A Time-period Choice Model for the Strategic Flemish Freight Model Based on Stated Preference Data

A Time-period Choice Model for the Strategic Flemish Freight Model Based on Stated Preference Data


Gerard De Jong, Significance, Marco Kouwenhoven, Significance, Kim Ruijs, Significance


A stated preference survey among receivers of goods has been carried out and is now used to estimate a time-period choice model which will be implemented in the Strategic Flemish Freight Model


A time-period choice model for the Strategic Flemish Freight Model based on stated preference data.

Gerard de Jong, Marco
Kouwenhoven en Kim Ruijs – Significance
Pieter van Houwe – MINT
Dana Borremans – Verkeerscentrum, Departement MOW, Flemish authorities

In transport model systems that are used in practice for forecasting and project appraisal, time period choice is usually missing. However, there is evidence, especially in passenger transport, that departure time choice is rather sensitive to changes in transport time and costs.
Transport models that do include a time-period choice are usually passenger models, such as the Dutch National Model system (LMS). In freight transport, time-of-day choice models are almost non-existent.
The current Strategic Flemish Freight Model version 1.6 (SVV) of the Flemish Traffic Centre does not contain an explicit time-period choice model either. But a new version (version 4.1) is being developed. In this version, a module will be implemented that can determine how many road freight vehicles will depart earlier/later in response to increasing transport times (i.e. congestion) and/or increasing transport costs (e.g. road user charging that is differentiated by time-of-day). This paper describes the development of this new time-period choice module of the SVV.
It is very hard to obtain revealed preference (RP) data on transport time and cost by time period of the day: these variables are difficult to measure directly, and transport time and transport cost are highly correlated. Furthermore, the transport costs vary only little over time periods since there are few areas that have road user charges that vary with time-of-day period. Therefore, we have based the time-period choice model on stated preference (SP) data. In the SP interviews we focussed on the receivers of goods (consignees). Industry experts and the (limited) scientific literature tell us that they usually determine the delivery windows of the goods, and that carriers are bound by the choices that the senders and receivers make.
Firms in Flanders receiving goods by road transport were selected from company registers and called by phone to check whether they are in scope and to ask them to participate in the SP survey. The stated preference interview itself was done by computer assisted personal interviewing (CAPI). About 25 pilot interviews were carried out, followed by a main survey of 150 firms. These were stratified by type of firm (manufacturers, wholesalers/warehouses and retailers) and by transport distance class for the typical transport that serves as the context and reference situation for the SP experiment.
Since we are interested in shifts away from the peak, if sufficed to sample shipments that are currently transported in the (morning or evening) peak. So in the interview we asked the respondents to describe a recent road-based shipment that was transported (at least partly) during a peak period and in the SP experiments they were asked to choose between two (hypothetical) alternative transports for this shipment. Each transport is described by the following characteristics:
• Transport time
• Transport cost
• The start and end of the delivery time window: this is the timeframe within the receiver wants the shipment to arrive at its final destination.

In the statistical design these presented attribute values are derived from four attributes: transport time, transport cost, width of the delivery time window and midpoint of the delivery time window.
The SP data have been used to estimate discrete choice models that explain the trade-offs between transport time, cost and earlier/later transports. This gives the basis for the time-period choice module that is implemented in the SVV. The implemented time-period choice model does not include specific time windows, but it gives time shifts for transports that depend on travel time and costs per period. It is made consistent with observed shares for the time periods and only produces changes relative to the base distribution: the module is applied in a pivot-point fashion. For the base year, a table for the observed distribution of transport over time periods is available.
In the paper we will present the existing literature on time-period choice models in freight transport, describe the questionnaire used and the SP experiment on time-period choice in freight in detail, report on the outcomes of the survey and the estimation results for the discrete choice models and present the module as implemented in the SVV version 4.1.


Association for European Transport