Empirical Bayes Estimation in Heterogeneous Coefficient Panel Models

2026-01-11

AI Summary

This study addresses the multidimensional heterogeneity in short panel data, encompassing heterogeneous intercepts, slopes, dynamic structures, and non-spherical error covariance. To this end, the authors develop an empirical Bayes G modeling framework that employs nonparametric priors to capture coefficient heterogeneity and propose a nonparametric maximum likelihood estimation algorithm based on Wasserstein–Fisher–Rao gradient flows. Under general conditions, they establish identification and consistency of the estimators and demonstrate that the empirical Bayes estimator achieves regret consistency. In an empirical application to income dynamics panel data, they find substantial slope heterogeneity negatively correlated with intercepts, alongside marked cross-sectional variation in both error variances and autoregressive coefficients. The proposed estimator significantly reduces out-of-sample mean squared prediction error compared to conventional methods.

Technology Category

Machine Learning: Bayesian LearningReasoning under Uncertainty: Relational Probabilistic ModelsIntelligent Robots: State Estimation

Application Category

Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphsUser Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating successEconomics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

We develop an empirical Bayes (EB) G-modeling framework for short-panel linear models with nonparametric prior for the random intercepts, slopes, dynamics, and non-spherical error variances. We establish identification and consistency of the nonparametric maximum likelihood estimator (NPMLE) under general conditions, and provide low-level sufficient conditions for several models of empirical interest. Conditions for regret consistency of the EB estimators are also established. The NPMLE is computed using a Wasserstein-Fisher-Rao gradient flow algorithm adapted to panel regressions. Using data from the Panel Study of Income Dynamics, we find that the slope coefficient for potential experience is substantially heterogeneous and negatively correlated with the random intercept, and that error variances and autoregressive coefficients vary significantly across individuals. The EB estimates reduce mean squared prediction errors relative to individual maximum likelihood estimates.

Problem

• heterogeneous coefficients
• panel data
• empirical Bayes
• nonparametric prior
• short panel

Innovation

• Empirical Bayes
• Nonparametric Maximum Likelihood Estimation
• Heterogeneous Coefficient Panel Models
• Wasserstein-Fisher-Rao Gradient Flow
• Regret Consistency