Post-Earnings Announcement Drift: How to Trade It
Nathan Njuguna
Post-earnings announcement drift is the tendency of a stock to keep moving in the direction of an earnings surprise for days or weeks after the report. You can still trade it, but recent studies find the drift in large US stocks is far smaller than it was, so the edge now sits mostly in the first few days and in small, costly-to-trade stocks.
This guide explains how the drift works, what the research says from 1968 to 2026, and what our own record of 1,182 closed earnings trades shows about holding periods. It also walks through a real JPMorgan Chase trade from July 2026, day by day, including what happened to the drift after the exit.
Key takeaways
- Post-earnings announcement drift (PEAD) means prices keep moving toward a surprise after the report, instead of adjusting all at once.
- Older studies found about 4% per quarter between the best and worst surprises. Recent work finds little or none in large stocks since 2006.
- An independent 2026 study found the 60-day gap fell from 2.6 points for 2023–24 announcements to 0.5 points for 2025–26.
- In our Earnings AI record, trades closed within a day averaged +6.81%, while trades held 8 to 30 days averaged −3.24%.
- Trading costs decide the result: a 7.9% yearly gross edge in one 2026 study fell to 2.1% at 0.30% per trade.
What is post-earnings announcement drift?
When a company reports earnings that are better or worse than expected, its share price jumps or drops. In an efficient market, that move would be complete within minutes. Post-earnings announcement drift describes what researchers found instead: prices of companies with good surprises kept rising for weeks, and prices of companies with bad surprises kept falling.
Post-earnings announcement drift (PEAD): the continued movement of a stock's price in the direction of its earnings surprise after the announcement. Earnings surprise: the gap between reported earnings and what analysts expected. Excess return: the stock's return minus the market's return over the same days.
Ray Ball and Philip Brown first described the pattern in 1968. For decades it was one of the most cited puzzles in finance, because it seemed to offer a simple, public signal: buy after big positive surprises, sell or short after big negative ones, and wait.
The usual explanations are that investors underreact to earnings news, and that small, costly-to-trade stocks are not worth the effort for professionals to correct.
How do researchers measure the drift?
Most studies follow the same three steps. First, they score each surprise. Second, they sort companies into groups, often five (quintiles) or ten (deciles), from the most negative surprise to the most positive. Third, they track each group's excess return over the following 60 trading days, about one quarter.
The surprise score is usually standardized unexpected earnings, or SUE: the surprise divided by how much the company's earnings normally vary. A $0.10 beat means more for a company whose earnings are very stable than for one whose earnings swing widely. Our guide to earnings surprise shows how to calculate both the simple percentage and SUE.
The headline result is the spread: the excess return of the top group minus the bottom group. A spread near zero means no drift. Livnat and Mendenhall (2006) found the drift is larger when the surprise is measured against analyst forecasts than against a simple model based on past earnings.
Does post-earnings announcement drift still exist?
It depends on the company size and the years studied. The early numbers were large. Later studies found much less, and the most recent ones disagree about whether the drift has come back.
| Study | Sample | Main finding |
|---|---|---|
| Ball & Brown (1968) | US companies, annual earnings | First evidence that prices keep moving after the report |
| Foster, Olsen & Shevlin (1984) | US quarterly earnings | About 25% a year before costs from buying the best and shorting the worst surprises |
| Bernard & Thomas (1989) | US quarterly earnings | About 2% per side over 60 days, roughly 4% per quarter or 18% a year |
| Livnat & Mendenhall (2006) | US, analyst forecasts | Drift is larger when surprises are measured against analyst forecasts |
| Martineau (2022) | US stocks, by size | No drift in large stocks since 2006; none in microcaps since 2016 |
| Nyllinge & Oldenburg (2025) | 220,228 announcements, 2005–2024 | Large-cap 60-day drift rose from 0.52% to 1.99% after 2020 |
| Subrahmanyam (2025) | US stocks, with and without microcaps | Drift is statistically significant with microcaps (t = 2.18) but not without them (t = 1.43) |
| Gupta (2026) | 43,111 US announcements, 2023–2026 | +1.7-point spread overall, falling to +0.5 for 2025–26 announcements |
Charles Martineau's 2022 paper in the Critical Finance Review is the strongest case that the drift is gone. His conclusion is direct:
"In modern financial markets, stock prices fully reflect earnings surprises on the announcement date, leading to the disappearance of post-earnings announcement drifts (PEAD)."
— Charles Martineau, University of Toronto, "Rest in Peace Post-Earnings Announcement Drift," Critical Finance Review (2022)
Not everyone agrees. In January 2026, the UCLA Anderson Review noted two 2025 papers, by Dickerson, Julliard and Mueller in the Journal of Financial Economics and by Hirshleifer, Peng and Wang in the Review of Financial Studies, that argue the drift is still alive. It also reported Avanidhar Subrahmanyam's finding that the effect depends heavily on microcaps, which make up about 3% of US market value.
The fairest summary is this. The drift is not a reliable large-cap strategy anymore. Where it exists, it is smaller, concentrated in small companies, and sensitive to how you measure it.
Is the drift shrinking in recent years?
The most recent data we found says yes. Gaurav Gupta's independent 2026 study covers 43,111 US earnings announcements from January 2023 to October 2026. It sorts companies into surprise quintiles and measures returns from trading day 2 to day 61 against the S&P 500.
The drift is shrinking: best minus worst surprises, 60-day return gap by company size
Announcements 2023–24Announcements 2025–26
Across the whole period, the best quintile beat the worst by 1.7 percentage points over 60 days, a statistically meaningful result. But the gap shrank sharply over time, from 2.6 points for announcements in 2023–24 to 0.5 points for 2025–26. For small companies it turned negative. Microcaps kept the widest gap, 4.2 points, but they are the costliest to trade; large companies kept 1.7 points.
Gupta also tested a signal based on the price reaction around the report rather than the earnings surprise. It was weaker: a 1.3-point spread that was not statistically reliable. That matters for anyone who buys simply because a stock jumped on earnings day.
A Stockholm School of Economics thesis by Nyllinge and Oldenburg (2025) points the other way for large caps. Across 220,228 announcements from 2005 to 2024, it found the 60-day large-cap drift rose from 0.52% in 2005–2020 to 1.99% after 2020. Different samples, methods and years give different answers, which is itself a warning against treating the drift as a sure thing.
What does the drift look like in a real trade? A JPMorgan example
TruthSayer's Earnings AI took this trade in its funded account before members received the alert. We tracked JPMorgan Chase (JPM) against the S&P 500 (SPY) for 60 trading days from the close before the report.
- Open before the report. JPMorgan reported its second quarter of 2026 before the market opened on Tuesday, July 14. Earnings AI bought 5 shares on July 13 at $334.33, a position of $1,671.65.
- Score the surprise. Adjusted EPS was $6.14 against a Zacks consensus of $5.59, a surprise of +9.8%. Other sources used other estimates, which changes the size of the surprise, as our earnings surprise guide shows.
- Measure the day-one reaction. JPM closed at $343.02 on July 14, up 2.54% from $334.53 the day before. SPY rose 0.37%, so the excess return was +2.17 percentage points.
- Watch the drift. Over the next nine sessions, JPM kept pulling ahead of the market. By July 28, day 11, its excess return reached +7.95 points. That extra +5.78 points after the first day is the drift.
- Exit and count the result. The position was sold on August 14, day 24, at $362.63. The gross gain was $141.50. After $9.98 in commission, the net gain was $131.52, or 7.87%. On that day JPM was 4.82 points ahead of SPY.
- Check what holding longer would have done. The lead kept fading. By September 22, day 50, JPM trailed SPY by 1.58 points. By October 6, day 60, it trailed by 4.98 points. A textbook 60-day drift trade would have ended behind the market.
Post-earnings announcement drift in practice: JPMorgan vs the S&P 500 after its Q2 2026 beat
The lesson is not that the drift is fake. JPM did drift after a solid beat, and the trade captured part of it. The lesson is that the drift peaked early and then reversed, so the holding period mattered more than the entry.
How long should you hold after earnings? What our data shows
We sorted every closed Earnings AI stock trade with a recorded holding period, 1,182 trades, by how long it was held. The difference between short and long holds is large.
Earnings AI: average return by holding period, 1,182 closed stock trades
Trades closed within a day averaged +6.81%, and 93.0% of them made money. Trades held two to seven days averaged +1.87%. Beyond that, the average turned negative: −3.24% for 8 to 30 days, −5.78% for 31 to 60 days and −2.77% for more than 60 days. Win rates fell to about 48% for the two longest groups.
Put together, the 621 trades closed within a week averaged about +4.71%. The 561 trades held longer averaged about −3.59%.
Holding time is partly a result, not only a choice. Earnings AI uses a trailing stop, so trades that work tend to be closed sooner, and trades that linger are often the ones where the report did not deliver. The data does not prove that holding longer causes losses. It does say that when the move has not come within a week, it has rarely come later.
This fits the research. If prices now absorb most of a surprise on the day of the report, the trade is mainly about the report itself, and whatever drift remains is short. For more on why live results like these differ from backtests, see our guide to backtested vs live-traded signals.
What makes a drift setup stronger or weaker?
No single number makes a good drift trade. The research points to a few conditions that tend to separate setups with some edge from setups with none.
| Condition | Stronger setup | Weaker setup | Evidence |
|---|---|---|---|
| Surprise size | Top or bottom fifth of surprises | Small beat or miss, near consensus | Sorting by surprise is how every study finds the drift |
| Surprise measure | Against analyst consensus, like for like | Against last year's EPS only | Livnat & Mendenhall (2006) |
| Company size | Smaller companies | Large companies | Martineau (2022); Subrahmanyam (2025) |
| Trading cost | About 0.10% per trade or less | 0.30% per trade or more | Gupta (2026) |
| Time since report | First days to first week | Weeks to months later | Gupta (2026); our Earnings AI hold data |
The conditions pull against each other. Small companies show the most drift but cost the most to trade. Large companies are cheap to trade but drift the least. That tension is the main reason the classic strategy is hard to profit from today.
How does trading the drift compare with other earnings approaches?
| Approach | What you need | Recent evidence | Fits the data? |
|---|---|---|---|
| Buy after a beat, hold 60 days | Patience, a list of surprises | Little or no drift in large stocks | No |
| Buy after a beat, hold up to a week | Daily attention, a time exit | Most remaining drift is early | Yes |
| Long best and short worst small-cap surprises | Shorting access, low costs | Largest spreads, highest costs | No for most individuals |
| Position before the report, exit quickly | A view on the report, small size | Earnings AI trades closed within a day averaged +6.81% | Yes |
| Ignore the report, hold for the long term | A view on the business | Not a drift strategy | Yes for long-term investors |
Our guide on how to trade earnings covers the before-the-report approach, including how to read the expected move from options prices.
Who is a drift strategy best for?
Best for active investors with low costs
If you can check positions daily during earnings season, trade at low cost and follow a time exit, a short drift hold after a large surprise is a reasonable thing to test. Keep the test small and measure excess return, not just gains.
Not a fit for buy-and-hold investors or small accounts
Long-term investors gain little from timing a few weeks of drift. Small accounts face high costs per trade: on the JPM position above, $9.98 in commission was about 0.6% of the $1,671.65 position. Gupta's edge was mostly gone at costs of 0.30% per trade.
How much does trading the drift cost in money and time?
Costs decide whether a small edge survives. Gupta's long-short portfolio of mid- and large-cap stocks earned 7.9% a year before costs. At 0.10% per trade it kept 6.0%, at 0.20% it kept 4.0%, and at 0.30% only 2.1%. Those costs include commissions and the bid-ask spread, the gap between the buying and selling price.
Taxes matter too. In the US, gains on shares held one year or less are short-term and taxed at ordinary income rates, according to the IRS. A strategy built on holds of days or weeks will produce almost only short-term gains.
Time is the other cost. Building a surprise list, checking each reaction against the market and managing exits takes daily attention during the busiest weeks of each earnings season. Our stock profit calculator helps with the per-trade math, including commissions.
What are the biggest risks and mistakes with drift trading?
Assuming a big beat means a long drift
Nvidia beat analyst estimates by 5.7% on August 26, 2026, reporting $2.22 per share against $2.10 expected by LSEG, and guided to about 70% revenue growth for its next fiscal year against about 45% expected. The stock rose 8.63% the next day while SPY rose 0.66%. By the second session, the gain from the report-day close had already shrunk to 3.61%. By November 6, Nvidia was 11.08% below its August 26 close, while SPY was down 3.99%.
Measuring raw return instead of excess return
A stock that rises 3% while the market rises 4% has not drifted up. Always subtract the market return over the same days, as we did for JPM.
Ignoring costs on small positions
A fixed commission is a large share of a small trade. Check your total cost per round trip before you start, and compare it with the 0.10% to 0.30% range in the research.
Holding without a time exit
Without a planned end date, a drift trade can turn into a long-term holding you never meant to own. Both JPM and Nvidia gave back their early excess returns.
Write the date of your time exit next to the order before you place it. If the stock has not pulled further ahead of the market by that date, close the position and move on.
Is "buy the beat and hold for 60 days" still good advice?
We do not think so. The classic rule came from data that ended decades ago, when information moved slower and trading cost more. Recent large-cap evidence, our JPM example and our own hold-period data all point the same way: whatever drift remains shows up in the first days, and the 60-day hold is where the edge is lost.
The common view treats the drift as the opportunity and the report as noise. The data suggests the opposite. The report itself is where most of the move happens, and the following week is a short window to confirm it, not a 60-day ride. In our record, trades closed within a week averaged about +4.71%, while those held longer averaged about −3.59%.
The 2025 papers that defend the drift, and the Stockholm thesis, deserve attention. But they sort thousands of stocks into portfolios, which is a different exercise from buying a handful of names after a beat.
The TruthSayer Drift Check: six steps before you act on a surprise
We use these six checks when a report surprises and the stock starts to move. The one-week time exit is our working rule based on our own data, not an industry standard.
- Score the surprise like for like. Compare adjusted EPS with the adjusted consensus, and note which data provider you used.
- Measure the reaction against the market. Subtract SPY's return from the stock's return on the report day. A beat that lagged the market is not a drift candidate.
- Check size and cost. Expect less drift in large companies. If your cost per trade is 0.30% or more, the published edge is mostly gone.
- Cap the position. Earnings AI puts no more than 5% of the portfolio in any one position and uses a trailing stop set at three times the stock's average daily range (ATR).
- Set a one-week time exit. If the excess return has not grown within seven days, close the trade.
- Log excess return. After 20 to 30 trades, check your average excess return after costs, not your win rate.
How does TruthSayer AI use earnings surprises and drift?
Earnings AI scans US-listed companies ahead of their reports and opens positions in a funded account, typically one to four hours before the report. It trades the surprise itself, not the drift afterward. Members see each alert only after the trade is placed, and every trade, winners and losers, stays on the public list.
As of November 8, 2026, the Earnings AI page shows a 31.85% return since inception against 35.09% for the S&P 500 over the same dates, so the strategy trails the index. It also shows a 72.8% win rate, an average hold of 29 days, a Sharpe ratio of 3.94 and a beta of 0.30.
TruthSayer AI claims compliance with GIPS®, the global performance reporting standard; see our GIPS reports and verification. For a checklist to apply to any signal provider, including us, read how to judge a signal service.
Earnings AI figures come from its public trade list and performance page on November 8, 2026: 1,182 closed stock trades with a recorded holding period, returns net of commissions as published. Bucket averages, win rates and excess returns are our calculations. JPM, Nvidia and SPY prices are the last regular-session 15-minute bar each day from Intrinio. Study results are as reported by their authors.
What should you do after an earnings surprise?
- You own the stock for the long term: read the report for what it says about the business. A few weeks of drift should not change a long-term decision.
- You bought after a beat: set the time exit now. If the stock is not further ahead of the market within a week, close it.
- You sold and it kept rising: compare with your plan, not with the best possible exit. JPM's lead peaked on day 11 and was gone by day 50.
- You want to test the idea: paper-trade 20 surprises first and record excess returns after realistic costs.
To see how signals like these compare with other approaches, read our guide to AI stock signals, or see every live trade, including the losers, on Earnings AI.
The bottom line: Post-earnings announcement drift is real in history but much weaker today, especially in large companies, and costs can erase what remains. If you trade it, act within days, measure returns against the market, and use a time exit. The TruthSayer Drift Check turns that into six steps: score the surprise like for like, measure the reaction, check size and cost, cap the position, exit within a week if nothing happens, and log excess returns.
Frequently asked questions
What is post-earnings announcement drift in simple terms?
It is the tendency of a stock to keep rising after a good earnings surprise, or keep falling after a bad one, instead of adjusting all at once. Researchers first described it in 1968. It suggests investors take time to absorb earnings news, though recent studies find much less of it in large companies.
Does post-earnings announcement drift still work?
Mostly not in large US stocks. Martineau (2022) found no drift in large companies since 2006. A 2026 study of 43,111 announcements found a 0.5-point 60-day gap for 2025–26 reports, down from 2.6 points for 2023–24. Some 2025 papers disagree, but costs usually eat what remains for individual investors.
How long does post-earnings drift last?
Classic studies measured it over about 60 trading days. Recent evidence suggests most of what remains happens within the first days. In our JPMorgan example from July 2026, the stock's lead over the market peaked on day 11 and had reversed by day 50. Our trades closed within a week did best.
How do you calculate standardized unexpected earnings (SUE)?
Subtract the expected EPS from reported EPS, then divide by the standard deviation of past surprises or forecast errors. The result shows how unusual the surprise is for that company. Researchers sort stocks by SUE into groups and compare the returns of the highest and lowest groups over the following weeks.
I bought after an earnings beat and the stock is flat a week later. Should I sell?
If your plan had a one-week time exit, follow it. In our record of 1,182 closed Earnings AI trades, those held 8 to 30 days averaged −3.24%, against +1.87% for two to seven days. A stock that has not pulled ahead of the market within a week rarely did so later.
I sold after earnings and the stock kept rising. Did I make a mistake?
Not if you followed your plan. Judge the decision on what you knew, not on the best possible exit. Early drift often reverses: JPMorgan led the market by 7.95 points on day 11 after its July 2026 report, then trailed it by 4.98 points by day 60. Log the outcome and keep your rules.
Sources
- Charles Martineau, Rest in Peace Post-Earnings Announcement Drift, Critical Finance Review (2022)
- UCLA Anderson Review, Is Post-Earnings Announcement Drift a Thing Again? (Jan 21, 2026)
- Gaurav Gupta, Is Post-Earnings Drift Still Tradable? (accessed Nov 8, 2026)
- Nyllinge & Oldenburg, master thesis on post-earnings announcement drift, Stockholm School of Economics (2025)
- Fink, review of post-earnings announcement drift research, University of Graz working paper 2020-04 (2020), covering Ball & Brown (1968), Foster, Olsen & Shevlin (1984), Bernard & Thomas (1989) and Livnat & Mendenhall (2006)
- Zacks Equity Research via Yahoo Finance, JPMorgan Chase (JPM) tops Q2 earnings estimates (Jul 14, 2026)
- CNBC, Nvidia second-quarter fiscal 2027 earnings report (Aug 26, 2026)
- Internal Revenue Service, Topic no. 409, Capital gains and losses (accessed Nov 8, 2026)
- Intrinio, US equity 15-minute price history for JPM, NVDA and SPY (Jul–Nov 2026)
- TruthSayer AI, Earnings AI performance and trade list (Nov 8, 2026)
This article is for educational purposes only and is not investment advice. Past performance does not guarantee future results.
Nathan Njuguna
Nathan Njuguna is Lead Data and Machine Learning Engineer at TruthSayer AI, specializing in quantitative research, machine learning, and data engineering. He develops and validates systematic trading strategies, builds predictive models, and designs backtesting frameworks so trading signals are statistically validated, reproducible, and ready for production.
