The tidy version of markets says a surprise gets priced in instantly. Someone reads the result, everyone agrees what it means, the price jumps once and settles. Clean, fast, done.
That is not quite what the record shows. After a genuinely big surprise, prices have tended to keep sliding in the same direction for weeks afterwards. The jump happens, and then a slow follow-through. This is called post-earnings-announcement drift, and it has been studied since the 1960s.
Think of a rumour spreading through a large office. It does not reach everyone the instant it is spoken. It moves desk by desk over days — and the mood shifts gradually rather than all at once.
The obvious question is: surprise compared to what? The most common answer is the analysts’ forecast — did the company beat what the professionals expected? That is fine, but it measures whether the forecasters were wrong, which is a different thing from whether the business changed.
There is an older approach that asks the simpler question. Compare this quarter to the same quarter a year ago. That comparison cancels out seasonality on its own: a retailer’s Christmas quarter is measured against its previous Christmas quarter, never against a quiet spring.
But a raw change is not enough, because a big number means different things for different companies. Some businesses earn almost the same amount every quarter. Others swing wildly by nature. A £1 change is an earthquake for the first and an ordinary Tuesday for the second.
So the change gets divided by how much that company’s results normally move. The result says how unusual this is for this business — not how big it is in absolute terms. A steady company posting a modest beat can score as a bigger surprise than a volatile one posting a huge jump.
- Compare to the same quarter a year ago, not to the last quarter.
- Divide by how much that company normally swings.
- The answer is in standard deviations: how strange is this, for them?
- Big numbers and big surprises are not the same thing.
The effect is real enough to have survived decades of scrutiny, though it has weakened as more people traded on it, and it was always strongest in smaller companies that fewer analysts follow.[1]
Which brings the honest caveats. Earnings arrive in a crowded few weeks each quarter, so hundreds of "separate" observations are really a handful of shared moments — the sample looks bigger than the evidence is. And the effect lives largely in small, thinly-covered companies; among the household names, it is much fainter.
Finisdom’s Earnings Drift page pools every quarter the tracked companies have reported since 2008, taken straight from their filings with the US regulator, and shows what actually happened over the following 5, 21 and 63 trading days — with the sample size on every number and the caveats printed alongside.