Seasonality in stock market returns

Research by Steve Heston

Seasonality in the economy is an old idea. It is expected that retail sales go up in November for Christmas, or that snow tires sell better in October than in May.

Seasonality in the stock market has also been explored before—historically, the stock market has gone up more in January than in other months, a phenomenon known as the January effect. But recent research by Steve Heston, associate professor of finance, is the first to show evidence of seasonality in one stock relative to other stocks. In a forthcoming paper he models a permanent seasonal effect of expected returns that is tied to the month of the year.

Heston and co-author Ronnie Sadka, University of Chicago, set out to detect seasonal variation in the cross-section of expected stock returns. The study uses almost forty years of monthly returns—from 1963 to 2002—for NYSE/AMEX-listed firms. Rather than being constant over an entire year or over a period of several months, the model allows expected returns to vary separately across different calendar months—each January, each February, each March, and so on. Applying this methodology to separate calendar months produced results that were both surprising and dramatic.

Heston formed winner-loser decile spreads based on the average monthly return of stocks over various historical periods and measured the returns over subsequent months. Then, to exploit seasonal variation, he sorted stocks based on their returns in each individual calendar months—historical Januaries, historical Februaries, and so on.

Heston found that stocks with high historical returns in a particular calendar month tend to have high future returns in that same calendar month, earning high expected returns every February or July or October, for example. These strategies produced statistically significant positive returns for up to 20 years. The results held true across industries, across different sizes of companies, and independent of earnings announcements. The seasonal effect is positive in all twelve calendar months, but it is particularly high in January and above average in October and December.

The results were so unexpected that when Heston first saw them he was convinced they were the result of a programming error. He called Sadka to ask him to write the code from scratch and re-run the data. “People have been looking for predictability in the stock market for so long, and finding, if any evidence, very weak evidence,” says Heston. “And suddenly we were finding that we could predict returns very easily using almost astrological methods.”

The consideration of seasonality has important implications for asset pricing models. Heston and Sadka find large differences across stock return in particular calendar months. But measuring returns over the whole year overlooks these large differences in expected return.

The study has implications for investment strategies as well. Previous studies have shown that stocks with high historical returns tend to have high expected returns. So an investment strategy that buys stocks with high historical returns and sells stocks with low historical returns should earn high expected returns in future months. Now it may be possible to design an investment strategy that captures the benefits of seasonal variation. The magnitude of the decile spread strategies described in the paper exceeds 60 basis points per month. These short-lived fluctuations in monthly expected return may not be an effective in forming long-term investment strategy, but they could be worthwhile for an active portfolio in certain situations.

“Imagine you’re going to sell stocks for your retirement, or to fund your child’s college education,” says Heston. “Some of those stocks went up a lot last October. You won’t want to sell those in September, because historically they will do well in October. Rather, you’d sell stocks that historically do poorly in October. If you’re going to pay a brokerage fee, you’d rather pay a brokerage fee to sell stocks that are likely to go down in value next month.”

Heston is working on a study that examines this effect in international markets in Japan, the United Kingdom, several European countries, and Canada. “Seasonality in the Cross-Section of Expected Stock Returns” is forthcoming from the Journal of Financial Economics. For more information about this research, contactsheston@rhsmith.umd.edu.