Toward a comprehensive, unified, framework for analyzing spatial location choice

In today’s world of increasing congestion and insufficient scope for infrastructural expansions, urban and transportation planners rely on the accuracy and behavioral realism of travel demand models to make informed policy decisions. The development of accurate and behaviorally realistic travel demand models requires a good understanding of individual travel behavior, and an important step toward this has been the development of the activity-based paradigm, which states that travel is a result of the desire to participate in activities at spatially scattered locations. Activity-based travel demand modeling systems essentially model the activity-travel patterns of individuals, which are characterized by several attributes such as activity purpose, location of activity participation and choice of mode. Of all these attributes, the choice of location of activity participation is one that has received relatively inadequate attention in the literature. On the other hand, the location of activity participation spatially pegs the daily activity-travel patterns of individuals. Accurate predictions of activity location are, therefore, key to effective travel demand management and air quality control strategies. Moreover, an understanding of the factors that influence the choice of location can contribute to more effective land-use and zoning policies. The broad objectives of this dissertation research are two-fold. The first objective is to develop a comprehensive econometric model of location choice for non-work activities that incorporates accuracy and behavioral realism in capturing different choice behaviors. This was achieved through the comprehensive introduction of heterogeneity in choice behavior, including observed and unobserved sources of inter- and intra-personal heterogeneity, spatial correlation, variety seeking and loyalty/inertial behavior, and spatial cognition. The estimation of such a flexible model typically requires the use of simulated maximum likelihood estimation (SMLE). The second broad objective of this research is to contribute toward improving the efficiency of the SMLE by comparing the performance of various quasi-Monte Carlo (QMC) sequences and their scrambled versions. Numerical experiments were designed and the Random Linear and Random Digit Scrambled Faure sequences are identified as the most efficient. Finally, all these research efforts contribute to the empirical estimation of non-maintenance shopping location choice models using panel data from the Mobidrive survey.