There is a long tradition in monetary economics of searching for a single policy variable perhaps a monetary aggregate, perhaps an interest rate that is more or less controlled by policy and stably related to economic activity. Whether the variable is conceived as an “indicator of policy” or a “measure of policy stance”, correlations between it and macro time series are taken to reflect the effects of monetary policy. Conditions for the existence of such a variable are stringent. Essentially policy choices must evolve autonomously, independently of economic conditions. Even the harshest critics of monetary authorities would not maintain policy decisions are unrelated to the economy. The line of work we extend builds on a venerable econometric tradition to emphasize the need to specify and estimate behavioral relationships for policy. The estimated relationships separate the regular response of policy to the economy from the response of the economy to policy, arriving at a more accurate measure of the effects of policy changes.
One sometimes encounters a presumption that models for “policy analysis” and “forecasting” are sharply distinct. A model useful for policy choice need not fit data well and well-fit models necessarily sacrifice economic interpretability. We do not share this presumption, and aim here to show that it is possible to construct economically interpretable models with superior fit to the data.
As the recent empirical literature on the effects of monetary policy has developed ways of handling more complex, multivariate data sets, a variety of models and approaches has emerged. Researchers have chosen different data sets, made different assumptions, and tended to emphasize the differences between their results and those of others, rather than the commonalities. This paper uses a single time frame and data set to test the robustness of results in this recent literature and to trace the nature and sources of differences in conclusions.
We analyze and interpret the data in ways that do not impose strong economic beliefs. The methods we employ permit estimation of large time series models, allowing more comprehensive analysis of the data. Some of the models integrate policy behavior with the banking system, the demand for a broad monetary aggregate, and a rich array of goods and financial market variables, providing a more complete understanding of the monetary transmission mechanism. Weak economic assumptions and large models combine to reveal difficulties with sorting out policy effects that other approaches fail to bring out.
The size of effects attributed to shifts in monetary policy varies across specifications of economic behavior. We show, though, that most of the specifications imply that only a modest portion (or in some cases, essentially none) of the variance of output or prices in the US since 1960 is attributable to shifts in monetary policy. Furthermore, we point out substantive problems in the models that imply large real effects, and argue that correcting these problems lowers the implied size of the real effects.
Another robust conclusion, common across these models, is that a large fraction of the variation in monetary policy instruments is attributable to systematic reaction by policy authorities to the state of the economy. This is of course what we would expect of good monetary policy, but it is also the reason why using the historical behavior of aggregate time series to un coverthe effects of monetary policy is difficult.
Download
What Does Monetary Policy Do?
