I explore whether time-series methods exploiting the long-run equilibrium properties of the housing market might have detected the disequilibrium in U.S. house prices which pre-dated the Great Recession as it was building up. Based on real-time data, I show that a VAR in levels identified as in Uhlig (2003, 2004) would have detected the disequilibrium with high confidence by the Summer of 2004, with the estimated extent of overvaluation peaking at about 15 per cent immediately before the crisis. These results demonstrate that disequilibria in the prices of at least one asset class-housing-can indeed be robustly detected as they are building up. Conceptually in line with Cochrane's (1994) analysis for consumption and GNP, and dividends and stock prices, a key factor in order to robustly identify the transitory component of real house prices is applying Uhlig-style identification to real rents, which are cointegrated with house prices, and are comparatively much closer to the common stochastic trend. Directly focusing on house prices themselves, on the other hand, produces less robust results.
Structural VARs unit roots cointegration long-run restrictions medium-run identification Great Recession housing bubbles