Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Group-Heterogeneous Changes-in-Changes and Distributional Synthetic Controls

Songnian Chen, Junlong Feng

TL;DR

This paper addresses group-level heterogeneity in causal inference by extending changes-in-changes (CIC) and distributional synthetic control (DSC) to settings with both individual- and group-level unobservables, introducing a group-level factor $V_{gt}$ that interacts with the individual unobservables $U_{igt}$. For CIC, it shows identification of heterogeneous quantile treatment effects on the treated through matching on $V_{gt}$ across large cross-sections or matched quantile conditions, yielding three practical identification scenarios. For DSC, it constructs a time-series, quantile-based synthetic control that aligns groups with comparable $V_{gt}$ using possibly different pre-treatment periods, supported by isometry conditions on the production function and time-trend considerations. The paper also provides testable implications, discusses time trends, and outlines implementable estimation procedures, positioning the approach as a practical, nonparametric toolkit for handling group heterogeneity in clustered data. Overall, the contributions offer robust identification and feasible estimation strategies for causal effects when group-level factors shape outcomes, with broad applicability to economics and policy analysis.

Abstract

We develop new changes-in-changes (CIC) and distributional synthetic controls (DSC) types of methods when there exists group-level heterogeneity. For CIC, we allow individuals to belong to heterogeneous groups, extending Athey and Imbens (2006) by finding appropriate control groups that share similar group-level unobserved characteristics to the treatment groups. For DSC, we show that the synthetic control units are not necessarily from the same period as in Gunsilius (2023); they may come from different periods in which they have comparable group-level heterogeneity to the treatment group. Implementation of these new methods is briefly discussed.

Group-Heterogeneous Changes-in-Changes and Distributional Synthetic Controls

TL;DR

This paper addresses group-level heterogeneity in causal inference by extending changes-in-changes (CIC) and distributional synthetic control (DSC) to settings with both individual- and group-level unobservables, introducing a group-level factor that interacts with the individual unobservables . For CIC, it shows identification of heterogeneous quantile treatment effects on the treated through matching on across large cross-sections or matched quantile conditions, yielding three practical identification scenarios. For DSC, it constructs a time-series, quantile-based synthetic control that aligns groups with comparable using possibly different pre-treatment periods, supported by isometry conditions on the production function and time-trend considerations. The paper also provides testable implications, discusses time trends, and outlines implementable estimation procedures, positioning the approach as a practical, nonparametric toolkit for handling group heterogeneity in clustered data. Overall, the contributions offer robust identification and feasible estimation strategies for causal effects when group-level factors shape outcomes, with broad applicability to economics and policy analysis.

Abstract

We develop new changes-in-changes (CIC) and distributional synthetic controls (DSC) types of methods when there exists group-level heterogeneity. For CIC, we allow individuals to belong to heterogeneous groups, extending Athey and Imbens (2006) by finding appropriate control groups that share similar group-level unobserved characteristics to the treatment groups. For DSC, we show that the synthetic control units are not necessarily from the same period as in Gunsilius (2023); they may come from different periods in which they have comparable group-level heterogeneity to the treatment group. Implementation of these new methods is briefly discussed.
Paper Structure (12 sections, 3 theorems, 22 equations, 2 tables)

This paper contains 12 sections, 3 theorems, 22 equations, 2 tables.

Table of Contents

  1. Introduction
  2. The Model
  3. Changes-in-Changes with Group Heterogeneity
  4. Testable Implications
  5. Distributional Synthetic Control with Group Heterogeneity
  6. Time Trend
  7. Comparison with the Classical Synthetic Control Methods
  8. Comparison with Synthetic Differences-in-Differences (SDID)
  9. Comparison with gunsilius2023distributional
  10. Implementation
  11. Changes-in-Changes
  12. Synthetic Control

Key Result

Theorem 1

Under Assumptions assum.did.mono-assum.did.support, for any fixed $(\tau_{U}^{*},\tau_{V}^{*})\in (0,1)\times (0,1)$, the following statements are true:

Theorems & Definitions (9)

  • Remark 1
  • Theorem 1
  • Theorem 2
  • Remark 2
  • Theorem 3
  • Remark 3
  • proof : Proof of Theorem \ref{['thm.did.id']}
  • proof : Proof of Theorem \ref{['thm.did.test']}
  • proof : Proof of Theorem \ref{['thm.sc.id']}