Moody's Talks - Inside Economics - AI Series: Can AI Fix the Deficit?

Episode Date: September 15, 2026

Ben Harris, Director of the Economic Studies Program at the Brookings Institution, joins the Inside Economics team to discuss his new research on how AI could shape the nation’s fiscal outlook. The bottom line: it’s complicated. AI has the potential to lift economic growth and ease pressure on the budget, but it won’t erase the difficult choices ahead on taxes, spending, and the nation’s long-term debt path.Hosts: Mark Zandi – Chief Economist, Moody’s Analytics, Cris deRitis – Deputy Chief Economist, Moody’s Analytics, and Marisa DiNatale – Senior Director - Head of Global Forecasting, Moody’s AnalyticsFollow Mark Zandi on 'X' and BlueSky @MarkZandi, Cris deRitis on LinkedIn, and Marisa DiNatale on LinkedInGuest: Ben Harris, Director of Economic Studies Program, Brookings InstitutionView our latest articles and research on AI- https://www.economy.com/ai-insight-hub Questions or Comments, please email us at [e-mail]. We would love to hear from you. To stay informed and follow the insights of Moody's Analytics economists, visit Economic View. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Welcome to Inside Economics. I'm Mark Zandi, the chief economist of Moody's Analytics, and I'm joined by one of my trusty coes, Chris DeReedies. Hey, Chris. Hi, Mark. Good to see you. Hey, Mark. Yeah, great to see you. You're good. Good to have you back. It's been... Thank you. Good to be back. Four or five weeks, I think, since we've had you on last. No? It's been two.

Two. Oh, two. It may feel like four to five, but it's only been two. It definitely felt like four to five. We really... I mean, I don't know. We really depend on you. When you're not here, like, we're let's complete chaos. You know, it's like Chris loses control, you know, sharp elbows, you know, all that kind of stuff. So glad to have you back. Deep depression. Yeah.

I'm glad you. You made it through the jobs, the jobs report without me. I talked about it. But we didn't play the game. It was, you know. No listener questions, you know. Yeah. So good to have you back, though.

It's good to be back. Yeah. It's nice to see you guys. And we've got a guest. We're going to go right to the guest. Ben Harris. Hey, Ben. Hey, Mark.

Thanks for having me. Yeah, Ben's the Director of Economic Studies at the Brookings Institution, and you've been on Inside Economics before. Welcome back. Thank you. You came back. Thank you. Yeah, it's really good to have you. And this is a special podcast because it's part of our artificial intelligence series.

We've been having a series of podcast around AI began with David Autor back. I don't know, Chris, what was it, a month ago now? Yeah. I think it was about a month ago. Seems like it. Yeah, talked about the job market. Then we had Daryl Spence of Capital Group talking about markets. That was pretty cool.

Then Michael Gukas, he's the chief economist of Construct Connect, and so spent a fair amount of time on data centers. And Ara Kara Zayen, but he's the lead economist at Ramp. I talked a lot about AI adoption. And Ben, we've got you talking about the fiscal implications of AI. And so thank you for coming on. and really appreciate your coming on like this. Hey, you were just describing to me at Brookings

that it's kind of a weird time to be a think tank. How so? Why do you think, why is it a weird time? Yeah, that's a great question. So I came to Brookings for the first time in the late 90s. It was my first job. I've come back for subsequent times. So this is my fifth time being here,

and I'm now running economic studies. Think tanks have been, I think, in a bit of an evolution over the past few decades. One, there's just more competition. I mean, when I came to Brookings in the 90s, there was not a ton of think tanks. And that was for Tudas place to be. We had, I think, enormous influence with Congress because in many ways, we weren't the only game in town, but there was, you know, much fewer think tanks. But secondly, the information flow has changed a lot. I think where members of Congress and other policymakers go for council has expanded.

And I think, you know, what it means to put out high quality research has changed before, you know, you might have a scholar sit down and write a book over the course of the year. And that book would have influenced over the course of a decade. But the rapidness of the news cycle has, I think, increased demands on think tanks. And also the funding models have changed. And so you've got a lot of foundations, which are kind of our bread and butter when it comes to funding, are now looking less at research organizations and more kind of do tanks. You know, they want to be part of the solution. And we're a research organization and we always will be. So it's a challenging time.

And then also you have an administration, which I think doesn't look at traditional sources of expertise. You know, we've had influence. We're nonpartisan. We've had influence in the Bush administration. We have influenced the Obama administration. But the Trump administration, I think, turns to different sources of knowledge. And so, you know, we do so want to be a resource for the Trump administration, but the doors are not as wide open as maybe they've been for other administrations. Got it, got it.

But you're putting out really high-quality research. And I know you put out a paper a couple months ago now, I think, on AI. and what it might mean for the fiscal situation. A lot of optimism around that, but we'll get back to the paper. I thought we'd begin the conversation. And I should ask, before we even get to the kind of the meat of the matter, do you use AI a lot in your work?

Is that something that's kind of part of your daily workflow? For me, I do. I mean, I don't really do coding anymore. I've got this terrific research assistant. But I use AI. Called Claude? Your assistant is called Claude? Yeah, exactly. I mean, yeah, it's interesting.

I went to a presentation at this great conference I go to in Jackson Hole. And for someone who is a high user of AI, and he had a personal assistant who he named. And his personal assistant was named Samantha. And he referred to her as though she were a real person. I use AI. At Brookings, we use AI a lot. I would say everyone under 30 uses AI, particularly for coding. And I would say roughly, I don't know, half of the scholars,

age 30 and over use it regularly. Got it. And we're having a lot of debate, you know, in our world about how to use AI appropriately, particularly in the context of writing. Do you find that kind of a tricky thing? Because AI can kind of take over if you let it. And I'm not sure that's a good thing or not. Yeah, you can't let it.

And I actually, for the first time, had AI admit to me that it was wrong when I was kind of prepping for this podcast. and it described a certain steepening of the yield curve over a certain time. And I said, it's not right. And it came back to say, yeah, yeah, I thank you for catching that. Yeah, I think it's really important that humans continue to learn how to write. I'm glad that our educational system has not given up on that because writing is how you reason and how you formulate your thoughts. And if we've surrendered that, then we've surrendered everything.

So I use it as a check on my writing, like an editor, but I still want human eyes on it. I see it as a supplement. So, Ben, before we dive in AI in the fiscal outlook, how would you characterize the fiscal situation? I mean, I know it feels dark, but how dark is it really? Well, there's a couple different ways to look in at this. Okay. The first is that you could say, look, I can look at Congressional Budget Office

or CBO projections over the next 30 years, and there's this wide structural gap that is actually growing over time. I think that CBO's projections for the next three decades, we get out to the mid-2050s, they're seeing a 2% gap between non-interest spending and revenues, but then you throw on top another seven percentage points of GDP for interest spending. We've got this like 9% point gap,

which is, I mean, you keep hearing this word, unsustainable thrown around, and that's not well defined, but that is certainly unsustainable. But there's a more positive way at looking at it, which is to say our country has a ton of resources. There was recently a piece put out by another think tank where they looked not at debt to GDP, but debt to wealth. And if you think about sort of national wealth, almost like a household, you know, do we have the resources to pay for the commitments we've made? Well, from that perspective, if we're doing okay, and we're actually kind of in the middle of developed countries as far as debt to wealth. We're around 22 percent. Japan is in the mid-40s. They've got a really high debt-to-wealth ratio,

but Norway is around 5 percent debt-to-wealth. So there is this argument that, yeah, our fiscal outlook looks unsustainable, but effectively the argument is if we want to change our tax code, we could close it. Yeah, so you're saying we have the resources that they're there, if we can figure out, politically how to tap them. That's the argument. Look, I don't, I don't really buy into that. Oh, you don't? You don't know, no, no.

I mean, I was just characterizing another argument. I mean, I am quite worried about our fiscal outlook. I think if you look what's happening in markets right now, there's real cause for concern. And we can go under the hood if you want to. But, you know, as you know, you're seeing a pretty steady runup in yields over the past, you know, half a year. And there's a lot of different potential explanations for that. But I'm worried about that.

I think you look at the data around auctions. You know, we have over 200 auctions a year. Those provide regular data points to go ahead and sort of measure demand. I think you're seeing weekend demand from investors. And then you're starting to see policymakers play games. And so, for example, Secretary Besson's announcement a few weeks ago that they're going to double the buybacks on treasuries. You know, that's surprise markets. I thought it was not a good sign.

And so there's these little causes for concern here and there. It's not catastrophic, but I've been worried about this for a long time. And I'm more worried today than I was 10 years ago. Yeah, I mean, I kind of look at, you mentioned some of the measures I look at. I look at the deficit to GDP. And right now that's 6%. You're saying if you look at the CBO projections out over the 10-year horizon, we get to 9% of GDP. then I look at the primary deficit to GDP,

and that's the deficit X, the interest payments on the debt. And right now that's 3%. And you're saying that the CBO thinks that might get a little bit better in the future, so down to two. But still pretty out of bounds in the context of a full employment economy, which is kind of sort of where we are. Then, of course, debt to GDP, and that's publicly traded debt to GDP's 100%, and the trend lines don't look.

And I've been following, the fiscal outlook for 35 years. I can't remember time when all three of those measures are kind of screaming the same thing. You know, we got a problem. It just feels like, you know, the outlook here pretty, pretty dark if we don't change something. No? No, definitely. I mean, and what's underlying all of this is not just the outlook for 10 years, which continues to deteriorate, deteriorate. I mean, relative to about 10 years ago, about 10 years ago, we thought we'd have about a one percentage point of GDP, primary deficit, and about a 3% spending on interest payments. Well, each of those metrics have gone up by about a percentage GDP.

I still think we have tools to get out of this, but as the stock of debt grows larger and larger and we're spending more and more on interest, the opportunity to pull ourselves out of this diminishes. The tools we need, you need a bazooka instead of a handgun. Yeah. And it just gets harder and harder to go way out. Yeah. I mean, if you look at the CBO's projections, you know, the kind of the budgeters for the federal government by a nonpartisan, excellent group of economists, some of whom probably are at Brookings now that you've been at CBO and times past and vice versa. They, the outlook is, I guess it's relatively saying when, right, the debt to GDP ratio, you know, the deficits are a lot.

are large, the primary deficits are large, the debt to GDP ratio continues to rise 10 years out in the 10-year budget horizon. It's, what, goes from 100% now to 120, 125 under current policy, I should say under current law. And yet the economy continues in that period to grow kind of two percentish, that's GDP growth per annum, which is kind of sort of the growth we have been getting, and they kind of just project that out into the future. Does that feel like a, you know, I mean, does that feel like that's a sustained, given what's what we know and where we're headed and what's going on, does that feel like a reasonable forecast, baseline forecast?

So it's really tough to forecast three decades out, but I want to make two comments about that. The first is that the reason why I brought up the trajectory in Treasury yields over the past several months is because I think it's really, really important to understand what's going on with investor demand for treasuries. And if we've got a 7% rate on the 10-year, you know, that's catastrophe staring us straight in the face. And we're at 4.75 or so right now. And so I think understanding why there seems to be weakening demand for treasuries is really important to understanding whether or not that's a reasonable forecast. And so what

you've seen over the past couple of years is what we refer to as, you know, Mark, and and everyone else here is term premium, well, the term premium is going up. And there's a lot that gets thrown into the term premium, but it basically means that investors are demanding more higher return for treasuries. And that doesn't really show any sign of stopping. And I'm not entirely sure whether or not that's driven by a shift in foreign demand away from treasuries or whether that's driven by increased concern around the Fed response

or potentially whether it's just driven by competition from other types of investment. I mean, there was this great bank note the other day that noted, for the first time, it was looking at all this extra investment in AI. And it said, for the first time, the private sector is crowding out the public sector. And that's a really important statement because usually you'd have the Treasury rate kind of set the baseline for private investment, and then you get a premium on top of that. But this could be going to go in the opposite direction where, you know, people are going to pay what they're going to pay to invest in AI. They're just so optimistic about it. And then Treasury kind of gets the residual. And so if you tell me what the rate is going to be on the tenure of the next 30 years,

then I can tell you whether or not that's a realistic projection. The second thing is we really don't know what AI is going to do to productivity. And that was kind of the crux of the paper that you mentioned. And there's a lot of optimism around here. You will not find that optimism at the CBO. So when CBO looks at productivity over the next several decades, their most recent long-term budget outlook had productivity growth at about one-term. percent average, which is less than we saw from the mid-90s to today.

And so they are not optimistic about AI's potential impact. Yeah, on that first point, I think what's really getting crowded out is non-AI private investment. So Treasury is going to get what the Treasury needs. You know, they're going to pay whatever is required. and of course AI is willing to pay whatever is required because the returns there are seemingly, you know, endless. So that leaves the everyone else kind of stuck, you know, paying a high rate that they can't make pencil out

and you just see less investment in non-AI. So it feels like that's already starting to happen. I mean, if you kind of look at the past year in investment spending, it feels like that's already starting to happen, you know, to some degree. It does feel like that's happening, but I'm a little confused around where the causation is. Is there so much optimism around AI because people are legitimately excited about what this might bring? Or is there so much investment in AI because we're just pretty sanguine about other prospects?

You know, like there is no other alternative. Oh, well, I don't know. My vote for the, it's AI euphoria. That's what I've known for. That's what it feels like to me. I don't know. But so the CBO forecast the baseline, the 2% per annum GDP growth and kind of reasonably stable interest rates. I think the 10-year treasury yield in the 10-year horizon is not much different on average from where it is right now, maybe, you know, 4.5%ish, you know, somewhere in there.

So they're pretty optimistic. You're saying that, you know, that assumes really no significant boost from AI, at least any additional. So it might be adding to business as usual productivity growth that you might call it, but there's no, you know, extra juice. And but that, you know, there is a lot of optimism around that, that, you know, we could see potentially much stronger productivity gains. And if we do, that that might, if not bail us out of our fiscal problems, certainly make them a lot easier. Do you do buy into that kind of perspective? Yeah, I mean, just on the CBO point, you're right. I mean, they lock in the 10-year rate at 4.4% for almost a whole long-term,

budget window. So it's consistent. A, they see pretty low productivity growth, but they also don't see higher investment in AI crowding out treasuries. You know, I don't know what's going to happen with AI and productivity. My best guess, and here's how I look at it, I'm size it relative to the internet expansion we saw in the late 90s, and say, A, will it be bigger or smaller than that, and will it happen faster or slower than that? You can look at work, for example, out of the Yale Budget Lab, where they look at labor market churn, and it looks like it's happening slightly faster than that

over the past four years or so since the introduction of chat GPT. But it's mostly similar to prior technological shock. So the advent and expansion of the personal computer in the 1980s, the expansion of the Internet in the late 90s, it looks pretty similar from a labor market perspective. I think about in terms of the timing of all of this, and I think that we were initially biased because of some of the euphoria,

really unjustified euphoria. One example I give is I had lunch with a tech CEO about a year ago who told me in six months the unemployment rate will be 18%. And so even giving him an extra six months, I mean, he was clearly widely off. You know, I think... Did you really say that? He said, in six months, the unemployment rate would go from four percentage to 18.

He picked the... He actually said 18%. He actually said 18%. And you see these similar kind of these outstanding claims. I mean, there are health experts who say death will become optional. You know, I mean, and I just don't think that that's going to happen. And I think it's going to take longer than some are projecting. I mean, just look at Amazon. I mean, Amazon started up as a retail bookseller, then it evolved into an everything retailer. And then it became a marketplace. And then you saw Amazon Web Services come about about 15 years ago and then became more of a service provider and cloud compute. And now Amazon is poised to become an AI company. But it took at least 25 years for Amazon to become the company that we now know. Was it transformed of? Yeah, of course it was. But not in two years or three years or five years.

years. And so I think it's actually good news for the labor market. What's bad news for the labor market is if we see this really sharp increase in labor replacing productivity that happens too fast for people to adapt. And I just don't think that's going to happen. So let me see if I can pin you down, though. So I mean, the CBO, when they do their forecast, their economic forecasts, and I should say I'm on the economic board of advisor. So there's a group of economists, and you may have been on that board at one time, Ben, as well. But this is a group that provides advice to the CBO with regard to their macroeconomic forecasting and, you know, other issues that come up that require, you know, economic input. And simply put, their forecast for GDP and productivity

largely reflects the consensus. So, you know, as far as you can discern the consensus. So if you look at the consensus view of economists, it, you know, certainly over the next couple, three years, it's 2%. You know, it's just 2% GDP growth. And then, you know, in the longer run, not many as many people do that, but you look, it's kind of sort of like 2%. There's some, you know, more AI optimists that are out there, but generally that's the case. So it feels like what CBO is doing is doing what they always do and they take the consensus forecast, which I guess is it's going to be wrong, but, you know, maybe that's the prudent thing when you're kind of doing the budget forecasting and the kind of political and policy environment that we exist that exist today.

Yeah, I think it's helpful to put, the way I like to look at it to kind of compare growth rates is to look at cumulative growth over a decade. And so in the 1990s, cumulative nominal growth was 72%. In the 2000s, it was 47%. And in the 2010s, it was 42%. So we basically went from cumulative growth of just over 70% to in the low 40s. CBO says that between 2026 and 2035, we're going to grow in the high 40s, a 48% cumulative growth. So they say, okay, it's going to look a lot like the 2000s. But you've got to think what's inherent in that projection, which is that all of these trillions in AI investment are not going to pan out in the ways that we thought.

And that may be true, and it may be the consensus. but I'm increasingly thinking it may look a bit more like the 1990s where we see cumulative growth of 72% rather than in the mid-40s. Oh, okay. So you're, you sound, I mean, obviously there's a boatload of uncertainty, which you, you know, make clear, and there's a wide distribution of possible outcomes. But, you know, push comes to shove,

you would think if things, judging relatives at the CBO's baseline, and by the way, those numbers you were quoted are nominal dollars. Right? There's not real. Nominal dollars. That it's going to more likely be the case that we see even stronger growth, something like what we saw during the Internet period in the late 90s and early 2000s. That's kind of your perspective. Yeah, I think we're going to look like the 1990s, which included the early 90s, which wasn't quite as rapid growth as the later 90s. But I think we're going to look like the 1990s rather than the two decades have followed. Got it, got it. Okay. So if we get that kind of growth, what does that mean for the fiscal situation? then are you much, does that mean we're assuming that policy doesn't adjust and adapt to the better revenues that that implies, does that mean that we should be less nervous about what's going on on the fiscal side or we should just be prudent and be cautious and hope for the best that things turn out better than it that are in our baseline? Yeah, so now you're getting kind of to the meat of the topic of our paper. And so what we want to do is look at projections of growth.

and then also account for the idiosyncratic nature of the current AI shock and say, if we grow like we did in the 1990s, but we account for the realities of AI, what's that going to look like? And so as a kind of baseline, we say, okay, what happens if we grow like the late 1990s? We looked at a lot of different papers and projections. We looked at Goldman projections. We looked at McKinsey projections. on Duran Asamaglu's projections. And so we say, okay, what if we have a nominal growth rate that looks like the late 1990s, but nothing else changes?

It's just pure growth. So we're talking about expansions and tax bases. Well, then it's good news. Then we get down to around zero primary deficits after six or seven years. We've still got about 4% debt to GDP in terms of interest payments. But when you get to zero primary deficits, what you really do is you give your economy and your budget the opportunity to grow your way out of its mess. Effectively, you stop digging.

So that's great news. If we get a late 90s like boom, six or seven years from now, zero primary deficits. My guess is investors would take some solace in our trajectory. You'd see fall in the 10-year, term premium to go down. We're kind of in good shape. Can I just push back for just a little bit? Because we've looked at those studies as well. And we've got our own forecasts.

And if I look at, and I look at the late 1990s, like the 95 to, you know, the Y2K, GDP growth, real GDP growth in that period was 3% brand. Productivity growth was, excuse me, was productivity growth, non-farm business productivity growth in that period was 3%. So we're going from 2% non-farm business productivity growth today to 3% in that period. And that would be consistent with like what Goldman thinks is going to happen here in terms of AI. Asi-Moglu has barely any growth, you know, additional growth from AI. It's maybe a tenth of a percent per annum, maybe two-tenths of a percent per annum.

Yeah, he's at like half a point over 10 years. So it's like basis point, basis point growth. But he has this paper of the economics of AI where he's kind of sizing others' views and talks about some of others' optimism. He is not an AI optimist. I see. So you're saying if I take the consensus of the forecast that for those that are doing these kind of work and we get that kind of growth kind of sort of what we got in the 90s, the 3% non-farm business productivity growth, by the end of the 10-year budget horizon, we're pretty close to zero primary deficit. We're kind of kind of there. Yeah, if you kind of take the consensus, exactly. If you take the consensus to the optimist, but not the consensus. But there's a lot of reasons not to be. The story does not end there.

Okay. And so there are five reasons when we started kind of modeling the different potential impacts of AI why you should temper your optimism. And I'll kind of go through those in turn. Okay, reason number one is increased longevity. And so while I would not describe myself as terribly optimistic about AI, I think it's going to look like the internet boom, I am pretty optimistic about health care delivery. I think we're all going to live a lot longer.

And that is great news for humanity. it is terrible news for the Social Security Trust Fund. And so just to kind of size it out in the projections out of the trust fund, it was a process I used to oversee when I was Assistant Treasury Secretary, they had reductions in age-specific mortality rates of 0.7% per year. We boosted it up to 2.0% per year. So a tripling in the reduction, if that makes sense. Now, historically, Japan coming out of World War II,

they were around 6% in a reduction. So there's historical precedence for this. So it's a pretty steep increase in longevity, but not unprecedented. The second shock that we went ahead and modeled was around shifts in the labor market, particularly reductions in labor force participation. And I think kind of the scenario that's the most applicable here, we had a three percentage point reduction in LFPR. So it kind of looks like the reduction we had during COVID. Now, some people say, look, you're going to have these huge exits from the labor market.

I don't think that's plausible. Three percentage point reduction during COVID felt like a good benchmark. That lets the participation, right, LFPR, the participation. And you're saying three percent over what percentage points over what period? So we just phase it in over four years. And then it stays three percent depressed for the rest of the 10 year. Got it. Okay.

Three percentage points of press. The third shock is related to the capital share. And you've heard a lot about this where national income is increasingly earned by capital rather than being earned by labor. And so we looked at new growth. And so within the CBO baseline, you have about 54% of income being earned by labor. We reduced that to 30%. And the reason why that's relevant for the budget is because, the marginal tax rate on capital is much lower than the marginal tax rate on labor.

So when you shift from labor to capital, you're sacrificing all of this revenue. Can I say, can I get those numbers again? So the labor share is currently what? 54%. And you have it going down to... 30%. Over a four-year period? We, yeah, we look at it over the budget window, and we phase this all in pretty quickly.

But yeah, it's over a four-year period. We face it. Okay. Okay. So the thinking is AI really upends kind of the income and wealth distribution, that the benefits of AI are accruing to capital and not labor. Yeah. And I should say those proportions are new growth. So it's not the labor share doesn't go from 54% down to 30%.

The share of new growth goes from 54% down to 30%. That makes more sense. Got it. Yeah. Yeah. Yeah, I mean, that type of shock would be, I mean, it would look dystopian. That happened overnight to everyone. Got it.

Quick question back to the LFPR. Are you assuming that across the age distribution or just at the tails or just for retirees, older workers? Yeah, we assume that across the age distribution. Okay. So prime age workers as well are reducing. Okay. Exactly. The next, the fourth thing that we measure is when the,

the, sorry, so we've got mortality. We've got the shift in labor share and the consequent impact on tax rates. We have the reduction in LFPR, which means more people on income support programs. So we look at traditional relationships between when people drop out of the labor market and when they claim SNAP, which is nutrition benefits, and when they claim Medicaid. So you've got a lot more people claiming those benefits. The irony in that, I would guess, is unemployment stays low, right? At least measured unemployment stays low. Measureed unemployment stays low, but what we care about is here for the purposes of the fiscal outlook is to draw on these programs.

Yeah, right. Interesting. Yeah. And then so the fifth thing that happens is that we think we may be in AI arms race with our adversary. So particularly with China. And so when we wrote this paper, we said we're going to look at a nominal increase, annual increase in defense spending. And the CBO baseline is 1.9% per year. We boasted up to 2.2% per year, which equates to about an extra $350 billion in spending over the budget window. And at times since we've published this paper, I would probably have doubled that. So you go ahead and you speak with defense experts. It really feels like we're on the cost of a lot of extra spending. If I had to put a point estimate on that right now. I'd say it's about $700 billion of the next 10 years. I mean, this is a wild guess mark. I don't want to, it's like false competition to say that, but it's a really big, it's a really big number. But in our paper.

That's the additional defense bending related to protecting the U.S. from China due to the AI. Yeah, the way I would say it is to winning the AI arms race with our adversaries. So this means like investing in new drone technology. Right. cyber protections for the military. It's just a ton of extra spending, maybe increasing our domestic infrastructure. And then the last point,

which is something we talked about from the outset, was increases in interest rates as Treasury has compete with AI. And we went ahead, we boosted yields across the curve by 35%. So you went from a 4.5% to something like, what, six percent? Yeah, on the 10 years. Yeah.

Yeah, over a 10-year period. Okay. Oh, wow. Okay. Interesting. So here's the punchline. Yeah. Things look much less rosy under the scenario when you account for all these

eneosyncratic nature about the AI shock. Yeah. And there's kind of a rule of thumb here. I'd love to give this a name for a rule of thumb, but I can't. I'm not clever enough to think of it. Did you ask Claude? I should ask Claude. Yeah, ask Clyde.

I'm just saying. Or Shirley or Samantha. But it turns out that the optimism should be tempered by about half. And so, you know, in the baseline, we're at around 6% deficits. We go down to around 2. If we're just going to look at historical growth like we had in the late 1990s, 2% deficits is a share of GDP. But once you account for these five shocks, after 10 years, deficits go down to around 4% of GDP. And the scenario includes all five shocks.

The scenario includes all five shocks. Got it. Right. Now, in general, and we haven't, you know, we ran a bunch of different simulations, but we haven't run everyone personally, everyone separately. I would say in general, however optimistic you're going to be, just cut it in half. So if you think instead of having late 90s like growth and if we're going to see this big boost in productivity. If you think it's going to be slower, it's going to look more like a hybrid between

2000s like growth in late 90s, okay, cut that optimism in half. It feels like you should cut it by more than that. At least, like, for example, if we're getting kind of the growth that you expect, that would be 3% productivity growth, you know, maybe a half a point for labor force that's three and a half plus 2% inflation. That's five and a Let's just round it up to six. You have a 10-year treasury yield, I guess it's six. That's what you're saying. So it's R is equal to G, basically.

So, okay. Yeah, and we're at about 5.6 or 5.8% for normal growth. So, yeah, R is close equal to G. And some of these things actually, like, we went to 20 years, the story may change. So just to put a number on it, if you have those types of mortality reductions that I described, the number of older Americans, so people age 65 and older, goes from about 74 million up to about 76 million. So you have an extra 2 million people drawing on Social Security and Medicare.

But these mortality reductions compound over time, right? So if you're going to go 20 years out, the Delta is going to be much bigger than 2 million. So part of this is informed by the budget window, which is just 10 years. But if we're all living much, much longer, right? Like that plays out over generations. Okay, just to make it concrete, and maybe you don't have the numbers at the top of mind, but I'll just give it a shot. So CBO says the publicly traded debt to GDP ratio is 100% today. I'm rounding, obviously.

I think it goes up to 125, 10 years from now under their assumptions, which are pretty benign, you know, punk kind of growth. The AI is not changing the world. you know, it may just help and maintain the kind of productivity growth we've been getting historically. That's the, and we get like four, as you said, four, a 10-year treasury yield over that period. So if we get your AI growth scenario, where is the debt to GDP ratio, abstracting from those five caveats, which we'll get to in a second, but with the growth rate, where do we get in terms of debt to GDP? Do you know offhand? I don't. I don't think we did.

gross debt calculations, my guess would be around 110 after 10. Okay, 110. So that, okay, one 10. And then you're saying if I throw in the, if we'd go to this, the scenario where these other forces are at work that are countervailing from mortality to labor force participation to the higher interest rates, so forth and so on, defense spending, we're going to get to like 1.15 or something like that. Yeah, 115, 118, that's my guess.

Got it. Okay. Great. Good. Perfect. And is that, so if I really pushed you, and you could say, no, Mark, I'm, you know, I'm not going to answer that question. But if you were sitting there doing the forecast, where would you land? You know, where would that, what debt to GDP ratio, you know, would be consistent with your thinking, your baseline thinking.

Again, with the caveat, boatload of uncertainty, wide distribution, but kind of in the middle of distribution, where do you think we land, you know, 10 years from now? So 10 years from now, well, this is interesting because what happens over the next 10 years is that the Social Security Trust Fund is exhausted in 23. Yeah. Right. Medicare is exhausted in 23. And so this sort of current policy assumption is not reasonable. Right. My guess is that we're going to learn a lot of lessons about investors' appetite for debt over the next couple years. Right. And we have a Fed, I mean, we didn't get into monetary. policy, but I think that this is a less hawkish fit than some people think.

Anyway, I guess, so if you're going to pinpoint me today on a projection, I think, gosh, I mean, probably not as optimistic as I laid out. I don't think we're on the cuss of a late 90s, like boom, but I think we're going to probably recapture, I don't know, 3% debt to GDP before these all deficits to GDP before these other factors come in, and then you cut it. and a half for these other other factors. So if CBO thought we were going to be at six percent deficits of GDP, we might be at four and a half after 10 years. Okay. So it's going to help. Your view is that this is going to be a positive. All net, net, net, there's a lot of cross currents, but net, it's a positive, but it doesn't bail us out that at the end of the day, we still have

some pretty significant fiscal issues we're going to have to deal with. Exactly. So I think it probably buys us one to two percent of GDP. in-depth reduction. Got it, got it. Hey, Marissa, a lot of moving parts there. Where would you put kind of enter in or push back or agree with? Well, Ben, I'm wondering, do you make any assumptions about the equity market here? Because if we're talking about AI playing out kind of like the 90s and the Internet bubble,

do you also assume there's an equity market bust at all around AI? or do you kind of expect things to just keep it on? Yeah, so we had to limit what we could do and couldn't do. We're modeling a ton of different things here. And so the way we modeled expansions and tax bases were we have this new growth in national income and then we distribute it to the corporate sector, to the non-corporate capital sector,

and then to labor because there are different tax rates on each one of those. But then we don't make assumptions about, equity markets. I mean, it's really just captured in the overall national income. We don't, we don't get, you know, quite that discreet. But you're right. And there's a couple of different things which could potentially happen, which I think are likely. I mean, one, the likelihood of a recession in the next 10 years is quite high at some point. Recessions happen. The second thing which could happen is that we do get a bit of a bust in the equity market, in part because

they're going to be winners and losers, the way there were winners and losers in the early 2000s from the dot-com bubble. And that's not modeled. So, you know, know, for those reasons, that's one of the reasons why I think we might be at one to two percent points, deficit reduction instead of four. Yeah, that makes sense. Chris, would you like to push back on anything that Ben said or maybe a little bit more? Chris is too. He's very polite, Ben.

He will, you know, I give him the opportunity to push back every single time and he says, no, I'm not going to push back. But maybe I have a question. Maybe. Ask a few questions. I'm wondering about the demographic assumptions, particularly around immigration and what the labor force looks like here. Are you assuming, right?

Because you could argue that, well, even if the predictive growth is pedestrian, given the headwind of demographic reduction, it's actually pretty good, right? It's actually contributing positively. It's just, you know, we have a significant headwind. But you could also assume that this immigration policy is going to reverse. There'll be more bodies in the future. And that certainly has an impact. So just curious what your assumptions are.

Yeah, so we took for the population assumptions, so this is part of population model. We can tell you how many people of each gender are working and how many people are alive. And so we stick with the immigration assumptions in the Social Security Trustees report, which, if I remember correctly, are pretty optimistic that this massive slowdown in immigration will be temporary. So we revert. Okay. But I mean, you're right. Like, I mean, just from a growth perspective, you know, I think that if we go from around an average of around a million net immigrants, which we've had over the past 20 years, to an average of zero. I mean, the growth implications for that are outstanding. And that's been a bit of a conundrum for the current economy was, you know, why the slowdown immigration hasn't had a bigger impact. And I actually think if you're squinting, you're starting to see some of that. I mean, construction wages are rising faster than others.

Arrigate consumption maybe isn't quite going as fast as we would have thought. And so the explanation to a lot of the questions around why the economy isn't operating the way we thought it would, the answer to that is we've completely stifled net immigration. And so I think that it was inherent in that question was, what if we take the Trump administration's immigration stance and we get, you know, J.E. Vance is the next president and we're sticking at zero net immigration or potentially even worse. what does that mean for growth? And my answer to that is that's a whole different model.

Yeah, I mean, that's a, it could go the other direction, too, though, right? I mean, we get to the other side of the kind of the current thinking around immigration policy. And we come to kind of the perspective that we need a rational immigration. You know, the labor market's going to be tight. We're not going to be able to fill these positions that we need workers to fill. they're non-AI related kind of positions. They're not going to be AIed out. And of course, we still need the best and the brightest

from the rest of the world to come here because that's our secret sauce. That's key to business formation and innovation, technological change. I mean, take a look at the folks that are running these AI companies, at least more broadly. Many of them are immigrants to the country. And we come and get a rational immigration policy.

we get back to that million or even a little bit more, but it's even more targeted to the right kinds of skills in labor. We could actually end up being in a better place, you know. So I think that doesn't feel like that's right now, but that could happen. And in addition, we have two million more people buying things. Yeah, exactly. Right. And so, you know, immigrants just aren't workers. They're also consumers.

And so if I was advising someone who said, look, I just want to grow the economy. Right. One answer is to grow the population. And it's tough to have people, you know, to boost fertility. There aren't great tools for doing that. And, you know, we may have people living longer. I think we will. And that's good news. But immigration's got to be part of that. Yeah. Got it. So those five different caveats that you, that are part of this alternative scenario, how would you rank order them, you know, in terms of their importance or however you want to rank order them? Like, what's the most, the thing that's singly, in your mind, the most important of those five or least important? Well, we did this exercise. Oh, you did? Yeah, mathematically, it's interest rate. It's a 35% increase in the yield curve, across the curve.

And that's just because our stock of debt is so dang high right now. You know, as you mentioned earlier, Mark, 100% debt to GDP. Even you get little small changes in interest rates, it really matters. And then the other four are roughly equal. But I would say the changes in mortality over time. are going to loom larger and larger as the reductions compound. Okay, so here we are, you know, even with AI and adding a lot to growth, it doesn't solve our fiscal problems. We're still left with a fiscal problem.

What do we do, Ben? I mean, you know, how would you counsel addressing this, this point? problem that we face, even in the context of the benefits of AI. So the first is to provide a target. And I think we've just got to get to zero primary deficits to give our economy a chance to grow out of this fiscal hole. That's got to be the target. You're hearing people on Congress coming up with 3% maximums. That's fine. It's close enough. But I would like to have zero primary deficits as kind of the mantra. And so the question is, how do you close the long-term gap by 2% of GDP?

And we've only got a handful of options. So on the revenue side, traditionally, corporate revenue has been around 1.7 to 1.9% of GDP because of the Trump tax cuts, mostly in 2017. And the extension of those, we've seen it fall closer to 1%. So I think there's room to raise that by about half a percentage point of GDP. So that's step one. boosts corporate revenue. So that's like the raise the corporate tax rate. Go back to something closer to what we had before.

Yeah, I mean, sort of split the difference. I think 35% is too high for most of my lifetime is either 34 or 35%. The headline corporate tax rate, 2017 Tax Act took it down to 21%, which was even a sharper reduction than corporate lobbyists were asking for. So something in the mid-to-high 20s, I think, is, is reasonable. And then you can do stuff on the international side

to discourage some of the games that are played. The second thing you can do is change the taxation of non-corporate capital. And so we've had these fairly low rates on capital for a long time, in part because we wanted

to preserve capital because capital is mobile. It's just as mobile as it's been before. But I think that there are there's plenty of savings to be an engine for our economy right now. There are a lot of loopholes in the taxation of capital. One of the biggest ones is that if you hold, as you guys know, as you hold assets until death, all the gains are excused. So there is some room.

Look, I don't think that the capital gains rate should be the same as on the individual side, probably not. But there is room to raise about half a percent of GDP by closing loopholes and maybe boosting rates by a little bit. I think that we have to do something on the spending side. The way that entitlements have been framed, and I get why they're framed this way, was when you reach a certain age,

you get covered with an annuity, a death benefit for your spouse, and pretty good health coverage. But the thing is, like over the last 25 years or so, the life expectancy at age 65 has gone up by about two years. So actually, Social Security and Medicare benefits have increased over the past 25 years. We're promising an extra two years of benefits. And so I would like to see an increase in the normal retirement age, particularly if people are going to be living longer.

But even if not, we've already begun to live longer. That gets you another around half percent of GDP. And the last piece of this, I think, is improved tax compliance. And one mystery that I'll never understand is why a week, we, weekend to IRS has become a partisan issue. Right. And I just don't get why honest taxpayers who are paying their tax bills are more pissed off that we've cut the number of auditors by sweeping levels.

Like, if I'm an honest taxpayer and I am, I'm furious right now by what's going on. And just to digress for one quick second, I think it has implications not just for revenue, but for economic growth. I mean, if people go into cash businesses because it means you don't have to pay taxes, there's real economic efficiency losses there. So anyway, that's the Ben Harris four-part plan, how to get us back to zero power deficits. I like it. A little bit on tax compliance, a little bit on the spending side, a little bit on our taxation of corporate and non-corporate capital. Well, I failed to mention this at the top because I know everyone knows you, but you were a key player on the economics team for President Biden.

And were you in the NEC, Ben? No, I was, so I was his chief economist. Chief economist. For about the last two years, the Obama administration, and then I stuck with him for the first Trump administration. And then after, you know, on inauguration day, I was sworn in at the Treasury. Treasury. Oh, that's right.

Yeah. Incomparable and terrific and wonderful Janet Yellen, where I was assistant Treasury secretary and chief economist at the Treasury Department. Got it, got it. So it feels like you're going to have a crack at this, hopefully, at some point. So I think my wife and daughters would hope not. You basically say goodbye to people at the, you know. Well, the nation would be very happy to see you, you know,

better to engage with your four point plan. It sounds pretty reasonable to me. Thank you. And, you know, Mark, at some point, you know, I can sit down and drop an immigration plan that is sensible and pro-growing. I'm with you. I'm with you. Because, you know, we're going to battle to the death on whether it's taxes or spending.

But if we can get growth from immigration, you know, rational immigration, not three million people across the border, you know, in a year. But, you know, something that we all feel good about. And I'm sure we came pretty close, didn't we back? When was that? 2012? 2013. We got like 2013. Between 67 and 69 votes in the Senate, which is great.

Can you imagine an immigration plan today that gets almost 70 votes? That's the sign. And Rubio was leading the charge. The Secretary of State was leading the charge at the time. So anyway, well, let's hope that you get another crack at this. And thank you so much for coming on. And I really appreciate it. Are you going to run these scenarios on an ongoing basis or is that one and done for you, do you know?

It sounds like something you should continue. Yeah, yeah. We've got like 45 different parameters. I'll send it over to you guys. If you want, you know, go ahead and run your preferred scenario. I'm happy to do that for you. Really? Yeah.

Really? Of course. Oh, okay. Well, we're going to definitely take you up on that offer. Absolutely. All right. I'll send it over. All right.

Very much appreciate that. Hey, Marissa, Chris, anything? Yeah, before we call it a podcast? So no value added tax, that's what I'm hearing. No carbon tax. No financial transactions tax. No tax on tokens. No tax on tokens.

Wow. That's big. All right. Wow. No change to the income tax bracket. We don't model policy changes. So we're just kind of assuming average marginal tax rates are the same, and we look at changes in national income. Got it.

Got it. All right. Well, thanks. Again, Ben, I really appreciate it. And that, I think we're going to call this a podcast. Hopefully you enjoyed it. And we'll talk to you soon. Take care now.