Nonparametric Modeling for the Time-Varying Persistence of Inflation

This article develops a novel nonparametric time-varying auto-regressive distributed-lag model to estimate and test the persistence of inflation. To characterize the temporal instability of persistence in the inflation process, our proposed model allows for time-varying coefficients with unspecified functional forms. The local linear estimation method is employed to estimate the coefficients, and a wild bootstrap method is used to construct their confidence bands. Monte Carlo simulation illustrates that the proposed estimation method works well in finite and large samples. Empirically, there appears to be a strong instability in the persistence of the inflation process in the United States — the level of persistence declined before the 2007-2009 global financial crisis (GFC) but then sharply rebounded and achieved the local maximum during the recent COVID-19 pandemic for the post-GFC period