How Researchers Measure What a Piece of News Is Really Worth
When a stock jumps 8 percent the day a company announces an acquisition, it is not obvious how much of that move was caused by the announcement versus ordinary market-wide noise. The event study method was built to separate the two, and it now underlies most of the rigorous evidence on whether markets absorb news quickly, slowly, or too early.
The core mechanism: isolating abnormal return
A stock's price moves every day for reasons that have nothing to do with company-specific news: the overall market rises or falls, interest rate expectations shift, sector sentiment changes. An event study is the standard statistical method for separating a stock's reaction to a specific announcement, an earnings report, a merger, a regulatory ruling, from that ordinary background noise. The method compares the stock's actual return around the event date to the return it would have been expected to earn based on its normal relationship to the broader market, using an estimation period well before the event to establish what "normal" looks like for that particular stock.
The workhorse tool for estimating that normal relationship is the market model: expected return = alpha + (beta x market return), where alpha and beta are estimated by regressing the stock's daily returns against the market's daily returns over a clean estimation window, typically 120 to 250 trading days ending well before the event so the event itself does not contaminate the estimate. Once alpha and beta are known, the abnormal return on any given day is simply abnormal return = actual return - expected return, the piece of the day's move the market model could not explain by ordinary market-wide movement, and which is therefore attributed to the event.
The market model described here is the most common baseline, but it is not the only one used in practice, and knowing the alternatives helps explain why two published studies of the same type of event sometimes report slightly different results. A simpler market-adjusted returns model skips the regression entirely and just subtracts the market's return from the stock's return each day, implicitly assuming a beta of exactly 1.0 for every stock, a shortcut that is faster to compute but less accurate for stocks whose true market sensitivity differs meaningfully from average. More sophisticated variants extend the market model to a multifactor benchmark, controlling for size and value exposure alongside overall market movement, since a small, statistically cheap stock's "normal" return pattern differs systematically from a large, richly valued one, and failing to account for that can make an ordinary factor-driven move look, incorrectly, like an event-driven abnormal return.
The math: abnormal return, CAR, and a significance test
Worked example 1: computing a single day's abnormal return. A company reports stronger-than-expected earnings. Using a clean prior estimation window, its market-model alpha is estimated at roughly 0.01% per day (essentially negligible) and its beta at 1.2. On the announcement day, the broad market index rises 0.5%, so the expected return is 0.01% + (1.2 x 0.5%) = 0.01% + 0.6% = 0.61%. The stock's actual return that day is 4.6%. The abnormal return is 4.6% - 0.61% = 3.99%, the portion of the day's move attributable to the earnings surprise rather than to the market's general direction that day.
Worked example 2: cumulating abnormal returns and testing whether the reaction is statistically real. Researchers rarely stop at a single day, because information can leak before an announcement or continue to be absorbed afterward. Suppose the same stock shows an abnormal return of 0.3% the day before the announcement (a possible sign of leakage), 3.99% on the announcement day itself, and 0.2% the day after (possible slow absorption). The cumulative abnormal return, or CAR, over this three-day window is 0.3% + 3.99% + 0.2% = 4.49%.
To determine whether this 4.49% figure reflects a genuine reaction rather than ordinary daily noise, researchers compare it against the stock's typical abnormal-return volatility, estimated from the same clean estimation window used for alpha and beta. Suppose that window shows a daily abnormal-return standard deviation of 1.0%. Under the null hypothesis of no real event effect, and assuming daily abnormal returns are independent, the standard deviation of a three-day cumulative abnormal return is 1.0% x sqrt(3) = 1.0% x 1.732 = 1.73%. The resulting t-statistic is 4.49% / 1.73% = 2.60, comfortably above the conventional 1.96 threshold used for a two-tailed test at the 5% significance level, so a researcher would conclude this reaction is very unlikely to be the product of ordinary noise alone.
It is worth walking through what happens when this test comes out the other way, since a null result is just as informative to a careful reader. Suppose a second, more routine announcement, a minor product update rather than an earnings surprise, produces a three-day CAR of just 0.6%, against the same 1.73% estimated standard deviation. The t-statistic is 0.6% / 1.73% = 0.35, far below the 1.96 threshold, meaning this small positive drift is statistically indistinguishable from noise and should not be interpreted as evidence the announcement moved the stock at all. This distinction, between a reaction large enough to clear a formal significance threshold and one that merely looks directionally positive on a chart, is exactly what separates a rigorous event study from an eyeballed price chart, and it is the reason academic event studies typically pool many events of the same type together rather than drawing conclusions from a single announcement, since averaging across a large sample sharpens the statistical test considerably further.
What decades of event studies show
Applied across thousands of earnings announcements, merger announcements, stock splits, dividend changes, and regulatory decisions, the event study method has produced one of the more durable and widely replicated patterns in empirical finance: for most routine announcement types, the bulk of the cumulative abnormal return appears within a very short window, often a single trading day, around the announcement itself, with comparatively little abnormal movement in the days immediately before (barring information leakage or insider trading, which event studies have also been used to help detect) and a documented but modest continuation afterward known as post-earnings-announcement drift, where stocks that beat expectations tend to keep drifting upward, and stocks that miss tend to keep drifting downward, for several weeks following the announcement.
This drift finding is one of the more persistent, mild anomalies in the market efficiency literature: it has shown up across multiple decades and multiple markets, yet it has proven difficult to trade profitably at scale once realistic transaction costs and the fact that the effect is concentrated in smaller, less liquid stocks are accounted for. Event studies focused on merger announcements have separately documented a fairly consistent pattern in which target company shareholders capture most of the announced deal premium almost immediately, while acquiring company shareholders often see a flat or mildly negative abnormal return, a finding that has become a standard reference point for evaluating whether a given acquisition was priced generously or conservatively for each side.
Event studies have also been used well beyond corporate announcements, extending to regulatory changes, index reconstitutions (what happens to a stock's price when it is added to or dropped from a major benchmark index), and even broader macroeconomic or policy announcements, such as central bank interest rate decisions, applied to entire sectors rather than a single company. The reconstitution literature in particular has produced a striking, well-replicated result: stocks added to a widely tracked index tend to see a measurable positive abnormal return around the announcement and effective dates, even though the addition itself changes nothing about the company's underlying cash flows, a pattern generally attributed to the mechanical buying pressure from index funds required to track the benchmark, and one of the cleaner pieces of evidence that trading flows themselves, separate from fundamental information, can move prices in a statistically detectable and temporarily persistent way.
Reading event-driven moves in a real portfolio
For an individual investor, the direct application of the full event study machinery, estimation windows, market models, formal significance tests, is more of an analytical background skill than a daily portfolio tool, but the underlying discipline transfers directly to how you should interpret news-driven price moves in your own holdings. When a stock in your portfolio jumps or drops sharply on a specific piece of news, the useful question is not simply "did the price move a lot" but "how much of that move is attributable to this company specifically, once the day's overall market and sector movement is subtracted out," which is precisely the abnormal-return calculation in worked example 1, informally applied.
The post-earnings drift finding also has a direct, practical implication worth internalizing: a large earnings-day reaction does not necessarily mean the market has finished digesting the news. Chasing a stock immediately after a large announcement-day jump, hoping to capture a piece of the continuing move, sits at the intersection of a real but modest, well-documented effect and a well-documented tendency for the effect to be smaller after realistic trading costs than it appears in academic samples built without those costs, which is a reasonable basis for caution rather than a reliable trading strategy.
The index reconstitution finding described above has a more directly usable implication for long-term investors, particularly those holding individual stocks alongside index funds. If a stock you own is added to a major benchmark, some short-term price appreciation from mechanical index-fund buying is a documented, plausible outcome, not something to interpret as new fundamental validation of the business. Conversely, a stock being removed from an index can see mechanical selling pressure unrelated to anything the company did wrong operationally. Recognizing this distinction, between a price move caused by trading flow and a price move caused by genuine new information about future cash flows, is the single most transferable lesson the event study literature offers an ordinary investor trying to make sense of a sudden, unexplained price swing in a specific holding. The same distinction is worth applying to a broader portfolio rebalance forced by any large, mechanical trading flow, a major benchmark provider changing its methodology, for instance, which can move an entire sector's pricing temporarily without any change to the underlying businesses involved.
Actionable breakdown
- Separate a stock's move into market-wide and company-specific components.
- Check whether abnormal movement began before the official announcement.
- Treat a single day's reaction as informative, not necessarily complete.
- Remember post-earnings drift is real but modest and cost-sensitive.
- Compare any merger reaction against the historical acquirer-target pattern.
Common pitfalls
- Attributing an entire price move to one announcement without adjusting for the market.
- Ignoring confounding events landing on or near the same date.
- Treating a large raw price move as automatically statistically significant.
- Assuming post-earnings drift is large enough to trade profitably after costs.
The bottom line
Event studies strip out ordinary market noise to reveal how much of a price move a specific announcement actually caused and how quickly the market absorbed it, and the same market-adjusted thinking is worth applying informally to any sharp move in your own holdings.
All articles · Are markets efficient · Stock analysis · Mutual fund and analyst performance · Random walks and the EMH