Making use of respondent reported processing information to understand attribute importance: a latent variable scaling approach

In recent years we have seen an explosion of research seeking to understand the role that rules and heuristics might play in improving the predictive capability of discrete choice models, as well as delivering willingness to pay estimates for specific attributes that may (and often do) differ significantly from estimates based on a model specification that assumes all attributes are relevant. This paper adds to that literature in one important way—it explicitly recognises the endogeneity issues raised by typical attribute non-attendance treatments and conditions attribute parameters on underlying unobserved attribute importance ratings. We develop a hybrid model system involving attribute processing and outcome choice models in which latent variables are introduced as explanatory variables in both parts of the model, explaining the answers to attribute processing questions and explaining heterogeneity in marginal sensitivities in the choice model. The resulting empirical model explains how lower latent attribute importance leads to a higher probability of indicating that an attribute was ignored or that it was ranked as less important, as well as increasing the probability of a reduced value for the associated marginal utility coefficient in the choice model. The model does so by treating the answers to information processing questions as dependent rather than explanatory variables, hence avoiding potential risk of endogeneity bias and measurement error.     

Correcting for endogeneity in behavioral choice models with social influence variables

While psychologists and behavioral economists emphasize the importance of social influences, an outstanding issue is how to capture such influences in behavioral models used to inform urban planning and policy. In this paper we focus on operational models that do not require explicit knowledge of the individual networks of decision makers. We employ a field effect variable to capture social influences, which is calculated as the percent of population in the peer group that has chosen the specific alternative. We define the peer group based on socio-economic status and spatial proximity of residential location. As in behavioral economics and psychology, the concept is that one is influenced by the choices made by one’s peers. However, using such a social influence variable in a behavioral model causes complications because it is likely endogenous; unobserved factors that impact the peer group also influence the decision maker, yielding correlation between the field effect variable and the error. The contribution of this paper is the use of the Berry, Levinsohn, and Pakes (BLP) method to correct the endogeneity in a choice model. The two-stage BLP introduces constants for each peer group to remove the endogeneity from the choice model (where it is difficult to deal with) and insert it into a linear regression model (where endogeneity is relatively easier to deal with). We test the method using a mode choice data set from the Netherlands and readily available software and find there is an upward bias of the field effect parameter when endogeneity is not corrected. The procedure outlined presents a practical and tractable method for incorporating social influences in choice models.     

Forecasting light-duty vehicle demand using alternative-specific constants for endogeneity correction versus calibration

We investigate parameter recovery and forecast accuracy implications of incorporating alternative-specific constants (ASCs) in the utility functions of vehicle choice models. We compare two methods of incorporating ASCs: (1) a maximum likelihood estimator that computes ASCs post-hoc as calibration constants (MLE-C) and (2) a generalized method of moments estimator that uses instrumental variables (GMM-IV) to correct for price endogeneity. In a synthetic study we observe significant coefficient bias with MLE-C when the price-ASC correlation (endogeneity) is large. GMM-IV successfully mitigates this bias given valid instruments but exacerbates the bias given invalid instruments. Despite greater coefficient bias, MLE-C yields better forecasts than GMM-IV with valid instruments in most of the cases examined, including most cases where the price-ASC correlation present in the estimation data is absent in the prediction data. In a market study of U.S. midsize sedan sales from 2002 – 2006 the GMM-IV model predicts the 1-year-forward market better, but the MLE-C model predicts the 5-year-forward market better. Including an ASC in predictions by any of the methods proposed improves share forecasts, and assuming that the ASC of each new vehicle matches that of its closest competitor vehicle yields the best long term forecasts. We find evidence that the instruments most frequently used in the automotive demand literature may be invalid.     

Comparison of updating techniques in transferability analysis of work trip mode choice models in developing countries

This study analyzes the performances of updating techniques in transferability of mode choice models in developing countries. A model specification, estimated in Ho Chi Minh City, was transferred to Phnom Penh. Naïve transfer and four updating methods associated with small sized samples were used in the transfer process and were evaluated based on statistical perspective and predictive ability. The study also illustrates the problems faced in model transferability development, due to the lack of available and suitable data in Phnom Penh. This lack is strongly related to different methods and structures applied in collecting the data. Simplified approaches to the difficulties are proposed in the study. The results show that updating ASCs, updating both ASCs and scale parameter, and use of combined transfer estimators all produce significant improvement, both statistically and in predictability, in updating the model. The last two methods have proven to be superior to the first method, owing to the inclusion of transfer bias considerations in the estimations. However, small data samples should not have large transfer bias when using combined transfer estimators. It is also concluded that naïvely transferring a model is not recommended, and Bayesian updating should be avoided when transfer bias exists. Copyright © 2010 John Wiley & Sons, Ltd.     

Incorporating spatial interactions in zero-inflated negative binomial models for freight trip generation

This paper formulates a spatial autoregressive zero-inflated negative binomial model for freight trip productions and attractions. The model captures the following freight trip characteristics: count data type, positive trip rates, overdispersion, zero-inflation, and spatial autocorrelation. The spatial autoregressive structure is applied in the negative binomial part of the models to obtain unbiased estimates of the effects of different regressors. Further, we estimate parameters using the full information maximum likelihood estimator. We perform empirical analysis with an establishment based freight survey conducted in Chennai. Separate models are estimated for trips generated by motorised two-wheelers and three-wheelers, and pickups besides an aggregate model. Spatial variables such as road density and indicator of geolocation are insignificant in all the models. In contrast, the spatial autocorrelation is significant in all of the models except for the freight trips attracted and produced by pickups. From a policy standpoint, the elasticity results show the importance of considering spatial autocorrelation. We also highlight the bias due to aggregation of vehicle classes, based on the elasticities.

     

Using census aggregates to proxy for household characteristics: an application to vehicle ownership

Traditionally, researchers studying transportation choice have used data either acquired from household surveys or broad, region-wide aggregates. At the disaggregate level, researchers usually do not have access to important variables or observations. This study investigates the potential usefulness of a proxy approach to modeling discrete choice vehicle ownership: substituting narrow area-based aggregate proxies for missing micro-level explanatory variables by accessing large, publicly maintained datasets. We use data from the 2000 Bay Area Travel Survey (BATS) and the contemporaneous U.S. Census file to compare three models of vehicle ownership, drawing area-wide proxies from increasing levels of aggregation. The models with proxies are compared with a parallel model that uses only survey data. The results indicate that the proxy models are preferred in terms of model selection criteria, and predict vehicle ownership as well or better than the survey model. Parameter values produced by the proxy method effectively approximate those returned by household survey models in terms of coefficient sign and significance, particularly when the aggregate variables are representative of their household-level counterparts. The proxy model with the narrowest level of aggregation achieved the best fit, coefficient precision, and percentage of correct prediction.

A comparative study of aggregate and disaggregate gravity models using Seoul metropolitan subway trip data

Abstract

Existing origin constrained and doubly constrained gravity models have not been compared, theoretically or empirically, in terms of their forecasting power. Due to the newly advanced technology of intelligent transport systems, the expanded data presently available have made various models more comparable in terms of forecasting power. This paper uses archived automatic passenger counting (APC) data for urban rail in the Seoul metropolitan area. The APC data contains information about each trip's origin, destination, ticket type, fare, and distance on a daily basis. The objective of this paper is to compare the goodness-of-fit of aggregate and disaggregate gravity modeling using these data. A Hyman aggregate gravity model is used as the aggregate model without the spatial effect. The disaggregate model adopts a multinomial logit as the destination choice model with the spatial effect. In general, while the formulation of aggregate and disaggregate gravity model models are similar, the calibration and parameter estimation methods of the two models are different. As a result, this empirical study demonstrates that the variation in goodness-of-fit and forecasting power largely depends on the estimation method and selected variables. The forecasting power of the disaggregate modeling approach outperforms that of the aggregate model. This paper further confirms that spatial arrangement plays important roles in gravity modeling.     

Sampling of alternatives in Logit Mixture models

Employing a strategy of sampling of alternatives is necessary for various transportation models that have to deal with large choice-sets. In this article, we propose a method to obtain consistent, asymptotically normal and relatively efficient estimators for Logit Mixture models while sampling alternatives. Our method is an extension of previous results for Logit and MEV models. We show that the practical application of the proposed method for Logit Mixture can result in a Naïve approach, in which the kernel is replaced by the usual sampling correction for Logit. We give theoretical support for previous applications of the Naïve approach, showing not only that it yields consistent estimators, but also providing its asymptotic distribution for proper hypothesis testing. We illustrate the proposed method using Monte Carlo experimentation and real data. Results provide further evidence that the Naïve approach is suitable and practical. The article concludes by summarizing the findings of this research, assessing their potential impact, and suggesting extensions of the research in this area.     

Choice of mode of transport for long-distance trips: Solving the problem of sparse data

Transport planning is usually based on models’ forecasts, but the reliability of their outputs depends so much on the quality of input-data they are fed with. Discrete-choice models are used to characterise travellers’ behaviour in choosing their transport mode. Their calibration process is usually based on data stemming from household survey campaigns. However, the modelling in multimodal and intermodal transport on an interurban level is far more complicated and costly than in the case of an urban area. An alternative way to reduce costs is achieved by designing a choice-based sampling strategy where household surveys are replaced by specific surveys for each transport mode. This strategy generates a non-random sample that has to be treated correctly during the estimation process. In principle, the sample does not represent population market quotas for each different transport option. Moreover, as a result of both physical and functional constraints, the survey period cannot cover all origin–destination pairs (O–D pairs) in an optimal way and, consequently, the above-mentioned bias also affects each different individual O–D pair or, at least, group of pairs. In order to overcome this problem, this study presents a new procedure derived from the introduction of maximum likelihood estimators. These estimators assume the original mode options in terms of population quotas and in terms of O–D groups of pairs. The procedure is based on the optimisation of an objective-function to correct the above-mentioned bias in a way similar to the estimators of samples based on different choice options. The method named DWELT estimates the parameters corresponding to each explanatory variable using mode shares for each O–D pair or group of pairs. DWELT has been successfully validated in the case study of the Madrid–Barcelona interurban corridor in Spain. This result allows to achieve a more flexible cheaper survey procedure for interurban transport planning activities. Therefore transport policy strategies could be better designed and tested with lower costs.     

Experience as a conditioning effect on choice: Does it matter whether it is exogenous or endogenous?

Previous choice studies have proposed a way to condition the utility of each alternative in a choice set on experience with the alternatives accumulated over previous periods, defined either as a mode used or not in a most recent trip, or the mode chosen in their most recent trip and the number of similar one-way trips made during the last week. The paper found that the overall statistical performance of the mixed logit model improved significantly, suggesting that this conditioning idea has merit. Experience was treated as an exogenous influence linked to the scale of the random component, and to that extent it captures some amount of the heterogeneity in unobserved effects, purging them of potential endogeneity. The current paper continues to investigate the matter of endogeneity versus exogeneity. The proposed approach implements the control function method through the experience conditioning feature in a choice model. We develop two choice models, both using stated preference data. The paper extends the received contribution in that we allow for the endogenous variable to have an impact on the attributes through a two stage method, called the Multiple Indicator Solution, originally implemented in a different context and for a single (quality) attribute, in which stage two is the popular control function method. In the first stage, the entire utility expression associated with all observed attributes is conditioned on the prior experience with an alternative. Hence, we are capturing possible correlates associated with each and every attribute and not just one selected attribute. We find evidence of potential endogeneity. The purging exercise however, results in both statistical similarities and differences in time and cost choice elasticities and mean estimates of the value of travel time savings. We are able to identify a very practical method to correct for possible endogeneity under experience conditioning that will encourage researchers and practitioners to use such an approach in more advanced non-linear discrete choice models as a matter of routine.

     

On accommodating spatial interactions in a Generalized Heterogeneous Data Model (GHDM) of mixed types of dependent variables

We develop an econometric framework for incorporating spatial dependence in integrated model systems of latent variables and multidimensional mixed data outcomes. The framework combines Bhat's Generalized Heterogeneous Data Model (GHDM) with a spatial (social) formulation to parsimoniously introduce spatial (social) dependencies through latent constructs. The applicability of the spatial GHDM framework is demonstrated through an empirical analysis of spatial dependencies in a multidimensional mixed data bundle comprising a variety of household choices – household commute distance, residential location (density) choice, vehicle ownership, parents’ commute mode choice, and children's school mode choice – along with other measurement variables for two latent constructs – parent's safety concerns about children walking/biking to school and active lifestyle propensity. The GHDM framework identifies an intricate web of causal relationships and endogeneity among the endogenous variables. Furthermore, the spatial (social) version of the GHDM model reveals a high level of spatial (social) dependency in the latent active lifestyle propensity of different households and moderate level of spatial dependency in parents’ safety concerns. Ignoring spatial (social) dependencies in the empirical model results in inferior data fit, potential bias and statistical insignificance of the parameters corresponding to nominal variables, and underestimation of policy impacts.     

The main determinants of the demand for public transport: a comparative analysis of England and France using shrinkage estimators

This study analyses the impacts of changes in fares, service supply, income and other factors on the demand for public transport on the basis of panels of English counties and French urban areas. The analysis is based on dynamic econometric models, so that both short- and long-run elasticities are estimated. Conventional approaches (i.e. fixed- and random-effect models) rely on the hypothesis that elasticities are the same for all areas. Having shown that this hypothesis is not valid for these data sets, the heterogeneity amongst areas is accounted for using a random-coefficients approach, and Bayesian shrinkage estimators.Estimated elasticities for France and England are compared, by using a common set of variables, similar time period and a common methodology. The results show a considerable variation in elasticities among areas within each country. The major conclusion is that public transport demand is relatively sensitive to fare changes, so that policy measures aimed at fare reduction (subsidisation) can play a substantial role in encouraging the use of public transport, thus reducing the use of private cars.     

Estimating GEV models with censored data

We examine the problem of estimating parameters for Generalized Extreme Value (GEV) models when one or more alternatives are censored in the sample data, i.e., all decision makers who choose these censored alternatives are excluded from the sample; however, information about the censored alternatives is still available. This problem is common in marketing and revenue management applications, and is essentially an extreme form of choice-based sampling. We review estimators typically used with GEV models, describe why many of these estimators cannot be used for these censored samples, and present two approaches that can be used to estimate parameters associated with censored alternatives. We detail necessary conditions for the identification of parameters associated exclusively with the utility of censored alternatives. These conditions are derived for single-level nested logit, multi-level nested logit and cross-nested logit models. One of the more surprising results shows that alternative specific constants for multiple censored alternatives that belong to the same nest can still be separately identified in nested logit models. Empirical examples based on simulated datasets demonstrate the large-sample consistency of estimators and provide insights into data requirements needed to estimate these models for finite samples.     

How,when and why integrated choice and latent variable models are latently useful

Integrated Choice and Latent Variable (ICLV) models are an increasingly popular extension to discrete choice models that attempt explicitly to model the cognitive process underlying the formation of any choice. This study was born from the discovery that an ICLV model can in many cases be reduced to a choice model without latent variables that fits the choice data at least as well as the original ICLV model from which it was obtained. The failure of past studies to recognize this fact raised concerns about other benefits that have been claimed with regards to the framework. With the objective of addressing these concerns, this study undertakes a systematic comparison between the ICLV model and an appropriately specified reduced form choice model. We derive analytical proofs regarding the benefits of the framework and use synthetic datasets to corroborate any conclusions drawn from the analytical proofs. We find that the ICLV model can under certain conditions lead to an improvement in the analyst's ability to predict outcomes to the choice data, allow for the identification of structural relationships between observable and latent variables, correct for bias arising from omitted variables and measurement error, reduce the variance of parameter estimates, and abet practice and policy, all in ways that would not be possible using the reduced form choice model. We synthesize these findings into a general process of evaluation that can be used to assess what gains, if any, might be had from developing an ICLV model in a particular empirical context.     

Simultaneous modeling of individuals’ duration and expenditure decisions in out-of-home leisure activities

In this paper, we propose an activity model under time and budget constraints to simultaneously predict the allocation of time and money to out-of-home leisure activities. The proposed framework considers the activity episode level, given that the activity is scheduled. Thus, the model considers the decision of the quantities for duration and expenditure spent during the activity. We use a flexible utility function and show how the simultaneous equations can be estimated by using structural equations model (SEM) estimation techniques to handle the endogeneity problem of time and expenditure. The estimation results are based on a large national leisure diary data set collected in 2008 in the Netherlands, which provides detailed information about time and money spent as well as timing and location attributes of the activities. The analysis reveals that socio-demographics, travel party, timing and location variables influence the duration and expenditure of activity episodes. It shows that various socio-demographic groups display different preferences in terms of the time and money spent on activities. The results also indicate substitution relationships between spending more time and money for various activity categories. Thus it is concluded that the analysis provides useful results for a better understanding of combined time and money allocation decisions for leisure activities.     

A latent segmentation based multiple discrete continuous extreme value model

We examine an alternative method to incorporate potential presence of population heterogeneity within the Multiple Discrete Continuous Extreme Value (MDCEV) model structure. Towards this end, an endogenous segmentation approach is proposed that allocates decision makers probabilistically to various segments as a function of exogenous variables. Within each endogenously determined segment, a segment specific MDCEV model is estimated. This approach provides insights on the various population segments present while evaluating distinct choice regimes for each of these segments. The segmentation approach addresses two concerns: (1) ensures that the parameters are estimated employing the full sample for each segment while using all the population records for model estimation, and (2) provides valuable insights on how the exogenous variables affect segmentation. An Expectation–Maximization algorithm is proposed to address the challenges of estimating the resulting endogenous segmentation based econometric model. A prediction procedure to employ the estimated latent MDCEV models for forecasting is also developed. The proposed model is estimated using data from 2009 National Household Travel Survey (NHTS) for the New York region. The results of the model estimates and prediction exercises illustrate the benefits of employing an endogenous segmentation based MDCEV model. The challenges associated with the estimation of latent MDCEV models are also documented.     

Modeling the relationships among urban passenger travel carbon dioxide emissions,transportation demand and supply,population density,and proxy policy variables

To support the development of policies that reduce greenhouse gas (GHG) emissions by encouraging reduced travel and increased use of efficient transportation modes, it is necessary to better understand the explanatory effects that transportation, population density, and policy variables have on passenger travel related CO₂ emissions. This study presents the development of a model of CO₂ emissions per capita as a function of various explanatory variables using data on 146 urbanized areas in the United States. The model takes into account selectivity bias resulting from the fact that adopting policies aimed at reducing emissions in an urbanized area may be partly driven by the presence of environmental concerns in that area. The results indicate that population density, transit share, freeway lane-miles per capita, private vehicle occupancy, and average travel time have a statistically significant explanatory effect on passenger travel related CO₂ emissions. In addition, the presence of automobile emissions inspection programs, which serves as a proxy indicator of other policies addressing environmental concerns and which could influence travelers in making environmentally favorable travel choices, markedly changes the manner in which transportation variables explain CO₂ emission levels.     

Comparing decision tree algorithms to estimate intercity trip distribution

Traditional trip distribution models usually ignore the fact that destination choices are made individually in addition to aggregated factors, such as employment and average travel costs. This paper proposes a disaggregated analysis of destination choices for intercity trips, taking into account aggregated characteristics of the origin city, an impedance measurement and disaggregated variables related to the individual, by applying nonparametric Decision Tree (DT) algorithms. Furthermore, each algorithm’s performance is compared with traditional gravity models estimated from a stepwise procedure (1) and a doubly constrained procedure (2). The analysis was based on a dataset from the 2012 Origin-Destination Survey carried out in Bahia, Brazil. The final selected variables to describe the destination choices were population of the origin city, GDP of the origin city and travel distances at an aggregated level, as well as the variables: age, occupation, level of education, income (monthly), number of cars per household and gender at a disaggregated one. The comparison of the DT models with gravity models demonstrated that the former models provided better accuracy when predicting the destination choices (trip length distribution, goodness-of-fit measures and qualitative perspective). The main conclusion is that Decision Tree algorithms can be applied to distribution modeling to improve traditional trip distribution approaches by assimilating the effect of disaggregated variables.     

Sampling of alternatives for route choice modeling

This paper presents a new paradigm for choice set generation in the context of route choice model estimation. We assume that the choice sets contain all paths connecting each origin–destination pair. Although this is behaviorally questionable, we make this assumption in order to avoid bias in the econometric model. These sets are in general impossible to generate explicitly. Therefore, we propose an importance sampling approach to generate subsets of paths suitable for model estimation. Using only a subset of alternatives requires the path utilities to be corrected according to the sampling protocol in order to obtain unbiased parameter estimates. We derive such a sampling correction for the proposed algorithm.Estimating models based on samples of alternatives is straightforward for some types of models, in particular the multinomial logit (MNL) model. In order to apply MNL for route choice, the utilities should also be corrected to account for the correlation using, for instance, a path size (PS) formulation. We argue that the PS attribute should be computed based on the full choice set. Again, this is not feasible in general, and we propose a new version of the PS attribute derived from the sampling protocol, called Expanded PS.Numerical results based on synthetic data show that models including a sampling correction are remarkably better than the ones that do not. Moreover, the Expanded PS shows good results and outperforms models with the original PS formulation.     

Determining the role of bicycle sharing system infrastructure installation decision on usage: Case study of montreal BIXI system

The traditional quantitative approach to studying Bicycle Sharing System (BSS) usage involves examining the influence of BSS infrastructure (such as number of BSS stations and capacity), transportation network infrastructure, land use and urban form, meteorological data, and temporal characteristics. These studies, as expected, conclude that BSS infrastructure (number of stations and capacity) have substantial influence on BSS usage. The earlier studies consider usage as a dependent variable and employ BSS infrastructure as an independent variable. Thus, in the models developed, the unobserved factors influencing the measured dependent variable (BSS usage) also strongly influence one of the independent variables (BSS infrastructure). This is a classic violation of the most basic assumption in econometric modeling i.e. the error component in the model is not correlated with any of the exogenous variables. The model estimates obtained with this erroneous assumption are likely to over-estimate the impact of BSS infrastructure. Our research effort proposes an econometric framework that remedies this drawback. We propose a measurement equation to account for the installation process and relate it to the usage equations thus correcting for the bias introduced in earlier research efforts by formulating a multi-level joint econometric framework. The econometric models developed have been estimated using data compiled from April 2012 to August 2012 for the BIXI system in Montreal. The model estimates support our hypothesis and clearly show over-estimation of BSS infrastructure impacts in models that neglect the installation process. An elasticity analysis to highlight the advantages of the proposed econometric model is also conducted.