Using a synthetic social network to improve leisure destination choice simulation

Abstract

In developed countries, in the last years, a continuous increase of the share of trips that are performed for leisure purposes could be observed. Various data sources further indicate that the most important motivation behind out-of-home leisure activities is social contact. On the other side, leisure remains very hard to model in simulation models, mainly because it depends highly on characteristics that are difficult to observe, such as heterogeneity of taste and the characteristics of leisure locations. As such, the best such models can do is to calibrate the level of noise added on top of classical distance and cost based utility of travel, such that simulated travel distances fit the data adequately. This paper makes a first step in using the social nature of leisure travel to decrease the level of noise needed to simulate leisure travel. Using a realistic synthetic population with a realistic social network for Switzerland, it implements a simple model where agents have individual preferences over activity locations and social contacts, and shows how this can help reproduce the observed travel distances. In particular, the traveled distances for visiting social contacts, one of the most important leisure types in Switzerland, are pretty insensitive to the scale of the error terms and come out of the structure of the social network.