Will AI help the Fed conquer inflation? With Austan Goolsbee

Hi listeners, our inaugural FT Weekend Festival in New York City is fast approaching, with a lineup featuring Paul Krugman, Martin Wolfe, Jillian Tet and Palenti Moore. Join on Saturday June 20th at Spring Studios or online. Register now and as a podcast listener, save 10% using our code FT Podcast. Don't miss it. The Federal Reserve gets a new chairman this month and he will have one hell of an intray. Inflation is up, jobs numbers are mixed and the specter of stagflation is looming. There is one reason to be optimistic though, artificial intelligence. What if AI turbocharges productivity, allowing interest rates to fall? This week I'm asking, will AI help the Fed conquer inflation? This is the economic show with Simea Cainz. This week I am delighted to be joined by Austin Gulsby, president of the Federal Reserve Bank of Chicago and speaking to me from California. Austin, hello. Hi Simea, great to see you again. Great to have you on. Okay, well, on this show we always start with a silly question. So on a scale of one to 10, how excited are you about AI and the effects on the US economy? Oh, in the short run, the long run, what are we talking about? However, why not both? However you want to answer, I gave a talk here at Stanford about a year and a half ago in which I highlighted that we had a string of faster than expected productivity growth and there was a kind of a debate among economists as you know, some saying this is just a one-off increase coming out of COVID and I highlighted it at least looked like maybe it was concentrated in industries that were users of AI if you interpreted broadly. So it's not probably big enough yet in terms of its adoption to explain the majority of that, but if that has added, let's call it a tens of a point to growth, maybe two tens of it came from AI. So I'm going to give it a point two. We can measure it in basis points if you want. I might give it 20 basis points. I'm sorry, I'm going to need a number from you on that scale of one to ten. So I guess that's a two. So in short run, not that excited. If its biggest advocates are correct and it's going to deliver massive productivity growth, that's tremendous. It'll make us rich and we'll have to deal with disruption problems, but overall it'll be great. Technological improvements, it's the thing that made us a richest major country in the world. Okay, but you just said if in the long run is your number higher than a two out of ten? Could be, could be way higher than two out of ten, but I want us to just be a little careful over hyping what the even in the medium run technological improvements do to the economy. And I'm old enough that back in the internet, what turned out to be bubble, I was doing research about the internet at that time. And we had a similar dynamic of massive stock market valuations on a very speculative basis. And I remember saying at the time, I think the internet is going to change the world. I don't think that means that we're going to have the NASDAQ going up 30% a year every year forever. But it's still going to be an tremendously important technological change could raise productivity and could form the basis of a bunch of jobs of the future. And all of that proved to be right. It's just a technology adoption is not instant. This isn't the first time we're doing this. New technologies that are disruptive of major industries and lead to shifting. That's not new at all. That's been going on for a long, long time. And I do think there are some lessons from the past that it's worth our applying. So you're at the Chicago Fed and you're speaking to folks in your district. So in terms of the kind of qualitative impressions you're getting of how AI is transforming life or not, what kind of thing are you hearing? People are all over the map and it varies a lot by industry. So software, the popular impression that it's fundamentally changing the labor market for coding and for tech people. I don't think that's wrong. I hear a lot in that space that it's both changing the nature of work for software engineers, but also changing the balance of do they want entry-level people, do they want just high-level people, etc. As you move into deep into traditional manufacturing and the Chicago district is kind of the heart of the Midwest. It has the most manufacturing of all the districts. They're still looking for what's the right use case. The prospects that it is massively disruptive of the labor market in the short run feels a little less applicable in heavy manufacturing so far in agriculture so far. The interesting case is a number of service sector industries that are high, normally high skill in healthcare, legal, accounting and those kind of spaces. Finance for sure, they're experimenting. Could this replace a lot of activity that they currently hire people to do, but writ on a larger scale? I think the same thing that drives you crazy with the AI overview that it's like, is the restaurant open today? Yes, it's open and then you drive there and it's not open. In several of these settings, the stakes are high enough that they can't really afford there to be mistakes. And so trying to figure out, is this converging on a no-mistake version? Or are we going to be dealing with this for a long time in which case the applications to some of the more sensitive stuff will be harder to envision? That's really interesting. Can I ask about specifically inflation? This is the big question. We've got some people saying that AI is good news and could allow for lower rates because it will put downward pressure on inflation. Others are more concerned. Where do you put yourself? I think the big question is your first question. Is AI going to increase productivity growth and make us massively more rich than we are today? Then there is a second grubbier question, which is more of my day-to-day world, which is, what is it going to do to interest rates? If we have lower inflation because of AI, is this going to be like the 1990s, which is perceived to be a time when we had fast productivity growth that came about from information technology, I am a little cautious for the following reason. If you go think through the economics of that, I think it makes a huge difference whether the productivity growth is a surprise that was not, nobody saw coming or was expected and people changed their behavior planning for it. In the mid-90s, the data did not show the productivity growth yet. Alan Greenspan was the Fed chair and he deduced that there must be productivity growth because corporate profits were really high. The unemployment rate was going down and there was no inflation. But by the end of the 90s, his tune had changed and he started saying, well, now the data were very much showing the productivity growth. People were talking about we've got a new economy. We'll never have a recession. Again, productivity growth is going to go forever and he began highlighting, wait a minute, if everybody expects that we're going to have this productivity growth, we might overheat the economy. The stock market might go up too fast and the Fed began raising rates pretty rapidly from 99 to 2000. They raised the rates six times in less than a year. I think that dynamic, there is a lot of wisdom in that. And to the extent that so far, AI's impact was not expected has landed on us and we're trying to have been trying to figure out is this for real or is this a one time that naturally reduces inflation. And I think the natural response to that is lower rates. But if this thing is predicted and if even half of what the advocates of AI comes true, there's going to be overwhelming. Remember future increases to your income, AI massively increasing our productivity over the next 10 years. That's just a wealth effect today. And we need to be on the lookout for overheating in the economy in the short run before that productivity bounty has actually showed up and you can easily overheat the economy, you could easily generate inflation in the short run unless the central bank raises the interest rate to try to shift some of that pulling forward bag. Let it stay there. Don't pull it forward. Let it's let it sit there in the background. So the more hype there is about what productivity is going to come from AI, the more we should be on guard for evidence of shifting to the future, whether that's spending wealth effect out of stock market, whether that's massive increases in data center investments leading to non AI industry saying, geez, the weasel price of electricity is through the roof. All of those are of the form in the short run, we got to think about the grubby question of whether inflation actually would go down from AI or whether that time shifting would lead it to go up. In the logger run, in our language that's called the R star, what's the natural rate of interest that keeps the economy in equilibrium. The R star goes up if the growth rate is higher. And so we're going to have to balance out those two things in the short run. Okay, well, there's a lot of things to unpack there. So let's take them on systematically. So I suppose one is, you know, you've got all this AI investment and there could be short term bottlenecks, right? There could be a huge increase in energy demand, right? And that could just push up the price of energy. So you could get those short term effects on inflation. There's a separate effect that you're describing, which is a kind of anticipation effect. So people are so excited about the future productivity gains from AI that they essentially pull consumption forward and you get a kind of overheating in the short run. Could you just as both spell out what you think the mechanism is there? I think those two mechanisms that you just described in my world, that's the same mechanism. It is if people are expecting the future to look very different and that their income is going to be higher in the future and we're going to have a productivity bounty in the future. That's going to go into equity prices today and you have seen it and that's what happened in the in the late 90s and people are going to want to smooth their income. If you have a big increase in your wealth, you start spending today and if you're a business, you start investing today because you want to have this payoff that's coming in the future. All of those things are pushing toward the possibility of overheating in the here and now and it hinges on this question of is the growth that comes from AI expected or does it land on us unexpectedly? And that makes the interest rate question more subtle than the broader question of is it good for us if AI increases the productivity growth rate? Yes, it's lovely. It will make us rich. The question of well, does that mean that inflation will go down and that the Fed can lower interest rates there? Like I say, not necessarily. It depends if that's fully expected. You got to watch out for the it changing changing people's behavior. Then there's a second stage, which is what I've just described has nothing to do with bubbles. I just described a thing that is not bubble related at all. It's just about the fundamentals. Then ask yourself, is this a bubble? Is the equity valuation pressure that upward that we've seen in the excitement justified? Is hard to answer that question in real time? It's also worth remembering other tech, massive infrastructure investment analogies like the rollout of fiber optic, the building out of the 3G, the 4G, the 5G mobile phone networks. They were huge expenses. They did not end up being massively profitable for the companies that did them because there was more competition and the competition kind of drove down the profits or what the economists would call the rents in that space. If you were nervous about valuation, I think it would be in this space of are there really going to be barriers to entry that are going to make these specific companies so profitable that it's going to justify billions of dollars of investment in the data centers or are the margins going to remain relatively small because there's going to be competition of all of these models against each other and it's going to end up more like the telecom companies. Yeah, I mean, in a sense, it's potentially quite good news, right? If they do end up like the telecom companies because it suggests that consumers are getting all of the benefit from the company. Yeah, that we get all the benefits. So that's what I say. And that was my observation about the internet in 1999 is I kind of thought, and a bunch of my research was about what are the consumer benefits from the information that we're able to gather and that would look at life insurance, you know, as people could go on the internet and compare prices. You see the price of term life insurance dropping pretty precipitously where they adopt the internet. That idea that the consumers may end up being the ultimate beneficiaries. Great. That's part of what makes new technologies wonderful for the economy. If you can correlate the private sector profits and the incentives to invest with delivering massive consumer benefit, that's one of the glories of a capitalist system, but it doesn't always work out like that. But just so I can be really clear about them, you know, the thing that you will worried about, right? So essentially a lot of enthusiasm about AI and you've got a lot of excitement and that is going into equity prices. And so one concern could be that people start spending based on that wealth that they've now perceived. Another one is that you've got an effect through the investment channel where companies are trying to get on the AI bandwagon. But I suppose it's sort of an empirical question of what do we know about the link between those things and inflation. Right. So, you know, the stock market wealth effect being inflationary that feels like an area where we should have empirical evidence. What does that say? There are two parts. What does it say in the past? And what is it saying right now? I would say in the past, I've been kind of a skeptic of the magnitude of the wealth effect on consumer spending. It exists. I've never thought it was giant. I'd say that empirical evidence taken collectively is that there is a wealth effect. And so if you start to see significant increases in valuation, you would probably see some some race spending. And my inner economist is not troubled by that. If you raise your lifetime income, you should raise your spending over your whole life. And if that means borrowing from the future, as long as you're pretty confident that you're going to have the higher income, that works fine. Both of the mechanisms you described, wealth effect on consumption and businesses increasing the investment go at this question that I'm concerned about, the short run bottlenecks that arise from the more you hype, how great it's going to be in the near future, the more convincing you are to people right now, we better go do even more of that. We got to build even more data centers. And then that brings up the second part, which is what if they're wrong? Well, what if it is a bubble or what if they're just not correct? And it's 2000. And the internet isn't going to grow at 20% per year for the next decade. We got in a big capital overhang where they laid the fiber off if they built all of the infrastructure for a world where the growth rate wasn't did not deliver what it was anticipated. Bubbles that pop or overhangs like that often tend to lead to recessions. So that's a second category of concern. But I can be convinced for sure, I think the evidence so far is they're not as much stock market effect and bottleneck effect as there could be. I think for all the talk about data center investment and its contributions to growth, I think the main driver of stability and growth in the US economy for the last two years has been a broad-based consumer spending. US consumer just chugging right along. That's the main thing that has kept growth high. The data center space is not as big as you would think. And what people tend to do is take all the capital expenditure on data centers and divide by GDP and say, ooh, look at how big that is, failing to remember that something like half the spending on data centers are imported products. And that doesn't count to GDP. It gets added as investment. Everybody gets subtracted as import. So the actual contribution of AI data centers to GDP growth, it's growing very fast, but it's not nearly as big as what it's some of its biggest advocates claim. >> Okay. Well, I just have one more question before we throw to a break, which is about bottlenecks, right? Because we've been talking about bottlenecks associated with AI and data centers, but there seems to be a rather bigger bottleneck hitting the global economy right now. >> Right. >> Which seems to me could quite easily overwhelm any of these short term pressures from AI. Do you have a sense for which direction this shock will end up pushing? I mean, espos, there are those who worry that it will lead to so much demand destruction, that the economy will be so hurt that you'll need to be cutting rates. Others worried about inflation being embedded into the system, which could call for higher rates. >> Yeah, look, that's, that's, I'm worried about all of those. I'm worried about all of those. Now, it's worth adding that the US since the 1970s became a much more significant energy producer. So I do think on the GDP side, if we had an extendably high oil price, that would probably bring back the frackers. And we went through periods in the US where a big driver of GDP growth was the capital investment of the energy production sector. You would probably at least get one part of the economy pushing back the other way. But the inflationary impact, I think, if this thing is extended, could be a problem. And when I'm out in the seventh district talking to manufacturers, which I was, as soon as the war began, I talked to the major auto manufacturers and several other industries. And I said, what does this do to the supply chain? And there was pretty much consensus that they didn't think it would do much, comma, as long as it didn't last too long. And they said, because they had some parts and supplies, it would have had built up inventories that it would be fine. But if it went for too long, as one of the auto executives said, remember, just in time manufacturing means that at any given moment, most of the components are not sitting in a warehouse. They are on a truck or on a ship or on a plane. The price of fuel makes a big difference. And if we stay at a hundred plus dollars a barrel for a year or for two years, we're going to start seeing disruptions in the shipping channels in there. There may be again reminiscent of COVID times. There may be a bunch of things that we just can't get access to. And that would be a big mess. That would be a really big mess. It's almost the exact opposite of an increase in the productivity growth rate. I'm still hopeful that that's not true. And certainly if you look at the markets, they don't seem to suggest that they think that it's going to be extended. But we, the Fed, were burned before in the recent past saying, oh, yeah, this is going to be transitory. It'll be gone by the end of the year. So I just want us to be circumspect for this to be piled on before the tariffs, the tariffs increase inflation. And that's supposed to go away. That was supposed to be a one time increase in prices that the inflation part would be transitory. And now we're piling this thing on before that went away. That makes me nervous that people start taking for granted. None on inflation is rising a lot. Okay. All right. Well, look, I want to go to a break now. But when we come back, I'm going to be asking more about the outlook for the Fed and how it will change under its new chair and Kevin Walsh. We are back for the break. Okay. I understand that you have worked with the new Fed Chairman, Kevin Walsh in the past. Do you have any fun stories? And in a wrestling match between the two of you, who would win? He's definitely got more hair. I worked with Kevin Walsh. He was a governor at the Fed when I was at the Council of Economic Advisors through the financial crisis 2009, 2010 going into 2011. I liked working with him a lot. I felt like in a way we had been fox hole buddies and though our backgrounds were totally different and we came to our jobs from a totally different way. I had a lot of respect for him. And when they said he's going to be the Fed chair, I think he's going to come with the same seriousness that I've been at the Fed for about three years. And before that, I was 30 years on the faculty at University of Chicago. I've been struck by how seriously everybody who sits around that table takes the job. They're not down there. It's not about elections. It's not partisan. It's about every person coming with their worldview. How do they think the conditions are and what do they think the outlook is? And that's that's how they're making the decisions and having seen Kevin Walsh under a lot of pressure. I think that's what that's what he's going to do. So I look forward to the discussion and the and the debates we're going to have. He's got some new ideas that he's bringing. I think it's good. I think it's good. Let's let's talk about those ideas and whether they balance shape communication. All of those we need to refresh our thinking all the time. Okay, well, let's talk about some of those ideas then. Maybe starting with communication. So there's been a suggestion that maybe Fed members should limit the number of speeches or interviews they give. Personally, that would be very bad for me. So I'm not in favor of that for the record. But walk me through that debate and where you sit on it. I've seen a number of discussions about communication that have been attributed to chair design, Walsh. Many of the ones I've seen have revolved around the release of the SCP dot plot, the use of forward guidance in the statement and forms of communication like that, where he's been somewhat skeptical. And I would say I kind of share some of his skepticism. I said skeptical things about it before I was ever at the Fed. And I've tried to remain consistent with that. I understand forward guidance where for people aren't familiar. Sometimes in the statements that come with rate decisions, the committee will say things like the committee does not think that it will cut rates until and it could specify a time or it could specify an economic condition and tell it sees XYZ thing happen. I've been a little skeptical of forward guidance and an environment where you're not at the zero lower bound. I understand when you can't cut the interest rate anymore, you're already at zero, you're scrounging around trying to find some tools of monetary policy. But we're not in that environment right now. And I just think it's not binding. What's the point? You're saying a thing that if the conditions change, everybody knows you might not do what you're pledging to do. And who are we to try to put some binding constraint on future committees? We don't know what the conditions are going to be there. So personally, I kind of applaud a rethink of all of those aspects of Fed communication. You're asking about a different part. And I haven't seen some specific proposal. I think the strength of the FOMC as embodied from the very beginning is there's a diversity of views. People coming from different backgrounds, people coming from different parts of the country. You can see it in the meetings at the last meeting. We had four dissents from statement. I don't view that as a weakness. I view that as a strength. Anytime you're going to make important decisions, it's better to not have everybody thinking the same way. So I don't think that the communication argument that if people have a disagreement, they should not express what it is. I don't understand that critique. Okay. My final question is what do you think it's going to be like having chair Powell stick around? I don't know. I'm quite fond of the guy. So I'm happy he's he's going to be there. So you know, move his chair. I assume. So now he's going to be there. There's a science seat. So he's probably going to be over here on my right where the governors are. It's probably going to be a strange dynamic. Nobody knows how if he is the chair, chair, wash wants to run the meetings. He expressed either in the hearings or in the run-up to the hearings that he'd like it to be a little less scripted, a little more debating kind of a format. I would that could that could be interesting. I don't know and I'm not supposed to talk about what anyone else thinks on on the committee only what I myself think. So I don't know what the motivations are gonna be of others, but like I say, people take the job really seriously. So I don't think that that part's gonna change. - Okay, well, Austin, thanks for joining me. - So May, you always keep us on our toes. Thank you for having me. (gentle music) And that is all for this week. You have been listening to the economics show with Samarcane's. If you enjoyed the show, then I would be eternally grateful if you could rate and review us wherever you listen. This episode was produced by Misha Frankel Dubal with original music from Breene Turner and Sound Design from Samantha Javinko. The senior producer is Michaela Tendera. I'm Samarcane's. Thanks for listening. (gentle music) [Music]