Microsimulation of household and firm behaviors : coupled models of land use and travel demand in Austin, Texas

Households and firms are key drivers of urban growth, yet models for forecasting travel demand often ignore their dynamic evolution and several key decision processes. An understanding of household and firm behavior over time is critical in anticipating urban futures and addressing transportation, land use and other concerns. Birth and death, migration and location choice are defining events in a household's and firm's life cycle, and a study of household and firm evolution requires the estimation and application of models for each of these. Such an exercise is hindered primarily by a lack of quality micro-data. This thesis develops a basic framework for modeling household and firm demographics using microsimulation. Year 2005 zonal household population and employment point data for the Austin, Texas region, coupled with various, more aggregate data sets, are used to simulate household and firm evolution over time and space. The model consists of household evolution, firm evolution, location choice and travel demand models. Household and firm simulation models are run at one-year time steps, in order to forecast Austin's future. The household simulation component is made up of models for birth (of children and of households), death of individuals (and other forms of household dissolution), migration, children leaving home, vehicle ownership, and location choice. These models are estimated using multinomial logit and Poisson specifications. The firm simulation component consists of firm birth, death, growth and location choice models. A Markovian process is assumed in order to anticipate firm growth and contraction (across firm-size categories), along with logit and Poisson model specifications for firm location choice. Firms are categorized based on number of jobs (6 categories) and industry sector (4 sectors) they belong to. Austin's household and commercial vehicle travel survey data were used to estimate trip generation and distribution models. Simulation results for multiple growth-rate scenarios suggest a roughly 180% increase in the Austin population over a 30-year period, 210% increase in vehicle ownership, a 230% increase in jobs, and more than a 300% increase in vehicle-miles traveled. When a 10-cent/mile flat-rate toll is applied over all links, the year 2035 VMT is predicted to be just 3% less than under the no-toll scenario. A fixed toll of 10-cents-per-mile shows a very low impact on VMT over a 30 year period than expected. To ensure a jobs-worker balance, the model may well merit greater synchronization of the population and firm synthesis models. The simulations also suggest a clear shift of firms and households towards more central zones, in part because of the cross-sectional nature of the data sets used to calibrate the location choice models and the lack of density restrictions or other reflections of land-availability constraints on new development. Essentially, households and firms exhibit a strong centralizing tendency, that Austin's land market simply cannot allow, due to space and other constraints on new building. Explicit expressions of such constraints should prove helpful in future implementations of this work. While microsimulation of urban systems is data and computing intensive, it provides a flexible tool for analyzing the impacts of various policy decisions as well as other, demographic, environmental and system changes. It allows transportation planners explore the potential responses of individuals to changes in their environments and predict the long-term implications of policy decisions. This thesis seeks to be a bridge for further integrated travel demand and land use models of this type.