wrong estimation of the true number of shifts in structural break models: Theoretical and numerical evidence

The aim of the paper is to consider the problem of selecting the number of breaks in the mean of a time series. Indeed, we prove analytically and show by a Monte Carlo study that some model selection criteria will tend to choose a spuriously high number of structural breaks when the process is trend-stationary without changes. The important question suggested by our results is that of distinction between trend-stationary process and random walk when modelling real data series.