Is the Canadian housing market ‘really’ exuberant? Evidence from Vancouver, Toronto and Montreal

ABSTRACT

Since the 2008 financial crisis, federal and provincial governments in Canada has consistently monitored its housing market. However, real estate markets in Canada seem to have significant regional imbalances with moderate evidence of overheating and price exuberance and strong evidence of over-evaluation for several quarters. In this perspective, this paper seeks to test the existence of real estate bubbles in Canada and their contagion. Our methodology, based on monthly data, relies on two steps: First, we build on a GSADF test developed by Phillips, Shi, and Yu (2015) to detect bubbles. Second, using a method developed by Greenaway-McGrevy and Phillips (2016), a non-parametric regression with time-varying coefficient, we verify if, in case of bubbles, there is a contagion among Canadian cities. Our results suggest that there are bubbles in the real estate markets of Toronto and Vancouver, while this is not the case for Montreal.

KEYWORDS:

• Exuberance
• bubble
• contagion
• gsadf test
• Canada

JEL CLASSIFICATION:

• C12
• G12
• R31

Acknowledgments

We are very grateful to Greenaway-McGrevy from the University of Auckland for the Matlab code. We also want to thank Les Oxley from the University of Waikato, Kevin Moran and Jean-Louis Bago from Laval University, Itamar Caspi from the bank of Israel, Valerie Grossman and Enrique Martnez-Garca from the Federal Reserve Bank of Dallas for their insightful comments on an earlier version. Any remaining errors are our own.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ For Montreal, data is only available for the period starting from January 2000.

² Here we used the notation of Deng et al. (2017).

³ As the real estate price exuberance in Vancouver and Toronto occurred between November 2003 and September 2004, the bubble migration analysis is performed for the period starting in November 2003. However, due to the rolling window approach, the first 36 months are excluded, and rolling estimates are computed for the period starting in 2007.