Comparing a Conventional Travel Demand Model to an Activity-based Travel Demand Model: a Case Study of Copenhagen

Comparing a Conventional Travel Demand Model to an Activity-based Travel Demand Model: a Case Study of Copenhagen


Eric Petersen, RAND Europe, NL; Goran Vuk, Danish Transport Research Institute, DK


This paper will offer some direct comparisons of an activity-based model and a conventional four-step model based on the same data source, namely data from Denmark?s national travel behaviour survey data for the period 2000-2005.


Over the past decade, there has been significant movement in bringing advanced activity-based models to fruition. This activity has taken place in North America (San Francisco, New York, Portland, OR, Columbus, OH and Toronto) and in Europe (Netherlands, Switzerland and Stockholm). While activity-based models offer a much more behaviourally-realistic approach to modelling travel, there has been some reluctance on the part of regional planning agencies to move away from conventional four-step models and embrace activity-based models. Some practitioners have pointed to the lack of direct comparisons between four-step models and activity-based models as one barrier to the wider acceptance of activity-based models. This paper will offer some direct comparisons of the two types of models, based on the same data source, namely data from Denmark?s national travel behaviour survey data (the TU dataset) for the period 2000-2005.

The ├śrestad Traffic Model (OTM) is a well-established model for the Greater Copenhagen Area. It is currently in the process of being revalidated with updated TU data, i.e. the new 2004 base tour matrices will be built and the demand model will be re-estimated. A parallel modelling process is in place to estimate an activity-based model called Copenhagen Model for Passenger Activity Scheduling (COMPAS). The process will lead to a gradual transition from OTM to COMPAS. The planning reasons for shifting to COMPAS are: better understanding of consequences of introducing road pricing in Copenhagen as well as congestion problems on the access roads to the city. The tour generation model components will be the first to be estimated and the first COMPAS model components to replace the OTM model components. Thus, this paper will focus on comparing the tour generation models between a conventional and activity-based approach.

A pure comparison between the two types of models is not meaningful, since activity-based models are structured differently. Activity-based models are typically tour-based and have far fewer non-home based tours than conventional models (the majority of which are trip based). In many instances, we need a methodology to compare total tours out of an activity-based model to the trips out of a conventional model and perhaps more importantly to compare total passenger miles by mode. Of course, the comparison is more direct if the conventional model is tour-based. OTM is a tour-based model, which does assist us in making comparisons between the models. Nonetheless, it is still possible that tours are categorized or defined differently. For instance, many activity-based models have an explicit escorting purpose, which would more typically be grouped together with other maintenance tours in a conventional, tour-based model. Thus, the raw number of tours by different purposes will probably not coincide completely between models.

Important comparisons that can be made between the models are looking at the sensitivities of travellers to changes in time and cost. If these differ between models and they might, the modeller needs to be able to explain how the structures of the model influence these values. This paper will set out a number of structural differences between OTM and COMPAS and track through the model how these differences affect estimation results.

The other important comparison that should be made between models is to show how they can be used to inform policy makers. In general, this is considered a strength of activity-based models, since they are more behaviourally-detailed. Activity-based models are built upon a better understanding of intra-household interactions and interdependencies that lead to joint activities and travel. In practical terms, the joint travel feature which can be built into activity-based models is critical for proper modelling of congestion pricing schemes. The nature of activity-based models also allows for more detailed investigations of policy options, such as varying parking charges by time of day. However, one major issue is that policy makers have come to expect certain output from travel models. They may not be equipped to incorporate new kinds of information or the information they seek may not be easily extracted from the model output. The final comparison between OTM and COMPAS will be to evaluate its ease of use for decision makers, particularly in these non-traditional areas.


Association for European Transport