Second Order Time Dependent Inflation Persistence in the United States : a GARCH-in-Mean Model with Time Varying Coefficients

In this paper we investigate the behavior of inflation persistence in the United States. To model inflation we estimate an autoregressive GARCH-in-mean model with variable coefficients and we propose a new measure of second-order time varying persistence, which not only distinguishes between changes in the dynamics of inflation and its volatility, but it also allows for feedback from nominal uncertainty to inflation. Our empirical results suggest that inflation persistence in the United States is best described as unchanged. Another important result relates to the Monte Carlo experiment evidence which reveal that if the model is misspecified, then commonly used unit root tests will misclassify inflation of being a nonstationary, rather than a stationary process