EXPLORING BUS PASSENGERS' REPEATED CHOICES OF OPERATORS: A MIXED NESTED LOGIT MODEL APPROACH USING PANEL DATA

The purpose of this study is to examine the selection of bus operators by using dynamic discrete choice models with panel data. Empirical data was collected via a panel questionnaire survey conducted among students from a major university in Taiwan. Factor analysis indentified three latent factors, namely on board amenities, bus station service, and operational performance. The result of the panel multinomial logit model shows that bus fare, latent service factors, and personal income are important determinants influencing the selection of bus operators. The panel nested logit models further indicate that distinct competitive structures exist among bus operators due to common unobserved components manifesting in operation characteristics. The panel mixed nested logit model that accommodates correlation among alternatives and parameter heterogeneity is our preferred model. The results can provide insights for bus operators to develop operational and marketing strategies.