Regional and Longitudinal Disparity of Housing Bubbles in US Markets: Evidence from GSADF Tests

Abstract

The US housing price run-up and its subsequent collapse are considered by many to be central elements in causing the global financial crisis. This paper aims to estimate housing bubbles in 140 US cities by identifying the origination and termination of the bubble, during the period from 1989 to 2017. The study introduced the generalized supremum augmented Dickey-Fuller (GSADF) test, an innovative and powerful bubble detection method, to evaluate the regional US housing bubbles. Estimation results show several important findings. (1) Forty-five out of the 140 studied cities had existing housing bubbles. Phoenix-Mesa-Scottsdale had 8 bubbles, which was the largest quantity. Nine cities had one bubble, which are Northern New Jersey–Long Island, Richmond, Honolulu, Sarasota-Bradenton-Venice, Boise City–Nampa, Pensacola–Ferry Pass–Brent, Ocala, Atlantic City, and Barnstable Town. The bubble duration of Northern New Jersey-Long was the longest, ranging from 1999 Q4 to 2007 Q4. (2) The time period of bubbles can be divided into three stages based on its beginning time and duration. There were 48 bubbles at the first stage from 1992 to 2002, 85 at the second stage from 2003 to 2008, and 12 at the third stage from 2009 to 2017. The housing bubbles were mainly concentrated in the second stage. (3) Significant regional disparity can be observed in the bubbles. The 45 cities where there were bubbles were divided into three groups according to the population. Urban population is positively correlated with the average duration of the bubble. The lessons drawn from the US contribute to an in-depth understanding of the market dynamics and evolution patterns of housing bubbles.