Work Departure Time Analysis Using Dogit Ordered Generalized Extreme Value Model

This paper applies a discrete choice modeling technique to investigate departure time choice for individual workers residing in the New York City metropolitan area. This effort is distinguished from previous departure time models primarily by the use of the dogit ordered generalized extreme value (DOGEV) model rather than the commonly used multinomial logit (MNL) model. The MNL model is restrictive in cases in which the continuous time needs to be discretized into departure time intervals. Because the MNL model treats departure time intervals as independent alternatives, the model cannot account for the ordering of the time intervals and their correlation. In contrast, the DOGEV model has two distinct features. First, it recognizes the ordinal nature of the departure time intervals by allowing them to be correlated in vicinity (i.e., time intervals that are close to each other in the ordering have error terms that are correlated). Second, it allows a worker's departure time choice to be captive or constrained to a particular departure time interval. The modeling approach was based on a behavioral analysis that explained the factors influencing work departure time decisions in a highly urbanized environment. The results of the model estimation provide valuable insights into the effects, on a worker's departure time choice, of socio-economic characteristics, employment characteristics, travel-related attributes, and land use and location attributes. Empirical application also shows that New York metropolitan travel survey data were well modeled by the DOGEV with both significant captivity and ordering components. In particular, evidence of ordering and proximate covariance in the choice set may suggest an additional source of misspecification in the existing departure time literature, which has focused largely on unordered discrete choice models, such as multinomial logit and nested logit.