What Event Studies Actually Measure

An event study asks a deceptively simple question: when something happens to a company — a spill, a lawsuit, a surprise filing — how much of the subsequent share-price move can we attribute to that event rather than to the market’s general mood?

The machinery, briefly

The standard approach compares the stock’s actual return over an event window with the return you would have expected given how the market moved. The difference is the abnormal return. Sum it over several days and you get a cumulative abnormal return, the headline number in most studies.

So far, so clean. The trouble starts with three assumptions that are easy to state and hard to satisfy.

Where the method quietly bends

First, the benchmark has to be right. If you benchmark a Canadian energy producer against a US index in a week when currency and commodity moves diverge, your “abnormal” return is partly a benchmark artifact. Same-country, same-currency benchmarks are boring — and correct.

Second, the event has to be a surprise. Markets price expectations continuously. If a regulatory penalty was telegraphed for months, the announcement-day return measures only the gap between the outcome and what was already priced in, not the cost of the penalty itself.

Third, the window has to match the information flow. A one-day window misses slow-developing events; a sixty-day window drowns the signal in noise. There is no universally right choice, only a defensible one — which is why serious studies report several.

Reading results like an adult

When someone shows you a cumulative abnormal return, ask three questions:

1. Abnormal relative to what? (The benchmark.)
2. Surprising relative to what? (The prior expectation.)
3. Measured over what window, and why that one?

An event study that answers all three is evidence. One that answers none is a chart with a story attached.