A regression, a window, and a question: did the price move more than the market explains? The method is standard. The choices inside it decide cases.
Start a conversation with Cournot, the Institute’s damages concierge, already scoped to event studies. Pick a starting point, or describe the dispute directly.
An event study asks whether a security's price reacted abnormally to specific information, and answers with a number and a significance level. The mechanics are settled. A control model, typically regressing the stock's returns on a market index and an industry index over an estimation period of clean trading days, establishes how the stock normally moves with its environment. On the event date, the model predicts the return the environment would have produced; the difference between the actual return and that prediction is the abnormal return; and the estimation period's residual volatility supplies the yardstick for whether that abnormal return is statistically distinguishable from noise. Inside the standard frame, every choice is consequential and contestable. The estimation window must be clean of the alleged fraud and representative of the stock's behavior. The event window must be long enough to capture the reaction and short enough to exclude everything else, and widening it after seeing results is the kind of choice that gets noticed. The index selection, the treatment of days with multiple news items, the handling of thinly traded securities where prices update sluggishly, and the significance threshold applied are each the subject of a literature and of recurring courtroom fights. The instrument's strength is that it disciplines argument: assertions about what moved the price become testable. Its limit is equally important: a single-day study has limited power, so an insignificant result is not proof of no effect, and both sides routinely overclaim in opposite directions on exactly that point.
Standard machinery, consequential choices.
The clean period that defines normal behavior. Contaminated or unrepresentative windows bias everything downstream.
Market and industry indices absorbing what the environment did. Index choice is argued, and results can turn on it.
The days over which reaction is measured. Chosen on principle before results, or attacked as chosen for results.
Actual return minus predicted. The quantity everything else exists to certify.
The abnormal return against the stock’s own noise level. Conventions are strong, and departures are conspicuous.
Earnings, guidance, analyst actions, market shocks on the same day. The standard fight, covered in loss causation.
How the Institute approaches an event study.
It appears at every stage of a securities case, doing different work at each.
A single-firm, single-day event study often lacks the power to detect real but modest price effects, so "not statistically significant" means "not proven by this instrument", not "disproven". Experts on both sides blur that line in opposite directions, and tribunals increasingly notice.
The ones made after the results were visible. The order of operations is the tell: an event window that happens to end exactly where the reaction stops helping, an industry index swapped for one that makes the abnormal return larger, dates added or dropped from the analysis without a principled reason stated in advance. Then the mechanics: whether the estimation period is clean of the alleged fraud, whether significance is computed with the stock's own volatility or something more convenient, whether days with confounding news are handled by a stated rule or by discretion. A well-run study documents its choices and their rationale before results; a results-driven study cannot, and the deposition questions that expose the difference are short.
Imperfectly, which is why confounded days are where cases concentrate. The event study measures the day's total abnormal return; it cannot by itself allocate that return between the corrective disclosure and the unrelated earnings miss announced the same morning. The available approaches are all arguments rather than algorithms: intraday analysis where trading data can time the reactions separately, analyst commentary parsing what the market understood, valuation of the confounding item from fundamentals so the residual can be attributed, and comparison to peer reactions on the same day. Each can be done rigorously, and none is mechanical. An expert who assigns the entire movement to their side's preferred cause without engaging the confound has produced the standard target for the other side's cross.
With decreasing force, and the honest analysis says so. The methodology assumes prices that update quickly in active trading. Corporate bonds trade sporadically and often by dealer quotation, so daily abnormal returns may be measuring staleness rather than reaction, and the analysis shifts toward trade-level data and wider windows, with efficiency itself harder to establish. Thinly traded stocks raise the same issue in milder form, along with statistical complications from nonsynchronous trading. For private instruments there is no price series and no event study; value must be built from fundamentals. The boundary matters doctrinally as well, since the fraud-on-the-market framework leans on efficiency, and counsel will know how the governing law treats each instrument class; the economics either supports the showing or it does not.
Through an inflation analysis that the event study certifies but does not complete. The study establishes which disclosures moved the price and by how much; damages then require a per-share inflation series over the class period: how much of the price, on each day, was attributable to the alleged misstatements. The common construction works backwards from the corrective disclosures, allocating the certified declines to the misrepresentations they revealed and carrying the inflation back through the period, with choices, constant dollar inflation, constant percentage, or something structured, that must be tied to the case's theory of what was concealed and when its significance changed. Statutory features, including the lookback cap, then shape the final arithmetic. Each trader's recovery depends on when they bought and sold against that series, which is where the class-wide model meets the claims process.
Describe the security and the dates. The Institute will help you see where the event study will be fought.