2 .S w ^ 2 V. u a o
111
O)
-c c 9 explain the cross-section of average returns. Based on market capitalization, Banz divided NYSE stocks into quintiles for the period 1931 to 1975. The smallest quintile (stocks with the smallest market capitalization) showed higher average returns than the other quintiles and indexes. Hawawini and Keim (1996) grouped NYSE- and Amex-stocks into ten value-weighted portfolios for the period 1962-1994. Their results showed a clear negative relation between size and p-. the smallest size portfolio offered 8.8 % higher returns than the largest one. As for the U.S. markets, numerous studies have documented the existence of size effect in other developed markets. Chan et al. (1991) showed a significant negative relation between size and expected returns for the Tokyo Stock Exchange. A size effect was documented by Hawawini and Viallet (1988) for the stock market of France, and by Corhay et al. (1988) for the UK. Recently, the book-to-market effect has gained the attention of researchers. The B/M ratio has been shown to have a significant predictive power with respect to the crosssectional average returns in some stock markets. The B/M effect was first documented by Stattman (1980). He discovered that average returns on U.S. stocks are positive related to the B/M ratio. Rosenberg, Reid, and Lanstein (1985) have confirmed this result. With a sample of all nonfinancial firms listed on the NYSE, Amex, and Nasdaq, Fama and French (1992) show that the relations between average return and size of book-to-market are the strongest. Moreover, they argue that "although the size effect has attracted more attention, book-to-market equity has a consistently stronger role in average retums."^^ Chan et al. (1991) have shown for Japan that the book-to-market ratio is statistically and economically the most important of the four variables they *^ See Fama and French (1992), page 428.
97
analyzed in order to explain the cross-sectional average return. The book-to-market effect has also been documented at the stock exchanges in London, Germany, France, and Switzerland (Hawawini and Keim (2000)). Fama and French (1998) confirmed the B/M effect for thirteen major countries. The tradition of earnings-related investment strategies is old, but Basu (1983) was the first to document it empirically. He studied the relation of the ratio £/P, market capitalization, and return on the NYSE for the period 1963-1980. Basu shows that, on average, high E/P stocks have higher risk-adjusted returns than low E/P stocks. In addition, the E/P effect is significant even after controlling for firm size. Fama and French (1992) also find a strong univariate relation between the average returns and the E/P ratio. For the Singapore Stock Exchange, Wong and Lye (1990) find that the size effect is apparently of secondary importance, when it is compared to the E/P effect. Regarding to stocks of both sections of the Tokyo Stock Exchange, Chan et al. (1991) show that high E/P stocks outperform low E/P stocks. However, there is less evidence of an E/P effect in other developed markets.^^ Another factor that is gaining in importance is Turnover. Amihud and Mendelson (1986) show that cross-sectional expected returns are an increasing function of the turnover for stocks traded in the U.S. markets. Brennan et al. (1996), Haugen and Baker (1996), and Teh and De Bondt (1996) confirm that stocks with higher trading activity tend to have lower expected returns. Hu (1997) finds that the Japanese stock market supports the results of Amihud and Mendelson (1986). Using a cross-sectional analysis, Hu discovered that stocks with higher turnover tend to have lower expected returns. He used data from the Tokyo Stock Exchange from 1976 to 1993, and he used stock turnover as a measure of liquidity. 4.2.2 Evidence of Emerging Markets The trade-off between the expected returns and their associated risk determines the benefits to an equity investor in the stock markets. Many factors must be considered for the assessment of this trade-off in the LAEM: the underlying factors determining the rate of return and its variability, the efficiency of domestic stock markets, the regulatory, accounting, and enforcement standards, the different forms of transfer risk
Hawawini and Keim (2000) present a complete survey of papers in which the E/P effect is examined.
98
(e.g., barriers to the repatriation of capital), the ability to invest in a country, taxes and other transaction costs (Claessens (1995)). In recent years some papers have analyzed the relation between risk and returns for emerging markets. They were carried out at country or firm level. On an aggregate country level, the most frequently cited papers are those of Harvey (1995), Bekaert (1995), Bekaert et al. (1998), and Fama and French (1998). Harvey (1995) documented that emerging markets exhibit high expected returns accompanied by high volatility. He also pointed out that the stock returns of a Latin American stock market are low correlated with other markets. Their low correlation with developed markets has an advantage: it reduces the unconditional portfolio risk of a global investor. In addition, he shows that the standard global asset pricing models fail to explain the cross-section of average returns in emerging markets and that the returns of these markets are more likely to be influenced by local information than in developed countries. On the other hand, Bekaert (1995) examined nineteen emerging equity markets and observed that stocks with higher >^ offer lower expected stock retums.^^ Forming two portfolios, Bekaert et al. (1998) examine a set of risk factors for several emerging markets.^'* They register a size and a price-to-book effect. The price-to-eamings effect, however, is not as clear as the former. Finally, Fama and French (1998) also tested for book-to-market, eamings-to-price, and size effects in sixteen emerging markets. Their results conform with the evidence on developed markets. Small stocks offer higher average returns than big stocks. Fama and French (1998) also found an average difference between annual dollar returns on the high and low book-to-market portfolios of 16.91% when countries are value-weighted and 14.13% when countries are equally weighted. The value-premium is less reliable when they sort according to eamings-to-price. On an aggregate stock level, the results of the different studies are contradictory. Rouwenhorst (1999) ranked stocks by country according to local P, S, prior six-month return, B/M, and TO. From 1982 to 1997 stocks were sorted each month into three equally weighted portfolios. For this period Rouwenhorst (1999) shows that the return Furthermore, he develops a return-based measure of market integration and analyzes the relation between this measure, other return characteristics, and investment barriers. He concludes that each emerging market shows different degrees of market integration and that the differences are not necessarily associated with barriers to investments.
99
factors in emerging markets are similar to those documented for many developed markets. On average across all emerging markets, stocks exhibit momentum and small stocks outperform large stocks, and value-stocks outperform growth stocks. Rouwenhorst (1999) does not find that average returns are related to share turnover. Claessens et al. (1995) also investigate the presence of some return anomalies in twenty emerging stock markets. They use monthly data from the EMDB. Their sample period depends on the stock market.*^ For each market the authors rank stocks at the beginning of each year by market capitalization. Thereafter, they assign stocks to one of four portfolios and calculate the return on these portfolios for the subsequent twelve months. At the end of each year, they recreate the portfolio scheme. In this way, they generate a return time series for each portfolio. Their results show that the Latin American markets may have a size effect, but it is not necessarily restricted to the smallest size portfolios. It was only in Mexico that the small portfolio had significantly higher annual returns than the portfolio with the largest stocks. Using an F-test for equality of returns across all portfolios, they also investigated whether the size effect is significant. Their results show that any return difference between the four portfolios was statistically significant at the 5% level. In summary, the results of Rouwenhorst (1999) and those of Claessens et al. (1995) contradict one another. Some of the reasons for the contradictions are: different sample periods, different numbers of portfolios, the frequency with which stocks were sorted, and differences in the computation of portfolio returns (equally or valueweighted). For these reasons, the causes that originate the contradictory results are discussed in this chapter.
4.3 Database Most of the data used in this chapter are from the EMDB of the IFC-S&P. For emerging markets, the EMDB provides current and historical statistics of series at stock, index, and market levels. Using different samples of stocks, the IFC-S&P calculates different indexes for each stock market. Among its wide range of indexes, the IFC Global Index (IFCG) and the IFC Investable Index (IFCI) are the most important for
*^ They sorted the countries by risk factors into portfolios. Then they computed equally and valueweighted returns for each portfolio and rebalance them four times per year. *^ The time series start in 1976,1986,1987, or 1990, depending on the country.
100
our purposes. For differences among these two indexes, see Figure 4.1, below. For both indexes, the IFC-S&P computes a vast number of alternatives.^^ IFCGobal(IFCG) Q
BroadBSt ccnsftuBntbase
D
Covers 60 % of tt ale xchang9 market C£pJtBl izafon
D
No acjustments forforeign irvestmentrestricioTB
Q
Daily calculEtiGn
•
Indistry W e)©sc£icdated morthly
IPC Inves table (IFCI) •
Forei gnl n vest men t restrlcio ns in corpora ed at oonstiLiert level
\j
9 z e a n d iqudity screens
•
Corporste cross-hold ng& gDvernment ownership adjusted
Q
Daily calculEt icn
G
Indistry hd e}esc£(ciJated nrx)rthly
IPCGF rentier Markets •
Introducedin 1996
D
20 markets
[]
IndudBd if trading occurs regUarly;a cfvitytypicsllythn
•
G r a d u d e b IFCGdaly series when iqiidity increases
Figure 4.1: IFC-S&P-Indexes Source: The IFC Indexes: Methodology, Definitions, and Practices, IFC, July 1999
For their indexes, the IFC-S&P selects stocks as follows. At the date a market qualifies as emerging, the IFC-S&P begins the stock selection process. The first step of this process is a survey of the market and of all listed companies and shares. While this survey is repeated each year during an annual review process, additions and deletions outside the annual review are nevertheless possible. In the second step, the guidelines and profiles are defined in order to capture the real market in the IFCG Indexes. To ensure it, the IFC-S&P uses three criteria: First, actively traded stocks; Second, target market share of 60-75% of total market capitalization; and Third, industry diversification. Finally, in order to graduate from index coverage, two criteria must be met: (1) GNP per capita for an economy should exceed the World Bank's upper income threshold
For a description of the indexes, see the EMDB Guide and The EFC Indexes: Methodology, Definitions, and Practices.
101
for at least three consecutive years, and (2) the investable market capitaHzation to GDP ratio must be near the average of developed markets for three consecutive years. The most popular indexes of the IFC-S&P are the Total Return Indexes, which are calculated in U.S. dollars (IFCG.TR.US) or in local currency (IFCG.TR.LC) for each country. The IFCG.TR.US are commonly used as a benchmark of the national stock markets since they are consistent across national boundaries. IFCG.TR.US indexes include all stocks used in the EMDB of the corresponding country. The IFCG.TR.US is computed by taking into account stock dividends and assuming that dividends are reinvested across the entire index portfolio in proportion to the capitalization of all stocks in the index. The IFCG.TR.US indexes are market capitalization weighted. Its period data are linked by the chained Paasche method. These indexes are calculated on a price only and a total return basis, using end-of-week and end-of-month data. In addition, it is common for firms in Latin America to issue more than one class of shares. Then, according to its trading activity, one or more classes of stocks are included in the IFCG-Index. Only the market capitalization of the selected classes are included; not the entire capital of the company in question. ^^ In addition to their indexes, the IFC-S&P also publishes data used for the computation of their indexes and valuation ratios. The principal time series published in the EMDB include: closing price, number of shares outstanding, number of shares traded, dividends, and new issues of shares. The IFC-S&P also computes three valuation ratios for each stock and index: Price-to-eamings, price-to-book-value, and sash dividend yield at a stock and index level.^^ Although the database of the IFC-S&P is very popular among researchers and practitioners, it suffers from three sources of bias. First, the EMDB has some data processing problems. For instance, sometimes the values are false or the database shows zero when the entries have insufficient digits. Second, the stock and country selection criteria force one to choose the most frequently traded and larger stocks of the most successful countries. This winners selection induces a reemerging bias (Goetzmann and
^^ The IFC-S&P excludes mutual funds and other investment trusts. ^* In: The IPC Indexes: Methodology, Definitions, and Practices is described how they are computed.
102
Gorion (1996)). Third, the IFC-S&P began collecting infonnation in mid-1981 and gathered data extending back to 1976 for the stocks of ten markets.^^ Most of the information used in the present chapter is from the EMDB. The following are taken from the EMDB for each country: closing price, outstanding shares, market capitalization, number of shares-traded (for the turnover), price-to-book value, priceto-eamings, the capital adjusted rate (CAR), and the IFCG.TR.US.^ Only interest rates are taken from the International Monetary Fund (the deposit rate for Argentina, Chile, Colombia, Peru, and Venezuela; the money market rate for Brazil; and the treasury bill rate for Mexico). The period analyzed runs from May 1986 to November 1999 for all countries, except Peru. The IFC-S&P began to collect information on Argentina, Brazil, Chile, and Mexico in 1981, for Colombia and Venezuela in 1984, and for Peru in 1993. However, the period analyzed here begins in May 1986 because the IFC-S&P began to publish P/BV and P/E at this time. Another reason we begin in 1986 is to avoid the first years of data for each country, since they suffer from a back-tracking bias. For Argentina, Brazil, Chile, and Mexico the IFC-S&P has published information on the firms selected in 1981 since December 1975. For these five years, the EMDB might suffer from a back-tracking bias, since successful firms in 1981 were not necessarily the same as in 1975. Indeed, a firm that was successful in 1976 and had faired badly by 1981 is not included in that database.
4.4 Characteristics of the Analyzed Stock Markets According to the IFC-S&P's definition of an emerging market (see Chapter 2), all Latin American stock markets are emerging markets. Unfortunately, the IFC-S&P collects information only on some Latin American markets and of those, information
This is one reason why the present study starts in 1985. Information dating back to 1982 is used only for the calculation of y^. ^ The IPC uses the last transaction price recorded at the stock exchange to determine closing price. The IFC-S&P adjusts book values between balance sheet report dates by the amount of capital raised by rights issues. In hyperinflationary economies, the IFC-S&P also adjusts earnings and book values. A number of adjustments to market capitalization are made by the IFC-S&P in order to approximate more closely the amount of a company's stock that is traded in the market. This includes recognizing statutory limits on foreign ownership, removing capital owned by the government, and removing ownership by other constituents (cross-holdings). In addition, the IFC-S&P makes adjustments when a company issues new shares, declares a rights issue, stock splits, or stock dividends. Furthermore, the IFC-S&P corrects its series for several kinds of unusual corporate actions. For a complete discussion of the ad-
103
on the smaller markets is incomplete. Therefore, the stock markets analyzed here number only seven: Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela. Some summary statistics for these markets are provided in Table 4.1. From Table 4.1 one can see that the number of firms and the market capitalization in each LAEM are small compared to some developed markets, such as the stock exchanges of New York, Japan, London, and Frankfurt (NYSE, TSE, LSE, and FSE). However, the number of stocks in the EMDB are comparable to the MSCI portfolios for developed countries.^^ It is also surprising that the capitalization in some Latin American markets is larger than one might have suspected.^^ The total market capitalization of the LAEM was US$ 434.4 billion at the end of June 1992. From June 1992 to June 1999 it grew 65.7%. Despite this growth, the weight of the LAEM on the total market capitalization of the IFCG Composite Index diminished from 35% in 1992 to 22% by the end of 1999.
4.5 Return Characteristics The monthly stock returns for the Latin American stock markets under consideration here will be analyzed in this section. In order to compare the returns across markets, they are calculated in US Dollars. In Table 4.2 some properties of the monthly returns are presented. The monthly mean returns range from 3.62% for Argentina to 1.35% for Peru. However, the standard deviation seems to be correlated with the mean returns. Argentina has the highest standard deviation, and Peru has the second smallest. Compared to developed markets, the average returns and standard deviation of the LAEM are often higher.
justments made to the series, see 'The IPC Indexes: Methodology, Definitions, and Practices," IPC, July 1999. ^^ See Harvey (1991). ^ Several factors have been important in the increase of capital flows going to the LAEM. The decline in international interest rates (Calvo, Liderman, and Reinhart (1993)), improved domestic policies and better growth performance (Chuhan et al. (1993)), as well as market liberalization (Claessens and Rhee (1994)), have encouraged the increase of capital flows to these markets.
104
-^ q^ ^ eg ' * OJ CM
y
^
hCO
' ^" 00 CO CM 00
-^
O CO 00 CO (7) CO
:2; o 9.
T-
co ^
00
*• •'
IT)
o
'^
C3>
JS
in
^
CO
"^
"^ K
LO
CM 00 CO O
J^
CO
-^
'^
1^ '
r>. r^
O)' LO
r^
in • ^
o> o
"^
CM CO ^
CM CO
11
00
CO C7)
co"
co" 00 in
CO in
% s
o
O) CD" O
CO
CO CM
o -§
TJ
1
O in
2
o ON ON
CO CO
III o 3 lU (0 'w
^
O
o o
o
S
g
in
o> 9
Si
CD o
o o
O)
o o
00
en o
00 CO
o o o
z:
o
00
^-
O
O
CO
E;5
o
CO
^:
CO
CO
1-
o"
9
00
3
CD
o o
^ 9"
00 00
CO
& a
"I•«t 1-
O CVJ CVJ
o"
o"
o o"
CVJ
o"
CO
^
00 CO 0 )
co"
in
00
CVJ 0
10
oo"
o" o o o o o"
in
s
3
o o o"
1-
°0
00
S
S
8
rt
$2
o'
CO (O
CO
o o o o o"
§
1-
'-
8
00 CO
O O"
TT-"
CO CVI
§
o"
o
o
CVI (O CM
c:>
CO
00 CO in
r^
in
y
CVI
o'
o"
00 00
00 0>
o o o"
•^ O
o"
ffi
c o
O
Ho" Hcvj Hco CO
o
y-
o o" o o o'
T-
CM
in
o"
o" 9 o" CVJ 1 O 1 - CM CVJ
o" o" o" o o CO ' ' "
^
^ o o o" 9" 9" Oi O _. CO TJ- CO O 11-
o" o" o" o o o CVI ;St i:^ lO 00 00
o o o o" 9" 9^ o o ol CD ^
CVJ 0 _ T-_
o" 9
9
CO
o o"
o"
8S|
8
CO in O -*
CO r^ T1-
o CVJ 1^ in
9
co"
T-"^
o"
M-
00 CVJ in
CO 00 00
CO 00 o
CD"
O"
S ts
o8
o o
t00 CM
O Q°
o o o o" o" o"
in
o
CM
''t
00 O) 00 O
CVI
in
o"
co"
CVI
> o
o c
'M
0)
W
CVI 0 0 "^ O
E c
s o
3
E
>< s
O
O '"»
o" o" 9
in
®
o
o
o o o c^ JP J^ CO J ^ CO CVJ O ^ "r-_
o" 9
0)
S
o
-9
3
o" (3
-S 2
^
-^
Q.
o
CO 0 0 - ^ "^
«B
•c 3
9
9. o o 9 o" o" o 1^
c ,s:f
h-;
0>
<J) h- 1^
in
o "^"
^1
" ^ !:;
cvT
8
!"«?f"
o
CO
in
? ;: CJ> -^
'•5
106
o
o o o o i^ in
08
H
o
h - h". CVI CVJ
o o
CD Oi 00 00
CVI
1 ^ :S
o" o" o"
2
in CVJ O
rl
o
CD CO O 00 TO
CO
!Q
I
CVI CO CJ)
CVJ CO
00 -^ T-
o "7
CD
Ho
o o
o o"
CO
"^
CO i n
T- 00 Osj i n CO i n Tj- i n CO o" o o" o ' 1 - CO CO CO T- hi n h- CO h- CO o~ o " o " o " o "
o
o
o
o
00
o" "0
-o o
i o
r- CO O
o o"
S 5 S oj « S ??J9^V 1^1 OOdI o" o" o" o" o" o" o
CO i n i n cj) CO cvj cvj ^ . ^ . O ^ - ^ . CD
o
a 0 0 0 0 0 0 0
5 2 ^ ° ^ ^ ;i!e>|>||N'UBdBr o" o" o" o"
CO i n i n N T- T- Tt "^ "^ CO Tt Tt '^t Q
o o o o o o o
1!
co LL CO i ^ =) Z)
§5
tt"
^ 3
:S 2
ON ON
CO o j 00 CO r^ T- CO m "^ -^ 1 - 15: Tt c^ cvj SOOS , o" o" o" o" o" o"
1—»
d^S
STl
13 C
^ 8. -^ fc ^ 0
u
o" o" o o' o" o
O
o 2 o ^ LO LU CL CO
H
i n ! ^ S ;= 5* S ^eisvOOdl o o" o" o' o" o" o"
o
i n CO C7> N. CO r^ r«co i n i n m "^t -^ i n o" o" o ' o" o o" o"
o
CO S S SD S S coV3l/M3 OOdI o" o" o" o" o o" o"
ooooooo o o U. o LL o LL o Li. o LJo Li. LL
CO
(0
(1) N ^
O 'x
0
< en o o 2 I > 109
these characteristics cause the LAEM to differ from other markets and because the LAEM offer independent samples, it is worthwhile to study the LAEM.
4.6 Methodology 4.6.1 Portfolio Approach The construction of portfolio is a common practice in empirical finance to analyze the link among stock returns and some variables such as 5, P/BV, P/E, and TO (see for example Fama and French (1992, 1993, 1998), Banz (1981), Hawawini and Keim (1996), Harvey (1995), Bekaert et al. (1998)). The portfolio approach implemented here is based on Rouwenhorst (1999) and Claessens et al. (1995) and it consists of 5 steps. (1) Stocks are ranked according to one observable firm variable (here, 5, P/BV, P/E or TO) at the beginning of a given period (here, of each month or year). (2) The stocks are grouped into the portfolios (here, three or six). (3) The returns (here, equally or value) are computed for each month and each portfolio. Using this process, a return time series is generated. (4) For each portfolio its average return, volatility, and its average TO, P/E, P/BV, 5, and P, are computed. And (5) the returns of the portfolios are compared and tested to determine whether their differences are statistically significant, that is, whether the link among stock returns and 5, P/BY, P/E, or TO is statistically significant. 4.6.2 Prehminary Computations Before beginning to compute returns, some preliminary computations are required. First, prices are adjusted for corporate actions, such as stock splits, stock dividends, rights issues, new issues of stock, stock cancellations for constituents, and the addition or deletion of constituents. Second, the returns of firms with several classes of shares are calculated. And third, a y^series is computed for each stock. The prices of each security are scaled by the CAR,^^ which corrects the prices for changes in the capitalization caused by a stock split, a stock dividend, or a rights issue. The simple returns are then calculated as the sum of the adjusted price appreciation and of the dividend return.
^^ Price and capital adjusted rates are from EMDB. The price is the price per share of the stock at the end of the "Last Trading Date." These are actual prices; they are not adjusted to changes in capitalization (for example, a two-for-one split).
110
r,
'
=——
'' - 1 ,
(4.1)
where ru is the return of the stock / in t. Pit and Pij.i are the price of stock / at t and t-1, respectively. CARit and CARitj are the capital adjust rate at time / and t-1 of stock /, and Dit are the dividends in t of /. It is common in the LAEM for firms to have several classes of shares. The returns of these firms are calculated as the returns of a value-weighted portfolio of the out98
standing securities.
''u=Lv^
^^."
(4.2)
where n^ is similar to (4.1), Caps,t.i is the market capitalization of class s in t-1 and TotCapit-i is the capitalization sum of all included share classes of firm / in t-1. A monthly P for each stock is estimated with the help of the modified method of Scholes and WilUams with one lag for each stock.^^ 'i. = ^ i + >^'*^..+^,
\
m,t'
m,t—\ I
where r^ and r^-i are like in Equation {\.\),Rm,t and Rm,t-i represent the return of the IFCG.TR.US Index in t and t-1, respectively; cor(Rm,t, Rm,t-i) is the correlation between the market returns in t and t-1, and Et and EM are the error terms in t and t-1, respectively. To compute j3, the returns of each stock are regressed on the IFCG.TR.US Index of firm's country of origin. Due to non-synchronous trading, one lag of the index is included. In Table 2 we saw that the LAEM have high autocorrelation coefficients at lag 1 but low at lag 2. This characteristic can be observed if two lags are used in the regression: in the first lag more than 50% of the y^ coefficients are significant, while for
This is done only with the share classes included in the EMDB. ^ In Zimmerman (1997) there is a complete discussion of the alternative methods for computing J3.
Ill
the second lag only 15% were. Therefore, only one lag should be taken into consideration in order not to diminish the efficiency of the estimated p. One lead has also been tested, but almost all y^"^ are insignificant. Zinmierman (1997) compares alternative methods for estimating ^ for German stocks. He noted that the Dimson P is as good as the modified P of Scholes and Williams (MSWB). I prefer to use the MSWB because, as Fowler and Rorke (1983) have convincingly shown, the derivation of the Dimson /?is incorrect. From 24 up to 48 months of prior historical returns are used to obtain the >^s.
4.7 Results In this section the relation between stock portfolio returns and 5, P/BV, P/E, and TO is analyzed for the seven Latin American markets of concern here. As explained above, the portfolios are first constructed and then their returns are computed. 4.7.1
3-Portfolio Approach
The relation between average stock returns and 5, P/B, P/E, and TO is conducted in this chapter at portfolio level. The stocks are listed according to their 5, P/BV, P/E, and TO at the beginning of either each month or each year. Then stocks are assigned to one of three portfolios. The top and bottom portfolios {PI and P3) each has 30% of the stocks, while the middle one (P2) has 40%. The next step is to calculate equally or value-weighted monthly^^ returns for each of the portfolios for the next month or subsequent twelve months. Finally, we derive a series of monthly returns from May 1986 to November 1999 for all portfolios. ^^^ In a similar way, a time series for portfolio capitalization, value-weighted P/BV, P/E, TO, and >^ is also calculated. 4.7.1.1
Value and Equally Weighted Stocks
Tables 4.5 to 4.8 report value and equally monthly returns, value-weighted TO, valueweighted P/BV, value-weighted P/E, value-weighted P, value-weighted 5, and other characteristics for the portfolios. These variables are calculated for each of the three portfolios for each Latin American stock market.^^^ The grouping procedure is re-
*^ Value-weighted returns are calculated only for portfolios sorted each month. ^"^ The sample periods for Colombia and Venezuela range from January 1987 to November 1999, and the sample period for Peru ranges from January 1993 to November 1999. ^^^ There are twelve portfolios per market, three for each of the four variables.
112
peated every month. Only in the case of P/E sorted portfolios is there a fourth PO portfolio for stocks with negative P/E. Portfolios Sorted by Size In Table 4.5 the value-weighted returns of portfolios with stocks sorted by S are shown. It can be seen that the average returns of small portfolios {PI) are higher than the returns of the large portfolios {P3) in five markets (Argentina, Brazil, Chile, Mexico, and Venezuela), whereas for the other two markets, the middle portfolios (P2) have the highest returns. Only the PI portfolios of Mexico and Venezuela have the highest returns with the smallest p. However, in any stock market the return differences between PI and P3 are significantly different from zero. The returns of equally weighted portfolios confirm the results of value-weighted portfolios: The return difference (P1-P3) is any stock market statistically different from zero. The results of the sample used here (April 1986 to November 1999) do not confirm the result of Rouwenhorst (1999) for the Argentinean and Mexican markets. One might wish to argue that the absence of a size effect in the LAEM is due to the stock selection criteria employed by the IFC-S&P. However, the differences between the three portfolios in terms of the market capitalization are overwhelming. For example, in Argentina the largest portfolio represents 82% of the market capitalization of the three portfolios, even though this portfolio contains less stocks than the middle one. Portfolios Sorted by P/BV Panel A of table 4.6. shows the results of Value weighted portfolios sorted by P/BV. In Brazil, Chile, Mexico, and Venezuela the PI portfolios have the highest returns, while in Argentina, Colombia, and Peru the P2 portfolios have them. Portfolios with the highest returns have the lowest fi only in two markets. In the other markets, portfolios have a similar /?. From the seven analyzed markets, only Brazil, Chile, and Mexico have a return difference (PI-PS) with a r-statistic higher than 1.5. However, the PI portfolio of Brazil has the highest y^, but the J3 differences between the PI and PS portfolios seems to be too small to explain the 2.7% return difference of both portfolios. These results are different from those obtained by Fama and French (1998) and
113
Claessens et al. (1996). One reason for the differences might be the way stocks are weighted in the portfoUos. Therefore, equally weighted portfolios are also calculated. Table 4.5: Summary Statistics for Portfolios Sorted by Size Panel A: Characteristics of Vaiue-Weighted Portfoiios Brazil
/^gentina Return t Turnover P/E P/BV Beta Capitalization N Return t Turnover P/E P/BV Beta Capitalization N Return t Turnover P/E P/BV Beta Capitalization N Return StdAbw t
PI (low)
P2
P3(high)
P1-P3
0.0387 1.8 0.0763 5,3 1,0 1,2 57191,6 8,8 0.0392 2,0 0.1963 12,6 1,3 1,1 375133,2 11,7 0.0358 2,2 0.0344 37,8 1,9 0,9 1986676,1 8,8 0.0029 0,1 0,3
0.0241 1,4 0.0405 5,5 0,6 1,1 136172,2 15,7 0.0360 2,2 0.0309 6,1 1,1 1,0 1269194,9 21,2 0.0206 1,5 0.0480 6,0 1,0 1,2 16590521,7 16,2 0.0035 0,2 0,3
Chile
1Colombia
0.0272 3,6 0.0076 11,2 1,5 1,2 199269,2 10,5 0.0204 3,1 0.0075 10,9 1,8 1,1 918467,7 14,6 0.0247 3,8 0.0098 14,6 2,1 1.1 3610211,1 10,6 0.0029 0,1 0.6
0.0078 0,9 0.0081 8,8 1,0 0,8 32333,0 6.6 0.0189 2,5 0.0068 15,1 1,7 0,9 268808,9 9,3 0.0181 2,2 0.0061 18,1 1.6 0,9 579445,5 6,7 -0.0103 0,1 -1.5
Mexico
Peru
0.0334 3,2 0.0463 9,9 1,2 0,9 296274,8 14,3 0.0244 2,6 0.0447 0,6 1,6 0,9 2060179,5 19,7 0.0304 2,9 0.0426 30,2 1.9 1.1 12277177,5 14,5 0.0030 0,1 0,4
Venezuela
-0.0007 -0,1 0.0745 10,4 1.2 1.2 3027,8 9.0 0.0020 0,2 0.0506 14,7 1.8 1.1 53715,3 12,3 0.0064 0,6 0.0226 33,6 3,5 1,0 696280,7 9,2 -0.0072 0,1 -0,7
0.0196 1,4 0.0223 15,6 1,5 1.0 35354,3 4,5 0.0149 1.2 0.0160 18,3 2,6 1.0 162478,3 5,7 0.0170 1,4 0.0216 17,8 2,7 1,0 478214,1 4.5 0.0026 0,1 0.3
Panel B: Returns of Equally-Weighted Portfolios Argentina Return StdAbw t Return StdAbw t Return StdAbw t Return StdAbw t
P1-P3
PI
P2
P3
0.0199 0,2 1,2 0.0533 0,3 2,0 0.0417 0,3 1,9 0.0334 0,2
L9-
Brazil 0.0067 0,2 0,5 0.0297 0,2 1,7 0.0346 0,2 2,1 0.0230 0,2 1,6
Chile
(0 0)
>-
^j a)\
^ •1 o cvi 00
X
X
X CM
X X X
X
1^
X
X
X
OH 0-
o o X X X X
X X
X
X
^S
X
o> X
1 I
X
X
X
X
PQ
X X •o
o
"5
X X X X X X X
X
X
X
X X
X
i
HI
X
X
5 ^
(0
(0
H 128
® 'a = o •>< E < CQ O O 2 CL
>
x5 < CQ
68il
6 Conclusions Before participating in a stock market, investors compare it with others. In the Hterature three factors are often considered important for the comparison: stock returns, their associated risk, and cost of trading. Accordingly, in the forgoing we have studied three questions with regard to the seven major Latin American stock markets: (1) How different are the financial, economic, and political conditions of the LAEM from the developed markets and what are the investment laws in the LAEM? (2) What is the cost of trading in the LAEM; is it higher than in the developed stock markets? And (3) does the return determination process in the LAEM differ from those processes documented for the developed markets? The first and second questions were studied in the first part of this thesis (Chapters 2 and 3), while the third question was addressed in the second part (Chapters 4 and 5). The seven emerging markets of Latin America studied here were Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela. They were selected in accordance with the IFC-S&P's definition of emerging markets. Although the seven major Latin American emerging markets fulfill the IFC-S&P requirements, they are nevertheless heterogeneous. Areas such as operational efficiency, quality of market regulation, supervision and enforcement, corporate governance practices, minority shareholder rights, transparency, level of accounting standards, and information levels vary substantially across these stock markets (see 2.5 above). The emerging markets are gaining in importance. Capital flows to the emerging stock markets, especially to Asia and Latin America, increased substantially in recent years (see 2.3). The interest of researchers in emerging markets has also increased; however, investigations analyzing stock returns, their associated risk and cost of trading, are still limited. In Chapter 2 the analyzed Latin American stock markets are introduced. The presentation centers on two questions: (1) What are the financial, economic, and political conditions in those countries? And (2) what are the investment laws there? Financial, economic, and political conditions are important aspects of a stock market since they influence the associated risk of stock returns. The first question is discussed through the political, economic, financial, and compounded risk indexes of the ICRG. To answer the second question, the Investment Law Index is developed and calculated. 158
In contrast to developed countries, political risk is more important in emerging markets. In developed countries, economic news is likely to have a more significant impact on the stock market than political news since the impact of the latter is less clear (Chan, Chui, and Kwok (1999)). However, the evidence shows the reverse in emerging markets. The ICRG considers political risk to be twice as important as economic and financial risk because it is related to the intention to pay, while economic and financial risk is associated with the ability to pay. The political events in Latin America over the past two decades have confirmed the ICRG's position. The political index shows great heterogeneity among the Latin American countries (see 2.4). Compared to developed countries (France, Germany, the UK, and the U.S.), the political risk of the LAEM is higher. However, most of the LAEM showed greater improvements than the developed markets did. Although the magnitudes differ, the LAEM showed an increase in the risk rating from 1986 to 1999.^^^ The countries with the highest improvement were Chile and Peru, which had increases of 71% and 59%, respectively. The 25% and 10% decreases of Venezuela and Colombia were the lowest of the region. The political risk indexes of Brazil and Mexico were the most stable of the LAEM and belonged among the highest over the whole analyzed period. The political risk index of Argentina increased significantly in the first part of the '90s but decreased strongly in 2001. The political correlation between the LAEM and developed countries is high. However, when the U.S. political risk index declines, the political indexes of the LAEM and other developed countries also diminish. The opposite does not occur. The economic risk ratings showed similar results regarding political risk. The relationship between stock returns and economic variables has been shown to be strong (see, e.g., Fama (1990), Balvers et al. (1990), and Lovatt (1996)). Therefore, the economic risk ratings of the ICGR were discussed in section 2.4. The economic risk ratings summarize the macroeconomic factors (current account to GDP, real GDP growth, inflation, budget balance to GDP, and GDP per capita). The trajectories of economic and political risk ratings show some similarities. (1) Chile, Mexico, and Peru have the lowest economic risk; (2) economic risk decreased during the analyzed period in most of the LAEM, especially in Chile and Brazil. The decreases were more
^^^ A higher risk rating means less risk.
159
evident from 1990 to 1994. Thereafter, the risk ratings diminished (risk increased) as a consequence of the crises in Mexico and Brazil. (3) Compared to developed markets, the LAEM have significantly lower economic risk ratings (i.e., are riskier); their volatility is much higher and they show faster improvements. The financial ratings, which aim to provide a measure of the countries' ability to pay, shows that: (1) the financial risk diminished over the last 15 years, (2) the volatility of the indexes vary significantly among the LAEM, (3) compared to the developed countries, the LAEM have a higher financial risk over the whole period. The financial risk of the LAEM diminished at higher rates than in France, Germany, the UK, and the U.S. Using the Composite Risk Index, we aimed to answer two questions: (1) How risky are the LAEM compared to the most developed stock markets? And (2) how different is risk across the LAEM? The CRI is the weighted sum of political, financial, and economic risk indexes (poUtical risk 50%, financial and economic risk each 25%). The CRI shows that (1) the LAEM are riskier than the four developed countries, but (2) in the LAEM risk diminished faster. (3) The LAEM are heterogeneous and can be divided into two groups, the first group consisting of the least risky countries (Chile, Mexico, and Colombia) and the second consisting of the riskiest countries (the riskiest being Peru, which is followed by Argentina, Brazil, and Venezuela). And (4) the CRI correlation between the LAEM is low but higher than with the U.S. or Germany. An Investment Law Index was developed and calculated in the second part of Chapter 2 (section 2.5). The aim of the IL-Index is to determine the differences as regards (1) shareholder rights, (2) law enforcement, (3) insider trading, and (4) investment barriers to foreigners across the LAEM, and how they compare to some developed markets. Contrary to La Porta et. al. (1998),^^^ the information on the existing shareholder rights, law enforcement, insider trading, and barriers to foreign investors was aggregated into the IL-Index. An index makes it easier for investors to interpret the information and compare it across countries.
They discussed in a separate way shareholder rights, law enforcement, and insider trading.
160
The IL-Index is calculated by averaging the values of the indexes on investor protection,^^^ law enforcement, insider trading, and barriers to foreign investors.^^'^The BLIndex shows that the laws stimulating investments vary in the Latin American countries analyzed. They fall into two groups: Chile, Argentina, Brazil, and Peru have the highest IL-Index, while Colombia, Mexico, and Venezuela have the lowest. Compared to the developed markets, the IL-Index of the first group is as high as that of Italy, France, and Germany, but lower than that of the U.S. and the UK. The presentation of the LAEM in Chapter 2 showed that: (1) the LAEM are heterogeneous but all are riskier than the developed countries and (2) investment incentives and restrictions in the former substantially differ from developed markets. For these reasons it is interesting to study whether the factors determining the stock returns in the LAEM are similar to those determining returns in developed markets. This is studied in Chapters 4 and 5. But before turning to this topic, the implicit trading costs in the LAEM is investigated in Chapter 3. Studying the trading costs is important since it can influence (1) the composition of globally efficient portfolios and (2) the international order flow (Domowitz et al. 2001). Trading a stock in one way costs 58 bp in Brazil and 61 bp in Mexico, while in Colombia and Venezuela it costs 97.5 bp and 134 bp, respectively. Compared to developed marks, trading costs in the LAEM are high. For example, one-way trade in Paris costs 29.5 bp, in Frankfurt 37.7 bp, in the U.S. 38.1 bp, and in London 54.5 bp. The one-way average trading costs in the Latin American markets is 86.9 bp, while in the four developed markets it is 40 bp. If these portfolios turn over twice per year, the annual average trading costs will be 347.6 bp in the LAEM and 159.8 bp in the developed markets. These represent 16% of the average annual portfolio returns for the LAEM and 12% for the developed markets. In Chapter 3 we studied and compared the determinants of implicit trading costs among the LAEM andfour developed markets. Specifically, the development state and the trading architecture were analyzed since, according to the literature, they deter-
La Porta et al. (1998) developed an antidirector rights index. The difference between this index and the SH-Index is that the SH-Index additionally includes mandatory dividends and the percentage of capital to call an extraordinary shareholder meeting. By including these additional rights, the SHIndexes of French-civil-law countries move closer to conmion-law countries. However, the differences are not important enough to have a significant impact on the IL-Index.
161
mine the implicit costs. Three questions were discussed in this chapter: (1) How heterogeneous are implicit trading costs among the LAEM? (2) How different are the implicit trading costs in the LAEM from those in the developed markets? And (3) which factors are responsible for the differences? To answer these questions, we designed two main indexes: Development State Index and Trading Architecture Index. With these indexes is possible to detect the factors responsible for the high trading costs in the LAEM. Since stock market development is a multifaceted concept, the DS-Index was calculated by averaging four market indicators: stock market size, liquidity, stock market concentration, and number of quoted firms. The DS-Index shows that the LAEM are heterogeneous (see Table 3.3). In Brazil, Chile, and Mexico the DS-Index tended to be higher than in Argentina, Colombia, Peru, and Venezuela. Compared to developed markets, the results show that the developed markets are more developed than the LAEM. The DS-Index's components confirm the foregoing results. (1) Stock market size is significantly lower in most Latin American markets. Of the LAEM, Chile is the only stock market that has a MC/GDP comparable to Germany's. (2) VT/GDP and TO vary across the Latin American markets. Compared to developed markets, the VT/GDP ratio is also much lower in the LAEM. (3) Concentration is higher in the LAEM. (4) The number of quoted firms also shows the heterogeneity of the LAEM and a great difference from the developed markets. For example, among the Latin American markets, Brazil has the highest number of quoted firms (486 in 1999), which is a half of those in Germany and only 5% of the firms trading in the U.S. The TA-Index was developed in section 3.4. By "trading architecture" we understand the participating elements that make possible the trading process. The trading process can be divided into three phases: the buying and selling of securities by means of intermediaries, the transmission of orders from intermediaries to the central order book, and the clearing and settlement of securities and funds. In order to include the three phases in the TA-Index, the Intermediary Index, the Trading System Index, and the Custody, Clearing and Settlement Index were first calculated. Then the TA-Index is obtained by means of these indexes. For each LAEM and the studied developed mar-
This way of weighting the components produce reasonable results. The method of weighting employed here is not discussed further in this thesis.
162
kets, the TA-, I-, TS-, and CCS-indexes are shown in Table 3.18. This table is also shown below. The participation of intermediaries is relevant for a trading system because they facilitate the transactions of stocks between buyers and sellers. In each LAEM, all of which are order-driven, there are two types of intermediaries: brokerage firms or brokers and market-makers. The introduction of market-makers in the LAEM is recent and is justified by the stock trading concentration in these markets. Since the functions (privileges and obligations) of brokerage firms do not vary across the analyzed markets, they are not included in the I-Index. For this reason, only the functions of marketmakers are taken into account. Demarchi and Foucault (1998) used four obligations and three privileges to describe market-makers. The I-Index is computed using these factors (see 3.4.1). The I-Index shows how the obligations and privileges of the market-makers varies across the Latin American exchanges. Chile and Peru have the highest index, followed by Brazil and Mexico. Colombia and Venezuela have the lowest I-Index. The duties of market-makers in Chile and Peru are similar to the duties of the animateurs of the Paris Bourse and, to a lesser extent, to those of the Betreuers of Frankfurt. The next computed index was the Trading System Index. The TS-Index is important because it influences the price discovery process, that is, transaction costs and risk. The TS-Index is computed by averaging the points of market segmentation (types of stock orders and segments of the stock market) and electronic trading system (functions of the electronic trading systems, the use of call auctions, and information transparency). In section 3.4.2 we described how the TS-Index is computed. Like the other indexes, the TS-Index also shows the great heterogeneity in the LAEM. The trading systems of Sao Paulo, Mexico, and Caracas have the highest TS-Index, while Colombia and Lima have the lowest. Buenos Aires and Santiago are in the middle. Compared to the trading systems of developed markets, the seven Latin American systems have index ratings below the three developed markets analyzed. The reason for this lies in the low trading segmentation of the LAEM, which may be caused by the small number of stocks quoted. The limited opportunities for trading and the absence of some functions of the ETS may increase trading costs, which in turn affects liquidity and the cost of capital in Latin America.
163
The third and final component of the TA-Index is the Custody, Clearing, and Settlement Index. The clearing and settlement phases, together with the custody phase, seems to be merely administrative; however, problems can arise if one of the two parties does not fulfill the contract. The CCS-Index is constructed using information on its three components: custody (ownership structure, additional services, the form in which records of ownership are held), clearing (trade capture, matching, confirmation, comparison and affirmation mechanisms, and the calculation of settlement obligations), and settlement (delivery versus payment, settlement assurance, and settlement time). In 3.4.3 we described each of these elements in detail. The TA-Index shows that these trading phases are as risky in Buenos Aires, Sao Paulo, Mexico, and Caracas as in Paris, Frankfurt, or New York. The process is riskier, however, in the Santiago, Colombian, and Lima exchanges. Finally, the TA-Index is computed by averaging the I-, TS-, and CCS-Indexes. The TA-Index shows that there exists a great heterogeneity in the LAEM. According to this index, Sao Paulo and Mexico have the most developed trading architecture among the Latin American stock markets. Buenos Aires, Lima, Santiago, and Caracas follow, while Colombia has the lowest TA-Index. Compared to the developed markets, the TA-Index of all Latin American exchanges is significantly lower. Finally, the TA-Index makes it clear that the factors responsible for the weaknesses of the Latin American trading architectures differ from market to market. Buenos Aires, Colombia, and Caracas - and, to a lesser extent, Sao Paulo and Mexico - have to reinforce the role of their market makers. The trading systems is particularly weak in Colombia and Lima. Custody, clearing, and settlement is weak in Santiago, Lima, and Colombia. According to the results of the designed indexes, the answers to the three guiding questions read as follows: (1) The development states and trading architectures of the LAEM differ among themselves. Of the LAEM, Brazil and Mexico have the highest DS- and TA-Indexes. Chile has a high DS-Index, but its TA-Index is low. Both indexes are lower in Argentina, Peru, Venezuela, and Colombia (in this order). Compared to developed markets, both indexes show that (2) the implicit trading costs are higher in all LAEM and that (3) the factors responsible for the higher costs differ among them. In Table 6.1 the elements that each LAEM should improve in order to diminish the implicit trading costs are marked with an X.
164
Table 6.1: Summary Results of Implicit Trading Costs Argentina Brazil Chile Colombia Mexico Peru Venezuela
Implicit Trading Cost DS-lndex MC/GDP Liquidity Concentration NQF
29,6
21,4
38,6
X X X X
42,2
X X X X
27,3
35,2
34,7
X X
X X X X
X X
TA-lndex l-lndex
X X X X Privileges -Obligations X X -TS-lndex Segmentation X X X ETS X X X X CCS-lndex Depository X X Clearing X X X X Settlement X X X -X means that the stock market belongs to the last 4 of the LAEM or are under the average - - means no information
X X X
--
In concluding Chapter 3, the robustness of the indexes was shown. Table 3.19 shows that the exchanges with the lowest implicit trading costs have the highest D- and TAIndexes. This reflects the importance of the selected components for each index and at the same time gives a clue to how they are related to trading costs. Chapter 4 had two main goals. The first was to investigate whether the factors explaining the return variation in the LAEM are similar to those documented in developed markets - specifically, whether the firm size and the ratios price-to-book value, price-to-eamings, and the turnover, influence the expected returns. The second goal was to reconcile the contradictory results of previous investigations that tested for anomalies at a stock level in the Latin American markets. There are at least four reasons why those studies came up with contradictory results: their different sample periods, different number of portfolios, different frequencies at which they sorted stocks, and different methods for the computation of portfolio returns (equally or valueweighted). To achieve our goals, we conducted a simple portfolio approach. Thus we sorted stocks by one of the variables into one of three or six portfolios, allocated the stocks
165
each month or year, and then calculated the returns either equally or value-weighted. We also compared the results of different sample periods. Chapter 4 was divided into five sections. The empirical evidence from developed and emerging markets was presented in the first section. Of the existing literature on market anomalies in emerging markets, the publications of Rouwenhorst (1999) and Claessens et al. (1995) are of particular importance. It is their contradictory results that have provided the impetus for this chapter. Rouwenhorst (1999) found that the return factors in emerging markets are similar to those documented for many developed markets. By contrast, Claessens et al. (1995) demonstrated empirically that the Latin American markets may have a size effect, but it is not necessarily restricted to the smallest sized stocks. To study the reasons for the contradictory results, the EMDB of the IFC-S&P is used and discussed in section 4.3. This database is employed because it publishes current and historical statistics of series needed and has proved reliable. Section 4.4 presents summary statistics of the LAEM, while 4.5 discusses some properties of the monthly returns. It is shown that (1) the average returns and standard deviation of the LAEM are often higher than in the developed markets and many other emerging markets, (2) the mean and median are different, (3) the stock returns are not normally distributed (except in the Chilean market), and (4) the cross-country return correlation was low in the 1980s but increased in the '90s. Before testing for the market anomalies in the LAEM, the univariate portfolio methodology is described in 4.6. The portfolio approach involves four steps: (1) At the beginning of each period (month or year), stocks are ranked according to one of the variables (5, P/BV, P/E, or TO). (2) Stocks are grouped into one of the portfolios (three or six). (3) For each portfolio, equally or value-weighted returns, its volatility, and its average TO, P/E, P/BV, 5, and y^are computed. Finally, (4) we test whether the return differences between the portfolios are statistically significant. The results do not replicate the evidence of developed markets (see 4.7). For the sample period, from May 1986 to November 1999, value and equally monthly returns of the 3-portfolio approach shows the following results: (1) For size-sorted portfolios, the return differences between PI and P3 are significantly different from zero in any stock market. One possible reason for the absence of a size effect could be the stock selection criteria employed by the IFC-S&P. However, the average of the P5-port166
folios of the LAEM represents 80% of the market capitalization of the three portfolios, even though P3 portfolios contain fewer stocks than P2 portfolios. Therefore, the selection criteria used by the IFC-S&P are not the reason for the absence of a size effect. (2) For P/BV-sorted portfolios the equal return difference is statistically different from zero only in Brazil, Chile, and Mexico. The average return of value-weighted portfolios confirms the results. (3) The return difference of portfolios sorted by P/E is significant in Brazil, Chile, Mexico, and Venezuela for value-weighted and equally weighted returns. (4) A statistically significant negative relation between stock returns and TO does not exist in any Latin American stock market. Contrary to our expectations, the return difference is significant in Chile and Venezuela, but with opposite sign. (5) Equally and value-weighted portfolios show slightly different results. Equally weighted portfolios usually have higher t-statistics of the return difference ^(Pi-P3) ^han value-weighted portfolios. The reason for this is that within a portfolio there also exist small and large stocks. If small stocks have higher returns (as is expected) and are assigned the same weight, the return of the portfolios that weight their stocks equally will have higher returns. This effect may be more extreme in small portfolios since the size of its stocks varies more. Finally, (6) the return difference of the portfolios sorted by P/BV or P/E is statistically significant in Brazil, Chile, and Mexico, and these countries also happen to have the best-rated accounting systems (see Chapter 2). These results do not coincide with Rouwenhorst (1999). One reason for the differences could be the different sample periods used in each study. Therefore, we tested for the analyzed anomalies in three different periods: (1) From the beginning^^^ to November 1999 (whole period), (2) from the beginning to April 1997 (the period used by Rouwenhorst (1999)), and (3) from May 1997 to November 1999. For each month, stocks were sorted by 5, P/BV, P/E or TO, and grouped into three portfolios. The results obtained show that (1) for the period used by Rouwenhorst (1999) coincide with his. (2) Returns of equally and value-weighted portfolios show slight differences. And (3) the statistical significance of the anomalies varies with the sample period in most of the stock markets (see Table 4.9).
'^^ Date at which the IFC-S&P begins to collect information for the corresponding stock market.
167
Another possible reason for the different results obtained by researchers could be the frequency at which they rebalance their portfolios. As is common, stocks are sorted on a monthly and annual basis. There are some facts that speak in favor or against one of them. For example, due to the fast-changing environment in emerging markets, portfolios should be revalanced frequently. On the other hand, as we showed in Chapter 3, trading costs are higher in emerging markets and thus sorting stocks once per year might be sufficient. For the LAEM the results show that the frequency at which stocks are sorted influences portfolio returns (see Table 4.10). The different number of portfolios into which stocks are sorted could be yet another reason for the contradictory results obtained by researchers. To test whether the number of portfolios influences the significance of the anomalies, stocks were also grouped into six portfolios. However, we did this only for Brazilian and Mexican stocks since the EMDB does not include enough stocks from the other markets. The return differences of the 6-portfolio approach are slightly higher than the return differences of the 3-portfolio approach (see Table 4.11). However, despite the small changes, the significant anomalies of the 3- and 6-portfolio approaches remain the same. In Table 4.12 we sunmiarize the results of Chapter 4. In sum, the variable that influences stock returns depends on the market, the period, the frequency at which stocks are sorted, and the way stocks are weighted. The only variable that might not influence the results is the number of portfolios into which stocks are sorted. The results obtained with the univariate portfolio approach left some questions unanswered: Would P be able to absorb the return variation if the fundamental variables were to become insignificant? Or if one of the fundamental variables were to become significant while >ff did not? Or do both of them explain the return variation? Or might a third factor even be necessary for such an explanation? These questions were examined in Chapter 5 by means of time series and cross-sectional methodologies. In Chapter 5 the joint roles of market returns \R^ -R^) ox market p and S, P/BV, P/E, and TO are studied. To test the robustness of the results, the two most important econometric methodologies of the cross-sectional analysis were used: (1) The time series regression approach developed by Black et al. (1972) and (2) the cross-sectional
168
approach developed by Fama and MacBeth (1973). For both procedures we tested whether the one-factor model captures the variation of stock returns and whether combinations of (/?„ - /?^) or market P, and firm S, P/BV, P/E, or TO better drive the cross-sectional differences in stock returns in the LAEM. To my knowledge, no previous study has investigated the commonality of the various effects for the Latin American stock markets. The analysis in Chapter 5 was carried out at portfolio and stock level rather than at country level. *^^ The analysis was carried out by means of the time series approach because this methodology has several advantages. ^^^ On the other hand, the analysis at a portfolio level has been shown to generate biases in statistical inferences (Lo and MacKinlay (1990)). For this reason, we used the cross-sectional approach proposed by Fama and MacBeth (1973) to test whether the market returns completely explain the realized stock returns at a stock level. The Fama-MacBeth methodology also has several advantages.^^^ However, according to Pettengil et al. (1995), this approach biases the relation between returns and P since it is based on expected and not on realized returns. Because realized returns are used for the regressions, a segmented relation exists between realized returns and p. As a consequence, it is necessary to test for a conditional relation between P and realized returns. Since our results were in line with Pettengil et al. (1995), we also tested the conditional CAPM. The results of the time series approach were presented first. For the regressions, we used a combination of the excess market return and portfolio return differences (P1-P3) of 5-, P/BV-, P/E-, and TO-sorted portfolios as explanatory variables. The variables that were explained are the excess returns (portfolio return - risk-free interest rate) of the S, P/BV, P/E, and TO portfolios of each country. Different variants of Equation (5.8) were regressed.
'^^ There are several reasons for taking this approach. Emerging market returns are influenced by local information (Harvey (1995)) since a significant number of barriers effectively segment the emerging markets from the global capital markets (Bekaert (1995)), and the purchasing power parity varies among the LAEM. '^^ (1) The slopes and the R^ present direct evidence for how well the factors capture the common variation of the stock returns; (2) it is easy to test whether the intercepts are zero; (3) the high volatility of stock returns in a market will not lower the power of the asset pricing test; and (4) in the EMDB there are not enough stocks to construct a two- or three-way grouping procedure. ^^Vl) With it additional risk measures beyond y^can be aggregated and (2) y^and the coefficients of the explanatory variables are updated periodically.
169
The findings from the time series methodology reveal a strong relationship between the realized stock returns and market excess returns or ^ and one of the additional factors, depending on the stock market. Table 5.1 shows the results of the one-factor model, which can be reduced to three: (1) the CAPM cannot always explain the average returns of the portfolios since many intercepts are significantly different from zero; (2) P is always highly significant, but in four stock markets it is lower in small portfolios {FI) than in large portfolios {P3)\ and (3) in each stock market PI portfolios sorted by turnover have lower p^ than P3 portfolios. Since 20% of the intercepts of the one-factor model are different from zero, it made sense to test for combinations of market >^ and the portfolio return differences (P1-P3) of S, P/BV, P/E, or TO. We carried out this test after first making certain that the premium between the factors represented by the return differences of the portfolios (P1-P3) were uncorrected. The results of the two-factor models can be summarized as follows: (1) The explained return variations {^) increases slightly; (2) fewer intercepts are statistically different from zero. Intercepts are more frequently significant and negative in P3 portfolios; this can be interpreted as an adjustment for the unexpected higher ps of P3 portfolios, (3) Pis still highly significant; (4) one- and two-factor models produce similar/?, thus the small improvements must come from the second factors. (5) The P/BV(P].p3) is the factor that was most often significantly different from zero and that had the highest average increase of R^. (6) The variables that have the greatest impact on Rp^fwary across the LAEM: in Argentina they are S(PI.P3), P/BV(PI.P3), and P/E(pi.p3)\ in Brazil T0(pi.p3) and P/BV(Pi.p3); in Chile S(PI.P3), P/BV(PI.P3), and T0(Pi.p3)\ in Colombia P/BV(pj.p3) and P/E(Pi.p3)\ in Mexico P/BV(Pi.p3) and S(Pi.p3)\ in Peru P/BV(PJ.P3), S(PI. P3), and P/E(Pi.p3)\ and in Venezuela P/BV(PI.P3) and T0(Pi.p3). (7) Across portfolios, the PI portfolios almost always have the lowest R^ and the P5's have the highest. The two-factor model showed that market returns and one of the mimicking factors were significant, depending on the stock market. The next question was whether more than one of the four fundamental variables have a significant impact at the same time on the expected returns. The results of the three-factor models showed that (1) none of the fundamental variables has a clear influence on the average returns. The results vary depending on the stock market; this variation may be caused by the heterogeneity
170
of the LAEM (see Chapters 2 and 3).^^^ However, (2) P/BW is the most or the second most important factor of the four variables in each country, and (3) TO is important for three markets, while the size and P/E is are important only in one market each. To check for the robustness of the time series methodology, the cross-sectional approach developed by Fama and MacBeth (1973) was also used. One-, two-, and fivefactor models were estimated. In keeping with Pettengil et al. (1995), we also calculated the conditional relation between P and realized returns. Regressions were calculated at a stock level only for Brazil and Mexico since there were not enough stocks from the other countries included in the EMDB. The Fama-MacBeth cross-sectional approach was carried out at stock level due to several disadvantages that the analysis has at portfolio level. ^^^ The results of the one-factor model using the Fama-MacBeth cross-sectional methodology are striking. In any stock market, ^ is significant. The reason for this result is that the market risk premia are often negative even if the expected risk premium is positive (Elsas et al. (1999)). We therefore estimated the conditional relation between P and realized returns (Pettengil et al. (1995)). The results of the conditional model were consistent with those obtained in the previous sections. In all LAEM the relation between /? and realized returns was significant. Adding factors such as S, PfBV, P/E, or TO to the conditional one-factor model did not cause any problems. Since the additional factors always have positive values, it was not necessary to test for a conditional relation between the additional factors and the realized returns. Using the conditional cross-sectional approach, we calculated some combinations of Equation (5.15). As with the time series approach, /? always remained significant and the signs of 5, P/BV, P/E, and TO were not always the expected. Furthermore, the results of the conditional cross-sectional model may not coincide with the results of the time series since the portfolio returns and the four variables are statistically related. However, since S and P/E are economically insignificant in all countries, both approaches should coincide.
^^' The results for each country were explained in detail in section 5.3.1.3. '^° (1) All within group coefficients (intercepts) are lower (higher) than the coefficients obtained from the full sample; (2) in most cases the coefficients (intercepts) are so small (high) that they do not provide support for the model; and (3) the I^ within regressions are also lower (Berk (1997)).
171
Overall, the results show that (1) the Latin American stock markets are heterogeneous and riskier than developed markets. (2) The state of development in trading architecture of the LAEM varies among these markets; and compared to develop markets, their implicit trading costs are higher. (3) The return determination process in the LAEM differs from those documented for developed markets and does not confirm the evidence found by Fama and French (1993) for the U.S. stock markets. For the LAEM the variable that influences stock returns depends on the market, the period, the frequency at which stocks are sorted, and the way stocks are weighted. Furthermore, the results of the time series and cross-sectional approaches coincide: the realized returns are explained by >S P/BV, and, to a lesser extent, TO. In Table 6.2 the results are summarized for each country and recommendations are also given there for improving the efficiency of their stock markets. For each country, the main results of this thesis are summarized in table 6.2.
172
N 2 CO -P
p
p
I I* 111 ill 111 l l
11
Q.
E
iJ I 1iff
i a
1 r li
II If
CO C2
>
11
8
^
E f
(0
E
i C
Is ?,?
w8
Q) Q.
|i
Si
> to
3
E -D © z OH
t j (0 (0 _0
H O
CO '3
M-
o u '>
E W
0
E
1
LJ_
2
LU
^ 2 =
,2
LU
(0 0
a.
o
§1
0 o
?i2
>
.2 g
S c 0
I-
a i" o cvi"
z
g
o F: ^ ;
CM CO T- Tf O CO O CVJ
CM CO 1 - - * O -^ O CM
o o o ^
o o o
o" o" o" 0"
T- 0> T- T-
8^.81
o" o" 9 CM
O" I I T- CO
^ - CO ^ - h-
O o
8 018 5
9 9 o 9*
o" 9 o" o"
^
«)
o o" 9 9
B •
9" 9" o o"
0
o
0
0
' - CM 9 . 01 o CM" 9 9
0
) O CO
o" 9 9 9 O
o
O
CM TCO O O" O" O" CM]
§
I
o £ e ^ .2 + II
2 I^O t t:U
o
X 111 M (S
I
!
i
o
CJ) O
o" o" o" o"
o o" o" o I
o"
5
S o
§S8§
'
o o o o1
i
z
aa
00 CO O CM O O" CM" O " CD
o" 1-" 9" 9"
o '^- oi
O
-
o §
- iJ? o- g
8S8^.
O O
CO Q ' CO O
m ^
is.
'1
5eS5§
CO CO ; r »'] CM 00 9 . "T.
o" 1-" 9 CO O) CO CO CO 00 CM 1 -
00 2S
•>- ; ^ o " ; i
9" v" o" 9*
rs. en Tt
^
9 s
5
I
o " T-"
SS! 00 O
C»
2
CO T, "^ O O)
^
i
^
5
^
i
•o
c
(0
S
9 T 9 9
S S O) CD O . " 1 o CM
CM T- CO 101 ^ CO CM CO
9 9 o -"
o'
col
CM" O*
i n 00 CM 0 C3 in"
s d d
o T- o 00 T- 00 CM CM d " CM" O CM"
O)
CM 0
o Sd^
o" ^ d ; :
o
8 ^. 8 ^. Q- , -•' o - r" v ^
8 -^ o c» 0" 9" 9" 71
o" d^ 9 T
I s
z
z
CM CO O CO O CM '-^ « J d d 9 CO ^
TCO
r o" o CM"
$
^
CO h-
^ 00
o"
o
T- 00 Q CM ID O
Q. o
o o o o
0) 05
Q-
^-
0>
10 O
^
o
ID
. o
§
CM ^ CO CO O CD o " o o " T-"
h4
o "1 o o>
o 00 o 00 o" o" o" o"
9. 9. o 9' "7 o" o |
O*
CO]
O
ID O CO
O
O
CM"
-
00 " 1 o ' o ^ - • ;
o" CM" 9 T
o c» o o
5 I
T - 0 0 i n CM CO
- CO
8^88
C3)
s S :* §
O CO O) CO ^ CD r - 10 •r-' IS." T-" OT
(B
'^-
o" o" o" 9
o" co" o"
•a c
o " -*" 9 ' CO
8«g^.
o" CM" O" O*
S
^ t2 § §
o" "* o" o"
o" 9 o" CM"
g •UJ (A
Tf ( 0 1
o" 1
S fi! a S
o" o" d" o" >
^ cj) CM g CM ID O , ID_
o" "f" 9 9
"-I
o +
as-
T-"
^sgs
§fc=
. 2 OQ
CM" O "
O •^
CO 8 CM
f-
o
(J>\
0 ^ 5 0
o" o" o* o"
9
I o ^
CM o" 9
II ^
I
O
C0|
SI
I 193
^ $ s X ) "S * , (50
05 ^,. 05 '^.'
o 5 ^ c ^2
°
e b
3 °2 °
o" 1^ T-" 1^
J3 c
CM
S - 05 T-_ • * o_ co_ d co' o" o"
O u. Q. OC
c ^ x: H
i S £ 2i 6 § « g^ ^
M
! J (u ^
^^
«
tu rrY' ^
o"°?-Sl
ftb
o •.- o O O O T-
2i -§
cu
«
ddd d
j« 05 O C * -
^
^^ 1 s ^ ^1 8> I g
'kl
^
05
-a ' §
1 - CM lO 05 CO T- T- •^^
o" CM"
l O N O N CMi-CMO d co" d co"
O CM § $ 9" 9" d 1
^OOJOO CMC0'^U5 d CM" d N"
O" T-"
gco^ooj T-" •^" T-" * 1
CO O CM O d^ co" o" CM"
'T. " I '". '"-I 9 V ° '^
^•«tCJ5C0 coi-coin d co" d co"
»-lOi-(D CMOifl'.d CM d co"
OC0CJ52 00 •^ CJ5 " . d 00" d S2
T-C0O5CMI O 05 00 05| T-" 00" d Tf"
o £ o d d d
9v9"9l
CO o r^
o. o « «
IP?
_e ^ .2
1 - CO T -
d'T-"dT-|
. 5 a,
%
5
gco
3:
^
t:
s? XI
c H
„
o
«
^
2 ^' H
«
c
-
-
1.1
ft,fc1^ zZ .2 ^
IH