Customarily, for housing markets, interest focuses on selling prices of properties at locations and times. Hedonic models are employed using property-level, neighborhood-level, and economic regressors. However, in hedonic modeling the fact that the locations and times of property transactions are random is ignored. Here, we focus on explanation of the locations of transactions in space and time, viewing them as a point pattern over space and time. Our contribution is to explain such a point pattern using suitable regressors. We examine two explanatory models, the nonhomogeneous Poisson process and the log Gaussian Cox process. We study a point pattern in the city of Zaragoza, Spain, over the years, 2006–2014. We argue for point level modeling since the process of property sales operates at that scale. We elaborate efficient computation for fitting the foregoing models to the Zaragoza data. We show how the modeling enables rich inference and extraction of novel stories for this market over this time period. In addition, we clarify the potential benefits of this modeling for brokers, buyers, and administrators. To our knowledge, this is the first application of formal space–time point pattern analysis to locations of urban real estate transactions.
- Empirical coverage
- Log-Gaussian Cox process
- Nearest neighbor Gaussian process
- Nonhomogeneous Poisson process
- Ranked probability scores