Post-Earnings Announcement Drift: The Market's Slow Reaction to News

A Discovery That Challenged Market Efficiency

In 1968, Ray Ball and Philip Brown published what would become one of the most cited papers in accounting research. Studying annual earnings announcements for NYSE-listed firms between 1957 and 1965, they noticed something that the prevailing theory said should not exist: stock prices continued to move in the direction of earnings news for months after the information became public. Positive earnings surprises preceded continued gains; negative surprises preceded continued losses. The pattern was unmistakable, and it pointed to a fundamental delay in how markets processed new information.

Two decades later, Bernard and Thomas (1989) transformed this observation into a rigorous empirical finding that the field could no longer dismiss. Their paper, "Post-Earnings-Announcement Drift: Delayed Price Response or Risk Premium?", used quarterly earnings data from 1974 to 1986 to demonstrate that the drift was systematic, predictable, and inconsistent with rational risk-based explanations. It was, as they argued, a genuine failure of price efficiency.

Measuring the Drift

Bernard and Thomas sorted stocks into deciles based on Standardized Unexpected Earnings (SUE), calculated as the difference between actual and expected earnings scaled by its historical standard deviation. The results were striking: stocks in the top SUE decile outperformed those in the bottom decile by roughly 4.2% over the 60 trading days following an announcement. Approximately half of this excess return materialized after the first ten days, well past the window when market anomalies could plausibly be attributed to announcement-day microstructure effects.

The drift pattern displayed a distinctive seasonal structure. Bernard and Thomas found that a disproportionate share of the predictable return occurred around the subsequent quarter's earnings announcement, consistent with a model where investors partially anchor to prior-quarter earnings. When the next quarter's results confirmed the earlier surprise, the market adjusted further, as though rediscovering information it had already received.

These cumulative abnormal returns were measured relative to size-adjusted benchmarks, ruling out the possibility that the drift merely reflected compensation for bearing momentum-related systematic risk.

Why Does the Drift Persist?

The persistence of PEAD across decades has generated two broad classes of explanation, and the evidence increasingly favors a combination of both.

The behavioral account centers on investor underreaction. Hirshleifer, Lim, and Teoh (2009) demonstrated that when investors face many simultaneous earnings announcements, limited attention amplifies the drift. Stocks reporting on high-volume announcement days exhibit larger and more prolonged PEAD than those reporting in isolation. DellaVigna and Pollet (2009) found a parallel result for Friday announcements, when investor engagement is typically lower. These findings connect PEAD directly to the broader literature on behavioral biases in investing, where cognitive limitations produce systematic pricing errors.

The friction-based account focuses on implementation costs. Ng, Rusticus, and Verdi (2008) showed that transaction costs absorb a large fraction of PEAD profits, particularly among smaller firms where the drift is most pronounced. Bid-ask spreads, market impact, and short-selling constraints create barriers that prevent arbitrageurs from trading away the mispricing. This explanation does not claim that the market is efficient in a theoretical sense; rather, it argues that the cost of exploiting the inefficiency exceeds the return for most participants.

The two accounts are complementary rather than competing. Behavioral underreaction generates the mispricing; trading frictions allow it to survive.

PEAD and Momentum: Distinct but Related

Post-earnings drift shares surface characteristics with the broader momentum factor, as both involve the continuation of prior price moves. However, the mechanisms differ in important ways. Momentum strategies select stocks based on past returns over formation periods of three to twelve months, capturing a broad range of information-driven and sentiment-driven price trends. PEAD is anchored specifically to a discrete information event, the earnings announcement, and operates over a shorter window of 30 to 90 days.

Livnat and Mendenhall (2006) showed that using analyst forecast errors rather than time-series models to calculate SUE produces a stronger drift signal, suggesting that the market's failure to fully incorporate analyst consensus is a key driver. This finding distinguishes PEAD from generic momentum: the predictability arises from a specific, identifiable information gap rather than from diffuse price continuation.

Empirically, momentum and PEAD exhibit low correlation. Portfolios constructed on earnings surprises retain significant alpha after controlling for Carhart four-factor loadings, confirming that PEAD represents a source of return that is largely independent of the standard momentum premium.

What Has Changed Since 1989

The landscape for exploiting PEAD has shifted considerably since Bernard and Thomas published their results. Algorithmic trading and quantitative hedge funds now react to earnings announcements within milliseconds, compressing the initial price adjustment window. Research by Chordia, Subrahmanyam, and Tong (2014) documented that the magnitude of PEAD declined over the 2000s as market efficiency improved for large and mid-cap stocks.

Yet the anomaly has not vanished. Among smaller, less-followed firms, where analyst coverage is sparse and institutional ownership is thin, the drift remains economically meaningful. The pattern has also been replicated in non-US markets, including Japan, the United Kingdom, and several emerging economies, suggesting that the underlying behavioral mechanisms are not culturally specific.

For quantitative researchers, PEAD occupies a special position: it is simultaneously one of the oldest known anomalies and one of the most enduring. Its survival across nearly six decades of scrutiny, replication crises, and technological transformation speaks to the depth of the cognitive limitations that produce it.