باز بودن، رژیم نرخ ارز و منحنی فیلیپس

Recent research suggests that the Phillips curve slope, measured using sacrifice ratios from the period 1961–88, is positively related to trade openness, contradicting the Romer [1993. Openness and inflation: theory and evidence. Quarterly Journal of Economics 108, 869–903.] hypothesis that disinflations are less costly in open economies. In this paper I consider sacrifice ratios and output–inflation trade-offs from 1981–98 and allow their dependence on openness to vary with the exchange rate regime. Sacrifice ratios are weakly negatively related to openness, but the strength of the relationship does not increase with exchange rate flexibility. Output–inflation trade-offs are negatively related to openness, and the strength of the relationship increases with exchange rate flexibility.

In this paper I provide new evidence on the relationship between openness to trade and the slope of the short-run Phillips curve. Models in which monetary policy expansions depreciate the real exchange rate predict that the increase in output associated with a unit increase in inflation is negatively related to openness, see for example Romer (1993). This implies that the sacrifice ratio and the output–inflation trade-off, which are both proxies for the Phillips curve slope, should be negatively related to openness. However, cross-country studies indicate very little support for this prediction, see Ball et al., 1988 and Ball, 1994 and Temple (2002). In a recent contribution Daniels et al. (2005) show that after controlling for central bank independence (CBI) and its interaction with openness, the effect of openness on the sacrifice ratio is positive, contradicting the predictions in Romer (1993). Daniels et al. argue that their findings are consistent with the models due to Razin and Yuen (2002) and Daniels and VanHoose (2006), which incorporate imperfectly competitive labour and product markets in an open economy framework. The main purpose of this paper is to consider two extensions of previous research. The first is to examine sacrifice ratios and output–inflation trade-offs calculated using data from the 1980s and 1990s. Most existing research uses the sacrifice ratios calculated by Ball (1994) and the output–inflation trade-offs reported by Ball et al. (1988). The sample periods used to calculate these indices vary slightly across countries, the earliest data being from the 1940s and the latest from the mid-1980s. New versions of these statistics for the period 1981–98 are calculated. During that time monetary policy arguably played a more important role in disinflation episodes than in earlier decades (Nelson, 2005) and, as I argue in Section 3, this is important because the theory proposed by Romer (1993) requires that movements along the Phillips curve result from monetary policy, rather than fiscal policy or some other determinant of aggregate demand. The second innovation is to allow the relationship between openness and the Phillips curve to depend on the exchange rate regime. The Romer hypothesis requires that unanticipated increases in the money supply lead to depreciation of the nominal and real exchange rates. This pushes up the relative price of imports, raises inflation and restricts the increase in output associated with monetary expansion. These effects will be stronger in more open economies and the responsiveness of output to inflation is therefore expected to decrease with openness. An idea emphasized in the paper is that the correlation between openness and the Phillips curve predicted by Romer (1993) is likely to be stronger under flexible exchange rate regimes.1 In order to allow for this possibility I estimate models that include an interaction between openness and the exchange rate regime. The results show that during the post-1980 period the sacrifice ratio is negatively related to openness, supporting the Romer hypothesis. This finding contrasts with the robust positive effect of openness on the sacrifice ratio estimated by Daniels et al. (2005) using data from earlier decades. As such it appears that the mechanisms determining the sacrifice ratio have changed during the post-war period. The relationship between openness and the sacrifice ratio does not vary systematically with the exchange rate regime in either the earlier or later sample periods, casting doubt on the argument that exchange rate adjustment contributes to the relationship between openness and the Phillips curve. However, regressions based on post-1980 estimates of the output–inflation trade-off yield negative coefficients on both openness and its interaction with the exchange rate regime indicator, and as such are consistent with the hypothesis that openness exerts a stronger effect on the Phillips curve under flexible exchange rate regimes. The rest of the paper expands on these points and is structured as follows. Section 2 summarizes the model due to Romer (1993), highlighting the role played by exchange rate adjustment, and discusses the recent contributions of Razin and Yuen (2002) and Daniels and VanHoose (2006). Section 3 argues for an updating of the sample period used to measure the slope of the Phillips curve and describes the data used in the empirical analysis. Section 4 presents results based on sacrifice ratios and Section 5 presents results based on output–inflation trade-offs. Finally, Section 6 summarizes the paper.

In column (1) of Table 2 the Ball et al. estimates of the output–inflation trade-off are regressed on OPEN, OPEN × EX, average inflation for 1948–86 (INF) and its square (SQINF).18 The last two variables are the key controls emphasized by Ball et al. (1988). OPEN and OPEN × EX are positively signed and insignificant, confirming the results of Ball et al. (1988) and Temple (2002).19 Columns (2) and (3) update the observation period to 1981–98 for all variables. In column (2) the openness coefficient is negatively signed but far from significant. The picture changes in column (3), however. After controlling for OPEN × EX the OPEN coefficient is negative and significant at the 15% level, and the OPEN × EX coefficient is negative and significant at the 5% level.20 The results imply that the marginal effect of OPEN on the output–inflation trade-off for a country one standard deviation above the sample average of EX is −0.86, which is significant at the 10% level using the joint standard error for the fitted coefficients on OPEN and OPEN × EX (the joint standard error is 0.497). In contrast, for a country one standard deviation below the sample average exchange rate regime score, the marginal effect of OPEN is insignificant at conventional levels.The negative relationship between openness and the output–inflation trade-off is clearly conditional on the interaction term featuring the exchange rate regime. In contrast, the negative relationship between openness and the sacrifice ratio did not depend on the interaction term. This finding is related to the set of controls used in each case. In the sample used to fit the output–inflation trade-off regressions the correlation between OPEN and OPEN × EX is −68%. This implies that the omission of OPEN × EX from the model biases the OPEN coefficient towards zero, as in column (2).21 In the sample used to estimate models for the sacrifice ratio the correlation between OPEN and OPEN × EX is −62%, but the additional controls included in those models account for the effects of OPEN × EX and therefore the negative coefficient on OPEN does not depend on the interaction term. If the additional controls in columns (2) and (3) in Table 1 are excluded, both OPEN and OPEN × EX are negatively signed and are significant at the 1% and 5% levels, respectively. Hence, controlling for factors such as the length and magnitude of a disinflation appears to be part of the explanation for the fact that the negative effect of openness on the sacrifice ratio is not conditional on an interaction between openness and the exchange rate regime. One objection to the results presented in columns (1)–(3) is that the observation periods often span changes in the Reinhart–Rogoff index and as such do not refer to individual exchange rate regimes, e.g. a country could be classified as 1 (completely fixed exchange rate) for 1981–89 and 2 (limited exchange rate flexibility) for 1990–98. The problem was less serious in the case of the sacrifice ratio regressions because the exchange rate regime was calculated over the course of a disinflation, typically lasting only a few years and less likely to span a change in the Reinhart–Rogoff index. In order to address this issue the longest interval from the post-1980 period for which just one of the five categories in the Reinhart–Rogoff scheme applied is determined. The output–inflation trade-off and each of the regressors are calculated for the period for which an unchanged exchange rate regime applies and the cross-country regressions re-estimated using the new data.22 The results are provided in column (4).23 The coefficients on OPEN and OPEN × EX are larger in absolute size than those in column (3) and are significant at the 5% level, providing somewhat stronger evidence for a negative effect of openness on the output–inflation trade-off than that documented in column (3). In columns (5) and (6) I examine the robustness of the results using alternative exchange rate regime classifications, which is important in this context given the crucial role played by the OPEN × EX term. In column (5) each of the variables are measured for the longest interval from the 1981–98 period for which the de facto exchange rate regime classification due to Levy-Yeyati and Sturzenegger (2003) takes a single value.24 The results are weaker than in column (4). Nevertheless, the coefficient for OPEN is negative and significant at the 15% level, whilst that for OPEN × EX is negative and significant at the 10% level. Hence, the main implications of the results based on the Reinhart and Rogoff (2004) exchange rate regime classification appear to be robust to using the alternative de facto classification. This is not the case when the de jure classification of Ghosh et al. (2003) is used in column (6). In this case both OPEN and OPEN × EX are positively signed and insignificant. This result is unsurprising given that de jure and de facto exchange rate regime classifications are known to convey very different information, see Reinhart and Rogoff (2004). In this particular application the correlation between the Ghosh et al. measure and that due to Reinhart and Rogoff is just 25%. To the extent that it is the actual degree of exchange rate flexibility (as opposed to the amount of flexibility that a country claims to maintain) that matters in determining the relationship between openness and the output–inflation trade-off the de facto classification is more appropriate in the present context.