Is Post-Earnings Announcement Drift a Thing? Again?

Considered dead, the phenomenon resurfaces in two studies — which are critiqued in a third paper

Wall Street once feasted on free lunches during earnings season. 

When a company announces earnings that beat (or miss) expectations, the price of its stock should jump (or fall) immediately to settle at its new fair value. At least that’s how efficient markets are supposed to work. But for decades, finance researchers documented that prices kept drifting, often for weeks or even months, before the news was fully priced in. It was as if investors were slow to process the news, creating an easy opportunity, for those who fully understood the results, to profit.

This pattern became known as post-earnings announcement drift, or PEAD. But in 2022, University of Toronto’s Charles Martineau reported that drift began disappearing from nonmicrocap stocks in 2001 and completely disappeared by 2006. Martineau brashly titled his paper “Rest in Peace Post-Earnings Announcement Drift.”

Stocks had started trading in decimals rather than fractions and this made electronic arbitrage more precise and profitable. Faster price discovery through arbitrage was further fueled by new regulation in 2005 that accelerated high frequency trading. For most stocks, prices began fully adjusting on earnings announcement day, marking the death of PEAD.

Yet, some market anomalies are like zombies; they refuse to stay buried. In 2025, two papers accepted for publication contradict Martineau’s 2022 finding, claiming PEAD is alive and well.

In a working paper, UCLA Anderson’s Avanidhar Subrahmanyam sets out to reconcile this contradiction. He argues that the new papers’ disagreement with Martineau’s findings largely reflect research design choices. Especially important is how PEAD is measured and whether microcap stocks, those with a low market value (in the bottom 20th percentile of NYSE market capitalization), are included in the sample.

Catch My Drift?

There are a lot of microcap companies, but they only represent about 3% of the overall stock market’s value. Subrahmanyam’s analysis illustrates that the debate turns on whether stocks forming such a small portion of market value — and that are generally illiquid and difficult to trade — should end up driving statistical tests of how “the market” behaves.

One of the two studies is by Dickerson, Julliard and Mueller and analyzes about 18 quadrillion factor model combinations for jointly pricing corporate bond and stock returns. It concludes that earnings drift remains a major market factor.

Subrahmanyam states that the study used an off-the-shelf earnings factor, essentially a trading strategy that buys stocks with positive earnings surprises and sells those with negative surprises, based on abnormal returns around earnings announcement days. (Past research finds that such an earnings factor is computed in a nonstandard way since it is based on abnormal stock price returns around earnings rather than accounting earnings surprises. In other words, it uses price movement instead of accounting numbers.)

Nonetheless, Subrahmanyam replicated the study using U.S. stock data from February 2001 to December 2024 and recreated the “earnings drift factor.” The critical step was running this strategy twice: once including all stocks, and once filtering out the microcap stocks — a step not included in the Dickerson, et al. paper, despite previous research asserting that drift appeared to only occur in microcap stocks.

Subrahmanyam’s results are telling. A t-statistic was used for measurement, which indicates whether a pattern is real or just random noise. Values of 2 or higher suggest a pattern is real, while those below 2 are likely to be just chance. When Subrahmanyam included all stocks, the drift showed a t-statistic of 2.18 — just barely clearing the threshold. But when he excluded microcaps, the t-statistic dropped to 1.43, well below meaningful significance. This suggests the inclusion of the microcaps was the reason PEAD appeared.

The other paper, by Hirshleifer, Peng and Wang, investigates the impact of social networks on the stock market. It finds that drift exists in the market with extremely high statistical confidence (a t-statistic around 14), but that greater social connectedness weakens its effect. Once again, the study did not filter out microcap stocks.

Subrahmanyam, using the results the authors presented in the paper, suggests the data within their tables indicate that, when proper controls, like firm size, are included, the results are not robust. Their social connection effect appears to just be a market cap effect, Subrahmanyam asserts.

Subrahmanyam urges researchers to prioritize sensible methods, like filtering out microcaps, over statistical complexity, to ensure anomalies are reliable.