Ben Harris on AI, Fiscal Sustainability, and the Resilience of the U.S. Economy

Ben Harris is vice president and director of economic studies at the Brookings Institution and previously served in senior economic policy roles at the Treasury Department and Council of Economic Advisers. Ben returns to the program to discuss why recent economic shocks have been less damaging than expected, the outlook for inflation and energy markets, whether AI can improve the U.S. fiscal outlook, why we should temper our AI optimism, how longer lifespans and higher interest rates could complicate the fiscal picture, and much more.

Subscribe to David's Substack: Macroeconomic Policy Nexus

Read the full episode transcript:

This episode was recorded on July 20th, 2026

Note: While transcripts are lightly edited, they are not rigorously proofed for accuracy. If you notice an error, please reach out to [email protected]. 

David Beckworth: Welcome to Macro Musings, where each week we pull back the curtain and take a closer look at the most important macroeconomic issues of the past, present, and future. I am your host, David Beckworth, a senior research fellow with the Mercatus Center at George Mason University, and I’m glad you decided to join us. 

Our guest today is Ben Harris. Ben is the vice president and director of economic studies at the Brookings Institution. He is also a veteran of the Treasury Department and the Council of Economic Advisers and is a return guest to the show, so check out his earlier appearance if you haven’t already. Ben, welcome back to the program.

Ben Harris: Thanks for having me.

Beckworth: It’s great to have you on. It’s about a year ago that you were on the show. We had a great conversation then about all things fiscal. We’ll return to that topic because it hasn’t gotten much better. [chuckles] It’s gotten worse, probably.

Harris: It’s gotten worse.

Reasons We Were Wrong

Beckworth: That’s a big, big deal for us. But you have a new perspective today to shed on this, and that is what role AI can help play toward that.

But, Ben, I have some other questions I want to raise with you because I was someone who was fairly certain last year, well, on Liberation Day, thereafter, that, man, these policies were going to cause serious havoc on the economy. I was saying this to people, and here we are more than a year later, and we haven’t seen the worst outcomes manifested. There’s been some strain, some stress, and so I was wrong. A number of us were wrong.

Now, some of you were more humble and probably more careful in your analysis. I bring this up because you’ve given some talks and written some notes about this very question. In fact, you have a paper titled “Four Reasons Trump’s Economic Agenda Hasn’t Tanked the Economy.” So, help me understand and make sense of what I thought was going to happen versus what actually happened.

Harris: Well, let’s just talk about the actual shocks and the— 

Beckworth: OK.

Harris: —policies we’re talking about. So, I think there are four big things that happened, mostly over the course of 2025. So, the first is with trade policy. I think one of the best ways to understand trade policy is through the average trade-weighted tariff. What’s the average tax rate on imports to the US? When Trump was inaugurated, it was around 2.5%, which felt like a big increase from where we’d been for most of my adult life, was around 1%. But to think it couldn’t go much higher was a failure of imagination.

Trump, of course, on Liberation Day, which is just a crazy title for a day in retrospect, but, took it up to the high 20s. Some people have it around 28%. Then you saw that reaction in the stock market and maybe a reaction in the bond market, and it came down, and then it went to around 15%. Then we had the Supreme Court ruling on IEEPA, and it looked like it went to around 12%. But the point is, we had this very volatile tax rate or tariff rate on imports. I think a lot of economists would have thought that that, even that by itself, would have caused a recession.

The second big shock was around net immigration. So, going back to the 2010s, for example, in any given year, we have about a million net immigrants to the United States. That spiked during the Biden administration for a bunch of reasons, but during 2025, the best estimates show that it probably went a little bit negative, within spitting distance of zero, and is projected to go even more negative over 2026. We have a lot fewer workers coming into the labor market and a lot fewer consumers living within the US border.

The third shock has to do with respect to the fiscal outlook and government debt. The president’s signature economic agenda—One Big Beautiful Bill or OBA or OBBB—that took on about $5 trillion in new debt. That was the third big shock. 

Beckworth: Yes.

Harris: And the fourth was related to Federal Reserve independence. You saw this unprecedented attack on Federal Reserve independence. You saw the DOJ investigating Lisa Cook. You saw the president’s own attacks, both verbally and using his own enforcement mechanisms, against Jay Powell.

You would have thought that the combination of these four things would have caused some big economic recession, but it didn’t. What I like to say is if you locked 100 economists in a room or a cave and let them out at the beginning of 2026 and described all these shocks and said, “What do you project for GDP growth?” I think everyone would have a negative in front of it. So, the big question is: Why didn’t that happen?

Beckworth: Yes. And you have the answers, right?

Harris: Well,I have the speculation.

Beckworth: OK.

Harris: I think one big thing is that the shocks were overestimated. If you look at the variation in the average trade-weighted tariff, they’re these big numbers going from 2.5% average tariff rate to the high 20s. But if you actually look at how much was collected, we’re talking about, I don’t know, maybe $250 billion in tariff revenue. That’s maybe about $200 billion more over the course of 2025 than we would typically collect. About $170 billion of that looks like it’s going to be refunded. So, we’re talking about tenths of a percentage point in GDP and extra revenue collected through tariffs, ultimately, when all is said and done, after the refunds went out. A lot of companies and CEOs expected the Supreme Court to rule the way it did, so they were planning on this money coming back.

Beckworth: OK.

Harris: So, when all is said and done, we introduced a lot of volatility here, but as far as sucking money out of the economy in the form of consumption taxes or tariffs, it wasn’t that big. 

With respect to the Federal Reserve independence, I think there was an expectation amongst markets—and I’d be curious to hear what you think, David—that the Supreme Court would step in and rule the way it did and protect the Federal Reserve. A lot of people I spoke to in financial markets are very well-informed. That was the expectation, that Powell would never really be fired, and that Scott Besson and other people around the president would step in and say, “Look, this is really going to rile markets if you start firing Federal Reserve chairs.” I think there was an expectation that this wouldn’t eventually happen. And so, the shock was overestimated in many—

Beckworth: OK.

Harris: —cases. A second explanation is that there were new stimuli that offset the negative impacts of these shocks. So, for example, data centers have been a real tailwind for the US macroeconomy, adding, I don’t know, maybe half a point to growth over the past year and a half or so. Also, I mentioned that the president’s One Big Beautiful Bill took on $5 trillion in new debt. Well, it also added, I don’t know, around half a point to disposable income over the course of 2026 when all these households got refunds that they weren’t going to get otherwise. That actually acted like a tailwind in the near term rather than a headwind.

I think when it comes to immigration, you saw a real slowdown in the labor market. The fact that we have a million fewer workers, the timing was right. I don’t think the president necessarily timed that around the macroeconomy, but it happened to be fortuitous. But there were all these stimuli and other conditions which offset these shocks. So, that’s the second explanation. 

The third potential explanation is that economists just got it wrong. And I think that there are—even after you control for the lesser size of the shocks, there are some surprises. For example, I think even in the labor market, we would have projected tighter labor markets in certain—

Beckworth: Right.

Harris: —occupations, like construction workers, for example. I remember asking labor economists, “Why aren’t wages for construction workers going through the roof when we have so many fewer immigrants coming to this country?” There wasn’t a great explanation, which brings me to the fourth potential explanation, which is that time will tell. We are actually now seeing construction wages start to outpace typical wages. So, some of our expected outcomes are happening over time. We’re starting to see a wage premium rise with respect to treasuries. We are seeing international investors starting to show some skepticism, maybe around the lack of independence. It’s possible that things will turn out the way we thought they would, but with a bit of a lag.

Beckworth: So, to paraphrase Milton Friedman, these shocks operate with a long and variable lag, some of them maybe, and we’ll find out [chuckles] over the next six months to a year.

Harris: Exactly. But I think when all is said and done, the shocks on economic perspective were really not as big as they seemed.

Beckworth: Yes. Now, is part of that because Trump had to cave, he had to negotiate, or he got sweet deals with certain corporations? I remember some of my friends who I said, “Oh, this is going to be bad. It’s going to be bad,” then it didn’t turn out bad, and they called me on it. I said, “Well, that’s because,” as you mentioned, “the tariff rates were lowered because of various reasons.” 

Why did they come down? Was it negotiations? Was it Apple coming into the White House and getting a sweet deal for phones coming in from China?

Harris: Well, because we had a tariff regime, until the Supreme Court ruled on IEEPA tariffs, that was controlled by one person, there’s only one person who really knows why [laughter] tariff rates came down, and that’s Donald Trump.

Beckworth: Yes.

Harris: There is speculation that the bond market had something to do with it. I think if you remember the few days after Liberation Day, the S&P came down, I don’t quite remember exactly, but on the order of 10%, and that didn’t seem to frighten the White House at all. But when the bond market started to freak out a little bit, [laughter] and you saw a pretty sharp rise in the 10-year, well, that’s when everything went back.

I think that the other surprise with respect to tariffs, specifically, was with the reaction from US trading partners. The reaction was pretty muted. Most trading partners came to the table, tried to negotiate these trade deals. A lot of the trade deals really weren’t that detailed and didn’t have much bite in terms of enforcement. I think the general approach from our trading partners was, “Look, let’s just appease the president with something which seems like a deal, like it’ll matter, but it won’t very much and we’ll get away from this.”

Beckworth: This too shall pass.

Harris: This too shall pass. 

Beckworth: Yes.

Harris: This bizarre negotiating strategy really wasn’t built into economic models.

Beckworth: And there are discussions about how some of the stuff we would buy from China is now coming from Vietnam, right? So, China’s sending stuff to Vietnam, then coming into the US.

Harris: Yes. That’s why I cited the difference between the average trade-weighted tariff and how much we actually collect.

Beckworth: OK.

Harris: And I think how much we actually collect, for reasons like you were talking around and around trading and obscuring the origin of the goods, is really important. There’s certainly some evasion that happens, understating the cost of the good and things like that. At the end of the day, I think we’re going to have to switch the source of how we look to the magnitude of the tariffs. We have to look at how many dollars are actually being collected.

Beckworth: Yes.I think this speaks to a bigger point, at least for me, and that is how integrated the global economy is and how hard it is to get away from that reality. There’s an argument to be made, maybe we should do more friend-sourcing, be more robust to supply shocks, and all of that, but it really is hard to completely unwind globalization when you are so integrated globally.

Harris: Yes, and for a lot of the goods, we’re fairly flexible, [laughter] except in times of a pandemic. I know we’re going to talk about the oil trade in a little bit, but when it comes to oil, you can switch sources pretty quickly. We saw that happening in 2022. There’s a lot of flexibility built into the system. I think the system is more resilient than it was 15, 20 years ago.

Beckworth: I like that. What you’re saying is, “Give markets a chance, and they can adapt and respond to some of these pressures, as long as they’re not too burdensome and too heavy.”

Harris: Yes. Particularly with respect to the United States, I think we’ve really learned how incredibly resilient [laughter] our country is.

Beckworth: Yes, we’re fortunate, huh?

Harris: We’re fortunate.

Beckworth: Yes.

Harris: It’s going to take something like a pandemic or a once-in-a-hundred-year financial crisis to throw us into recession.

Stubborn Inflation

Beckworth: OK, well, let’s talk about another development that’s been interesting to watch. Then we’ll get to your paper on AI and the fiscal outlook. That is inflation. Inflation’s been stubbornly high or above the Fed’s target of 2%. Last month, it came in a little bit lower, but it’s been a real struggle, right? The Fed has not hit its target for six years now. Part of the story has been inflation tied to oil. I know oil is considered a transitory factor [chuckles] until it isn’t. Part of that’s tied to some of the developments we’ve seen around the world, the war in the Middle East, Iran, the Strait of Hormuz . . .

And so, we have all of this happening. You’re someone who’s followed these markets closely. You had a recent discussion about this. Maybe you could fill us in on what should we expect? Because I think you’ve talked about maybe we haven’t seen the full effects yet of what’s happening.

Harris: Yes, so, my model for inflation right now over the near term, looking over the next year or so, is we’re at about 2.3–2.5% underlying inflation—so, higher than we’d like, but not dramatically higher than we would like.

Beckworth: Yes.

Harris: But then you have to layer on oil shocks, AI, and tariffs, and for all the reasons we just talked about, tariffs are a much lesser shock than we would have thought. I’m having a really hard time sizing the AI shock. I think a lot of people are. But then oil, oil is this big question mark, and that’s one reason why headline inflation had been so high. Anyone who’s filled up their tank knows that. So, what’s going on with oil markets? Well, alright, can I throw a bunch of numbers at you?

Beckworth: Please do.

Harris: OK.So, out of the Strait of Hormuz, there are about 15 million barrels of crude coming out per day. About 6–6.5 million barrels can be circumvented through these pipelines.

Beckworth: OK.

Harris: Saudi Arabia has a pipeline. UAE has a pipeline. It goes into the Red Sea, the Gulf of Oman. So, already, you’re taking that 15 million down to about 8.5 in a shortage in terms of daily exports. Then on top of that, you’ve got some oil is coming out of the strait. In the first half of July, for example, we saw much more active shipments coming out of the strait. It was recently slowed to a trickle, but it’s this ebb and flow.

In the paper you referenced, we thought it’d be about 2.5 million barrels per day. Now, you’re up to 9 million barrels per day, either going through these pipelines or coming out of the gulf anyway, because Iran says it’s OK, or because you’ve got some really brave ship captains who are willing to take their chances. Then, so, you’re only talking now about 6 million barrels per day or so in shortage. 

What Robin Brooks and I did in that paper was we applied traditional elasticities around supply shocks. We said, “Look, worst-case scenario is you get to about $125 per barrel, maybe $150 if things are really tight for the price of Brent.” Now, that’s problematic, but that’s not necessarily a global recession. 

Beckworth: Right.

Crude Is Not Really the Problem

Harris: My view—and I think that markets generally share the same view—is that crude is really not the problem. The real problem is what you’re trying to do with that crude is you’re trying to refine it and sell it as diesel or jet fuel or a ton of other distillates. What has happened over the past six months or so is that we’ve seen a contraction in refining capacity. You’ve seen about 10% of global refining capacity just get destroyed. A lot of this has come out of Russia because of Ukrainian drone strikes. I think Ukrainians just had to escalate the war to another level, and it was clear that sanctions weren’t working. We can get into that, certainly. You saw some destruction in refining capacity in the gulf. Not only will this last for some time because it takes a fairly long time to rebuild a refinery, but this is where you’re seeing all of the stress.

Energy experts tend to look at crack spreads, which is the difference between the price of crude and the price of the refined product, and those are through the roof right now. And so, the problem isn’t really crude. The problem is refined product, and I think everyone’s pretty concerned about diesel and particularly looking at demand coming out of Europe. Ironically, because you have seen the destruction [of] the ability to refine crude oil, that’s also lessened some of the pressure on crude. Before we were saying you’re up to around 6 or 7 million barrels—

Beckworth: Interesting.

Harris: —in shortage, it’s probably even less than that because refineries aren’t able to refine as much crude. The problem really isn’t crude, and I think you’re seeing that reflected in markets. The problem is really refined product, and that’s going to be with us for quarters or years, not months.

Beckworth: So, that capacity that’s been restricted, does that affect jet fuel more, diesel, or ordinary gasoline that we would buy for our vehicles?

Harris: I think that diesel has been—

Beckworth: Diesel, OK.

Harris: —the most affected. Russia put in place a ban on diesel exports. It had already sharply curtailed diesel exports by about 75% leading into the ban, so in some ways it just solidified what we already knew. But you’re also seeing that with respect to jet fuel, which is why airline prices have gone up fairly substantially, but diesel is the real concern.

Beckworth: And we care about that because that’s what drives our trucks across the interstate system in the US. Groceries, food, retail.

Harris: Yes, exactly.  And we care about it because, A, some people buy diesel directly—

Beckworth: That’s true.

Harris: —but also because it’s a passthrough into other consumer goods. Some of the estimates that I’ve seen—I don’t do the calculations personally—but is that the Strait of Hormuz impasse, or war—I don’t really know what to call it—conflict—has added about six-tenths of a percentage to headline inflation over the course of 2026. Of course, it didn’t start until several months and about 20 basis points to core. So, you’ve seen a passthrough rate of about one-third. So, it matters. It matters for food prices. It matters—

Beckworth: Sure.

Harris: —for manufacturing. It matters for transportation, but mostly it matters for when you’re buying diesel directly.

Beckworth: But you’re suggesting it could get worse because we’re going to hit these bottlenecks or these limits on the refined product.

Harris: No, I think that we’ve already seen most—

Beckworth: Oh, we’ve seen the worst. OK.

Harris: —of the worst of it. Now, the Ukrainian drone strikes could intensify. Or who knows what’ll happen with the war in the Middle East. Honestly anyone’s guess is as good as anyone else’s. If you see more destruction of refineries in the Middle East, yes, it could get worse, but it feels like we’ve reached this equilibrium.

Beckworth: Oh, good, OK, I was getting worried over here. [chuckles] So, ceteris paribus, you’re saying the worst is behind us because we’ve destroyed the capacity in the past. We saw the price shocks from that. So, it’s the price-level effect going forward.

Harris: Yes, that’s my assumption unless you see—Russia’s not exporting any diesel, right?

Beckworth: Right, right.

Harris: You can’t go any lower than zero. Maybe they could add demand. They could start trying to buy it. You are seeing these competitions across different countries. Asian buyers tend to lead and set the price. But, yes, my guess is that most of the worst has already been realized.

Beckworth: Oh, great. So, Ben, this raises an interesting question in my mind about some of the research from the past of Jim Hamilton. He used to really make the case that recessions are driven by oil shocks. But it seems like, increasingly, as we are less reliant on oil—or oil as a percentage of GDP seems to be falling, at least that’s what I recall—that maybe oil isn’t as consequential as it used to be? Is that a takeaway from this experience, or not?

Harris: Yes, that’s a big takeaway.

Beckworth: OK.

Harris: We’re more of a services-based economy than we were in the 1970s or 1980s. We have more alternative sources of energy. I think that one thing that may come out of this is that, just from a national security perspective, we may want to invest in some more wind turbines and solar panels.

Beckworth: Oh, yes.

Harris: Janet Yellen, when I worked for her at the Treasury Department, had a saying—I’m probably going to butcher it, but, “Putin doesn’t control the wind and the sun.” Look, a lot of our adversaries produce oil.

Beckworth: Yes.

Harris: And maybe it would be a good idea to diversify away from that.

Beckworth: So, alternatives are part of a national security imperative, I guess, moving forward.

Harris: One hundred percent.

A Novel Sanctions Regime

Beckworth: I want to go to some of the research that you and Robin Brooks have done on the shadow fleet of Russian ships carrying oil. I want to bring this up because, as you mentioned earlier, Ukraine has become very effective at hitting these ships. They’ve destroyed capacity inside Russia, but they’re also destroying the fleets in the oceans, the navy of the Russians, but also these ships carrying it. So, in some ways, Zelensky and the Ukrainian military, are they taking your advice to heart [chuckles] and shutting it down? Is this a referendum on your proposal?

Harris: Not at all. But let me explain why. When I was at the Treasury Department, I and a team of other officials and staff helped design what I think was fairly novel at the time, which was a sanctions regime that capped the price of Russian oil. To my knowledge, no one has ever tried this before. In really simple terms, if you went back to look at, for example, the sanctions against Iran, we, the United States, tried to go ahead and limit the quantity or the queue.

Beckworth: Yes.

Harris: Right? We said, “Look, we’re really trying to drive down Iranian export levels to zero.” That was the goal.

When Russia went ahead and invaded Ukraine, the price of oil spiked to over $120 a barrel in forward markets. Markets were very concerned about a disruption. This is coming out of COVID. The macroeconomy was very fragile. We didn’t know if there’d be subsequent waves of the pandemic. 

At the Treasury Department and in the Biden White House, we tried to design another sanctions scheme that would punish Russia but not threaten the global economy. And so, what that scheme did is said, “Rather than limit queue, rather than try to drive down exports to zero, we’re going to try to limit the price—”

Beckworth: OK.

Harris: “—at which Russia receives on every barrel of oil or distilled product.”

In fact, I would lead delegations to Europe, and we had to negotiate with our allies in Europe over exactly what the price cap was going to be. We actually had three different price caps: one on crude, one on high-value refined product, and one on low-value refined product. But the notion, I think, was pretty novel, which is, if you’re in a macroeconomic environment where you’re worried about what Jim Hamilton worried about, which was how an energy shock could impact the macroeconomy, well, then you have an alternative to completely shutting off—

Beckworth: OK.

Harris: —oil exports from a major oil producer. And that alternative is to limit the price at which it can receive on its exports. Now, how do you do that? You do that by controlling the services that you need in order to go ahead and trade a barrel of oil. So, you need a tanker. You’re going to have to go ahead and insure that tanker. You’re going to oftentimes need to finance that through a bank. You’re going to have to have that tanker flagged by a country. So, there are all of these services that go into it. At the time, we controlled all of these services. Western companies, not just the US, but the UK, Europe, Japan, we controlled about 90% of the oil trade that was coming out of Russia.

What we said was, “If you want to use these services, you’re going to have to go ahead and do it at the prices that we prescribe.” And it worked for a while. There was this really big spread between the market value of oil, if it came from anywhere but Russia, and what Russia received. When all is said and done, we probably limited Russia from getting tens of billions of dollars in extra revenue,m and you did not get the global recession that I think some had feared. There were banks that were saying we’d be at an average price of oil of $285 or higher.

Beckworth: Wow.

Harris: Major banks and reputable analysts. That is an instant recession. 

Beckworth: Right.

Harris: That never happened. But what also happened is that Russia became really good at circumventing— 

Beckworth: OK.

Harris: —the price cap. It did that by buying up hundreds of tankers, often sold by Western companies, and trading with countries not in the EU and not in our coalition, but increasingly with China and particularly India. And so, you’ve got this race with Russia with respect to the price cap. 

One thing that’s been unfortunate is that the Trump administration has sat on the sidelines. By the last day of the Biden administration, we had sanctioned 216 oil tankers; the Trump administration has sanctioned zero. The rest of the G7 has lowered the price cap on oil; the Trump administration has not moved at all. It’s still at $60. And in fact, the Trump administration also granted really sweeping exceptions or licenses on the import of Russian oil after it engaged in this conflict with Iran. So, the Trump administration has been fairly friendly to the trade of Russian oil. I still think the price cap concept could work, but it’s going to take more resolve on behalf of the United States.

Beckworth: This is different than what Zelensky is doing. I just brought up Zelensky because he called it a “long-run economic sanction” when they hit these ships and these refineries. It is very different. This is about controlling the price of the Russian oil export versus actually affecting the quantity that they’re exporting.

Harris: Exactly. What Zelensky is doing—and I understand why he’s doing it. His country’s under attack. I think he probably showed terrific patience waiting this long. 

Beckworth: Yes.

Harris: But he’s doing effectively what we did to Iran in the 2010s, which is trying to drive down oil exports to zero.

Beckworth: Altogether, yes. Very interesting. I’ll be curious to see how this all unfolds. Ukraine seems to be at the cutting edge of military technology. They recently sent some naval drones over that unloaded these AI-driven little mini tanks that went and cleared some fields. No humans involved altogether, but a two-stage process of nonhuman military technology going places. This is just like watching the movie Terminator, but in real time, [chuckles] in our lifetime. So, crazy.

Alright. Let’s go to your paper about fiscal sustainability. It’s titled “Can AI Restore Fiscal Sustainability in the US?” Before we jump into your paper, maybe walk us through and remind us how dire the fiscal outlook is for the United States.

Harris: Yes. So, I think, given the pace of change, maybe it’s best just to look over the next 10 years or so.

Beckworth: Let’s do that, yes.

Harris: OK, so, in just really simple terms, we run deficits of around 6% of GDP in every year. We’re projected to run those deficits. Really, the key thing to look at is not the total deficit number, but after you subtract out interest payments, what we call the primary deficit.

Beckworth: OK.

Harris: That’s around 2% of GDP. We spend about 4% of GDP on interest payments, and we’re still running this 2% of GDP structural deficit year after year for the next 10 years. So, if you want to give the economy any real chance of growing its way out of this mess, you need to have a primary deficit of zero or maybe a surplus of about 1% of GDP.

Now, the One Big Beautiful Bill that we talked about earlier made this worse by about 1% of GDP. If we were talking a year and a half ago, instead of saying 6% deficits, I would have said 5. Instead of 2% primary deficits, I would have said 1. From that perspective, it doubled the medium-term problem.

Beckworth: Wow.

Harris: If we want to give our economy any chance of growing out of this, we have to have primary deficits of around zero.

Beckworth: We’re now paying more on the debt than we do defense. Is that right? It’s one of the fastest-growing categories?

Harris: It’s growing quickly for a couple reasons. One, our debt just keeps growing and growing, [laughter] so the stock keeps growing and growing, right?

Beckworth: Yes.

Harris: If you double your mortgage, you’re going to pay more in debt, even if interest rates don’t change. The second thing that’s happened is that what we call the risk premium on treasuries is going up. What investors are demanding in order to invest in the United States, in order to lend to the United States, has been increasing from—we were at record lows a few years ago, and we’ve seen that inch up. I don’t know why exactly investors are demanding more, but I’ve got to think that some of the volatility we’re seeing from policymakers is part of it.

Beckworth: One last question on the fiscal situation before we move on to your paper specifically. But the current administration is issuing lots of T-bills, I think disproportionately more T-bills. I think there was some of this discussion in the previous administration as well. Right now, it may not be a big problem, but the concern is that if you keep doing this, at some point, you’re going to load on more interest rate risk. If rates do go up in the future, you’re going to have a larger interest rate payment on your debt, right?

Harris: Yes. It depends on the average duration of the debt you issue. If you’re feeling pretty good about the future, I think you want longer duration. If you’re not feeling so good, you want shorter duration. One of the things which is complicating this is the prevalence of stablecoin, which will rely on bills in order to operate. In order to have a judgment about the duration issuance with the Treasury Department, you really have to have a judgment on how much extra demand you think the stablecoin regime will add. I don’t have a good sense of that. I think some of the financial markets think it’ll add around $2 trillion. If that’s the case, then this policy doesn’t seem so concerning.

Beckworth: OK.

Harris: But, look, there’s nothing to stop investors from demanding more and more to invest in treasuries. I will tell you, I was in Europe a few days after the president threatened to invade Greenland, meeting with executives and investors, and they were really eager to pull as much money as they possibly could out of the United States. How could you possibly continue to invest in a country which is threatening to invade you? Right now, the 10-year is at, I don’t know, 4.5%, but there’s certainly no magical law—

Beckworth: Right.

Harris: —which prohibits it from being 6% or even higher.

Beckworth: I think we talked about this last time. I’m surprised it’s so low, honestly, [chuckles] given everything going on. Hopefully, I can have faith in the markets that they’re making the right call here. Things will get better. Maybe there’ll be reform. Maybe higher inflation will cause the public to want to have meaningful change. We’ll see what happens. Let’s go to your paper, which maybe can give us hope. Actually, after you read it, you’re not so sure that AI is going to restore fiscal sustainability in the US. Give us the bird’s-eye view of your paper. What is it trying to address, and what are the big takeaways?

Temper Your AI Optimism

Harris: There are some real techno-optimists in general. 

Beckworth: Yes.

Harris: Not necessarily with respect to the federal budget, but in general, as far as what AI will do to the US economy. One example I give is I had lunch with a fairly prominent tech executive. This is about a year ago. He said, “Look, in six months, the unemployment rate is going to b e 18%.”

Beckworth: Wow, 18%.

Harris: Yes. And so, there was all this optimism. Now, that sounds like pessimism because you’re focusing on the unemployment rate, but the optimism is that this is going to be transformative, and just sharply reduce our need for labor, right? We’re going to be able to turn over a lot of the tasks to AI. Maybe we can work three or four days a week instead of five. Now, that hasn’t happened. The unemployment rate is still quite low, in the low fours. 

But there is this optimism around what AI will do to the US economy. If you think we’re going to grow so fast that we don’t really need labor in the way we did in the past, then you’ve got to transfer that optimism to the federal budget. What we tried to do in this paper with my coauthors, Neil Mehrotra and William Overcash, was try to scope the size of this possible shock.

Beckworth: Now, we’ve had big transformations before. We’ve had electrification, computers, the internet. But you make an argument this is going to be potentially different, very different than those. How so?

Harris: Yes, so, we tried to get into the weeds a bit with this technology. We tried to ask how, precisely, will this change the US economy, and are there certain things we’re not thinking about? We start with what would look like a shock that resembled the 1990s, where you grow much faster than you thought. You get these fairly sizable productivity gains associated with increased use of the internet.

That’s our base-case scenario, where if the shock that we’re experiencing right now looks a lot like the late 1990s, well, then good news, we’re going to grow our way out of this fiscal situation, we’re going to go down to deficits of around 2%, and we’re going to get primary deficits of around zero. So, Elon Musk, you’re right. We’re going to grow our way out of it. But there are a lot of reasons to think that this shock will be different, and that’s what we’re trying to make the case for.

Beckworth: You look at several different scenarios where this would be different, like demographics, labor markets, healthcare, even defense. How can AI be fundamentally different from these previous general-purpose technologies that we’ve just talked about?

Harris: Yes.We came up with five reasons why you should temper your optimism a bit. I’m going to give you the punchline up front—

Beckworth: Please do, yes.

Harris: —which is that in general—this is a rule of thumb—I think however much you think that AI will drive down deficits, just cut that in half because of all these idiosyncratic reasons with respect to AI. So, if it’s going to drive deficits from 6% to 2% well, it’s probably going to be more like 4%. If you think it’s going to only drive down deficits from 6% to 4%, it’s probably going to be more like 5. Just cut your optimism in half.

The takeaway is not to be optimistic. The takeaway is, “Temper that optimism a little bit.” Let’s start off by talking about health. I’m actually really optimistic about AI’s potential to impact healthcare delivery and lower mortality rates, particularly in older Americans, for a bunch of different reasons. We already have pretty good evidence that is five, seven, in some cases, 10 years old, that AI improves healthcare delivery by making it more personal, it improves the pharmaceutical industry by reducing the time to market and by making the drugs more effective, and that it potentially can lower mortality by increasing our ability to diagnose with diagnostics. 

For example, diabetic retinopathy, which is when you are potentially having diabetes, and you’re having problems with your retinas. We have studies that are roughly 10 years old that show that AI can already have an impact on our ability to diagnose that. We have some studies that show if you switch to personalized medicine, which, I think AI allows us to deliver personalized medicine much better, that it reduces the prevalence of going to an emergency room by about 3 percentage points. Now, going to emergency room is really expensive, so if you can reduce the prevalence by 3 percentage points, there are potential cost savings there.

But there are also potential gains in longevity. What I like to say is that human lifespan is our most precious commodity. What is great news for humanity is terrible news for the federal budget. In our most optimistic scenario with respect to longevity, the number of Americans aged 65 and older goes from about 74 million to about 76 million. That means 2 million more people collecting Social Security, 2 million more people on Medicare, and this just compounds over time. We only took a 10-year outlook. The first reason to temper some of the optimism is actually good news: We might live much longer. Now, that means more draws on these entitlement programs. 

The second reason to be somewhat less optimistic has to do with investment in AI, and this productivity gain that we’re potentially on the cusp of, which means that we’ve got all these investments in the economy that are so much more productive. Investors are clamoring to invest more in Nvidia and data centers and everything else, but that drives up the equilibrium interest rate. Like we just talked about, we’re paying a lot more in interest. The stock of debt is growing. This is actually probably the biggest reason that we find to temper some of your optimism is that equilibrium interest rates—

Beckworth: Right, right.

Harris: —or neutral rates of interest could be substantially higher. In our most extreme scenario, we raise all interest rates by about 35% across the yield curve. Another reason to temper some of the optimism is around what could happen in the labor market. As we just talked about, we might need a lot fewer workers, which means more people who’ve dropped out of the labor market, who are on income support programs, programs like SNAP, programs like Medicaid for some, and that just means higher federal spending.

A fourth reason to temper some of that optimism has to do with national income. If we think that capital’s going to start capturing a greater share of this growth than labor, well, capital’s taxed less in the tax code than labor is, which means less in federal revenues, so we account for that as well. Then finally—and this was driven by discussions with defense experts—there’s the potential for an AI-driven arms race. I think is—

Beckworth: Lovely.

Harris: Yes, exactly. [laughter] This is actually purely bad news.

Beckworth: OK, yes.

Harris: This is wasted technology—

Beckworth: Right, right.

Harris: —but you could imagine, for example, a US-and-China arms race with respect to AI and how that’s integrated into our respective militaries. I will say, since we drafted this paper, I’ve actually become more convinced that we’ll see these big, big explosions in military spending related to AI.

Beckworth: Yes.

Harris: Now, some people think, “Oh, no, it’s going to be so much more efficient,” but if it’s more efficient for us, it’s more efficient for our adversaries.

Beckworth: Right, right.

Harris: I’m pretty convinced we are on the cusp of an AI-driven arms race.

Beckworth: I was listening to another podcast where they were talking about the military implications. We tried to rein in and limit the use of Claude Mythos, for example, because it had this potential to hack into anything and destroy anything. They said, “Look, this is great, but China’s only six months behind on its own version of Mythos.” They’re going to open up to everybody. They’re not doing what we’re going to do. There is this unnerving, disturbing race. Who has the most effective AI? Who gets there first, and what can they do with it? You can imagine AIs duking it out and who becomes the dominant AI. That is not a very pleasant image.

Harris: Exactly. In our scenario, we had about $35 billion a year being devoted to this. If we start getting worried about the safety of our borders, it’s going to be a lot higher than that.

Beckworth: There was a TV show on, I don’t know, maybe a decade ago, called Person of Interest. Did you ever watch that? [chuckles]

Harris: I’m aware of it, yes.

Beckworth: Basically, there was a sentient AI, a good one. Then you find out near the end of the show there’s a bad one, and these two AIs, they duke it out. They have their own people and groups. It’s this incredible story. Now, I don’t know if what you’re saying is [as] dystopian as that, but it is something to think about. What happens if these things get out in the wild and some party that doesn’t have benign interest uses them?

Harris: Absolutely. Time will tell on this one. I can say that I think that some of the fantastic claims about AI are starting to get dialed back, even relative to a year ago. This shock to me is looking a lot like the shock in the late ’90s.

Longer Lives, But Larger Debt

Beckworth: OK, well, that’s good to hear because we talked earlier about Ukraine having naval drones that offload robots that go and clear the field of machine guns. Throw in a little nefarious AI to that [chuckles], and we have Terminator, right? The makings of a movie. 

So, let’s go back to a little more realistic, though, concerns, because I’m throwing things way out there. Let’s go back to the health concerns that you outlined. You mentioned on many fronts, it’s great. AI is great for health. You get personalized care, you’re diagnosed better, things are faster, but then it’s going to get more costly. You said 2 million more folks age 70 and above. Is that what I heard?

Harris: Yes. The way we implemented this was through faster reductions in mortality rates. We started off with the Social Security Trustee’s assumption, which is that for each age-specific mortality rate, reduces at 0.7% per year.

Beckworth: OK.

Harris: If you had a 5% chance of dying when you’re 80-something, it goes down to 49 point-something in the first year. What we said was, “OK, what if that triples in our most extreme example?” You’re talking about 3% average reductions in mortality a year. This is about half what we saw as far as reductions in Japanese mortality coming out of World War II. It’s not totally unheard of, but it is a pretty rapid acceleration in reduction in mortality rates.

Beckworth: Another way of saying that is that the population’s going to age. It’s going to get older and older, fewer younger people. Is that fair?

Harris: Yes. I think your chance of dying each year is going to go down.

Beckworth: OK.

Harris: You don’t have a very high chance of dying if you’re younger than 60 or 65 in the United States, so these reductions in mortality don’t matter very much for people who are not retirement age. 

Beckworth: Right.

Harris: What it means is you’re going to live a lot longer.

Beckworth: I bring this up because there’s been some recent exchanges on X about what is the effect of demographic decline in the US? There’s a recent paper out by some prominent economists who run contrary to all conventional wisdom, say, “Oh, well, shrinking population actually leads to higher real GDP for the working-age populations.” It’s a shocking finding. I’m not sure I totally buy everything, how they do it and what it finds. But then, I’ll mention this one person, Jesús Fernández-Villaverde. He responded to them indirectly, I think, without mentioning names.

He goes, “This misses the point. Even if you get higher per capita GDP of the working-age population from a shrinking population”—and to be clear, their argument is the population shrinks, we invest more in labor-saving technologies. This is the claim they’re making. He goes, “Even if that’s true, there is a political economy thing. As the population ages, the older folks are not going to give up what they feel they’re entitled to. There’s more rent-seeking. Everything from housing—they’re going to hold onto the housing stock. They’re not going to want to take cuts in their entitlements.”

I guess to add more concern to what you’re saying here, if we do have a longer-lived, older population, I think there’s a real political-economy concern here because they’re going to continue to demand these entitlements that older generations want.

Harris: Yes. The question is, what goes along with longer lifespans?

Beckworth: Yes, exactly.

Harris: Are we going to continue to promise a pretty fruitful retirement at age 65 if people are living on average to 105? The answer to that has got to be no. It just puts too much pressure on young people, unless we’re willing to just go crazy when it comes to net immigration and start trying to keep that ratio the same. 

We have seen fairly sharp increases in longevity over the past 25 years or so. I think on average, you’re expected to live about three additional years today than you were at the turn of the century, but we haven’t changed the promise at all. We’re still promising you full benefits at age 65. 

Beckworth: Yes.

Harris: And that’s happened slowly, and, so, I think there hasn’t necessarily been a true conversation around changing that commitment. But if we see some sharp changes in life expectancy, I would expect that to happen.

Beckworth: OK, so you’re saying then there’s a potential fix to this concern that if we live that long and there’s that much of a demand on the younger generation, there will be some revolt, some push for change, and we will see that age pushed maybe to 70 or something else.

Harris: Yes. It’s not just with respect to public benefits. It’s also with respect to housing.

Beckworth: Yes.

Harris: There are a lot of young people waiting to buy their first home. One nice thing about better healthcare delivery is that older people can age in place where they want to age. I have a book out on retirement. I think maybe I talked about it the last time on this podcast, but people don’t like going to nursing homes. They don’t like living in institutions. They like staying in their home. That’s great news for people. They can continue to do that.

The combination, I’ll say, of AI and robotics might completely revolutionize this. Maybe when you turn age 65, you purchase a home care robot who takes care of you for as long as you want. But we’re not seeing a lot of new studies in the United States. I think one of the biggest concerns for younger people is getting to a home. My only point here—and there’s also jobs, too. You don’t want to be a young person waiting for that job to open up. We may need to redesign the social contract.

AI and the Future of Interest Rates

Beckworth: There are a lot of areas where this could have an implication: housing, jobs, and so forth. 

OK, let’s go back to the interest rate point you made. Your point is this: As AI grows rapidly, or should it grow rapidly and push up real GDP growth, well, so will the neutral or real interest rate go as well. If you think of debt-to-GDP as one way to think of debt sustainability, yes, the numerator’s going up. You’re hoping that the denominator grows even faster, but it’s not clear that it will.

Harris: It’s not clear that it will.

Beckworth: Yes.

Harris: Yes. I think the way that I think about it is that you’ve got a constant number of dollars that’s competing for different investments. And as alternative investments to treasuries start looking better and better, you’re going to have to pay more on those treasuries to start attracting investors who are willing to take a risk for the adjustment there.

Beckworth: Yes.

Harris: This was a really tough one. We had to look at traditional relationships between productivity shocks and the neutral rate, and there is a traditional relationship there. So, in one scenario, we boosted all interest rates across the yield curve by 35%. That was our more extreme scenario where you get these really sharp increases in productivity. In the more mild scenarios, it was half that.

Beckworth: What happens when you have the really sharp increases? Does the debt burden really explode then?

Harris: Well, it depends. I think that the interest rates increase, and you’re paying more for every dollar, but you’re also expanding these tax bases.

Beckworth: OK, so GDP’s getting larger too.

Harris: Yes, so it depends. Think about this as a race between interest payments and the tax bases. Purely with respect to capital, if the S&P goes up to, I don’t know, twice what it is today, think of all the extra capital gains that are coming in. Think of what’s happening to corporate profits. Corporate revenues go from around 2% of GDP up to 4% of GDP. Well, then the fact that we’re paying an extra percent on interest doesn’t really matter because we’ve doubled that with respect to corporate tax revenues. So, there are all these different factors going on. I will say the model we built, which I’m really proud of, actually has 45 different variables in it—

Beckworth: Wow.

Harris: —and you can just go ahead and build your—

Beckworth: You can tweak—

Harris: —build you future. Yes.

Beckworth: Again, this goes back to the question of demographics, though. If the economy is growing really rapidly, so, the tax base is growing, incomes are flush, do you think people will be more comfortable paying more taxes, or are they getting even stronger entitlement? “This is my money.” You think during times of robust economic growth, people are happier, more content. They’re more open to immigration. A lot of good things come from economic growth. Would they be more willing to pay more taxes or to pay a bigger share to help pay down the debt?

Harris: OK, so, I have to answer your question with a question, which I hate doing, but I’ve got to do it.

Beckworth: OK.

Harris: OK. So, is AI technology labor-replacing, or is it labor-augmenting? If I’m a radiologist, am I out of a job now, or am I able to be twice as productive?

Beckworth: I see. It depends on what your outcome is, whether . . . 

Harris: It depends on what your outcome is.

Beckworth: Yes.

Harris: I think that’s a real fear, which is that you’ll see the combination of AI and robotics—which I think is an important distinction—will displace labor across the economy. We’re not just talking about call centers and paralegals. We’re talking about people who actually use their bodies to perform work. If that happens across the labor market, then who will be the beneficiaries? Not workers, or not many workers. It’ll be the owners of capital.

If I’ve now been fired from a job, if I’m in a two-earner household, and previously I was a plumber, and now a robot can come into a house and find a leak, and now I’ve got to rely on my spouse, who’s maybe a nurse and still has a job, do I want to pay more in that scenario? Absolutely not. Particularly when the Elon Musks of the world are now $15 trillion richer.

Beckworth: I guess the question is then, do the Elon Musks of the world, the capital owners, are they more willing to pay more to the common cause? You’re not certain they will be.

Harris: I think it depends on the executive. I think that some executives realize, “Look, this may be a historic shock.” There is the desire to give back and even shape this shock a little bit, not just let it happen. That’s where we are.

Beckworth: Yes.

Harris: We’re at square zero with respect to AI regulation.

Policy Implications

Beckworth: Yes. Let’s go to then some of your policy implications for these findings. Let me just throw one out that just naturally bounces off what we’ve been saying. But if there’s a way to get people to have a bigger piece or stake in the equity of America, Inc. I know Trump has introduced these, I don’t know, baby bonds, just to get people have some share of S&P 500 index fund or something. Is that a possible solution?

Harris: Yes. This is driving a conversation around the government taking equity stakes in tech companies, which makes me really uncomfortable—

Beckworth: Same here.

Harris: —just from a, I don’t know, from a common-sense perspective. How do you have a government that collects revenue but also owns a massive share in a company? Is giving away this equity in the company seen as being a way to circumvent any sort of regulation or give sweetheart deals? Are we going to get to a point where it’s just a race for who can give the government the most? If one tech company gives 5%, and another decides to come and give 8%, will that get favorable treatment? This is really bad. It’s starting to feel a lot like China, [laughter] and it doesn’t feel like capitalism.

Beckworth: No, no, no, I’m not at all advocating that.

Harris: No, no, I know you’re not. So, I’m building up. I’m building up.

Beckworth: OK.

Harris: There is a better answer here.

Beckworth: OK.

Harris: Rather than the government owning stakes in companies, American households should own more stakes in companies themselves.

Beckworth: Yes.

Harris: Congress actually passed a bill that I think has gone under the radar. It’s the SECURE 2.0 Act, and it makes changes to something called the Saver’s Credit. What it does is it provides low- and middle-income workers with a higher match on any money they put into their 401(k).

Beckworth: OK.

Harris: That costs—I’m not sure how much we’re giving up in revenue for that, but probably in the order of $10 billion. You could imagine something like the Saver’s Match on steroids, where we’re saying, “Look, if we’re going to have the shift from labor to capital in the US economy, let’s just all own more capital.” Not the government owning stakes of companies—

Beckworth: Right.

Harris: —but let markets work, right? If you want to invest in S&P, that’s fine. If you want to be a bit riskier and invest in individual tech companies, that’s your choice as a household as well. But right now, we give around $250 billion in tax breaks to US workers, but almost all that goes to the top 30% or so. You can imagine saying, “Look, let’s put in place policies where we can own more of the stock market, particularly for workers who are more likely to be affected.”

Beckworth: So, an example of this would be if you have a retirement account with work, you could get more matched based on what you donate. 

Harris: Exactly.

Beckworth: More than 3, 4 percent.

Harris: What the SECURE 2.0 legislation does, is it says for low- and some middle-income workers, if you put in a dollar, you’re going to get 50 cents back on that deposited directly into your 401(k). A really easy thing to do is just boost that up for one-to-one match—

Beckworth: OK.

Harris: —or even higher.

Beckworth: Alright. What are some of the other policy implications of this work?

Harris: We’ve had a long-time a discrepancy between the tax on labor and tax on capital. We may want to equalize that a little bit. I think that one of the justifications—there are a lot of justifications for it, but one of the justifications would be that we want to attract more capital to invest. You can look at how much the tech companies are investing in hyper-scaling and other things. There’s really no shortage of capital right now. We could probably see an equalization between labor and capital tax rates. That’s certainly something we could do.

I think we’re going to have to think more about the safety net, particularly for people who lose their jobs. We have a really messy unemployment insurance system. In some cases, it was even designed to fail. As someone who was a senior official at the Treasury Department during COVID, I can tell you that we struggled a lot with trying to design this program on the run. I know that the Trump administration also struggled with that before that. You saw the Paycheck Protection Program. We just handed out trillions because we had this poorly designed system.

OK, well, now, we have a little bit of advance warning. It’s not going to be like COVID, where it materializes over weeks. We know we could have this labor market shake-up. We should start planning for it and thinking, do we have the right support system for workers if they lose their jobs?

Beckworth: Any other implications, or are those the main ones?

Harris: There are just big social implications. 

Beckworth: Yes.

Harris: University professors and teachers, they don’t really know how to assess students anymore. There are big implications as far as who controls the information. 

Beckworth: Yes.

Harris: There’s this democratic aspect to Google even though you’ve got search algorithms and everything. You’re presented with different information. Some of these LLMs are just giving you a singular answer that is taken as gospel. Do we want to have more transparency around that? There are lots of interesting questions.

Beckworth: And probably new ones will emerge as we get more and more immersed in an AI-driven world.

Harris: Definitely.

Beckworth: I guess, if you had to rank, what is your top proposal? What would be the thing you really want to push? You gave me a menu of options here. Which one do you think you would like, and then, what is the most practical one? For example, I suspect equalizing capital taxation with labor [chuckles] might be a bit of an uphill battle politically. Would creating this matching grant be easier, do you think?

Harris: Yes. I had a paper about this with one of my former deputies, Natasha Sarin, who leads the Yale Budget Lab, where we went ahead and did some simulations as far as ways to increase the Saver’s Match in a revenue-neutral way. This just feels like an obvious answer for policymakers.

Beckworth: Yes.

Harris: Let’s just get low- and middle-income workers more invested in the stock market. We’ve got an infrastructure for doing it. You could have a one-line piece of legislation that just changes the match rate from 50% to 100%. That alone is a win. 

Beckworth: That seems like it’d be a bipartisan issue, right?

Harris: Yes. SECURE 2.0 was a bipartisan bill, so it doesn’t have to be partisan at all.

Beckworth: Yes, well, this is a very promising note to end the show on here. [laughter] Let’s make more Americans involved in America, Inc., have them have a stake in it. Any final parting words you want to share with the audience before we wrap up?

Harris: I think the last thing I’ll say is that when it comes to AI, no one really knows what will happen moving forward. But looking back, I just want to turn down the temperature a little bit—

Beckworth: OK.

Harris: —because I can tell you that looking backwards, this has not been the shake-up that it seems like. You really have to squint to see where there have been job losses. It’s even an open question about whether or not this will lead to job losses versus job gains.

Beckworth: Yes.

Harris: I get the anxiety. I have three daughters. I’m worried for them in the labor market moving forward. They’re all teenagers right now. I even feel the anxiety, but if you’re willing to be analytical about it and look what’s happened in the labor market so far, unless you’re a 24-year-old programmer, there are probably not a lot of reasons to be pessimistic.

Beckworth: On that high note, our guest today has been Ben Harris. Ben, thank you for coming back on the program.

Harris: My pleasure. Thanks for having me.

Beckworth: Macro Musings is produced by the Mercatus Center at George Mason University. Dive deeper into our research at mercatus.org/monetarypolicy. You can subscribe to the show on Apple Podcasts, Spotify, or your favorite podcast app. If you like this podcast, please consider giving us a rating and leaving a review. This helps other thoughtful people like you find the show. Find me on Twitter @DavidBeckworth and follow the show @Macro_Musings.