Beyond Words: AI Shows Facial, Vocal Cues of Central Bankers Can Impact Financial Markets

A picture is worth a thousand words. In the world of central bankers, it can even move global financial markets.

Central bankers like Fed Chairman Jerome Powell and European Central Bank (ECB) President Christine Lagarde parse their comments about monetary policy carefully because the capital markets hang on their very words.

A recent paper written by researchers from the ECB and German academics show that even facial expressions and tone of voice can have an impact on stocks, bonds and inflation.

These findings emphasize an important and yet overlooked aspect of monetary policy: Markets respond not only to what central bankers say but also to how they say it.

For their paper, “The Emotions of Monetary Policy,” the researchers used machine learning models and advanced facial and vocal recognition technologies to analyze emotional cues from Lagarde and her predecessor, former ECB President Mario Draghi.

They looked at 100 ECB post-policy press conferences between 2012 and 2023, which includes the time when central banks around the world cut interest rates drastically during the pandemic and the return to normalcy.

The researchers discovered that inadvertent emotional cues shown by ECB central bankers during press conferences after their monetary policy meetings correspond to movement in stocks, bond yields and how far the inflation rate travels from the ECB’s target of 2%.

Here are some examples of their findings:

• A happy facial expression from Draghi amplifies the impact of a hawkish statement — indicating tighter monetary policy that translates to a period of higher interest rates — and causes lower stock prices and higher bond yields. This ironic pairing led researchers to call the phenomenon “killing with kindness.”
• Happy expressions from both Draghi and Lagarde have translated to lower bond yields and a weaker euro, which means interest rates are coming down. Conversely, expressions of anger or heightened vocal cues heightened market reaction, signaling tighter monetary policies.
• A higher level of vocal intensity from the central banker leads to higher bond yields, falling stock prices and a stronger euro against the dollar.

For finance professionals, this research underscores the importance of incorporating emotional analytics into their policy analysis. Understanding the role of emotional expressions can provide a competitive edge to asset managers, traders and analysts.

Read more: Fed Prioritizes FedNow, Delays Fedwire Funds Service Changes

The researchers, from the ECB and the Justus Liebig University and THM Business School in Germany, collected video recordings of ECB press conferences from its official YouTube channel. These press conferences featured Q&A sessions with journalists to explain monetary policy decisions.

The researchers applied SHORE facial recognition software to analyze and quantify emotional expressions such as happiness, anger, sadness, surprise, disgust and fear on a minute-by-minute basis. They also looked at valence, or whether an emotion was positive or negative.

They also used transformer-based machine learning models for vocal analysis to measure vocal patterns of the two ECB presidents. The researchers classified vocal emotions into three buckets: whether they sounded or happy or not, how intense and engaged they were, and how assertive, authoritative or confident they sounded.

The study used natural language processing techniques to gauge sentiment and compare verbal and nonverbal communication.

Then they matched the emotional cues to the words the two ECB presidents were saying and then watched how the markets reacted. The study looked at the bond yields of Germany, France, Italy and Spain, the euro-dollar exchange rate, the BUND Future, inflation and the Eurostoxx 50 index of 50 Eurozone blue-chip stocks from 11 countries.

The researchers made sure to separate out distinct emotional differences between Draghi and Lagarde — the former had higher sadness levels, and Lagarde was happier but could get angrier as well, as well as differentiate between causal and correlational impact on the financial markets.

While the results show that emotions do amplify market reactions, a newly released paper from the National Bureau of Economic Research and the University of Essex (U.K.) cautioned against using large language models and machine learning to classify emotional cues in central bank speeches.

The February 2025 paper “Automated Detection of Emotion in Central Bank Communication: A Warning,” argued that even advanced language models struggle with understanding subjectivity, especially in the nuanced and complex communication of central bankers.

“Large language models outperform traditional models on some occasions; however, the results depend on a number of choices,” wrote authors Nicole Baerg and Carola Binder.

“We therefore caution researchers from depending solely on such models even for tasks that appear similar,” they added. “Our findings suggest that central bank communications are not only technically but also subjectively difficult to understand.”

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.

The post Beyond Words: AI Shows Facial, Vocal Cues of Central Bankers Can Impact Financial Markets appeared first on PYMNTS.com.