Can Investors Create Profitable Trading Strategies Based on Academic Findings?

“There have been a ton of studies around earnings announcements, and the evidence is pretty compelling,” says Matthew Lyle, an associate professor of accounting information and management at Kellogg. However, “these studies aren’t necessarily designed to mimic how traders actually act or what information they truly incorporate into their decisions.”

So Lyle and coauthor Teri Lombardi Yohn, at Emory University, set out to see whether investors could use the research findings around earnings announcements to make profitable trades. They worked hard to keep the investor’s perspective in mind. As they write in their paper, they wanted to create “a method that more closely resembles implementable trading strategies” than past research has done.

They did this by creating portfolios populated by historical trading data in order to mimic a plausible investor portfolio and then applying a “train-test approach” (common in machine learning) to simulate the various ways a real investor might attempt to trade on earnings-announcement news. This allowed the researchers to discern how an investor might best gain returns from the stock-movement patterns that tend to either precede or follow earnings announcements.

They looked in particular at three general findings from previous academic research to see how useful each of them might be to a typical investor.

Lyle and Yohn’s analysis found that some—but not all—panned out as a useful trading strategy. But more broadly, their findings supported the idea that actual investors could profit from understanding these stock-movement patterns.

“The paper addresses, I believe for the first time, the question, ‘Is something actually economically there when it comes to the research findings about earnings announcements?’” Lyle says. “And the answer is yes.”

Simulating Real Investors

Lyle says the most challenging part of the study was designing it with an investor’s perspective at the forefront—and accounting for what they are, and are not, likely to know in the moment as they make trading decisions.

For example, in creating their portfolios and designing their train-test approach, the researchers elected to assume that the imagined investor was aware of the three key patterns around earnings announcements that have been documented in past research. These findings were:

1) After a company reports surprising earnings, that company’s stock tends to drift for a few days or even weeks—up for positive surprises, down for negative ones.

2) A company’s stock price generally gets a boost in the lead-up to earnings announcements, perhaps because more investors are thinking about it.

3) Earnings tend to be more positively surprising when an earnings announcement is moved up from its expected date—and more negatively surprising when the announcement is delayed, suggesting that firms appear more eager to share good news than bad.

Lyle and Yohn’s imagined investor might wonder, could an investment strategy trade off of all three patterns at once? Should one or two take primacy over the others? Lyle says previous research hasn’t addressed these questions.

Often studies assume that investors have no knowledge of the earnings-announcement patterns found in previous research. This seemed unlikely to Lyle and Yohn. At the same time, Lyle argues, many of the common approaches used in these studies also assume that an investor has too much knowledge.

For example, many of these studies rely on what is called an “event time” approach, which groups together events that happened within a given time period—and treats those events as though they happened at the same time.

Event time is clever for helping researchers make sense of stock-market movements, Lyle says, “but from an investor’s perspective, as you go into today, you only have the information based on today, and based on the past, to actually trade on. You don’t get to see tomorrow.”

Lyle and Yohn sought to address the limitations of this “look-ahead bias” by instead relying only on what an investor would have actually known on any given day.

The researchers also had their simulated portfolios hold at least 30 stocks. This was intended to mimic the real-life diversification that many investors are mindful of, as opposed to less diversified and less realistic portfolios that academics often use.

In their portfolios, the researchers’ train-test approach simulated various ways an investor might attempt to trade on earnings-announcement news. In some portfolios, for example, they allowed for a long lag after earnings news hit before trading a stock—since some past research has suggested that a stock’s post-earnings announcement drift can last several weeks. In others, they moved more quickly. From there, the researchers were able to zero in on the most profitable strategies.

A Dominant Pattern

The researchers’ findings reveal some interesting insights for implementing trading strategies around earnings announcements.

First, they found that a stock’s post-earnings-announcement drift pattern can generally be profitably traded on for the first one to four days. This is a much smaller window for trading on earnings-announcement updates than suggested in previous research.

What’s more, it appears that investors are best off when they only pay attention to the post-earnings-announcement drift patterns—rather than also trying to incorporate the patterns in stocks’ movement just before these announcements or in the event of a rescheduling. Lyle acknowledges that this surprised him; he had assumed that the most profitable strategies would be the ones that managed to incorporate more than one of these documented patterns.

In the end, he says, he believes the non-post-announcement drift patterns do not prove incrementally useful because it’s essentially a trade-off: when an investor chooses to invest, that investor must give up something, and since the patterns overlap across different firms for much of the time, it behooves investors to choose the one that tends to produce the most stock movement, which is the post-earnings-announcement drift pattern.

The underlying reasons why this pattern was the dominant one were not addressed by this paper—and Lyle hopes that the study helps to provide a new way of examining the reasons in future research.

A Message to Future Researchers

Lyle believes his findings have important messages to convey.

The first is for investors: even after designing a methodology that integrated many of the constraints that real investors deal with, Lyle and Yohn still found potentially profitable strategies in patterns surrounding earnings announcements. To Lyle, this suggests that “some of the results in the academic literature do appear to be real and remarkably robust.”

Lyle says the other central message of his paper is for academics: it may be time to update some of the common methodologies used in this realm of research that are preventing results from being generalizable to real investors.

“I think if we want to say something about how much an investor is leaving on the table, it’s important for us to provide some guidance on how they would actually trade based on the patterns identified in academic research.”