1. Introduction

After the Asian financial crises in the late 1990s, there is a prevailing suggestion that developing countries should adopt either a pure floating exchange rate or a hard peg regime to maintain capital flow stability. Following the suggestion⁽¹⁾ of the IMF, most of the major Latin American countries, such as Mexico, Brazil, Colombia and Chile implement pure floating and inflation targeting (FIT) policies. In 1999, all of them officially announce that they switched to an FIT regime⁽²⁾, although none of the Latin American FIT countries did really let their currency float the way assumed under a conventional FIT arrangement ^[10]. Evidence shows that central banks in these countries set their target interest rate in response to nominal exchange rate (NER) movements. Given that the NER is an important transmission mechanism for monetary policy, FIT central banks can use the interest rate to influence inflation through its effects on the NER ^[33].

However, the central banks in Latin America FIT countries intervene in the foreign exchange (FX) market regularly to affect NER movements through adjusting interest rates. Such behaviors are documented as “fear of floating” ^[7], albeit there are several arguments about this explanation⁽³⁾. First, all central banks have made explicit statements that they do not pursue exchange rate targets ^[10]. This implies that interventions are only undertaken when the exchange rate go far beyond the fundamental equilibrium. Second, the RERs in the FIT countries are substantially more volatile than in the countries with a stable and competitive real exchange rate (SCRER)⁽⁴⁾, such as Chile with the crawling bands (1985-1995) or Argentina with the managed floating after the convertibility crisis (2003-2008). In the other Latin America FIT countries, significant appreciation trends were also observed. When the inflation targets were threatened, central banks did not hesitate to induce NER appreciation to meet them. Therefore, investigating the long-run relationship between the real exchange rates (RERs) and real interest rate (RIR) differentials among Latin American FIT countries have important implication for policy makers and international investors.

Previous research on the long-run relationship between the RERs and RIR differentials mainly focuses on the developed economies, few studies examine the market in emerging Latin America countries. The theoretical framework of the law of one price (LOP) and uncovered interest parity (UIP) illustrates that there should be a systematic relationship between the RERs and RIR differentials, although the empirical evidence in support of the classical economic theories is controversial. Two earlier seminal papers by Campbell and Clarida ^[9] and Meese and Rogoff ^[32] both conclude that the RER movements cannot be explained by shifts in RIR differentials. Later, Coughlin and Koedijk ^[13], Blundell-Wignall and Browne ^[3] and Edison and Pauls ^[17] argue that the RERs and RIR differentials may be cointegrated given an extended sample period. After around 10 years, by using a longer period of time series data and advanced statistical techniques, Chortareas and Driver ^[12] reveal that the RERs and RIR differentials cointegration holds better for small open economies rather than G7 countries. More recently, both Nakagawa ^[35] and Kim ^[30] are in favor of a long-run linkage between the RERs and RIR differentials with the help of advanced dynamic statistical approaches.

This paper contributes to the previous research in the following ways. First, this is the first paper testing the long-run relationship between the RERs and RIR differentials in Latin America emerging markets, while previous research has overwhelmingly focused on major developed countries. Second, we investigate the cointegration and causality relation country-by-country to draw inferences on individual-specific characteristics; therefore, our methodology is quite different from the extant pooled panel approaches which could miss individual unique properties. Finally, we investigate the RER-RIR relationship in a dynamic way by constructing a Vector Error Correction Model (VECM) allowing for deviations from the long-run equilibrium. Thus, our model provides valuable information on how the equilibrium error in each series could be adjusted by the changes in subsequent periods.

To test for the long-run linkage hypothesis, we retrieve the monthly data over 1993-2009 to perform both Johansen cointegration test and Granger causality test for each nation. The empirical results show that the long-run linear relationship between the RERs and RIR differentials is supported in 3 out of 6 Latin America economies: Argentina, Chile and Columbia. Moreover, the causal relationship between the RERs and RIR differentials is detected for another 3 Latin America emerging countries: Brazil, Mexico and Venezuela, although the pattern of causality is mixed and country-specific. In sum, the causal relationship between the RERs and RIR differentials is found in most of the FIT countries at the 5% level over the past 17 years. The findings of this paper have important implications for policy makers and international investors in emerging markets.

The rest of the paper is organized as follows. Section 2 presents the literature reviews; Section 3 describes the data; Section 4 explains the methodology with a brief review of the theoretical background. In section 5, empirical evidence is reported and interpreted simultaneously. Section 6 discusses the limitations and concludes the paper.

2. Literature Review

The long-run relationship between the RERs and RIR differentials has drawn considerable attention since 1980s. The most well-known papers are those of Campbell and Clarida ^[9] and Meese and Rogoff ^[32]. Campbell and Clarida ^[9] examine whether the RER movements can be explained by shifts in RIR differentials. They find that expected RIR differentials have simply not been persistent enough, and their innovation variances are not large enough to account for much of the fluctuation in the dollar’s real exchange rate. Meanwhile, Meese and Rogoff ^[32] attempt to detect the cointegration relationship, and they find that they cannot reject the null hypothesis of no cointegration between long-term the RERs and RIR differentials. They suggest that a variable that may be omitted from the relationship, possibly the expected value of some future real exchange rate, may have a large variance which, if included, would lead to the finding of cointegration. This conjecture of an important missing variable is also in line with the empirical results of Campbell and Clarida ^[9].

While in 1990s, the contradicting empirical results of no cointegration between the RERs and RIR differentials become suspect because the time span used in the previous research is not persuasive. Coughlin and Koedijk ^[13] and Blundell-Wignall and Browne ^[3] argue that the RERs and RIR differentials may be cointegrated when the sample period is extended. By using more recent data to extend the sample period, Coughlin and Koedijk ^[13] find cointegration for the Mark/Dollar exchange rate. The ability of Blundell-Wignall and Browne ^[3] to find cointegration is due to the inclusion of the difference in the share of the cumulated current account relative to GNP in the relevant countries. In addition, Edison and Pauls ^[17] re-assess the relationship between the RERs and RIRs over the period 1974-1990 and conclude that there might be a long-run relationship between these variables, but it cannot be verified. Although most of the final conclusions in the 1990s showed little empirical evidence in support of a systematic relationship, the researchers also agree that the robustness of the results varies across exchange rates, time periods, and the measures of expected inflation rates.

However, in the 21^st century, with the help of longer period of time series data and advanced statistical techniques, more and more empirical evidence on the hypothesized link incline to support the sticky-price exchange rate theories. Chortareas and Driver ^[12]’s tests are conducted with a panel of 18 OECD economies in the post-Bretton Woods era. When they divide the sample into two subsamples--the G7 and eleven small open economies, they find strong evidence in favor of cointegration for the panel of small open economies. In contrast, there is weak evidence of cointegration in a panel that consists purely of the G7 economies. Their results of supporting a long-run relationship between the RERs and RIR differentials appear to be more positive. More recently, Nakagawa ^[35] conjectures that the empirical difficulty is due to a failure to recognize nonlinearity in real exchange rate adjustment. When he introduces threshold nonlinearity into a traditional model to take account of a transaction cost-induced “band of inaction” for price adjustment, only outside the band will exhibit mean reversion and bear an association with RIR differentials. Meanwhile, Kim ^[30] examines the link between the RERs and RIR differentials for traded and non-traded goods with a systematic method in a dynamic seemingly unrelated (SUR) regression for panel data. His empirical result shows that the linkage between the RERs and RIR differential is more favorable for traded goods than for general and non-traded goods.

Recent studies have a special interest in the link between exchange rates and interest rates in emerging market economies, this interest is further spurred by the fact that many of these economies have recently introduced changes in their foreign exchange and monetary policies, moving towards an inflation targeting framework which operate officially under flexible exchange rate regimes⁽⁵⁾. In the case of EMEs, both theoretical and empirical work should take into consideration the specificities of these economies regarding the behavior of interest rates and exchange rates. Authors such as Calvo ^[5], Calvo and Reinhart ^{[6, 7]} and Eichengreen have insisted that there are large discrepancies between advanced economies and EMEs. These differences include credibility problems, the presence of liability dollarization, a high degree of exchange rate pass-through⁽⁶⁾ and non-stationarity in the inflationary process. In this paper, we contribute to the previous research by performing advanced statistical techniques to investigate the cointegration and the causality relation between the RERs and RIR differentials for 6 major Latin American economies to draw inferences on policy strategies and portfolio diversification benefits for investors and policy makers in emerging markets.

3. DATA

The monthly data of the interest rate and exchange rate are collected for 6 Latin American emerging countries and the U.S. over the period 1993:M1 to 2009:M12⁽⁷⁾. The sample currencies include ARS (Argentine peso), BRL (Brazilian real), CLP (Chilean peso), COP (Colombian peso), MXN (Mexican peso) and VEF (Venezuelan Bolívar). The Consumer Price Index (CPI) for each nation is identified as the percent change over the same period of the previous year. In addition, we set the monthly CPI of the year 2000 as 100. The monthly exchange rate for each country is quoted as foreign currency units per unit of the U.S. dollar (USD). Further, we use annualized average rate on 3-month negotiable certificates of deposit as the benchmark interest rate. The interest and exchange rate data are retrieved from DataStream (Thomson Financial Limited, 2010). To check the data accuracy, we plot each time series and visually inspect the data for errors according to relevant literatures on structural breaks. The CPI data are drawn from the International Financial System (IFS), further details and the methodology of the CPI data specification are available on the International Monetary Foundation (IMF, 2010) official website.

4. Methodology

Dornbusch’s ^[16] seminal paper claims that goods market prices adjust sluggishly in response to anticipated disturbances and to excess demand. Consequently, less than perfectly anticipated monetary innovations can cause temporary deviations in the real exchange rate from its long-run equilibrium value. Further studies on sticky price assumptions are investigated by Frankel ^[20], Campell and Clarida ^[9] and Meese and Rogoff ^[32], etc. The interpretation of the empirical tests in this paper depends mainly on three assumptions, expressed in equations (1), (2), and (4) below. First, equation (1) is a feature of a broad class of sticky-price rational-expectations monetary models by Obstfeld and Rogoff ^[37]. It assumes that any temporary deviation of the real exchange rate from its flexible-price equilibrium value is expected to vanish at a constant rate (in the absence of further shocks). That is, the logarithm of the real exchange rate q_t follows:

(1)

where q_t=s_t + p_t* - p_t, s_t is the logarithm of the nominal exchange rate (domestic currency per foreign currency unit), p and p* are the logarithm of the domestic and foreign price level at time t, respectively. E_t(.) is an expectation operator at time t, is the real exchange rate that would prevail at time t if all prices were fully flexible, and θ (0< θ <1) is a speed adjustment parameter.

In general, and will equal, unless there are no real shocks or all real shocks are random-walk processes. If the ex-ante Purchasing Power Parity (PPP) holds in the long run, it also assumes that

(2)

Substituting equation (2) into equation (1) obtains

(3)

The third important building block is the Uncovered Interest Parity (UIP) relation, which can be written as

(4)

where is the k^th period foreign (domestic) nominal interest rate at time t. It implies that

(5)

where the k-period real interest rate r_t,kcan be obtained by subtracting the expected inflation rate from the nominal interest rate i_t,k.

Substituting equation (5) into equation (3) yields

(6)

Relaxing equation (4) by allowing for an exogenous risk premium will add a forcing factor to equation (6). Equation (6) relates the real exchange rates to the real interest rate differentials and to the flexible-price RERs. Thus, sticky price theories of exchange rate determination predict that an increase in domestic RIR relative to foreign rates gives rise to an appreciation of the domestic currency. However, in the empirical literatures, a link between the RERs and RIR differentials estimated by Meese and Rogoff ^[32] can be written as the following version of the equation

(7)

where α_i is a constant term for a domestic country i (i=1,…, N) and a composite of inflation forecast errors.

The assumption of the classical regression model necessitates that both the {q_it} and {r_it,k-r*_it,k}sequences be stationary and the errors have a zero mean and finite variance. In the presence of nonstationary variables, there might be what Granger and Newbold ^[24] call a spurious regression. In other words, the unit root issue arises in the context of the standard regression model.

The results from such spurious regressions are meaningless in that all errors are permanent. In this case, it is often recommended that the regression equation be estimated in first differences as follows

(8)

Since {q_it}, {r_it,k-r*_it,k}and {ε_t} each contains unit roots, the first difference of each is stationary. Hence, the usual asymptotic results apply. In section 5.4, we will check the stationarity property for each RER and RIR differential sequence.

If the nonstatioanry{q_it} and {r_it,k-r*_it,k}sequences are integrated of the same order and the residual sequence is stationary, {q_it} and {r_it,k-r*_it,k}sequences are cointegrated. This relation can be investigated by conducting Johansen ^[28] cointegration rank test.

If the cointegration between {q_it} and {r_it,k-r*_it,k}sequences does not exist, a bivariate VAR can be constructed to draw inferences about causal relationships. In this circumstance, If both coefficients in the VAR model are statistically significant, it implies bi-directional long-run causality ^{[26, 29]}.

5. Empirical Results

The real exchange rate {q_t} sequence is defined as the nominal exchange rate (s_t) that is adjusted by the ratio of the foreign price level (p_t*) to the domestic price level (p_t). In logarithmic form, it can be shown as

(9)

where p_t refers to each nation’s Consumer Price Index (CPI) and p_t* refers to the CPI of the U.S. in time t relative to a base year (The CPI of the year 2000 is set as 100).

The RIR differential is defined as the difference between the RIR in each Latin American economy and that of the United States. RIR is the nominal interest rate subtracting the inflation rate. Mathematically, it can be written as

(10)

where i and λ are the nominal interest rate and inflation rate for Latin American countries; i*and λ* are the nominal interest rate and inflation rate for the United States. Δi is the real interest rate differential.

Statistical inspection of data clearly highlights the explosion in inflation experienced during the early 1990s in Argentina, Brazil, and Venezuela. This evidence is also confirmed by the high sum of squared deviations in the last column of the summary statistics that are displayed in Table 1. The data elucidates that there is clear evidence of rejection of normality. Moreover, all series display significant skewness and excess kurtosis. Therefore, the null hypothesis of normal distribution is overwhelmingly rejected for all the series of the RERs and RIR differentials by the Jarque-Bera test ^[27] at the 5% level. We further eliminate the outliers (most of them are in early1990s) which may affect the robustness of the estimation, yielding 204 observations for each time series.

Table 1. Summary of Descriptive Statistics of the RERs and RIR Differentials of Latin America Countries

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View next table

To formally test for the presence of unit roots in each series, both the Augmented Dickey-Fuller (ADF) ^{[14, 15]} and Ng and Perron ^[36] tests are employed to examine the stationary property of the data. Because the ADF test is known to suffer potentially severe finite sample power and size problems, the Ng-Perron ^[36] unit root test, building on some of their own work ^[38] deals with both of these problems. To improve finite sample performance, we employ the Ng and Perron ^[36] approach as a preferred alternative to the traditional ADF test. The testing equation can be written as

(11)

where α is a constant, β is the coefficient on a time trend and p is the lag order of the autoregressive process. To determine the optimal lag length p, a 12-month maximum lag length is imposed, and Schwarts Information Criterion (SIC) is followed to select the optimal lag augmentation.

Table 2. Unit Root Tests of the RERs and RIR Differentials of Latin America Countries

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

View next table

Examining the ADF test results in Table 2, there is no evidence to suggest that the RER and RIR differential series are stationary in level at the 5% level except that the RIR differential of Argentina is marginally significant. Furthermore, we test for the stationary properties in the first difference of all the RER and RIR differential series with intercept and trend, we overwhelmingly reject the null of having a unit root for the all series at the 1% level by ADF test. Seemingly, we conduct the Ng-Perron test and get the similar results. We cannot reject the null of non-stationarity in level at the 5% level for all series except that the RIR differentials for Brazil and Mexico are significant at the 10% level. However, in the first difference, we uniformly reject the null of having a unit root for all the RER and the RIR differentials at the 1% level. In sum, all the selected RER and RIR differentials series over the sample period are following I(1) processes.

The method of Johansen ^[28] involves investigation of the p-dimensional vector autoregressive (VAR) process of k^th order:

(12)

where Δ is the first-difference lag operator, Y_t is a ( p×1) random vector of time series variables with order of integration less than or equal to one, I(1). μ is a ( p×1) vector of constants, Γi are ( p×p) matrices of parameters, ε_t is a sequence of zero mean p-dimensional white noise processes, and Π is a ( p×p) matrix of parameters. The rank (r) of Π contains information about long-run relationships among the variables. The tested hypothesis is that the number of cointegrating vectors is at most r (r=1, . . . , p), we conduct Johansen rank tests by using both the maximum-eigenvalue and trace test statistic on the above unrestricted VAR model.

An inspection of Table 3 tells us that cointegration between the RERs and RIR differentials exist in 3 out of 6 Latin America countries: Argentina Chile and Columbia. For the other 3 Latin American economies, no cointegration is found between the RERs and RIR differentials at least at the 5% significant level.

Table 3. Johansen Cointegration Tests Between the RERs and RIR Differentials of Latin America Countries

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

View next table

If two variables, Y_t and X_t, are cointegrated, then there must exist an error-correction representation of the form

(13)

where Δ is the first differenced operator (i.e., Δ Yt = Y_t - Y_t-1), e_it follows iid with zero mean and constant variance, Z_t-1 and Z^*_t-1 are the lagged residuals obtained from the following cointegration regressions:

(14)

Since we find significant cointegrations between the RER and RIR differentials in Argentina, Chile and Columbia, we further employ the above VECM models to investigate the speed of adjustment to equilibrium in these three markets. The results are shown in Table 4.

Table 4. VECM Estimations of the RERs and RIR Differentials of Latin America Countries

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

View next table

Table 4 confirms that the long-run cointegration holds for Argentina, Chile and Columbia over the sample period from January 1993 to December 2009. Comparing the adjustment coefficients of real interest rate differentials to those of the real exchange rates in Argentina, Chile and Columbia, we can readily conclude that the RIR differentials have an adverse response to the previous period’s deviation from long-run equilibrium, while the RERs have a positive response to the previous period’s deviation from long-run equilibrium. However, the RIR differentials are less responsive to their last period’s equilibrium error in all these three economies; while the responsiveness of the RERs to their last period’s equilibrium error is greater in magnitude. Only 0.36%, 1.76% and 0.23% of the equilibrium error in the RIR differentials could be adjusted by the subsequent period’s changes, while 1.07%, 2.84% and 2.96% of the equilibrium error in the RERs could be adjusted by the subsequent period’s changes in Argentina, Chile and Columbia, respectively. This is not surprising because the foreign exchange markets tend to react to the market information and policy shocks much quicker than the monetary market. Therefore, the short run dynamic deviations from the long-run relationship are more likely to be influenced by the RERs rather than the RIR differentials in Argentina, Chile and Columbia.

If the cointegration does not exist, the following bivariate VAR can be used to draw inferences about causal relationships. In equation (12), If both β₀ and β₁ are statistically significant, this indicates bi-directional long-run causality ^{[26, 29]}. If β₀ = β₁ = 0, there is no long-run equilibrium relationship between Y and X, then it reduces to the standard Granger (1983) causality test. The following equations are used to explore the subtle relations between Y and X

(15)

where Y_t is the real exchange rate, X_t is the real interest rate differential. Rejecting H₀: α₂₁=α₂₂=…=α_2k=0 implies that the RIR differentials Granger cause the RERs. Similarly, Rejecting H₀: β₁₁=β₁₂=…=β_1k=0 implies that the RERs Granger cause the RIR differentials.

We further employ a bivariate VAR model to explore the causal relationships between the RERs and RIR differentials of Brazil, Mexico and Venezuela. Table 5 illustrates that a causal relationship does exist in all the other Latin America countries, albeit the pattern of causality is mixed. Based on the Granger causality test results, we uniformly reject the null of no Granger causality between the RERs and RIR differentials in Brazil, Mexico and Venezuela at the 1 % level. It elucidates that interest rate changes in these countries could cause foreign exchange market fluctuations; on the other hand, the currency markets also have substantial impacts on the domestic money market. In conclusion, the detection of causal relationships between the RERs and RIR differentials in most of these Latin America countries indicates that the changes in RIR differentials and the changes in RERs are mutually affected in the long run, although the pattern of interactions is mixed and country-specific in Brazil, Mexico and Venezuela.

Table 5. The Granger Causality Tests Between the RERs and RIR Differentials of Latin America Countries

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

View next table

Table 6. The Evolution of Exchange Rate Stabilization Programs in Latin America Countries

Download as

PowerPoint Slide

Larger image(png format)

Tables index

Veiw figure

View current table in a new window

View previous table

6. Conclusion

This paper investigates the long-run cointegration relationship between the RERs and RIR differentials in 6 major Latin America emerging markets during the inflation targeting period. The empirical results show the long-run cointegration between the RERs and RIR differentials holds for Argentina, Chile and Columbia over 1993-2009. Moreover, the causal relationship between the RERs and RIR differentials is also supported for the other three Latin America countries: Brazil, Mexico and Venezuela, albeit the pattern of causality is mixed and country-specific. The results are robust and largely significant at the 5% level over the sample period.

We find either cointegrations or causal relations between the RERs and RIR differentials in most of the Latin American economies during the inflation targeting period. The reasons could be explained as follows: First, the Latin American financial markets have switched to a FIT regime since early 1990s, thus, most of the major Latin American economies have undergone currencies stabilization progresses (see Table 6). For example, Chile and Argentina have begun a gradual adjustment process in exchange rate, while other countries have also implemented some adjustment programs to eliminate their relatively high inflation trends. Second, intervention in foreign exchange (FX) has been regularly undertaken by several Latin American countries, such as Mexico and Venezuela. Third, many countries implement inflation targeting program. Since 1999, Brazil, Chile, Columbia and Mexico have fully implemented inflation targeting programs, although weaker forms of inflation targeting measures have been undertaken since early 1990s. For example, the soft inflation targeting policy in Chile started in 1991, and followed by Columbia and Mexico in 1995. ^[34]. Under such a regime, central banks in these countries set their target interest rate in response to nominal exchange rate (NER) movements. In addition, local currency operations in credit markets are gradually being displaced by dollar denominated operation in Latin America ^[40], for both deposits and loans. In sum, combining the stylized facts in Latin American financial markets, our empirical findings may have more important implications of portfolio diversification for policy makers as well as international investors in emerging markets.

Future studies can be conducted by employing more advanced statistical techniques, such as nonlinear models, threshold integration or panel cointegration approaches allowing for structural break. Moreover, one may better explain the dynamics between the RERs and RIR differentials for each country by observing the specific omitting variables between the relationships and test the possibilities of applications for other emerging markets. In addition, there are big discrepancies in inflation rates among Latin America countries in early 1990s, which call for some adjustments, such as the exchange rates of Brazil and inflation rates in most of the countries. Finally, we start the data from January 1993. Further studies can be done by extending the sample period back to 1980s if the adjustment on inflation rates is appropriate.

References

[1]	Alfaro L. 2002. On the political economy of temporary stabilization programs. Economics and Politics, 14,133-16.
	In article		CrossRef
[2]	Bahram A, Allender M, Raffiee K. 2003. Emerging markets and uncovered interest parity. Atlantic Economic Journal,. 31 (3):291-291.
	In article		CrossRef
[3]	Blundell-Wignall A, Browne F. 1991. Increasing financial market integration: real exchange rates and macroeconomic adjustment. OECD Department of Economics and Statistics Working Paper No. 96.
	In article
[4]	Ca' Zorzi M, Hahn E, Sánchez M. 2004. Exchange rate passthrough in emerging markets. Paper prepared for the BOFIT Workshopon Emerging Markets, Bank of Finland, Helsinki.
	In article
[5]	Calvo, G. 2001. Capital markets and the exchange rate: with special reference to the dollarization debate in Latin America, Journal of Money, Credit and Banking, 33: 312-334.
	In article		CrossRef
[6]	Calvo G, Reinhart C. 2001. Fixing for your life. In Collins, S. and Rodrik D. (Eds.), Brookings Trade Forum: 2000, Brookings InstitutionPress, Washington, DC, 1-57.
	In article		PubMed
[7]	Calvo G, ReinhartC. 2002. Fear of floating.Quarterly Journal of Economics107(2):379-408.
	In article		CrossRef
[8]	Calvo G, Vegh C. 1999. Inflation Stabilization in Chronic Inflation Countries, in: Taylor JB, Woodford M. (eds.) Handbook of Macroeconomics 1C, (North-Holland, Amsterdam) 1531-1614.
	In article
[9]	Campbell JY, Clarida RH. 1987. The dollar and real interest rates. Carnegie-Rochester Conference Series on Public Policy, 27: 103-140.
	In article		CrossRef
[10]	Chang R. 2008. Inflation targeting, reserve accumulation, and exchange rate management in Latin America. Banco de la República de Colombia Borradores de Economía, 487.
	In article
[11]	Chinn MD. 2006. The (partial) rehabilitation of interest rate parity in the floating rate era: longer horizons, alternative expectations, and emerging markets. Journal of International Money and Finance, 25: 7-21.
	In article		CrossRef
[12]	Chortareas GE, Driver RL. 2001. PPP and the real exchange rate–real interest rate differential puzzle revisited: evidence from non-stationary panel data. Bank of England Working Paper No. 138.
	In article
[13]	Coughlin C, Koedijk K. 1990. What do we know about the long-run real exchange rate? St. Louis Federal Reserve Bank Review72: 36-48.
	In article
[14]	Dickey DA, Fuller WA. 1979. Distribution of the estimators for autoregressive time series with a unit root. Journal of American Statistical Association74: 427-431.
	In article
[15]	Dickey DA, Fuller WA. 1981. Likelihood ratio statistics for autoregressive time series with a unit root, Econometrica, 49: 1057-1072.
	In article		CrossRef
[16]	Dornbusch R. 1976. Expectations and exchange rate dynamics. Journal of Political Economy. 84:1161-1176.
	In article		CrossRef
[17]	Edison HJ, Pauls BD. 1993. A re-assessment of the relationship between real exchange rates and real interest rates: 1974-1990. Journal of Monetary Economics. 31(2), 165-187.
	In article		CrossRef
[18]	Engle R, Granger C. 1987. Co-integration and error correction: representation, estimation, and testing. Econometrica. 55: 251-276.
	In article		CrossRef
[19]	Fischer S.2001. Exchange rate regimes: is the Bipolar-view correct? Journal of Economic Perspectives. 15(2): 3-24.
	In article		CrossRef
[20]	Frankel J. 1979. On the mark: a theory of floating exchange rates based on real interest differentials. American Economic Review69: 610-622.
	In article
[21]	Frenkel R, Rapetti M. 2008. Five years of competitive and stable real exchange rate in Argentina, 2002-2007. International Review of Applied Economics. 22(2): 215-226.
	In article		CrossRef
[22]	FrenkelR, Rapetti M. 2010. A concise history of exchange rate regimes in Latin America, Center for Economic and Policy Research, Washington, D.C.
	In article
[23]	Frenkel R, Ros J. 2006. Unemployment and the real exchange rate in Latin America. World Development. 34 (4):631-646.
	In article		CrossRef
[24]	Granger CWJ, Newbold P. 1974. Spurious regressions in econometrics. Journal of Econometrics. 2: 111-120.
	In article		CrossRef
[25]	Granger CWJ. 1983. Co-integrated variables and error-correcting models. UCSD Discussion Paper 83-13.
	In article
[26]	Granger CWJ. 1988. Some recent developments in a concept of causality. Journal of Econometrics. 39:7-21.
	In article		CrossRef
[27]	Jarque CM, Bera AK. 1980. Efficient tests for normality, homoscedasticity and serial independence of regression residuals. Economics Letters. 6 (3): 255-259.
	In article		CrossRef
[28]	Johansen S. 1996. Likelihood-based inference in cointegrated vector auto-regressive models, Oxford University Press, Oxford.
	In article
[29]	Jones JD, Joulfaian D. 1991. Federal government expenditures and revenues in the early years of the American republic: Evidence from 1792 to 1860. Journal of Macroeconomics. 13(1): 133-155.
	In article		CrossRef
[30]	Kim J. 2007. Real exchange rates and real interest differentials for sectoral data: a dynamic SUR approach. Economics Letters. 97(3): 247-252.
	In article		CrossRef
[31]	Lütkepohl H. 2007. Econometric analysis with vector autoregressive models. Economics Working Papers, European University Institute.
	In article
[32]	Meese RA, Rogoff K. 1988. Was it real? The exchange rate interest rate relation 1973-1984. Journal of Finance. 43: 933-948.
	In article		CrossRef
[33]	Mishkin F, Schmidt-Hebbel K. 2002. One decade of inflation targeting in the world: what do we know and what do we need to know? in Norman L. and Raimundo S. (eds) Inflation targeting: design, performance, challenges, 171-219.
	In article
[34]	Mollick AV, Torres RC, Carneiro FG. 2008. Does inflation targeting matter for output growth? Evidence from industrial and emerging economies. Policy Research Working Paper 4791. The World Bank.
	In article
[35]	Nakagawa H. 2002. Exchange rates and real interest differentials: implications of nonlinear adjustment in real exchange rates. Journal of Monetary Economics. 49(3): 629-649.
	In article		CrossRef
[36]	Ng S, Perron P. 2001. Lag length selection and the construction of unit root tests with good size and power. Econometrica. 69: 1519-1554.
	In article		CrossRef
[37]	Obstfeld M, Rogoff K. 1984. Exchange rate dynamics with sluggish prices under alternative price-adjustment rules. International Economic Review. 25 (1):159-174.
	In article		CrossRef
[38]	Perron P, Ng S. 1996. Useful modifications to some unit root tests with dependent errors and their local asymptotic properties. Review of Economic Studies. 63: 435-463.
	In article		CrossRef
[39]	Reinhart C, Rogoff K. 2004. The modern history of exchange rate arrangements: a reinterpretation.Quarterly Journal of Economics. 119: 1-48.
	In article		CrossRef
[40]	Rodriguez C. 1994. Interest rate in Latin America. Centro de EstudiosMacroeconomicos de Argentina (CEMA) 98.
	In article
[41]	Sánchez M. 2005. The link between interest rates and exchange rates: do contractionary depreciations make a difference? European Central Bank Working Paper Series 548.
	In article

Notes

¹The IMF suggests that apure floating regime is a better institutional setting (Fischer, 2001).

²Despite their public statements about their exchange rate regime choice, both Chile and Colombia have already been utilizing annual targets of inflation since 1990 and 1991 respectively. Peru had been usingamanaged floatingpolicyjointly with a monetary regime since the early 1990s.

³This means that, despite the recently proclaimed switch tofloating exchange rates, the evidence seems to suggest a reversion to some degree of exchange rate management, albeit one which seems to be less tight than before the crisis. In this regard, some analysts have found considerable discrepancies between the de jure exchange rate classifications andthede facto regimes (Reinhart and Rogoff, 2004).

⁴The RER under the Brazilian FIT has been 259% and 209% more volatile than under the managed floating in Argentina and the crawling bands in Chile,respectively (Frenkel and Rapetti,2010).

⁵Some Asian emerging countries have declared that their currencies have floated in post-Asian-crisis period, accompanied by a switch to inflation targeting. Such moves were taken by South Korea in 1998, Indonesia in 2000, Thailand in 2000, andthePhilippines in 2001. In Latin America, inflation targeting has been adoptedbyChile in 1990(together with an exchange ratefloat only since 1999), Mexico and Colombia in 1999, Brazil in 2000, and Peru in 2002. Among Eastern and Central European countries, EU new member states Czech Republic and Poland have also moved to comparable monetary and exchange rate policy frameworks (in 1998 and 1999, respectively), while South Africa and Israel count among other middle-income inflation targeters.

⁶Ca'Zorzi et al. (2004) find that not all EMEs display degrees of exchange rate pass-through above those seen in advanced economies. In particular, while pass-through tends to be high in countries in Eastern and Central Europe and Latin America, it is relatively low in many Asian economies.

⁷The early 1990s marked the turning point in the recent economic development of Latin America with the implementation of debt relief programs and the resumption of capital inflows to the region(Frenkel and Ros,2006).