Modeling household interactions: a mixed model approach

Abstract

From an international research perspective, an activity-based view on transportation has become standard today. The aim of activity-based models is to predict which activities will be conducted where, when, for how long, with whom and with which transport mode. The required information for these models is gathered by means of the activity-diary surveys. These diaries provide information at the individual level, but next to that, also a household survey needs to be filled out. One should therefore acknowledge that both levels can play an important role in trying to predict each of the responses stated above. Very often, though, only household variables are taken into account in an attempt to model a response at an individual level, without actually accounting for the clustering that is present. Another option that is pursued sometimes is to account for this information at different levels by posing constraints on the underlying model. This paper gives a better solution by incorporating the correlation between the household and the individual level directly by means of mixed models. The mixed model strategy is applied to two important responses of the Flemish travel survey of 2000, i.e. daily travel time and daily travel distance. The employment status, gender and age are shown to be important individual characteristics, while the number of children and the household income are significant household variables. Moreover, the results show that up to 30% of the variation in travel time and/or distance can be attributed due to the fact that people live in households, so this may not be ignored any longer.