A Revealed Preference Time of Day Model for Departure Time of Delivery Trucks in the Netherlands

Abstract

This paper presents one of the first discrete choice models regarding truck departure time for road freight transport using revealed preference electronic trace data. The data containing 1447 logistical routes was obtained from GPS units in trucks of a large European retailer in June 2016. Detailed historical link speed data of the road network is used to compute trip durations for non-chosen departure times for each specific recorded route. A baseline multinomial logit model is initially estimated solely based upon trip duration as an explanatory variable. Next, product type is added as an observed variable, improving the multinomial logit model fit. Finally, the model is flexibly extended to incorporate unobserved heterogeneity by nesting departure time alternatives into time blocks for morning, afternoon, and night, further improving the model fit.