Time-Varying Inflation Target and Unbiased Taylor Rule Estimation

Taylor rule estimation can be biased if the central bank's inflation target is counterfactually held fixed. First, we identify in the standard New Keynesian model a downward bias affecting the OLS estimate of the policy response to inflation. Based on simulations from a similar model with positive trend inflation, we show that the policy response to inflation is nearly half the true value with large inflation target fluctuations, as purportedly exhibited by the Fed during the 1960s and 1970s. Second, we implement a procedure for estimating DSGE models using a novel population-based MCMC routine known as parallel tempering. Applying this procedure to a medium-scale DSGE model with positive trend inflation, we do not find evidence of a bias when jointly estimating model parameters and the unobserved time-varying inflation target.