Oil Shocks and Aggregate Macroeconomic Behavior: The Role of Monetary Policy: A Reply

Oil Shocks and Aggregate Macroeconomic Behavior:The Role of Monetary Policy Ben S. Bernanke (bio), Mark Gertler (bio), and Mark W. Watson (bio) JEL codes: E32, E50 Keywords oil shocks, monetary policy Hamilton and Herrera (HH) have provided an interesting comment on our 1997 paper (Bernanke, Gertler, and Watson 1997 BGW). We take the opportunity offered to us by the editors to respond briefly. The goal of BGW was to show that the magnitude and shape of the economy's response to a particular exogenous shock will typically depend critically on how monetary policy makers choose to react to the shock. As a consequence, we argued that assessments of the importance of monetary policy for real activity should take into account the systematic portion of policy (i.e., the policy rule) as well as the unsystematic component (i.e., monetary policy shocks.) The specific type of exogenous shock that we considered in BGW was a sharp increase in the price of oil, of the magnitude observed during several episodes in the 1970s. Using a modified VAR framework, we considered counterfactual scenarios in which monetary policy (represented by the level of the federal funds rate) does not respond to an oil price shock. We found that the adverse effects of an oil price shock on output are reduced considerably when the endogenous response of the funds rate is "shut off." Indeed, our point estimates suggested that the endogenous response of monetary policy accounted for virtually all the negative impact of the oil shock on End Page 287 output (though, as we discuss in the paper, there is considerable sampling uncertainty about the true response). HH raise several legitimate issues about our analysis. First, they correctly point out that our counterfactual exercises typically correspond to shifts in the funds rate of hundreds of basis points, changes out of the range of historical experience. They ask whether the Federal Reserve could engineer such large swings in the nominal funds rate in practice. In our view, given a plausible degree of nominal price rigidity and the Fed's monopoly of the supply of bank reserves, the Fed could certainly alter the path of the nominal funds rate by the amount required by our counterfactuals. The more debatable point, however, is whether the parameters in our modified VAR can reasonably be treated as invariant to policy shifts of the type we considered, which involved permanently shutting off the response of the funds rate to the exogenous shock. In our original paper we tried to address this issue by incorporating a limited amount of structure in the model. In this note we will report results from a counterfactual scenario in which the monetary policy response to the change in oil prices is not eliminated, but only delayed for several quarters. It seems plausible to us that a purely transitory deviation from the usual policy rule would not significantly affect the structure of the economy (that is, the quantitative effect of the Lucas critique should be small). As we describe below, in this alternative simulation we find that monetary policy's endogenous response accounts for about half of the depressing effect of an oil price shock on output—less than we found in BGW, but still quite significant. The second issue HH raise is the potential sensitivity of our point estimates to assumptions about lag length. Because we were estimating a seven-variable VAR system with monthly data, we were concerned about the implications for sampling error of including too many lags and hence an excessive number of parameters. Accordingly, we used an Akaike information criterion (AIC) to determine lag length, which resulted in our choice of seven monthly lags. Of course, while the use of fewer lags reduces sampling uncertainty, it also introduces the possibility of omitted variable bias. In their comment, HH test the 7-lag model against models with 12 and 16 lags. Based on conventional F-tests, they find that the extra lags enter significantly, so that the 7-lag model is rejected. HH also show that the point estimates of the models with longer lags imply a smaller contribution of endogenous monetary policy to the total effect on output of an oil...