| dc.description.abstract | Random utility maximization (RUM)-based discrete choice models are widely used for analyzing individual choice behaviour across diverse fields, including transportation, economics, market research, healthcare, and environmental science. The random components of the utility functions used in these models comprise both random noise and systematic errors. The systematic errors, in turn, comprise the following sources of variability or stochasticity: (a) unobserved heterogeneity in tastes or preferences of the decision-makers, (b) analyst’s errors in measurement of the choice alternative attributes and choice environment, (c) inherent stochasticity in alternative attributes and choice environment, (d) omission of influential variables from the model specification, and (e) other errors in model assumptions and specification. It is useful to disentangle these different types of systematic errors from the random noise. Doing so helps reduce estimation bias, offers a deeper understanding of individual’s choice behaviour, and results in more credible policy insights. The choice modelling field has witnessed several methodological advances on characterizing and disentangling these systematic errors from the random noise. However, the simultaneous identification of these different sources of stochasticity has been hitherto under-explored.
This dissertation seeks to formulate and apply novel discrete choice modelling frameworks that simultaneously account for various sources of stochasticity in travel choice models. Specifically, the following new models are formulated and applied to understand traveller choice behaviour, drawing on empirical data from India and the United States of America (USA):
1. An integrated choice and stochastic variable modelling framework with random coefficients that allows the analyst to simultaneously accommodate stochasticity in alternative attributes (to represent analyst’s measurement errors in those attributes) and random coefficients on such attributes. We use this framework to show that ignoring either source of stochasticity – stochasticity in alternative attributes or unobserved heterogeneity in response to the attributes – results in models with inferior goodness-of-fit and a systematic bias in all parameter estimates. This is demonstrated using simulation experiments for two different travel choice settings, one involving labelled mode choice alternatives and the other involving unlabelled route choice alternatives. The mode choice simulations reflect commute travel in Bengaluru, India and the route choice simulations reflect freight truck movement in Florida, USA. In addition, an empirical analysis is presented in the context of route choice of trucks using empirical data from Florida, to highlight the importance of accommodating both sources of variability – stochasticity in travel times and random heterogeneity in response to travel times.
2. A choice modelling framework that can be used to simultaneously infer alternative attributes and the corresponding coefficients, as well as stochasticity in both – without the help of external measurement data on alternative attributes – using mixed logit models on pooled revealed preference (RP) and stated preference (SP) choice datasets. The hypothesis is that SP data, which is typically free of variability in alternative attributes (since the attribute values are carefully constructed and presented to the respondent or decision-maker), allows the identification of heterogeneity in the coefficient on alternative attributes. As the response heterogeneity is identified from SP data, the RP attributes (and the variability therein) may be identified using RP choice data collected from the field. To test this hypothesis, first, a theoretical examination is performed to examine the feasibility and conditions of parameter identification for different specifications and distributional forms of alternative attributes and the corresponding coefficients. Next, simulation experiments are conducted to examine the efficacy of the proposed approach in identifying stochasticity on any one of the alternative attributes (for multiple alternatives) and random coefficients on that and any other alternative attributes. In addition, an empirical application is presented in the context of commute mode choice in Bengaluru, India, to demonstrate the importance of recognizing stochasticity in mode-specific in-vehicle travel times along with the random coefficient on in-vehicle travel times. It is worth noting here that in the context of pooled RP-SP datasets, no external source of measurement data (for example, measurement data for the stochastic variable) is required to identify the distribution of the stochastic variable.
3. A choice modelling framework for pooled revealed preference and stated preference (RP-SP) data to examine the influence of the interaction between crowding levels and stochastic in-vehicle travel times (IVTT) on travel mode choice. Using this framework, simulation experiments are conducted to examine the bias in parameter estimates and distortion in the willingness to pay measures if travellers’ sensitivity to transit travel time is not specified as dependent on crowding levels in transit systems. In addition, an empirical application to commute mode choice in Bengaluru, India, is presented which highlights the influence of crowding levels (and its interaction with travel time that is specified as stochastic) on travellers’ choice of public transit modes.
Methodological advancements in discrete choice modelling have accelerated empirical research on travel behaviour in many parts of the world. However, the progress in travel behaviour research has not been uniform across the world, with many countries in the global south lacking adequate empirical data on travel behaviour. There is a need for more empirical data and studies on travel behaviour to aid transport planning and decision-making in these cities. This dissertation attempts to fill such substantive gaps through a RP-SP mode choice data collection effort in Bengaluru, India. Moreover, the empirical mode choice models developed in this dissertation contribute to the literature on empirical analysis of travel behaviour in Indian cities. | en_US |
