Homogeneous Groupings of Metropolitan Housing Markets

Abstract

In this paper, we use clustering techniques to identify structural relationships among U.S. housing markets and develop a bootstrapping procedure to test whether associations between cities are significant. The method allows the creation of meaningful "groups" of cities. These groups are useful for purposes of diversification, and for identifying appropriate hedging proxies for city-specific futures instruments. A clustering algorithm, K-means, is applied to the 1977-1992 returns to housing price indices in 30 metropolitan U.S. housing markets. It demonstrates strong regional differences in housing price fluctuations. When three groups are specified, we find a West Coast group, an East Coast group, and a central U.S. group. When more groups are specified, the West Coast divides into two clusters that are not north and south, and Texas cities separate from the central U.S. group. Using bootstrap methods, we reject the hypothesis that these groupings are a result of random associations. © 1994 Academic Press. All rights reserved.