Exploring the role of worker income and workplace characteristics on the journey to work

Traditional trip distribution processes that rely heavily on gravity models fail to capture how the characteristics of individuals or the heterogeneity in the attributes of attraction zones may influence the accessibility to jobs and, therefore, journey-to-work patterns. Different approaches, such as destination choice models, are not generally applied because of limited data availability and calibration requirements. This paper proposes an alternative approach to overcome this challenge by combining a utility-based measure of accessibility and a maximum range of commuting distance to predict the journey-to-work patterns of individual worker-agents using an open-access database. A multinomial logit model is estimated and an agent-based model is developed using data from the Census Transportation Planning Products (CTPP) 5-year database. The proposed methodology is demonstrated using a case study based on Tippecanoe County, Indiana, and the results are compared to a double-constrained gravity model. Results indicate that the utility functions derived from the CTPP database can replicate the aggregated journey-to-work patterns by income levels. Furthermore, it was found that the utility functions for low-, middle-, and high-household income groups could be different. Finally, while a calibrated gravity model could produce a trip-length distribution and average commuting distance more similar to observed data, the destination choice model provides more insights into the trip patterns across different household income groups, thereby providing a better basis for policy analysis.