If indeed this inflation hysteria has passed, its peak was surely late January. Even the stock market liquidations that showed up at that time were classified under that narrative. The economy was so good, it was bad; the Fed would be forced by rapid economic acceleration to speed themselves up before that acceleration got out of hand in uncontrolled consumer price gains.

On February 1, the Atlanta Fed’s GDPNow tracking model was moved up to predict 5.4% GDP growth in Q1. It was the perfect top for the hysteria; a blowout estimate that was every bit as hollow as the narrative. First, the narrative:

The economy is on track to put up blockbuster growth numbers in the first quarter, according to the latest forecast from the Atlanta Fed…

If the forecast holds, it would be the best quarter since the Great Recession ended in 2009. The previous highest was third quarter of 2014, which hit 5.2 percent.

The GDP tracking model was, in the beginning, a good idea. To be able to predict the big aggregate in close to real-time was a considerable draw. Unfortunately, it has proven just as difficult as ever. Demonstrating over time instead that there is quite a bit of noise on these shorter timescales, it’s the very problem that led to the creation of GDP (derived from GNP before it) in the first place.

To be fair to CNBC, the outlet publishing the article quoted above, these flaws were pointed out in the article. In particular, GDPNow over the past year and a half has fallen frequent victim to the ISM and sentiment. According to the ISM and other PMI’s, the economy is booming in a way it hasn’t since the turn of the millennium. According to GDP, not even close. The Atlanta Fed’s model has trouble reconciling these vast differences.

Part of the problem especially in the latter stages of 2017 was the introduction of artificial elements in the aftermath of Harvey and Irma. For a project like GDPNow, the shorter scale for measurement meant a susceptibility to distortions. This is particularly true when carrying estimates of output from one quarter to start off the next.