Since the 1920s, traders have discussed a strange phenomenon: Stock returns tend to be lower on Mondays.
It’s known as the “weekend effect,” confirmed by various studies but never explained, though traders have put forth different theories: Companies release bad news after markets close on Friday, and it moves stock prices on Monday. Traders are depressed about returning to work after the weekend. Treasury securities are auctioned on Mondays, drawing investors away from stocks.
But none of these explanations seems adequate for causing such a long-ranging effect.
Geoffrey Smith, clinical associate professor of finance, along with Russell Robins of Tulane University, have discovered that there’s a good reason the “weekend effect” can’t be explained: It no longer exists. Despite an ongoing stream of research and papers on the subject, the “weekend effect” actually disappeared in 1975.
Letting the data speak
Smith and Robins describe their work and methods in a new academic paper published in Critical Finance Review titled, “No More Weekend Effect.”
“Researchers have spent a lot of time trying to explain the ‘weekend effect,’” Smith says. “Our belief was, what if it’s not an anomaly? What if it’s just something specific to the periods researchers looked at?”
Computer analysis proved intuition right. Using data from the Center for Research in Security Prices, the researchers examined returns for 1,000 stocks, going all the way back to 1926 and up to 2014.
Previous studies used much narrower time frames. A well-known 1980 study considered the years from 1953 to 1977, and another in 1981 looked at the period of 1962 to 1978.
Using a method called the Quant/Andrews test, Smith and Robins began by recording the average return of stocks in the first 100 Mondays of their data set. Then they took a new average of the first 101 Mondays, comparing it to the previous 100-Monday average. They did the same for the first 102 Mondays, the first 103 Mondays, and down the line for all 4,200 Mondays in the sample, comparing each new average to the original one of the first 100 Mondays.
This method reveals whether the average return changed in a way that was statistically significant and if so when that happened.
Smith and Robins found that for the entire study period of 1926 to 2014, there was a statistically significant “weekend effect” — stock returns were down on Mondays by an average of 12.33 points.
But they also found a significant “break” in the data beginning in 1975. For the period from 1926 through 1974, Monday returns averaged down by 18.1 points. But from 1975 to 2014, they averaged down by only 5 points, a tiny amount that isn’t statistically significant.
“It’s effectively zero,” Smith says. “After 1975, the ‘weekend effect’ went away.”
In fact, the anomaly had already disappeared in the early 1980s when the primary studies affirming its existence were published.
“All these studies that try to explain this weird ‘weekend effect’ are explaining something that disappeared in 1975,” Smith says.
Back in the early ’80s, when the most consequential studies were done, computers were lumbering behemoths. The technique Smith and Robins used didn’t exist. Data going back to 1926 may not have been available, either. Since those studies were done, researchers have tested a number of other time periods, but none as comprehensive Smith and Robins’.
“Our study was agnostic on period. We wanted to let the data speak,” Smith says. “Our statistical break point of 1975 is what the computer found, not something we selected.”
Smith says he wasn’t surprised by his conclusion because the “weekend effect” never made sense to him in the first place.
“If everyone knows Monday’s such a bad day, why don’t people only buy on Monday, when prices are low? That’s why the ‘weekend effect’ was so troublesome.”
An anomaly that has no valid reason to exist should go away on its own over time.
“And it did,” Smith says. “But people are still writing about it and testing it. People still believe the ‘weekend effect’ is true.”
Smith and Robins tested the “weekend effect” because it’s the best-known trading anomaly, but it’s by no means the only one. There’s also a “January effect,” a “September effect,” a “Super Bowl Indicator” (an NFC win means the stock market will be up for the year; an AFC win means you better sell), a “hemline indicator” (the market rises and falls in accordance with fashionable skirt lengths), a “Daylight Savings anomaly” (disrupted sleep patterns after the start of Daylight Savings Time cause markets to fall), and even an “aspirin indicator” (aspirin sales rise when stock market losses literally give people headaches).
And then there’s the “October effect,” which takes into account the Panic of 1907, the October crash that started the Great Depression in 1929, and the Black Monday crash of 1987.
The “October effect” is also known as the “Mark Twain effect,” named after Twain’s character Pudd’nhead Wilson, who said: “October. This is one of the peculiarly dangerous months to speculate in stocks. The others are July, January, September, April, November, May, March, June, December, August, and February.”
Smith’s research suggests Pudd’nhead Wilson was on the right track.
Though they haven’t studied the “October effect,” Smith and Robins have done research for a related paper that they hope to publish this summer. It’s about the “January effect” (a seasonal increase) and the “turn-of-the-year effect” (a pattern of increased trading volume and higher stock prices in the last week of December and the first two weeks of January). Using research techniques similar to those they used for the “weekend effect,” they found that both effects disappear over time.