Modelling pedestrian go-home decisions: A comparison of linear and nonlinear compensatory, and conjunctive non-compensatory specifications

This paper presents the main findings of an application of several models to predict the go-home decision of pedestrians in shopping streets. Two compensatory multinomial logit models, one with a linear utility function of time and the other with a nonlinear utility function of time, and a non-compensatory conjunctive model are specified. Data about pedestrian behaviour in a major shopping street in Beijing served as input for model estimation. The conjunctive model performs best, suggesting that pedestrians use simplifying heuristics to decide when to end the shopping trip and go home. In addition, the nonlinear multinomial logit model outperforms the linear model, indicating that marginal utility of time decreases with increasing time.