Pub Date : 2026-01-13DOI: 10.1016/j.jocm.2025.100585
Francisco Bahamonde-Birke , C. Angelo Guevara
Empirical evidence has established the existence of a phenomenon known as the decoy effect, which suggests that including irrelevant alternatives into a choice-set may affect the way in which the original alternatives are evaluated. In this paper we explore different ways to characterize the decoy effect and offer an in-depth discussion on the theoretical and empirical implications of the different modeling approaches. We also consider a stated preference experiment for which we model the phenomenon according to the assumptions of regret theory, emergent value, and prospect theory. Based on theoretical and empirical considerations, our results suggest that models based on prospect theory seem to outperform alternative behavioral paradigms to model the decoy effect.
{"title":"Which rubber duck makes the best decoy? Considering the decoy effect on the basis of different behavioral theories","authors":"Francisco Bahamonde-Birke , C. Angelo Guevara","doi":"10.1016/j.jocm.2025.100585","DOIUrl":"10.1016/j.jocm.2025.100585","url":null,"abstract":"<div><div>Empirical evidence has established the existence of a phenomenon known as the decoy effect, which suggests that including irrelevant alternatives into a choice-set may affect the way in which the original alternatives are evaluated. In this paper we explore different ways to characterize the decoy effect and offer an in-depth discussion on the theoretical and empirical implications of the different modeling approaches. We also consider a stated preference experiment for which we model the phenomenon according to the assumptions of regret theory, emergent value, and prospect theory. Based on theoretical and empirical considerations, our results suggest that models based on prospect theory seem to outperform alternative behavioral paradigms to model the decoy effect.</div></div>","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"58 ","pages":"Article 100585"},"PeriodicalIF":2.4,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145977675","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-09DOI: 10.1016/j.jocm.2025.100586
Peio Alcorta, Petr Mariel
We derive the Hessian matrix and asymptotic variance of linear-in-parameters latent class logit models. These analytical solutions enable precise estimation of standard errors, significantly improve post-estimation computational efficiency, and may offer deeper theoretical insights into model behaviour and stability.
{"title":"On the asymptotic variance of latent class logit models for discrete choice applications","authors":"Peio Alcorta, Petr Mariel","doi":"10.1016/j.jocm.2025.100586","DOIUrl":"10.1016/j.jocm.2025.100586","url":null,"abstract":"<div><div>We derive the Hessian matrix and asymptotic variance of linear-in-parameters latent class logit models. These analytical solutions enable precise estimation of standard errors, significantly improve post-estimation computational efficiency, and may offer deeper theoretical insights into model behaviour and stability.</div></div>","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"58 ","pages":"Article 100586"},"PeriodicalIF":2.4,"publicationDate":"2026-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145927114","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-23DOI: 10.1016/j.jocm.2025.100584
Wooje Seong , Hyunhong Choi , Yoonmo Koo
{"title":"Semi-dynamics of electric vehicle adoption based on strategic consumer choices: Preference statement-guided forward-looking behavior based on individual expectations","authors":"Wooje Seong , Hyunhong Choi , Yoonmo Koo","doi":"10.1016/j.jocm.2025.100584","DOIUrl":"10.1016/j.jocm.2025.100584","url":null,"abstract":"","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"58 ","pages":"Article 100584"},"PeriodicalIF":2.4,"publicationDate":"2025-12-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841393","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-11DOI: 10.1016/j.jocm.2025.100582
Jeff Tjiong , Thijs Dekker , Stephane Hess , Marek Giergiczny , Manuel Ojeda-Cabral , Mikołaj Czajkowski
Stated choice surveys commonly used in public policy appraisal – such as in transport or environmental economics – often contrast a ‘free’ status quo alternative against a range of (policy) interventions which can be implemented at a cost. Limited attention has, however, been paid to the fact that the ‘free’ nature of the status quo (SQ) alternative may make the SQ alternative overly attractive due to the zero-price (ZP) effect. The ZP effect is a well-established notion in behavioural economics explaining the phenomenon that individuals tend to over-react to free alternatives. We present an experimental design setup allowing the separation of the ZP effect from the SQ effect together with the identification of non-linear sensitivities to costs. Choices made by students between different mobile broadband packages are used for illustrational purposes. Our analysis shows that the ZP effect is significant and the observed preference to remain in the SQ is largely due to the ZP effect. In practice, this may lead to biased welfare estimates for public policy packages if the ZP effect is not explicitly accounted for. Moreover, we also show that misspecification of the functional form for cost can lead to significant bias in WTP estimates and the ZP and SQ effects.
{"title":"Capturing zero-price effects in stated choice surveys: implications for willingness-to-pay and welfare","authors":"Jeff Tjiong , Thijs Dekker , Stephane Hess , Marek Giergiczny , Manuel Ojeda-Cabral , Mikołaj Czajkowski","doi":"10.1016/j.jocm.2025.100582","DOIUrl":"10.1016/j.jocm.2025.100582","url":null,"abstract":"<div><div>Stated choice surveys commonly used in public policy appraisal – such as in transport or environmental economics – often contrast a ‘free’ status quo alternative against a range of (policy) interventions which can be implemented at a cost. Limited attention has, however, been paid to the fact that the ‘free’ nature of the status quo (SQ) alternative may make the SQ alternative overly attractive due to the zero-price (ZP) effect. The ZP effect is a well-established notion in behavioural economics explaining the phenomenon that individuals tend to over-react to free alternatives. We present an experimental design setup allowing the separation of the ZP effect from the SQ effect together with the identification of non-linear sensitivities to costs. Choices made by students between different mobile broadband packages are used for illustrational purposes. Our analysis shows that the ZP effect is significant and the observed preference to remain in the SQ is largely due to the ZP effect. In practice, this may lead to biased welfare estimates for public policy packages if the ZP effect is not explicitly accounted for. Moreover, we also show that misspecification of the functional form for cost can lead to significant bias in WTP estimates and the ZP and SQ effects.</div></div>","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"58 ","pages":"Article 100582"},"PeriodicalIF":2.4,"publicationDate":"2025-12-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145712328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-01DOI: 10.1016/j.jocm.2025.100583
Niousha Bagheri , Milad Ghasri , Michael Barlow
This paper presents RUM-NN, a neural network framework that is fully consistent with the Random Utility Maximisation (RUM) theory and designed to flexibly model discrete choice behaviour under a wide range of error distributions. RUM-NN contributes a flexible estimation approach to accommodate arbitrary error distributions. This enables the modelling of choice probabilities even when closed-form solutions are unavailable, accommodating arbitrary error structures, including correlated and non-conventional distributions. The proposed RUM-NN is introduced in both linear and non-linear structures. The linear version of RUM-NN retains interpretability similar to traditional econometric models, while the nonlinear extension enhances predictive flexibility by capturing complex relationships in the utility function. The performance of RUM-NN in parameter recovery and prediction accuracy is rigorously evaluated using synthetic datasets through Monte Carlo experiments. Additionally, RUM-NN is evaluated on the Swissmetro and the London Passenger Mode Choice (LPMC) datasets with different sets of distribution assumptions for the error component. The results demonstrate that RUM-NN under linear utility structure and IID Gumbel error terms can replicate the performance of Multinomial Logit model, but relaxing those constraints yields to superior performance for both Swissmetro and LMPC datasets. By introducing a novel estimation approach aligned with statistical theories, this study empowers econometricians to harness the advantages of neural network models. To facilitate the implementation of RUM-NN, a Python library has been developed and made publicly available.
{"title":"A neural estimation framework for discrete choice models with arbitrary error distributions","authors":"Niousha Bagheri , Milad Ghasri , Michael Barlow","doi":"10.1016/j.jocm.2025.100583","DOIUrl":"10.1016/j.jocm.2025.100583","url":null,"abstract":"<div><div>This paper presents RUM-NN, a neural network framework that is fully consistent with the Random Utility Maximisation (RUM) theory and designed to flexibly model discrete choice behaviour under a wide range of error distributions. RUM-NN contributes a flexible estimation approach to accommodate arbitrary error distributions. This enables the modelling of choice probabilities even when closed-form solutions are unavailable, accommodating arbitrary error structures, including correlated and non-conventional distributions. The proposed RUM-NN is introduced in both linear and non-linear structures. The linear version of RUM-NN retains interpretability similar to traditional econometric models, while the nonlinear extension enhances predictive flexibility by capturing complex relationships in the utility function. The performance of RUM-NN in parameter recovery and prediction accuracy is rigorously evaluated using synthetic datasets through Monte Carlo experiments. Additionally, RUM-NN is evaluated on the Swissmetro and the London Passenger Mode Choice (LPMC) datasets with different sets of distribution assumptions for the error component. The results demonstrate that RUM-NN under linear utility structure and IID Gumbel error terms can replicate the performance of Multinomial Logit model, but relaxing those constraints yields to superior performance for both Swissmetro and LMPC datasets. By introducing a novel estimation approach aligned with statistical theories, this study empowers econometricians to harness the advantages of neural network models. To facilitate the implementation of RUM-NN, a Python library has been developed and made publicly available.</div></div>","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"57 ","pages":"Article 100583"},"PeriodicalIF":2.4,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145684484","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-21DOI: 10.1016/j.jocm.2025.100581
Ambuj Sriwastava, Peter Reichert
The elicitation and quantification of preferences of individuals or aggregated preferences of stakeholders or samples of the population are crucial for decision support. This can be done by statistically evaluating the results of discrete choice inquiries using a parameterized value function. When doing this with Bayesian inference, the specification of a prior can be challenging as it may be difficult to find similar cases to transfer knowledge. This makes it particularly important to be informed about the sensitivity of the results to the choice of the prior. This can be done by computing posteriors for different plausible priors and analyzing differences between them. This is infeasible for a large number of priors. This paper proposes the application of Density Ratio Classes, which sandwich non-normalized prior densities between specified lower and upper functional bounds. In this study, differences among posteriors resulting from priors in such a class are analyzed by comparing marginal posterior credible intervals. We compute “outer” credible intervals that range from the minimum of all lower bounds to the maximum of all upper bounds of marginal posterior credible intervals with the same quantile bounds resulting from the priors in the density ratio class. The methodology is easy to implement and only requires one Markov chain of the posterior resulting from the upper function. We provide an R package “DRclass” that supports such implementations. Theoretical considerations and three case studies provide illustration and guidance about constructing the prior density ratio class.
{"title":"Sensitivity analysis of Bayesian estimates of value function parameters to priors using imprecise probabilities","authors":"Ambuj Sriwastava, Peter Reichert","doi":"10.1016/j.jocm.2025.100581","DOIUrl":"10.1016/j.jocm.2025.100581","url":null,"abstract":"<div><div>The elicitation and quantification of preferences of individuals or aggregated preferences of stakeholders or samples of the population are crucial for decision support. This can be done by statistically evaluating the results of discrete choice inquiries using a parameterized value function. When doing this with Bayesian inference, the specification of a prior can be challenging as it may be difficult to find similar cases to transfer knowledge. This makes it particularly important to be informed about the sensitivity of the results to the choice of the prior. This can be done by computing posteriors for different plausible priors and analyzing differences between them. This is infeasible for a large number of priors. This paper proposes the application of Density Ratio Classes, which sandwich non-normalized prior densities between specified lower and upper functional bounds. In this study, differences among posteriors resulting from priors in such a class are analyzed by comparing marginal posterior credible intervals. We compute “outer” credible intervals that range from the minimum of all lower bounds to the maximum of all upper bounds of marginal posterior credible intervals with the same quantile bounds resulting from the priors in the density ratio class. The methodology is easy to implement and only requires one Markov chain of the posterior resulting from the upper function. We provide an R package “DRclass” that supports such implementations. Theoretical considerations and three case studies provide illustration and guidance about constructing the prior density ratio class.</div></div>","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"57 ","pages":"Article 100581"},"PeriodicalIF":2.4,"publicationDate":"2025-10-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145362556","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-18DOI: 10.1016/j.jocm.2025.100578
Cheng-Yu Hung , Peter Kurz , Roger A. Bailey , Joel Huber , Greg M. Allenby
The validity of using conjoint analysis to conduct an economic evaluation of product characteristics rests on the inclusion of brand names, prices, and an outside “no-choice” option in the choice task. The no-choice option is included in case respondents determine that some other offering, not included in the conjoint choice task, is preferred to those that are included and that it would be better to hold onto their money and not make a purchase at that time. Selecting the no-choice option assumes that respondents have some level of knowledge of the value and prices of goods in the market. In this paper, we show that survey respondents may lack this information and make inferences about market prices from the conjoint exercise itself. This learning effect is especially problematic for new products for which a set of reference prices do not yet exist, but can also be problematic in established markets that are familiar. We discuss results from two sets of conjoint experiments, one in a new product category conducted in three countries in Europe, and another in an established category in the United States involving three experimental conditions that inform respondents about products and prices available in the marketplace. We find that the lack of knowledge of competitive offerings and prices affects estimates of brand values but not the value of other product features. In addition, we discuss aspects of how a well-designed conjoint study mitigates the effects of this type of learning in conjoint analysis.
{"title":"Re-examining the no-choice option in conjoint analysis","authors":"Cheng-Yu Hung , Peter Kurz , Roger A. Bailey , Joel Huber , Greg M. Allenby","doi":"10.1016/j.jocm.2025.100578","DOIUrl":"10.1016/j.jocm.2025.100578","url":null,"abstract":"<div><div>The validity of using conjoint analysis to conduct an economic evaluation of product characteristics rests on the inclusion of brand names, prices, and an outside “no-choice” option in the choice task. The no-choice option is included in case respondents determine that some other offering, not included in the conjoint choice task, is preferred to those that are included and that it would be better to hold onto their money and not make a purchase at that time. Selecting the no-choice option assumes that respondents have some level of knowledge of the value and prices of goods in the market. In this paper, we show that survey respondents may lack this information and make inferences about market prices from the conjoint exercise itself. This learning effect is especially problematic for new products for which a set of reference prices do not yet exist, but can also be problematic in established markets that are familiar. We discuss results from two sets of conjoint experiments, one in a new product category conducted in three countries in Europe, and another in an established category in the United States involving three experimental conditions that inform respondents about products and prices available in the marketplace. We find that the lack of knowledge of competitive offerings and prices affects estimates of brand values but not the value of other product features. In addition, we discuss aspects of how a well-designed conjoint study mitigates the effects of this type of learning in conjoint analysis.</div></div>","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"57 ","pages":"Article 100578"},"PeriodicalIF":2.4,"publicationDate":"2025-10-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145321469","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-04DOI: 10.1016/j.jocm.2025.100576
Stephen McCarthy, Fatemeh Naqavi, Anders Karlström
This paper applies recursive logit (RL) to model activity-trip chaining behaviour. We present a comparison between two approaches to applying the RL model in this context. In the first ‘sequential’ approach, agents form a trip chain by making a sequence of joint choices of activity location (i.e. trip destination) and travel mode, ending the chain by choosing to return home. The second ‘dynamic’ approach adds a time variable. Its agents form a full-day activity/travel schedule by making a sequence of choices either to continue the current activity for a fixed timestep or make a joint choice of new activity location and travel mode. We estimate parameters for both models using data from a Stockholm travel survey and validate model simulations against observed data. The models reproduce patterns of observed behaviour beyond their estimated parameters, including different types of trip chains and the spatial distribution of activities. While the dynamic model is advantageous in its ability to predict agent schedules, reflect time-varying travel conditions and endogenously represent space–time constraints, it does not surpass the simpler sequential model on mutual areas of trip chaining behaviour. We conclude that the RL model is well-suited to model trip chaining behaviour, and that the simpler sequential approach may be appropriate for many modelling purposes.
{"title":"Recursive logit models for dynamic versus sequential trip chaining","authors":"Stephen McCarthy, Fatemeh Naqavi, Anders Karlström","doi":"10.1016/j.jocm.2025.100576","DOIUrl":"10.1016/j.jocm.2025.100576","url":null,"abstract":"<div><div>This paper applies recursive logit (RL) to model activity-trip chaining behaviour. We present a comparison between two approaches to applying the RL model in this context. In the first ‘sequential’ approach, agents form a trip chain by making a sequence of joint choices of activity location (i.e. trip destination) and travel mode, ending the chain by choosing to return home. The second ‘dynamic’ approach adds a time variable. Its agents form a full-day activity/travel schedule by making a sequence of choices either to continue the current activity for a fixed timestep or make a joint choice of new activity location and travel mode. We estimate parameters for both models using data from a Stockholm travel survey and validate model simulations against observed data. The models reproduce patterns of observed behaviour beyond their estimated parameters, including different types of trip chains and the spatial distribution of activities. While the dynamic model is advantageous in its ability to predict agent schedules, reflect time-varying travel conditions and endogenously represent space–time constraints, it does not surpass the simpler sequential model on mutual areas of trip chaining behaviour. We conclude that the RL model is well-suited to model trip chaining behaviour, and that the simpler sequential approach may be appropriate for many modelling purposes.</div></div>","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"57 ","pages":"Article 100576"},"PeriodicalIF":2.4,"publicationDate":"2025-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221364","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-04DOI: 10.1016/j.jocm.2025.100580
Kun Huang, Xin Ye, Mengyi Wang
Estimating a valid correlation matrix with structural restrictions presents significant challenges, particularly in ensuring positive definiteness and enforcing zero-correlation constraints. Traditional approaches, such as the Cholesky decomposition, often suffer from numerical instability and convergence failures in these settings. This paper introduces a novel Cholesky-based parameterization that effectively addresses these issues by allowing zero constraints while maintaining positive definiteness and unit diagonal elements. Through extensive Monte Carlo simulations, we demonstrate that the proposed method outperforms the existing spherical parameterization approach, achieving superior convergence rates, enhanced estimation accuracy, and robustness under high-correlation scenarios. An empirical application on non-commuters’ activity participation in Shanghai further validates the practical effectiveness of the proposed method, showcasing its ability to capture complex behavioral relationships while ensuring stable estimation. The results suggest that the proposed parameterization provides a reliable and computationally efficient alternative for correlation matrix estimation in multivariate models.
{"title":"A special Cholesky-based parameterization for estimation of restricted correlation matrices","authors":"Kun Huang, Xin Ye, Mengyi Wang","doi":"10.1016/j.jocm.2025.100580","DOIUrl":"10.1016/j.jocm.2025.100580","url":null,"abstract":"<div><div>Estimating a valid correlation matrix with structural restrictions presents significant challenges, particularly in ensuring positive definiteness and enforcing zero-correlation constraints. Traditional approaches, such as the Cholesky decomposition, often suffer from numerical instability and convergence failures in these settings. This paper introduces a novel Cholesky-based parameterization that effectively addresses these issues by allowing zero constraints while maintaining positive definiteness and unit diagonal elements. Through extensive Monte Carlo simulations, we demonstrate that the proposed method outperforms the existing spherical parameterization approach, achieving superior convergence rates, enhanced estimation accuracy, and robustness under high-correlation scenarios. An empirical application on non-commuters’ activity participation in Shanghai further validates the practical effectiveness of the proposed method, showcasing its ability to capture complex behavioral relationships while ensuring stable estimation. The results suggest that the proposed parameterization provides a reliable and computationally efficient alternative for correlation matrix estimation in multivariate models.</div></div>","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"57 ","pages":"Article 100580"},"PeriodicalIF":2.4,"publicationDate":"2025-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145267465","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-03DOI: 10.1016/j.jocm.2025.100579
Kyungah Kim , Jongsu Lee , Junghun Kim
Setting an appropriate reference point is crucial in reference-dependent choice modeling, as it directly influences the reliability of utility estimates and the interpretation of consumer decision-making. However, many prior studies have relied on generalized or fixed reference points—such as status quo or past experiences—without accounting for individual-level heterogeneity. To address this limitation, this study proposes a reference-dependent artificial neural network (RD-ANN) that integrates the structure of reference-dependent choice models into a neural network framework. RD-ANN is designed to learn individual- and alternative-specific reference points based on consumer and alternative attributes, thereby providing a flexible and data-driven approach to reference point estimation. Empirical validation using smartphone and automobile choice data shows that RD-ANN outperforms benchmark models in various predictive performance metrics including accuracy, recall, precision, and F1 score. The model also captures behavioral patterns such as brand loyalty and status quo bias more effectively. In the empirical contexts considered, RD-ANN was found to better reflect consumer heterogeneity and may help provide more accurate estimates of price sensitivity compared to models using a fixed status quo reference point. These findings suggest that the proposed approach offers a promising direction for integrating behavioral theory and machine learning in discrete choice modeling.
{"title":"Beyond the status quo: Leveraging reference-dependent theory in a neural network for consumer choice analysis","authors":"Kyungah Kim , Jongsu Lee , Junghun Kim","doi":"10.1016/j.jocm.2025.100579","DOIUrl":"10.1016/j.jocm.2025.100579","url":null,"abstract":"<div><div>Setting an appropriate reference point is crucial in reference-dependent choice modeling, as it directly influences the reliability of utility estimates and the interpretation of consumer decision-making. However, many prior studies have relied on generalized or fixed reference points—such as status quo or past experiences—without accounting for individual-level heterogeneity. To address this limitation, this study proposes a reference-dependent artificial neural network (RD-ANN) that integrates the structure of reference-dependent choice models into a neural network framework. RD-ANN is designed to learn individual- and alternative-specific reference points based on consumer and alternative attributes, thereby providing a flexible and data-driven approach to reference point estimation. Empirical validation using smartphone and automobile choice data shows that RD-ANN outperforms benchmark models in various predictive performance metrics including accuracy, recall, precision, and F1 score. The model also captures behavioral patterns such as brand loyalty and status quo bias more effectively. In the empirical contexts considered, RD-ANN was found to better reflect consumer heterogeneity and may help provide more accurate estimates of price sensitivity compared to models using a fixed status quo reference point. These findings suggest that the proposed approach offers a promising direction for integrating behavioral theory and machine learning in discrete choice modeling.</div></div>","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"57 ","pages":"Article 100579"},"PeriodicalIF":2.4,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221365","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号