Event Study Analysis Made Simple

How Event Studies Work

An event study isolates the effect of a specific event by comparing observed returns to a counterfactual. During an estimation window — typically 120 to 250 trading days before the event — a return model establishes what "normal" returns look like. This counterfactual baseline is the foundation of all event study inference.

The abnormal return measures how far the actual return deviates from the model prediction. If the model predicts 0.2% but the stock returns 2.8%, the abnormal return is +2.6 percentage points. Cumulating daily abnormal returns across the event window yields the Cumulative Abnormal Return, or CAR.

"The usefulness of [event studies] comes from the fact that, given rationality in the marketplace, the effects of an event will be reflected immediately in security prices."

According to MacKinlay (1997), cited over 15,000 times, the choice of return model is critical for reliable inference. EventStudy supports 15 expected return models — including the Market Model, Fama-French three- and five-factor models, Carhart four-factor model, and GARCH(1,1). These models cover over 90% of designs found in published event study research.

Each return model is paired with 12 parametric and non-parametric test statistics. These tests ensure that conclusions are robust to common violations such as event-induced variance and cross-sectional correlation. Available tests include the Patell Z-test, BMP test, Generalized Sign Test, and Rank Test.

As documented by Kothari and Warner (2007), researchers have published over 2,000 event studies since Fama, Fisher, Jensen, and Roll introduced the methodology in 1969. Understanding the core concepts below — abnormal returns, cumulative abnormal returns, and modern difference-in-differences estimators — is essential for designing studies that produce credible, replicable results.

"In an efficient capital market, security prices fully reflect all available information."

Abnormal Return (AR)

According to Brown and Warner (1985), the abnormal return is the difference between the actual return of a security on a given day and the expected return predicted by a benchmark model. A stock that returns 3.5% when its model predicts 1.2% has an abnormal return of +2.3 percentage points. This isolates the price movement attributable to the event itself.

EventStudy supports 15 models for estimating expected returns, including the Market Model, Fama-French three- and five-factor models, and GARCH(1,1). The choice of model depends on the asset class, event type, and sample size. Researchers can compare results across multiple models to test robustness.

Cumulative Abnormal Return (CAR)

The Cumulative Abnormal Return is the sum of daily abnormal returns over the event window. A CAR of +5.3% over the window [0, +10] at p < 0.01 means the security gained 5.3 percentage points more than expected in the ten days following the event. The CAR is the most commonly reported statistic in event study research.

Researchers test whether the CAR is statistically significant using parametric tests such as the Patell Z-test and BMP test. Non-parametric alternatives like the Generalized Sign Test and Rank Test provide robustness when return distributions are non-normal. EventStudy includes all 12 test statistics for both single-event and multi-event designs.

Difference-in-Differences (DiD)

As shown by Callaway and Sant'Anna (2021), traditional TWFE estimators produce biased estimates under staggered treatment timing. Difference-in-Differences is a quasi-experimental design that compares changes in outcomes between a treatment group and a control group before and after an event.

Modern DiD estimators — including Callaway-Sant'Anna, Sun-Abraham, and de Chaisemartin-D'Haultfoeuille — handle staggered treatment timing and heterogeneous treatment effects. EventStudy's R package includes all six estimators and can handle datasets with 100+ treatment cohorts and thousands of units.