Library Post-Earnings Announcement Drift
The Anomaly That Outlived Its Benchmark
Wall Street’s oldest anomaly isn’t that investors ignore earnings. It’s that understanding an earnings report takes time.
In 1968, Ray Ball and Philip Brown published An Empirical Evaluation of Accounting Income Numbers and turned earnings analysis from theory into measurement. Their finding was uncomfortable for the efficient-market view: stocks kept moving in the direction of the earnings news after the news was already public.
That continuation became Post-Earnings Announcement Drift, or PEAD. Companies reporting unexpectedly strong results tended to keep outperforming. Companies reporting unexpectedly weak results tended to keep lagging.
The entire question is what counts as unexpected.
What Ball and Brown actually found
They had no standardized analyst-estimate database. So they built an expectation from a company’s own earnings history and a market-based model, then separated the part investors could reasonably have anticipated from the part that was genuinely new.
Three results came out of it.
Earnings changes moved with stock returns, confirming that accounting results carried information investors cared about. Prices frequently started moving before the announcement, because investors were already learning from industry data, competitors, management commentary, and the economy. And only a modest share of the total move happened around the announcement itself — the rest kept arriving afterward.
William Beaver, publishing the same year, came at it from the other side: earnings announcements produced unusually heavy trading volume and unusually high return volatility. The releases plainly contained new information. Later work found no sign that quarterly reports had become less informative as other financial data grew more available; if anything, the abnormal volume and volatility around them increased.
Foster, Olsen, and Shevlin documented the drift again in 1984. Bernard and Thomas, in 1989 and 1990, showed it was consistent with a genuinely delayed reaction rather than an artifact of how abnormal returns were being calculated.
One report, two drifts
It helps to split PEAD into two windows that behave differently.
Initial drift
The first phase starts after the earnings gap and runs roughly five trading days.
The gap is the market’s first interpretation — headline numbers, guidance, and whatever management said on the call. Initial drift is what happens once investors read the actual release, analysts rebuild their models, and institutions adjust size.
When the report is genuinely better than expected, the move accelerates. When it is worse than investors had assumed, the selling continues past the first reaction.
The gap tells you how the market reacted. The initial drift tells you whether that reaction survived contact with the details.
This window is what the Earnings Whispers Power Rating is built for, and it is measured from the post-announcement opening price through the fifth session — excluding the gap, deliberately.
Longer-term drift
The second phase can run most of the quarter, until the company reports again and resets the story.
This one depends on more than the EPS line. Analysts revise estimates. Investors reassess whether the results were high quality and repeatable. Management speaks at conferences. Customers, suppliers, and competitors keep supplying evidence about the industry.
A strong report backed by raised guidance, improving demand, expanding margins, and rising estimates can outperform for months. A weak one backed by cuts does the reverse.
The Earnings Whisper Grade measures this window, weighting earnings quality and durability over the immediate reaction.
More on the Earnings Whisper Grade · More on the Earnings Quarter
Why the market drifts at all
The standard explanation is underreaction, and the mechanism is mundane: an earnings release contains far more than one number. Revenue, margins, cash flow, guidance, expenses, segments, customer activity, and the outlook all have to be weighed. A company can beat the estimate and cut guidance. Another can miss and reveal that the business has turned.
The market prices the headline fast. It prices the meaning slowly.
Attention appears to matter a great deal. Research on Friday announcements — when the fewest people are paying attention — found the immediate response was about 15% smaller while the delayed response was roughly 70% larger, with lower abnormal volume alongside it.
The same pattern shows up on crowded reporting days. When many companies report at once, the immediate price and volume reaction is weaker and the subsequent drift is stronger. In one study the 60-day abnormal return spread between high and low earnings surprises reached 7.14% on crowded days, against a statistically insignificant 2.30% on quiet ones.
The drift grows when the market has more trouble processing the information — not when the information is more surprising.
Limited attention is unlikely to be the whole story. Behavioral bias, transaction costs, short-selling constraints, disagreement about interpretation, and the plain risk of arbitraging a single stock all plausibly contribute. The durable lesson is narrower: publicly available and immediately understood are not the same thing.
The benchmark problem
As analyst forecasts became widely collected, the consensus estimate became the default definition of a surprise. Above consensus was a beat. Below was a miss.
But the published consensus is not the expectation sitting in the stock price.
Some estimates were filed weeks earlier. Analysts often privately expect a company to clear their own published number without revising it. Investors fold in guidance, recent industry data, peer results, and channel checks that never touch the estimate database at all.
Which produces the most familiar outcome of earnings season: a company beats consensus and the stock falls. The result was better than the published number and worse than what investors were actually expecting.
This matters for measuring PEAD, not just for trading it. If the benchmark is set too low, a reported beat may not be good news, and a drift measured against it will look weaker than it is. Modern research finding that PEAD has faded for large caps since the mid-2000s is measuring against consensus.
More on what happened to the drift
Whisper forecasts
Whisper forecasts emerged as the alternative: unofficial EPS numbers circulating among investors and traders, meant to capture what the market genuinely expected rather than what the estimate database recorded.
In 1999, Mark Bagnoli, Messod Beneish, and Susan Watts published Whisper Forecasts of Quarterly Earnings per Share, comparing 943 whisper forecasts against 3,546 First Call analyst forecasts across 127 companies between January 1995 and May 1997.
Whispers were on average more accurate than the First Call forecasts and were better proxies for the market’s expectations. Strategies built on the relationship between the two generated abnormal returns.
Part of that edge was timing — whispers appeared closer to the announcement and could absorb more recent information. But when the researchers restricted the comparison to analyst forecasts issued within five days of the report, the whisper advantage weakened without disappearing.
The study had real limits. The sample was small and 91% of the companies were technology firms. It established that whispers carried information beyond the consensus. It did not establish that any given whisper was superior in every industry.
The point it did settle has held up: the market’s real expectation can differ materially from the official consensus.
The Earnings Whisper number
Earnings Whispers began publishing Earnings Whisper numbers in 1998 — not as another average of analyst estimates, but as an attempt to measure the expectation embedded in the market, drawing on available forecasts, analyst behavior, historical tendencies, and investor expectations around the release.
The consensus tells you whether a company cleared the published Wall Street number. The Earnings Whisper number is meant to tell you whether it cleared the number that was actually priced in.
More on the Earnings Whisper® number
The Earnings Whisper Drift
Average cumulative return from the post-announcement open through the fifth trading day. The earnings gap is excluded.
The middle group beat the consensus. Headlines could accurately have called those results better than expected, because against the published estimate they were. The stocks drifted lower anyway — which is the cleanest available evidence that the consensus was not the bar investors had actually set.
What starts the initial drift
Initial drift usually begins when the opening reaction fails to capture the full distance between the report and expectations.
A company clears both the consensus and the Whisper, raises guidance, and shows demand strengthening. The stock gaps up — but analysts and institutions still need days to move estimates and positions, and that buying arrives after the open.
The reverse: a company beats consensus, misses the Whisper, guides cautiously, or reports lower-quality earnings. The stock holds up initially because the headline says beat, then weakens as investors work through the details.
Sentiment amplifies both. Bearish positioning into a genuinely strong report means the market is caught wrong — short covering, upgrades, and new buying compound the move. Overwhelmingly bullish positioning means even good results have a much higher bar to clear.
What sustains the longer one
Initial drift becomes longer-term drift when the report changes the expected path of future results.
A one-time accounting benefit can manufacture a beat without improving the business, and that surprise rarely supports a quarter-long move. A report showing accelerating demand, more recurring revenue, expanding margins, better cash flow, or materially higher guidance forces analysts to raise numbers across several future quarters — and that revision process takes weeks.
The strongest positive drift tends to stack up:
- The company cleared realistic expectations, not just the published consensus
- Guidance and commentary support higher future estimates
- Earnings quality is strong and comes from the core business
- Analysts revise estimates higher after the report
- Sentiment leaves room for expectations to improve
- The stock shows relative strength after the initial reaction
Negative drift is the mirror image: disappointing results, weaker guidance, falling estimates, deteriorating trends, and expectations that were too high walking in.
A framework, not a rule
PEAD does not mean buy every beat and sell every miss.
Some companies beat because expectations were walked down first. Some miss on timing that says nothing about the business. A stock can report superb results and fall because the price already assumed better. And rates, economic data, industry trends, and overall market direction can bury a company-specific signal entirely.
What PEAD offers is a read on when the probabilities favor continuation. For the initial drift, that means the relationship between the report, real expectations, sentiment, positioning, and the opening reaction. For the longer drift, it means earnings quality, guidance, revisions, and whether the report actually changed the outlook.
The first reaction is only the first reaction
The gap gets the attention because it is immediate and dramatic. It is also the market’s first attempt at pricing news it has had minutes to read.
Ball and Brown showed prices keep moving after the information is public. Analyst forecasts made the surprise easier to quantify. Whisper forecasts made the benchmark closer to what investors actually expected.
The anomaly was never that investors ignore earnings. It is that reading an earnings report takes time — and that the yardstick you measure the surprise against decides what you find.
Then the company reports again, expectations reset, and it starts over.
Drift figures average the cumulative return from the first opening price after each earnings announcement through the fifth subsequent trading day, and exclude the earnings gap itself. The sample covers approximately 117,000 published Earnings Whisper numbers from August 7, 1998. Day 4 is not collected; the chart’s horizontal axis is spaced by trading day rather than by data point. Figures exclude transaction costs, dividends, and slippage, and are not a guarantee of future results.