In this paper we develop a simple procedure which delivers tests for the pres-ence of a broken trend in a univariate time series which do not require knowledgeof the form of serial correlation in the data and are robust as to whether theshocks are generated by an I(0) or an I(1) process. Two...