Frequently Asked Questions
Everything you need to know about event study methodology, our tools, and how to apply them in your research.
6 Questions
General
What is an event study?
An event study is a statistical method used to measure the impact of a specific event on the value of a financial security. It works by comparing actual stock returns around the event date to the returns that would have been expected without the event (the "normal" or "expected" return). The difference is the abnormal return, which quantifies the event's effect.
Who uses event studies?
Event studies are widely used by academic researchers in finance, accounting, and economics; by regulators and antitrust authorities assessing market impact; by litigation consultants quantifying damages; and by corporate finance professionals analyzing M&A, earnings announcements, regulatory changes, and other corporate events.
What data do I need for an event study?
You need three things: (1) a list of events with firm identifiers and event dates, (2) daily adjusted closing prices or returns for each firm in your sample, and (3) daily adjusted closing prices or returns for the market index (e.g., S&P 500). For multi-factor models like Fama-French, you also need factor return data.
How long does an event study take to run?
With prepared data, running an event study takes seconds in the R package or web app. The most time-consuming part is usually data preparation -- gathering clean, adjusted price data and defining your event list. For a typical study with 50-200 events, the computational analysis completes in under a minute.
Are there free tools for event studies?
Yes. EventStudy provides a free and open-source R package, a free browser-based web app (powered by WebAssembly), and a Google Sheets template. The R package and web app include all major return models and statistical tests at no cost.
What is the difference between AR, CAR, AAR, and CAAR?
AR (Abnormal Return) is the difference between the actual return and expected return for a single firm on a single day. CAR (Cumulative Abnormal Return) is the sum of ARs over the event window for one firm. AAR (Average Abnormal Return) is the cross-sectional mean of ARs across all firms on a given day. CAAR (Cumulative Average Abnormal Return) is the sum of AARs over the event window.
6 Questions
Methodology
Which expected return model should I use?
The Market Model is the most common choice and works well for most studies. Use the Fama-French 3- or 5-Factor Model when you need to control for size, value, profitability, or investment factors. GARCH models are appropriate when volatility clustering is a concern. The Mean Adjusted Model is suitable when no suitable market index is available.
Which test statistic should I choose?
Start with the Patell Z-test or BMP test for your primary results. The BMP test is robust to event-induced variance changes, and the Kolari-Pynnonen test additionally corrects for cross-sectional correlation. Always report at least one non-parametric test (e.g., the Generalized Sign Test or Rank Test) as a robustness check, since non-parametric tests do not assume normality.
How long should my estimation window be?
The standard recommendation is 120-250 trading days. A 250-day (one year) estimation window provides stable parameter estimates for most models. Shorter windows (120 days) may be necessary when firms have limited trading history. The estimation window should not overlap with other events affecting the same firm.
What is event-induced variance and why does it matter?
Event-induced variance refers to the increase in return volatility around the event date. Standard parametric tests like the cross-sectional t-test can be biased when event-induced variance is present, leading to over-rejection of the null hypothesis. The BMP test (Boehmer, Musumeci, Poulsen, 1991) and Kolari-Pynnonen test correct for this by using standardized abnormal returns.
Can I study multiple events for the same firm?
Yes, but you need to ensure that event windows do not overlap. Overlapping event windows violate the independence assumption required by most test statistics. If events are close together, you can shorten the event window, exclude overlapping events, or use the Calendar-Time Portfolio approach, which is robust to cross-sectional dependence.
What is the difference between short-run and long-run event studies?
Short-run event studies examine abnormal returns over a window of a few days around the event (typically -10 to +10 trading days). Long-run event studies measure abnormal performance over months or years (e.g., 12-36 months) using methods like BHAR or calendar-time portfolios. Long-run studies are methodologically more challenging due to compounding effects and model misspecification.
5 Questions
R Package
How do I install the EventStudy R package?
Install from GitHub using devtools: devtools::install_github("sipemu/eventstudy"). The package requires R >= 4.0.0 and will automatically install all dependencies.
Is the package available on CRAN?
The package is currently distributed via GitHub. Install it with devtools::install_github("sipemu/eventstudy"). GitHub distribution allows for more frequent updates and faster bug fixes.
What dependencies does the package have?
The EventStudy R package depends on standard CRAN packages including zoo, xts, sandwich, and lmtest for time series and robust standard errors, plus ggplot2 for plotting. All dependencies are installed automatically when you install the package.
Can I run panel Difference-in-Differences with the R package?
Yes. The R package includes six modern DiD estimators for staggered treatment designs: TWFE, Callaway-Sant'Anna, Sun-Abraham, de Chaisemartin-D'Haultfoeuille, Borusyak-Jaravel-Spiess, and Gardner. These methods produce consistent estimates even with heterogeneous treatment effects.
What export formats are supported?
The R package can export results to styled Excel workbooks (multi-sheet with formatting), LaTeX tables (ready for academic papers), CSV files, and publication-quality PDF/PNG plots. You can also access all results programmatically as R data frames for further analysis.
4 Questions
Web App
Is my data safe in the web app?
Yes. The Event Study App runs entirely in your browser using WebAssembly. No data is uploaded to or processed on any server. All computation happens locally on your machine, so your data never leaves your browser.
What file formats does the web app accept?
The web app accepts four separate CSV files (comma-separated, with YYYY-MM-DD date format) or a single Excel (.xlsx) workbook with named sheets: events, firm_data, market_data, and optionally factor_data. Built-in validation catches formatting errors before analysis begins.
How does the web app compare to the R package?
The web app offers full feature parity for standard event studies: 10 return models, 11 test statistics, interactive charts, and Excel/PNG export. For advanced features like panel DiD, synthetic control, intraday analysis, GARCH models, and power simulation, use the R package.
Does the web app work on mobile devices?
The web app is designed for desktop use. While it will load on tablets and large mobile screens, the interactive charts, data tables, and result panels are optimized for screen widths of 1024px or wider. We recommend using a laptop or desktop computer for the best experience.
3 Questions
Google Sheets
What return models are supported in the Google Sheets template?
The Google Sheets template supports the Market Model, Market Adjusted Model, and Mean Adjusted Model. These three models cover the majority of standard event study applications and work well for most academic and professional analyses.
Are there data limits in the Google Sheets template?
The template is optimized for single-event studies with up to approximately 500 trading days of data. For multi-event studies with larger datasets, we recommend the R package or web app, which can handle thousands of events efficiently.
Can I use the template in Microsoft Excel?
The template is designed for Google Sheets and uses Google Sheets-specific functions. While you can download it as an Excel file, some formulas may not translate perfectly. For Excel users, we recommend the R package or the web app, which can export styled Excel workbooks.
4 Questions
Research & Academic Use
Can I use EventStudy tools for my thesis?
Yes, absolutely. Our tools are widely used in bachelor's, master's, and doctoral theses. The R package is particularly well-suited for academic work because it produces reproducible results and supports LaTeX export. Many students start with the web app for exploration and switch to the R package for their final analysis.
How should I cite the EventStudy R package?
You can cite the EventStudy R package using the citation information provided in the package documentation. In R, run citation("EventStudy") after loading the package to get the recommended BibTeX entry for your bibliography.
What is the minimum sample size for an event study?
There is no strict minimum, but statistical power increases with sample size. For multi-event studies, 30+ events is a common guideline for parametric tests to perform reliably. For single-event studies, the length of the estimation window (typically 120-250 days) provides the statistical basis. Use the power analysis feature in the R package to determine the sample size needed for your specific research design.
How do I handle confounding events?
Confounding events -- other events occurring during the event window -- can bias your results. Common approaches include: (1) excluding firms with confounding events, (2) narrowing the event window to reduce the chance of contamination, (3) using a control group design, or (4) using multivariate regression on CARs to control for confounding factors.
Still Have Questions?
Explore our documentation for in-depth explanations, or try the web app to see event studies in action.
Ready to run your first event study?
Jump straight into the Event Study App and follow along with your own data.