Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Is low inflation really causing the decline in exchange rate pass-through? Reginaldo P. Nogueira Júnior Fundação João Pinheiro, Brazil Miguel A. León-Ledesma University of Kent, UK (Preliminary version) This draft: August, 2008. Abstract: Recent literature has argued that exchange rate pass-through (ERPT) has been declining in many countries after the adoption of inflation targeting regimes, following the change of the inflation environment. We use a state-space model to estimate the time-path of ERPT for some developed and emerging market economies, allowing for the ERPT coefficient to depend on lagged inflation. We then apply likelihood ratio tests in order to determine causality between inflation and ERPT. The results reinforce the view that the impact of exchange rate on prices has been declining, but offer only weak evidence that lower inflation is causing it. Keywords: Exchange Rate Pass-Through, State-space Models, Causality Tests. JEL Classification: E31, E42, E52, E58, F31, F41. Acknowledgments: We would like to thank John Driffill, Mathan Satchi, Eduardo Sanchez and Andrew Harvey for their insightful comments. Usual disclaimer applies. Address for correspondence: Reginaldo P. Nogueira Júnior. Fundação João Pinheiro. Alameda das Acácias, 70. Belo Horizonte, Minas Gerais. Brazil. 31.275-150. E-mail: [email protected]. 1 Is low inflation really causing the decline in exchange rate pass-through? 1. Introduction The pass-through from exchange rate changes into domestic inflation seems to have been declining in many countries in recent years, especially after the adoption of Inflation Targeting (IT) regimes (see, for example, Campa and Goldberg, 2005; Bailliu and Fujii, 2004; Gagnon and Ihrig, 2004; Bouakez and Rebei, 2007; Choudhri and Hakura, 2006). The basic explanation found in the literature for this decline is that it is a by-product of the low inflation environment of the 1990s. In this paper we take part of this debate, and present further econometric evidence on the decline of exchange rate pass-through (ERPT) for some developed and emerging market countries that adopted IT. We apply a methodology that allows us to test directly whether the low inflation environment brought about by the adoption of IT is causing the decrease in ERPT. Taylor (2000) was the first to provide an interpretation of the declining ERPT related to the lower inflation environment of the last decade. He argued that with staggered prices firms are more likely to pass-through cost changes, including those from the exchange rate, when inflation is high. In this sense, ERPT would be endogenous to a country’s inflation performance. Campa and Goldberg (2005), Gagnon and Ihrig (2004), Choudhri and Hakura (2006), among others, have analyzed this relationship, finding a positive correlation between ERPT and inflation indicators. Mishkin and Savastano (2001), Leiderman and Bar-Or (2000), and SchmidtHebbel and Tapias (2002) argue that ERPT depends on the monetary policy credibility, and therefore it is likely to decline overtime as the central bank’s commitment as an inflation-fighter becomes clearer. This hypothesis is particularly important for countries that adopted IT, as the reaction of inflation to a nominal shock should depend on how much inflation expectations are in line with the inflation target. Hence, the impact of a nominal shock on inflation, such as a currency depreciation, should be lower the higher the monetary policy credibility. 2 Choudhri and Hakura (2006) relate the credibility hypothesis for the declining ERPT with Taylor’s (2000) hypothesis, arguing that regimes that make a stronger effort in stabilizing short-run inflation so it keeps in track with its long-run target, are able to maintain low inflation rates. This view once again links the degree of ERPT in an economy to its long-run inflation performance. Most of the evidence of the declining ERPT in the literature is provided by splitting the sample estimations, as in Campa and Goldberg (2005), Gagnon and Ihrig (2004) and Choudhri and Hakura (2006), or by rolling regressions, as in Reyes (2003). Those techniques do not provide precise timing of the exchange rate parameter shift. It is also often the case with rolling regressions that the timing of the parameter shifts crucially depends on the size of the windows. Three exceptions to these approaches are Kim (1990) that applies the Kalman Filter to US data until the mid-1980s, Amstad and Fischer (2005) whose approach is an application of event-study procedures used in empirical finance to Swiss data from 1993:05 to 2005:05, and Sekine (2006) that estimates ERPT for 6 industrial economies using a time-varying parameter with stochastic volatility model. The latter finds that ERPT into consumer inflation has declined over time for all economies analyzed, and that there is high correlation between the estimated ERPT and the observed inflation rate. Empirical investigations on the causes of this decline encounter the difficulty that ERPT is an unobservable variable. In this sense, checking whether inflation is driving down ERPT is not an easy task. Moreover, the previous literature has tested Taylor’s (2000) hypothesis that the inflation environment is inducing a decrease in ERPT by analyzing cross-country correlations between inflation and ERPT. The problem with this procedure is that, as widely acknowledged, correlations do not imply causality, and a positive correlation between the estimated ERPT and the rate of inflation may imply not that lower inflation induces a lower ERPT, but that a lower ERPT provides a better environment for low inflation. Following these issues, in this paper we model ERPT by means of a state-space model, allowing for time variation in the coefficients. State-space models have been used in the literature to model unobserved variables, such as rational expectations, measurements errors, missing observations, and permanent income. In this paper we use a state-space specification in order to model the unobservable variable ERPT. The state- 3 space model presents a flexible structure that allows testing the casual relationships postulated in the literature, in terms of Granger causality tests. In this sense the main advantages of using a state-space model are that it allows for modelling unobserved variables and testing for temporal causality. We develop a backward-looking Phillips curve augmented with exchange rate and import prices. In our specification we allow ERPT to depend on lagged inflation. We then apply likelihood ratio tests in order to determine the direction of causality between inflation and ERPT. To our knowledge there is no other study that applies such tests for these variables. An additional advantage of our methodology is that we observe the evolution of ERPT without imposing any break in the series, letting the data itself tell us what is happening with ERPT and its relationship with inflation. Our results show that ERPT has been gradually declining for all the economies under consideration, as suggested by the previous literature. Regarding the direction of causality between inflation and ERPT, the results provide only weak evidence in favor of Taylor’s (2000) hypothesis, as for only a minority of the countries analyzed there seems to be bi-directional causality between inflation and ERPT. The remainder of the paper is structured as follows. Section 2 briefly discusses the recent literature on the declining ERPT issue. Section 3 presents the methodology applied. Section 4 shows the results. Section 5 concludes. 2. Literature review There is plenty econometric evidence that the exchange rate pass-through (ERPT) has been declining in many countries in recent years. As this decrease occurred in a period of declining inflation, many studies started to consider the possibility that the ERPT is endogenous to the inflation regime. In favor to this argument lies the fact that most of the economies for which there is evidence of a falling ERPT have adopted an Inflation Targeting (IT) regime at some point during the 1990s, and have successfully used this monetary framework to drive down inflation to low and stationary levels. In a popular paper, Taylor (2000) was the first to provide a theoretical model relating the decline of ERPT observed in empirical investigations to the low inflation environment of the 1990s. Taylor (2000) explained this relationship in terms of a model 4 of firm behavior based on staggered price setting and monopolistic competition. As firms set prices for several periods in advance their prices respond more to cost increases due to exchange rate movements if cost changes are perceived to be more persistent. As regimes with higher inflation tend to have more persistent costs, he argued that a high inflation environment would thus tend to increase ERPT. In other words, ERPT would be endogenous to a country’s inflation performance. Campa and Goldberg (2005) argue that an important implication of Taylor’s (2000) argument is that there is a virtuous circle wherein low inflation leads to reduced ERPT, less inflationary implications of monetary expansions, and continued low ERPT. This circle, however, is rather fragile, as it can be overcome by adverse supply shocks. Campa and Goldberg (2005) tested Taylor’s (2000) hypothesis in terms of correlation analysis between the estimated ERPT and a set of macro and micro variables, and concluded that although the argument has some statistical merit, it is not of first importance for the low and medium inflation countries of the OECD. For those countries they argued that the lower ERPT was brought about by a change in the composition of the import bundle1. Choudhri and Hakura (2006) also found strong evidence of a positive correlation between ERPT and the average inflation for 71 countries, including developed and emerging market economies, as well as some countries that adopted IT. They explored the influence of other macroeconomic variables as well, but found that average inflation dominates in explaining differences in observed ERPT. Similar results were found by Ca’Zorzi, Hahn and Sanchez (2007) for a sample of emerging market economies. They have observed that emerging markets with moderate rates of inflation tend to have a degree of ERPT compatible to developed economies. Bailliu and Fujii (2004) have presented evidence from a panel-data set of 11 industrialized countries that adopted IT. Their findings support the hypothesis that ERPT declines with a shift to a low-inflation environment brought about by a change in the monetary policy regime. More specifically, the results suggest that ERPT declined following the inflation stabilization that occurred after the adoption of IT in the early 1 Choudhri and Hakura (2006) argued that Campa and Goldberg (2005) have analysed ERPT to import prices and not to consumer prices. Hence, they analysed the price behaviour of foreign firms, which may not be strongly related to the home inflationary environment. In this sense, evidence on the ERPT to domestic prices would provide a more appropriate test of Taylor’s (2000) view. 5 1990s, but did not decline following a similar episode in the 1980s. Bailliu and Fujii (2004) argued that a potential explanation for this finding is that changes in the monetary policy regimes implemented in the 1990s were perceived as more credible than those carried out in the 1980s. Mishkin and Savastano (2001), Leiderman and Bar-Or (2000), Cespedes and Soto (2005) and Schmidt-Hebbel and Tapias (2002) have also argued that ERPT depends on the credibility of the monetary policy. The basic hypothesis is that ERPT is likely to decline over time as the country’s anti-inflationary commitment becomes clearer2. As discussed by Reyes (2003) this finding is particularly important for countries that adopted IT. In a credible regime, expectations are more likely to be in line with the authority’s inflation target and therefore will be less influenced by short term exchange rate movements. Levin, Natalucci and Piger (2004), Schimidt-Hebbel and Tapias (2002), Minella et al. (2003) and others have tested the inflation response to nominal shocks, such as those from the exchange rate, in IT regimes. These studies are generally based upon the analyses of the impulse responses of VAR models. The results show that shocks are weaker and less persistent under IT frameworks, suggesting a reduced role for priceindexation and reinforcing the importance of the inflation targets to anchoring inflation expectations and, as consequence of that, to lowering ERPT3. Choudhri and Hakura (2006) provide a link between the low inflation and the high credibility hypothesis for the falling ERPT. They derive a negative association between ERPT and the degree to which monetary policy offsets short-run price deviations from the long-run target. This association basically arises because the ERPT reflects the expected effect of monetary shocks on current and future shocks. A regime that reacts aggressively to price deviations lowers ERPT by weakening the expected future effect of shocks. As regimes that make a stronger effort in stabilizing short-run inflation are also able to maintain low inflation in the long-run, they argue that in empirical analysis the rate of inflation can be used as an indicator of the aggressiveness of monetary policy response to short-run price fluctuations. 2 Eichengreen (2002) argued that emerging market economies tend to have a high ERPT as their institutions lack credibility. His argument is that under imperfect credibility the market takes transitory exchange rate shocks as permanent, thus influencing the degree of ERPT. 3 Levin, Natalucci and Piger (2004) and Minella et al. (2003) actually suggest using the degree of ERPT as a proxy of monetary policy credibility. 6 A similar argument is used by Gagnon and Ihrig (2004) that explored the relationship between inflation, monetary policy credibility and ERPT for 20 industrial countries, half of them following an explicit IT regime. They have looked at the link between ERPT and parameters estimated from Taylor-type monetary policy rules, but failed to find a robust relationship. The hypothesis they wanted to test was that when a central bank acts aggressively to stabilize inflation it tightens policy to offset any inflationary effect from a rise in import prices. When the market realizes the central bank’s intentions, they are less likely to pass-through cost increases, including those coming from the exchange rate. Nevertheless, they were able to find similar results to those of Choudhri and Hakura (2006), showing a strong correlation between the estimated ERPT and the inflation environment, taken as the mean and the standard deviation of the rate of inflation. 3. Methodology A wide variety of time-series models can be written and estimated as special cases of a state-space specification. Extensive examples of applications of state-space models can be found in Harvey (1989). We use a state-space specification to model a time-varying ERPT, as well as the variables that drive its dynamics. There are three main benefits of representing a dynamic system in state-space form. First, the state-space model allows unobserved variables (the state variables) to be estimated with the observable model. Second, state-space models can be analysed using the powerful Kalman Filter recursive algorithm, which is commonly used to estimate time-varying coefficient models4. And third, a very important feature of state equations is it flexibility, as they may contain exogenous variables and unknown coefficients, and may also be nonlinear in these elements. A state-space model usually consists of two sets of equations, the measurement equations and the state equations. The Kalman filtering approach provides optimal estimates for state variables based on the information from the two sources, the measurement and the state equations. 4 The Kalman Filter can be described as a tool that enables the estimation of the state variables and the parameters in a time-varying parameter model using maximum likelihood. 7 We estimate ERPT by means of a state-space model composed of the following system of equations: n ∆ Pt = α 1t ∑ ∆ Pt − i + α 2 t ∆ P * t −1 + α 3 t ∆ y t + α 4 t ∆ e t −1 + ε 1t (1) α1t = α1t −1 + ε 2t (2) α 2t = α 2t −1 + ε 3t (3) i =1 α 3 t = α 3 t −1 + ε 4 t (4) n α 4t = α 4 t −1 + β 1 ∑ ∆ Pt − i + ε 5 t (5) i =1 Where ∆P is the inflation rate, ∆P* is the change in the price of imports, ∆e is the exchange rate change, and ∆y is the output growth5. The terms ε t , 1…5, are independent normally distributed errors, with zero mean and constant variance. Equation (1) is the measurement equation, and equations (2) to (5) are the state equations. When estimating the model we included more than one lag of the inflation rate for some countries due to autocorrelation6. The α ’s, 1…4, are estimated state coefficients. The system is estimated using the Kalman Filter technique. Note that we imposed a unit-root in the state equations. As discussed by Sekine (2006) and Tombini and Alves (2006) this is a very standard procedure in the literature on state-space modelling for allowing possible level breaks or trend patterns to be captured7. Equation (1) captures the traditional arguments of the literature on ERPT, and is similar to that estimated in several papers, as Campa and Goldberg (2005), Choudhri and Hakura (2006) and others. The equation represents a backward-looking Phillips 5 The problem of smoothing output to obtain output gap, in special through the use of the traditional HP Filter is discussed in several papers, as King and Rebelo (1993), Harvey and Jaeger (1993), and Cogley and Nason (1995). It is, for example, widely recognised that the smoothing parameter in the HP Filter may be different for different countries. Besides, ad-hoc de-trending processes may eliminate valuable information from the data. Nevertheless, we have also estimated the model using HP-filtered and Bandpass filtered output gaps, and the results were similar. Following this, we have opted to report only the results using output growth. 6 The additional lags were included in equations (1) and (5). 7 When estimating the model we have also tested the inclusion of a constant in the measurement equation, but it proved to be statistically insignificant, and was excluded from the model. We also estimated the model with the state variables following an AR(1) process, but the state series found were too noisy to be of any economic sense. 8 curve, controlling for the exchange rate and the price of imports. As demonstrated by Campa and Goldberg (2005), empirical specifications that seek to isolate ERPT should introduce controls for the foreign costs, as without such controls the measured relationship is a statistical correlation without specific economic interpretation in terms of ERPT8. In the specification of the exchange rate coefficient we included a coefficient of lagged inflation rate. In this sense ERPT coefficient is a function not only of its past value, but also of the inflation environment. This specification allows us to analyse the explanatory power of lagged inflation over the ERPT, and thus if it can be considered to be a temporally causal variable for the decline in ERPT. To obtain time-series for the state variables we applied the Kalman Smoothing procedure9. The smoothing uses all the information in the sample to provide smoothed estimates of the states and of the variances. This procedure differs from the Kalman Filtering in the construction of the state series, as this technique uses only the information available up to the beginning of the estimation period. 3.1 Granger causality tests The previous literature has analysed correlations between the ERPT and the rate of inflation. The basic procedure has been the estimation of the ERPT for a large number of countries, and then using the estimated ERPT as dependent variable in a crosscountry regression, controlling for a set of macroeconomic variables (see, for example, Baqueiro, Diaz de Leon and Torres, 2003; Choudhri and Hakura, 2006; Gagnon and Ihrig, 2004). This approach provides correlations and cannot be informative about the causality between the variables. In addition, cross-country regressions may problems of heterogeneity among the different countries included in the model, making its results not as robust as those obtained in a time-series context. Providing evidence of causality is of utmost importance with respect to ERPT and inflation issue, as it may be that low ERPT is causing low inflation, and not the contrary as suggested by the literature. 8 Campa and Goldberg (2005) show that the price of imports in local currency is basically a function of the price of imports in foreign currency, the exchange rate, and the importer’s mark-up. 9 Suppose that we observe the sequence of data up to time period T. The process of using all this information to form expectations at any time period up to T is known as smoothing. 9 Following the system of equations (1) to (5) presented before, we use Likelihood ratio tests for checking the null hypothesis H0: α4 = 0 and H0: β1 = 0. The first null hypothesis states that the exchange rate coefficient is equal to zero, thus its rejection means that ERPT Granger causes inflation. The second null hypothesis checks if the previous causality also runs in the other way, i.e. if inflation causes ERPT, hence representing a formal test for Taylor’s (2000) proposition. The steps for testing the causalities are as follows. First, we estimate the unrestricted model and obtain the log-likelihood. Then, we estimate a restricted model, with the coefficient under consideration equal to zero, and obtain the new loglikelihood. To the test the hypothesis that the coefficient is indeed equal to zero, we apply the following test: λ = 2 x (ULL – RLL) (6) Where ULL is the unrestricted log-likelihood and RLL is the restricted loglikelihood. When the sample size is large, it can be shown that the likelihood ratio test statistic λ follows the chi-square distribution with degrees-of-freedom equal to the number of restrictions imposed by the null hypothesis. If the statistic λ exceeds the critical value at the chosen level of significance, the null hypothesis is rejected, in which case the coefficient is not zero, and hence the variable belongs to the regression model. This implies that the variable improves the model’s prediction. Quarterly data was collected for 16 countries that adopted IT that may be divided in two groups: the first one comprises developed economies (Australia, New Zealand, Norway, Switzerland, Sweden, Iceland and United Kingdom), and the second one is composed of emerging markets (Brazil, Czech Republic, Mexico, Israel, South Korea, Thailand, Chile, Hungary and Poland). The period of estimation for each country and the dates of adoption of IT can be seen in Table 1. Data was obtained from the IMF International Financial Statistics database. The inflation rate is the rate of growth of the consumer price index. Exchange rate data is the change of the national currency per unit of dollar. A positive variation means depreciation of the national currency, and a negative one means appreciation. Output 10 data is the quarterly real GDP growth10. Import prices are defined as the change in the index of dollar price of imports for each country. This data was not available for Mexico and the Czech Republic, so as a proxy of the foreign costs faced by those countries we used the change in the index of international commodities. The data on real GDP and inflation were seasonally adjusted. Unit root tests show that the variables included in the model can be treated as stationary11. 4. Results As discussed in the methodology we estimated a time-varying ERPT in terms of a statespace model, using the Kalman Filter to obtain the time-path of the state variables. Table 2 shows some basic results of the estimated model. Figure 1 plots the estimated ERPT into consumer prices, together with the rate of inflation. As discussed before the time-series of the state coefficients were constructed by smoothing the states estimates. Our time-varying parameter model suggests that ERPT has indeed declined over time, but particularly after the adoption of the IT framework. The figures also suggest that this decline took place somewhat more gradually than when compared to estimates obtained from rolling regressions, such as those provided by Reyes (2003) and Sekine (2006), although it holds resemblance to those12. An interesting point is that it is possible to see the impact of exchange rate and confidence crises in influencing the degree of ERPT as, for example, Brazil (2002), Mexico (1995) and Czech Republic (1997). Furthermore, it is clear that there is a strong positive correlation between inflation and ERPT, which is in line with findings of the previous literature. However, as discussed before, instead of just observing correlations, as in the previous literature, our approach to estimating ERPT allows us to analyse casual temporal patterns. The results are presented in the Table 3. 10 In the case of Brazil the industrial production index was used. Due to data unavailability, Iceland’s model was estimated without the output control variable. 11 We applied ADF, KPSS and Phillips-Perron unit root tests. The results are not reported here for reasons of space, but are available upon request. 12 Sekine (2006) observes that rolling regressions tend to yield abrupt changes depending on whether or not a specific sample is in the window. Because of this dependence on the size of windows, rolling regressions do not provide precise timings of the changes in parameters. In addition to this, smoothed series tend to produce more gradual changes than filtered ones. However, as argued by Sims (2001), smoothed series provide more precise estimates of the actual time variation than filtered ones. 11 A salient result of the likelihood ratio tests is that for most countries in our sample we were able to reject the null that the ERPT coefficient is equal to zero, and in most cases at the 5% confidence level. This result makes it clear that inflation depends on ERPT, or in other words, that ERPT Granger causes inflation. Taylor’s (2000) hypothesis that a low inflation environment leads to a lower ERPT means that this causality should ran also in the opposite direction. However, this result applies only to a few countries. In other words, we were able to determine causality from ERPT to inflation for all the economies, but bi-directional causality between the two variables for only a small part of the sample. In summary our results reinforce the view of the previous literature that ERPT has been declining over time. We observed that this is true for all the economies in our sample. However, with respect to the causes of the decrease in ERPT our results provide only weak evidence in favour of the argument of Taylor (2000), Choudhri and Hakura (2006), Gagnon and Ihrig (2004), Baqueiro, Diaz de Leon and Torres (2003) and others, that advocates the importance of the lower inflation environment to reducing ERPT. This hypothesis seems to be more appropriate for developed economies than for emerging markets. In this sense, our results suggest that the correlations found in the previous literature between inflation and ERPT did not mean exactly that a low inflation regime is causing a low ERPT, but that a low ERPT is causing a lower inflation environment. This finding highlights the importance of further econometric investigation on the causes of the decline in ERPT observed in the countries that adopted IT in the 1990s. 5. Conclusion In this paper we have presented evidence on the role of low inflation on the decline of the exchange rate pass-through (ERPT) for some developed and emerging market economies that adopted Inflation Targeting (IT). Instead of just analysing correlations between ERPT and inflation, we have provided evidence on temporal causality between these variables. We developed a state-space model of a Phillips curve, allowing for time variation of the exchange rate parameter, and observed a gradual decline of the ERPT 12 after the adoption of IT. We modelled the ERPT as a function of its past value and lagged inflation. We then applied likelihood ratio tests, which provided evidence of causality between inflation and ERPT. For almost all the economies under consideration we found that ERPT is an important variable in explaining the rate of inflation, but only for few countries the causality also run the other way. In this sense further investigation on this issue should be necessary in order to shed some light on the macroeconomic determinants of ERPT. References Amitrano, A., de Grauwe, P. and Tullio, G. (1997). “Why has inflation remained so low after the large exchange rate depreciations of 1992?” Journal of Common Market Studies. 35. pp. 329-346. Amstad, M. and Fischer, A. (2005). “Time-varying pass-through from import prices to consumer prices: evidence from an event study with real time data.” Federal Reserve Bank of New York. Staff Report 228. Bailliu, J. and Fujii, E. (2004). “Exchange rate pass-through and the inflation environment in industrialized countries: an empirical investigation”. Bank of Canada. Working Paper 21. Ball, C. and Reyes, J. (2004). “Inflation Targeting or Fear of Floating in disguise: the case of Mexico”. International Journal of Finance and Economics. 9. pp 49-69. Baqueiro, A., Diaz de Leon, A. and Torres, A. (2003). “Fear of Floating or Fear of Inflation? The role of the exchange rate pass-through”. BIS. Working Paper 19. Bouakez, H. and Rebei, N. (2007). “Has exchange rate pass-through really declined in Canada?” Journal of International Economics, forthcoming. Ca’Zorzi, M., Hahn, E. and Sánchez, M. (2007). “Exchange rate pass-through in Emerging Markets”. European Central Bank. Working Paper 739. Campa, J. and Goldberg, L. (2005). “Exchange rate pass-through into imports prices.” The review of Economics and Statistics. 87. pp. 679-690. Cespedes, L. and Soto, C. (2005) “Credibility and inflation targeting in an emerging market: the case of Chile”. Central Bank of Chile. Working Paper 312. 13 Choudhri, E. and Hakura, D. (2006). “Exchange rate pass-through to domestic prices: does the inflationary environment matter?” Journal of International Money and Finance. 25. pp. 614-639. Eichengreen, Barry (2002). “Can emerging markets float? Should they inflation target?” Central Bank of Brazil. Working Paper 36. Ganapolsky, E. and Vilan, D. (2005). “Buy foreign when you can: the cheap dollar and exchange rate pass-through”. Economic Review, Jul. (Atlanta, Ga.) Gagnon, J. and Ihrig, J. (2004). “Monetary policy and exchange rate pass-through”. International Journal of Finance and Economics. 9. pp. 315-338. Goldfajn, I. and Werlang, S. (2000). “The pass-through from depreciation to inflation: a panel study”. Central Bank of Brazil. Working Paper 05. Kim, Y. (1990). “Exchange rate and import prices in the US: a varying-parameter estimation of exchange rate pass-through”. Journal of Business and Economic Statistics. 8. pp. 305-315. Levin, A., Natalucci, F. and Piger, J. (2004). “The macroeconomic effects of Inflation Targeting”. Federal Reserve Bank of St. Louis Review. Jul. pp. 51-80. Leiderman, L. and Bar-Or, H. (2000) “Monetary Policy Rules and Transmission Mechanisms under Inflation Targeting in Israel.” Bank of Israel. Working Paper 71. Minella, A. et al. (2003). “Inflation targeting in Brazil: constructing credibility under exchange rate volatility”. Journal of International Money and Finance. 22. pp. 1015-1040. Mishkin, F. and Savastano, M. (2001). “Monetary policy strategies for Latin America.” Journal of Development Economics. 66. pp. 415-444. Park, Y. (2001). “Fear of Floating: Korea’s exchange rate policy after the crisis”. Journal of the Japanese and International Economics. 15. pp. 225-251. Reyes, J. (2003). “Exchange Rate Pass-through Effect, Inflation Targeting and Fear of Floating in Emerging Economies”. University of Arkansas, Mimeo. Schimidt-Hebbel, K. and Tapias, M. (2002). “Inflation Targeting in Chile”. The North American Journal of Economics and Finance. 13. pp. 125-146. Sekine, T. (2006). “Time-varying exchange rate pass-through: experiences of some industrial countries”. BIS. Working Paper 202. 14 Sims, C. (2001). “Comments on Sargent and Cogley’s Evolving Post World War II US Inflation Dynamics”. NBER Macroeconomics Annual. 16. pp. 373-379. Taylor, John (2000). “Low inflation, pass-through and the pricing power of firms”. European Economic Review. 44. pp. 1389-1408. Tombini, A. and Alves, S. (2006). “The recent Brazilian disinflation process and costs”. Central Bank of Brazil. Working Paper 109. 15 Table 1: Estimation period Countries Period Australia 1981:1-2006:4 (1993:2) Brazil 1995:1-2006:4 (1999:2) Chile 1986:1-2006:4 (1991:1) Czech Republic 1994:4-2006:4 (1998:1) Hungary 1996:1-2006:4 (2001:2) Iceland 1983:3-2006:4 (2001:1) Israel 1990:1-2006:4 (1992:1) Mexico 1980:3-2006:4 (1999:1) New Zealand 1988:1-2006:4 (1990:1) Norway 1981:1-2006:4 (2001:1) Poland 1995:3-2006:4 (1998:4) South Korea 1983:1-2006:4 (1998:2) Switzerland 1980:4-2006:4 (2000:1) Sweden 1981:1-2006:4 (1993:1) Thailand 1993:3-2006:4 (2000:2) United Kingdom 1980:3-2006:4 (1992:4) Notes: the dates in parenthesis are the dates of adoption of Inflation Targeting. Table 2: State-space model estimation results Australia Brazil Chile Czech Republic Hungary Iceland Mexico New Zealand Norway Poland Israel South Korea Switzerland Sweden Thailand United Kingdom Log Likelihood -154.1082 -91.0541 -161.8151 -99.8528 -67.7258 -170.9993 -233.8218 -87.6329 -163.8045 -57.3341 -136.0843 -135.6297 -68.9071 -144.9307 -88.0339 -100.4403 SIC 3.4549 4.2778 4.4320 4.8864 4.1122 3.9283 4.7640 2.7050 3.6413 3.1403 4.4368 3.3827 1.6228 3.2784 3.8464 2.2031 Notes: Figures in the first column is the Log Likelihood of the Maximum Likelihood estimation of the model. The second column is the Schwarz Information Criterion (SIC) of the estimated model. 16 Table 3: Granger Causality Tests Australia Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Brazil Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Chile Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Czech Republic Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Hungary Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Iceland Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Mexico Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation New Zealand Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Norway Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Poland Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Israel Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation df 3 4 λ-stat. 3.001 17.932** df 1 2 λ-stat. 0.044 5.220* df 3 4 λ-stat. 9.529** 11.863** df 2 3 λ-stat. 0.384 12.279** df 3 4 λ-stat. 4.528 16.663** df 1 2 λ-stat. 9.507** 2.776 df 1 2 λ-stat. 0.169 42.039** df 1 2 λ-stat. 0.306 7.809** df 3 4 λ-stat. 14.418** 34.701** df 1 2 λ-stat. 5.036** 11.542** df 1 2 λ-stat. 0.739 4.428* 17 South Korea Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Switzerland Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Sweden Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation Thailand Inflation does not Granger Cause Pass-through Pass-through does not Granger Cause Inflation United Kingdom df 3 4 λ-stat. 4.075 29.726** df 1 2 λ-stat. 1.531 9.076** df 3 4 λ-stat. 1.923 12.121** df 1 2 λ-stat. 3.039* 19.500** df λ-stat. Inflation does not Granger Cause Pass-through 1 2.516 Pass-through does not Granger Cause Inflation 2 20.313** Notes: ** indicates significance at the 5% confidence level and * indicates significance at the 10% confidence level. The λ-statistic follows the Chi-square distribution with degrees of freedom equal to the number of restrictions imposed by each null hypothesis. 18 Figure 1: Inflation and ERPT Australia Brazil 7 Inflation Inflation ERPT ERPT 0.06 3.5 0.12 6 0.05 3.0 0.04 2.5 0.03 2.0 0.02 1.5 0.01 0.00 1.0 5 0.10 4 0.08 3 0.06 2 -0.01 0.5 1 -0.02 0.0 -0.03 0.04 0 -0.04 -0.5 0.02 -1 1980 1985 1990 1995 2000 1995 2005 2000 Chile 2005 Czech Republic 5 8 Inflation Inflation ERPT 0.25 ERPT 0.5 7 0.4 6 4 0.20 0.3 5 0.2 4 3 0.15 0.1 3 0.0 2 0.10 2 -0.1 1 1 0.05 0 0.00 -0.2 0 -0.3 -1 -0.4 1985 1990 1995 2000 1995 2005 2000 Hungary Inflation 2005 Iceland ERPT Inflation ERPT 0.5 5 0.15 20 0.4 4 0.10 15 0.3 3 10 0.05 0.2 2 5 0.00 0.1 1 0 -0.05 0.0 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 1985 1990 Mexico 35 Inflation 1995 2000 2005 New Zealand 0.7 ERPT 30 0.6 3.5 Inflation ERPT 0.150 3.0 0.125 2.5 25 0.5 0.100 2.0 20 0.4 1.5 15 0.075 0.3 1.0 0.050 10 0.2 0.5 5 0.025 0.1 0.0 0.0 -0.5 0 -5 1980 0.000 1985 1990 1995 2000 2005 1990 1995 2000 2005 19 Norway Poland 4 0.3 Inflation 5 ERPT Inflation ERPT 0.125 0.2 3 4 0.100 0.1 2 0.0 3 0.075 1 -0.1 0 2 0.050 1 0.025 -0.2 -1 -0.3 -2 -0.4 1980 1985 1990 1995 2000 0.000 0 2005 1996 1997 1998 Israel 1999 2000 2001 2002 2003 2004 2005 2006 2007 South Korea 8 Inflation ERPT Inflation 0.25 ERPT 0.23 7 4 0.22 0.20 6 0.21 5 3 0.20 4 0.19 3 0.18 0.17 2 0.15 2 0.10 1 0.05 0.16 1 0.15 0 0.00 0 0.14 -1 -1 0.13 1990 1995 2000 -0.05 0.12 2005 1985 1990 1995 Switzerland 2000 2005 Sweden 2.5 4 Inflation ERPT 0.075 Inflation ERPT 0.075 3 2.0 0.050 1.5 0.050 2 1 1.0 0.025 0.025 0 0.5 0.000 0.000 0.0 1980 1985 1990 1995 2000 2005 -1 1980 1985 Thailand Inflation 1990 1995 2000 2005 United Kingdom ERPT 4.0 Inflation 0.03 ERPT 0.2 2.5 3.5 0.02 2.0 0.1 1.5 3.0 0.01 2.5 1.0 0.0 2.0 0.00 1.5 0.5 -0.1 0.0 1.0 -0.01 0.5 -0.5 -0.2 -0.02 0.0 -1.0 -0.5 1995 2000 2005 1980 1985 1990 1995 2000 2005 Notes: The graphs show the evolution of inflation (dashed line, left scale) and exchange rate pass-through (black line, right scale). 20