How to Do an Event Study for a Finance Thesis

How to Do an Event Study for a Finance Thesis

A step-by-step, survivorship-free workflow for the most defensible kind of empirical finance thesis.

Why an Event Study Suits a Thesis

An event study is one of the most thesis-friendly designs in empirical finance. It has a clear question, a known timeline, and a standard methodology that an examiner already trusts. You measure how prices behave in a window around a specific, dated event, which gives you a sharp hypothesis and a clean identification rather than a vague predictive claim. The methodology, from event windows to abnormal returns, is the backbone of Systematic Event-Driven Trading.

The catch is that the standard methodology assumes clean data, and most of the ways a thesis goes wrong are upstream of the statistics, in how the sample was built.

The Workflow, Step by Step

Define the event and window. Choose a dated event, such as an equity offering, and fix the estimation and event windows in advance. Build a survivorship-free sample. Include companies that were later delisted, because excluding them biases the result, as our piece on survivorship bias shows. Use point-in-time data. Align prices and any conditioning variables to what was known on each date, following our guide to point-in-time data.

Compute abnormal returns. Subtract an expected-return benchmark to isolate the event effect. Test and report honestly. Check significance, then stress the result with realistic costs before claiming it. A full data-driven example of this end to end is in our approach to shorting de-SPACs.

The Steps Where Theses Fail

Most of these failures are invisible in the output, which is why an examiner probes the data construction rather than the regression.

Sourcing the Event Data

The hardest practical step is usually building the dated event set without lookahead. Alphanume's dilution events dataset provides financing events parsed from filings with accurate disclosure dates, and the historical market cap dataset supplies the point-in-time size used for conditioning and weighting. Starting from a clean event feed lets you focus the thesis on the analysis rather than the data engineering.

A Worked Example

Suppose your thesis studies returns around at-the-market offering disclosures. You fix a short event window and a prior estimation window, assemble every issuer that disclosed such a program including those that later delisted, align prices to each disclosure date, and compute cumulative abnormal returns. You then repeat the analysis after subtracting realistic borrow and trading costs to show whether the effect survives implementation.

Reported this way, the thesis is hard to attack. An examiner can see that the window was fixed in advance, that survivors did not bias the sample, that no future information leaked in, and that the effect was tested against costs. Those are exactly the four questions an event-study examiner asks, answered before they are posed.

Defending the Thesis

A thesis defense on an event study tends to follow a predictable path, and preparing for it is mostly about the data. Expect questions on how delisted firms were treated, whether any variable could have used future information, why the windows were chosen, and how robust the effect is to costs and specification. Each of those is answerable in advance if the study was built with discipline, which turns the defense into a confirmation rather than an interrogation.

It helps to include a short robustness section that varies the window length and the cost assumptions. Showing that the result holds across reasonable choices preempts the most common line of attack and demonstrates the maturity examiners are looking for in a finance thesis.

How to Choose

Get the sample right and the statistics follow. For a finance thesis, fix your windows in advance, build a survivorship-free and point-in-time sample, compute abnormal returns, and report with realistic costs. An event study done with that discipline is one of the most defensible projects a student can submit.