Post-Earnings Announcement Drift (PEAD)
Buy stocks that beat earnings, hold ~60 days to capture the documented drift.
Inconclusive, not a clean fail — PEAD is a genuine, heavily-documented academic anomaly, but it is not validatable on the free data we had.
- Why it fails
- Free earnings feeds cap history at four quarters per ticker, confining the entire test to one bull-market regime where the apparent drift was largely market beta. A clean read needs paid IBES/Compustat data — this is 'untestable on free data here,' not 'PEAD is fake.'
- When / how it stopped
- Tested 2026-06-25. The free-data ceiling forced a single ~2024–2026 window: PF 2.20 and Sharpe 3.40 on paper, but the placebo passed only at a razor-thin margin and concentration/walk-forward gates failed.
Post-earnings announcement drift is one of the oldest and most robust anomalies in finance: after a company beats earnings, its stock tends to keep drifting upward for weeks. Bernard & Thomas (1989) documented it formally, and it has survived decades of replication. So this entry is different from the others — it is not a clean fail.
We attempted a pre-registered version — buy EPS beats, hold ~60 days — but ran straight into a data wall:
- The free earnings feed is hard-capped at four quarters (~1 year) per ticker. Every tradable event therefore landed in a single ~2024–2026 bull regime — no down market, no regime diversity, no decade of decay to measure.
- On paper it posted PF 2.20, Sharpe 3.40, −6.5% drawdown and nominally cleared 9 of 11 gates. But those passes were artifacts of one tape. The placebo passed only at a razor-thin margin (real PF 2.20 vs placebo p95 of 2.10) — meaning a basket of random same-week reporters captured roughly 96% of the “drift.” In a rising market, anything held 60 days drifts up.
- The two gates that did fail — year concentration (77% of P&L in one year) and walk-forward (IS PF 3.45 → OOS 1.44) — fail precisely because there is only one regime to see.
So the honest verdict is inconclusive: untestable on free data here. PEAD is a genuine anomaly; we simply could not isolate it from market beta without point-in-time, survivorship-free, multi-year data with consensus EPS and guidance — i.e. paid IBES/Compustat. We are not claiming PEAD is fake. We are saying our free-data validation program cannot test it cleanly.
The full pre-registered report, gate scorecard and charts are in our validation write-up.
→ Read our full validation report: /strategy/pead
Sources
- Our validation report — PEAD, inconclusive on free data (placebo razor-thin, single regime)
- Bernard & Thomas (1989), 'Post-Earnings-Announcement Drift', Journal of Accounting Research
Frequently asked
Does post-earnings announcement drift work in 2026?
PEAD is a real, heavily-documented academic anomaly — but we could not validate it on free data, so our verdict is inconclusive, not a fail. Free earnings feeds cap history at four quarters per ticker, which confined our entire test to a single ~2024–2026 bull regime. In that window the apparent drift was largely market beta: a basket of random same-week reporters captured nearly all of it. A clean read needs paid IBES/Compustat data.
Why is PEAD inconclusive instead of retired?
Because the binding problem was the data, not the strategy. The free feed hard-caps earnings history at four quarters, so there was no regime diversity, no down market, and no decade of decay to measure — the experiment was never testable end to end. The numbers it produced (PF 2.20, Sharpe 3.40) all came from one bull tape, the placebo passed only at a razor-thin margin, and two gates failed on single-regime concentration. We are explicit: this is untestable on free data here, not evidence that PEAD is fake.
Not investment advice — your mileage may vary, but the burden of proof is on the person claiming an edge. This entry describes general research and published evidence (or its absence), not a recommendation. See the full disclaimer.