|
on Discrete Choice Models |
| By: | Omar Abdel Haq; Amitabh Chandra; Tomáš Jagelka; Erzo F.P. Luttmer; Joshua Schwartzstein |
| Abstract: | Large Language Models (LLMs) are trained on a prodigious corpus of human writing and may reveal human preferences over characteristics of life courses, such as income, longevity, and working conditions. We present OpenAI's GPT-5.4 and a broadly representative sample of Americans with pairs of life stories and ask them to choose the life they would prefer for themselves. A person's choice is better predicted by the LLM's choice than by another person's choice over the same stories, and LLM valuations of several life attributes are similar to those derived from human responses. Our results suggest that LLM responses offer a scalable and cost-effective complement to existing methods for studying human preferences. |
| Keywords: | Generative AI, preference estimation methods, choice experiments, survey validation |
| JEL: | D0 H0 I0 |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:crm:wpaper:26134 |
| By: | Abdel Haq, Omar (Harvard Business School); Chandra, Amitabh (Harvard Business School and Harvard Kennedy School); Jagelka, Tomáš (University of Bonn); Luttmer, Erzo (Dartmouth College); Schwartzstein, Joshua (Harvard Business School) |
| Abstract: | Large Language Models (LLMs) are trained on a prodigious corpus of human writing and may reveal human preferences over characteristics of life courses, such as income, longevity, and working conditions. We present OpenAI's GPT-5.4 and a broadly representative sample of Americans with pairs of life stories and ask them to choose the life they would prefer for themselves. A person's choice is better predicted by the LLM's choice than by another person’s choice over the same stories, and LLM valuations of several life attributes are similar to those derived from human responses. Our results suggest that LLM responses offer a scalable and cost-effective complement to existing methods for studying human preferences. |
| Keywords: | generative AI, preference estimation methods, choice experiments, survey validation |
| JEL: | D0 H0 I0 |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18634 |
| By: | Shuhua Si |
| Abstract: | Economic choices are often stochastic: the same person may make a different choice when facing the same alternatives repeatedly. Standard models assume that the degree of randomness reflects the size of utility differences, but choice inconsistencies could also reflect difficulty comparing alternatives. Recent studies estimate such comparison difficulty (or "complexity") by fitting functional forms to aggregate choice data under a representative agent assumption. However, aggregate data could violate standard models of random choice simply because of heterogeneity in preferences, even in the absence of variation in comparison difficulty. This paper develops a revealed preference framework, collective rationalizability, that tests for variation in comparison difficulty from aggregate data while explicitly accounting for heterogeneity. The framework characterizes whether violations of standard models can be explained by comparison difficulty alone, heterogeneity alone, or require both. I provide a statistical test with finite-sample inference and apply the method to two existing experiments. In both cases, heterogeneity alone explains observed failures of stochastic transitivity well, demonstrating that comparison difficulty can be not only theoretically but also empirically confused with heterogeneity in aggregate data. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.01850 |
| By: | Federico Echenique; Alireza Fallah; Baihe Huang; Michael I. Jordan |
| Abstract: | Aligning large language models (LLMs) to human preferences typically relies on aggregating pooled feedback into a single reward model. However, this standard approach assumes that all labelers share the same underlying preferences, ignoring the fact that real-world labelers are highly heterogeneous and usually anonymous. Consequently, relying solely on binary choice data fundamentally distorts the learned policy, making the true population-average preference unidentifiable. To overcome this critical limitation, we demonstrate that augmenting preference datasets with a simple, secondary signal -- the user's response time -- can restore the identifiability of the population's average preference. By modeling each decision as a Drift-Diffusion Model (DDM), we introduce a novel, consistent estimator of heterogeneous preferences that successfully corrects the distortions of standard choice-only labels. We prove that our estimator asymptotically converges to the true average preference even in extreme cases where each anonymous labeler contributes only a single choice. Empirically, across both synthetic and real-world datasets, our method consistently outperforms standard baselines that otherwise fail and plateau at a bias floor. Because response times are essentially free to record and require zero user tracking or identification, our results bring promises and open up new opportunities for future data-collection pipelines to improve the social benefit without requiring user-level identifiers or repeated elicitations. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.06987 |
| By: | Vedunka Kopecna (Institute of Economic Studies, Faculty of Social Sciences, Charles University, Prague, Czech Republic & Charles University, Environment Center); Inaki Veruete Villegas (Institute of Economic Studies, Faculty of Social Sciences, Charles University, Prague, Czech Republic & Charles University, Environment Center) |
| Abstract: | This paper assesses the long-term macroeconomic and environmental impacts of climate policies in the Czech Republic, a coal-dependent economy, under the EU´s Fit-for-55 package. Using a hybrid dynamic Computable General Equilibrium (CGE) model, we integrate a bottom-up electricity module with technology-specific detail and a discrete choice module capturing consumer preferences for vehicle technologies. The model, formulated as a mixed complementarity problem in GAMS, accounts for capacity constraints in power generation and endogenizes vehicle fleet evolution based on choice probabilities. We evaluate two scenarios: With Existing Measures (WEM), reflecting current policies, and With Additional Measures (WAM), which includes coal phase-out, expanded renewables, and the introduction of ETS2. Results show that WAM leads to more than 60% reduction in power sector CO2 emissions by 2040 and 80% battery electric vehicle (BEV) adoption by 2050. However, green investments under WAM do not balance out structural shifts - especially in fossil-related sectors - negatively influencing GDP. This integrated top-down and bottom-up modeling approach offers a robust framework for evaluating economy-wide effects of climate action. Findings inform cost-effective and socially balanced decarbonization strategies for Czech and EU policymakers. |
| Keywords: | Hybrid CGE model; Green Transition; Climate policies; Energy and transport |
| JEL: | C68 D12 D58 H22 H23 Q43 Q52 R42 |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:fau:wpaper:wp2026_06 |
| By: | Ohyun Kwon; Mario Larch; Jangsu Yoon; Yoto V. Yotov |
| Abstract: | We implement an instrumental-variable Poisson pseudo-maximum likelihood estimator with high-dimensional fixed effects (IV-PPML-HDFE). To correct for incidental parameter bias, we use a split-panel jackknife (SPJ) routine with bootstrapped standard errors. Monte Carlo simulations across the three most common fixed-effect structures confirm that SPJ reduces the mean absolute bias by 42% and raises mean bootstrap confidence-interval coverage from 69% to 92%. We provide a robust and user-friendly 'ivppmlhdfe' package, and deploy it in three empirical applications to establish the validity and usefulness of our methods. |
| Keywords: | Poisson pseudo-maximum likelihood, instrumental variables, high-dimensional fixed effects, incidental parameter problem, gravity model, split-panel jackknife. |
| JEL: | C13 C23 C26 F14 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:ces:ceswps:_12641 |
| By: | Jikai Jin; Vasilis Syrgkanis |
| Abstract: | Offline evaluation of language models from usage logs is biased when model choice is confounded: the same user-side factors that influence which model is used can also influence how its output is judged, so raw comparisons of logged scores mix self-selected populations rather than estimating a common quantity of interest. A small randomized experiment can break this bias by overriding model choice, but in practice such experiments are scarce and costly. We study a three-source design that combines a large confounded observational log (OBS) for scale, a small randomized experiment (EXP) for unconfounded scoring, and an offline simulator (SIM) that replays candidate models on cached contexts. Our main result is an identification theorem showing that the randomized experiment and the simulator are together enough to recover causal model values; the observational log enters only afterward, to reduce estimation error rather than to make the causal comparison valid. Six estimator families are evaluated in a controlled semi-synthetic validation and in two real-task cached benchmarks for summarization and coding. No family dominates every regime; relative performance depends on the amount of unbiased EXP supervision and on how closely the target reward aligns with OBS-derived structure. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.01311 |
| By: | Yasunori Okumura |
| Abstract: | This note proposes a simple polynomial-time method for constructing an ex ante stable school-choice lottery satisfying equal treatment of equals. The method applies the ETE reassignment to a constrained efficient stable matching and yields a lottery that is not ordinally dominated by any other ex ante stable lottery. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.06721 |
| By: | L. Kaili Diamond; Ben Gilbert |
| Abstract: | Interaction effects are often economically central in environments where structural dynamic estimation becomes computationally infeasible. Under fixed group membership and sparse within-group interaction structure, the Bellman operator admits a block-diagonal decomposition that allows high-dimensional dynamic programs to be solved through independent group-level subproblems while preserving the original structural problem exactly. The result applies to a class of dynamic discrete choice models in which interactions are confined within stable local groups and state transitions depend only on within-group conditions. We apply the framework to replacement decisions across 14, 344 GPU node locations in the Titan supercomputer, where operating environments differ systematically across cage positions. The structural estimates reveal significant spatial coordination: both neighboring failures and recent local replacement activity increase replacement incentives. Accounting for these interaction effects materially shifts predicted replacement timing and reveals significant misoptimization costs in benchmarks that assume conditional independence. More broadly, the results show how exploiting sparsity in interaction structures can make fully structural estimation feasible in large-scale networked systems without relying on simulation-based auxiliary moments or numerical approximation. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.04592 |
| By: | yoshida, ken |
| Abstract: | This study examines whether impatience measured by hypothetical monetary intertemporal choices is reflected in an observable consumption behavior, namely candy consumption. Participants first answered a series of hypothetical intertemporal-choice questions involving both monetary gains and monetary payments and then completed a candy-consumption task. Realized behavior was classified from the reported completion status at the end of the task as licking the candy to the end, biting it before finishing, or being censored after 20 minutes, while completion time was recorded from QR-code scans at the beginning and end of the task. Using a strict sample restricted to respondents who passed consistency and attention checks, we find suggestive but meaningful evidence that participants who bit the candy displayed greater impatience in monetary choices than those who licked it to the end. The average difference across all questions is modest, and the sharper contrast in some payment questions should be interpreted as exploratory. Participants who bit the candy also completed the task significantly faster. In addition, self-reported usual eating style is moderately consistent with realized behavior in the task. These findings suggest that candy-consumption behavior may serve as a behavioral indicator broadly consistent with impatience, while also indicating that it should not be interpreted as a structural estimate of an individual's discount rate. |
| Keywords: | Time preference; intertemporal choice; impatience; candy consumption; experimental economics |
| JEL: | C91 D91 |
| Date: | 2026–03–31 |
| URL: | https://d.repec.org/n?u=RePEc:pra:mprapa:128522 |
| By: | Peter Reinhard Hansen; Chen Tong |
| Abstract: | The convolution of a Gaussian and a Cauchy distribution, known as the Voigt distribution, is widely used in spectroscopy and provides a natural framework for modeling heavy-tailed measurement noise. We derive analytical expressions for its density, score, Hessian, and conditional moments using the scaled complementary error function, enabling stable maximum likelihood estimation without numerical convolution, finite-difference derivatives, or pseudo-Voigt approximations. The conditional expectation of the latent Gaussian component is governed by a redescending location score, so extreme observations are automatically discounted rather than propagated. This structure motivates the Gauss-Cauchy Convolution (GCC) filter for state-space models with Gaussian latent dynamics and heavy-tailed measurement errors. In an application to log realized volatility for the Technology Select Sector SPDR Fund, the GCC filter separates persistent latent variation from transient measurement noise and improves on Gaussian, Student-$t$, Huber, and related robust alternatives. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.01665 |
| By: | Fang, Ximeng; Innocenti, Stefania; Vogt, Sonja (Faculty of Business and Economics, University of Lausanne) |
| Abstract: | Scalable behavioural interventions often struggle to engage the cognitive and psychological mechanisms that underlie durable changes in preferences and habits. This study provides a proof of concept for an underexplored intervention format: edutainment through video games. Partnering with a large video game company, we develop a game that embeds educational content on sustainable food consumption into an entertaining storyline. In a pre-registered field experiment (N = 4, 034 UK adults), participants are randomly assigned to play either one of three treatment versions of the game or a control version without environmental content. Real-world food choice behaviour is measured through incentivised online supermarket tasks. Relative to the control group, treated participants select grocery baskets with 20% lower environmental impact immediately after gameplay, an effect that remains at 8–10% in a follow-up 2–3 weeks later. Behavioural change results from a combination of knowledge gains, short-term salience and preference change. Strikingly, effects were particularly persistent among subjects with low baseline sustainability. Further evidence suggests that the intervention was effective partly because it provided an enjoyable experience and affected a rich set of beliefs and attitudes, including personal norms, efficacy, and perceived social norms. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:amz:wpaper:2026-13 |
| By: | Nayoung Lee; Hyungsik Roger Moon; Martin Weidner |
| Abstract: | This paper studies a simple dynamic linear panel regression model with interactive fixed effects in which the variable of interest is measured with error. To estimate the dynamic coefficient, we consider the least-squares minimum distance (LS-MD) estimation method. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.02311 |