Library · Causal inference & statistics

What Event Study Designs got wrong, from 18 theses

Summary written by a model from the records below: The records document methodological limitations and failures when applying event study designs across policy evaluation, corporate finance, and asset market settings. Researchers frequently encounter parallel trend violations, treatment effect heterogeneity under staggered timing, noisy cumulative abnormal return windows, and challenges in defining event timing or valid benchmark controls. These records come from graduate theses, 2021 to 2026, each labeled with its degree where the repository states it, and each links to its thesis. They were extracted by language models reading the full text, so treat each as a lead to read, not a verdict.

18
theses
10
institutions
19
failure records
5
failure modes

Plus 3 problems the authors left open, listed at the end.

Failure modes, by theses reporting them

Select a mode to jump to its records.

Pre-existing differential trends and pre-treatment divergence violate the parallel trends assumption
8
8 theses · 3 institutions · 9 records
Cumulative abnormal return event windows and adjustments introduce noise and distort time series properties
3
3 theses · 3 institutions · 3 records
Two-way fixed effects dynamic event studies suffer from treatment effect heterogeneity and cross-period contamination
3
3 theses · 2 institutions · 3 records
Decentralized asset settings lack reliable benchmark controls and exhibit non-overlapping event timing
2
2 theses · 2 institutions · 2 records
Event studies fail when disclosures cannot be tied to precise unexpected dates or electronic dissemination
2
2 theses · 1 institutions · 2 records

Cumulative abnormal return event windows and adjustments introduce noise and distort time series properties

3 theses · 3 institutions

Longer event windows introduced excessive noise that obscured abnormal return effects, while pre-event detrending caused return differences across groups to lose consistency. Standard cumulative abnormal return models also failed to evaluate long-term gains and could not adjust for time series properties such as autocorrelation, non-stationarity, seasonality, or drifts.

Tried and failed

long-horizon event study methodology applied to firm equity value post-regulatory events. Outcome: no signal. Reason: longer event windows introduced too much noise, obscuring statistically significant abnormal return effects

Studies In The Political Economy Of Anti-Money Laundering Regulation · Penn · PhD

Considered and rejected

Considered and rejected: Rejected cumulative abnormal returns (short-term CAR event study) as the dependent variable because the study required evaluating long-term synergy gains rather than short-term market cash flow expectations.

Essays on the role of experience and learning in post-aquisition performance of emerging market multinational enterprises · Nottingham · PhD

Considered and rejected

Considered and rejected: Standard event study methodology based on CARs was rejected because it cannot adjust for autocorrelation, non-stationarity, seasonality, cyclical changes, or drifts in time series

The effects of supply chain disruptions due to hurricanes on stock price. · Cranfield · PhD

Two-way fixed effects dynamic event studies suffer from treatment effect heterogeneity and cross-period contamination

3 theses · 2 institutions

Standard two-way fixed effects event study designs with leads and lags caused cross-period contamination and cohort treatment effect heterogeneity under staggered adoption. These model structures introduced estimation bias, produced spurious pre-trends, and mechanically eliminated never-treated comparison units.

Considered and rejected

Considered and rejected: Rejected standard two-way fixed effects event study designs with leads and lags due to cross-period contamination under staggered treatment adoption.

The Trade Impact of Diplomatic Relations in Developing Countries: The Choice between China or Taiwan · University of Oregon · PhD

Considered and rejected

Considered and rejected: Rejected standard two-way fixed effects (TWFE) dynamic event study estimation due to cohort treatment effect heterogeneity contamination that introduces bias and spurious pre-trends.

A QUANTITATIVE ANALYSIS OF OWNERSHIP-INDUCED QUALITY GAPS IN THE LONG-TERM CARE SECTOR: INFLUENCES OF OWNERSHIP CONVERSIONS, SELF-REPORTING, REGULATORY REFORMS, AND THE COVID-19 PANDEMIC · Georgia Tech · PhD

Considered and rejected

Considered and rejected: Rejected standard Two-Way Fixed Effects event study because leads/lags mechanically eliminate never-treated units and fail under treatment effect heterogeneity.

Technology, Violent Conflict, and the Determinants of Migration · Georgia Tech · PhD

Decentralized asset settings lack reliable benchmark controls and exhibit non-overlapping event timing

2 theses · 2 institutions

Traditional event study methodologies failed in cryptocurrency settings due to small sample sizes and non-overlapping event dates across individual coins. Researchers also rejected synthetic control assets and broad market indices for decentralized finance event studies because reliable benchmark counterfactuals were unavailable.

Considered and rejected

Considered and rejected: Rejected traditional event study methodologies when examining PoS transition volatility due to small sample size and non-overlapping event dates across coins, adopting an extended GARCH model with a novel Chi-squared joint test statistic instead.

Essays on Financial Innovations : Expected Returns and Volatility in the Cryptocurrency Market · University of Vaasa · PhD

Considered and rejected

Considered and rejected: Rejected relying on synthetic control assets or market indices for DeFi event study DiD analysis due to lack of reliable benchmarks; used real correlated counterfactual governance tokens instead.

Computational Methods for Analyzing Decentralized Finance Ecosystems Rechnergestützte Methoden zur Analyse dezentraler Finanzökosysteme · TU Wien · PhD

Event studies fail when disclosures cannot be tied to precise unexpected dates or electronic dissemination

2 theses · 1 institutions

Event studies could not measure market reactions when broad keyword categories failed to map to specific unexpected event dates. Similarly, examining physical document disclosures failed because information was not electronically disseminated to market participants prior to online posting.

Tried and failed

event study of physical document disclosures applied to asset price and volume reactions. Outcome: no signal. Reason: information was not publicly disseminated electronically to market participants prior to online posting

Hidden Mandatory Disclosure · Penn · PhD

Considered and rejected

Considered and rejected: Rejected using generic keyword categories ('social', 'political', 'people') for the empirical event study, because they could not be mapped to specific unexpected event dates.

Essays In Information Economics · Penn · PhD

Left open by the authors

Problems the authors named and did not get to.

Left open

Perform an event study analyzing systemic risk and spillover effects from specific stablecoins and volatile assets following liquidation cascades or oracle failures. Blocker: None

Blockchain and Decentralized Finance: Assessment of Risks, Opportunities, and Innovative Monetary Systems · University of Camerino · PhD

Left open

Conduct an event study estimating the impact of live-streaming sales events on offline brick-and-mortar stores. Blocker: Proprietary or restricted commercial data on live-streaming sales events and offline store transactions

Essays on Network Coalition, Online Price Signaling, and E-commerce Effects on Offline Stores · Cornell · PhD

Left open

Conduct an event study of the PoW-to-PoS protocol transition using event windows centered on announcement dates instead of implementation dates. Blocker: None

Essays on Financial Innovations : Expected Returns and Volatility in the Cryptocurrency Market · University of Vaasa · PhD

Who has worked on this

Authors of the public theses cited on this page, most-cited first.

Related methods in Causal inference & statistics

Bootstrap Resampling & PermutationFixed Effects Panel MethodsPropensity Score MatchingLongitudinal Growth & Latent Curve ModelingDifference-in-Differences DesignSensitivity & Uncertainty Analysis

Checking a claim in this area?

We can run the same search on any method or claim. If nothing turns up, we will say so, and that proves nothing on its own.

Ask about a method