متغیرهای اطلاعاتی برای سیاست های پولی در مدل ساختاری برآورد شده از منطقه یورو

A small scale new keynesian model for the euro area is estimated with maximum likelihood under the assumptions of imperfect information and discretionary monetary policy. The estimated parametrization of this widely used dynamic stochastic model unveils the monetary authorities’ objectives and the information content of two indicator variables: monetary aggregates and real unit labour costs. The results highlight a significant policy concern about interest-rate smoothing and inflation; almost no concern for output gap stabilization emerges. Regarding indicator variables, unit labour costs provide information on potential output that is helpful for stabilization policy; no useful information role emerges for monetary aggregates.

Dynamic stochastic models of the new keynesian variety have acquired a solid position in the analysis of monetary policy.1 Only recently, though, economists have attempted to estimate the structural parameters of these models, a key step for the credibility of the policy advice they were ultimately designed for. This paper applies a solution method by Svensson and Woodford (2003) to solve and estimate a new keynesian model for the euro area under the assumptions of discretionary monetary policy and imperfect information. To the best of our knowledge, this is the first attempt at simultaneously estimating the economic structure, the preference weights and the information structure within a new keynesian framework. The results contribute to this branch of the literature along three dimensions. First, previous quantitative analyses dealing with imperfect information in new keynesian models proceed by separating the estimation of the structural and information model parameters into two sequential stages (e.g. Ehrmann and Smets, 2003; Coenen et al., 2005). Such a separation is in principle problematic because, as shown by Svensson and Woodford (2003), the dynamics of all the variables depend on both sets of parameters when information is imperfect. An advantage of the maximum likelihood (ML) method pursued here is that it allows us to estimate the economy's structural and information parameters simultaneously, as predicated by the underlying economic theory. A comparison of results highlights important differences between our approach and the two-stage procedure. Second, the analysis provides an estimate of the weights attached by the monetary authority to policy objectives (the volatility of inflation, the output gap and interest rate adjustments). This distinguishes the paper from previous estimation exercises, e.g. Ireland (2004b) for the United States or Smets and Wouters (2003) for the euro area, in which a “simple” instrument rule is used to describe monetary policy. The results quantify the relevance of the various objectives and illustrate the extent to which actual policy can be described in terms of these targets. Finally, the incomplete information framework is used to assess the usefulness of two indicator variables often discussed in the practice of monetary policy: monetary aggregates and real unit labour costs. The paper is organized as follows. The next section introduces a dynamic stochastic monetary policy model and describes the imperfect information problem. Section 3 estimates the model following a method proposed by Sargent (1989), McGrattan (1994) and Ireland, 2004a and Ireland, 2004b. Section 4 discusses the main differences with respect to a two-stage estimation method. The model implications concerning the optimal policy rule and the usefulness of indicator variables are the subject of Section 5. The main findings are summarized in a concluding section.

This paper used maximum likelihood to estimate a small scale dynamic stochastic model for the euro area. The model accounts for imperfect information about the state of the economy and allows us to assess the role of monetary aggregates and unit labour costs as indicator variables for monetary policy. Moreover, by assuming optimal discretionary policy, the estimates provide a characterization of policy in terms of the relative weights assigned to the targets: inflation, the output gap and the volatility of the short term interest rate.19 Three main results emerge from the paper. First, the structural estimation implemented by this paper allows us to identify some biases that emerge in a two-stage procedure that separately estimates the structural parameters and the information structure. The comparison shows that neglecting the simultaneity problem, as the two-stage approach does, leads to a significant underestimation of the magnitude of measurement errors, especially about inflation, and affects the estimates of the Phillips curve slope. In particular, the inflation elasticity to the output gap turns out to be much smaller with ML than with the two-stage method. Second, the estimated weights for the targets of monetary policy suggest that interest-smoothing is assigned a very large weight, followed by inflation stabilization and lastly by the output-gap target, whose weight among policy objectives turns out to be very small.20 The estimates indicate that a data-consistent characterization of optimal policy requires the output gap not to be an important policy target. The significant role detected for this variable by several reduced-form interest rate rules probably reflects the fact that the output gap is a leading indicator of inflation, not a target in itself. Lastly, the analysis shows that the M3 monetary aggregate does not help towards a better identification of the state variables of interest for stabilization policy. A more useful role emerges for the unit labour costs indicator, which contains information on potential output that helps to reduce the volatility of the output gap. Several working assumptions stipulated by our analysis are worth further investigation. Our paper considered a simple model without capital accumulation and focussed on nominal price rigidity. One extension includes integrating the analysis to account for capital accumulation, sticky wages and capital adjustment costs. Another one involves solving the model by a second-order approximation—e.g. relying on non-linear methods such as those considered by Fernández-Villaverde and Rubio-Ramírez (2004). In this case, uncertainty and imperfect information would influence optimal policy since the certainty equivalence principle would cease to apply.