Empirical Evaluation of Alternative Model Specifications to Predict Commuting Flows

Abstract

This study is devoted to empirical and modeling aspects on how characteristics of spatial structure influence commuting flows. Within a doubly-constrained framework, results from a competing-destinations formulation are evaluated and compared to results from the traditional gravity model. The evaluation depends critically upon the specification of within-zone journeys-to-work. Specific labor-market characteristics are found to be significant to explain how workers are absorbed in diagonal elements of the trip-distribution matrix. We also find that the parametric specification of the accessibility measure is important, and that the competing-destinations formulation is superior to the traditional gravity model.

The Spatial Transferability of Parameters in a Gravity Model of Commuting Flows

“…Labour market characteristics are reflected by the demand for labour originating from the firms in a specific zone, relative to the supply of labour originating from the zone. The results presented in Thorsen and Gitlesen (1998) supported a hypothesis which states that the relative frequency of within-zone journeys-to-work is high in a zone where employment is low relative to the labour force.…”

“…This is since the transportation network being considered in this paper includes road tolls and ferry connections. This was not the case in the version found in Thorsen and Gitlesen (1998). Both models belong to the gravity modelling tradition.…”

“…The model used in this paper is taken from Thorsen and Gitlesen (1998). In their paper they test several model specifications.…”

“…When agglomeration forces are dominant, the sign of the parameter ρ in Equation (1) will be positive, while the parameter takes on a negative value if competition forces are dominant. Thorsen and Gitlesen (1998) tested a hypothesis that special care should be taken with regard to potential benefits of residing and working in the same zone, represented by the additive constant exp(µδ ij ) to the diagonal elements of the trip distribution model specification. This approach was found to contribute significantly to the explanatory power of the model, suggesting that the option of residing and working in the same zone should be specifically accounted for in a model explaining commuting flows.…”

“…Despite this, the LR test remains popular in the literature. Thorsen and Gitlesen (1998) utilise the relative number of wrong predictions (RNWP) measure to compare model specifications. It is an appealing measure because its interpretation is extremely intuitive.…”