Incorporating Spatial Dependencies in a Multinomial Logit Model

In this thesis, commuter's transportation choices for their Home-To-Work travel (HTWT) are analyzed with respect to the companies' characteristics in Belgium. The main focus is on mobility management measures, which are mobility policies whose main aim is a sustainable transport system, as well as external factors related with municipality characteristics and the infrastructure of the area. Taking into account spatial autocorrelation, which is referred to the dependency among companies over space, a mixed multinomial logit model involving random effects was considered. The results show that 'Guaranteed ride home' and 'Organization of carpooling' are efficient mobility management measures for promoting carpooling. In addition, fixed work schedules increase carpool use compared with car use. On the contrary, flexible work schedules contribute more to increase metro-tram-bus and train usage. Also, the existence of bus and train stations nearby the company is a major factor for employees to choose public transportation for their HTMT. Finally, regarding mobility management measures, financial incentives motivate more commuters to travel by bicycle instead of by car. However, external factors also have a large impact on commuter's behavior regarding cycling. In particular, the slope of the road network and job density are negatively related with cycling. The random effects in our model show that there are some unobserved factors that differ among boroughs which have an impact on employees' choices regarding train usage