Search

The primary metric for measuring electric vehicle (EV) adoption growth is new car sales. However, to enable mass market penetration, EV adoption in the used car market will play a crucial role. The used vehicle market is relatively under-studied or has been studied mostly for specific regions. This project analyzed US national consumer expenditure survey data that tracks households' expenditure on vehicle acquisition and operation. The study aim is to understand new versus used vehicle choice behavior and the consequent cost of vehicle ownership, with the larger aim of determining how much households who generally buy used vehicles can gain or lose if they transition from a used internal combustion engine vehicle (ICEV) to a used EV. A choice model and cluster analysis showed that ownership of used vehicles is influenced by family size, income, housing tenure, and age. For lower-income renters, current vehicle ownership and purchase costs tend to constitute a high fraction of their household income, raising concerns related to equity and suggesting that these households in particular should be considered in policies to encourage the EV transition. Moreover, while at present the average price paid for a used ICEV is approximately $18,000, the price of a comparable used EV can range between $14,000 (e.g., lower electric range Nissan Leaf) to $50,000 (high-range Tesla), suggesting the need for incentives to encourage the used EV market.

Assessing the efficacy of algorithms plays a pivotal role in advancing various fields, both in theory and practice. Unlike the predictive models, due to the intricate relationship between decisions and the underlying data‐generating processes, the evaluation of decision algorithms cannot directly rely on real data. Hence, a simulator becomes indispensable for appraising decision algorithm effectiveness. In this paper, we aim to leverage assortment decisions, a widely used application in revenue management, to illustrate the utilization of a machine learning‐based simulation. The process can be summarised as: we utilize the modified Transformer‐based choice model, acting as a simulator, to generate a synthetic dataset that mimics consumer purchasing behavior. After training the MNL, DeepFM, and DeepFM‐a models, all of which can swiftly provide assortment decisions in real‐time, we utilize the simulator to evaluate the revenue generated by each assortment prescribed by different choice models. This approach mitigates the challenge of validating decision models that alter real‐world observed data. To show the benefit of such a simulation approach, we have conducted various numerical studies. These studies aim to examine the impact of outside option attractiveness, data size, the number of features, and cardinality. Admittedly, due to the close alignment between the simulator and complex consumer purchase choice datasets, some numerical observations may be challenging to explain. Nevertheless, by employing the simulator, we are able to contrast the differences between the MNL and DeepFM/DeepFM‐a models, shedding light on their respective model misspecifications. [ABSTRACT FROM AUTHOR]

The increasing diversity of vehicle type holdings and growing demand for BEVs and PHEVs have serious policy implications for travel demand and air pollution. Consequently, it is important to accurately predict or estimate the preference for vehicle holdings of households as well as the vehicle miles traveled by vehicle body- and fuel-type to project future VMT changes and mobile source emission levels. Leveraging the 2019 California Vehicle Survey data, this report presents the application of a utility-based model for multiple discreteness that combines multiple vehicle types with usage in an integrated model, specifically the MDCEV model. The model results suggest the important effects of household demographics, residence location, and built environment factors on vehicle body type and powertrain choice and usage. Further the predictions associated with changes inbuilt environment factors like population density can inform the design of land-use and transportation policies to influence household vehicle holdings and usage that can in turn impact travel demand and air quality issues in California.View the NCST Project Webpage

This study analyses the factors that affect farmers' decisions to quit farming, using the İzmir region of Türkiye as a case study. A sample of 195 farmers was surveyed to identify the factors influencing this decision. Factors such as the farm type, farm size, agricultural subsidies, land ownership, and age of farmers were found to significantly affect the decisions of the farmers. Sensitivity analysis was performed to examine the changes in the selected variables. The sensitivity analysis showed that young farmers, under financial pressure and tenancy uncertainty, are likely to stay in farming when they receive agricultural subsidies. Thus, targeted financial subsidies can help sustain the livelihood of young farmers efficiently. This study reveals the important relationship between policy interventions and the long-term economic viability of farming as a livelihood. This relationship is crucial for the overall viability of agriculture and the well-being of rural communities. Overall, the findings from this study will inform the development of policies to strengthen the resilience and sustainability of rural farming communities in Türkiye. [ABSTRACT FROM AUTHOR]

Augmenting the Modeling Power of the Multinomial Logit Model Using choice models to capture customer choice behavior has steadily become the common practice in revenue management. Discrete choice models allow one to capture the fact that customers substitute among the offered products, so if a particular product is not offered, then a portion of the customers interested in this product will substitute into a suitable available alternative. The multinomial logit model is one of the most commonly used choice models to capture customer choice in practice. This choice model is based on the utility maximization framework. A customer associates random utilities with the products, as well as the no-purchase option, in which case, the customer purchases the product with the largest utility, as long as its utility exceeds that of the no-purchase option. In "Assortment Optimization under the Multinomial Logit Model with Utility-Based Rank Cutoffs," Bai, Feldman, Topaloglu, and Wagner augment the modeling flexibility of the multinomial logit model, where a customer also leaves without a purchase if she cannot find one of her top few choices. They develop algorithms to find revenue-maximizing assortments. We study assortment optimization problems under a natural variant of the multinomial logit model where the customers are willing to focus only on a certain number of products that provide the largest utilities. In particular, each customer has a rank cutoff, characterizing the number of products that she will focus on during the course of her choice process. Given that we offer a certain assortment of products, the choice process of a customer with rank cutoff k proceeds as follows. The customer associates random utilities with all of the products as well as the no-purchase option. The customer ignores all alternatives whose utilities are not within the k largest utilities. Among the remaining alternatives, the customer chooses the available alternative that provides the largest utility. Under the assumption that the utilities follow Gumbel distributions with the same scale parameter, we provide a recursion to compute the choice probabilities. Considering the assortment optimization problem to find the revenue-maximizing assortment of products to offer, we show that the problem is NP-hard and give a polynomial time approximation scheme. Because the customers ignore the products below their rank cutoffs in our variant of the multinomial logit model, intuitively speaking, our variant captures choosier choice behavior than the standard multinomial logit model. Accordingly, we show that the revenue-maximizing assortment under our variant includes the revenue-maximizing assortment under the standard multinomial logit model, so choosier behavior leads to larger assortments offered to maximize the expected revenue. We conduct computational experiments on both synthetic and real data sets to demonstrate that incorporating rank cutoffs can yield better predictions of customer choices and yield more profitable assortment recommendations. Funding: The work of Y. Bai and H. Topaloglu was partially supported by the National Science Foundation [Grant No. CMMI-1825406]. Supplemental Material: The e-companion is available at https://doi.org/10.1287/opre.2021.0060. [ABSTRACT FROM AUTHOR]

Risk Sensitivity and Firm Power: Price Competition with Mean2-Variance Profit Objective Under Multinomial Logit Demand Firms with varying degrees of risk sensitivity perceive the same stochastic profit prospect differently. It is important to understand the equilibrium pricing behaviors under risk aversion and their implications on firms' survival and market health. The authors identify a power index, which is the ratio of product attractiveness to risk sensitivity, and show that the set of profitable firms is power index ordered. Firms with a favorable power index value earn a positive profit in a price competition, and others are driven to zero profit. Interestingly, although high risk aversion handicaps a firm, this paper shows that moderate risk aversion may give a firm an advantage over an otherwise equivalent competitor that is less risk sensitive. The authors also establish that in an equilibrium with market entry and exit, the power index is generalized to the ratio of product attractiveness and an entry cost-adjusted risk sensitivity measure. This paper is the first in the literature to address a risk-sensitive price competition under the multinomial logit choice model, with each participating firm maximizing a risk-adjusted profit objective. We find that, at equilibrium, a subset of firms earns a positive profit, whereas others are driven to zero profit, contrasting with the risk-neutral equilibrium in which all firms earn a positive profit regardless of quality and cost. We identify a power index—the ratio of effective product attractiveness to risk sensitivity—and show that the set of profitable firms is power index ordered. Risk aversion drives firms toward a more aggressive equilibrium pricing strategy and intensifies competition. However, the relative profit impact across firms is not monotone; although high risk aversion handicaps a firm, moderate risk aversion may place a firm at an advantage over an otherwise equivalent competitor that is less risk sensitive, contrary to intuition. In an equilibrium with market entry cost, we show that the set of active firms at equilibrium follows a generalized power index order that depends on entry cost. Furthermore, although the power index is decreasing in risk aversion, the generalized power index may initially be increasing in risk aversion. Supplemental Material: The online appendix is available at https://doi.org/10.1287/opre.2023.2466. [ABSTRACT FROM AUTHOR]

Microtransit is an emerging, technology-enabled, on-demand transportation mode whereby small shuttles provide shared rides through flexible routing and scheduling in response to customers’ requests for rides. Given its potential to address the equity and accessibility needs of the public, public transportation agencies are experimenting with this service to fill gaps in traditional transportation in the US. However, why some people are interested in microtransit while others are not remains an open question. For people who have never used it, what factors could work as facilitators or barriers in their willingness to adopt microtransit? Who are the early adopters of microtransit? Guided by the theory of planned behavior, this study aims to fill the gap in knowledge by conducting a large-scale survey of microtransit adopters and users of other means of transportation in the Sacramento area of California in 2021. This study focuses on the microtransit service SmaRT Ride (SR), operated by the Sacramento Regional Transit District (SacRT). Focus groups and interviews were conducted before the largescale online survey to gather preliminary information, help develop survey questions, and improve understanding of research findings, given the novelty of microtransit. Discrete choice models, including binary logit and ordered logit models and latent class analysis, were employed to explore barriers to and facilitators of SR adoption, willingness to use it, and underlying subgroups of early adopters. Important findings include that people who like fixed-route transit are less likely to adopt microtransit. Social support plays an important role in explaining the willingness to use microtransit. The analysis reveals three salient classes of microtransit users: travel time savers with environmental awareness, riders with a neutral mindset, and pro-SR and travel cost savers.View the NCST Project Webpage

In this article, we present software that is suitable for use with Stata's choice modeling suite of commands, which begin with cm. Within the context of choice models, we focus on best–worst data. In such data, respondents are presented a set of choices and are required to select a best and a worst choice from among the alternatives. Optionally, respondents may indicate an opt-out choice, in which no best or worst choice exists in the choice set. Such data are simplified versions of experiments in which respondents rank all the choices. Once best–worst data are collected, there are specific types of data expansions that analysts use to take advantage of both explicit and implicit information. The commands described in this article support data expansion and model estimation. [ABSTRACT FROM AUTHOR]

Nonprofit organizations (NPOs) play a central role in many economies in the form of private entities serving a public purpose. Strengthening the fundraising capabilities of NPOs can have a large impact on their survival and effective functioning. NPOs typically elicit financial contributions through multiple forms of giving, such as donation and membership. These options enable individuals to express their altruism by giving in one or multiple forms. The authors develop a utility-based multiple discrete-continuous model that provides insights into potentially large differences in individuals' giving behaviors. Through Bayesian Gaussian processes, the model also incorporates changes in givers' preferences for forms of giving. The authors apply their model to five years of individual giving data. They find that the effects of lifetime, recency, seasonality, and appeals on donation and membership options change nonmonotonically over time and in distinctive ways. The authors demonstrate that the model estimates help predict who will give in more than one form in the future as well as build appeal targeting strategies. The model also shows that fundraising attempts should emphasize participation rather than amount, and that long-lapsed members are still worth pursuing for renewal, whereas long-lapsed donors are less productive for repeat giving. [ABSTRACT FROM AUTHOR]

This study analyzes the use of ride-hailing and bus services from a multi-modal user perspective by jointly modeling individuals’ monthly frequency of use of both bus and ride-hailing services. A bivariate ordered probit model is estimated to capture the influence of socio-demographic and attitudinal characteristics as well as unobserved factors that may simultaneously influence the travel frequency by both modes. We use data collected in a survey of travelers from Porto Alegre, Brazil. The results characterize the profile of local frequent ride-hailing users as young, medium-income, self-employed individuals with a propensity toward the use of technology and a low inclination for car ownership. Most importantly, we observe that after controlling for unobserved factors that simultaneously contribute to higher trip frequencies by both modes, frequent bus users demonstrate a lower propensity toward ride-hailing use.

On-demand mobility services including transportation network companies (also known as ridesourcing and ridehailing) like Lyft and Uber are changing the way that people travel by providing dynamic mobility that can supplement public transit and personal-vehicle use. However, TNC services have been found to contribute to increasing vehicle mileage, traffic congestion, and greenhouse gas emissions. Pooling rides ⎯ sharing a vehicle by multiple passengers to complete journeys of similar origin and destination ⎯ can increase the average vehicle occupancy of TNC trips and thus mitigate some of the negative impacts. Several mobility companies have launched app-based pooling services in recent years including app-based carpooling services (e.g., Waze Carpool, Scoop) that match drivers with riders; pooled on-demand ride services (e.g., Uber Pool and Lyft Shared rides) that match multiple TNC users; and microtransit services (e.g., Bridj, Chariot, Via) that offer on-demand, flexibly routed service, typically in larger vehicles such as vans or shuttles. However, information on the potential impacts of these options is so far limited. This research employs a general population stated preference survey of four California metropolitan regions (Los Angeles, Sacramento, San Diego, and the San Francisco Bay Area) in Fall 2018 to examine the opportunities and challenges for drastically expanding the market for pooling, accounting for differences in emergent travel behavior and preferences across the four metropolitan regions surveyed. The travel profiles, TNC use patterns, and attitudes and perceptions of TNCs and pooling are analyzed across key socio-demographic attributes to enrich behavioral understanding of marginalized and price sensitive users of on-demand ride services. This research further develops a discrete choice model to identify significant factors influencing a TNC user’s choice to pool or not to pool, as well as estimating a traveler’s value of time (VOT) across different portions of a TNC trip. This research provides key insights and social equity considerations for policies that could be employed to reduce vehicle miles traveled and emissions from passenger road transportation by incentivizing the use of pooled on-demand ride services and public transit

Higher-occupancy self-driving shuttles could bring about the benefits of vehicle automation—improved safety, parking cost savings, greater mobility to those who cannot drive, and stress relief for drivers. At the same time, these shuttles would not bring the potential drawbacks of self-driving vehicle ownership, such as increases in vehicle miles traveled and associated energy use. Because they can only currently operate in relatively simple and closed environments, self-driving shuttles are likely to be deployed earlier than personal self-driving vehicles in open road environments. However, acceptance of the new technology remains uncertain. Whether people will use these services will be largely influenced by their attitudes toward self-driving technology.Researchers at the University of California, Davis surveyed residents and employees of the West Village area of the UC Davis campus during the three-month pilot deployment of a self-driving, electric shuttle to understand attitudes toward self-driving technology. The researchers then applied existing theories of technology adoption to model how attitudes of residents and employees influenced their acceptance of the shuttle service. This policy brief summarizes findings from that research and provides policy implications of self-driving shuttles.View the NCST Project Webpage

Self-driving vehicles, as a revolution in mobility, are emerging and developing rapidly. However, public attitudes toward this new unproven technology are still uncertain. Given the significant influence of attitude toward a new technology on the intention to use it, the question arises as to why some people are in favor of this technology whereas others are not. Additionally, questions about the key attitudes influencing self-driving technology acceptance, where these attitudes come from, and how they interact with each other have not yet been addressed.This study aims to explore these research questions based on data collected from people who live or work in the West Village area of the University of California, Davis, campus after a self-driving electric shuttle was piloted in this area. structural equation modeling was employed to explore interactions between attitude elements. The results show that affect—i.e., liking or enthusiasm for self-driving shuttles—strongly explains the acceptance of self-driving technology. A higher level of affect could be formed by strengthening an individual’s trust. Additionally, trust works as an important mediator between perceived risk, usefulness, and ease of use on both affect and intention to ride self-driving vehicles. Perceived risk captured more security and functional concerns, reflecting uncertainty around current self-driving technology. The model identified important bi-directional influences between trust and affect. Significant effects of mental and physical intangibility were also shown, but each works differently on cognitive beliefs. Individuals’ socio-demographic, lifestyle, and mobility characteristics also exert influences on attitude and self-driving technology acceptance.View the NCST Project Webpage

Increasing transit use has many benefits, including reducing traffic congestion and greenhouse gas (GHG) emissions. However, riders need to be able to get to a station in order to use transit. Walking is an option only for those within a limited radius of a station. Driving to a station may be feasible for some, but providing sufficient parking can be expensive and land intensive. The rise of ride-hailing companies such as Uber and Lyft presents a new opportunity for bridging the “first-mile” gap to high quality transit. Transit agencies are beginning to launch pilot projects to test public-private partnerships with ride-hailing companies to increase access to transit.This policy brief summarizes findings from researchers at UC Davis who used existing modeling tools and data to understand the potential market demand for a first-mile transit access service in the San Francisco Bay Area. They modeled the likelihood of commuters who drive alone to switch to using ride-hailing and the Bay Area Rapid Transit (BART) rail system to get to work based on travel time, cost, and distance to a BART station. They explored the magnitude of change in overall travel time and cost for travelers who switch from driving alone to using ride-hailing and BART, as well as potential changes to vehicle miles traveled (VMT) and GHG emissions at both the regional and station level.View the NCST Project Webpage

We consider the cardinality-constrained assortment optimization problem under the nested logit model where there is a constraint that limits the number of products that can be offered within each nest. The problem is known to be intractable if the nest dissimilarity parameters are larger than one or there is a no-purchase alternative within each nest. Although these conditions often come up in practice, the existing solution approaches cannot handle them. We propose a solution method to obtain heuristic assortments with provable worst-case performance guarantees that hold even when the nest dissimilarity parameters are larger than one or there is a no-purchase alternative within each nest. We obtain a tractable upper bound that can be used to assess the practical performance of our solution approach. Computational experiments indicate that the heuristic assortments perform very well, with optimality gaps being smaller than 1% on average. Our analysis also provides sharper performance bounds for the unconstrained assortment optimization problem under the nested logit model. [ABSTRACT FROM AUTHOR]

The goal of this study is to assess and quantify the potential employment accessibility benefits of Shared Autonomous Mobility Service (SAMS) commute modes across a large diverse metropolitan region considering heterogeneity in the working population. To meet this goal, this study employs a welfare-based (i.e. logsum-based) measure of accessibility, obtained via estimating a hierarchical work destination-commute mode choice model. The employment accessibility logsum measure incorporates the spatial distribution of worker residences and employment opportunities, the attributes of the available commute modes, and the characteristics of individual workers. This research further captures heterogeneity of workers using latent class analysis (LCA). The LCA model inputs include the socio-demographic characteristics of workers to subsequently account for different worker clusters valuing different types of employment opportunities differently. The accessibility analysis results indicate: (i) the accessibility benefit differences across latent classes are modest but young workers and low-income workers do see higher benefits than high- and middle-income workers; (ii) there are substantial spatial differences in accessibility benefits with workers living in lower density areas benefiting more than workers living in high-density areas; (iii) nearly all the accessibility benefits come from the SAMS-only mode as opposed to the SAMS+Transit mode; and (iv) the SAMS cost per mile assumption significantly impacts the magnitude of the overall employment accessibility benefits.

The public sector and the private sector, which includes automakers and charging network companies, are increasingly investing in building charging infrastructure to encourage the adoption and use of plug-in electric vehicles (PEVs) and to ensure that current facilities are not congested. However, building infrastructure is costly and, as with road congestion, when there is significant uptake of PEVs, we may not be able to “build out of congestion.” We modelled the choice of charging location that more than 3000 PEV drivers make when given the options of home, work, and public locations. Our study focused on understanding the importance of factors driving demand such as: the cost of charging, driver characteristics, access to charging infrastructure, and vehicle characteristics. We found that differences in the cost of charging play an important role in the demand for charging location. PEV drivers tend to substitute workplace charging for home charging when they pay a higher electricity rate at home, more so when the former is free. Additionally, socio-demographic factors like dwelling type and gender, as well as vehicle technology factors like electric range, influence the choice of charging location.

Marketing research relies on individual-level estimates to understand the rich heterogeneity of consumers, firms, and products. While much of the literature focuses on capturing static cross-sectional heterogeneity, little research has been done on modeling dynamic heterogeneity, or the heterogeneous evolution of individual-level model parameters. In this work, the authors propose a novel framework for capturing the dynamics of heterogeneity, using individual-level, latent, Bayesian nonparametric Gaussian processes. Similar to standard heterogeneity specifications, this Gaussian process dynamic heterogeneity (GPDH) specification models individual-level parameters as flexible variations around population-level trends, allowing for sharing of statistical information both across individuals and within individuals over time. This hierarchical structure provides precise individual-level insights regarding parameter dynamics. The authors show that GPDH nests existing heterogeneity specifications and that not flexibly capturing individual-level dynamics may result in biased parameter estimates. Substantively, they apply GPDH to understand preference dynamics and to model the evolution of online reviews. Across both applications, they find robust evidence of dynamic heterogeneity and illustrate GPDH's rich managerial insights, with implications for targeting, pricing, and market structure analysis. [ABSTRACT FROM AUTHOR]

The authors propose and empirically evaluate a new hybrid estimation approach that integrates choice-based conjoint with repeated purchase data for a dense consumer panel, and they show that it increases the accuracy of conjoint predictions for actual purchases observed months later. The key innovation lies in combining conjoint data with a long and detailed panel of actual choices for a random sample of the target population. By linking the actual purchase and conjoint data, researchers can estimate preferences for attributes not yet present in the marketplace, while also addressing many of the key limitations of conjoint analysis, including sample selection and contextual differences. Counterfactual product and pricing exercises illustrate the managerial relevance of the approach. [ABSTRACT FROM AUTHOR]

ObjectivesTo identify consumers' preferences over care settings, such as physicians' offices, emergency rooms (ERs), urgent care centers, retail clinics, and virtual physicians on smartphones, for minor illnesses.Data sourcesA survey conducted between 9/27/16 and 12/7/16 emailed to all University of California, Irvine employees.Study designParticipants were presented with 10 clinical scenarios and asked to choose the setting in which they wanted to receive care. We estimated multinomial conditional logit regression models, conditioning the choice on out-of-pocket costs, wait time, travel time, and chooser characteristics.Data collection5451 out of 21 037 employees responded.Principal findingsOut-of-pocket costs and wait time had minimal impact on patient's preference for site of care. Choices were driven primarily by the clinical scenario and patient characteristics. For chronic conditions and children's well-visits, the doctor's office was the preferred choice by a strong majority, but for most acute conditions, either the ER (for high severity) or urgent care clinics (for lower severity) were preferred to the office setting, particularly among younger patients and those with less education.ConclusionsPatients have several alternatives to traditional physicians' offices and ERs. The low impact of out-of-pocket costs suggests that insurers interested in encouraging increased utilization of alternatives would need to consider substantial changes to benefit structure.

The paper compares the profitability of personalized pricing policies that are generated from different models of demand and using different data inputs. We study the profitability of personalized pricing policies in a setting with consumer-level panel data. To compare pricing policies, we propose an inverse probability-weighted estimator of profits, discuss how to handle nonrandom price variation, and show how to apply it in a typical consumer-packaged good market with supermarket scanner data. We generate pricing policies from Bayesian hierarchical choice models, regularized regressions, neural networks, and nonparametric classifiers using different sets of data inputs. We find that the performance of machine learning methods is highly varied, ranging from a 30.7% loss to a 14.9% gain relative to a blanket couponing strategy, whereas hierarchical models generate profit gains in the range of 13−16.7%. Across all models, information on consumers' purchase histories leads to large improvements in profits, whereas demographic information has only a small impact. We find that out-of-sample fit statistics are uncorrelated with profit estimates and provide poor guidance toward model selection. History: Avi Goldfarb served as the senior editor, and Günter Hitsch served as associate editor for this article. Supplemental Material: Data files are available at https://doi.org/10.1287/mksc.2022.1387. [ABSTRACT FROM AUTHOR]

Motivated by online labor markets, we consider the online assortment optimization problem faced by a two-sided matching platform that hosts a set of suppliers waiting to match with a customer. Arriving customers are shown an assortment of suppliers and may choose to issue a match request to one of them. After spending some time on the platform, each supplier reviews all the match requests she has received and, based on her preferences, she chooses whether to match with a customer or to leave unmatched. We study how platforms should design online assortment algorithms to maximize the expected number of matches in such two-sided settings. We establish that a simple greedy algorithm is 1/2-competitive against an optimal clairvoyant algorithm that knows in advance the full sequence of customers' arrivals. However, unlike related online assortment problems, no randomized algorithm can achieve a better competitive ratio, even in asymptotic regimes. To advance beyond this general impossibility, we consider structured settings where suppliers' preferences are described by the multinomial logit and nested logit choice models. We develop new forms of balancing algorithms, which we call preference-aware, that leverage structural information about suppliers' choice models to design the associated discount function. In certain settings, these algorithms attain competitive ratios provably larger than the standard "barrier" of 1 − 1 / e in the adversarial arrival model. Our results suggest that the shape and timing of suppliers' choices play critical roles in designing online assortment algorithms for two-sided matching platforms. This paper was accepted by Omar Besbes, revenue management and market analytics. Supplemental Material: The data files and online appendices are available at https://doi.org/10.1287/mnsc.2022.4464. [ABSTRACT FROM AUTHOR]

Choice models (CM) are proposed in the field of tourism recommender systems (TRS) with the aim of providing algorithms with both a theoretical understanding of tourist's motivations and a certain degree of transparency. The goal of this work is to overcome some of the limitations of current state‐of‐art algorithms used in TRSs by providing: (1) accurate preferences, which are learnt from user choices rather than from ratings, and (2) interpretable coefficients, which are achieved by means of the set of estimated parameters of CM. The study was carried out with a gastronomic data set generated in an ecological experiment in the tourism domain. The performance of CM has been compared with a set of baseline algorithms (rating‐based and ensembles) by using two evaluation metrics: precision and DCG. The CM outperformed the baseline algorithms when the size of the choice set was limited. The findings suggest that CM may provide an optimal trade‐off between theoretical soundness, interpretability and performance in the field of TRS. [ABSTRACT FROM AUTHOR]

Exit choice in emergency evacuation is of utmost importance. In a multi-exit scenario, the choice of emergency exit leads to a safer destination. Evacuees can choose other exits also. Understanding the factors affecting the exit choice is important. In this study, an exit choice experiment is conducted to understand the factors affecting the exit choice. Partial immersive virtual reality is used to record the choices of individuals in twelve different scenarios. An efficient choice design is made for the scenario creation. Unreal engine, a game engine with better graphics, particle system, and blueprints is used to give participants a virtual experience of the fire emergency in a metro station. Participants are partially immersed in a wide computer screen for this experiment. The recorded exit choice information is used to calibrate different multinomial logit models. The results indicate that the choice of emergency exit in different scenarios is not same for all participants. The factors such as distance, queue time, fire and smoke negatively impact the exit choices.

We present an employer-side perspective on remote work through the pandemic using data from top executives of 129 employers in North America. Our analysis suggests that at least some of the pandemic-accelerated changes to the work location landscape will likely stick; with some form of hybrid work being the norm. However, the patterns will vary by department (HR/legal/sales/IT, etc.) and by sector of operations. Top three concerns among employers include: supervision and mentoring, reduction in innovation, and creativity; and the top three benefits include their ability to retain / recruit talent, positive impact on public image and their ability to compete. An Ordered Probit model of the expected April 2024 work location strategy revealed that those in transportation, warehousing, and manufacturing sectors, those with a fully in-person approach to work pre-COVID, and those with a negative outlook towards the impact of remote work are likely to be more in-person-centered, while those with fully remote work approach in April 2020 are likely to be less in-person-centered. Lastly, we present data on resumption of business travel, in-person client interactions and changes in office space reconfigurations that employers have made since the beginning of the pandemic. [ABSTRACT FROM AUTHOR]

Gift giving generates high revenues for retailers. It is also marked with significant welfare, or deadweight, loss in that givers tend to pay more than the receivers’ valuation. Previous research has attributed this discrepancy to givers’ inaccurate predictions of the receivers’ preferences. This research demonstrates that reduced price sensitivity is another important source of the deadweight loss: givers use gift prices to signal the importance of their relationship with the receiver. In order to demonstrate this mechanism, the authors develop a new Bayesian gift-choice model that captures both preference predictions as well as the signaling value of price. The model is estimated on two choice-based conjoint studies for gift giving that allow for the manipulation of the giver’s uncertainty about the receiver’s preferences. Both studies show the strong signaling value of price, especially when givers are uncertain about receivers’ preferences. Decomposition of the deadweight loss shows that the signaling value of price is an important source of welfare loss, especially in markets with heterogeneous prices. These findings have key implications for the gift industry. [ABSTRACT FROM AUTHOR]

This paper aims at developing new models to combine the best of two paradigms: the performance of ensembles and the transparency of choice models. Specifically, the work explores several blind methods to build ensembles of single choice-based models. They were fit using a dataset that includes rational, emotional and attentional variables, all of them gathered from humans during a choice task. Two main strategies, 1-Learner and N-Learners type ensembles, were analyzed in terms of their accuracy as well as frequency of best-model metrics. The results point out the superior performance of N-Learners ensembles and show the potential of personalized arrangements. [ABSTRACT FROM AUTHOR]

Sparsely populated areas tend to be poorly served by Fixed Line and Schedule (FLS) public transport systems as the operation of a regular bus line is not economically viable for such areas. Therefore, introducing a Demand Responsive Transport (DRT) to partially replace FLS can result in increasing mobility service accessibility and inclusion. In this paper, a mixed-integer linear problem (MILP) is proposed to design an integrated FLS and DRT network for a transport operator. Passengers behavior is implicitly incorporated in our proposed approach via a discrete choice model. In addition, a tailored Adaptive Large Neighborhood Search (ALNS) coupled with tabu search and simulated annealing are introduced. We test our algorithm on real instances from a public transport operator in the Netherlands. The proposed algorithm can solve the problem up to 170 times faster than the MILP within 4% to 10% gap. Our proposed resolution approach investigates the temporal and spatial feasibility of deploying these integrated mobility systems based on the service level and provides recommendations to public transport operators. • An integrated FLS and DRT integrated transport system is considered. • A choice-driven MILP mathematical model is presented. • An ALNS algorithm coupled with tabu search and simulated annealing is introduced. • Computational time and optimality gaps are reported for proposed methods. • Results are validated using a case in the Netherlands. [ABSTRACT FROM AUTHOR]

The design of new products is a matter of great importance that can directly affect profitability and competitiveness in modern companies. For this reason, the selection of a final product design needs to consider at least four factors of importance: anticipated market demand for the product design, preference heterogeneity among consumers, decision-maker preferences, and fuzzy preference information in the design criteria. This paper proposes a frequency-based preference choice model that considers all the above factors and can be used with algorithms that solve the optimal product design problem using the share of preference frequency criterion. The choice model introduced in this paper is based on the multicriteria outranking approach, and its predictive accuracy is optimized with genetic algorithms. The proposed genetic algorithm is compared with Interior Point OPTimizer, a software package for large-scale non-linear optimization. The experiment results demonstrate that the proposed method achieves near-optimal solutions in reasonable computational time and significantly outperforms the runtime compared algorithm. [ABSTRACT FROM AUTHOR]

This report presents the results of a project to develop a truck vehicle/fuel decision choice model for California and to use that model to make initial projections of truck sales by technology out to 2050. The report also describes the linkage of this model to a broader scenarios model of road transportation energy use in California to 2050. A separate report provides the authors detailed assumptions about truck technologies, fuels, and projections to 2050 that are inputs to this choice modeling effort. The need for low carbon trucking in California, as in other states and countries of the world, is outlined in IPCC reports and the Paris Agreement. An 80% reduction in energy-related CO2 emissions worldwide is targeted in that agreement. For trucks to contribute anywhere near this level of reduction, new, zero emissions technologies, such as electric and hydrogen fuel cell trucks, would need to be adopted at a large scale and at a rapid pace, both unprecedented for trucks anywhere in the world to date. Many truck models create new technology market penetration scenarios through minimizing cost or in an ad-hoc manner. This model utilizes a fleet decision choice process based on real world factors identified through discussions with trucking fleets. These factors include capital and operating costs, uncertainty (risk), model availability, refueling inconvenience, green PR (perceived benefit of environmentally beneficial technologies), and various incentives. The authors have developed a spreadsheet structured as a nested multinomial logit model that monetizes these factors to calculate a generalized cost. The authors have attempted to estimate the value of these factors to different types of fleets using a series of interviews, initial survey work, a truck choice workshop, and finally expert judgment and “basic logic” on how various factors might be valued now and in the future. The factors drive the choice analysis and are highly uncertain and likely highly variant across fleet types and even fleets within a type (early adopter, late adopter, in between), so the authors use a scenario approach to explore how this uncertainty could affect their results and projections. The authors created four scenarios and variants: 1) a business as usual (BAU), 2) a zero-emission vehicle (ZEV) mandate requiring the market share of ZEVs to reach 25% by 2050 (ZEV scenario 1a), 3) the same scenario but with a low penalty assumed for refueling time and (ZEV scenario 1b) 4) a ZEV mandate requiring the market share of ZEVs to reach 50% by 2050 (ZEV scenario 2). The authors also look at some policies that could help to spur sales growth among ZEV technologies in order to reach specific targets.View the NCST Project Webpage

In this study, the authors use existing modeling tools and data from the San Francisco Bay Area (California) to understand the potential market demand for a “first” mile transit access service and possible reductions in vehicle miles traveled (VMT) (and thus GHGs) at both the regional and station level. The authors develop a project scenario that targets commuters who drive alone but could take rail (Bay Area Rapid Transit, or BART) to work. The authors use travel time, cost, and distance data from San Francisco Bay Area travel models, Google and BART APIs, and TNC fares to explore the magnitude of change in overall travel time and cost for travelers who switch from driving alone to using TNC and BART to travel to work. The analysis indicates that 31% of the identified drive-alone trips could reduce generalized costs (travel time and monetary costs) by switching to TNC and BART. If all travelers who could benefit from traveling by TNC and BART, did in fact switch from drive-alone travel, about 40 thousand new BART trips could be generated and over a half a million miles of VMT avoided during the morning commute period. Most of these trips experienced relatively high levels of cost and VMT savings, which may be more likely to motivate behavioral change. Examination of cost savings by income level and vehicle availability suggests that the new service is more likely to benefit lower income households with fewer vehicles. Sensitivity analyses indicate that even with increased TNC fares and waiting times, there is still a relatively large number of trips that could benefit from switching from driving to TNC and BART. Tests of a shared TNC BART access service suggest cost savings and VMT reductions even when travel time costs increase by 60%. Analysis of benefits by station indicate locations with relatively high market potential (i.e., trips and cost saving) and environmental benefits (VMT reduction) for early pilot implementation. These results can be used to estimate potential TNC fare subsidies to increase performance by station to achieve project objectives, such as equitable access, increased BART ridership, and reduced VMT.View the NCST Project Webpage

• Propose model to consider probabilistic preliminary preference (PP) in unlabeled choice context. • Both mixture and latent class structure considered. • Simulation analysis confirms parameters can be estimated using MLE. • Proposed model applied to parking choice dataset and has better fit to data than existing models. • Analysis reveals loss aversion and inertia (or variety seeking) are present in parking choice data. When selecting among a set of alternatives, individuals often compare each alternative with reference to some preliminary preference (PP). The preliminary preference could be the individual's preferred global option (i.e., their ideal choice) or the most ideal option within the current choice set. Previous research has shown that loss of an attribute relative to that of the PP has a stronger impact on the individual's choice compared to gain of that attribute—a phenomenon commonly known as loss aversion. To incorporate this loss aversion into choice modeling, the PP of the individuals must then be known; however, such information is generally not available when observing individual's choice behavior. To overcome the lack of knowledge of an individual's PP, this study proposes a probabilistic approach where all alternatives in the choice set have some potential of being an individual's PP. A multinomial logit (MNL) based methodology, named the latent preliminary preference (LPP) model, is used to estimate the PP probabilities entirely as a function of the decision-maker characteristics and each alternative's attributes, which overcomes the complexity that arises when choices are unlabeled. In addition, the methodology accommodates inertia that some individual's may have toward their PP, as well as the desire to try new options. Moreover, this study proposes randomizing the parameters of LPP model to account for random taste heterogeneity of individuals that would arise in a panel dataset, giving rise to random parameter latent preliminary preference (RPLPP) model. The ability of maximum likelihood inference method to retrieve the parameters of fixed parameters LPP model is demonstrated on synthetic data. The statistical significance and bias in estimated parameters are also demonstrated. The proposed model is then applied on a parking preference survey to examine the presence of loss aversion and inertia/variety seeking behavior among the respondents. [ABSTRACT FROM AUTHOR]

Exit choice in emergency evacuation is of utmost importance. In a multi-exit scenario, the choice of emergency exit leads to a safer destination. Evacuees can choose other exits also. Understanding the factors affecting the exit choice is important. In this study, an exit choice experiment is conducted to understand the factors affecting the exit choice. Partial immersive virtual reality is used to record the choices of individuals in twelve different scenarios. An efficient choice design is made for the scenario creation. Unreal engine, a game engine with better graphics, particle system, and blueprints is used to give participants a virtual experience of the fire emergency in a metro station. Participants are partially immersed in a wide computer screen for this experiment. The recorded exit choice information is used to calibrate different multinomial logit models. The results indicate that the choice of emergency exit in different scenarios is not same for all participants. The factors such as distance, queue time, fire and smoke negatively impact the exit choices. • Exit choice behaviour in a metro station is studied using partial immersive virtual reality. • Unreal engine (with better graphics and particle system) is used for the study. • MNL model is calibrated with the choice data collected. [ABSTRACT FROM AUTHOR]

Integrating inventory and assortment planning decisions is a challenging task that requires comparing the value of demand expansion through broader choice for consumers with the value of higher in-stock availability. We develop a stockout-based substitution model for trading off these values in a setting with inventory replenishment, a feature missing in the literature. Using the closed form solution for the single-product case, we develop an accurate approximation for the multiproduct case. This approximated formulation allows us to optimize inventory decisions by solving a fractional integer program with a fixed point equation constraint. When products have equal margins, we solve the integer program exactly by bisection over a one-dimensional parameter. In contrast, when products have different margins, we propose a fractional relaxation that we can also solve by bisection and that results in near-optimal solutions. Overall, our approach provides solutions within 0.1% of the optimal policy and finds the optimal solution in 80% of the random instances we generate. This paper was accepted by David Simchi-Levi, optimization. [ABSTRACT FROM AUTHOR]

Long-distance and intercity travel generally make up a small portion of the total number of trips taken by an individual, while representing a large portion of aggregate distance traveled on the transportation system. While some research exists on intercity travel behavior between large metropolitan centers, this report addresses a need for more research on travel behavior between non-metropolitan areas and large metropolitan centers. This research specifically considers travel from home locations in northern New England, going to Boston, New York City, Philadelphia, and Washington, DC. These trips are important for quality of life, multimodal planning, and rural economies. This research identifies and quantifies factors that influence people’s mode choice (automobile, intercity bus, passenger rail, or commercial air travel) for these trips. The research uses survey questionnaire data, latent factor analysis, and discrete choice modeling methods. Factors include sociodemographic, built environment, latent attitudes, and trip characteristics. The survey, designed by the University of Vermont Transportation Research Center and the New England Transportation Institute, was conducted by Resource Systems Group, Inc. in 2014, with an initial sample size of 2560. Factor analysis was used to prepare 6 latent attitudinal factors, based on 70 attitudinal responses from the survey statements. The survey data were augmented with built environment variables using geographic information systems (GIS) analysis. A set of multinomial logit models, and a set of nested logit models, were estimated for business and non-business trip mode choice. Results indicate that for this type of travel, factors influencing mode choice for both business and non-business trips include trip distance; land use; personal use of technology; and latent attitudes about auto dependence, preference for automobile, and comfort with personal space and safety on public transportation. Gender is a less significant factor. Age is only significant for non-business trips.View the NCST Project Webpage

Two‐sided platforms have become omnipresent. In this context, firms compete not only for customers but also for flexible self‐scheduling workers who can work for multiple platforms. We consider a setting where two‐sided platforms simultaneously choose prices and wages to compete on both sides of the market. We assume that customers and workers each follow an endogenous generalized attraction model that accounts for network effects. In our model, the behavior of an agent depends not only on the price or wage set by the platforms, but also on the strategic interactions among agents on both sides of the market. We show that a unique equilibrium exists and that it can be computed using a tatônnement scheme. The proof technique for the competition between two‐sided platforms is not a simple extension of the traditional (one‐sided) setting and involves different arguments. Armed with this result, we study the impact of coopetition between two‐sided platforms, that is, the business strategy of cooperating with competitors. Motivated by recent practices in the ride‐sharing industry, we analyze a setting where two competing platforms engage in a profit sharing contract by introducing a new joint service. We show that a well‐designed profit sharing contract (e.g., under Nash bargaining) will benefit every party in the market (platforms, riders, and drivers), especially when the platforms are facing intensive competition on the demand side. However, if the platforms are facing intensive competition on the supply side, the coopetition partnership may hurt the profit of at least one platform. [ABSTRACT FROM AUTHOR]

Assortment optimization is an important problem arising in various applications. In many practical settings, the assortment is subject to a capacity constraint. In "Capacitated Assortment Optimization: Hardness and Approximation," Désir, Goyal, and Zhang study the capacitated assortment optimization problem. The authors first show that adding a general capacity constraint makes the problem NP-hard even for the simple multinomial logit model. They also show that under the mixture of multinomial logit model, even the unconstrained problem is hard to approximate within any reasonable factor when the number of mixtures is not constant. In view of these hardness results, the authors present near-optimal algorithms for a large class of parametric choice models including the mixture of multinomial logit, Markov chain, nested logit, and d-level nested logit choice models. In fact, their approach extends to a large class of objective functions that depend only on a small number of linear functions. Assortment optimization is an important problem that arises in many practical applications such as retailing and online advertising. In this problem, the goal is to select a subset of items that maximizes the expected revenue in the presence of (1) the substitution behavior of consumers specified by a choice model, and (2) a potential capacity constraint bounding the total weight of items in the assortment. The latter is a natural constraint arising in many applications. We begin by showing how challenging these two aspects are from an optimization perspective. First, we show that adding a general capacity constraint makes the problem NP-hard even for the simplest choice model, namely the multinomial logit model. Second, we show that even the unconstrained assortment optimization for the mixture of multinomial logit model is hard to approximate within any reasonable factor when the number of mixtures is not constant. In view of these hardness results, we present near-optimal algorithms for the capacity constrained assortment optimization problem under a large class of parametric choice models including the mixture of multinomial logit, Markov chain, nested logit, and d-level nested logit choice models. In fact, we develop near-optimal algorithms for a general class of capacity constrained optimization problems whose objective function depends on a small number of linear functions. For the mixture of multinomial logit model (resp. Markov chain model), the running time of our algorithm depends exponentially on the number of segments (resp. rank of the transition matrix). Therefore, we get efficient algorithms only for the case of constant number of segments (resp. constant rank). However, in light of our hardness result, any near-optimal algorithm will have a super polynomial dependence on the number of mixtures for the mixture of multinomial logit choice model. [ABSTRACT FROM AUTHOR]

The availability of individual-level transaction data allows retailers to implement personalized operational decisions. Although such decisions have been around for several years now in online platforms, recent technological developments open new opportunities to extend similar practices to bricks-and-mortar settings (e.g., by using electronic price tags to show different prices to different customers or by using beacon-based technology to send promotion offers to targeted customers). In "Personalized Retail Promotions through a DAG-Based Representation of Customer Preferences," Jagabathula, Mitrofanov, and Vulcano propose a back-to-back procedure for running customized promotions in retail operations contexts, from the construction of a nonparametric choice model where customer preferences are represented by directed acyclic graphs to the formulation of the promotion optimization problem. The empirical validation of their proposal on real supermarket data shows the promising performance of their approach over state-of-the-art benchmarks. We propose a back-to-back procedure for running personalized promotions in retail operations contexts, from the construction of a nonparametric choice model where customer preferences are represented by directed acyclic graphs (DAGs) to the design of such promotions. The source data include a history of purchases tagged by customer ID jointly with product availability and promotion data for a category of products. In each customer DAG, nodes represent products and directed edges represent the relative preference order between two products. Upon arrival to the store, a customer samples a full ranking of products within the category consistent with her DAG and purchases the most preferred option among the available ones. We describe the construction process to obtain the DAGs and explain how to mount a parametric, multinomial logit model (MNL) over them. We provide new bounds for the likelihood of a DAG and show how to conduct the MNL estimation. We test our model to predict purchases at the individual level on real retail data and characterize conditions under which it outperforms state-of-the-art benchmarks. Finally, we illustrate how to use the model to run personalized promotions. Our framework leads to significant revenue gains that make it an attractive candidate to be pursued in practice. [ABSTRACT FROM AUTHOR]

• We combined stated preference (SP) and revealed preference (RP) data to assess crossing pedestrian behavior. • We evaluate the impact of individual and trip characteristics, built environment, control strategies, and traffic flows on pedestrian crossing decisions. • The origin and destination of the pedestrian trip influence the decision. • We estimate marginal substitution rates for attributes of the built environment and the crossing area. Considering the high crash rates involving pedestrians on urban roads, it is highly relevant to understanding pedestrian crossing behavior. This paper is the first to combine stated preference (SP) and revealed preference (RP) data to evaluate the impact that individual attributes, trip characteristics, built environment, strategies to prevent unauthorized crossing, and traffic flows have on pedestrians crossing decisions in an urban context. SP and RP surveys were designed and collected in Barranquilla (Colombia) near pedestrian bridges or signalized intersections where direct crossings and a high concentration of pedestrian fatalities related to traffic accidents exist. A logit model was estimated using the data enrichment paradigm. Results show that pedestrians weigh risks and costs when choosing how to cross the road. The trajectories observed in the RP component suggest that people prefer direct crossings; nevertheless, pedestrian bridges and signalized intersections can be attractive alternatives if their location matches the origin or destination of the crossing, and no detour is needed to use them. Waiting time; safety; the fine imposed for jaywalking; personal security, and previous decisions are also variables that influence pedestrian behavior when crossing urban roads. These results can be helpful to urban planners and decision-makers interested in proposing appropriate pedestrian infrastructure. The data pooling technique and the inclusion of a cost-related variable (i.e., fine) allowed computing the willingness to pay and marginal substitution rates for attributes of the built environment and other characteristics associated with the crossing decision. Also, the inclusion of several crossing alternatives and situations allowed assessing pedestrian crossing preferences under different scenarios. [ABSTRACT FROM AUTHOR]

We compare the performance of two approaches for finding the optimal set of products to display to customers landing on Alibaba's two online marketplaces, Tmall and Taobao. Both approaches were placed online simultaneously and tested on real customers for one week. The first approach we test is Alibaba's current practice. This procedure embeds thousands of product and customer features within a sophisticated machine-learning algorithm that is used to estimate the purchase probabilities of each product for the customer at hand. Our second approach uses a featurized multinomial logit (MNL) model to predict purchase probabilities for each arriving customer. In this way, we use less sophisticated machinery to estimate purchase probabilities, but we employ a model that was built to capture customer purchasing behavior and, more specifically, substitution patterns. Our experiments show that despite the lower prediction power of our MNL-based approach, it generates significantly higher revenue per visit compared with the current machine-learning algorithm with the same set of features. We compare the performance of two approaches for finding the optimal set of products to display to customers landing on Alibaba's two online marketplaces, Tmall and Taobao. We conducted a large-scale field experiment, in which we randomly assigned 10,421,649 customer visits during a one-week-long period to one of the two approaches and measured the revenue generated per customer visit. The first approach we tested was Alibaba's current practice, which embeds product and customer features within a sophisticated machine-learning algorithm to estimate the purchase probabilities of each product for the customer at hand. The products with the largest expected revenue (revenue × predicted purchase probability) are then made available for purchase. Our second approach, which we developed and implemented in collaboration with Alibaba engineers, uses a featurized multinomial logit (MNL) model to predict purchase probabilities for each arriving customer. We used historical sales data to fit the MNL model, and then, for each arriving customer, we solved a cardinality-constrained assortment-optimization problem under the MNL model to find the optimal set of products to display. Our field experiments revealed that the MNL-based approach generated 5.17 renminbi (RMB) per customer visit, compared with the 4.04 RMB per customer visit generated by the machine-learning-based approach when both approaches were given access to the same set of the 25 most important features. This improvement represents a 28% gain in revenue per customer visit, which corresponds to a 4 million RMB improvement over the week in which the experiments were conducted. Motivated by the results of our initial field experiment, Alibaba then implemented a full-featured version of our MNL-based approach, which now serves the majority of customers in this setting. Using another small-scale field experiment, we estimate that our new MNL-based approach that utilizes the full feature set is able to increase Alibaba's annual revenue by 87.26 million RMB (12.42 million U.S. dollars). [ABSTRACT FROM AUTHOR]

This paper examines the impact of omitted variable bias (OVB) within consumer choice models on engineering design optimization solutions. Engineering products often have a multitude of attributes that influence consumers' purchasing decisions, many of which are difficult to include in revealed-preference models due to a lack of data. Correlations among these omitted variables and product attributes included in the model can bias demand parameter estimates. However, engineering design optimization studies typically do not account for this bias. We examine the influence consumer choice OVB can have on design optimization results. We first mathematically derive how OVB propagates into optimal design solutions and characterize properties of optimization problems that affect the magnitude of the resulting error in solutions. We then demonstrate the impact of OVB on optimal designs using a case study of automotive powertrain design optimization. In the case study, we estimate two sets of choice models: one using only "typically observed" vehicle attributes commonly found in the literature, and one with an additional set of "typically unobserved" attributes gathered from Edmunds.com. We find that the model with omitted variables leads to, in some scenarios, substantial bias in parameter estimates (5-143%), which propagates up to 21% error in the optimal engine size. [ABSTRACT FROM AUTHOR]

We study a stylized queueing model motivated by paid express lanes on highways. There are two parallel, observable first-come, first-served queues with finitely many servers: one queue has a faster service rate, but charges a fee to join, and the other is free but slow. Upon arrival, customers see the state of each queue and choose between them by comparing the respective disutility of time spent waiting, subject to random shocks. This framework encompasses both the multinomial logit and exponomial customer choice models. Using a fluid limit approximation, we give a detailed characterization of the equilibrium in this system. We show that social welfare is optimized when the express queue is exactly at (but not over) full capacity; however, in some cases, revenue is maximized by artificially creating congestion in the free queue. The latter behavior is caused by changes in the price elasticity of demand as the service capacity of the free queue fills up. [ABSTRACT FROM AUTHOR]