Earnings season: what moves, what does not, and the drift that died

Earnings season is the one part of the stock calendar that is scheduled, public and violent. Every quarter a listed company reports a completed period, and the law says how: when the company "publicly releases material nonpublic information" about results, it furnishes that release to the SEC on Form 8-K, Item 2.02, within four business days — in practice the same morning or evening. Almost every company does it when the market is closed, before the open or after the close, which is why the first trade after earnings is so often a gap: the stock does not move to its new price, it reopens there.

This page is about what that does to a trader and what the research actually supports. It is on the stocks shelf, but the desk trades futures, and the honest reason the page exists is that the desk's members ask about it — and that the evidence here disagrees with some of what gets taught.

The day itself: a premium for holding through it

The most robust finding is the least dramatic. Savor and Wilson (Journal of Finance, 2016) sorted stocks by whether they were scheduled to report that week, 1974–2009, and found that announcers earned more: an equal-weighted portfolio long the announcers and short everyone else returned about 0.39% per week (t-statistic 14), roughly 0.23% value-weighted, and the effect was "extremely persistent across stocks over horizons going up to 20 years." Their reading is that it is a risk premium — announcement days carry systematic risk, because what one company says moves beliefs about the economy — and the market pays for bearing it.

Two things to notice before anyone turns that into a trade. The premium is for the week, across hundreds of stocks, netted long and short; a single name on a single day is a coin with a fat tail on each side. And it is paid for taking the gap, not for predicting it. There is no sentence in that paper about knowing which way a stock will go.

The drift that died

The thing textbooks still teach is post-earnings announcement drift: after a positive surprise the stock keeps rising for weeks, after a negative one it keeps falling, as if the market were slow to absorb the news. Bernard and Thomas documented it in 1989 and it became one of the most-cited anomalies in finance.

Martineau re-ran it through 2019 and titled the result Rest in Peace Post-Earnings Announcement Drift. His sentence: "For large stocks, PEAD have been non-existent since 2006 but has only disappeared recently for microcap stocks." In the modern data, prices "fully reflect earnings surprises on the announcement date." The 60-day post-announcement coefficient on the surprise, for everything above the smallest fifth of NYSE stocks, went from 0.003 in 1984–1990 to −0.002 in 2016–2019 — not smaller, reversed. Among microcaps it had faded to a statistically insignificant 0.002 by the same years.

So the mechanism a drift trader is counting on — the market being slow — is the mechanism the market fixed. What is left of "earnings momentum" lives in stocks too small and illiquid for the costs to let you have it. The factors page has the same story at larger scale: an anomaly that is published, easy and liquid gets arbitraged until it is none of those things.

The "expected move"

Before a report, the options market prices a straddle expiring just after it, and the straddle's price is a forecast of how far the stock will move in either direction — the implied move, usually quoted as a percentage. ORATS, which measures this across hundreds of names every season, reported in 2019 that the actual move divided by the implied move "hovered around 85% for all weeks" of the seasons they tracked, and hit 114% in one unusually wild week. Read plainly: on average the market slightly overestimates the size of the move, and sometimes badly underestimates it. An implied move is a range the stock usually stays inside and sometimes blows through — a useful size for a stop that will not be respected, which brings us to the part that costs people money.

A stop does not work through a gap

A stop order is an instruction that becomes a market order when the price touches the stop. If the stock closes at 100 with your stop at 95 and reopens at 80, the stop triggers at the open and fills near 80. FINRA's notice on the subject says it in its own words: "a stop order becomes a market order when the stop price is reached," and "the price at which a stop order ultimately is executed may be very different from the investor's" stop price. In a sharp decline "a sell stop order also is more likely to result in an execution well below the stop price."

The desk's own position-sizing page sizes a trade from the stop distance, and the whole method assumes the stop can be hit near its price. Across an earnings release that assumption is false by construction: the loss you can take is not "stop distance times shares," it is "whatever the reopening price is, times shares." A trader who sizes an earnings hold like an ordinary trade has taken a position several times larger than the one on the ticket. The honest sizing for a held-through-earnings position starts from the implied move, treats it as the stop distance that applies, and then remembers that the implied move is an average the actual move exceeds in a meaningful share of weeks — ORATS' figures above are a ratio of averages, not a promise about any one name.

What the research does not say

It does not say which way a stock will go after a beat; it does not say a "beat" moves a stock up (the market reacts to the surprise against what it already priced, and to guidance, not to the headline); it does not say the announcement premium is available to a single stock on a single day; and it does not say a smart trader can tell, from the numbers, what the opening print will be. The desk has read the papers above in full and quotes them because they are the evidence that exists, not because they contain a trade.

Where the desk stands

The desk trades an index future, and an index does not have an earnings date. The gap risk described here is why its hands do not hold single stocks and why the overnight page treats any holding across a scheduled release as a sizing problem first. A member who trades stocks through earnings is trading something the desk's tools were not built to size; the arithmetic on this page is the arithmetic that applies.

• Index funds vs stock picking: what fifty years of data say
• Momentum, value and quality: the factors with evidence, and their drawdowns
• Trend following on stocks — what moving-average rules did and did not do

← Stocks · All of Learn