Momentum by day is not the same as momentum by night

The more you look at the momentum effect, the more interesting it becomes. It is easy to dismiss momentum as simply investors following past winners and herding them into the same kind of stocks. And investor herding is definitely a big part of the story, but by looking at the momentum effect in more detail, you can find out a lot about financial markets and human nature.

I have written before about the very different behaviour of markets during the day when people trade with each other, and during the night when markets are closed but news arrives about stocks or the economy. Because there is no human interaction during the night, share price reactions to news overnight are entirely rational. Only during the daily trading when humans deal with other humans do we see irrational market effects emerge.

Also, more recently, I have written about an experiment that tried to identify the main reason why momentum effects exist. That analysis concluded that it was limited attention by investors that leads to momentum effects. Not all the news is incorporated into the share price of a company immediately because investors only have so much time on their hands to digest the constant stream of news. This limited attention hypothesis also explains why the momentum effect is stronger with smaller stocks and stocks of companies that are less popular with investors.

But when people are paying attention to the news, they not only incorporate that news into the share price, but they typically overreact to that news. This is why there is the dreaded short-term reversal effect. If you simply buy the stocks with the best performance over the previous 12 months, you will be quite disappointed, because those stocks will be dominated by stocks that have done really well last month. But because investors overreact to news, stocks that have done really well in recent history tend to decline in price because that overreaction is eventually corrected by investors, pushing share prices lower. That is why typically, an up month in markets is followed by a down month and vice versa. And that is why it is a bad idea to check share prices or your portfolio too often. You are just starting to see the short-term up and down of the market instead of the overall trend.

Momentum investors thus typically calculate the return of a stock over the last 12 months minus the return of the last month, in order to reduce the negative impact from short-term reversals.

But wait a minute. If share price overreaction comes from investors trying to incorporate news about a company, wouldn’t it be possible to eliminate the short-term reversal effect by looking at overnight returns only and excluding daytime returns?

Yashar Barardehi and his colleagues performed an interesting experiment. They took all US-listed stocks from 1963 to 2019 and calculated the returns of these stocks during the day (when markets were open) and overnight, as well as the usual 24-hour full-day return. Then they formed momentum portfolios and looked at their performance in the subsequent weeks and months.

The chart below shows the results but requires a bit of explanation. The bars show the alpha of momentum strategies vs a classic 3-factor model including systematic market returns, value effects, and size effects. The horizontal axis shows how many months of past returns have been used to form the momentum portfolios. 1 month means that one buys the 10% stocks with the highest return last month and shorts the 10% stocks with the lowest return last month. As you can see, no matter how you form the momentum portfolios, they all show negative returns in the following month, but portfolios formed on overnight price changes alone have a much, much smaller loss than portfolios formed on 24-hour returns or portfolios formed on intraday returns.

Then you go on to formation periods of 7 and 12 months where portfolios are formed based on the previous 7 and 12 months’ return minus the last month (this is the conventional approach to momentum investing). You see how portfolios formed on 24-hour returns or intraday returns suddenly outperform the market. This switch from underperformance to outperformance shows that these portfolios show a short-term reversal. The portfolios formed on overnight returns alone show no such short-term reversal. Because markets overnight are more rational and not driven by humans, the all too human effect of overreaction is not present and short-term reversal disappears.

Now, if you go even further out and form your momentum portfolio based on past 3- or 5-year returns, you see that portfolios formed on 24-hour returns and on overnight returns show underperformance again. That is the long-term reversal effect where companies that do well for a long time, tend to revert to the mean. This long-term mean reversion is visible in both 24-hour momentum and in overnight momentum, but not in intraday momentum portfolios where the overreaction of human traders creates a persistent push higher for share prices.

So once again, we are left to conclude that if you take humans out of markets, they would be more rational and more efficient, and the momentum effect would largely disappear. This is the world of economics textbooks, modern portfolio theory, and traditional finance. But unfortunately, I don’t think we are going to take humans out of markets anytime soon. And that means that the momentum effect and short-term reversals are not going to disappear either.

Momentum effects depending on different styles of momentum

Source: Barardehi et al. (2022)