CoChat

Revisiting Event-Study Designs: Robust and Efficient Estimation

Open access
Borusyak, Kirill, Jaravel, Xavier, Spiess, Jann · 2024 · Review of Economic Studies

The DOI matches its Crossref record; 57 of the 57 references it cites were found in a registry or an index; 6 of the 26 sources its claims cite were checked for support.

Abstract

We develop a framework for difference-in-differences designs with staggered treatment adoption and heterogeneous causal effects. We show that conventional regression-based estimators fail to provide unbiased estimates of relevant estimands absent strong restrictions on treatment-effect homogeneity. We then derive the efficient estimator addressing this challenge, which takes an intuitive “imputation” form when treatment-effect heterogeneity is unrestricted. We characterize the asymptotic behaviour of the estimator, propose tools for inference, and develop tests for identifying assumptions. Our method applies with time-varying controls, in triple-difference designs, and with certain non-binary treatments. We show the practical relevance of our results in a simulation study and an application. Studying the consumption response to tax rebates in the U.S., we find that the notional marginal propensity to consume is between 8 and 11% in the first quarter—about half as large as benchmark estimates used to calibrate macroeconomic models—and predominantly occurs in the first month after the rebate.

Overview

CoChat Summary

TL;DR

The paper develops a framework for staggered-adoption difference-in-differences with heterogeneous treatment effects, showing that conventional TWFE event-study estimators can be biased, under-identified, or spuriously identify long-run effects, and deriving a robust efficient estimator with an imputation form. In an application to U.S. 2008 tax rebates, the imputation estimator yields a first-quarter notional MPC of 7.8–11.4%, about half benchmark estimates, with spending response concentrated in the first month.

Key findings

  • Conventional static TWFE specifications can place negative weights on treatment effects: in the paper’s two-unit example, the estimand is τ_static = τ_A2 + 1/2 τ_B3 − 1/2 τ_A3, so the long-run effect enters with weight −1/2.
  • Fully dynamic specifications are under-identified when there are no never-treated units: for any κ, the path {τ_h + κ(h+1)} fits the data equally well; Proposition 5 shows weighted sums of effects with K_it ≥ H̄ are not identified under Assumptions 1 and 2.
  • The efficient estimator takes an imputation form under unrestricted heterogeneity: estimate fixed effects on untreated observations only, impute untreated outcomes for treated observations, then take a weighted sum of actual-minus-imputed outcomes; the paper derives consistency, asymptotic normality, conservative clustered standard errors, and a robust OLS-based pre-trend test using untreated observations only.
  • In the tax-rebate application, the preferred cumulative MPX for NielsenIQ products is $30.5 (3.4% of the average rebate) in the first month; Table 3 reports total MPX 24.8–36.6% vs. 50.8–74.8% for Broda–Parker, non-durable MPX 6.9–10.2% vs. 14.1–20.8%, and notional MPC 7.8–11.4% vs. 15.9–23.4%. Efficiency comparisons show dCDH confidence intervals about 50% longer and Sun–Abraham 2–3.5 times longer than the imputation estimator; simulations find alternative robust estimators’ standard deviations 1.3–3.6 times higher with spherical errors.

Methods

  • Panel DiD framework with staggered, absorbing binary treatment; the estimand is a pre-specified weighted sum or average of heterogeneous treatment effects, under parallel trends, no anticipation, and optionally restricted treatment-effect models.
  • Derives the efficient linear unbiased estimator under spherical errors, its imputation representation, asymptotic properties under clustered errors and a Herfindahl condition, conservative variance estimation, and a robust pre-trend test.
  • Application revisits Broda and Parker’s (2014) event study of the 2008 U.S. tax rebates using NielsenIQ Homescan data: 21,760 households (21,690 with disbursement-method information) over 52 weeks of 2008, with rebate timing quasi-randomized by SSN; compares binned OLS, unbinned OLS, and the imputation estimator, plus a Monte Carlo study based on the same data.

Limitations

  • Stated in the paper: all event-study designs assume a restrictive parametric model for untreated outcomes; the paper does not evaluate when such designs are ex ante appropriate, does not consider estimation robust to violations of parallel-trend-type assumptions, and does not consider finance-style high-frequency event studies without period fixed effects.
  • Stated/evident: the efficiency result uses spherical errors as a benchmark, though it can be relaxed to known heteroskedasticity or dependence; inference under unrestricted treatment-effect heterogeneity is conservative rather than exact and requires an auxiliary parsimonious model of treatment effects, with leave-out versions proposed for finite-sample overfitting.
  • Evident from the text: asymptotic conditions such as the Herfindahl condition can be restrictive, and the approach is least restrictive when the number of time periods is fixed or grows slowly; without never-treated units, long-run effects beyond the observed event-time gap are not identified.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1093/restud/rdae007

Cites

57 of 57 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

3 Valid · 6 Not valid · 14 Not checked · 175 Own results and interpretations

126 more claims

Create a free account to see all 198 claims and to verify the 1 reference still open.

See all claims — free account
  • It may state two separate claims as one. 53
  • Borderline: one of the tests for a substantial claim is unclear. 1
  • Probably not a substantial claim. 14
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

4 more claims about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

Show all 57 references

Research works printed but not linked: 1

  1. MacKinlay, 1997, Even Studies in Economics and Finance, Journal of Economic Literature, XXXV, 13

Cited by

Items 1,301–1,400 of 2,172

Metadata

Authors
Borusyak, K., Jaravel, X., & Spiess, J.
Affiliations
  • UC Berkeley ,
  • Center for Economic and Policy Research (US)
  • Stanford University (US)
Corresponding author
Jaravel, Xavier
Venue
Review of Economic Studies
Published
2024-02-06
Volume
91
Issue
6
Pages
3253-3285
ISSN
  • 0034-6527 (ISSN-L)
  • 1467-937X
OpenAlex
W3196836501
Cited by
2,280
References
39
Citations by year
2026: 858 · 2025: 736 · 2024: 368
FWCI
673.30
Type
journal article
Language
en
Open access
Locations
  • The Review of Economic Studies (publishedVersion, cc-by)
  • London School of Economics and Political Science Research Online (London School of Economics and Political Science) (acceptedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
Funders
  • University of Chicago
  • Booth School of Business, University of Chicago
  • Economic and Social Research Council (ES/T014431/1)
Topics
  • Physical Sciences › Mathematics › Statistics and Probability › Advanced Causal Inference Techniques
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Spatial and Panel Data Analysis
  • Social Sciences › Economics, Econometrics and Finance › General Economics, Econometrics and Finance › Monetary Policy and Economic Impact
Keywords
event study design, staggered treatment adoption, treatment effect heterogeneity, efficient estimators, asymptotic properties, identification tests, time-varying control, tax rebates, marginal propensity to consume, consumption response

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-09-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-09-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.