National and regional housing vacancy : insights using Markov-switching models

We examine homeowner vacancy rates over time and space using Markov-switching models. Our theoretical analysis extends the Wheaton (1990) search and matching model for housing by incorporating regime-switching behavior and interregional spillovers. Our approach is strongly supported by our empirical results. Estimations, using constant-only as well as Vector Autoregressions, allow us to examine differences in vacancy rates as well as explore the possibility of asymmetries within and across housing markets, depending on the state/regime (e.g., low or high vacancy) of a given housing market. Estimated vacancy rates, conditional on the vacancy regime, which are found to be persistent, vary across regions in all Markov-Switching Vector Autoregression (MS-VAR) models. Models allowing for interregional effects via lagged vacancy rates and controls relating to migration tend to perform better than models lacking this feature. These models track vacancies well. Noteworthy is their performance during the Great Recession/Financial Crisis. The importance and diversity of interregional effects are demonstrated, and vacancies in a specific Census region are affected by vacancies in other regions. Moreover, the sizes of these effects depend on the vacancy state of the specific region