Testing for Bubbles in Housing Markets: A Panel Data Approach

A bursting bubble in a housing market can have a severe negative impact on consumption and GDP. Hence, it is of interest to identify a presence of the bubble in a timely fashion. Existing tests often rely on the relationship between house prices and their corresponding fundamentals, e.g. rents. These tests typically employ standard univariate unit root methodology and require relatively long time series, which precludes a timely testing. We therefore combine panel data tests for unit roots, cointegration and Granger causality using shorter time span data on house prices and rents in the US metropolitan areas. For our full sample, we find that there is no relationship between house prices and rents in levels but there is one in first differences. Also, a ëbubble indicatorí, which is one whenever there is no statistical relationship between levels of our two variables, is defined and applied to overlapping ten-year periods. This indicator shows that one period of possible bubble occurred in the late 1980s and another in 2001-2003.