Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Estimating Treatment Effects in Panel Data Without Parallel Trends

Shoya Ishimaru

TL;DR

The paper develops a flexible framework for estimating treatment effects in panel data without relying on parallel trends, by exploiting repeated untreated outcomes as noisy measurements of multidimensional unobservables. It provides nonparametric identification results under broad conditions and extends to QTT, treatment-effect heterogeneity, and staggered adoption, with estimation feasible via ML or semiparametric methods when the unobservable dimension is modest. An empirical application on job displacement shows that standard DID likely overstates long-run earnings losses, with the proposed approach delivering substantially smaller effects by accounting for complex unobserved heterogeneity. The work highlights the importance of flexible unobserved-heterogeneity modeling for credible causal inference in panel data, while noting data requirements and computational complexity as practical considerations.

Abstract

This paper proposes a novel approach for estimating treatment effects in panel data settings, addressing key limitations of the standard difference-in-differences (DID) approach. The standard approach relies on the parallel trends assumption, implicitly requiring that unobservable factors correlated with treatment assignment be unidimensional, time-invariant, and affect untreated potential outcomes in an additively separable manner. This paper introduces a more flexible framework that allows for multidimensional unobservables and non-additive separability, and provides sufficient conditions for identifying the average treatment effect on the treated. An empirical application to job displacement reveals substantially smaller long-run earnings losses compared to the standard DID approach, demonstrating the framework's ability to account for unobserved heterogeneity that manifests as differential outcome trajectories between treated and control groups.

Estimating Treatment Effects in Panel Data Without Parallel Trends

TL;DR

The paper develops a flexible framework for estimating treatment effects in panel data without relying on parallel trends, by exploiting repeated untreated outcomes as noisy measurements of multidimensional unobservables. It provides nonparametric identification results under broad conditions and extends to QTT, treatment-effect heterogeneity, and staggered adoption, with estimation feasible via ML or semiparametric methods when the unobservable dimension is modest. An empirical application on job displacement shows that standard DID likely overstates long-run earnings losses, with the proposed approach delivering substantially smaller effects by accounting for complex unobserved heterogeneity. The work highlights the importance of flexible unobserved-heterogeneity modeling for credible causal inference in panel data, while noting data requirements and computational complexity as practical considerations.

Abstract

This paper proposes a novel approach for estimating treatment effects in panel data settings, addressing key limitations of the standard difference-in-differences (DID) approach. The standard approach relies on the parallel trends assumption, implicitly requiring that unobservable factors correlated with treatment assignment be unidimensional, time-invariant, and affect untreated potential outcomes in an additively separable manner. This paper introduces a more flexible framework that allows for multidimensional unobservables and non-additive separability, and provides sufficient conditions for identifying the average treatment effect on the treated. An empirical application to job displacement reveals substantially smaller long-run earnings losses compared to the standard DID approach, demonstrating the framework's ability to account for unobserved heterogeneity that manifests as differential outcome trajectories between treated and control groups.
Paper Structure (34 sections, 2 theorems, 45 equations, 2 figures, 3 tables)

This paper contains 34 sections, 2 theorems, 45 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Econometric Framework
  3. Setup and Assumptions
  4. Examples
  5. Standard DID
  6. Linear and Nonlinear Factor Models
  7. Hidden Markov Model
  8. Identification
  9. Estimation
  10. Extensions
  11. Identifying the Quantile Treatment Effect on Treated
  12. Identifying Treatment Effect Heterogeneity
  13. Staggered Treatment Adoption
  14. Empirical Illustration
  15. Standard Approach as a Benchmark
  16. ...and 19 more sections

Key Result

Theorem 1

Under Assumptions 1--3 and either Assumption 4 or 5, the ATT parameters $\{\theta_{t}^{\text{ATT}}\}_{t=1}^{T_{\text{post}}}$ are identified from the joint distribution of $\left(D_{i},Y_{i,-T_{\text{pre}}},\ldots,Y_{i,T_{\text{post}}}\right)$.

Figures (2)

  • Figure 1: The Estimated Earning Losses from Displacement (Standard DID)
  • Figure 2: Estimated Earnings Losses from Displacement Across Different Years

Theorems & Definitions (2)

  • Theorem 1
  • Theorem 2