Tapes Tell All: The Hidden Messages in Earnings Calls

For public corporations, the quarterly meetings between their executives and analysts, investors and the media to discuss the company’s financial results and future prospects are red-letter dates. Short-term fortunes are at stake. To the uninitiated, these earnings calls—more like live-streamed webcasts these days—are a mix of theatrical performance and a round of Texas Hold’em.

After all, executives are carefully coached to present scripted remarks crafted by investor relations professionals and designed to pump up good results or play down missed expectations. Little is left to chance. But like well-schooled poker players, analysts and investors search for signs—the bead of sweat, the forced laugh—that reveal just how good the executive's hand really is.

And in the middle sits Blake Steenhoven.

Steenhoven, assistant professor of accounting at Smith School of Business, is part of a new wave of researchers who employ unconventional techniques to study the earnings call phenomenon. Drawing on a huge body of recorded meetings, they use textual and acoustic analysis software and machine learning to identify executives’ hidden vocal cues that unintentionally reveal their physiological or emotional states. They then examine potential connections between the vocal cues and a firm’s financial outcomes.

“These raw acoustic measurements have been used in a number of applications, like understanding the emotional states of speakers, identifying deception and even in medical diagnoses,” says Steenhoven. “Research in accounting has started analyzing vocal cues of managers and has found that they can explain stock returns and predict firm performance and managers’ career outcomes.”

Faster voice, lower voice

Early research in the field involved a person listening to recordings and assigning subjective scores for a speaker’s speech patterns. Now, low-cost software allows for automated extraction of vocal cues from audio recordings, providing objective measurements of the speaker’s vocal pitch, how the pitch varies, volume and other sound tidbits.

Steenhoven uses these techniques to analyze sound waves from recordings of subjects in a lab and participants in actual earnings calls. In one of his controlled experiments, 108 MBA students acted as CEOs of a fictitious craft brewing company. According to the scenario, the company had unexpectedly low earnings for the quarter but had planned to raise its full-year guidance. In mock calls with investors, the CEO-actors read scripts either trying to persuade prospective investors to make an investment in the company or current investors not to sell their position.

When he analyzed the acoustic data, Steenhoven found that the “executives” spoke faster and increased their pitch, pitch variation, volume and volume variation when persuading prospective investors to invest in the company. They used lower pitch and pitch variation when persuading current investors not to sell their position. Curiously, female executives—but not male—lowered their volume when trying to reassure current investors.

The message from this study is clear: Executives who communicate results in an earnings call—or other high-stakes situations—should understand what their voices can reveal. “They may consider varying their vocal cues based on the patterns I identify when trying to persuade others to act or to not act,” Steenhoven says. “Given evidence that analysts and investors scrutinize these cues, these findings may also help them understand how these signals will be interpreted.”

Can coaching work?

Understanding vocal cues is one thing; actually changing them successfully is quite another. Coaching executives on vocal delivery can have unintended consequences, according to a study Steenhoven conducted with Kristina Rennekamp (Cornell University) and Brian White (University of Texas at Austin). Steenhoven got the idea for the study from conversations he had with investor relations officers. They told him that executives are often trained to moderate non-verbal behaviours in presentations to investors to prevent the executives from unintentionally conveying emotion.

To see what would happen if people were coached to suppress emotion in their voice, the research team recruited 94 MBA students to act as CEOs. Some were given voice coaching; others not. Then they all recorded mock earnings calls with investors and analysts. The researchers arranged for almost 1,200 people to listen to the recordings and rate how emotional and natural the executives sounded.

Vocal coaching proved effective: the disclosures by coached speakers were judged as sounding less emotional than uncoached disclosures. Acoustic analyses showed that vocal coaching caused speakers to lower their pitch, pitch variation, volume, volume variation and rate of speech. Pitch variation, in particular, was associated with how listeners perceived a speakers’ emotion. 

Unfortunately, the coached disclosures were also perceived as sounding less natural, and that had serious consequences. With further research, Steenhoven discovered that, because the coached disclosures sounded less natural, people perceived the executives as less competent and trustworthy—two important components of their credibility. “Vocal coaching could potentially do more harm than good if it increases perceptions that a speaker sounds unnatural or inauthentic,” he says.

A vocal cue arms race

Steenhoven is at an early stage of research. He feels most of his work to date sets the groundwork for what’s to come. But the results illustrate the importance of paying attention to vocal cues when refining your messages. Just like with written communication, Steenhoven says, vocal cues should be tailored to the specific message you’re trying to communicate, the audience you're communicating with and your goals. 

“A lot of our understanding around vocal cues assumes that speakers are not in control of what they’re doing with their voice,” says Steenhoven, who knows all about the art of persuasion from his years on the debate team in high school and college. “But that’s not quite true. Among other things, people actively manipulate their voices to clarify content, to signal their attitudes and stances, to project confidence, sincerity or dominance and to influence others.”

The biggest corporations and investment houses likely already know this and use the same techniques as Steenhoven. He says asset managers increasingly use text mining and sentiment analysis software, for example, to guide their investment decisions—analyzing and acting on disclosures in real time. Not surprisingly, public firms now use similar software to understand how best to present their quarterly and annual reports and other disclosures. It’s an arms race based on artificial intelligence.

“Investors are constantly looking for an edge and becoming more sophisticated,” Steenhoven says. “There are hedge funds that fly drones over parking lots to help estimate sales numbers. And they're now looking at managers’ verbal and non-verbal behaviours in these settings. That’s a great concern for firms. They want to make sure they’re not disclosing anything they shouldn’t.”