The idea that a company’s business is seasonal is nothing new. What is new is trying to understand why the market is not properly pricing for this seasonality. The authors suggest that investors and analysts overweight the more recent lower-earnings quarter, leading to a more pessimistic forecast, and subsequently underweight the positive seasonality quarter. For those companies that exhibit seasonality in their earnings, the median analyst correctly forecasts 93% of this seasonal shift in earnings, missing only 7%. Although this finding shows that analysts are taking the seasonal nature of the earnings into account, they are not fully adjusting their forecasts and properly accounting for earnings seasonality.

Understanding the seasonality of a company’s earnings is fairly straightforward, but calculating this seasonal adjustment is challenging because the models are extremely sensitive to inputs. Owing to the seemingly straightforward nature of this seasonality, investors and analysts may not fully appreciate its complexity and may thus believe that it is easy to solve. By holding this simplistic view, which discounts the true difficulty of accurately forecasting seasonality, market participants are likely to be susceptible to behavioral biases—in particular, what is called the recency effect.

The recency effect is the tendency of people to be most influenced by what they have last heard or seen. Not surprisingly, investors suffering from the recency effect will be more likely to overweight recent lower earnings compared with the higher seasonality earnings from the year-ago period. Related to the recency effect—and perhaps an additional factor contributing to the mispricing—is the availability heuristic, which operates on the notion that something that can be recalled must be important.

Consistent with the predictions of the recency effect and the availability heuristic, when recent earnings are lower, the seasonality effect is larger. The authors find monthly excess returns of 65 bps in an equal-weighted portfolio and 76 bps in a value-weighted portfolio, both significant at the 1% level.

In contrast, when companies report higher recent earnings, the returns are lower: The authors find monthly excess returns of 28 bps in an equal-weighted portfolio and 6 bps in a value-weighted portfolio. Demonstrating that the seasonality effect is larger when companies have had lower recent earnings, the authors suggest that recent low earnings make investors unduly pessimistic about positive seasonality quarters. When a company has record earnings in the last 12 months, the seasonality effect is not present. These results are consistent with investors’ focusing excessively on recent events, implying that market participants are not fully appreciating the longer-term patterns in seasonal earnings.

The authors provide additional evidence that similar repeating events also exhibit abnormal stock returns for companies that are expected to announce earnings and dividends, dividend increases, and various types of stock splits. To the degree that investors make the same miscalculation, markets can be expected to misprice assets and thus abnormal returns may be anticipated.

The authors use earnings data from Compustat, stock price data from CRSP (excluding all companies with a stock price under $5 a share), analyst forecasts from I/B/E/S, and data from Kenneth French’s website.

To measure the variability of earnings seasonality, the authors use five years of quarterly data to calculate how often earnings are higher in a given quarter and by how much. By looking back quarterly for five years, they can determine what they refer to as the earnrank. The earnrank is simple to construct and is both replicable and transparent. It has three important benefits: The earnrank is not affected by negative earnings, it is not affected by large variability in earnings, and it is not sensitive to trends in overall earnings growth because it is constructed using several years’ worth of quarterly earnings.

Thinking that other variables may be contributing to the abnormal returns, the authors control for and examine several factors, including the dividend month premium, which is the notion that companies have abnormally high returns in months when they are predicted to pay a dividend. The authors also examine the effects of other variables that could affect returns—log market capitalization, log book to market, momentum, and last month’s return—finding these variables to be nonexplanatory. They analyze a number of risk-based explanations for abnormal returns that fail to explain, including firm-specific information, increased volume, and idiosyncratic volatility.

Adding to the growing body of evidence of the powerful nature of behavioral biases and heuristics, the authors’ results are consistent with the behavior of people who have vast amounts of information but do not process that information in the most rational way.

Although data are ubiquitous, our ability to act rationally on those data is another matter. Daniel Kahneman has stated, “We think, each of us, that we’re much more rational than we are.”