Work-Trip Mode-Choice Models for South Africa

Abstract

This paper describes a number of individual-choice models of work-trip mode-choice, calibrated in the course of a bus study in two communities of Metropolitan Johannesburg. First, it is found to be useful to reduce the alternative set from eight modes to four modes, without loss of estimation power. Second, a number of alternative specifications of costs and times are examined, from which the chosen model is selected as one containing five such variables. In addition, a variable of car competition is found to add significantly to the explanatory power of the model. Finally, recently-documented procedures are applied to the selected model specification to pool the results of two random subsamples and to correct the choice-based biases in the alternative-specific constants. It is found that the coefficients of the South African model fall well within the range of coefficients determined in the U.S.