Turning Words into Data, Prattle Watches the Fed

  • Sunday, April 26, 2015

By David Nicklaus – St. Louis Post-Distach

When the Federal Reserve updates its policy statement Wednesday, investors around the world will examine it for clues about the direction of interest rates. A small firm in St. Louis will be scrutinizing the statement too, but not with human eyes. Prattle Analytics’ computers will run the 500 or so words through a textual analysis algorithm and distill them to a single number that tells clients whether the central bank is more hawkish or dovish.

Hawkish, in this context, would mean the Fed is eager to raise short­-term interest rates; dovish would mean it is content to keep them at essentially zero. A move in either direction is a big deal to traders in the $160 trillion global debt market. Mostly, they react to news services’ instant headlines, which focus on a few key words in the central bank’s statement. This can be dangerous. In December, traders noticed that the words “considerable time” remained in the Fed statement. They briefly bid up bond prices, assuming that the Fed remained as dovish as it had been for months. The market sold off minutes later after traders realized that the full statement was more hawkish.

Evan Schnidman, Prattle’s chief executive and co­-founder, says his algorithm correctly picked up the hawkish tilt. It produces a score in less than 10 seconds, about the same time it took traders to read the incomplete headlines. “The focus on the phrase ‘considerable time’ caused a lot of people to lose a lot of money in December,” Schnidman said. “A lot of headline traders called us that day and said they were interested in our data.”

Prattle grew out of Schnidman’s Ph.D. dissertation at Harvard University on how central banks communicate. He was surprised by markets’ fixation on a few words out of hundreds. “I felt like people were leaving data on the table,” he said. “Words are data, and people were not looking at words as carefully as they could.”

Schnidman co-­founded Prattle with William MacMillan, who had been a postdoctoral researcher at Washington University. They first thought they would sell forecasts and economic analysis, but that market was tough to crack without a track record.

What they really had to sell, the pair realized, was data. With the help of a $50,000 Arch Grant and money from private investors, MacMillan set up a data development operation at the Cambridge Innovation Center in St. Louis.

Schnidman is establishing an office in CIC’s original Boston location, and he says Prattle may eventually need a New York presence.The firm publishes data on 18 central banks around the world, from Switzerland to India to Taiwan. It covers every statement, speech or other formal communication by a central bank official.

For the Federal Reserve, that’s almost one data point per business day. Fed officials’ speeches have been slightly more dovish in recent weeks, so Prattle is predicting a similar bias in this week’s policy statement. To be precise, its Fed sentiment indicator stands at minus 0.28, on a scale where negative means dovish and positive numbers are hawkish. If Wednesday’s statement scores below minus 0.2, Schnidman says, a rate increase at the next meeting in June is highly unlikely. Ironically, the Fed itself may benefit from having its lengthy communications boiled down to a single number. It doesn’t want its intentions to be misunderstood, as happened during the “taper tantrum” of 2013. If Prattle’s big­ data approach can keep that from occurring again, it will be welcomed by traders and central bankers alike.

Read article on St. Louis Post-Dispatch



The Prattle Team,
click here to subscribe to the signal blog