تغییرات نظام و ثبات منحنی فیلیپس رو به عقب، در اقتصاد باز

We assess the stability of open-economy backward-looking Phillips curves estimated over two different exchange rate regimes. We calibrate a new-Keynesian monetary policy model and employ it for producing artificial data. A monetary policy break replicating the move from a Target-Zone regime to a Free-Floating regime implemented in Sweden in 1992 is modeled. We employ two different, plausibly calibrated Taylor rules to describe the Swedish monetary policy conduct, and fit a reduced-form Phillips curve to the artificial data. While not rejecting the statistical relevance of the Lucas critique, we find that its economic importance does not seem to be overwhelming.

Since the publication of the seminal paper by Lucas (1976), many researchers have explicitly embedded forward-looking expectations in their policy models. One of the fields that has been intensely affected by this push towards micro-foundation is the monetary one (e.g. Woodford, 2003). Interestingly, a different strand of this literature (e.g. Rudebusch and Svensson, 1999, Rudebusch and Svensson, 2002, Ball, 1999, Ball, 2000, Onatski and Stock, 2002, Laubach and Williams, 2003 and Fagan et al., 2005) has relied on ad-hoc backward-looking frameworks. In fact, backward-looking models tend to offer quite a good fit of the data. Moreover, their impulse responses closely resemble those stemming from structural VARs, an issue that pure forward-looking models have some troubles in dealing with (Estrella and Fuhrer, 2002). Evidently, at a theoretical level backward-looking models are affected by the Lucas (1976) critique. The argument goes as follows. If agents are forward-looking they will adjust their expectations once a policy change is credibly announced. As a consequence, reduced-form coefficients will be unstable under a change in the policy regime. Then, a policy analysis performed with reduced-form coefficients may be severely mis-leading. This is true in principle: but how important is this change in practice? While some researchers have undertaken empirical efforts to answer this question in a closed-economy set-up (e.g. Lindé, 2001, Estrella and Fuhrer, 2003 and Rudebusch, 2005), to the best of our knowledge the only contribution dealing with this issue in an open-economy framework dates back to Taylor (1989). This is somewhat surprising, given the increasing openness in terms of trade and flows of resources conveyed in the international financial markets observed in several countries in the last decades (Lane, 2001).1 The aim of this paper is that of ‘updating’ Taylor's (1989) contribution. Taylor (1989) employs an estimated macro-model for simulating the shift from a ‘fixed’ to a ‘flexible’ nominal exchange rate in some industrialized countries. Once done so, he fits some reduced-form schedules (mainly demand and supply curves) to such simulated data, and compares the estimated parameters under the first regime to those estimated under the second one. Taylor (1989) observes that the differences in magnitude between those parameters are not really large, and concludes that the Lucas critique does not find a large support in the data. Notice that, in performing his analysis, Taylor (1989) does not use any statistical tool for assessing the stability of the estimated coefficients. We refine the contribution by Taylor (1989) along two main dimensions. First, we employ a modern new-Keynesian DSGE open-economy monetary policy model in the spirit of the one proposed by Svensson (2000). In this model, the monetary policy makers follow a Taylor-type rule and aim at minimizing the oscillations of inflation, real GDP, and nominal exchange rate around their targets. As a second difference with respect to Taylor (1989), we assess both the statistical and the economic relevance of the Lucas critique. In this sense, we line up with Rudebusch (2005) in acknowledging the importance of employing a formal test for assessing the instability of the estimated schedules from a statistical perspective. Importantly, as underlined by Rudebusch (2005) this is not the only dimension one may want to consider in this kind of exercise. In fact, the macroeconometrician is also called to evaluate the pros and cons of employing a reduced-form model estimated under a policy break from an economic perspective. To do so, we compare the mean values and the distributions of the estimated coefficients in the two different subsamples we focus on. For our exercise to be interesting, we need to calibrate our model in a plausible fashion. Interestingly, Sweden experienced a dramatic change in regime in 1992, when it moved from an exchange rate target-zone (TZ hereafter) to a flexible exchange rate regime (FF henceforth). Cúrdia and Finocchiaro (2005) split the Swedish sample 1980Q1–2003Q3 into two subsamples, i.e. 1980Q1–1992Q4 (TZ) and 1993Q1–2003Q3 (FF), estimate regime-specific Taylor rules, and find that there are remarkable differences between the two sub-sets of estimated values.2 We calibrate our Taylor rules with their point estimates in order to simulate a ‘plausible’ regime shift, i.e. a regime shift historically occurred in an open-economy such as Sweden. We calibrate the rest of our new-Keynesian model by borrowing parameter values from the recent empirical literature on monetary policy models for the Swedish economy, i.e. Leitemo and Söderström, 2005 and Lindé et al., 2004, and Cúrdia and Finocchiaro (2005). In our exercise, we focus our attention on OLS estimated reduced-form Phillips curve. We do so to contribute to the recent discussion on inflation dynamics and its formalization, discussion that has led some authors to prefer the ‘accelerationist’ version of the Phillips curve (e.g. Mankiw, 2001 and Estrella and Fuhrer, 2002) to the micro-founded, expectations-equipped new-Keynesian schedule (e.g. Woodford, 2003). Our results acknowledge the statistical importance of the Lucas critique, i.e. the reduced-form Phillips curve we estimate tend to be unstable. However, the magnitude of the changes in the estimated coefficients leads us to think that the Lucas critique might not necessarily be of overwhelming importance for policy analysis purposes, at least as far as the Swedish case is concerned. The paper is structured as follows. Section 2 outlines the small macro-model we employ to produce the simulated time-series of interest. Section 3 contains an explanation of the steps we implement to perform our econometric exercise. In Section 4 we present our findings, whose robustness is discussed in Section 5. Section 6 concludes, and References follow.

This paper aims at assessing the stability of reduced-form Phillips curves in presence of a policy break in the nominal exchange rate regime. We employ a new-Keynesian small scale open-economy dynamic stochastic model allowing for imperfect exchange rate pass-through and endogenous persistence in inflation, the output gap, and the nominal exchange rate for simulating such policy break. The model is carefully calibrated in order to plausibly represent the dynamics of an economy that experienced a dramatic policy regime change in the 1990s, i.e. Sweden. To assess the relevance of the Lucas critique in this context we estimate a reduced-form Phillips curve and assess its stability. Our results tend to support the statistical importance of the Lucas critique. Indeed, some signs in instability in the estimated Phillips curve seem to emerge. However, when looking at the estimated coefficients of the inflation curve, the differences under the two regimes are far from being large, both in terms of average realizations and from a distributional perspective. Therefore, our exercise does not tend to support the economic importance of the critique, and corroborates the conclusions by Taylor (1989) and Rudebusch (2005).