Correlated Random Coefficient Distributions in Linear Panel Models

TL;DR

This paper develops a method to identify and estimate the distribution of correlated and uncorrelated random coefficients in linear panel models, allowing for heterogeneity analysis without restrictive error structure assumptions.

Contribution

It introduces a novel identification approach and a two-step sieve estimator for distributions of random coefficients in panel data, applicable to both regular and irregular designs.

Findings

Estimated distributions show substantial heterogeneity in household elasticities.

Heterogeneity includes a significant share of near-zero and negative elasticities.

Results suggest households respond variably to income and expenditure changes.

Abstract

We consider a static linear panel model with both correlated and uncorrelated random coefficients, where the former can depend arbitrarily on observable regressors while the latter are independent of them. We provide sufficient conditions for identification of the distributions of the random coefficients without imposing restrictions on the time-series structure of the error terms in short panels. Our framework applies to regular and irregular designs. The distribution of the correlated coefficients follows via a deconvolution argument. In irregular designs, identification relies on a stayer-based argument exploiting near-singular realizations of the regressor matrix. We develop a two-step minimum distance sieve estimator, with tuning parameters selected by cross-validation. In an application to calorie-expenditure elasticities using data from the randomized evaluation of a conditional cash transfer program, we interpret the estimated distributions by program status as distributions of regime-specific structural calorie-expenditure elasticities. The estimated densities themselves reveal substantial heterogeneity in household-specific elasticities, with nontrivial mass concentrated near zero and a non-negligible share of negative realizations. This heterogeneity implies that responses to income or expenditure changes are not uniformly positive and vary widely across households. Taken together, these features support a framework in which households adjust along both quantity and quality margins, rather than conforming to a homogeneous Engel-curve response.