Integration of Freight Network and Computable General Equilibrium Models

Over the past 30 years very significant progress has occurred in the understanding and modeling of passenger trip making behavior over networks. Yet, corresponding advances in understanding and modeling of freight transportation decision making over inter-regional, inter-modal networks have not occurred. In fact the most recent large scale U.S. freight network model is able to predict equilibrium network link volumes agreeing with Federal Railway Administration (FRA) density codes (reported data describing annual tonnages on every physical link of the rail system) with a frequency of only about 60% (Friesz et al., 1981; 1983a; 1983b; 1985). This is poor performance since density codes denote upper and lower bounds for link volumes; the difference between those upper and lower bounds is frequently of the same order of magnitude as the predicted volumes themselves. Poor as this accuracy is, it is substantially greater (about three times greater) than that reported for earlier models (Bronzini, 1980) and was achieved by straight-forward extensions of the urban passenger network modeling paradigm. Still greater accuracy may be obtained from a model designed specifically for freight applications from the outset.