Capturing interdependencies in tour mode and activity choice : a co-evolutionary logit modelling approach

Abstract

Paper presented at the 23rd Annual Southern African Transport Conference 12 - 15 July 2004 "Getting recognition for the importance of transport", CSIR International Convention Centre, Pretoria, South Africa. The paper aims to analyse the interdependencies in work-tour choice facets, specifically mode and activity choice. Activities may be inserted before, in-between and/or after the work activity resulting in the formation of complex work-tours. Traditional modelling approaches assume that tour-decisions are being made simultaneously or in some predefined order. Both these assumptions have inherent shortcomings such as defining a discrete set of choice alternatives or wrong estimation of parameters in the case of hierarchical estimation. To address the questions of interdependent tour-choice facets, the paper proposes the co-evolutionary methodology. The methodology holds implications for both the estimation and prediction phase of modelling. Separate utility models are estimated for each choice facet with the other choice facets used as independent variables. Estimated parameters thus represent the influence of the other choice facets. Prediction involves interactively updating predicted possibilities until a pre-defined convergence is reached (which solves the problem of circularity between linked decisions). Under the assumptions that individuals make least uncertain decisions first, the methodology provides for clarification on the order of decisions. The empirical analysis uses detail, disaggregate travel-activity dairy collected in the Amsterdam region, The Netherlands, collected as part of a study into activity-travel patterns with public transport, undertaken for the Dutch Government. The results reveal that mode choice is significantly influenced by intermediate activities while intermediate activities are less influenced by mode choice. Also, before, in-between and after intermediate activities correlate with distinctly different transport, land use and socio-demographic characteristics. Considering the order of decisions, it was found that, in the majority of cases, intermediate activity choice rank higher up in the decision hierarchy while transport mode ranks rather low. The finding lends support to the hypothesis that intermediate activities might not be as discretionary as sometimes believed and that mode choice is determined, in most cases, by activity choice and not vice versa. The findings of the research, while using data from the Netherlands, are potentially relevant to South Africa and hold implications for data collection, model specification and, ultimately, transport policy. South African travel surveys are mostly of the Origin-Destination which focuses on the separate trips with little activity or tour information collected. As a result, model specification does not incorporate activity or tour decisions. Given the importance of tours in structuring daily activity-travel behaviour, this might lead to unrealistic assumptions and invalid policies about travel behaviour, in specific mode choice and trip generation.