Tabular Foundation Models and the Unity of Economic Behaviour

2026-08-07

AI Summary

This study addresses the long-standing fragmentation of behavioral models across core domains in economics—such as risk, time, loss, valuation, and social choice—which has impeded the development of a unified explanatory framework. The authors propose an integrated approach that combines frozen tabular foundation models with structured random utility models, leveraging cross-domain choice data collected through unified experimental protocols. The foundation model predicts behavior in held-out domains based on choices observed in other domains, and its learned representations serve as the basis for a single random utility specification. This method uncovers a shared preference structure across domains, disentangling a common choice space, systematic utility functions, and stochastic components. The resulting model substantially outperforms a median benchmark, retains high predictive accuracy, generalizes to domains excluded from estimation, and faithfully reproduces the covariation patterns of individual-level behavioral metrics.

Abstract

Economics uses different behavioural models for risk, time, losses, valuation, and social choice. I study a unified choice experiment in which the same decision makers face all these domains. I hide a decision maker's choices in one domain and ask a frozen tabular foundation model to recover them from that decision maker's choices elsewhere and labelled choices by other participants. The foundation model improves on the training-sample median, and the gain disappears when visible choices are shuffled across decision makers. I then estimate one random-utility model over the foundation model's learned representation. This structural model applies the same utility function in every domain, retains most of the foundation model's reduction in prediction error, predicts domains excluded from utility estimation, and reproduces how behavioural measures co-move across people. The resulting model separates three objects: a learned common choice domain, one systematic utility function on that domain, and one random component that generates stochastic choice on observed menus.

Problem

• unified choice model
• economic behaviour
• tabular foundation models
• cross-domain decision making
• behavioural consistency

Innovation

• tabular foundation model
• unified economic behavior
• random utility model
• representation learning
• cross-domain prediction