Post-earnings announcement drift, usually shortened to PEAD, is the name for a documented research pattern in which returns after an earnings announcement have been related to the direction of the earnings news. It is an anomaly to investigate, not an instruction to trade: the result depends on how the surprise, date, universe, execution and holding period are defined (Bernard and Thomas, Brown and Pope).
The important distinction is easy to miss. An earnings announcement is an event; PEAD is a claim about a rule applied after that event. A chart of a stock rising after a report is not evidence that the rule existed before the outcome.
What post-earnings announcement drift means
PEAD describes a tendency investigated in academic studies: after earnings news, subsequent returns have sometimes continued in the same broad direction as the surprise. The classic literature framed the question as either delayed price response or compensation for risk, rather than treating the observed relation as a settled mechanical cause (Bernard and Thomas, Ball and Brown).
An earnings surprise must therefore be defined before any test begins. It might mean reported earnings relative to a forecast, relative to an earlier value, or a standardised measure built from a chosen data set. Those are different inputs, and substituting one for another after seeing results is a form of rule-shopping.
announcement data available
↓
pre-defined surprise measure
↓
pre-defined ranking or threshold
↓
entry after an executable timestamp
↓
pre-defined exit, costs and delisting treatmentWhy the label is not the edge
The label compresses several choices that can change a result: the securities included, the source and timestamp of expectations, the event date, the way surprises are ranked, the delay before entry, the exit rule and the costs. Research on revenue surprises also treats the relationship between revenue news and earnings news as a separate conditioning choice, which is exactly why a vague “earnings beat” rule is not enough (Jegadeesh and Livnat, Bernard and Thomas).
A practical test also needs a definition of what information was known when. A database can make revised estimates, late filings, delisted shares or clean corporate-action histories look simpler than they were. The same basic problem appears in any event study: a result is only as credible as its point-in-time inputs.
This is why PEAD is not interchangeable with generic momentum. Momentum ranks returns over a chosen past window. PEAD begins with a corporate disclosure and then asks whether the information in that disclosure adds anything once the full trading rule is fixed (Ball and Brown, Jegadeesh and Livnat).
The criticism belongs in the method
A published anomaly is evidence worth examining, not a promise of an implementable return. Brown and Pope explicitly examine sample survival and bid-ask spreads in relation to reported drift; Bernard and Thomas present competing interpretations of the pattern (Brown and Pope, Bernard and Thomas).
That leaves several open questions for any current implementation:
| Question | Why it can change the conclusion |
|---|---|
| Was the surprise observable at the claimed decision time? | A late or revised input can create look-ahead bias. |
| Are delisted and acquired shares in the universe? | Removing inconvenient histories can flatter a result. |
| Can the order be filled after the release? | Spread, slippage and liquidity can change an event-driven rule. |
| Was the threshold selected before the test? | Repeatedly trying cut-offs makes the chosen one less persuasive. |
| Does the pattern persist in unseen periods? | A historical relation can weaken after discovery or under a new market regime. |
The criticism is not a reason to ignore the literature. It is the reason to turn its broad finding into a falsifiable implementation rather than a story about a handful of charts.
How you'd actually test PEAD
Start with a written research protocol. Specify the market and share universe, the exact earnings and forecast data vendor, the timestamp that makes each observation eligible, the surprise formula, the portfolio construction, the holding period, rebalancing, corporate-action treatment and every trading cost. Lock that protocol before inspecting a performance chart.
Then preserve time order. Build the rule on one period and judge it on genuinely unseen periods; do not allow later restatements or revised consensus data into the earlier decision set. Out-of-sample testing explains why a held-out period matters, while survivorship bias shows why the missing names matter too.
Next, report the result as a range of predefined robustness checks, not one winning threshold. Change the eligible universe, delay, holding period and cost assumptions within a documented plan. If a result only survives one precise combination, that fragility is a result.
realbacktesting is a trading-software studio for cTrader built around verifiable research rather than a neat curve. The same standard applies here: audit the data, costs and order path before deciding what an event-driven backtest means.
Frequently asked
Is PEAD the same as an earnings surprise?
No. The earnings surprise is an input whose definition must be fixed. PEAD is the research claim that a fully defined rule based on that input may show related subsequent returns (Bernard and Thomas, Jegadeesh and Livnat).
Does PEAD prove that markets ignore earnings?
No. The research literature contains competing explanations, including delayed response and risk-related interpretations. An observed return pattern alone does not settle that dispute (Bernard and Thomas, Brown and Pope).
Can a price chart test PEAD?
No. A credible test needs a pre-defined surprise input, point-in-time availability, an executable entry and exit, realistic costs and an unseen evaluation period. A chart picked after an announcement cannot supply those controls.
Takeaway
PEAD is useful because it forces a hard question: not whether earnings mattered after the fact, but whether a rule using only timely, tradable information survives the details that charts hide.