Over the last 15 years, agent-based simulation models have been used to explore pressing issues in housing markets related to the determinants and dynamics of urban sprawl, gentrification, disaster risk, and housing bubbles. While their methodology is rapidly evolving, fundamental challenges remain regarding empirical estimation of buyer’s willingness to pay for a property, especially as researchers often only have data on property transactions. A parallel question relates to the role that hedonic statistical models, which estimate incremental contributions to value of property and neighbourhood attributes, can play in the development and evaluation of empirical agent-based land-market models.
We illustrate two alternative approaches to empirically establish bidding rules for buyers in our agent-based model, designed to explore the effects of a soon-to-be-launched light rail transit line (LRT) on residential land use in Waterloo Region, Canada. The first approach uses a state-of-the-art spatial econometric hedonic model combined with a household budget constraint to estimate buyer demand for properties. The second approach, in the development stages, combines an updated spatial econometric model with information on buyer stated preferences, demographics, and resources, from a recent buyer-seller survey conducted by our group. We illustrate alternative approaches to demand estimation that are tightly coupled to microeconomic theory, drawing on the concept of Rosen’s second-stage demand estimation. Our survey also contains detailed data on transportation preferences and mode choice, allowing us to explore the impacts of these factors on housing demand. We also discuss the important role of qualitative research to our work, which has helped us understand constraints to supply and emerging market segmentation dynamics. As our survey took place in a highly dynamic market (a bubble of about a year’s duration), we expect that our modelling results will also shed new light on the bidding dynamics that contribute to housing bubbles.