The dark side of a data-driven Fed
Data are central to Fed policymaking, and that's a good thing. However, data are imperfect and do not speak for themselves, so sometimes they lead us astray.
Today’s post is mainly for paid subscribers. First, for everyone, I discuss the Atlanta Fed’s GDPNow. Behind the paywall, I talk about inflation expectations from the Michigan Survey.
A key mantra at the Fed is that its policy decisions are data-driven, as they should be. The Fed makes a serious effort to ground its analysis, forecasts, and decisions in data. Even so, data differ in quality, often send contradictory messages, and can be hard to interpret. More often than not it’s how we use the data that can get us in trouble.
GDPNow is in the spotlight now
We have an enormous amount of data about the economy. And nowcasting models are an approach to summarizing them. The basic idea is to put a bunch of macroeconomic data in a statistical model and see what it pushes out. The Atlanta Fed’s GDPNow is one example. And it’s gotten a lot of attention recently.
Its current forecast for no GDP growth in the second quarter is gloomy. It’s much worse than professional forecasters. Plus, after a decline in GDP in the first quarter, their forecast would get us uncomfortably close to the recession rule of thumb of a two-quarters decrease in GDP.
GDPNow was in a recent Wall Street Journal piece, “Recession probabilities soar as inflation worsens:”
One closely watched model—the Federal Reserve Bank of Atlanta’s GDPNow tracker—estimates that gross domestic product is on track to remain unchanged at an annual rate over the three months through June 30.
I do not worry about the Fed getting too worked up over GDPNow, though they are watching it. The recent solid growth in consumer and business fixed investment should outweigh GDPNow. That said, GDPNow is a useful exercise and could be right.
It’s mainly the (change in) the change in inventories that is driving their forecast. Inventories were one of two factors behind the decline in the first quarter GDP.
Inventories may be a big drag on GDP growth again. And GDPNow incorporates monthly data so far; however, data on inventories are hard to parse. I looked at some available data and didn’t see red flags.
My concern is that GDPNow is cherry-picked out and given media coverage because it fits the recession narrative. Dig into the forecast details on its webpage, and you might think twice about it. See their FAQs. The Fed knows how to use this one.
Below the paywall, I talk about inflation expectations from the Michigan Survey and why they should not have played a leading role in the Fed’s 75-basis-point increase this month. My discussion draws on my work with the Michigan Survey at the Fed and my research with it.