Pro Take: JPMorgan’s Fedspeak Evaluator Is Unsure About This Week’s Rate Decision

Investors in the federal-funds rate futures market put a 23% probability on a quarter-point increase of the Fed’s benchmark rate this week, and 77% chances on the rate staying the same, according to CME Group data Monday afternoon.

JPMorgan’s tool, which was launched in late April, gives a “hawk-dove score” to each Fed official based on his or her speeches, and Fed Chairman Jerome Powell has the highest hawk ranking on average over the past year. “Hawkish” describes central bankers who back tighter monetary policy to fight inflation, while “dovish” describes officials who favor easier policy to support economic and job growth.

Powell has indicated the Fed could pause its rate increases to assess stresses in the banking system and the effect of its 10 rate hikes since early 2022. But if the Fed does hold steady this week, another rate increase could come later as the central bank looks to bring inflation down to its 2% target.

JPMorgan declined to comment on its new tool, but a research note by the nation’s largest bank describes the scoring model. Its results are available to the bank’s clients.

“Over the past few decades, central bank communications have become as important as policy actions themselves. This has raised the scrutiny placed on policy statements and speeches,” JPMorgan says.

The idea of evaluating Fed speeches with a computer has been around for several years. JPMorgan first considered a tool to evaluate the speeches in 2019. Recent advances in technology made the project viable, the bank says.

JPMorgan says it has trained its scoring model to think like a central banker through speech patterns, based on the evaluation of about 4,000 sentences. JPMorgan has loaded Fed speeches dating to 1998 into the model.

The bank believes computer scoring can be more objective than evaluations by journalists or economists—and faster. The core natural language processing technology is similar to what enables consumers to interact with automated customer service over the phone.

Over time, the model could incorporate artificial intelligence. “Technology is evolving rapidly and our tool will evolve in stride. We are already working to upgrade our model to the latest ChatGPT environment,” JPMorgan says.

The bank’s hawk-dove score and other computer-based efforts in evaluating so-called Fedspeak may be more objective than journalists and economists. But time will tell whether computers do a better job of figuring out what Fed officials are saying about their intentions on interest rates.

Write to [email protected]