Don't bet on holiday effect to boost market returns

First, he debunked the "weekend effect," a decades-long belief among stock traders that market returns are lower on Mondays.

Then he went after the “January effect" (a seasonal increase) and the "turn-of-the-month effect" (higher stock prices at the beginning of the month).

Now Geoffrey Smith, clinical associate professor of finance, is publishing a paper in the Journal of Wealth Management that proves there’s no such thing as the “holiday effect,” which was thought to produce higher returns on the day before Christmas, New Year’s, and the other 10 stock market holidays.

Smith laughs at the notion that he’s a professional grinch determined to take all the fun out of trading. “The puzzle is, why do these things exist? All these effects are not supposed to be there in the first place — they’re anomalies. They’re either due to the market not understanding something or to statistical noise. We’re applying a constancy test to find out,” he explains.

Comprehensive methodology

To determine whether the holiday effect is real enough to be tradeable, Smith and his colleague Russell Robins, a business professor at Tulane University, started afresh with their research, using a very long time range — from 1926 to 2017 — and data from three different sources.

The first source was the Center for Research in Security Prices (CRSP), which collects stock price information from two composite indexes, and NYSE/Amex and NASDAQ, both of which weight returns according to the market capitalization of the stocks they contain. These data serve as a proxy for the stock market as a whole, giving large-cap issues the greater representation.

The second data source came from the comprehensive sets of portfolio information assembled by Dartmouth Finance Professor Kenneth French, who divides stock portfolios into five quintiles according to the market cap size of their stocks. Robins and Smith used data only from the largest-cap and smallest-cap tranches.

Finally, the researchers created another set of data by subtracting French’s highest-quintile portfolio returns from those of the lowest quintile. “That has the effect of creating a portfolio in which you’ve invested zero dollars,” Smith says. “If you invest zero dollars, you should get zero return. If you consistently get more than that, it would indicate there’s something to the holiday effect.”

Another trading myth bites the dust

Robins and Smith found that there had indeed been a holiday effect — a runup of up to 14 times the normal returns on the day before a trading holiday — just as previous researchers had discovered. But their analysis revealed for the first time that the effect is no longer a market force. It disappeared after 1978 from both the overall-market CRSP data and the large-cap portfolio data. In the other two data sets, which contained a larger proportion of small-cap stocks, there was enough of an effect to be statistically significant during some periods, but it wasn’t consistent enough to place bets on.

“The data suggest that the holiday effect is not tradeable because it’s not constant or stable — it bounces around a lot,” Smith says. In addition, the study reveals a slight decline in the holiday effect over time, suggesting that as traders caught onto it, it stopped working.

“If a lot of people bought to trade the pre-holidays, there would be a runup in price, which would eliminate the extra profit and eventually have the effect of killing off the holiday effect,” Smith says. And that’s pretty much what happened.

The fact that the effect has persisted longer in portfolios with a greater proportion of small-cap stocks also makes sense.

“Often these unusual effects we call anomalies show up based on firm size. Small firms, in particular, are the ones that behave strangely — that’s why it’s interesting to test different sizes,” Smith says. Small issues are more difficult to trade, making for less trading activity. That might explain why the holiday effect hasn’t yet disappeared from the two smaller-portfolio data sets. But it likely will over time.

Potent algorithm

To analyze the data for this study and their previous work, Robins and Smith applied an algorithm developed by econometricians Jushan Bai and Pierre Peron in the late 1990s. In the early 2000s, the algorithm was used to study macroeconomic data such as GDP output and production. “We decided to use it to test the constancy of stock returns,” says Smith.

The researchers may use the algorithm to examine other purported market effects, though Smith isn’t saying which ones. “We want to extend it as much as we can and keep looking for unusual effects we can test,” he says. Though Smith doesn’t know of any current applications. He says the algorithm could also be valuable to businesses and governments.

“You could use it to study the change in the number of passengers on an airline route. Or you could study the effect of a policy change — for example, to see whether a change in the minimum wage causes unemployment. It works for any data you can line up over time,” Smith says.

Score one for efficient markets

Though it is limited in scope, Robins and Smith’s research provides additional support for the efficient markets hypothesis, a key tenet of modern financial theory. According to the theory, stock prices always trade at their fair value because they reflect all the information available about them at the time. As a result, it’s impossible for trading techniques based on technical or fundamental indicators (such as the holiday effect) to outperform the market, except possibly for a brief window of time.

Though the efficient markets theory is widely accepted, some traders and economists dispute it. Based on his finance background as well as his own research, Smith is mostly a believer.

“I think markets are mostly efficient. If you hunt far enough, you can probably find exceptions. But if you scale your trading, they probably won’t last,” he warns.

By Teresa Meek