Assessing Portland's Smart Growth: A Comprehensive Housing Supply and Location Choice Modeling Approach

Abstract

There are extensive empirical studies on the impacts and effectiveness of Smart Growth policies; however, very few of them consider the perspective of individual decision makers and, to this author's knowledge, none have studied developers as location-aware decision-making agents. This study tries to fill this gap partially by assessing the impacts of Portland's smart growth policies on developers' location choice behavior with developer-based location choice models.The dissertation has two purposes. By assessing the impacts of Smart Growth policies on individual home developer's location choice, it provides a micro-and behavioral foundation for the understanding of Smart Growth policies. As a bi-state metropolitan area located on the border between Oregon and Washington, the Portland region provides a unique environment that allows my research to examine whether home developers react The three-step new housing supply and location choice forecast model seem to be able to capture the basic trend of housing market and land development in the Portland region.Three different aggregate housing supply forecast models, an conditional time series regressive model, a unconditional time series regression model, and an auto-regression iii integrated moving average (ARIMA) model were tested and their advantages and disadvantages were discussed. Both the SFH and MFH project synthesis models can simulate housing projects well for a forecast year.Three location choice models were developed to allocate synthesized housing projects into space. The three models are characterized separately as: (1) assumed market homogeneity and atomization of development projects; (2) deterministic market segmentation and synthesis of projects by size; and (3) probabilistic market segmentation and synthesis of projects by size, using a latent class approach. Examination of forecast results shows that all three models can successfully capture the basic spatial pattern of housing development in the region; however, the spatial distribution of MFH development is lumpier and more unpredictable. While Models 2 and 3 are more sophisticated and make more sense from a theoretical perspective, they do not return better forecast results than Model 1 due to some practical issues. Models 2 and 3 would be expected to perform better when those practical issues are solved, at least partially, in future research.iv ACKNOWLEDGEMENTS