Investigating the impact of satisfaction indicators on the efficiency of choice models: New evidence from Lebanon

Abstract

New approaches to choice modeling which aim at increasing model realism through increased model complexity to capture different aspects of human behavior and latent psychological constructs have been increasingly applied in recent transportation research efforts. One of the most recent approaches is the inclusion of happiness or satisfaction indicators which function as measures of utility. In previous work using revealed preference data, adding satisfaction indicators resulted in gains in efficiency. This paper explores whether gains in efficiency also hold in a stated preference context of mode choice where the data is more controlled and the satisfaction that is measured includes both satisfaction with a usual commuting mode and expected satisfaction for a non-routine commuting mode. The paper also tests whether gains in efficiency lead to any improvements in prediction accuracy. Finally, the paper also presents a scenario analysis for the prediction of changes in satisfaction as a function of changes in car and bus attributes. The results indicate that the addition of satisfaction indicators to the model indeed improve model efficiency in line with other findings from previous research utilizing the same framework. Such findings encourage choice modelers to collect satisfaction indicators in revealed and stated preferences surveys and to subsequently include these indicators in choice models to increase model efficiency. However, the gains in efficiency do not lead to improvements in prediction in the study context probably due to the controlled nature of the stated preference data. © 2016 Elsevier Ltd