Digitizing the Vibe Economy for Fun and Profit

The Vibe Economy, an idea that how one feels about current financial conditions is more important than what those conditions actually are, is heading into a new phase. And now it’s going high-tech.

Rather than rely on anecdotal evidence from random consumers and questionable opinion surveys, the business press is looking to up its game with the help of an algorithm.

Bloomberg Economics recently announced the creation of what they call a Fed Sentiment Index, which purports to statistically categorize headlines regarding Federal Reserve Board actions as an indicator of future activity. In a world of bad ideas, this one has to be near the top of the list.

According to Bloomberg, the index is based on a natural language processing algorithm, which used artificial intelligence to develop models to understand and process natural language. The algorithm is “trained” to read news media headlines on speeches from Fed members and press conferences and then “scores” them on a scale from “ultra-hawkish” – meaning a predilection by Federal Open Market Committee members to raise interest rates – to “ultra-dovish” and more inclined to cut rates.

On the surface, it sounds like a good idea. With all the news about interest rates and Fed monetary policymaking filling the airwaves and digital servers, finding a way to bring a sense of order to the reporting should produce a more coherent sense of what the Fed is thinking and what it is likely to do in the future. While the idea has merit, the execution leaves a lot to be desired.

For most people, the use of technology is seen as a positive, removing the emotional bias that humans bring to serious topics like money and interest rates. But the dispassionate, statistical veneer of the Bloomberg index is just that – a bit of polish on a digitally enhanced system that masks its flaws. And there are plenty of flaws.

The biggest problem with the headline-mining algorithm is the headlines themselves. No matter how “objective” a journalist claims to be, headlines are a framing device for a story, designed to signal to the audience the significance of the story. But that significance is determined not by the algorithm but by the author. And in a world of clickbait headlines and engagement farming in digital media, cherry picking headlines doesn’t provide a transparent foundation for drawing any conclusions about Fed thinking.

With multiple news platforms competing for eyeballs and knowing the reading public gets its news mainly from skimming headlines, the index is based not on the substance of a story about the Fed, but how splashy the headline is.

At its heart, the Bloomberg index confirms one of the earliest truisms of the computer age. The end product is only as good as the data collected. In the case of the Fed Sentiment Index, that data is suspect for a variety of reasons. Or as the old-timey computer nerds used to say more succinctly than eloquently, garbage in, garbage out.