peadsuepost-earnings-driftbehavioral-financeneglected-stocks

Post-Earnings Announcement Drift (PEAD, 盈余公告后漂移)

The empirical regularity that stocks continue to drift in the direction of an earnings surprise for weeks to months after the announcement. Discovered by Ball & Brown (1968) and formalised by Bernard & Thomas (1989). The drift is the canonical release channel through which informationally slow cross-sectional signals — including Abnormal Investor Visits (AIV) — realise their alpha.

Definition

Let \text{SUE}_{i,t} = standardised unexpected earnings of firm i at earnings announcement t, typically defined as the year-over-year change in quarterly EPS scaled by the standard deviation of unexpected earnings over the prior 8 quarters. Sort firms into deciles by SUE each announcement; the long-short decile-10 minus decile-1 portfolio earns a positive risk-adjusted return that accumulates over 60 trading days post-announcement — the drift.

Headline numbers (US, post-1980 literature)

  • Decile 10 − Decile 1 CAR(0, +60): ~8–10% raw; Sharpe ~0.4–0.6 in the original sample.
  • Abnormal volume: ~2–3× normal volume on the announcement day; abnormal volume does not predict drift, but analyst revision does.
  • Persistence: drift lasts 40–80 trading days, with most of the return earned in the first 20.

Bernard & Thomas (1989) argue ~50% of the drift is mechanical — predictable from prior-quarter SUE (time-series momentum in earnings). The remainder is news impoundation.

Why does PEAD exist?

Three (non-mutually exclusive) mechanisms:

  1. Investor inattention / limited processing capacity — Hong, Lim & Stein (2000) "bad news travels slowly": under-covered firms adjust slower.
  2. Asymmetric trading costs — Johnson & So (2018) Asymmetric trading costs prior to earnings announcements: bid-ask spreads widen more before positive surprises than negative ones (or vice-versa), creating an asymmetric cost-of-immediacy barrier.
  3. Risk-based — ball-and-brown-style risk-factor explanations have largely failed to absorb the full drift in the post-1990 sample.

In the A-share setting, institutional attention is the dominant explanation. So Wang Zhang 2024 Investor Corporate Visits Table 7 Panel B shows that the AIV hedge strategy earns 3.6–4× more return on the next earnings announcement day than on non-announcement days; 2.9% of the next-6-months raw return is earned on a single earnings day. The implication: AIV is identifying pre-announcement underpricing, and PEAD is the channel through which that underpricing corrects.

Cross-references

Mechanism Reference Connection to AIV
Limited attention / underreaction Hong, Lim & Stein (2000) AIV concentrates in neglected firms
Asymmetric trading costs Johnson & So (2018) Co-authored by Eric C So (AIV co-author)
Mechanical time-series SUE momentum Bernard & Thomas (1989) Independent of AIV
News impoundation Engelberg, McLean & Pontiff (2018) Drift concentrated on news days; matches So-Wang-Zhang Table 7

Empirical implementation

Standard PEAD test:

  1. Compute \text{SUE}_{i,t} at each earnings announcement date t.
  2. Sort into deciles; long decile 10, short decile 1.
  3. Hold for 60 trading days; rebalance as new announcements arrive.
  4. Compute market-, size-, B/M-adjusted alphas.

For the Abnormal Investor Visits (AIV) test:

  1. Compute AIV at end of month m from on-site visit data.
  2. Long top AIV quintile, short bottom; hold for 1 month (or up to 10 months per Table 5 results).
  3. Decompose returns into announcement-day vs. non-announcement-day contributions (Table 7 Panel B).
  4. If the announcement-day share is disproportionately high, the signal is consistent with pre-announcement underpricing + PEAD-style correction.

Limitations

  • Crowded: PEAD is one of the most-replicated and most-decayed anomalies. Post-2000 Sharpe is roughly half the original sample's.
  • Crisis-period attenuation: the drift collapses in high-volatility regimes when uncertainty about fundamentals dominates.
  • Earnings quality: PEAD works less well on firms with high accruals, high restatement risk, or low earnings quality.
  • A-share specifics: SZSE / SHSE disclosure rules and T+1 settlement interact with the standard US-style PEAD calendar; a careful A-share replication requires alignment with disclosure date (not announcement date).

与 schema 化反应信号的对照 (2026 AlphaSignals 解读)

把 PEAD 的"surprise 方向 + 修正幅度"拆成结构化反应字段(主线影响 / 结果方向 / 未来 12m EPS 影响)后,单独的 earnings reaction 信号 多头腿税后 Sharpe 仅 0.17;而把同样的信息用于延长已建仓位(position management)而非独立开仓,Earnings Cycle 多头腿 Sharpe 达 1.49 / IR 1.23。

含义:PEAD 在市场吸收反应信息后衰减极快;真正稀缺的是事前判断(见 Earnings Preview Signal)。把 reaction 用作 position sizing 乘数 — 确认则延长、中性则退出、反向则减仓 — 比把它当作独立 alpha 更稳。

来源:summaries/202608-analyst-forecast-event-alpha/202608-analyst-forecast-event-alpha

See also