Event Study Glossary

This glossary defines the key terms used throughout the Event Study Methodology Guide. Each entry provides a concise definition and links to the documentation page where the concept is covered in detail.

A

Abnormal Return (AR)

The difference between a stock's actual return and its expected (normal) return on a given day. Abnormal returns measure the event's impact on the stock price. A positive AR means the stock performed better than expected; a negative AR means it performed worse. See Test Statistics Overview.

Average Abnormal Return (AAR)

The cross-sectional average of abnormal returns across all firms on a given event day. AAR measures the average effect of the event on day t across the entire sample. It is computed as AAR_t = (1/N) * Sum of AR_i,t. See AAR & CAAR Tests.

Adjusted Closing Price

The stock's closing price corrected for stock splits, dividends, and other corporate actions. Event studies must use adjusted prices to avoid artificial return spikes. See Data Preparation.

Alpha

In the Market Model, alpha is the intercept of the regression of stock returns on market returns. It represents the portion of the stock's return not explained by market movements. In portfolio analysis, alpha refers to risk-adjusted excess return. See Expected Return Models.

B

Benchmark

The reference return used to compute expected returns. Common benchmarks include market indices (S&P 500), matched firms, and factor model predictions. The benchmark defines what "normal" performance looks like in the absence of the event.

Beta

In the Market Model, beta is the slope coefficient measuring the stock's sensitivity to market returns. A beta of 1.2 means the stock tends to move 1.2% for every 1% move in the market. Beta is estimated during the estimation window. See Expected Return Models.

BHAR (Buy-and-Hold Abnormal Return)

The difference between the compounded return of the event firm and the compounded return of a benchmark over a holding period. BHAR reflects the actual investor experience and is used in long-run event studies. See Long-Run Event Studies.

BMP Test

The Boehmer-Musumeci-Poulsen (1991) test for multi-event studies. It standardizes abnormal returns by firm-specific estimation-window variance, then uses the cross-sectional standard deviation for inference. Robust to event-induced variance changes. See AAR & CAAR Tests.

C

CAAR (Cumulative Average Abnormal Return)

The sum of average abnormal returns over the event window. CAAR measures the total average impact of the event across all firms. CAAR(t1, t2) = Sum of AAR_t from t1 to t2. See AAR & CAAR Tests.

Calendar-Time Portfolio

A method for measuring long-run abnormal returns that forms a portfolio of all event firms each calendar month and regresses portfolio returns on a factor model. The intercept (alpha) measures abnormal performance. Naturally handles cross-sectional correlation. See Long-Run Event Studies.

CAR (Cumulative Abnormal Return)

The sum of abnormal returns for a single firm over the event window. CAR measures the total price impact of the event on one firm. CAR_i(t1, t2) = Sum of AR_i,t from t1 to t2. See AR & CAR Tests.

Confounding Event

Any event other than the one being studied that occurs within the event window and affects the stock price. Confounding events bias abnormal return estimates. They should be identified and either controlled for or lead to exclusion of the affected observation.

Cross-Correlation

Correlation of abnormal returns across firms, typically caused by overlapping event windows or shared event dates. Cross-correlation violates the independence assumption of standard tests and leads to inflated test statistics. See Choosing the Right Test Statistic.

Cross-Sectional t-Test

A multi-event test that divides the average abnormal return by its cross-sectional standard error. Simple and intuitive, but not robust to event-induced variance. See AAR & CAAR Tests.

D

DCC-GARCH

Dynamic Conditional Correlation GARCH. A model that allows both return volatility and the correlation between stock and market returns to vary over time. Used when the stock-market relationship is unstable. See Expected Return Models.

E

Estimation Window

The period before the event used to estimate the expected return model parameters. Typically 120-250 trading days. The estimation window must not overlap with the event window. See Window Selection.

Event Date (Day 0)

The date on which the event occurs or is announced. All other dates in the event study are measured relative to the event date. Day -1 is the day before the event; day +1 is the day after.

Event-Induced Variance

The increase in return variance that often occurs around the event date. Events like earnings announcements and M&A deals typically cause higher volatility. Standard tests that ignore this phenomenon tend to over-reject the null hypothesis. See Variance-Based Tests.

Event Study

A statistical method that measures the impact of a specific event on a company's stock price by comparing actual returns to expected returns. Widely used in finance, economics, and law. See Introduction.

Event Window

The period around the event date during which abnormal returns are measured. Defined by start and end days relative to the event date (e.g., [-5, +5] for an 11-day window). See Window Selection.

Expected Return

The return a stock is predicted to earn in the absence of the event, based on a model estimated during the estimation window. The abnormal return is the difference between the actual and expected return. See Expected Return Models.

F

Fama-French Models

Multi-factor asset pricing models. The 3-Factor model includes market, size (SMB), and value (HML) factors. The 5-Factor model adds profitability (RMW) and investment (CMA). Used as expected return models when size and value effects need to be controlled. See Expected Return Models.

G

Gap Period (Guard Period)

The buffer between the estimation window and the event window, typically 5-20 trading days. Prevents pre-event information leakage from contaminating the estimation of normal returns. See Window Selection.

GARCH

Generalized Autoregressive Conditional Heteroskedasticity. A model that captures time-varying volatility (volatility clustering) in return data. Useful as an expected return model when return variance is not constant. See Expected Return Models.

Generalized Sign Test

A non-parametric test that tests whether the proportion of positive abnormal returns exceeds the expected proportion estimated from the estimation window. Unlike the simple Sign test, it adjusts for asymmetric return distributions. See AAR & CAAR Tests.

K

Kolari-Pynnonen Test

A multi-event test that extends the BMP test to handle both event-induced variance and cross-sectional correlation (event clustering). Recommended when events share common dates. See AAR & CAAR Tests.

L

Log Return (Continuously Compounded Return)

The natural logarithm of the price relative: r_t = ln(P_t / P_{t-1}). Log returns are additive over time and approximately normally distributed. See Return Calculation.

M

Market Adjusted Model

An expected return model that assumes the stock's expected return equals the market return. No estimation window regression is needed. AR = R_stock - R_market. See Expected Return Models.

Market Model

The most common expected return model. It regresses stock returns on market returns during the estimation window to estimate alpha and beta. Expected return during the event window is alpha + beta * R_market. See Expected Return Models.

Mean Adjusted Model

An expected return model that uses the stock's own average return during the estimation window as the expected return. No market index is needed. AR = R_stock - mean(R_stock in estimation window). See Expected Return Models.

N

Non-Parametric Test

A test that does not assume a specific distribution for abnormal returns. Examples include the Sign test, Generalized Sign test, and Rank test. More robust to outliers and skewness than parametric tests, but generally less powerful when normality holds. See Choosing the Right Test Statistic.

Normal Return

Synonym for expected return. The return a stock would have earned in the absence of the event. The term "normal return" is used interchangeably with "expected return" in the event study literature.

P

Parametric Test

A test that assumes abnormal returns follow a specific distribution (typically normal). Examples include the Cross-Sectional t-test, Patell Z, and BMP test. More powerful than non-parametric tests when the distributional assumptions hold. See Choosing the Right Test Statistic.

Patell Z Test

A multi-event test that standardizes each firm's abnormal return by its estimation-window standard deviation before aggregating. Accounts for cross-firm heterogeneity in return variance. Not robust to event-induced variance changes. See AAR & CAAR Tests.

Power (Statistical Power)

The probability of rejecting the null hypothesis when it is false (i.e., detecting a real abnormal return). Power depends on sample size, event window length, abnormal return magnitude, and test statistic choice. See Power Analysis.

R

Rank Test

A non-parametric test that ranks event-window abnormal returns against estimation-window residuals. Robust to non-normality and outliers. Uses magnitude information (unlike the Sign test). See AAR & CAAR Tests.

S

Sign Test

A non-parametric test that counts the proportion of positive abnormal returns and tests whether it exceeds 50%. Ignores the magnitude of abnormal returns. Robust to outliers but low power. See AAR & CAAR Tests.

Simple Return (Arithmetic Return)

The percentage price change: R_t = (P_t - P_{t-1}) / P_{t-1}. Simple returns are intuitive and aggregate correctly across portfolios, but they are not additive over time. See Return Calculation.

Standardized Abnormal Return (SAR)

An abnormal return divided by its estimated standard deviation from the estimation window: SAR = AR / sd(AR). Standardization puts all firms on a common scale, accounting for differences in return volatility across firms.

Synthetic Control

A method that constructs a counterfactual ("synthetic") version of the treated unit from a weighted combination of untreated units. Used when there is only one treated firm or when traditional event study assumptions do not hold. See Synthetic Control.

T

Test Statistic

A numerical value computed from abnormal returns to test whether the event had a significant impact. The test statistic is compared to a critical value or used to compute a p-value. See Test Statistics Overview.

Trading Day

A day on which the stock exchange is open for trading. Event study dates are expressed in trading days, not calendar days. Weekends and public holidays are excluded. Day 0 is the event trading day.

W

Wild Bootstrap

A resampling method for constructing p-values and confidence intervals that is robust to heteroskedasticity and non-normality. In event studies, wild bootstrap is used when standard asymptotic tests are unreliable due to small samples or non-normal returns. See Inference & Robustness.

What Should I Do Next?

• Event Study Methodology Guide — start from the beginning
• Data Preparation — prepare your data for analysis
• Expected Return Models — choose the right benchmark model
• Choosing the Right Test Statistic — select appropriate tests for your study