New evidence on the specification and performance of neoclassical gravity models in the study of urban transportation

Abstract

Gravity models with power and exponential resistance functions are derived using constrained neoclassical utility maximization. Both equations are then fitted to a set of weekday motor vehicle traffic data describing trips in the Los Angeles, California, area for the year 1966, using multiple linear regression analysis. The equation with the power resistance function gives the better fit for all twenty origin areas analyzed. This result ensues because the reciprocal of the power resistance function is more flexible; it can vary from concave to highly convex, but the exponential cannot. Analysis of estimated regression coefficients shows that the great majority of tripmakers was constrained by a convex time cost function rather than the concave money cost of travel in all but two of the poorest origin areas. The implications of these results for the understanding of automobile travel in urban areas are discussed.