Time-Varying Correlation in Housing Prices

In the wake of the housing crisis, credit rating agencies have received much blame, particularly for the statistical tools they used to measure correlations in housing prices in different locations. Several studies have proposed alternative statistical models, but to date, all such approaches assume that correlations remain constant over time. This paper argues that, regardless of the correlation patterns built into such statistical models, correlations might strengthen during times of financial turmoil. Consequently, mortgage-backed securities might have been appropriately diversified during "normal" times, but less so during extreme market swings. Using monthly data on housing prices in four major U.S. cities, the main findings confirm that housing prices do, indeed, exhibit correlations that change over time, and more importantly, those correlations appear to strengthen in the midst of market turmoil