|
on Discrete Choice Models |
| By: | Daniel F. Villarraga; Ricardo A. Daziano |
| Abstract: | Discrete choice models (DCMs) are used to analyze individual decision-making in contexts such as transportation choices, political elections, and consumer preferences. DCMs play a central role in applied econometrics by enabling inference on key economic variables, such as marginal rates of substitution, rather than focusing solely on predicting choices on new unlabeled data. However, while traditional DCMs offer high interpretability and support for point and interval estimation of economic quantities, these models often underperform in predictive tasks compared to deep learning (DL) models. Despite their predictive advantages, DL models remain largely underutilized in discrete choice due to concerns about their lack of interpretability, unstable parameter estimates, and the absence of established methods for uncertainty quantification. Here, we introduce a deep learning model architecture specifically designed to integrate with approximate Bayesian inference methods, such as Stochastic Gradient Langevin Dynamics (SGLD). Our proposed model collapses to behaviorally informed hypotheses when data is limited, mitigating overfitting and instability in underspecified settings while retaining the flexibility to capture complex nonlinear relationships when sufficient data is available. We demonstrate our approach using SGLD through a Monte Carlo simulation study, evaluating both predictive metrics--such as out-of-sample balanced accuracy--and inferential metrics--such as empirical coverage for marginal rates of substitution interval estimates. Additionally, we present results from two empirical case studies: one using revealed mode choice data in NYC, and the other based on the widely used Swiss train choice stated preference data. |
| Date: | 2025–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2505.18077 |
| By: | Yoosoon Chang; Joon Y. Park; Guo Yan |
| Abstract: | This paper revisits the slope consistency of QMLE for binary choice models. Ruud (1983, \emph{Econometrica}) introduced a set of conditions under which QMLE may yield a constant multiple of the slope coefficient of binary choice models asymptotically. However, he did not fully establish slope consistency of QMLE, which requires the existence of a positive multiple of slope coefficient identified as an interior maximizer of the population QMLE likelihood function over an appropriately restricted parameter space. We fill this gap by providing a formal proof for slope consistency under the same set of conditions for any binary choice model identified as in Horowitz (1992, \emph{Econometrica}). Our result implies that the logistic regression, which is used extensively in machine learning to analyze binary outcomes associated with a large number of covariates, yields a consistent estimate for the slope coefficient of binary choice models under suitable conditions. |
| Date: | 2025–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2505.02327 |
| By: | Stelios Arvanitis (Department of Economics, AUEB) |
| Abstract: | This paper utilizes a Banach-type fixed point theorem in a functorial context to develop Universal Choice Spaces for addressing decision problems, focusing on expected utility and preference uncertainty. This generates an infinite sequence of optimal selection problems involving probability measures on utility sets. Each solution at a given stage addresses the preference ambiguity from the previous stage, enabling optimal choices at that level. The Universal Choice Space is characterized as a collection of finite-dimensional vectors of probability distributions, with the mth component being an arbitrary probability measure relevant to the mth stage of the problem. The space is derived as the canonical fixed point of a suitable endofunctor on an enriched category and simultaneously as the colimit of the sequence of iterations of this functor, starting from a suitable object. |
| Keywords: | Expected utility, ambiguity of preferences, infinite regress, enriched category, endofunctor, canonical fixed point, initial algebra, colimit, universal choice space |
| JEL: | D81 |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:qed:wpaper:1534 |
| By: | Jian, Wenze; Zhong, Ziqi |
| Abstract: | This study investigates how eco-positioning strategies influence consumers’ evaluations of fashion brands, their willingness to pay for eco-friendly fashion products, and their sustainable fashion consumption intentions. Based on the Theory of Planned Behavior and the Value-Belief-Norm Theory, this study constructs an integrated analysis framework. Data were collected through a structured online experiment, wherein participants completed three randomized experimental modules, each testing a distinct dependent variable. Within each module, participants were independently assigned to different eco-positioning stimuli. The results indicate that eco-positioning significantly affects brand evaluation and purchase intention, with process-related eco-positioning having a stronger effect. High brand familiarity enhances the effectiveness of eco-positioning strategies. Strong eco-positioning remarkably increases consumers’ willingness to pay, with perceived environmental sustainability playing an important mediating role. Additionally, sustainable fashion consumption intention under eco-positioning advertising is markedly higher than that under other advertising conditions, with environmental concern and fashion involvement acting as key moderating factors. |
| Keywords: | fashion marketing; eco-positioning; consumer perception; sustainability |
| JEL: | L81 |
| Date: | 2025–05–21 |
| URL: | https://d.repec.org/n?u=RePEc:ehl:lserod:128137 |
| By: | Yuexin Liao; Kota Saito; Alec Sandroni |
| Abstract: | This paper studies when discrete choice data involving aggregated alternatives such as categorical data or an outside option can be rationalized by a random utility model (RUM). Aggregation introduces ambiguity in composition: the underlying alternatives may differ across individuals and remain unobserved by the analyst. We characterize the observable implications of RUMs under such ambiguity and show that they are surprisingly weak, implying only monotonicity with respect to adding aggregated alternatives and standard RUM consistency on unaggregated menus. These are insufficient to justify the use of an aggregated RUM. We identify two sufficient conditions that restore full rationalizability: non-overlapping preferences and menu-independent aggregation. Simulations show that violations of these conditions generate estimation bias, highlighting the practical importance of how aggregated alternatives are defined. |
| Date: | 2025–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2506.00372 |
| By: | Wooyong Lee |
| Abstract: | I study panel data linear models with predetermined regressors (such as lagged dependent variables) where coefficients are individual-specific, allowing for heterogeneity in the effects of the regressors on the dependent variable. I show that the model is not point-identified in a short panel context but rather partially identified, and I characterize the identified sets for the mean, variance, and CDF of the coefficient distribution. This characterization is general, accommodating discrete, continuous, and unbounded data, and it leads to computationally tractable estimation and inference procedures. I apply the method to study lifecycle earnings dynamics among U.S. households using the Panel Study of Income Dynamics (PSID) dataset. The results suggest substantial unobserved heterogeneity in earnings persistence, implying that households face varying levels of earnings risk which, in turn, contribute to heterogeneity in their consumption and savings behaviors. |
| Date: | 2025–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2505.01600 |
| By: | Mingshi Chen; Tracy Xiao Liu; You Shan; Shu Wang; Songfa Zhong; Yanju Zhou |
| Abstract: | Choice consistency with utility maximization is a fundamental assumption in economic analysis and is extensively measured across various contexts. Here we investigate the generalizability of consistency measures derived from purchasing decisions using supermarket scanner data and budgetary decisions from lab-in-the-field experiments. We observe a lack of correlation between consistency scores from supermarket purchasing decisions and those from risky decisions in the experiment. However, we observe moderate correlations among experimental tasks and low to moderate correlations across purchasing categories and time periods within the supermarket. These results suggest that choice consistency may be characterized as a multidimensional skill set. |
| Date: | 2025–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2505.05275 |
| By: | Carlos Alos Ferrer; Michele Garagnani |
| Abstract: | Surveys remain crucial tools for measuring societal preferences, but their reliability is limited by noise and bias in respondent data. We introduce a novel non-parametric method that leverages response times to reveal group preferences and rank preference strength across different populations. We validate the approach and apply it to key socio-economic questions using large representative surveys. The method complements traditional survey analysis techniques, providing clear indicators of when standard analyses may be inadequate and when response time data can yield additional insights. Importantly, our method also quantifies response biases, allowing researchers to adjust for systematic distortions in survey data. |
| Keywords: | Survey Data, Revealed Preference, Response Times, Response Bias |
| JEL: | D11 D81 D83 D87 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:lan:wpaper:424225030 |
| By: | Modibo K. Camara; Nicole Immorlica; Brendan Lucier |
| Abstract: | If people find it costly to evaluate the options available to them, their choices may not directly reveal their preferences. Yet, it is conceivable that a researcher can still learn about a population's preferences with careful experiment design. We formalize the researcher's problem in a model of robust mechanism design where it is costly for individuals to learn about how much they value a product. We characterize the statistics that the researcher can identify, and find that they are quite restricted. Finally, we apply our positive results to social choice and propose a way to combat uninformed voting. |
| Date: | 2025–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2505.19570 |
| By: | Henrike Sternberg (Technical University of Munich, TUM School of Social Sciences and Technology & TUM School of Management, Munich School of Politcs and Public Policy & Friedrich-Alexander-Universität Erlangen-Nürnberg); Janina Isabel Steinert (Technical University of Munich, TUM School of Social Sciences and Technology & TUM School of Medicine and Health, Munich School of Politcs and Public Policy); Tim Büthe (Technical University of Munich, TUM School of Social Sciences and Technology & TUM School of Management, Munich School of Politcs and Public Policy & Duke Univery, Sanford School of Public Policy) |
| Abstract: | This paper examines how inequality aversion shapes public support of international redistributive policies. We investigate this question in the context of the global allocation of vaccines during the Covid-19 pandemic, using online survey data from incentivized behavioral games and a discrete choice experiment conducted with German citizens in April 2021 (N=2, 402). We distinguish between aversion to advantageous inequality (others worse off, the ’guilt’ parameter) and aversion to disadvantageous inequality (others better off, the ’envy’ parameter). These two forms of inequality aversion shape German citizens’ attitudes towards the cross-country allocation of resources in distinct ways: While higher levels of the guilt parameter significantly increase respondents’ likelihood to prioritize an equitable vaccine allocation, the envy parameter is associated with lower support thereof. These findings suggest that inequality aversion matters for citizens’ support of redistribution beyond the national level and emphasize that distinguishing between both forms of inequality aversion is crucial. |
| Keywords: | Distributional preferences; Inequality aversion; International inequality; Covid-19 pandemic; Support for vaccine donations; Survey experiment |
| JEL: | C83 D63 D91 H87 I14 I18 |
| Date: | 2025–07 |
| URL: | https://d.repec.org/n?u=RePEc:aiw:wpaper:43 |
| By: | Pedro M. Gardete; Daniela Schmitt; Florian Stahl |
| Abstract: | This paper investigates how subscription pricing affects usage intensity, a key performance driver for firms operating under subscription-based business models. We analyze data from an online news publisher, a setting in which promotional pricing is commonly employed to attract new subscribers, though its broader effects remain ambiguous. Standard economic intuition suggests that lower-paying subscribers derive lower utility and thus consume less. In contrast, we document that promotional subscribers, on average, consume substantially more than those paying regular price, even after accounting for differences in churn behavior. This empirical pattern is inconsistent with simple demand models and points to the importance of taking unobserved heterogeneity into account. We develop and estimate an empirical model of subscription and consumption behavior, showing that, because subscription costs are sunk at the time of consumption, it is possible to recover the correlation between consumption levels and consumers’ unobserved willingness to pay. We use the model to recover the underlying consumer parameters and to evaluate the impact of alternative pricing policies on both subscription revenues (via customer acquisition) and advertising revenues (via subsequent consumption). Our findings highlight the economic value of understanding how price shapes not only who subscribes, but also how much they engage with the product. |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:unl:unlfep:wp674 |
| By: | Lei Bill Wang; Sooa Ahn |
| Abstract: | This paper examines why eligible households do not participate in welfare programs. Under the assumption that there exist some observed fully attentive groups, we model take-up as a two-stage process: attention followed by choice. We do so with two novel approaches. Drawing inspiration from the demand estimation for stochastically attentive consumers literature, Approach I is semiparametric with a nonparametric attention function and a parametric choice function. It uses fully attentive households to identify choice utility parameters and then uses the entire population to identify the attention probabilities. By augmenting Approach I with a random effect that simultaneously affects the attention and choice stages, Approach II allows household-level unobserved heterogeneity and dependence between attention and choice even after conditioning on observed covariates. Applied to NLSY panel data for WIC participation, both approaches consistently point to two empirical findings with regard to heterogeneous policy targeting. (1) As an infant ages towards 12 months and beyond, attention probability drops dramatically while choice probability steadily decreases. Finding (1) suggests that exit-prevention is the key for increasing the take-up rate because once a household exits the program when the infant ages close to 12 months old, it is unlikely to rejoin due to low attention. A value-increasing solution is predicted to be effective in promoting take-up by reducing exit probability. In contrast, an attention-raising solution is predicted to be ineffective. (2) Higher educated households are less attentive but more likely to enroll if attentive. Finding (2) suggests that running informational campaigns with parenting student groups at higher education institutions could be an effective strategy for boosting take-up. |
| Date: | 2025–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2506.03457 |
| By: | Carlos Alos Ferrer; Johannes Buckenmaier; Michele Garagnani |
| Abstract: | Economic decisions are noisy due to errors and cognitive imprecision. Often, they are also systematically biased by heuristics or behavioral rules of thumb, creating behavioral anomalies which challenge established economic theories. The interaction of noise and bias, however, has been mostly neglected, and recent work suggests that received behavioral anomalies might be just due to regularities in the noise. This contribution formalizes the idea that decision makers might follow a mixture of rules of behavior combining cognitively- imprecise value maximization and computationally simpler shortcuts. The model delivers new testable predictions which we validate in two experiments, focusing on biases in probability judgments and the certainty effect in lottery choice, respectively. Our findings suggest that neither cognitive imprecision nor multiplicity of behavioral rules suffice to explain received patterns in economic decision making. However, jointly modeling (cognitive) noise in value maximization and biases arising from simpler, cognitive shortcuts delivers a unified framework which can parsimoniously explain deviations from normative prescriptions across domains. |
| Keywords: | Cognitive Imprecision, Strength of Preference, Noise, Decision Biases, Belief Updating, Certainty Heuristic |
| JEL: | D01 D81 D87 D91 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:lan:wpaper:423483206 |