Does PEAD still exist in large caps?

After an earnings surprise, do large-cap stocks continue to drift in the surprise direction over the next sixty trading days (post-earnings announcement drift), and is the Q5−Q1 long–short spread still statistically detectable?

Universe: current S&P 100 tickers (survivorship-biased by construction). For each name we clean Yahoo Finance earnings dates (US-listed only; up to Yahoo's 100-quarter cap), estimate a market model on SPY, and form CARs for the announcement window [0, +1] and drift window [+2, +60] (truncated before the next print). Surprises are standardised (SUE). Within each calendar quarter we assign SUE quintiles and measure mean drift CAR; the Q5−Q1 spread uses quarter-clustered standard errors.

Live panel numbers below come from GET /earnings/panel (rebuilt weekly).

Per-stock tests on the Earnings page report BMP, Kolari–Pynnönen adjusted BMP, and a simple cross-sectional t on raw CARs. Narrative wording uses |t| < 1.96 as not significant, 1.96–2.5 as borderline significant, and > 2.5 as significant. Surprise/price series are winsorised at the study stage; the pooled panel uses the cleaned event list after those filters. Thin calendar quarters (<5 names) are dropped before quintile assignment.

The pooled quintiles and Q5−Q1 spread are computed under two abnormal-return definitions: the market model (estimated α, β on SPY) and a market-adjusted specification (α = 0, β = 1). Toggle between them above; a material gap between the two would warn that conclusions hinge on how systematic risk is removed.

• Survivorship bias: today's S&P 100 only — names that left the index are absent.
• Yahoo Finance timestamps and EPS figures can be missing, revised, or Street vs GAAP mismatched.
• Consensus definition is whatever Yahoo reports as the estimate; guidance and revenue are ignored.
• Large caps only; PEAD historically stronger in smaller names.
• No transaction costs, short-sale constraints, or capacity — not a tradeable backtest.

In today's S&P 100, there is no reliable post-earnings announcement drift. Stocks with the biggest positive surprises did not go on to outperform those with the biggest negative surprises over the following 60 trading days, and the difference is small and not statistically significant.

The result holds under both market-model and market-adjusted returns. Under market-adjusted returns every quintile drifts upward, which reflects survivorship bias (today's index members are past winners) rather than any earnings effect, because the gap between the top and bottom quintiles stays close to zero.

There is also no sign that the drift has weakened over time: it is small and insignificant both before and after 2015. This fits the view that PEAD is concentrated in smaller, less liquid stocks. In the most heavily traded large caps, the market appears to absorb earnings news within the first day or two.

The main caveat is survivorship bias: the sample only includes companies in the S&P 100 today. Not investment advice.