Doubly Robust Identification for Causal Panel Data Models

TL;DR

This paper introduces a doubly robust approach for identifying and estimating causal effects in panel data, combining assumptions about treatment assignment and confounders, which enhances robustness and flexibility.

Contribution

It develops new identification assumptions for panel data and proposes estimation methods that are doubly robust, bridging cross-section strategies with panel data analysis.

Findings

New identification assumptions for panel data causal inference

Doubly robust estimation methods applicable to panel data

Enhanced robustness in causal effect estimation

Abstract

We study identification and estimation of causal effects in settings with panel data. Traditionally researchers follow model-based identification strategies relying on assumptions governing the relation between the potential outcomes and the observed and unobserved confounders. We focus on a different, complementary approach to identification where assumptions are made about the connection between the treatment assignment and the unobserved confounders. Such strategies are common in cross-section settings but rarely used with panel data. We introduce different sets of assumptions that follow the two paths to identification and develop a doubly robust approach. We propose estimation methods that build on these identification strategies.