Correcting for endogeneity due to omitted crowding in public transport choice using the Multiple Indicator Solution (MIS) method

Abstract

Crowding levels are very relevant for the analysis and evaluation of the performance of public transport as they strongly affect the level of service and the overall perceived quality of the system. However, crowding is not an easy variable to measure and, hence, demand models often tend to ignore or use abstract proxies for it. In this paper, we assess the Multiple Indicator Solution (MIS) method in a Stated Preference (SP) experiment where crowding conditions were displayed to the respondent but are artificially omitted in the estimation of a curtailed model to cause endogeneity. Results provide evidence that the MIS method can be used to control for a wide range of omitted attributes in SP data. We also discuss the potential application of this approach to Revealed Preferences (RP) models of public transport by asking suitable post-trip questions to users. Two MIS variations were applied to this SP case study and both provided outcomes that were superior to those of the curtailed model. We enrich the analysis with the aid of Monte Carlo simulation. Results suggest that potential problems may arise in the presence of neglected interactions and if indicators are only weakly correlated with the omitted attribute. For the SP case study analysed, only the former issue seems to play a role in the results. The article finishes by discussing the implications of these findings for the correction of endogeneity on SP and RP data on public transport and suggesting future lines of research in this area.