This paper discusses the differences between Multinomial Logit, Random Coefficients Mixed Logit and Latent Class Choice models in terms of taste heterogeneity, correlation between coefficients and elasticities. Formulae for correlation and elasticities are presented for the Latent Class Choice model, and an empirical analysis is conducted to highlight the differences between the models on real world data. The paper highlights a number of advantages of the Latent Class Choice model, notably in terms of linking taste heterogeneity, inter-coefficient correlation and elasticities to socio-demographic attributes.

Latent class model structures present a flexible extension of standard choice modelling approaches in the context of the representation of taste heterogeneity. A latent class model divides the population into a number of classes with differences in sensitivities across classes. With the class membership being treated as a latent component, a probabilistic class-allocation model is used, meaning that an individual has a non-zero probability of belonging to each of the different classes. The class-membership probabilities are a function of socio-demographic attributes of the decision-maker, meaning that any taste heterogeneity across respondents can be linked directly to variations in socio-demographic attributes across individuals. This can be a major asset in the interpretation of model results as well as in forecasting. While to some extent taste heterogeneity can also be linked to socio-demographic indicators in a more standard modelling framework, the degree of flexibility (in terms of the extent of heterogeneity) is generally quite limited, and the specification search can be a very tedious task. Despite these very appealing characteristics, latent class approaches are used relatively rarely in the field of travel behaviour research, and have been largely upstaged by the increasing popularity of the Mixed Logit model. However, it should be said that latent class models have very significant advantages in interpretation over the Mixed Logit model. Indeed, being able to link taste heterogeneity to sociodemographic indicators is clearly preferable to simply knowing that a given sensitivity follows a certain (assumed) random distribution in the sample population. This paper builds on ongoing work by the authors in further developing latent class methodology and highlighting potential advantages over continuous mixture models in terms of flexibility as well as interpretation. We first illustrate once again how the marginal utilities (and by extension willingness-topay indicators such as the valuation of travel time savings) are a direct function of socio-demographic attributes used in the class-allocation model and show how latent class models make less restrictive shape assumptions than is the case with most Mixed Logit specifications. As a first contribution, we then derive formulae for the correlation between individual taste coefficients in latent class structures and show how this correlation similarly is a function of the socio-demographic attributes used in the classallocation model. From this, an analyst can for example conclude that for specific subgroups in the population, the time and cost coefficients are negatively correlated, while for others, the correlation may be positive. This is a crucial advantage over the Mixed Logit model, which only produces a fixed measure of the correlation between two randomly distributed coefficients. As a next step, we show that the same principle applies to the elasticities in latent class models, where these can again be expressed as a function of the socio-demographic attributes used in the class-allocation model. In the applied part of the paper, we make use of stated choice data for departure time and travel mode collected in the West Midlands region of the United Kingdom in 2003. Our results shows that the latent class model obtains significant gains over the basic MNL model, where these gains in log-likelihood are comparable to those obtained by a Mixed Logit model with a fully specified covariance structure between random coefficients. However, the real advantages come at the interpretation stage. Here, our analysis illustrates the relationship between socio-demographic indicators and the covariance structure and elasticities in a latent class model. This information is not only useful for the analysis of taste heterogeneity but can also provide significant advantages in forecasting. Additionally, the results show significant differences in the covariance results between the latent class and Mixed Logit models.

Latent class model structures present a flexible extension of standard choice modelling approaches in the context of the representation of taste heterogeneity. In this study evidence is provided of the flexibility of the latent class approach and the potential advantages over other model structures. As a first contribution, formulae are derived for the correlation between individual taste coefficients in latent class structures and show how this correlation is a function of the socio-demographic attributes used in the class-allocation model. From this, an analyst can for example conclude that for specific subgroups in the population, the time and cost coefficients are negatively correlated, while for others, the correlation may be positive. This is a crucial advantage over the Mixed Logit model, which only produces a fixed measure of the correlation between two randomly distributed coefficients. As a next step, it is shown that the same principle applies to the elasticities in latent class models, where these can again be expressed as a function of the socio-demographic attributes used in the class-allocation model. In the applied part of the study, use is made of stated choice data for departure time and travel mode collected for the Dutch National model. The results showed that the latent class model obtains significant gains over the basic MNL model, where these gains in log-likelihood are comparable to those obtained by a Mixed Logit model with a fully specified covariance structure between random coefficients. However, the real advantages come at the interpretation stage. Here, the analysis illustrates the relationship between socio-demographic indicators and the covariance structure and elasticities in a latent class model. This information is not only useful for the analysis of taste heterogeneity but can also provide significant advantages in forecasting. Additionally, the results show significant differences in the covariance results between the latent class and Mixed Logit models. As an example, while the Mixed Logit model gives a correlation of 0.41 between the travel time coefficients for car and train, in the latent class models, this correlation ranges from 0.43 to 0.9 depending on socio-demographic characteristics. For the covering abstract see ITRD E145999

A comprehensive travel demand modeling framework is proposed to identify and model the urban development decisions of firms and the mobility, activity, and travel decisions of individuals and households. The framework is also used to develop a system of models that can be used by decisionmakers and planners to evaluate the effects of developments in information technologies and changes in the transportation system. The implementation of an operational model system based on this framework is envisioned as an incremental process starting with the current best practice of disaggregate travel demand model systems. To this end, an activity-based model system is presented as the first stage in the development of an operational model system.

This paper presents a general methodology and framework for including latent variables—in particular, attitudes and perceptions—in choice models. This is something that has long been deemed necessary by behavioral researchers, but is often either ignored in statistical models, introduced in less than optimal ways (e.g., sequential estimation of a latent variable model then a choice model, which produces inconsistent estimates), or introduced for a narrowly defined model structure. The paper is focused on the use of psychometric data to explicitly model attitudes and perceptions and their influences on choices. The methodology requires the estimation of an integrated multi-equation model consisting of a discrete choice model and the latent variable model’s structural and measurement equations. The integrated model is estimated simultaneously using a maximum likelihood estimator, in which the likelihood function includes complex multi-dimensional integrals. The methodology is applicable to any situation in which one is modeling choice behavior (with any type and combination of choice data) where (1) there are important latent variables that are hypothesized to influence the choice and (2) there exist indicators (e.g., responses to survey questions) for the latent variables. Three applications of the methodology provide examples and demonstrate the flexibility of the approach, the resulting gain in explanatory power, and the improved specification of discrete choice models.

We discuss the development of predictive choice models that go beyond the random utility model in its narrowest formulation. Such approaches incorporate several elements of cognitive process that have been identified as important to the choice process, including strong dependence on history and context, perception formation, and latent constraints. A flexible and practical hybrid choice model is presented that integrates many types of discrete choice modeling methods, draws on different types of data, and allows for flexible disturbances and explicit modeling of latent psychological explanatory variables, heterogeneity, and latent segmentation. Both progress and challenges related to the development of the hybrid choice model are presented.

This paper presents conceptual and methodogical frameworks for inclusion of latent factors as explanatory variables in choice models. The method described provides for explicit treatment of the psychological factors affecting the decisionmaking process by modeling them as latent variables. Psychometric data, such as responses to attitudinal and perceptual survey questions, are used as indicators of the latent psychological factors. The resulting approach integrates choice models with latent variable models, in which the system of equations is estimated at the same time. This simultaneous estimation of the model structure represents an improvement over sequential methods, as it produces consistent and efficient estimates of the parameters. Three applications of the methodology give examples and demonstrate the flexibility of the approach, the resulting gain in explanatory power, and the improved specification of discrete choice models.

We review the case against the standard model of rational behavior and discuss the consequences of various ‘anomalies’ of preference elicitation. A general theoretical framework that attempts to disentangle the various psychological elements in the decision-making process is presented. We then present a rigorous and general methodology to model the theoretical framework, explicitly incorporating psychological factors and their influences on choices. This theme has long been deemed necessary by behavioral researchers, but is often ignored in demand models. The methodology requires the estimation of an integrated multi-equation model consisting of a discrete choice model and the latent variable model system. We conclude with a research agenda to bring the theoretical framework into fruition.

The issue of travelers' adoption of an advanced traveler information system (ATIS) and willingness to pay for such information services is addressed. A case study is presented of SmarTraveler, an ATIS that provides, via telephone, real-time location-specific traffic and transit information in the greater Boston area. The model is an integrated system of discrete choice and latent variable models. It predicts travelers' frequency of use and subscription under varying pricing scenarios. Two models are presented: one for current SmarTraveler users, and one for nonusers. The SmarTraveler usage rate is modeled as a function of payment method and pricing, travelers' travel and socioeconomic characteristics, and their attitudes and perceptions toward ATIS. The data used in model estimation included willingness-to-pay scenarios involving two methods of payment: a flat monthly fee, and a charge per call. It was found that for nonusers, the higher the expected benefit from an ATIS, the higher the willingness to pay. This expected benefit is a latent variable indicated via the importance placed by individuals on ATIS attributes such as reliability, relevance, and coverage. For users, the utility of SmarTraveler is affected strongly by the users' level of satisfaction with the service. A modeling framework is developed that captures response biases and presents figures on willingness to pay for an innovative ATIS actually implemented in the market.

This paper introduces new forms, sampling and estimation approaches fordiscrete choice models. The new models include behavioral specifications oflatent class choice models, multinomial probit, hybrid logit, andnon-parametric methods. Recent contributions also include new specializedchoice based sample designs that permit greater efficiency in datacollection. Finally, the paper describes recent developments in the use ofsimulation methods for model estimation. These developments are designed toallow the applications of discrete choice models to a wider variety ofdiscrete choice problems.

This paper develops a model system for assessing market penetration and usage rates of Advanced Traveler Information Systems (ATIS). The choice models developed predict travelers’ awareness, trial use, and repeat use for SmarTraveler, an ATIS implemented in the Boston area. The travelers’ attitudes toward travel information and perceptions of ATIS service attributes, such as quality and relative advantage over conventional information sources, are incorporated in the modeling framework. This research contributes to the literature on modeling awareness, and use decisions of ATIS using an integrated system of discrete choice and latent variable models.

The emergence of new information technologies and recent advances in existing technologies have provided new dimensions for travel demand decisions. In this paper we propose a comprehensive travel demand modeling framework to identify and model the urban development decisions of firms and developers and the mobility, activity and travel decisions of individuals and households, and to develop a system of models that can be used by decision makers and planners to evaluate the effects of changes in the transportation system and development of information technologies (e.g. various tele-commuting, tele-services and Intelligent Transportation Systems).

An infrastructure-performance-deterioration model predicts the performance of infrastructure facilities such as bridges, railroad, and highways as a function of explanatory variables such as inherent infrastructure characteristics (material properties, construction quality), ambient climate, usage of the facility, etc. However, there is no unambiguous approach to measuring directly the performance of the facility, and hence we consider performance to be unobservable (latent). The problem of developing performance-deterioration models includes the definition of the aforementioned unobservable performance in terms of the measurable distress measures of the facility, and simultaneously relating the performance to the explanatory variables. In this paper, we extend previous research to include user costs (costs accruing to the users of the infrastructure facility) in the modeling framework. Hence, an integrated performance and user-cost model system is developed, and a case study is conducted on a highway example using data from Brazil compiled by the World Bank. Although, the case study is on highways, the methodology is general and applicable to any deteriorating facility with measurable distress measures and explanatory variables.

A highway performance prediction model predicts the performance of highway pavements as a function of explanatory variables such as pavement characteristics, ambient climate, usage of the system, and so on. However, there is no unambiguous approach that can be used to directly measure the performance of the highway pavement. Performance is considered to be unobservable (latent). The problems with developing performance deterioration models include the definition of the aforementioned unobservable performance in terms of the observed or measurable distress measures of the system and simultaneously relating the performance to the explanatory variables. Previous research is extended by exploring the existence of a two-component performance measure for highway pavements: a latent variable to represent functional performance and another variable to represent the structural integrity or structural performance of the pavement. A case study is conducted on a data set from Brazil compiled by the World Bank.

The assumptions about the form and type of the random utility components in a discrete choice model will have a tremendous impact on both the estimability and predictive power of a given model. Early research in discrete choice modeling saw an overwhelming use of multinomial logit and nested logit error structures, due to the closed -form expressions available for calculating probabilities. However, increasing computational power and use of simulation techniques has led to the growing popularity of mixed multinomial logit, which allow for an almost limitless amount of complexity and variety in specifying underlying consumer behavior. The choice modeler thus has competing alternatives for capturing subst itution patterns. A closed-form GEV model (multinomial logit, nested logit, cross -nested logit) may be easier and faster to estimate, while a mixed logit formulation, which will be more complex and computationally intensive, may be able to more accurately capture underlying choice behavior. In this paper, we examine the problem of behaviorally analogous yet statistically different discrete choice models and the degree to which simpler model forms are able to capture more complex behaviors. We present criteria for evaluating alternative model str uctures in different situations and t hen use this framework to analyze simulated data of a five alternative mode choice setting . Results from this analysis suggest that if random parameters are insignificant , closed form models are more