44 Pages Posted: 21 Sep 2016
Date Written: September 20, 2016
We propose a unit root test for time series with a nonparametric trend component using a rolling window scheme with overlapping blocks. For both fixed-b and small-b blocklength asymptotics we derive limiting distributions that are robust under heteroskedasticity. By pre-whitening the series we account for autocorrelation. In a Monte Carlo simulation with slowly varying trend components both the fixed-b and small-b tests yield good size and higher power for near integrated alternatives than conventional unit root tests. An application to quarterly inflation rates of 22 countries provides evidence that inflation rates are stationary around a slowly varying trend.
Keywords: Unit root testing, nonlinear trends, heteroskedasticity, inflation
JEL Classification: C12, C14, C22, E31
Suggested Citation: Suggested Citation