Artículos

 

Exchange rates, market efficiency and purchasing power parity: Long-run tests for the Latin American currencies*

 

Edgar Ortiz** Alejandra Cabello*** Raúl de Jesús**** Robert Johnson*****

 

** Profesor de Finanzas, UNAM. Miembro del SNI nivel II. Editor en jefe de Frontiers in Finance and Economics. Correo electrónico: edgaro@servidor.unam.mx

*** Profesora de Finanzas, Programa de Posgrado en Ciencias de la Administración de la UNAM. Miembro del SNI. Correo electrónico: alecr@servidor.unam.mx

**** Profesor de Finanzas, Facultad de Ciencias, UNAM. Correo electrónico: edgaro@servidor.unam.mx

***** Profesor de Economía, Universidad de San Diego. Correo electrónico: johnson@sandiego.edu

 

Fecha de recepción: 2 de mayo de 2005.
Fecha de aceptación: 2 de junio de 2005.

 

Abstract

In efficient markets current prices reflect all available information. Past prices do not contain any useful information for predicting future prices or for realizing extraordinary gains. This principle, known as the weak hypothesis of informational market efficiency, has been incorporated into Purchasing Power Parity (PPP) theory to overcome its limitations in the intertemporal analysis of exchange rate adjustments to inflationary trends. Overall, recent studies dealing with exchange rates from developed countries validate their efficiency. Research for the case of developing economies is rather limited. The present study analyzes empirically the efficiency of the exchange rates markets from 15 Latin American Countries for the period 1970-2000. Based on the enhanced PPP model, two regression analyses and a unit root test are applied.

Key words: exchange rates, efficient markets, purchasing power parity, Latin America.

 

Resumen

En mercados eficientes, los precios corrientes reflejan toda la información disponible. Los precios pasados no contienen ninguna información útil para predecir precios futuros y realizar ganancias extraordinarias. Este principio, hipótesis débil de la eficiencia informativa de los mercados, ha sido incorporado a la teoría de la paridad del poder adquisitivo (PPA), a fin de sobreponer sus limitaciones en el análisis intertemporal del ajuste de los tipos de cambio a las tendencias inflacionarias. En general, estudios de mercados de divisas de los países desarrollados validan su eficiencia; sin embargo, la investigación para el caso de los países en vías de desarrollo es limitada. En este trabajo se analiza la eficiencia de los mercados de divisas de 15 países latinoamericanos para el periodo 1970-2000. Basándose en el modelo ampliado de PPA, se aplican dos modelos de análisis de regresión y uno de raíz unitaria.

 

Résumé

Dans des marchés efficients, les prix courants reflètent toute l'information disponible. Les prix du passé ne contiennent aucune information utile pour prédire les futurs prix ni pour obtenir des bénéfices extraordinaires. Ce principe, hypothèse faible de l'efficience informationnelle des marchés, a été incorporé à la théorie de la parité du pouvoir d'achat (PPA), afin de surmonter ses limitations dans l'analyse inter-temporelle de l'ajustement des taux de change aux tendances inflationnistes. En général, les études sur les marchés de devises des pays développés valident son efficience. L'investigation en ce qui concerne les nations en voie de développement est limitée. Dans ce travail, on analyse l'efficience des marchés de devises dans 15pays latino-américains pour la période 1970-2000. Se basant sur le modèle amplifié de PPA, on applique deux modèles d'analyse de régression et un modèle de racine unitaire.

Mots clés: taux de change, marchés efficients, parité du pouvoir d'achat, Amérique Latine.

 

Resumo

Nos mercados eficientes, os preços correntes refletem tuda a informação disponível. Os preços do passado não comtêm nenhuma informação útil para predizer preços futuros e obter lucros extraordinários. Este princípio, hipótese fraca da eficiência informacional dos mercados, foi incorporado à teoria da paridade do poder aquisitivo (PPA) afim de sobrepôr suas limitações no análise inter-temporal do ajuste dos tipos de câmbio às tendências inflacionárias. Em geral, estudos sobre os mercados de divisas dos paises desenvolvidos validam a sua eficiência. A pesquisa para o caso dos paises em desenvolvimento é limitada Neste trabalho analiza-se a eficiencia dos mercados de divisas em 15 paises da América Latina para o período 1970-2000. Baseando-se no modelo ampliado de PPA, aplicam-se dois modelos de análise de regressão e um de raiz unitária.

Palavras chave: tipos de câmbio, mercados eficientes, paridade do poder aquisitivo, América Latina.

 

Introduction

Empirical evidence demonstrates the failure of Purchasing Power Parity theory (PPP) to hold in the short run, while evidence in the long run has been mixed. A problem of many tests was an inadequate specification of PPP as a dynamic intertemporal theory. Roll's extension of this theory (1979) based on efficient markets (EPPP), overcomes that shortcoming. With few exceptions, studies carried out for the cases of the developed countries support the efficient markets view of PPP. No studies have been carried out for the case of developing countries. The purpose of this study is to investigate whether EPPP holds for the case of the Latin American currencies. Exchange rate and inflation rate data gathered from the International Monetary Fund's International Financial Statistics covers the period January 1970 to December 2000. The paper is organized as follows. Section I reviews the issues concerning PPP and the empirical evidence related to efficient purchasing power parity tests. Section II presents the model and the data, underlying the hypotheses to be tested derived from the efficient purchasing power parity propositions. Section III presents the empirical results. Two regression tests and a unit root test are performed; the first regression aims at determining whether or not past exchange rates, adjusted for inflation rates, contain any information to predict future spot rates. The second regression tests whether real exchange rates follow a martingale process, which is then complemented with unit root test to determine whether the series are stationary. prior to these tests, the basic stochastic characteristics of the exchange rates series are examined. The conclusions, in Section IV, offer some suggestions for policy making.

 

Purchasing power parity and efficient exchange markets

For the international investor, risk in international capital and money markets is closely associated with exchange rates. For corporations operating internationally, transactions and economic risk are also determined by exchange rates. Finally, macroeconomic performance depends on exchange rate stability and timely adjustments mainly to avoid overvaluation of the domestic currency, followed by drastic devaluation adjustments. Purchasing Power Parity (PPP) is the first well-developed, but very controversial theory of exchange rate determination in international finance (Taylor and Taylor, 2004). According to the traditional PPP theory, as originally defined by Cassel (1916; 1921), in perfect goods and financial markets identical goods must have the same real price everywhere. Otherwise commodity arbitrage will take place (Law of One Price). Assuming that every country consumes the same basket of goods, this theorem also applies to the national price indices. In other words, the variation in the exchange rates for two currencies should be equal to the inflation differential in the two countries over a period of time, equal in magnitude but opposite in sign (relative version of PPP).

PPP is a fundamental concept in international economics and also has important implications both for the financial manager of international portfolios as well as for the financial corporate manager. It is also an important guideline for policy makers from central banks; their adjustments to exchange rates should respond promptly to inflation differentials with their main trade partners to avoid extended overvaluations or undervaluations that undermine their currencies and lead to large delayed adjustments that become the root of financial crises. Although PPP is supposed to hold in the long run, short term deviations from PPP induce cross-border transfers of commodities and capital. Most models of exchange rate determination are based largely on the long-run validity of the PPP proposition (Dornbusch, 1976; Mussa, 1982; Abuaf and Jorion, 1990; and Cuddingham, 1998; Sarno and Taylor, 2002; Coakley et al, 2005, among others). The adoption of flexible exchange rates since the early 1970's induced theoretical and empirical research, refining existing models about exchange rate determination. PPP provides an easy and inexpensive way of making medium-to-long-run predictions about exchange rate movements. Sustained deviations of the current real exchange rate from its long-run equilibrium level create economic exposure for the firm, excessive exchange rate risk for international investors, and great macroeconomic fragility towards external shocks, which might end in severe currency and financial crises. There is no practical reason why the equilibrium real exchange rate should not vary through time as sustained by PPP. The path of the real exchange rate compatible with the attainment of internal and external equilibrium is affected by changing world conditions, productivity improvements, adjustments to trade barriers, and changes in taxation, among other factors (Edwards, 1989). Globalization has led to an increased importance of capital flows, particularly foreign direct and portfolio investments as determinants of international reserves and exchange rate levels (Agénor and Hoffmaister, 1998; Bohn and Tesar, 1998; Goldberg and Klein, 1998; and Ortiz, 2000).

One of the most extensive reviews of the earlier tests of PPP was undertaken by Officer (1976). Since then evidence has been accumulating that demonstrates PPP's failure to hold in the short run. For instance, Frenkel (1976, 1981), Hakkio (1982), Krugman (1978), Dornbusch (1980; 1985), Broadberry (1987), Edison (1987), Murray and Papell (2002), and Taylor (2002) all confirm this result.¹ Considering the fact that many studies asserted that real exchange rates behave like a random-walk, Roll (1979) argued that a problem of most tests on PPP is an inadequate specification of PPP as a dynamic intertemporal theory. He formulated a superior theory of PPP from an efficient markets perspective based on international commodity arbitrage, i.e. the efficient markets PPP (EPPP).² Later, Adler and Lehmann (1983) developed another version of the efficient markets PPP based on financial arbitrage in bonds. Empirical evidence for EPPP can be found in the works of Roll (1979), Darby (1980), Adler and Lehmann (1983), Koveos and Seifert (1985), Pippenger (1986), Huang (1987), Witt, Jr. (1992), Taylor, 2002 and others. On the whole, the empirical evidence supports the efficient markets view of PPP for most industrialized countries. A notable exception is Huang (1987) who reports that expected nominal exchange rate changes appear to deviate systematically from expected inflation rate differentials supporting the presence of time-varying risk premia in foreign exchange markets. More recently, Abuaf and Jorion (1990) re-examined the evidence for PPP using a first-order auto-regression model in a multivariate setting. They show that long run PPP might indeed hold, even when there are substantial short term deviations from the parity condition. Examining the Australian case, Olekalns and Wilkins (1998), estimating a fractionally integrated arma model, find that PPP does have relevance for the long run behavior of the exchange rate.

Previous studies have by and large been restricted to early time periods and especially to industrial countries.³ The purpose of this paper is to investigate whether EPPP, as identified by Roll (1979),⁴ holds for the case of the Latin American currencies, during the period January 1970-December 2000. During the last three decades of the 20th Century the Latin American countries underwent recurrent crises characterized, among other things, by severe exchange rate imbalances followed by drastic government-determined adjustments, complemented by market adjustments. Moreover, during the period under analysis, exchange rate authorities from the area adopted several exchange rate regimes attempting to cope with recurrent balances in their economies, as well as to respond to the challenges derived from opening up their economies. Towards the end of the century freer market exchange rate activity was promoted by their policy makers. In the long run, exchange rates in Latin America should have adjusted taking into account their inflation rates' differential with their main commercial and financial partner, the United States of America.

 

Efficient purchasing power parity models and data

EPPP is based on the constraint that, in efficient markets, the real return to an investor from intertemporal speculation in goods is anticipated to be zero. This paper investigates three testable implications of the efficient markets hypothesis, as suggested by Roll (1979). The EPPP hypothesis stipulates that all available information is utilized by the market participants such that the present spot exchange rate contains all the information to predict the future spot rate adjusted for the inflation differential. Lagged prices contain no information useful in predicting future prices or obtaining extraordinary gains; that is, from a dynamic perspective, predictability based on past information means a departure from equilibrium conditions. The first testable version of EPPP can be expressed in a regression format as follows:

where X_t is the natural logarithm of the spot exchange rate adjusted for the inter-country inflation differential in period t (i.e., X_t = In S_t - DI_t where DI_t is the difference in the continuously compounded inflation rate between the home country and the foreign country. S_t-1 is the spot exchange rate in period t-1. The efficient markets version of PPP would be supported if equation (1) results in the b₁ coefficient being equal to unity and the other coefficients being zero.

Second, EPPP also implies that the real exchange rates follow a martingale process. Therefore deviations from PPP from one period to the next should be serially independent (Adler and Lehmann (1983). Equation (2) can be used to test this hypothesis:

where Y_t is the difference between the rate of change in the spot exchange rate (In S_t - In S_t-1) and the inter-country inflation differential (DI_t ) in period t (i.e., Y_t = (In S_t - In S_t-1) - DI_t).

The random walk hypothesis implies that the b_i (i = 1,...,6) coefficients should be zero for all i. Both equations, (1) and (2) are estimated to test the relevance of EPPP for the Latin American currencies.

Finally, updating EPPP, if the time series of changes in the exchange rate follow a martingale process, it should therefore be characterized by a random walk process; the time series should be a non-stationary series. Thus, to support the EPPP we should be able to prove that the changes in real exchange rates have a unit root. The Augmented Dickey-Fuller Test (ADF) and the Phillip-Perron test are used to test this hypothesis. The two statistics test for a unit root in the univariate representation of a time series. For a series Y_t the ADF test (Dickey and Fuller, 1979) consist of a regression of the first difference of the series against the series lagged k times as shown in equation (3):

The null and alternative hypotheses are: H₀: .λ = 0; H₁: λ < 1; acceptance of the null hypothesis implies non-stationarity. To control for higher-order correlation in a series the ADF approach adds lagged differenced terms on the right side of the equation. Similarly, the Phillip-Perron test (1988) aims at controlling for higher-order serial correlation in a series making a correction to the t-statistic of the λ coefficient of the AR(1) regression to account for the serial correlation on ε. Unit root tests have become useful due to their increased testing power. Important long-term tests of PPP have been recently carried out by Lothian and Taylor (1996) and Cuddington and Liang (1998). The former conclude that PPP is valid in the long run for bilateral real rates of exchange. The evidence by Cuddington and Liang contradicts those findings; using a two hundred-year series for dollar-sterling real rates they find that, choice in the lag length might influence the results,⁵ or else deterministic trends and structural breaks may give rise to non-stationarity. However, their findings are limited to real exchange rates. This study extends the unit root test to the series of changes in exchange rates to complement the martingale test proposed by equation (2).⁶

The primary source of data for this study is the International Monetary Fund's International Financial Statistics, which includes end-of-month exchange rates relative to the U.S. dollar and end-of-month consumer price indices. The exchange rate data used in testing the efficient markets hypotheses cover the period January-1970-December 2000.⁷ Data was gathered for 15 countries: Argentina, Bolivia, Brazil, Colombia, Costa Rica, Chile, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Paraguay, Peru, Uruguay, and Venezuela.⁸ Exchange rate series were adjusted for changes in the numeraire taking place during the period under analysis. The original January 1970 price of the domestic currency for the dollar was maintained as a point of reference to evaluate and compare changes throughout time. Inflation rate series were adjusted to a uniform January 1970 base.

 

Empirical results

Basic Statistics

Tables 1 to 3 summarize the main stochastic characteristics of the Latin American currencies on a long-term basis.⁹ Statistics are shown for the bilateral local currency in relation to the U.S. dollar. Table 1 shows the basic statistics for the local currency price of dollars in nominal terms. Table 2 shows those statistics in real terms. Table 3 shows the basic statistics for the series of changes (returns) in the real exchange rates from the Latin American countries in the sample. The data includes a total of 372 monthly observations. To apprehend fully the nature of the series, these statistics are shown in terms of the original price of the dollar in the local currency. Two meaningful situations can be identified: a) exceedingly high changes in the price of the dollar from January 1970 to December 2000 coupled with large volatilities; and, b) lack of normality.

Moreover, the evidence shows that the largest Latin American countries, Argentina Brazil, Mexico and to a lesser extend Colombia, Chile, peru and Venezuela suffered the largest exchange rate changes. In nominal terms, in the case of Argentina, the price of the dollar was 3.50 pesos (minimum for the period), ending with an unbelievable maximum price of 100 000 000 000 (old) pesos per dollar; the average for the period was 33 956 905 017, and the standard deviation amounted to 46 359 116 966 points. These facts, as a lesson from the past, show the unsustainability of the one peso per dollar followed during the decade of the 1990s. The case of Brazil is even more dramatic. As shown in Table 1, its currency varied from a minimum of 4.42 cruzeiros per dollar to a maximum of 2 963 974 397 190 (old) cruzeiros per dollar.¹⁰ The average was 415 736 595 660 and the standard deviation was situated at 810 563 628 758 points. Finally, Mexico was the least affected of the three large Latin American economies. In nominal terms the dollar price changed from a minimum of 12.49 to a maximum of 10 174.50 pesos per dollar. The average was 2 368.31 pesos per dollar and the standard deviation amounted to 3 210.26 points. The medium sized Latin American economies (Colombia, Chile, Peru and Venezuela) followed similar, but somewhat less dramatic patterns of change in their exchange rates. The most severe changes took place in Peru. Its exchange rate in old nominal soles changed from a minimum of 38.70 to a maximum of 353 100 000 soles per dollar; the average amounted to 76 758 095 soles per dollar and the standard deviation was 118 968 812 points. Finally, it is worth noting that the smaller Latin American countries presented less severe changes. Furthermore, those from Central America experienced rather mild changes. The case of El Salvador best exemplifies this case. In nominal terms, the price of the dollar in colones changed only from a minimum of 2.50 to a maximum of 9.30; the average was 4.89 and the standard deviation was 2.76 points. However, the small South American countries, rather resembled the patterns present in the large Latin American economies. In particular, in the case of Chile, its exchange rate changed from a minimum of 11.68 pesos per dollar in January 1970 to a maximum of 572 680 (old) pesos per dollar by the end of the period. The average amounted to 197 144.68 pesos per dollar and the standard deviation was 183 291.81 points.

The exchange rate series for the Latin American currencies are also characterized by a lack of normality. As shown in Table 1, in all cases the Jarque-Bera statistics confirm the absence of normality. In this respect, the fact stands out that seven countries —Brazil, Colombia, Ecuador, Honduras, Mexico, Uruguay, and Venezuela— have leptokurtic curves, while the remaining 8 countries —Argentina, Bolivia, Costa Rica. Chile, El Salvador, Guatemala, paraguay and peru— have platykurtic curves. Finally, all the 15 countries in the sample have curves skewed to the right.

Similar statistical behavior is present in the case of real exchange rates and changes in real prices. APPArently, in the long run, exchange rate maladjustments to inflationary trends appear to be small for all Latin American currencies. Comparing the nominal rate with the adjusted (real) inflation rate differentials, slight overvaluations can be discerned for the cases of Argentina, Bolivia, Chile, Mexico and peru; minimal cases of undervaluation are present in the cases of Brazil, Ecuador, El Salvador and Uruguay. Colombia, Costa Rica, Guatemala, Honduras, paraguay and Venezuela had real rates on target with inflationary changes. Highlighting the case of Brazil, measured in old cruzeiros, in real terms the dollar should have appreciated from 4.34 units to 2 939 778 687 830 cruzeiros per dollar, which is slightly below 0.82 percent, the (overvalued) nominal value reported earlier. Finally, examining the series of changes in the exchange rates (returns for the foreign exchange investor), the large volatilities in the Latin American foreign exchange markets are confirmed. In real terms the standard deviation for monthly variations of exchange rates varied between 2.039 points in the case of Colombia to a maximum of 14.365 points in the case of Argentina.

 

Efficient PPP Tests

The estimated coefficients and the results of the hypothesis tests pertinent to equation (1) are presented in Table 4. The null hypothesis, formulated in accordance with EPPP, is that the coefficient of the previous period's spot exchange rate (t-1) is equal to one and that the coefficients of other past exchange rates adjusted for inflation are equal to zero. The t-statistic can be used to test the significance level for each individual coefficient. An F statistic and Chi-square tests can be utilized to test the hypothesis that b₁ = 1 and b_i = 0 (i > 1).

These results provide only weak support for the efficient markets version of PPP. The coefficient for the spot exchange rate in the previous period is close to unity (i.e., b₁ = 1), and aPPArently statistically significant, only for the cases of five countries: Brazil (1.0127), Ecuador (0.9967), Mexico (1.0449), Paraguay (1.0568), and Uruguay (0.9644). Only in the case of Ecuador are all remaining coefficients, b₂ to b₇, fairly close to zero, near to 0.05 in six cases and near to 0.07 in one case. However those coefficients are not statistically significant. For the cases of Brazil, Mexico, Paraguay and Uruguay there are some coefficients, b₂ to b₇, that are significantly distant from zero. For example, in the case of Brazil b₂ = 0.1701, b₃ = -0.3728, and b₄ = 0.1679. In the case of Mexico, b₂ = 0.2146 and b₅ = 0.2008. Most of these coefficients are statistically significant. Moreover, the high R square and insignificant t-tests for the remaining bi coefficients signal multi-colinearity. Therefore, we can convincingly reject the hypothesis that b₂ = b₃, --- b₇ = 0. This is confirmed by the Wald test. As shown in Table 4 the F and Chi-square statistics decisively reject the null hypothesis at a one percent significance level. Thus, on a long-term basis, past spot rates adjusted for inflation from previous months seem to contain some information about current spot exchange adjusted for inflation. This reflects the fact that authorities from Latin American central banks have tended to control foreign exchange markets and have also adjusted exchange rates with significant delays. Similarly, adjustments based on market activity, or on the application of different exchange rate regimes have often been unable to keep pace with inflation and overall economic conditions. Thus, past exchange rates contain valuable information to predict current spot rates. This would clearly be the case with pegged exchange rates and various forms of sliding currency adjustments. Indeed, especially before the debt crisis, Latin American countries that pegged their currency to the dollar, themselves later adopted different exchange rate regimes, from a tight, dirty float to programmed sliding rates. By the 1990's many economies tended to promote free exchange markets, or else to adopt some currency convertibility schemes, namely currency boards as in the case of Argentina.¹¹

Finally, it is worth noting that for the cases of Argentina, Costa Rica, Chile, El Salvador, Guatemala, Honduras, and Venezuela, the coefficient bl is significantly higher than one; and for the cases of Bolivia, Colombia and Peru, this coefficient is significantly lower than one. In addition, some other coefficients, b₂ to b₇, for all these ten countries are significantly above zero, and most show high t-statistics. In short, for the 15 Latin American countries under study, the first efficient purchasing power hypothesis must be rejected. Their markets are inefficient and past (monthly) exchange rates do contain some valuable information about the current spot rate.

Table 5 summarizes the results of the tests of equation (2). The null hypothesis that the real exchange rates differentials follow a martingale process is not supported for the Latin American case, on a long term basis. Indeed, 11 countries in the sample show several b₁ to b₆ coefficients greater than zero. This is the case of Argentina, Bolivia, Brazil, Colombia, Costa Rica, Chile, Ecuador, Mexico, Peru, Uruguay and Venezuela. Similarly, most coefficients relatively close to zero are not statistically significant. It is worth noting the cases of Argentina, Brazil and Mexico. In the case of Argentina four coefficients depart significantly from zero (b₁, b₂, b₃ and b₅) and are statistically significant; the remaining coefficients b₄ and b₆ are near to zero, but are not statistically significant. For Brazil two coefficients, b₁ and b₂, are not close to zero and their t-statistic is significant; the remaining coefficients are relatively close to zero but are not statistically significant. Finally, in the case of Mexico four coefficients depart from zero, b₁, b₂, b₂, b₃ and b₆, and their t-statistic is greater than two; the remaining two coefficients, b₄ and b₅ , are close to zero but are not statistically significant. Countries for which all coefficients b₁ to b₆ are close to zero are El Salvador, Guatemala, Honduras and Paraguay. Nevertheless in none of these cases is the t-statistic significant. These results indicate that the difference between the rate of change in the spot exchange rate and the inter-country inflation differential is correlated for the Latin American currencies, considering a six-month reference horizon for traders in the exchange markets. The F-statistic and the Chi-square test from the Wald test confirm this result. According to this test, the hypothesis that all of the coefficients in equation (2) are equal to zero can be rejected at a 1% level of significance. However, it is worth noting that the R square statistic is very low in all cases.

The unit root tests confirm the previous results. As shown in Table 6, for all fifteen Latin American currencies in the sample, the real price series have a unit root. The t- statistic for both the ADF and Phillips-Perron test are smaller than the critical value needed to reject the null hypothesis at either one percent, five percent or 10 percent levels of significance. These series therefore follow a random walk process and are therefore non-stationary, However, the series of changes in exchange rates (Table 7), also analyzed in equation (2) are stationary; the series do not follow a random walk process, again for the case of all Latin American currencies.

Furthermore, the existence of stationarity in the real exchange rate series can be explained by the findings by Cuddingham and Liang (1998); in some cases, this is due to the presence of time trends and structural breaks. This possibility is consistent with the behavior of the Latin American currencies, considering the recurrent crises and stop-go patterns of growth characterizing their economies during the last three decades of the 20th Century. It is worth mentioning that applying a Chow break point test, five countries (Argentina, Bolivia, Costa Rica, El Salvador, and Peru) show a rupture in their series of real exchange rates in the 1980's. Meanwhile, for the remaining 10 countries the rupture in that series took place during the last decade of the last century. Table 8 shows these results. Thus, although there are some regional similarities in the long-run behavior of exchange rates in the Latin American countries, particularly evident in Tables 1-3, results from Tables 4-8 underlie the heterogeneity of the region.¹² Finally, it must be pointed out that the unit root test for the Latin American currencies differs from that presented by Kahn and Parikh (1998) for the South African case. Despite drastic changes in exchange rate policy, they found no evidence of unit root non-stationarity, and the behavior of the real exchange rate was stable, not constant.¹³

In sum, the three tests applied to the Latin American currencies, for the period 1970-2000, do not support the EPPP theory. In terms of efficiency, pegging, excessive control over the their exchange rates, and delayed adjustment of the exchange rate vis-à-vis the U.S. dollar, exchange markets in the region have been made inefficient. Past prices and past changes in the exchange rate seem to contain some useful information about the present levels of Latin American exchange rates. Furthermore, the empirical evidence is in disagreement with the results for Latin American black market exchange rates, as reported by Koveos and Selfert (1985) and by Diamandis (2003) for parallel markets. Using market exchange rates, reported for the case of 15 Latin American currencies, the results are not favorable to their conclusion that the efficient markets version of PPP appears to be the appropriate framework for many currencies in Latin America; exchange rate markets in the region are inefficient. Seemingly, by the year 2000, exchange rates have practically adjusted to past inflationary trends, as shown by the evidence from Tables 1 and 2. However, the financial literature gives ample evidence of rather delayed, drastic, crisis-generating exchange adjustments traditionally being implemented by exchange rate authorities from the region. Only during the past decade has the market become an important mechanism to adjust exchange rates in line with inflation differentials in the rest of the world.

 

Conclusion

This paper has investigated whether the efficient markets version of Purchasing Power Parity theory holds for Latin American currencies for the 1970-2000 period. Two tests of the EPPP with seemingly unrelated regressions were used and, in addition, two unit root tests were applied. In general, the empirical evidence obtained does not favor the EPPP. Results suggest inefficient foreign exchange markets in the region, resulting both from weak exchange rate policies and weak foreign exchange market development. Concerning exchange rate policies, the evidence also suggests that the various exchange rate regimes adopted by governments from the region, ranging from tightly controlled markets, to managed sliding mechanisms, and fully or nearly fully free markets, were insufficient and inappropriate to deal with the extraordinary changes that the Latin American economies underwent during the last three decades of the 20th Century. Furthermore, contrary to prior evidence that the efficient markets version of PPP generally holds, this conclusion cannot be generalized for the Latin American case, for the period under study. Finally, although the evidence from the econometric tests also suggests some regional similarities in long run exchange rate behavior, the results obtained also underscore the heterogeneity of the region; patterns of maladjustment in exchange rates differ from country to country and the structural breaks that can be related to such processes are also different. The evidence also implies the need to strengthen foreign exchange market activity as a means to maintain more stable exchange rates and avoid cyclical economic crises which in Latin America have consistently been triggered by currency crises. In this respect, since markets might be inefficient in themselves, exchange rate authorities should also complement market activity by implementing timely, dynamic adjustments based on close monitoring of inflation rate differentials with their main trade partners. These policies should be complemented by the creation of exchange rate derivatives to overcome the limitations of the incomplete markets which still characterize many Latin American countries. Since the region's currencies have been subject to tight government controls, but have been moving towards freer markets in response to their recurrent economic crises and to the challenges of globalization, further research will be necessary in the near future (with more accumulated data) to test PPP for these economies in the long run, particularly determining optimal structural breaks to examine and compare PPP market adjustments from recent periods vis-a-vis PPP adjustments resulting from previous exchange rate regimes.

 

Bibliography

Abuaf, N. y Ph. Jorion, "Purchasing power parity in the long run", Journal of Finance 45 (1), 1990, pp. 157-154.         [ Links ]

Adler, M. y B. Lehmann, "Deviations from purchasing parity in the long run", Journal of Finance, 38 (4), 1983, pp. 1471-1478.         [ Links ]

Agénor, P-R. "Capital flows and the real exchange rate: analytical framework and econometric evidence", en Glick, Reuven (editor), Managing Capital Flows and Exchange Rates. Perspectives from the Pacific Basin, Cambridge: Cambridge University Press, 1998, pp. 100-140.         [ Links ]

Apte, P., P. Sercu y R. Uppal, "The exchange rate and purchasing power parity: extending the theory and tests", Journal of International Money and Finance 23 (4), June 2004, pp. 553-571.         [ Links ]

Baum, C., J.T. Barkoulas y M. Caglayan, "Long memory or structural breaks: can either explain non-stationary exchange rates under the current float?" Working Paper, Department of Economics, Boston College, 2000.         [ Links ]

Bohn, H. y L.L. Tesar. "US portfolio investments in Asian capital markets", en Glick, Reuven (editor), Managing Capital Flows and Exchange Rates. Perspectives from the Pacific Basin, Cambridge: Cambridge University Press 1998, pp. 43-72.         [ Links ]

Breuer, J.B., "An assessment of the evidence on purchasing power parity", en Williamson, J. (editor), Estimating equilibrium exchange rates, Washington, D.C.: Institute for International Economics, 1994.         [ Links ]

Broadberry, S.N., "Purchasing power parity and the Pound-dollar rate in the 1930's", Econometrica, 54 (February), 1987, pp. 69-78.         [ Links ]

Cassel, G., "The Present Situation of the Foreign Exchanges", Economic Journal 26, 1916.         [ Links ]

----------, The World's Monetary Problems, London: Constable & Co., 1921.         [ Links ]

Cheng, B.S., "Beyond the purchasing power parity: testing for cointegration and causality between exchange rates, prices, and interest rate", Journal of International Money and Finance 18, (6), December 1999, pp. 911-924.         [ Links ]

Choi, I., "Unit root tests for panel data", Journal of International Money and Finance 20 (2) April, 2001, pp.249-272.         [ Links ]

Coakley, J. et al., "Purchasing power parity and the theory of general relativity: The first tests", Journal of International Money and Finance, vol.25 (2), March, 2005, pp. 293-316.         [ Links ]

Cuddingham, J.T. y H. Liang, "Re-examining the purchasing power parity hypothesis over two centuries", Working Paper, Department of Economics, Georgetown University, 1998.         [ Links ]

Darby, M.R., "Does purchasing power parity work?" Working Paper 607, National Bureau of Economic Research, 1980.         [ Links ]

Diamandis, P.F., "Market efficiency, purchasing power parity, and the official and parallel markets for foreign currency in Latin America", International Review of Economics and Finance, 12, (1), 2003, pp. 89-110.         [ Links ]

Dickey, D. y W. Fuller, "Distribution of the estimators for autoregressive time series with unit root", Journal of the American Statistical Association, 74, 1979, pp. 427-431.         [ Links ]

Dornbusch, R., "Expectations and exchange rate dynamics", Journal of Political Economy, (84), 1976, pp. 1161-1176.         [ Links ]

----------, "Exchange rate economics: where do we stand?", Brookings Papers on Economic Activity, 1, 1980, pp. 143-206.         [ Links ]

----------, "Purchasing power parity", Working Paper 1591, National Bureau of Economic Research, 1995.         [ Links ]

Edison, H.J., "Purchasing power parity in the long run: A test of the dollar/pound exchange rates (1890-1978)", Journal of Money, Credit and Banking, 19 (August), 1987, pp. 376-387.         [ Links ]

Edwards, S., Real Exchange Rates, Devaluation, and Adjustment, Cambridge: MIT Press, 1989.         [ Links ]

Elton, E.J. et al., Modern Portfolio Theory and Investment Analysis, New York: John Wiley & Sons, 2003.         [ Links ]

Engel, C., "Long-run PPP may not hold after all", Journal of International Economics 51 (2), August 2000, pp. 243-273.         [ Links ]

Fleissig, A.R. y J. Strauss, "Panel unit root tests of purchasing power parity for price indices", Journal of International Money and Finance 19 (4), August, pp. 506.         [ Links ]

Frenkel, J.A. "A monetary approach to the exchange rate: Doctrinal aspects and empirical evidence." Scandinavian Journal of Economics, (78), 1976, pp. 200-224.         [ Links ]

----------, "Flexible exchange rates, prices and the role of news: lessons from the 1970's", Journal of Political Economy, 89 (August), 1981, pp. 665-706.         [ Links ]

Ghartey, Edward E., "Random walk as a universal test of weak-form foreign exchange market efficiency: A proof", Frontiers in Finance and Economics, 1 (1), 2004, pp. 37-45.         [ Links ]

Goldberg, L.S. y M. Klein, "Foreign direct investment, trade and real exchange rate linkages in developing countries", en Glick, Reuven (editor), Managing Capital Flows and Exchange Rates. Perspectives from the Pacific Basin, Cambridge: Cambridge University Press, 1998, pp. 73-100.         [ Links ]

Hakkio, C.A., "A re-examination of purchasing power parity: a multi-country and multi-period study", Working Paper 865, National Bureau of Economic Research, 1982.         [ Links ]

Huang, R. D., "Expectations of exchange rates and differential inflation rates: further evidence of purchasing power parity in efficient markets", Journal of Finance, 42 (1), 1987, pp. 69-79.         [ Links ]

Ishii, S. et al., Exchange Rate Arrangements and Foreign Exchange Markets. World Economics and Financial Surveys. Washington, D.C.: International Monetary Fund, 2003.         [ Links ]

Johnston, B. et al., Exchange Rate Arrangements and Currency Convertibility. World Economics and Financial Surveys. Washington, D.C.: International Monetary Fund, 1999.         [ Links ]

Kahn, B. y A. Parish, "Does purchasing power parity survive political shocks in South Africa?", Weltwistchafliches Archive, 134 (1), 1998, pp. 99-116.         [ Links ]

Kargbo, J.M., "Cointegration tests of purchasing power parity in Africa", World Development, 31 (10) October 2003, pp. 1673-1685.         [ Links ]

Koveos, P. y B. Seifert, "Purchasing power parity and black markets", Financial Management, (Autumn), 1985, pp. 40-46.         [ Links ]

Krugman, P.R., "Purchasing power parity and exchange rates: another look at the evidence." Journal of International Economics, 8 (August), 1978, pp. 397-407.         [ Links ]

Lothian, J.R. y M.P. Taylor, "Real exchange rate behavior: the recent float from the perspective of the past two centuries", Journal of Political Economy, 104 (3), 1996, pp. 488-509.         [ Links ]

Mkenda, B.K., "An empirical test of purchasing power parity in selected African countries-A panel data approach", Working Papers in Economics, núm. 39, Götemborg University, 2001.         [ Links ]

Murray, C.J. y D.H. Papell, "The purchasing power parity persistence", Journal of International Economics, 56, 2002, pp. 1-19.         [ Links ]

Mussa, M.L., "A model of exchange rate dynamics", Journal of Political Economy, 90, 1982, pp. 74-104.         [ Links ]

Ng, S. y P. Perron, "Lag length selection and the construction of unit root tests with good size and power", Econometrica, 69, 2001, pp. 1519-1554.         [ Links ]

Officer, L., "The purchasing power parity of exchange rates: a review article", International Monetary Fund Papers, 83 (May), 1976, pp. 1-60.         [ Links ]

Olekalns, N. y N. Wilkins, "Re-examining the evidence for long-run purchasing power parity", The Economic Record, 74 (224), 1998, pp. 54-61.         [ Links ]

Ortiz, E., "La Inversión extranjera de portafolios en los mercados de dinero y capital de México y su impacto en la crisis mexicana", en Irma Manrique (coordinadora), Arquitectura de la crisis financiera, México, Miguel Ángel Porrúa, 2000, pp. 191-220.         [ Links ]

Parikh, A. y E. Wakerly, "Real exchange rates and unit root tests", Weltwirtschaftliches Archiv, vol.136 (3, 2000), pp. 478-90.         [ Links ]

Perron, P., "Further evidence on breaking trend functions in macroeconomic variables", Journal of Econometrics, 80, 1997, pp. 355-385.         [ Links ]

---------- y T. Vogelsang. "Non-stationarity and level shifts with application to purchasing power parity", Journal of Business and Economic Statistics, 10 (3), 1992, pp. 301-320.         [ Links ]

Phillips, P.C. y P. Perron, "Testing for a unit root in time-series regression", Biometrica, vol. 74, 1988, pp. 335-346.         [ Links ]

Pippenger, J., "Arbitrage and efficient market interpretations of purchasing power parity: theory and evidence", Economic Review, Federal Reserve Bank of San Francisco, Winter 1986 (1), pp. 31-47.         [ Links ]

Roll, R., "Violations of purchasing power parity and their implications for efficient international commodity markets", en Sarnat, M. y G. P. Szego (editores), International Finance and Trade, vol. I, Cambridge, MA: Ballinger Publishing Company, 1979, pp.133-176.         [ Links ]

Sarno, L. y M.P. Taylor, "Purchasing power parity and the real exchange rate", IMF Staff Papers, Vol.49 (1), 2002, pp. 65-105.         [ Links ]

Taylor, A.M. y M.P. Taylor, "The PPP debate", Economic Perspectives, vol.18 (1), 2004, pp. 135-158.         [ Links ]

----------, "A century of purchasing power parity", The Review of Economics and Statistics, vol. 84 (1), 2002, pp. 139-150.         [ Links ]

Witt, J. A., "The long-run behavior of the real exchange rates: a reconsideration", Journal of Money, Credit and Banking 24, 1992, pp. 72-82.         [ Links ]

Wu, J-L.; L-J. Tsaic y S-L. Chend, "Aire real exchange rates non-stationary? The Pacific Basin perspective", Journal of Asian Economics 15 (2) , April 2004, pp. 425-438.         [ Links ]

 

Notas

* An earlier version of this work was presented at the VIII Foro de Investigación, Congreso Internacional de Contaduría, Administración e Informática, octubre 22-24, 2003, organized jointly by Facultad de Contaduría y Administración from Universidad Nacional Autónoma de México and Asociación Nacional de Escuelas y Facultades de Contaduría y Administración (ANFECA). Proceedings of this Foro were published in electronic means (CD). The authors also wish to acknowledge and thank valuable comments received from Joan B. Anderson, University of San Diego; Ephraim Clark, Middlesex University London and Graduate School of Management, Lille, France;Vincent Dropsy, California State University, Fullerton; Dilip K. Ghosh, Rutgers University, and Francisco López Herrera, Universidad Nacional Autónoma de México. We also thank the observations and suggestions from three anonymous referees.

¹ An assessment of PPP studies can be found in Breuer (1994), Pippenger (1986) and Taylor and Taylor (2002).

² An important contribution from modern financial economics is the study of market efficiency from an informational point of view. Essentially, it means that current prices from financial assets fully reflect all available information. Three hypotheses have been advanced in this respect: the weak hypothesis, semi-strong hypothesis, and strong hypothesis. The weak hypothesis, or return predictability hypothesis, indicates that in efficient markets, information from historical prices is fully reflected in current prices. Past prices do not provide information to predict current prices and obtain extraordinary gains. The semi-strong hypothesis maintains that all publicly available information, including fundamentals about the economy, is fully reflected in financial asset prices. The strong hypothesis maintains that all information, whether public or private, is fully reflected in financial asset prices. These hypotheses therefore indicate different degrees of informational efficiency implying, in turn; efficiency in the allocation of resources. Numerous econometric tests can be used to test these hypotheses. Most frequently these tests have been used in the context of Modern Portfolio Theory (MPT), as the lack of efficiency means prices and returns can be predicted to obtain extraordinary gains. Important extensions of these hypotheses include the analysis of price and return patterns from financial assets to determine disequilibria whether caused by market forces, or weak policy-making due to economic or political causes. Thus, efficiency tests on capital markets, derivatives markets, interest rates and exchange rates have also become powerful evidence for analyses carried out by policy —and normative-oriented economists, political economists and economic historians. For an excellent review of informational market efficiency see Elton, et al. (2003).

³ An exception is Roll (1979) who examined 23 countries, including Argentina, Brazil, Chile, Mexico and Venezuela (1957-1976). Another notable exception is Koveos and Seifert (1985) who tested EPPP for the Latin American black market currencies during the period April 1973-March 1983. The work by Mkenda (2001) applied a panel data approach for selected African countries using annual data for the period 1965-1996.

⁴ Financial arbitrage is not included in our tests because emerging capital markets are only recently integrating themselves into global financial activity. Furthermore, there is no available long-run data for emerging markets' international transactions in bonds. We must also point out that recent research on efficiency depart from Roll's EPPP, stressing co-integration analysis (with key variables such as futures rates) and unit root tests, but mostly for recent time periods due to data availability from developing countries. (For recent PPP unit root and co-integration analysis see: Cheng, 1999; Engel, 2000; Fleissig and Strauss, 2000; Parikh and Wakerley, 2000; Choi, 2001; Diamandis 2002; Apte, et al, 2004; Wu, 2004; Kargbo, 2005). These econometric models are powerful tests for long run equilibrium analysis, but leave aside intertemporal analysis like the regression analyses proposed by Roll and applied in this study. Nevertheless, because this work also deals with a long-run assessment we complement our empirical work with unit root tests.

⁵ On this issue see: Ng and Perron (2001).

⁶ Related to this study, it is worth noting that employing the martingale process and the definition of rational expectations, Ghartey (2004) proves that the pure random-walk spot exchange rate is an adequate means to universally test foreign exchange market efficiency.

⁷ Actually, the exchange series data includes December 1969, end of period, i.e., opening price for January 1970, to have a complete series of 12 months of changes in the exchange rate for all years during the period under analysis.

⁸ Countries from the Region not included in the sample due to the lack of continuity in the series available, as a result of their political problems, are: Dominican Republic, Nicaragua and Haiti. In addition, Cuba is not a member of the International Monetary Fund and the U.S. dollar is the means of exchange in Panama.

⁹ All tables presented in this study are original, presenting results from the econometric tests carried out by the authors with the econometric software E-Views.

¹⁰ The cruzeiro changed its denomination several times during the period under analysis. Currently the Real is the official Brazilian currency. We maintained the cruzeiro denomination in this section only for analytic purposes.

¹¹ See: Johnston and others (1999), and Ishii (2003) and others, from this International Monetary Fund publication, for a fine assessment of exchange rate arrangements and foreign exchange markets during the last two decades.

¹² The structural breaks present in Table 8 also underscore the singularity of each country, as well as stressing the importance of historical analysis and policy-oriented studies, as pointed out above, note 2.

¹³ These two facts suggest the need for further studies on EPPP for the Latin American currencies, with full identification of optimal breakpoints. Because the breakpoints for the real exchange rates indicate short analysis periods, for the last decade, this study does not include further research on the EPPP. On the issues concerning unit root test and structural breakpoints see: Perron and Vogeslang (1992), Perron (1997), and Baum, Barkoulas and Caglayan (2000).