Size improvement of the KPSS test using sieve bootstraps

It is widely known that size distortions of the so-called KPSS stationarity test, introduced in Kwiatkowski et al. (1992), become severe with persistent data. We propose the sieve bootstrap introduced by Bühlmann (1998) as an appropriate bootstrap for dependent processes, to obtain notable size improvement of the KPSS test. Our simulation studies demonstrate that sieve bootstraps can be effective in refining the finite-sample size performance.

► Sieve bootstraps are employed to improve the size performance of the KPSS test. ► Simulation studies confirm that sieve bootstraps can be effective. ► Application to the Nelson-Plosser US macroeconomic series is presented.