تغییر رژیم و اندازه گیری سیاست های پولی

This paper applies regime-switching methods to the problem of measuring monetary policy. Policy preferences and structural factors are specified parametrically as independent Markov processes. Interaction between the structural and preference parameters in the policy rule serves to identify the two processes. The estimates uncover policy episodes that are initiated by switches to “dove regimes,” shown to Granger-cause both NBER recessions and the Romer dates. These episodes imply real effects of monetary policy that are smaller than those found in previous studies.

Beginning with Hamilton's (1989) study of business cycle dynamics, regime-switching methods have proven extremely useful in a wide range of applications in macroeconomics and finance.1 This approach also holds promise for the measurement of monetary policy, since policy is typically regarded in terms of fluctuations between persistent regimes involving stronger or weaker anti-inflationary postures. Commonly used vector autoregression (VAR) methods for measuring policy cannot capture such persistent policy regimes, as these methods can identify only highly transitory policy shocks.2 This paper conducts an exploratory study of the use of regime switching for estimating monetary policy preferences. Our strategy is to avoid theoretical details by adopting a stylized model of policy determination that captures inflation/unemployment tradeoffs in a simple way. This approach allows for straightforward resolution of estimation issues, and our results may be viewed as a first assessment of the usefulness of regime switching for monetary policy measurement. The model posits that the policymaker is constrained by a standard expectations augmented Phillips curve. The Phillips curve contains a parameter that follows a two-state Markov process, reflecting periodic shifts in the natural rate of unemployment. The policymaker adopts an inflation target that embodies tradeoffs between inflation and unemployment, captured by a preference parameter that follows an independent two-state Markov process. The latter process switches between a “dove regime,” in which the policymaker more readily accommodates increases in the natural rate, and a “hawk regime,” in which there is less accommodation. Characterizing the policy process relies on the fact that a rise in the natural rate leads to a larger increase in the inflation target when the preference parameter is in the dove state, relative to the hawk state. Since the policy process is uncorrelated with the reduced-form residuals, the natural rate process can be distinguished probabilistically from the policy process, making it possible to estimate both processes. Estimates of the model are obtained by means of Gibbs sampling using monthly data over the period 1965:3–1999:2. Highly persistent natural rate and policy processes are estimated, each having statistically distinct state values. Further, we obtain estimates of the posterior expected values of both the natural rate and policy parameters over the sample period, providing a picture of the evolution of natural rate and policy regimes. The policy process, in particular, displays three “dove episodes”—one each in the late 1960s, mid-1970s, and an interval around 1980. These episodes correspond closely to the onset of NBER recessions, as well as to the dates identified by Romer and Romer, 1989 and Romer and Romer, 1994 as reflecting policy tightening by the Federal Reserve. The three episodes follow a basic pattern: A switch to the dove regime first occurs, followed roughly a year later by a Romer date and then a recession. A switch back to the hawk regime occurs after another year. Switches toward the dove regime are shown to Granger-cause both recessions and the Romer dates. This suggests that monetary policy regimes are driven by shifts toward looser policy, initiating a process of policy reversal that takes roughly two years. In other words, monetary policy is driven by persistent “dove switches.” The implications of these policy episodes for output, prices, and other variables are assessed by means of a VAR that treats the estimated posterior expected values of the natural rate and policy parameters as exogenous variables. Using the estimated VAR, we study the dynamic effects of a stylized policy episode in which the policy parameter switches to the dove regime for 24 periods and then switches back to the hawk regime. The onset of the dove regime initiates a steady rise in prices, while output begins to decline after a year. Prices fall after the hawk regime is restored, and output bottoms out about a year later. The switch to the dove regime also induces a sharp rise in the federal funds rate, and the federal funds rate jumps upward again halfway through the dove episode. The onset of our dove regime shares a number of characteristics with a positive federal funds rate shock in the standard VAR model, including the increase in price levels following the shock. In the present case, however, there is no “price puzzle,” since the switch actually represents a loosening of policy. Our policy episode generates a significantly smaller decline in output than that associated with federal funds shocks in standard VAR models. Moreover, the cumulative increase in unemployment associated with the restoration of the hawk regime, relative to the corresponding reduction of inflation (the so-called “sacrifice ratio”), is only 0.87, less than half the value found in previous studies. Overall, our results suggest that the real effects of monetary policy may be less significant than previously believed. Section 2 presents the model, estimates are given in Section 3, and comparisons with NBER recessions and the Romer dates are carried out in Section 4. Implications of policy episodes and the sacrifice ratio are considered in Sections 5 and 6, respectively, and Section 7 concludes.

In this paper, we apply regime-switching techniques to the measurement of monetary policy regimes. Using a stylized model of inflation/unemployment policy tradeoffs, we obtain estimates that reveal highly persistent processes of policy preferences and economic structure, switching between distinct states. The estimated posterior expected values of the policy parameter trace out episodes involving switches to a dove regime for about two years, followed by reversion to a hawk regime. The switches that initiate these episodes Granger-cause both NBER recessions and the Romer dates, suggesting that incidents of monetary tightening might be best regarded as responses to earlier dove shocks. Our estimated policy episodes imply smaller effects on real variables than have been obtained in previous studies using different policy measures. Our model may be extended to allow regime switches to depend on the duration of regimes or economic variables. The methodology can be applied to a broader set of variables that may influence policymaker preferences, including employment, output, and financial market variables such as interest rates, and it may be applied to other countries. Policymaker objectives may be combined with policy instruments to create a synthetic analysis linking policy regimes with the particular instruments used to implement these regimes. We have relied on a bare-bones natural rate model that has allowed us to obtain sharp estimates, but that also raises valid questions of robustness. A more theoretically complete model would incorporate explicit utility maximization by a forward-looking policymaker, lags in implementation of policy targets, and a richer structure of expectation formation by private agents. The econometric implementation of such a model represents a challenging and, in view of our results, potentially fruitful avenue for future research.