Normally, I write posts that are useful for a broad group of investors, but today I want to focus on a behavioural anomaly in markets that is probably only useful to professional market makers and traders. It is the observation that trading volume in stocks increase on days other stocks report earnings if the stocks in question have in the past reported earnings on the same day.
Earnings reports are relatively random within an earnings season. There are some patterns in the sense that each earnings season in the United States starts with the big banks reporting over several days before we move on to other sectors. But by and large, on each day of an earnings season companies from many different sectors that have nothing to do with each other report their results.
The problem (or simply, the reality) is that we have a pretty flawed memory. I have explored that topic in some detail some time ago, but recently, Constantin Charles from USC has examined trading behaviour of retail investors. What he could show is that if investors traded in one stock, they were more likely to trade in other stocks that were associated with it in their memory. And that had nothing to do with the companies being in the same sector. It was simply a reflection if the investors had bought the shares at more or less the same time in the past.
Now, he doubled up with a study that focuses on trading volumes in stocks that have seemingly nothing to do with each other. Assume you have two companies, A and B, and that historically these two companies have reported earnings on the same day every quarter. In the mind of investors, these two stocks start to occupy the same space. If you have several children and you call one child using the name of her sibling, you know what I mean. Because in your memory your children are occupying the same space, your memory sometimes mixes up the names of them and if you are tired or don’t pay attention, you may call your child by the wrong name.
Now, one of these two companies that historically have reported earnings on the same day publishes some fundamental news, but the other doesn’t. For example, company A may report another set of earnings but company B has scheduled its earnings release for whatever reason a week later. What happens is that trading volume in the shares of company B increase even though there is no fundamental news on company B. The association of company B with company A in the minds of investors triggers excess trading volume.
It is completely bizarre and to be sure, markets correct this “mistake” within two trading days so you can’t make money with it in a trading strategy but the volume is still elevated for a day or two. Charles showed in his study that this effect is not binary in the sense that it only happens when two stocks have historically reported earnings on the same day. He could show that is two companies have historically reported earnings one or two days apart, the excess volume still appeared. On the other hand, if two companies reported as part of a busy day, the association in the minds of investors was less intense than if fewer companies reported on the same day in the past. So, at the height of the earnings season when many companies report each day, the effect becomes smaller, but at the tail ends (start of the earnings season or last few weeks of the earnings season), the effect becomes larger.
As I said, it isn’t helpful if you want to become a better investor but it sure is one of the weirdest behavioural effects I have ever seen.
Excess volume of companies historically reporting within a few days of each other
Source: Charles (2022)
Seems possible that this is the result of investors putting on pairs trades around earnings announcements?
i.e. imagine that traders check which companies are reporting overnight & go long/short on the companies according to their expectation for the announcement. Then they hang on to the position until the next report date for one of the two companies, at which point they consider the pair trade to be completed & unwind both legs, generating "excess" volume on the non-announcer.