A0659
Title: Panel regressions with latent factors and two-way heterogeneous treatments: A unified perspective
Authors: Arturas Juodis - University of Groningen (Netherlands)
Martin Weidner - University of Oxford (United Kingdom) [presenting]
Abstract: Panel regressions are revisited with unobserved heterogeneity through the lens of variance-weighted average treatment effects. Building on established results for cross-sectional OLS and one-way fixed effects panels, it is shown that two-way panel estimators with latent factors, such as the interactive fixed effects estimator, also converge to interpretable estimands under minimal assumptions. Specifically, even when the factor structure is misspecified, the estimator targets a variance-weighted average of unit-time-specific treatment effects, generalizing insights from prior studies. This unified perspective bridges the treatment effects literature and the factor regression literature, highlighting which estimators admit meaningful interpretation in heterogeneous panels without relying on parametric restrictions. Implications are also outlined for empirical practice and suggest directions for robust inference under latent interactive structures.
