Testing for Bubbles in Housing Markets : A Panel Data Approach

We employ recently developed cross-sectionally robust panel data tests for unit roots and cointegration to find whether house prices reflect house related earnings. We use U.S. data for Metropolitan Statistical Areas, with house price measured by the weighted-repeated-sales index and cash-flows by market tenants' rents. In our full sample period, an error-correction model is not appropriate, i.e. there is a bubble. We then combine overlapping ten-year periods, price-rent ratios, and the panel data tests to construct a bubble indicator. The indicator is high for the late 1980s, early 1990s and since the late 1990s. Finally, evidence based on panel data Granger causality tests suggests that house price changes are helpful in pedicting changes in rents and vice versa