Unit Root Testing with Slowly Varying Trends

A unit root test is proposed for time series with a nonparametric trend component using a pooled regression of overlapping blocks. The class of trend functions considered includes any boundedly differentiable trend function with finitely many breaks. Limiting null-distributions of the pseudo t-statistic of the pooled regression are derived under two different block asymptotics. Small-b asymptotics yields a standard normal distribution and under fixed-b asymptotics a functional of Brownian motions is obtained. A nuisance parameter correction provides heteroskedasticity robust tests and serial correlation is accounted for by pre-whitening. For both tests a Monte Carlo study with slowly varying trends yields both good size and improved power results when compared to conventional unit root tests