Estimating inflation persistence by quantile autoregression with quantile-specific unit roots

In this paper we study inflation persistence, which is a key feature of inflation dynamics, related to how quickly a stationary inflation process reverts to its long-run equilibrium after a shock. Emerging economies with high inflation persistence need to adjust macroeconomic policies in a significant way to price shocks (e.g., at the cost of substantial output decrease), since these shocks can affect expectations and inflation for a much longer period. We propose a novel way to estimate inflation persistence by using a quantile autoregression (QAR) model, which allows for asymmetric dynamics and quantile-specific unit roots. An empirical exercise with Brazilian data from January 1995 to May 2017 illustrates the method. The results indicate that inflation is globally stationary, but exhibits non-stationary behavior in 28% of the observations. In addition, shocks occurring when inflation is higher seem to have greater dissipation time compared to shocks that occur when inflation is lower.