Testing Long Memory in the Presence of Structural Breaks: An Application to Regional and National Housing Markets

Abstract

We use parametric and semi-parametric methods to explore the effects of structural breaks on long memory processes in nine US regional and national housing prices over the period from January 1991 to February 2014. The results reveal multiple structural breaks and differential break points across regions. The regional break points do not coincide with the national break suggesting a spatial pattern of the underlying determinants of regional housing prices. We find long memory in regional housing prices using the entire sample period, but the results are generally reversed in sub-samples (regimes) that incorporate structural breaks. Our evidence suggests that failure to account for structural breaks when testing for long memory can lead to incorrect inferences. The results, which proved robust to model specifications, have important implications for policy prescriptions, for market efficiency, and for the integration of regional housing markets.