Analysing the degree of mode captivity in a multi-modal travel behaviour using stated preference data

Various mode choice models have been developed in the past, using the SP data, in order to forecast the mode shares under the hypothetical travel environment. However, little has been done to simultaneously analyse the influence of mode captive users, in order to determine their relative influence on the forecasted modal split for the study area. This paper presents a framework developed to analyse the degree of mode captivity in a multi-modal travel behaviour of the survey respondents, and to determine the influence of these users on the forecasted mode shares.\ud \ud Computer-based stated preference (SP) surveys were conducted in Southern Redland Shire, Queensland, presenting the respondents with eight randomly generated SP mode choice games. Based on these responses, an individual was determined to be a mode captive or mode choice user. The set of mode captive users was further split into car captive and PT captive users. \ud \ud Out of the 2007 respondents surveyed, approximately 60% were determined to be car captive users; i.e. not perceiving to switch to any travelling alternative of car, shown to them in the SP survey. Nested logit models were then estimated, using the mode choice data only, for four trip purposes of work, shopping, education and other trips. The trip purposes were further categorised on the basis of two trip lengths, regional and local trips, resulting in the estimation of eight unique mode choice models to forecast the mode shares of the targeted population. The model specification developed for the mode choice module comprised of the hypothetical travelling alternatives to car, namely bus on busway, walk on walkway and cycle to cycleway. The bus on busway mode further associated a set of five transit access modes of feeder bus, walking and cycling to busway, park and ride, and kiss and ride.\ud \ud For analysing the degree of car captivity in the travel behaviour of the region, multinomial logistic regression equations were developed, based on the three socio-demographic characteristics of household size, number of vehicles per household and age-group of the individuals, along with the level-of-service parameter of trip length. Logistic regression runs were conducted for each trip purpose to determine the significant parameters influencing the three possible outcomes of an individual; being a mode choice, car captive or PT captive user. \ud \ud The analysis showed that the attribute of number of vehicles per household served as the driving determinant for the traveller type outcome of an individual for each trip purpose. With a unit increase in the number of vehicles in the household, the likelihood of the traveller being a mode choice or PT captive user was found to reduce substantially for each trip purpose. Various probability functions were also tested with varying values of parameters, in order to observe the possible changes in the outcomes of traveller type