Event Study Methodology
- Introduced by
- A. Craig MacKinlay; Stephen J. Brown & Jerold B. Warner
- Year
- 1997
- Type
- Abnormal-return estimation pipeline for valuing corporate events
- Family
- Empirical corporate finance / strategy performance
What does it do?
Event study methodology measures the stock-market reaction to a discrete corporate event by isolating the portion of a firm's return that cannot be explained by normal market movements.
Under semi-strong market efficiency, new information about an acquisition, earnings announcement, alliance, CEO change, or regulatory shock is impounded into prices almost immediately, so the abnormal return around the event date is a clean, forward-looking estimate of the event's value consequences. A. Craig MacKinlay's 1997 survey codified the canonical pipeline -- define the event and windows, estimate a normal-return benchmark, compute abnormal returns, accumulate them into a cumulative abnormal return (CAR), and test significance. Brown and Warner's 1985 study established the statistical properties of these procedures with daily data, showing when simple methods are well specified and how variance and clustering must be handled. The method is the workhorse for linking strategic decisions to shareholder value.
The intuition behind the method
Suppose a firm announces it is buying a rival. Its stock moves -- but markets move every day for reasons that have nothing to do with this deal. The question is: how much of today's jump is about the acquisition, and how much is just the market drifting up or down? Event study methodology answers this by first learning, from a quiet pre-event period, how this stock normally tracks the overall market. That gives a prediction for what the return 'should' have been on the announcement day given how the market did. The gap between the actual return and that prediction is the abnormal return -- the market's verdict on the news itself. Add up those abnormal returns over the few days around the announcement and you get the cumulative abnormal return, a dollar-weighted opinion poll of investors on whether the decision creates or destroys value. Because prices react fast and aggregate the views of many informed traders, this small window can reveal what years of accounting data struggle to show.
History
Event study methodology grew out of efficient-markets research in the late 1960s -- Fama, Fisher, Jensen, and Roll's stock-split study and Ball and Brown's earnings work established that prices respond rapidly to information. Through the 1970s and 1980s the technique became the standard tool of empirical corporate finance and, increasingly, of strategic management for valuing corporate decisions. Brown and Warner's 1980 and 1985 papers were pivotal, using simulation to characterize the statistical behavior of abnormal-return tests and to show that simple market-model procedures with daily data are generally well specified provided clustering and variance shifts are handled. A. Craig MacKinlay's 1997 Journal of Economic Literature survey consolidated the field into the canonical estimation-and-testing pipeline now taught and applied, and the method remains central to research linking strategy to shareholder value.
When to use it
Use an event study when you can identify a discrete, dated, and largely unanticipated event and you want a market-based estimate of its value consequences for publicly traded firms. It is the method of choice in strategy and finance for evaluating acquisitions, divestitures, alliances, new-product or R&D announcements, executive succession, litigation, and regulatory changes, because traded prices give an immediate, forward-looking valuation that accounting measures lag. The approach requires reasonably efficient and liquid markets, clean and precise event dates, an uncontaminated estimation window, and ideally a single event per firm-window so that the reaction can be attributed. It is poorly suited to long-horizon questions, to thinly traded securities, to events that are heavily anticipated or confounded by simultaneous news, or to private firms, where market prices are unavailable and the central identifying assumption fails.
Limitations and considerations
- Validity hinges on market efficiency and on the event being unanticipated; leakage or gradual learning smears the reaction and biases short-window estimates.
- Confounding events in the same window cannot be disentangled, so attributing the abnormal return to a single decision requires careful event screening.
- Event-date clustering creates cross-sectional dependence that inflates test statistics unless explicitly corrected, as Brown and Warner demonstrated.
- Long-horizon abnormal-return measurement is sensitive to the benchmark model and to skewness, making conclusions about persistent value creation fragile.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Delivers a forward-looking, market-based valuation of a decision that is available immediately, rather than waiting years for accounting performance to materialize.
- Rests on a transparent, replicable pipeline with well-understood statistical properties documented by Brown and Warner for daily data.
- Isolates firm-specific effects by removing market-wide movements through the market model, improving the power to detect event effects.
- Scales across many firms and event types, letting researchers test average value effects and cross-sectional drivers of the reaction.
Applications
- Estimating the shareholder value created or destroyed by mergers and acquisitions for acquirers and targets.
- Valuing strategic alliances, joint ventures, and partnership announcements as discrete information events.
- Measuring the market's reaction to new-product introductions, R&D breakthroughs, and innovation announcements.
- Assessing the stock-price impact of CEO succession, board changes, and other governance events.
- Quantifying how regulatory rulings, litigation outcomes, and policy shocks affect affected firms.
How it works
How it works
The analysis begins by pinning down the event and partitioning calendar time into non-overlapping windows. The event date (tau equals zero) is the moment information reaches the market -- typically the first public announcement, not the closing date. MacKinlay distinguishes an estimation window before the event, used to learn the normal-return model, from an event window bracketing the announcement, over which abnormal returns are measured. The windows must be disjoint so that the event itself does not contaminate the benchmark, and the event window is kept short to limit confounding news while still capturing leakage and delayed reaction.
Formula
\tau=0\ \text{(event date)},\quad \text{estimation }[T_0,T_1],\ \text{event }[T_1+1,T_2],\ \text{disjoint}Over the estimation window the analyst fits a model of the security's normal return -- the return expected absent the event. MacKinlay treats the market model, a regression of the firm's return on a market index, as the standard choice because it removes market-wide movements and improves power over the simpler constant-mean or market-adjusted models. The estimated intercept and slope (alpha and beta) capture how the stock ordinarily co-moves with the market, providing the counterfactual against which the event-window returns are judged.
Formula
R_{it}=\alpha_i+\beta_i R_{mt}+\varepsilon_{it},\quad t\in[T_0,T_1]With the benchmark in hand, the abnormal return on each event-window day is the actual return minus the return predicted by the fitted normal-return model. This residual is the heart of the method: it strips out ordinary market exposure so that what remains reflects firm-specific, event-related information. Brown and Warner showed that for daily data this straightforward construction is generally well specified, and they characterized the data properties -- thin trading, non-normality, and shifting variance -- that the analyst must watch when forming these residuals.
Formula
AR_{it}=R_{it}-(\hat\alpha_i+\hat\beta_i R_{mt})Abnormal returns are then accumulated across time and across firms. Summing a single firm's abnormal returns over the event window yields its cumulative abnormal return (CAR), the total value effect of the event for that firm; averaging CARs across the sample of events gives the mean cumulative abnormal return, the typical market response to that class of decision. MacKinlay emphasizes that aggregation over the window captures both anticipation and delayed reaction, and aggregation across firms sharpens inference about the average economic significance of the event type.
Formula
CAR_i(t_1,t_2)=\sum_{t=t_1}^{t_2} AR_{it},\qquad \overline{CAR}=\tfrac{1}{N}\sum_{i=1}^{N} CAR_iFinally the analyst tests whether the aggregated abnormal return differs from zero. The standard test statistic divides the mean CAR by its estimated standard error and, under the null of no abnormal performance, is approximately standard normal. Brown and Warner's simulations established the size and power of such tests with daily returns and warned that event-date clustering induces cross-sectional correlation, so that tests assuming independence can be badly oversized -- motivating standardized, cross-sectionally adjusted, or nonparametric statistics when events share calendar time.
Formula
J=\frac{\overline{CAR}}{\sqrt{\widehat{\operatorname{Var}}(\overline{CAR})}}\ \xrightarrow{\,H_0\,}\ \mathcal N(0,1)Common pitfalls
- Dating the event by the deal completion or news-coverage date rather than the first information release, which mis-times the window and dilutes the measured reaction.
- Letting the estimation window overlap the event or other firm news, contaminating the alpha and beta used as the counterfactual.
- Applying tests that assume cross-sectional independence to events clustered in calendar time, producing spuriously significant CARs.
- Stretching the event window so wide that unrelated news enters, or so narrow that anticipation and delayed reaction are missed.
Frequently asked questions
What is the difference between an abnormal return and a cumulative abnormal return?
An abnormal return is the firm's actual return on a single day minus the return predicted by its normal-return benchmark, so it captures the event-related surprise on that day. A cumulative abnormal return (CAR) sums those daily abnormal returns over the event window to give the total value effect of the event for that firm. MacKinlay treats aggregation over the window as essential because the market's reaction often spans several days through anticipation and delayed adjustment; averaging CARs across firms then yields the typical response to a class of events and is the quantity most strategy studies report and test.
Why use the market model instead of just looking at the raw return?
Raw returns mix the event's effect with market-wide movements that have nothing to do with the firm's decision. The market model regresses the firm's return on a market index over a clean estimation window, learning how the stock normally co-moves with the market, and then subtracts that expected co-movement from the event-window return. MacKinlay notes this removes a major source of variance and so increases the power to detect the event's true effect relative to the constant-mean or market-adjusted alternatives. The abnormal return that remains is the part of the return attributable to firm-specific, event-related news.
Why does clustering of event dates threaten the statistical tests?
When many sample firms experience the event on or near the same calendar date -- common for regulatory or macro shocks -- their abnormal returns are correlated through shared market conditions. Brown and Warner showed that standard tests assuming cross-sectional independence then understate the true standard error, so the test statistic is too large and CARs look significant when they are not. The remedy is to use standardized cross-sectional statistics, portfolio or calendar-time approaches, or nonparametric tests that account for the dependence, rather than treating clustered events as independent observations.
Related methods
Sources
MacKinlay, A. C. (1997). Event Studies in Economics and Finance. Journal of Economic Literature, 35(1), 13-39.
Open DOI referenceCopied.Brown, S. J., & Warner, J. B. (1985). Using Daily Stock Returns: The Case of Event Studies. Journal of Financial Economics, 14(1), 3-31.
Open DOI referenceCopied.