Micro-geographic housing dynamics: Trajectories and tipping points

Keywords: housing lifecycle, neighborhood change, unsupervised learning, natural language processing, early warning system
Abstract Type: Paper Abstract

Abstract

In this article, we introduce an approach for estimating trajectories and tipping points for changes in housing characteristics at a micro-geographic scale. The approach is based on point-level property listing data which we classify using an unsupervised text classification algorithm grounded in theoretical concepts around housing lifecycles. Classified listings are then used in a k-means algorithm to summarize the mixture of housing developments in micro-geographic areas. To enhance our understanding neighborhood housing lifecycles, we examine sequences of housing mixture classes that these micro-geographic areas transition through over time. Finally, we estimate tipping points to gain a better understanding of when a neighborhood is likely to transition to another housing mixture class. Our suggested approach can serve as an ‘early warning system’ for planners and local stakeholders of when neighborhoods are about to experience significant change. The spatiotemporal granularity of the data allows for real-time indication of such changes.