Tyler Cowen serves as interlocutor in “Austan Goolsbee on Central Banking as a Data Dog: The data-driven Fed President on why human judgment still matters in monetary policy” (“Conversations with Tyler,” June 25, 2025, video, audio, and transcript). As Cowen notes, Goolsbee “had a long-standing teaching post at the University of Chicago, served in the Obama administration, and now is president of the Chicago Fed.”
On whether he is a “hawk” or a “dove” in monetary policy matters:
I was a data guy, as you know, in the field of economics. Soon as I got there [to the Chicago Fed], there’s all this pressure from the press and from others: “Are you a dove? Are you a hawk?” I used to say, “Look, I’m not one of the birds. I’m in the data dogs.” The first rule of the data dogs is, there’s a time for walking and a time for sniffing, and knowing the difference between those. I would say that discipline of academic economics — getting into the data — is super useful. …
[B]efore you can conclude anything, you’ve got to get a taste of, is this a supply shock or is this a demand shock? In a way, a lot of the machinery of central banking and macro analysis, let’s call it, is oriented around demand. I’m not disputing that, in the past, that has been the source of the most frequent business cycle variations, but I’ve tried to caution everybody in weird moments, like when you’re getting major developments on the supply side — whether they’re labor supply, or supply chain, or productivity growth, or a number of things that are hitting the supply side. Maybe all bets might not be off, but the training sample LLM version of being a central banker is going to be prone to hallucination problems because it’s going to give you things that are wrong because supply shocks might be driving inflation, not demand.
On high housing prices:
[O]ur district in Chicago is heart of the Midwest — most of Wisconsin, Iowa, Illinois, Indiana, Michigan. I’m out talking to business people. I’m talking to individuals, and overwhelmingly, what you hear is despair — I would even call it despair — about the cost of housing. That housing — they can’t move. This is not just in cities where you could argue a lot of it maybe has to do with building codes and zoning. We went out to the Iowa Farm Bureau, and in rural Iowa, I asked them, “What’s the biggest problem?” They said, “Attracting workers.” I said, “Why is it so hard to attract workers?” They said, “Because they can’t afford to buy housing.”
I’ve spent a long time trying to think that through. It’s not wrong that it’s just more extreme. … If you take the 12 years before COVID, house price inflation was 3.5 percent or 4 percent a year, and goods price inflation was actually deflation of around 1 percent a year. The relative price of housing has been rising 4 percent, 5 percent per year for a decade and a half. It doesn’t take a PhD by any means to recognize that something compounding at 5 percent a year is going to add up to a big number. …
I don’t fully understand why the relative price of housing has been trending upward like this. I find it hard to explain. I have a paper, you might’ve seen, with Chad Syverson, that’s about negative productivity growth in the construction industry over long periods of time, which is itself a puzzle. Maybe that’s part of it. Some component of it may be regulatory in nature, but as I say, you see it in rural areas, too, where the land use regulation is not as prevalent. I think that’s a real puzzle.
On the econ nerd joy of attending meetings of the Federal Open Market Committee:
[W]hen we go to the FOMC meeting, I love hearing what the other presidents and the governors have to say. I said with no irony, “I consider the FOMC to be the world’s greatest deliberative body at this point.” No offense to the US Senate or to anyone else, it’s an amazing group. If you’re an econ nerd, you go into that room, and it is just about the coolest thing there is on this planet. The shades come down. There’s a giant table, and they go around the table …
On the prospects for AI technologies:
[A] lot of the grandest dreams for AI and rapid adoption, I think, are premised on extrapolating a growth rate. To the extent that some of the improvement in AI is not coming from improvements of AI theory or new algorithms, but instead from bigger and bigger data sets and more and more computing power, data sets and computing power have diminishing returns that will kick in pretty quickly. So, it would behoove us to remember that there was a time 15 years ago when self-driving cars were improving so rapidly that people were predicting that within five years, there would not be a single professional driver in the United States.
Then, basically, what happened is, the rate of improvement didn’t go negative; it didn’t go to zero. It just slowed way down. Now, we’re having some self-driving taxis in different cities, but we’re nowhere near what, 15 years ago, there was a group of real advocates who said, by 15 years from now, people are going to be like, “Dad, what do you mean people used to drive their own car? How dumb were people?” We’re still a long, long way from that.
