Multinomial Choice Research Articles

A Joint Mixed-outcome and Instrumental Variable Model for Addressing Endogeneity in Multinomial Choice and Its Consideration Set Formation

A Joint Mixed-outcome and Instrumental Variable Model for Addressing Endogeneity in Multinomial Choice and Its Consideration Set Formation

Moment inequalities for multinomial choice with fixed effects

This paper proposes a new approach to identification of the semiparametric multinomial choice model with fixed effects. The framework employed is the semiparametric version of the traditional multinomial logit with the fixed‐effects model (Chamberlain (1980)). This semiparametric multinomial choice model places no restrictions on either the joint distribution of the random utility disturbances across choices or their within group (or across time) correlations. We show that a novel within‐group comparison leads to a set of conditional moment inequalities. Our main finding shows that the derived conditional moment inequalities yield the sharp identified set for the random utility covariate index, while avoiding the incidental parameter problem. Specializing this result to the binary choice case shows that Manski (1987)'s conditional moment inequalities still lead to sharp bounds without restrictions on covariates.

Semiparametric Bayesian estimation of dynamic discrete choice models

We propose a tractable semiparametric estimation method for structural dynamic discrete choice models. The distribution of additive utility shocks in the proposed framework is modeled by location-scale mixtures of extreme value distributions with varying numbers of mixture components. Our approach exploits the analytical tractability of extreme value distributions in the multinomial choice settings and the flexibility of the location-scale mixtures. We implement the Bayesian approach to inference using Hamiltonian Monte Carlo and an approximately optimal reversible jump algorithm. In our simulation experiments, we show that the standard dynamic logit model can deliver misleading results, especially about counterfactuals, when the shocks are not extreme value distributed. Our semiparametric approach delivers reliable inference in these settings. We develop theoretical results on approximations by location-scale mixtures in an appropriate distance and posterior concentration of the set identified utility parameters and the distribution of shocks in the model.

Patient preferences, referral process, and access to specialized care. Is patient choice constrained?

ABSTRACT In most developed countries, patients have been encouraged to elect their preferred choice of health care provider. However, this is different for specialized care, where the patient’s referral could be defined as a two-stage decision process and their options are pre-selected by their general practitioner (GP). In this study, we estimate patient preferences while controlling for the pre-selection procedure, and we investigate whether patients are actively choosing their provider for cancer care. The French national hospital discharge database (Programme de Médicalisation des Systèmes d’Information, PMSI – MCO 2017) has been used for investigation. We estimated a multinomial choice model when choice sets are in fact unobserved, which is assumed to identify patient preferences, in a revealed preferences framework. Our findings provide evidence that patients consider factors other than distance to select their provider. The patient – hospital distance as well as the specialization profile of providers appears to be internalized in the pre-selection process, while patients rather consider waiting times, hospital quality, and other provider attributes to make their final choice. We also found that patients would be treated in higher-quality hospitals if they had the opportunity to choose among all available providers.

Endogeneity in weakly separable models without monotonicity

We identify and estimate treatment effects when potential outcomes are weakly separable with a binary endogenous treatment. Vytlacil and Yildiz (2007) proposed an identification strategy that exploits the mean of observed outcomes, but their approach requires a monotonicity condition. In comparison, we exploit full information in the entire outcome distribution, instead of just its mean. As a result, our method does not require monotonicity and is also applicable to general settings with multiple indices. We provide examples where our approach can identify treatment effect parameters of interest whereas existing methods would fail. These include models where potential outcomes depend on multiple unobserved disturbance terms, such as a Roy model, a multinomial choice model, as well as a model with endogenous random coefficients. We establish consistency and asymptotic normality of our estimators.

Modelling driver's response to demand management strategies for electric vehicle charging in Australia

Electric vehicles (EVs) can help decarbonise transport as long as the energy sector successfully balances the electricity supply with the demand and maximises the use of renewable resources. For that, demand-side management strategies are necessary to encourage users to charge at certain times through time-of-use (ToU) tariffs or to control the load provided for charging (i.e., supplier-managed charging). However, adapting to different charging times can disrupt users' schedule flexibility. This study investigates consumer preferences for smart charging technology and control (user-managed or supplier-managed) and responses to progressive ToU tariff discounts for guiding changes in EV charging time in Australia. We analyse the potential of ToU tariffs in shifting demand to late-night hours and around midday when there is a peak in solar energy generation. Based on a sample of 994 drivers (including 97 EV owners), we estimated a multinomial choice model to identify key predictors of individual preference for smart charging management and a bivariate ordered model to investigate consumer response to time-of-use discounts. The results show that activity-travel behaviour is an important predictor of both demand-side management strategies. Consumers willing to change EV charging time in response to ToU tariffs are likely to have more flexible schedules, while those with more time constraints seek the practical benefits of smart charging. Current EV owners have higher propensities than potential adopters to choose supplier-managed smart charging and charge around midday in response to ToU tariffs. This indicates that trials with current EV owners may overestimate acceptance of these strategies. Our findings show that while monetary incentives can successfully shift an important share of consumers to night-time hours, these incentives are not very effective in shifting to midday charging. Synergy in formulating transport and energy demand strategies may be necessary to facilitate such a shift.

The odds of South African agriculture using wildlife ranching to adapt to climate change

In this paper, we explore the role of wildlife in climate change adaptation, especially in areas used predominantly for livestock production in South Africa. Using a sample of 3 449 wildlife and livestock ranches, we estimate a multinomial choice model of various ranching options in these areas. The results indicate that mixed wildlife-livestock ranches are less vulnerable to climate change when compared to ranches with only wildlife or only livestock. However, given the ranching options, the revenues of ranches with only wildlife are higher compared to other ranches. The results further show that an increase in temperature will influence more livestock farmers, especially those farming with cattle, to change their land use from livestock to wildlife ranching. Using climate models, we establish that livestock farmers in the Eastern Cape province of South Africa will be affected most by climate change and will subsequently change their land use.

A Bayesian instrumental variable model for multinomial choice with correlated alternatives

Endogeneity and correlated alternatives are major concerns to be addressed in travel behavior analysis. However, these issues have rarely been dealt with simultaneously in advanced discrete choice models. This study proposes a multinomial probit model that incorporates the instrumental variable method, namely, a fully parametric instrumental variable model for a multinomial choice. The proposed model has the following three characteristics: (1) it allows binary and/or continuous endogenous variables; (2) it allows any number of instrumental variables in each alternative; and (3) it allows positive and/or negative correlations between any choice alternatives. For parameter estimation, we also propose a Bayesian Markov chain Monte Carlo algorithm that can be accommodated in more extended model structures. The simulation study demonstrates that the proposed model addresses endogeneity while allowing correlations between the choice alternatives. Meanwhile, the simulation also implies that the users need to pay attention to the setting of the prior distribution when an endogenous variable of interest is binary, even if the sample size is moderate. The proposed model will be a useful tool in disciplines in which both endogeneity and correlations between choice alternatives are major concerns.

Multinomial Choice with Social Interactions: Occupations in Victorian London

Abstract We study the importance of social interactions on occupational choice in Victorian London using a multinomial choice model within an incomplete social network. Individuals form heterogeneous rational expectations about their peers' behaviors, taking into account their characteristics and the strength of their ties. We show the conditions under which the endogenous, exogenous, and correlated effects can be identified and a unique equilibrium can be established, Using a novel data set, we proxy social groups by parish boundaries and strength of ties by geographic distances, Our results show the importance of the endogenous effects and reveal distinct effects by occupation.

Heterogeneity in general multinomial choice models

Different voters behave differently at the polls, different students make different university choices, or different countries choose different health care systems. Many research questions important to social scientists concern choice behavior, which involves dealing with nominal dependent variables. Drawing on the principle of maximum random utility, we propose applying a flexible and general heterogeneous multinomial logit model to study differences in choice behavior. The model systematically accounts for heterogeneity that classical models do not capture, indicates the strength of heterogeneity, and permits examining which explanatory variables cause heterogeneity. As the proposed approach allows incorporating theoretical expectations about heterogeneity into the analysis of nominal dependent variables, it can be applied to a wide range of research problems. Our empirical example uses individual-level survey data to demonstrate the benefits of the model in studying heterogeneity in electoral decisions.

Identification and Estimation of Multinomial Choice Models with Latent Special Covariates

Identification of multinomial choice models is often established by using special covariates that have full support. This article shows how these identification results can be extended to a large class of multinomial choice models when all covariates are bounded. I also provide a new -consistent asymptotically normal estimator of the finite-dimensional parameters of the model.

The use of multinomial choice analysis in international business research

Many strategic decisions made by firms involve a choice among several discrete alternatives. International Business (IB) scholars are often interested in modelling the factors that potentially influence these (multinomial) choices: these factors might include not only characteristics of the firms making the choices but also attributes of the alternative choices. This paper provides a succinct and intuitive introduction to the possible applications of multinomial choice models in IB research. We briefly outline the theory behind discrete choice modelling, and then explain how multinomial choice models may be estimated (including how the datasets need to be formatted) and how the significance of the coefficient estimates as well as the diagnostic statistics may be tested and interpreted.

Partial Identification and Estimation of Semiparametric Ordered Response Models with Interval Regressor Data*

AbstractIn many micro‐data studies, the dependent variable often involves ordered categories and at least one regressor is measured by the interval rather than the precise value. This paper considers partial identification of such an ordered response model when point identification fails. We show the identified set of non‐intercept coefficients is the intersection of those for composite binary response models. We also propose a generalized modified maximum score set (GMMS) estimator. A practical implication of our finding is researchers can shrink the identified set and obtain more precise inference by designing as many as categories of response in a questionnaire during data collection. Another advantage is our theoretical finding can be used to infer the identified region in the multinomial choice model. A Monte Carlo study is conducted to illustrate the main finding in a finite sample. Finally, we apply GMMS estimator to a job satisfaction study using US data with the interval income.

Sparse assortment personalization in high dimensions

The data-driven conditional multinomial logit choice model with customer features performs well in the assortment personalization problem when the low-rank structure of the parameter matrix is considered. However, despite recent theoretical and algorithmic advances, parameter estimation in the choice model still poses a challenging task, especially when there are more predictors than observations. For this reason, we suggest a penalized likelihood approach based on a feature matrix to recover the sparse structure from populations and products toward the assortment. Our proposed method considers simultaneously low-rank and sparsity structures, which can further reduce model complexity and improve its estimation and prediction accuracy. A new algorithm, sparse factorial gradient descent (SFGD), was proposed to estimate the parameter matrix, which has high interpretability and efficient computing performance. As a first-order method, the SFGD works well in high-dimensional scenarios because of the absence of the Hessian matrix. Simulation studies show that the SFGD algorithm outperforms state-of-the-art methods in terms of estimation, sparsity recovery, and average regret. We also demonstrate the effectiveness of our proposed method using advertising behavior data analysis.

A Bayesian Instrumental Variable Model for Multinomial Choice with Correlated Alternatives

A Bayesian Instrumental Variable Model for Multinomial Choice with Correlated Alternatives

Reading Between the Stars: Understanding the Effects of Online Customer Reviews on Product Demand

Problem definition: Many studies have examined quantitative customer reviews (i.e., star ratings) and found them to be a reliable source of information that has a positive effect on product demand. Yet the effect of qualitative customer reviews (i.e., text reviews) on demand has been less thoroughly studied, and it is not known whether (or how) the sentiment expressed in text reviews moderates the influence of star ratings on product demand. We are therefore led to examine how the interplay between review sentiment and star ratings affects product demand. Academic/practical relevance: Consumer perceptions of product quality and how they are shared via customer reviews are of extreme relevance to the firm, but we still do not understand how product demand is affected by the quantitative and qualitative aspects of customer reviews. Our paper seeks to fill this critical gap in the literature by analyzing star ratings, the sentiment of customer reviews, and their interaction. Methodology: Using 2002–2013 data for the U.S. automobile market, we investigate empirically the impact of star ratings and review sentiment on product demand. Thus, we estimate an aggregated multinomial choice model after performing a machine learning–based sentiment analysis on the entire corpus of customer reviews included in our sample. We take advantage of a quasi-exogenous shock to establish a causal link between online reviews and product demand. Results: We find robust empirical evidence that (i) review sentiment and star ratings both have a decreasingly positive effect on product demand and (ii) the effect (on demand) of their interaction suggests that the two components of reviews are complements. Positive sentiments in text reviews increase the positive effect of ratings when the effect of ratings is decidedly positive while they also compensate for the tendency of consumers to discount extremely high star ratings. Managerial implications: The firm should pay greater attention to quantitative and qualitative customer reviews to better understand how consumers perceive the quality of its offerings.

Optimal linear discriminators for the discrete choice model in growing dimensions

Manski’s celebrated maximum score estimator for the discrete choice model, which is an optimal linear discriminator, has been the focus of much investigation in both the econometrics and statistics literatures, but its behavior under growing dimension scenarios largely remains unknown. This paper addresses that gap. Two different cases are considered: p grows with n but at a slow rate, that is, p/n→0; and p≫n (fast growth). In the binary response model, we recast Manski’s score estimation as empirical risk minimization for a classification problem, and derive the ℓ2 rate of convergence of the score estimator under a new transition condition in terms of a margin parameter that calibrates the level of difficulty of the estimation problem. We also establish upper and lower bounds for the minimax ℓ2 error in the binary choice model that differ by a logarithmic factor, and construct a minimax-optimal estimator in the slow growth regime. Some extensions to the multinomial choice model are also considered.

Determinants of Government Membership at the Subnational Level: Empirical Evidence from Large Cities in Germany (1999–2016)

AbstractCoalition governments prevail at the European subnational level. Although some studies explain the formation of subnational government coalitions, we know little about the determinants of individual parties' likelihood of joining such coalitions. This article aims to fill this gap in empirical and theoretical ways. It shows that an important institutional constraint matters for political actors' strategies when forming subnational coalitions: the party affiliation of the directly elected head of the executive. Being the party of the head of the executive or being ideologically close to that party significantly increases a party's likelihood of joining a coalition. The empirical evidence results from multinomial choice models using a novel data set on subnational parties' likelihood of joining 92 coalition governments at the local level in Germany between 1999 and 2016. The findings have substantive implications for subnational institutional settings resembling ‘mixed’ political systems (i.e. neither purely presidential nor purely parliamentarian).

Butterfly biodiversity in the city is driven by the interaction of the urban landscape and species traits: a call for contextualised management

ContextUrbanisation is an environmental filter for many species that leads to community homogenisation, with a few species inhabiting isolated patches (e.g. public and private gardens and parks) embedded within the urban landscape. Promoting biodiversity in urban areas requires understanding which species traits allow species to survive the urban landscape.ObjectivesThe objective of this study was to assess how species traits and landscape factors combine to allow species functional groups to live in the city.MethodsWe used butterfly count data collected by volunteers in 24 gardens of Barcelona city, during 2018 and 2019. Species were clustered in functional groups according to their traits. We applied a multinomial choice model to test for the effect of the landscape on the different functional groups.ResultsThree functional groups became prevalent in the city while a fourth, containing most sedentary specialist species, was filtered out. Although the observed groups had similar species richness, abundances varied depending on urban landscape characteristics. Specialist sedentary specialists and medium mobile species were all favoured by patch connectivity; while the presence of mobile generalist species was only enhanced by habitat quality. Our results indicate that butterfly communities are more diverse in highly connected gardens.ConclusionsOur study highlights the need of contextualised management with actions accounting for the species functional groups, rather than a management focused on general species richness. It demonstrates that urban landscape planning must focus on improving connectivity inside the city in order to diversify the community composition.

Community-embedded experiential learning and adoption of conservation farming practices in Eastern and Southern Africa

In low-income country agriculture, there is frequently limited information about the comparative effectiveness of alternative extension modalities for delivering information on new farming methods. This paper analyzes the impact of community-embedded farmer education programs on the adoption of conservation farming. A multinomial choice model is applied on survey data from Kenya, Malawi, Mozambique, and Tanzania to estimate the relative risk ratios (RRR), a measure of the probability of adopting the conservation agriculture practices studied, conditional on attending specific extension activities in the farmers' community. The results show that the likelihood of adoption was generally high as measured by RRRs between 10 and 14. Suggestive of the effectiveness of experiential learning methods, attending training sessions had consistently higher RRRs of adopting conservation farming practices. The positive role of community-based extension programs potentially means that the dissemination of conservation agriculture requires investments by governments and development organizations in community learning sites. Investments in programs that promote experiential, hands-on and farmer self-learning can be consequential in enabling the diffusion of better farming methods.