Lunch Money with Paul Krugman and Heather Cox Richardson

Arindrajit Dube on Wages

I often think of labor economics as a role model for the field: a subfield in which theory is disciplined by evidence and (most) researchers are willing to listen to that evidence even when it challenges their preconceptions. And hardly anyone does modern labor economics as well as UMass Amherst’s Arindrajit Dube, who has an excellent new book out. I talked with him about that book and the state of labor more generally. . . . TRANSCRIPT: Paul Krugman in Conversation with Arindrajit Dube (recorded 6/18/26) Paul Krugman: One of the most satisfying parts of economics, which doesn’t get as much attention as it should, is labor economics. It’s obviously important. Most of us work for a living, or at least pretend to work for a living. But also it is a field, a subfield you might say; more scientific than almost anything else in economics, really evidence-based. You’ve had multiple revelations where the data have actually changed the way people, myself included, have thought about stuff. And among the most effective, prominent practitioners of modern labor economics is Arin Dube [https://substack.com/@arindube], who has a new book called The Wage Standard [https://arindube.com/the-wage-standard-book/]. And I thought we’d take a break from all the other stuff going on and talk about Arin’s work [https://arindube.com/published-articles-and-book-chapters/]. So hi, Arin. Arindrajit Dube: Hi Paul, nice to see you. Krugman: Yeah, welcome to my virtual studio. Why don’t you talk just a little bit about The Wage Standard and what you’re trying to do, and then we can get into the broader labor economics issues? Dube: Yeah. So, I wrote a book. Here it is. Krugman: By the way, we mostly don’t do that in economics; we write 5,000-word articles. Dube: Exactly. Paul, of course, you’ve written many amazing books. But economists don’t usually write books. We publish articles. Krugman: That’s right. Dube: And so it was actually a big deal for me to sort of think about, did I want to write a book? And I kind of went for a number of years and I said, like, “Oh, well, I’m not writing this book for other economists as a main audience,” though of course, I’m very happy for other economists to read it, but I wanted to try to have a broader conversation, and I needed to be clear that I wanted to know what I was going to say in that conversation. And so here’s basically the main point of the book. The main argument is that Americans deserve a raise, that most American workers actually could get paid more and should get paid more. And there are really good reasons to think that. You know, the market has not delivered what could be a sustainable but higher wage for those in the bottom and lower part of the income distribution. So that’s basically the core idea. And I try to bring in what we know about the research that I think has really blossomed in the last decade or two decades on a bunch of topics when it comes to understanding the labor market. I was writing this book at the beginning of the pandemic and especially 2021. And it was really interesting because this was one of the more remarkable episodes in the labor market that really highlighted a lot of things that I was actually talking about in the book. Of course, it did it in a very messy way, because there were lots of things happening during that time. But it made for a very interesting process where I felt like I was writing the book and the world was writing itself outside, which was both exciting and challenging. Krugman: Okay. I said that labor economics has been revelatory. When I was not young, but younger, I think most economists circa 1990 would have thought of the labor market as just being a market of supply and demand. And where they crossed determines wages, and there’s nothing much you can do about it. And if you try to change it, you do so at your peril; bad things will happen. And as you say in the book, and in many of your writings [https://arindube.com/published-articles-and-book-chapters/] that I’ve been following on all this stuff, that’s something that really, really changed. You want to talk about what happened? Dube: Yeah. So, one really interesting thing is to think about how wages are set. And we could start with the basic supply and demand story, which basically is that there’s demand for workers of different skills and then there’s supply. And depending on the supply and demand conditions, you’re going to have different wages, a different skill price. And let me be clear, I think there’s a lot of important aspects of that that actually matter, but it’s also incomplete. Because here’s the thing: if the market really worked like the textbook supply and demand story, basically workers of a particular type would just get paid the same—that’s the skill price. But in reality, it turns out companies have a substantial degree of discretion in setting pay. And you can start to see this by just looking across companies hiring similar workers, but choosing to pay someone different. One simple example to start with is FedEx and UPS. Workers may be driving very similar routes delivering similar packages, but it turns out FedEx pays lower than UPS. UPS has maybe 37% of the workers; a few years back, they were paying less than $20 an hour. For FedEx, it was more like 60%. And so, of course, that’s just one example, but you have others. Like, look at Walmart versus Target. It turns out that Walmart tends to be paying somewhat lower than many of its other similar, large retail competitors. And the list goes on. But this is not a new observation. Labor economists who were studying this in the mid-20th century had gone and collected surveys and understood that, you know, factories in the same labor market could be paying different wages. But here’s what was not fully convincing: how do we know that it’s not maybe somewhat of a different skill mix? Maybe these companies are similar, but they’re hiring somewhat different types of workers the pay difference reflects that. So that argument held sway for decades until we had better data. And this is where what you say about labor economics, I think, really is right. And part of that has been our ability to really get much more granular and high-quality data, including administrative data linking pay for virtually most people in the labor market. And you can track them as they go from company to company. So you could say, “Hey, actually, what happens if the same person moves from Walmart to Target? Do you see they’re getting a higher pay?” Because you’re holding their skill set constant there. And so this kind of data and this sort of research design helped establish that actually, no, it turns out there is a substantial amount of variation in pay that comes from companies choosing different types of pay policies. And that’s a big part of the argument in my book, more broadly, that there are choices we have made. You know, if we wanted to go back and look to see what’s happened to productivity and what’s happened to wages since 1980, productivity has grown much more strongly than wages—maybe not as strong as it did in the postwar era, but nonetheless, it grew a lot more than the pay for the typical worker, certainly pay for those at the bottom. And one of the arguments that I make is that this reflects choices made in a variety of places, and that starts from choices at a corporate level, different companies choosing different pay policies, all the way to policies that are being made by state and federal government. But the core part of it is like, why does that make any sense? It doesn’t make much sense to talk about companies choosing pay policies if the market is just your supply and demand. There’s no role for saying, “Are you doing the high-wage strategy or a low-wage strategy?” That’s a nonsensical question in a perfectly competitive market. But it’s an absolutely sensible question to ask when companies have some degree of wage-setting power. You know, economists have a funny word for this, right? Monopsony [https://www.aeaweb.org/articles?id=10.1257/aer.20200678&from=f]. It’s a funny word. But the basic idea is really straightforward. You know, companies are making a choice there. You could go for a higher wage strategy or you could go for a lower wage strategy. Now, if you’re paying lower wages, you are going to have some more people quit and you’re going to have a somewhat harder time recruiting new workers. But the key thing is, it doesn’t mean that if you pay below a hypothetical market wage, everyone bolts, right? So you actually face a meaningful tradeoff of exactly how much more to pay or how much less to pay, and different companies end up choosing different amounts. And this is also where—and this is even more recent, really in the last, you know, 5 to 7 years—we have seen a really big increase in research on the topic of monopsony, so we can really better understand exactly how much wage-setting power companies have. And it just sort of turns out that if a company’s choosing to pay, let’s say, a 10% lower wage, they’re going to have higher quits. Maybe about 14% higher quits. I just finished doing a review for the Journal of Economic Literature, and that’s basically where it sort of lands, and the quit rate is just not super sensitive to wage. So this gives employers a degree of discretion. And they’re going to do a couple of things that are important. First, different companies may choose different strategies. That is what creates these differences across companies. And the way companies have made those choices has really been different in the arc of history. Krugman: Okay. So that’s where actually I came in on this topic, which was a classic paper by Claudia Goldin and Bob Margo [https://www.nber.org/system/files/working_papers/w3817/w3817.pdf]. You know, I grew up in a world very different from the world where you grew up, with much more equal wages than we have now. But it turns out that wasn’t something that gradually evolved. It happened in a few years, basically during the New Deal and World War II: the Great Compression. Dube: Absolutely. Yeah. And so that’s a story that has been told. But I also tell it with sort of a labor market focus. And a key part of that was actually creating a set of collective bargaining institutions, starting with the National Labor Relations Act; we had an upsurge in union organizing. And I highlight some more recent work that has been really careful to try to actually understand the causal effect of that unionization, for example, on the wage structure—work by Henry Farber and coauthors that really documents this very carefully. And it’s not just in the National Labor Relations Act. It’s also during the war. The Roosevelt administration actually helped encourage an increase in unionization. And that had a lasting impact on pay setting. So this is basically where, after the end of the war, we had what is called the Treaty of Detroit, which was the landmark agreement, as coined by Fortune magazine, between United Auto Workers and the big three automakers, which spills over into the nonunion sector and other parts of the economy through this pattern bargaining process. But all of that created something very different than we had in the early 20th century. It basically created a set of mechanisms that helped ensure wages stayed relatively well tethered to overall productivity. And wages, both at the bottom and the middle, stayed tethered to the top. There were lots of issues. I don’t want to romanticize the 1950s or early 60s. But when it came to how wages were determined, it just meant you had broader based prosperity. Krugman: So in the wage structure there are social institutions that set norms and so that’s part of it; the thing is much more sort of a surface on which you can move back and forth based on institutions. That was one of the lessons I took from the Great Compression. And now you’re saying that there’s much more of that. And also that you can get away with it. I would say that if somebody now proposed something like what happened during the New Deal and the war, The Wall Street Journal would be running nonstop, fire-breathing editorials about how this will destroy the economy and lead to mass unemployment. And your point is that it doesn’t, because of the range of discretion that companies have in setting wages. Dube: Exactly. And those range of discussions in some cases evolved and were forged in the fire of union organizing and militancy in the ‘30s and ‘40s, and other times. There are ones that come up in an era where it’s largely nonunion workplaces that are expanding—for example, Walmart in the 1980s—and in an era when there’s very different ideologies about how businesses should behave. So the entire shareholder primacy revolution that sort of happens in the ‘70s and ‘80s, turns out had a real impact on how wages were set. I talk about this in the book. Research by Daron Acemoglu from M.I.T. and coauthors find a really interesting fact. So it turns out that actually, most businesses are not run by people with a business school degree. I actually didn’t know that. Even today, that’s actually the case. But the share that actually have a CEO with a business school degree has been rising quite, quite steadily. So what happened, for example, in the ‘80s or the ‘90s, when a company moved for the first time to a CEO with a MBA? Sometimes it’s because maybe someone retired or even died, you know? It sounds kind of grim, but actually it makes for a good natural experiment where, almost like by random, you introduce a CEO with a MBA for the first time. And what’s really interesting is that it leads to a very clear reduction in pay: about a 6% reduction in pay for workers overall, and about a 9% reduction for blue-collar workers. So the labor share falls by about five percentage points. That’s the amount of money going to workers versus owners. And of course, CEO pay rises. Now you may say, well, maybe that happens, and that’s just like the cost of running the business better, right? MBAs probably raise productivity. Wrong. It has no effect on productivity compared to comparable businesses. So it’s purely a rent transfer, as we say. Meaning, you’re taking money from one group and giving it to the other. In this case, the money is going towards owners of capital and high-income managers, and away from the workers, especially blue-collar workers. Krugman: Wow. I always thought that the Harvard Business School was evil, but I didn’t realize it was quite that evil. So that’s pretty impressive. That’s really a significant impact on sort of the nature of our society that comes from almost an academic doctrine. Dube: Absolutely. This is sort of like ideology. It’s ideology, not skills that is explaining this important change here. And, in fact, this turns out to have played a non-trivial role in the fall in the labor share in the United States, for example. Krugman: That’s a really funny thing for me. Economists are supposed to be hard-headed, but in fact, if you really look at the data, and really do economic science, it says that ideology matters a lot. Dube: That’s right. And that’s one of the most important things. The late economist Alan Krueger once actually told me—well, he told us on Twitter in a conversation with me—that the idea that core theory is falsifiable and testable is a really big idea. And that is exactly right. Because if you start with saying, “Well, I’m pretty sure the labor market works this way,” and then I come and tell you, “Oh, actually, you know, it turns out this MBA CEO comes in and pay falls,” so you’d say, “Well, there’s got to be a really good explanation for that that is consistent with my model.” But it’s certainly not because the model is false, because it can’t be. And that basically highlights, in some ways, the conversations we had about the role of the minimum wage, which is something we could talk about as well. Krugman: I want to come back to wage structure for a second. When I say that labor economics is especially good or virtuous, or in some way special, it’s because there’s really this use of natural experiments where something happens and just looking at it—at least on a couple of major occasions—it has contradicted what most economists believed. And I do want to come back to wage structure, but minimum wage is the classic. It’s an extraordinary story. You could probably tell it better than I can. To some extent it’s where you came in, but it’s definitely where Alan Krueger and David Card came in. So let’s talk about that. Dube: Yeah. So maybe one thing just as a background for listeners: the United States, of course, introduced a minimum wage as part of the Fair Labor Standards Act in the 1930s, and during the ‘40s, ‘50s, ‘60s, and even ‘70s, the minimum wage was updated fairly regularly. You could have a Republican president or a Democratic president, or Congress, but it was generally updated and kept up with sort of like the typical or the median wage and even overall productivity and so on. That all changed in 1980, when Ronald Reagan came into power and he didn’t increase the minimum wage; he refused to, because he thought this was a bad idea. And this was also a time in the early ‘80s when, of course, we had real, still high inflation. So the combination of the fact that the nominal minimum wage just stayed put and there was inflation meant the actual real value of the minimum wage fell a lot. And so that had a really important impact on wage inequality at the bottom. It reduced pay for roughly the bottom 30 to 40% of the workforce. And so we went for basically a decade almost at this time without raising the minimum wage. And we have now had several of these long stretches. The most recent one is particularly long: it’s 17 years since we have actually raised the minimum wage. And so that’s a very dysfunctional way to set policy. But here’s the silver lining. The silver lining of dysfunctional policies is that you have natural experiments. So what happened starting in the ‘80s is that states started to come in and raise their own minimum wage. And so you started to create all of these little natural experiments. And this is really what began this literature—it’s called the new minimum wage literature—which started to look to see, ‘hey, New Jersey raised its minimum wage in 1992, but look, neighboring Pennsylvania did not. Eastern Pennsylvania and New Jersey are not super different; they’re right next to each other. There’s a lot of similarities, maybe sharing similar types of economic shocks and so forth. Why don’t we compare to see what happened?’ And this is exactly what Alan Krueger and David Card did [https://davidcard.berkeley.edu/papers/njmin-aer.pdf]. They went and surveyed fast-food restaurants on both sides of the state border, and then went back a year later and said, “Well, let’s take a look. What happened? Didn’t we actually see a lower number of jobs in New Jersey?” And what they found really shocked the profession. It turns out, not so much. In fact, not really anything we can see. And, you know, this was really kind of an earth-shattering discovery, because it challenged the core model of the labor market: the labor market is supply and demand, that’s it, there’s not much more to it, just like any other market. And this was really hard to square with it. And I think this led to kind of an emergence of a whole literature. And there are also things written that are very critical and, you know, not very polite about Card and Krueger. But, you know, it led to a lot of debate and also follow-up work, which is the way science progresses, if it’s doing the results that they’re replicated— Krugman: Yeah, the results have been replicated now many times, and you’ve done a fair bit of that. Because there are so many states and so much asynchronous minimum wage increases that you get results. And people might say, “Oh, it’s just fast-food workers in New Jersey.” But it turns out that we have now lots and lots of evidence that says, hey, these minimum wage hikes do not actually seem to cost jobs, or at least not significantly. Right? Dube: Yeah. So I think that my sort of contribution to the literature in our 2010 paper could be probably summarized by the word “many.” We see many of these and for many years, not just one short impact. And what we found was very much along the lines of what Card and Krueger had found. And even more recently, we updated that with more data, and we’re continuing to find very similar effects. In fact, just a couple weeks ago, I put out a Substack post [https://substack.com/@arindube/p-201524189] that really sort of leverages, in some ways, an important fact related to what I said—that we’ve not raised the federal minimum wage for 17 years, and that means 20 states have today a $7.25 an hour minimum wage, which economically is sort of equivalent to not having any minimum wage. It’s so low that it barely affects anyone. So we’re running this basically just more than a generation-long experiment where you have about half the country—a little less than half the country—with essentially no minimum wage, while the other half raised it sometimes quite substantially, or comparable to some of our European peer countries. And that creates this very sharp divide. But it also creates a divide that makes it very easy to see what is going on, because you don’t have to do a lot of fancy, you know, econometric statistics to really tell. Just plot, for example, as I do: what’s the restaurant wage in these two groups of states? Well, it turns out there’s a big gap that’s opened up, like maybe an 8 or 9% average earnings gap for restaurant workers. What happened to restaurant employment? It looks pretty much like a flat line. They’ve been growing very similarly. Per capita, restaurant employment has been very similar. And that just makes it very hard to look at that very simple fact and say, “No, I’m pretty sure it’s killing a lot of jobs,” because where is (the data that proves) it? I do a bunch of other things, but this sort of highlights how, for a very long, long stretch of time, we’ve split the country in some ways in half. And by the way, some of these states that have raised the minimum wage have also been more Republican-leaning. A lot of times when the minimum wage is on the ballot, it’s in red and purple states. In fact, this week in Oklahoma for a variety of reasons it didn’t pass, but it has passed in Nebraska, Florida, Arizona, and so on and so forth. So I think this sort of highlights, in some ways, one of the partial successes because we have been able to raise the minimum wage in about half the country. And as we have learned more, I think it has led to policymakers actually experimenting with potentially higher minimum wages. And that has, I think, helped create and raise wages at the bottom, partly offsetting the growth in inequality that had occurred over decades after 1980. Krugman: So I read the Substack post and I noticed that you had some, I would say discreetly acerbic comments for some of the people who refused to believe it. Or maybe it was a later comment of yours. But there have always been some economists who keep on insisting that this cannot be right, either because they believe in Econ 101 and that demand curves slope down, or at least implicitly, a little bit of a political critique because obviously a pro-minimum wage argument or something that seems to say that raising the minimum wage is okay has a kind of political side. But what’s actually striking is how little of that there is—that labor economics makes economics look good in the sense that if you have kind of overwhelming empirical evidence that contradicts people’s preconceptions and maybe even their political slant, people actually mostly go with the evidence. Am I being too idealistic? Dube: I think that’s generally right. I think in general, people have certainly updated their views. It’s not that there’s only a single answer to what does the minimum wage do, regardless of how high it is or something like that; it’s going to differ. And so, there are disagreements like, “Well, where is the turning point?” But that’s part of good science. But to be clear, there will be studies that claim that no, actually the minimum wage always causes job losses. And even just this week, there was one that sort of argued that if you don’t control for population differences, if you just look at the number of jobs, well, the number of jobs in California has grown less than Texas. Most economists, of course, look at what share of people are actually working—that’s the employment rate. But if you simply look at the number of jobs, that actually might suggest that it’s falling. Now, here’s the thing: it has been falling in these minimum-wage-raised states compared to the 20 states that haven’t raised it for four and a half decades. That’s largely driven by college-educated workers, because, of course, we have more college-educated workers moving to the Sunbelt. So, I think this is sort of a silly argument, but it is an argument that has been made. But it goes to show that there will always be studies. But if you look at the body of evidence overall, it suggests that the typical study finds very small employment effects, and especially in studies published in the last ten years, it’s basically around zero. And I think that has had an impact. And I think economists have sort of updated—I would say probably especially younger scholars. Sometimes, you know, as we get older, maybe it becomes harder for some of us to revise our priors, but younger scholars are therefore really important. Krugman: Yeah. I occasionally find people digging up some old quote of mine where I said minimum wages reduce employment, and it’s a 30 or 35-year-old quote, and I get to use the line, “When I see new evidence, I change my mind. What do you do, exactly?” There was a flurry of stuff showing up in my inbox claiming that California raised the minimum wage and it’s a disaster, and the evidence is in. But I guess the evidence actually goes the other way now, right? So what happened in California? Dube: Yeah. So here’s the interesting thing. California established a sector-wide minimum wage for the fast-food workers, higher than the overall minimum wage. So this is a case where this is applying for larger chains with 60 or more locations across the country to have a $20 minimum wage. And at that time, I think the minimum wage was $16 overall in California. So what’s interesting is this is much higher. And it’s also partial coverage, meaning, you know, only part of the low-wage workforce is covered. So you could actually imagine there’d be more theoretical reasons to expect a more negative employment effect, because you can switch—maybe you can relabel workers who are delivery workers as, like, outsourced and so forth, and not covered. So anyway, well, you’ve now had about five studies that have looked at it, including one that I did. And, you know, there are some differences across the studies, but really, it turns out a big part of that is what kind of data is used, in a really surprising way. So there are two kinds of administrative data sources that are really government data accounting based on actual payroll records: the QCEW and the QWI. And I know this is going into the weeds a bit, but it just turns out that one better captures the number of jobs at a point in time, and then the other looks at how many people are in a particular pay period. Now, this increase in wages also raises turnover because these are much better jobs now, so you have less people cycling through the same number of positions. And so there’s one data set that looks at a whole pay period; it seems to find a small reduction in employment. The other looks at a point in time and finds no change. And it turns out this is driven by the fact that these jobs begin so much better: people are not quitting so there’s just a lot lower turnover. But generally speaking, the overall range suggests that the employment effects were quite small—small positive in some cases, small negative depending on exactly how you do it—very large wage effects, and a very sharp reduction in turnover. So even in this very specific and very sharp and high minimum wage increase that serves as an experiment, if you will, it doesn’t show any clear predictions and projections about job losses so far. Krugman: Okay. I want to cycle back just for a couple of minutes to the wage structure issue, where, again, there’s this kind of historical story which says that the United States became relatively egalitarian because of New Deal era and 1940s policies, and then became a lot less equal. It’s funny. I always blame what happened after 1980 on Ronald Reagan, but you’re saying it’s partly the Harvard Business School, but there’s also cross-national comparisons. Talk to me about Sweden and then maybe I’ll weigh in. Dube: Well, I think we’ve both been writing about Europe and both visiting there. And so I was in Sweden for a while and partly talking about this book and also doing some of my research. What’s really interesting is that Sweden, of course, has been historically held up as sort of an egalitarian country, but it’s also gone through quite a bit of reforms in the ‘90s and 2000s, including scaling back partly some of the welfare state. And so I was really curious, like, where are they in terms of inequality? And it turns out that, yeah, if you look at their tax and transfer, they actually redistribute less than they used to. But the starting point, which is how much inequality do you have to begin with from the pay structure, that is still much lower than most other high-income countries. And the United States, of course, is the other extreme. So, just one example: the gap between someone at the 90th percentile and the 10th percentile—that kind of is a good measure of wage inequality—between like the early ‘90s and today, it went maybe from 1.8 in Sweden to 2.2, a little bit of an increase. In the US, starting off much higher to begin with, it went from like 3.7 to 4.8. And it actually increases even more if you look at a broader time horizon. So it’s just a really important thing to understand: like, why is that? And we can go back to, well, is it because the Swedes are just a lot more similarly skilled between each other? Because that would have to be the reason. Or is there something else? It turns out it’s mostly something else, and that has to do with collective bargaining. And this is also a really important aspect of where people don’t fully also appreciate one really interesting and important fact, which is that in the United States, when we ask, “Is your job covered by union contract?” that question is almost the same as asking, “Are you a union member?” And of course, union membership in the US, maybe in the private sector, having something like, you know, 35% back in the ‘50s, is today like 6%. And so barely anyone overall is covered in the private sector by a union contract. But here’s the interesting thing: if you went to France and asked what share of the workforce are union members overall, it’s like 10%. But 98% of jobs are covered by a union contract, right? Because what you have is sectoral bargaining. And this is a key thing which I talk about in the book. Sectoral bargaining was something that the US never really had. We basically had organizing and negotiating between the union and the employer at a company-by-company, sometimes store-by-store or factory-by-factory level, versus in a lot of our peer economies, what happens is workers and their representatives bargain with the employer and their representative at a sectoral level and at a national setting. Krugman: Basically, sectoral level means that instead of getting a wage agreement with XYZ contractors, you got a wage agreement with the whole construction industry. And so even workers who are not members of unions, even workers who work at companies that have hardly any union members get the benefit of the negotiation. And so, Sweden’s an interesting case where they actually have high union membership. Dube: Yeah. And Nordic countries generally, partly because of the way unions help provide some additional benefits, including unemployment benefits—that makes it more rewarding to actually join a union. But their coverage rate is even higher. And in countries like France or Austria, the coverage rates are substantially higher. So as a result, we have seen wage inequality not rise as much in a lot of other countries. And in Sweden, it’s actually been particularly low, and they’ve actually been able to retain it. And so that is a really important contrast. So one of the things that I talk about in the book is that we can’t get to sectoral bargaining at the national level without a substantial change in labor law. And look, the reality is that past attempts at changing and reforming labor law have not fared well. But the good news is that we can actually get to pay standards at the industry or sector level state-by-state. And what’s even more interesting is we actually have started to see some of this already, and this really leans on a model that actually now comes from a different continent: Australia. Australia has basically a national-level setting of wage floors by industries and, within industries, by different types of jobs. And that’s done not through collective bargaining—they have collective bargaining on top of that—but this is basically a sector-wide floor that’s set. And again, Australia has lower wage inequality, substantially lower than the United States. So, I talk about what the U.S. might look like if we had states do something similar. And like I said, I started to write this book in 2021. I actually had put out a survey proposal back in 2019. But in the last five years, we have a number of states that have started to implement some of this. For example, Minnesota has a sector-wide board that has representatives from workers and employers and the government to set pay in the nursing home sector. We have California that has a healthcare-wide minimum wage. Even more recently in the state of Washington we have a childcare sector board that just in the coming months will be issuing a set of wage floors in that sector. So we’re starting to see experimentation like this. And that’s important because if we’re trying to rebuild wages, not just at the very bottom that the minimum wage can really hit, but also those towards the middle, especially in the childcare or healthcare sectors, these kinds of jobs, you can actually raise pay there through these sectoral initiatives. And I’m very excited to see more being done along these lines, especially because, you know, I don’t know what can be done in Washington, DC right now. But we don’t have to necessarily wait around for a better day to come in DC. We can actually start doing some of this now, more or less. Krugman: So, it’s like the minimum wage is where half the states can do a lot on this broader issue of a more equal and better wage structure, even if things are totally stymied in Washington. Dube: That’s right. And that’s one of the nice things about federalism in the U.S., that we do actually experiment at the state level. And in the best cases, some of the better experiments actually get adopted. It could also be that some not-so-great experiments are done and get adopted. But that’s the nature of democracy. Krugman: Yeah. One of the areas where you really did a lot of the research and it was revelatory, but also, in a weird way, something where I found a lot of my sort of lefty friends not willing to believe it, was about wages post-COVID. So, let’s talk about that for a second. What happened? Dube: So, around 2021 and 2022, of course I looked at wages like any labor economist. I started to look around and find something that was puzzling because, as we’ve known for a long time, wages have been rising faster at the top than the middle and the bottom. And this is the growing wage inequality story. But it was looking like wages right after COVID, when we were reopening, a lot of people didn’t have jobs, especially in the hospitality sector—we’d sort of shut down part of the economy. So if in January 2020 someone said, “We are going to shut down some parts of the economy for a while, especially with low-wage workers, and then we’re going to reopen,” it’s like—here’s your quiz. If I could have given my class this question, like, “What do you think? What’s your prediction about what will happen to wages for low-wage workers?” I would have said wages would probably fall due to lower demand. And instead, it looked like wages were rising more at the bottom. And so this is what David Autor—my coauthor on this along with Annie McGrew—and I called The Unexpected Compression [https://www.nber.org/system/files/working_papers/w31010/w31010.pdf], meaning the compression of wages, reducing inequality—which is exactly what happened in the aftermath of the reopening after COVID, and led to a surprising amount of wage growth at the bottom. And it reduced maybe a quarter to a third of the increase in wage inequality that had occurred between 1980 and 2019. And so this was really very, very striking. And we asked, well, why? And the reason is because we had a very tight labor market. There were a lot of job openings chasing workers and, as a result, it increased workers’ leverage. And it’s not just that there was more demand for workers—that’s true—but we also saw people leaving jobs. So we had quits from particularly low-paid jobs. This goes back to the issue of different companies with different pay policies: well, companies that were actually going for a low-wage strategy found it harder to hold on to those workers, and wages actually then rose more there. And this is the increasing of intensification of competition in the labor market that actually really helped boost wages. In many ways, this was like: if we want the market to actually work well for workers, you need the market to be relatively tight. And in writing the book, what I’ve found was that, it just turns out between 1980 and 2019—up to just before the pandemic—there were about seven years of a tight labor market. We used to spend a lot more time with tight labor markets in the postwar era before 1980 than we did since. And this turns out to be another important part of that equation of: what did it take to have broad-based wage growth? Those seven years—if I just, like, snap my fingers and just erase those like some evil genius villain, what would happen? Well, if I went to the top of the pay distribution, it would make very little impact; the average wage growth would fall from 1.1 to 1%. Not much change. At the bottom, it would go from already a small 0.3% average real wage growth to zero. So the entirety of the wage growth at the bottom between 1980 and 2019 happened in a handful of years that was basically close to full employment: the late 1990s and the late 2010s. Under Trump I, those years also saw significant compression. And this is why the post-pandemic period was a really important one. But it’s also very messy because, as we know, this was also a time of a large increase in inflation, a chunk of which was, by the way, global in nature. But nonetheless, people were very reasonably unhappy about it. So it makes for a difficult thing to extract the signal from noise. And this is why in the book, I really highlight also why even these other periods in US history were so important in actually raising wages, highlighting really the critical pillar that full employment plays if we are trying to rebuild the wage standard. Krugman: Okay. What do you see happening now? My comment sections are full of, “Oh, it’s a K-shaped economy —the top is rising, the bottom is falling.” And people really refuse to admit that the compression ever happened. But also there are all these fears about AI. Everybody wants to know what AI is going to do, and nobody can honestly say that they know. But do you have any views on where we’re going right now? Dube: Yeah. So the easiest part of that to answer is just to start with wages. The good news is that much of the compression that we saw has remained. The bad news is that the last year and a half has seen some take-back. Basically we have seen lower wage growth at the very bottom. The particularly bad news is, of course, from this year, when higher inflation has erased, as of now, pretty much the entirety of the real wage growth since Donald Trump took office. And so, that’s really bad. That’s not just at the bottom, but just generally. And so I think wages are not doing great right now and that part is largely just an unforced error of where we are today with having raised inflation, literally having caused a supply shock—inflation purely out of discretion, right? But yeah, the other part—and this is the longer part and harder to say—is what we see not within pay, not wage inequality. Wage inequality has been an important part of inequality overall in the last 50 years. But wealth and the division between capital and labor. And looking into the future, that’s where my worries lie: where are we going? And I guess, the worrisome part of me thinks that, broadly, there are two possible ways that the current AI structure can go. My modal view is probably that I think it’s going to lead to moderate productivity gains. And how well that translates into wage growth partly depends on what we do in our other policy and institutional choices. But I think it can potentially be a source of possible wage growth. The other—and these are two very polar cases—well, this is going to be the singularity. I tend to be skeptical of that view of an artificial general intelligence that really just dramatically transforms the world as we know it. It’s possible—anything is possible—but the other possibility is that actually there’s a bubble and then it bursts, and that leads to a downturn. And that downturn could be harmful. So, there are all of these possibilities and I, of course, don’t know which it might be. But there are risks on both ends where what I do know—and this is what I sort of talk a little bit about in the book—is that, again, it goes back to the word “choices.” I don’t think we need to think about what AI does as something that just happens to us. We can choose to have institutions and a governance structure that can regulate that. You know what’s interesting, going back to Sweden, I was talking to folks in the labor movement there, and they, of course, have contractual language that requires negotiations over technology, and that includes AI. Where that goes is unclear at this time—it’s still early days—but that’s the kind of thing that we need to think about. So imagine having sectoral boards in the health care sector that, among other things, also sort of has regulatory language around how AI is used and how it can affect the workforce. So we need to think creatively, of course at the national level, but even more locally if necessary, about what that governance looks like, and understanding that this is part of the choice that we can make and not simply, you know, take the technology as just a force of nature that we just have to live with. Krugman: Okay. So, choices. We can actually shape our future. Probably won’t, but can. Anyway, thanks so much for talking to me. And I’m sure we’ll want to come back in a couple of years and see how all of this played out. Dube: Sounds great. Get full access to Paul Krugman at paulkrugman.substack.com/subscribe [https://paulkrugman.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4]

Power and Geopolitics After Trump

Transcript Hi everyone. Instead of a regular post today, I’m going to put up a video. There are a number of reasons why I feel like doing that instead of the usual. One of them is that this is a dry run for a talk that I will be giving virtually later today. There’s a conference on the economics of digital transformation taking place in Croatia, although I’ll be doing it remotely. And they have asked me to talk about global power, geoeconomics, and Europe. Those are all themes that I’ve been thinking about quite a lot. And today’s miniature talk is an opportunity to try talking through those themes. And the way I want to structure it is as what has changed, at least in the way that we all now understand the world, since, well, basically since Donald Trump returned to power. That’s an American-centric point of view, if you like, but it’s kind of a natural bracket. And of course, everything really has changed, mostly not for the better, under Trump. And it has, as it turns out, big implications for Europe as well. So let me just try to get into that. Start by talking about the world as it seemed to be at the beginning of 2025. There were, and still are, three great economic superpowers in the world: China, the United States, and the European Union, in that order. If we measure GDP in 2024 at purchasing power parity, which is basically just adjusting for differences in national price levels, you had China with a GDP of something like $37 trillion, the United States with something like 29 trillion and the EU with something like 28 trillion. That last bit may be a bit of a surprise — maybe all of it is a surprise to some people — but yes, in terms of the actual amount of stuff it produces the Chinese economy is now substantially bigger than the US economy. And the European economy is almost the same size as the US economy. If you think that Europe is backward and poor and helplessly dependent, it’s not. It is an economic superpower. And in fact, by this measure, Europe has basically maintained this position of being about comparable to the United States for a long time. This is a whole other topic that I’ve been writing about and will continue to write about in the future. In that world, basically, two things were really kind of striking. One is that the United States seemed to perceive itself as being a dominant power, even though China was bigger and even though Europe was about the same size, and Europe acted as if it seemed to perceive itself as not being in the same league, as being not a superpower at all. All of that may be changing, and events are part of the reason, so let’s talk about the events. Now, the most obvious: the United States just lost a war. Just lost it bigly, as Trump used to say. It’s an astonishing story. We went up against Iran, which was definitely not a major military power or a major economic power, a sort of middle-ranked power, if that, and utterly failed to achieve our war goals. In the process, we inflicted a lot of damage on the world economy and depleted our stocks of high-tech weapons that will take years to replace. Altogether, immense damage was inflicted on Iran, but Iran has clearly emerged stronger. The United States has emerged humiliated. The attempts by Trump and minions to pretend that it was a victory don’t help. They only make the United States look not just humiliated but delusional. So that’s a big deal. It has large implications for US power and influence going forward as well. To explain those implications, it’s helpful to talk about one of the other things that really dramatically changed with Trump coming back into office, which was trade policy. The United States began really seriously trying to throw its weight around. Liberation Day, the tariffs on everybody, basically trying to pressure all of the world into giving us various kinds of concessions. Give us what we want or we won’t let you sell in our market and everybody needs to sell in our market. Okay, what we learned from now well over a year of trade war is that U.S. power in that dimension is substantially less than certainly than Trump appeared to believe it was. And just in general, trade, leverage and trade negotiations, leverage in trade disputes has less to do with market access than a lot of people assumed and more to do with supply chains, with getting stuff that you use in your economy, means of production, not in the sense of capital, but intermediate inputs or just inputs in general. The nation that has more ability to strangle its rivals by cutting off supply chains is the one that has the upper hand. So it turns out, and we had already learned this from the trade stuff, that China with its dominant position in rare earths and some other crucial industrial materials actually had a stronger hand than the United States. Yes, we have a big market, but loss of a market can be offset to some extent by domestic stimulus, domestic support programs. Not having crucial industrial materials is not so easy to make up for. So we learned that the power in international trade disputes in a fundamental sense reflects power over supply, not power over demand, which is something economists have always tried to say. The point of trade is not to sell. The point of trade is to get stuff. You sell as a way to pay for things that you get from other countries. But now we have it demonstrated very obviously in real life. So that in itself meant that we’ve had a blow to the perception of US power. It turns out the US market is not almighty; access to the US market is not anything like as powerful a tool as we thought and Chinese strangleholds over key inputs are much more important. And then of course we’ve seen that even more graphically demonstrated by war with Iran and it turns out that Iran’s ability to disrupt traffic through the Strait of Hormuz was a really huge empowering point, and it was the kind of thing that the United States really didn’t think about, and certainly the Trump administration didn’t think about. And it shows the true rules of global economic power, because largely Iran was able to win this war through economic power rather than strictly military action, the rules of economic power are not what a lot of people thought they were. Who benefits from that? Well, obviously China. What we’ve seen now is that in terms of a global power competition, China has demonstrated that they have substantial power over supply chains. They’ve also demonstrated that they can weather a cutoff of oil pretty well. And global power is a zero-sum game. So the United States, by weakening itself, by showing that we don’t have the ability to impose our will militarily, we don’t even have the ability to keep international shipping routes open, has emerged as just a much less formidable player, which means that China by comparison looks better. Add to that the fact that the United States has been erratic and unreliable. Our current leadership just doesn’t understand that a reputation for doing what you promised, honoring your agreements, is itself a source of power, and we have done an enormous amount to undermine that. Not news to anybody. Trump looks much weaker. America looks much weaker. To a certain extent, China is the beneficiary of all that, at least in terms of power. Now, of course, life is not all about power. And in the end, you don’t run a country to maximize global power. Maybe the Chinese do. I’m not sure about that. But in any case, it’s not a zero-sum game in terms of living. But in terms of power, it is a zero-sum game. And the United States share of that power, however you measure it, is clearly down as a result of the war. Europe is a little bit interesting here. Europe played essentially no role in any of this. Europe wasn’t involved, obviously, in the war. Europe didn’t do very much at all except to suffer. Still, one thing that is kind of important is that Europe — at least to some degree, not really through emergency responses but just through the general way that the Hormuz shock played out — Europe demonstrated or some European countries demonstrated that they can be much more independent of global hydrocarbon resources than they have been. Europe is not a major oil producing area. It has some, but not a lot. It’s not a major gas producing area anymore. It’s essentially a very resource poor economy relative to the size of its GDP, relative to its population. But it is an economy that increasingly relies on renewable energy. And those countries that have gone especially far in relying on renewables weathered this really well. That’s the lesson of Spain’s ability to ride through this with very little rise in electricity costs compared with some other countries. Italy, which has very little in the way of renewables and is very heavily reliant on natural gas for electricity generation, Italy did much worse. But Spain has given an illustration of how the renewable energy revolution — solar plus batteries is what really runs Spain now — has made Europe more independent and can make it more independent still in a world economy where control of natural resources used to be really critical and it’s becoming increasingly less critical. So that’s actually a point in Europe’s favor. That’s one piece Another piece of this is that Europe has always, in my lifetime, literally, and from a bit before my lifetime, Europe has always been far less of a global power player than you would expect given its sheer economic weight. Now that’s partly because Europe doesn’t exist as a political entity. though it’s more of one than it used to be; the common market has gradually turned into something more than that and Europe is able in some important ways to operate as one and is finding ad hoc ways of cooperating more. But it was always in a secondary position very much — or tertiary position given the rise of China — largely because the United States in addition to having a big economy was overwhelmingly the dominant military force. Now until just the other day there was never a question that the United States would use its military force against Europe; but Europe depended on the United States. Europe’s defense, its security, all depended on the United States. Okay, now where are we? The United States is quite simply just less credible as a security guarantor, not just because of crazy stuff where we threaten Denmark over Greenland, and not just because we’re erratic all the time, but because we’ve just demonstrated that our military capability is a lot less than we thought it was. The United States could not batter Iran into doing what it wanted. It could not keep the Strait of Hormuz open. So U.S. military preeminence is a lot less intimidating, also a lot less reassuring if you thought you had America on your good side than it used to be. And on the other hand the prospect that Europe might be able to defend itself, achieve its own security without the United States, looks a lot stronger than it did not very long ago. And that’s not just because of the war in Iran but also because of the war in Ukraine. Now, there are many, many horrifying things that have happened under Trump. One of the ones that is particularly horrifying to some of us is the abandonment of Ukraine, the clear tilt towards siding with Putin in his attempt to destroy a democratic nation. The United States basically stopped giving any aid to Ukraine at all. almost as soon as Trump took office. U.S. aid of all kinds, but especially, of course, military aid, is all gone. But a funny thing has happened. Ukraine is still standing. If anything, the war seems to be tilting in its direction. Now, that reflects partly the fact that Europe did step up. particularly with economic aid: Europe has filled the gap, pretty much, that the United States left so the flow of money to Ukraine continues. But it’s also because war has changed. To the extent that the United States appeared to be essential it wasn’t just the money — we knew that Europe could come up with some money — but it appeared that what would what How could Ukraine defend itself without U.S. weapons? Well, it turns out that in this age of drone warfare that Ukraine can mostly defend itself. Actually, what they can’t really stop is Russian missiles that destroy civilian targets, which is horrifying, but it doesn’t appear to really work in terms of altering the military balance. And Ukraine has developed its own suite of weapons, and quite aside from the fact that Ukraine is hanging in there, this says that one of the sources of perceived US superpower status— super duper power? versus Europe is a mere superpower? — was that, well, we had the weapons, that we had the technology, that even if Europe could come up with the money, they needed U.S. weapons to be effective, as did Ukraine. And if the United States cut off the flow of weapons, what could you do? You really could not stand without all of those sophisticated, high-tech weapons that only the United States knew how to produce. Well, those weapons are kind of looking obsolete right now. Not entirely, but we just saw Iran do a lot of damage with drones that the United States didn’t appear prepared to stop. And the United States, with all of its super-duper weapons, was not able to suppress them. We had the spectacle of million-dollar patriots shooting down $30,000 Shaheds. This is not a good look. And Ukraine has become a major arms producer ,has become in many ways the expert in this new age of drone warfare. The Europeans are picking up some of that, and there’s a lot of new cooperation on weapons with Ukraine. But maybe the most important thing to say is that, well, that special U.S. advantage, because we had the weapons and no one else did, it’s not much of an advantage now that it appears that those weapons are largely obsolete. Not totally, of course. The Ukrainians would really love to get more Patriot missiles to stop some of those Russian missiles that are destroying 11th century churches and so on. But the balance has shifted in a way that means that the United States is not indispensable at any level. We’re not indispensable financially, and we’re not even indispensable militarily. It’s like we have the world’s best cavalry in an age of machine guns. What good does that do? Okay the Chinese presumably have immense capacity. Chinese dominance of manufacturing means that on almost any dimension China is the super super duper power, they’re really way out in front. But there’s much more parity between between Europe and the United States than there was because the United States doesn’t really have economic dominance and we don’t have military dominance anymore. We dominated an age of warfare that now appears to be behind us. So where does Europe stand here? In a rational world, the rise of China and the coordinated, concerted, efforts of the United States and Europe to deal with that rise would be the central story of geopolitics in the year 2026. Unfortunately, things are not rational. And so we have a belligerent, erratic United States with Europe largely on its own. But Europe being on its own is not nearly as impossible to imagine as it used to be. This is a world that has tilted towards China. That’s probably the biggest story. But it is also, in effect, tilted towards Europe because it’s tilted away from us here in the United States. Take care. Get full access to Paul Krugman at paulkrugman.substack.com/subscribe [https://paulkrugman.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4]

Lunch Money with Paul Krugman and Heather Cox Richardson

Thank you Michael Scarmack [https://substack.com/profile/29305883-michael-scarmack], Scotland Explained [https://substack.com/profile/375810392-scotland-explained], Oakbridges.ca [https://substack.com/profile/134432056-oakbridgesca], Kim G [https://substack.com/profile/32893660-kim-g], Cathy Stein [https://substack.com/profile/318670950-cathy-stein], and many others for tuning into my live video with Heather Cox Richardson [https://substack.com/profile/4875576-heather-cox-richardson]! Join me for my next live video in the app. Get full access to Paul Krugman at paulkrugman.substack.com/subscribe [https://paulkrugman.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4]

Talking With Azeem Azhar

I last spoke with Azeem, the proprietor of Exponential View, 18 months ago — ancient history on this subject. So we revisited the state of AI. . TRANSCRIPT: Paul Krugman in Conversation with Azeem Azhar (recorded 6/12/26) Paul Krugman: Hi everyone. Paul Krugman back on my usual schedule of recording interviews. And today I’m talking with Azeem Azhar [https://substack.com/@exponentialview], who I spoke to in January 2025, basically centuries ago in AI time. And with AI on everybody’s mind, I thought it would be good to revisit. I should say Azeem is an independent researcher and founder of Exponential View [https://substack.com/@exponentialview], which is one of the top tech Substacks out there. So hi, welcome to another conversation. Azeem Azhar: Yeah, thank you, Paul. And it has been eighteen months, also known as one and a half centuries in AI time since we spoke. Krugman: Yeah. Let me ask sort of the dumbest question: what is this thing called AI? How does it do what it does? I mean, even skeptics have to admit that it’s really impressive how it’s sort of leapt over all of the previous barriers. How is this happening? Azhar: You know, I think we’re still figuring it out. I think of AI ultimately as a machine that does certain things, and it’s been built by passing first millions, then billions, then tens of billions, hundreds of billions of trillions of words of human output through a neural network to give it some sense of how humans have thought about the world. And because it operates at dimensions well beyond the form of space and time, it seems to be able to find relationships between quite complex concepts. And I think we’ve all had that experience, whether we’ve been using Chat GPT or Claude over the last two or three years, that it seems to be able to recognize things that are quite deeply related that don’t immediately spring to mind. And in the last year and a half or so, the labs have started to train the AI models not just on words in books, but actually on tasks, like, “what is the set of things that you do to write a piece of code that does something?” “What is a set of things you do to use a piece of software in an enterprise?” And they’ve tried to train those models on those particular tasks. Essentially it’s aping what we do, and they use various mathematical tools like reinforcement learning where the model notionally gets a reward. Of course it’s not a reward the way you and I think of it because it’s a machine. Paul Krugman: Right. Azhar: And so that’s what it is. It’s sort of reflecting back, but also I think discovering some really deep relationships in the world that we might not spot, you know, prima facie as humans. Paul Krugman: Brad Delong [https://braddelong.substack.com/p/note-to-self-my-views-this-afternoon] calls it “a vast stew of linear algebra,” which makes some sense to me because I think that Pagerank with Google was the last thing I actually understood. And that’s the eigenvector with the largest eigenvalue. Not that anybody needs to know that, but this is like a million times bigger, right? Azhar: That’s basically it. Yeah. Krugman: But it’s sort of not what artificial intelligence was supposed to be, right? Azhar: No, not at all. I mean, I sometimes go back and look at the TV series of the seventies that I grew up with as a child, and they’ll always have an AI in the spaceship. Space 1999 had an AI you could talk to. And it was very precise, it was very clipped, and it did things and got things right. And there was a sense that you could trust it. But you’d never think to say, as I sometimes do now, you know, “Find me five analogies to help make this point.” I use it as a brainstorming partner, or I give it tracts of my book, the book that I’m writing, and say, you know, “How would Paul Krugman criticize this argument?” And I get suggestions that I then work through by hand? I don’t think we really imagined it would look like that. Krugman: Yeah. In sci-fi it would talk in a monotone and would be relentlessly logical. And in fact these models are unpredictable, they’re sometimes temperamental, they’re not reliable. That’s probably one of the big problems. It’s not at all what we imagined. Azhar: It’s not at all and this point about reliability is so complex. A couple of months back, one of the versions of Anthropic’s Claude came out and I found it so sycophantic that it became unhelpful because I like these things to help me on hard problems and to challenge me. So I switched back to Chat GPT, which has always been a little bit less friendly. And what’s going on there, Paul, is that because we don’t really have a good theory about how to build these. They are developed almost like in a petri dish and nudged in particular directions so they take the shape that we expect them to take. And to use an economist term, they improve non-monotonically with every release. So you’ll see the latest release of an Anthropic model, and there are maybe twenty or thirty public benchmarks that they’re measured against, like how well they summarize text and how well they write software code. And the next version of the model won’t necessarily be better at everything than the previous version, because you lose something in order to get it. And that’s the complexity that the labs are wrestling with. Krugman: Wow. Okay. Second naive question. I don’t think I’m a Luddite. I’ve always been happy to adopt technologies, but maybe I’m incurious on some of these things. I tend to pick up things like mathematical techniques, as needed, because I see something that could be useful. Now, I’m using NotebookLM to extract tables from PDFs, that sort of thing. But what should I be doing? I have friends who are using Claude a lot, but I can’t quite figure out what particularly agentic AI should be doing for me. Azhar: You know, I’m really sympathetic to that because I have the same issue. These tools have been developed by software developers in a really particular part of the world, which is Silicon Valley, where the culture really revolves around the art of the programmer. And so if you have a programmer’s day and you think in coding terms and you have programming workflows, it becomes really obvious what you do with a really advanced AI tool. I do a lot of research, some of it qualitative, some of it quantitative, and in such a world, those workflows don’t match the way that I think through problems. And so the way that I get around this is that I do look at things on Twitter or X as it’s called because people are sharing tips. And I often just ask the models, you know, “What could I do with you given that I’m trying to do this thing? I’m trying to solve this problem.” And it will come back and give me a suggestion. And I have had some success with agents. So I have an agent called R. Mini Arnold. So R is a play Isaac Asimov’s robots. They’re all called R. Arnold is after the good Terminator in Terminator 2, played by Arnold Schwarzenegger, who protects humanity. And R. Mini Arnold is available on my WhatsApp and it’s available on email. Krugman: Okay. Azhar: And it has access to a whole set of resources. It can browse the web, it can access LinkedIn, it can access Twitter, it can look at my library of PDFs of research that I’ve downloaded. And I can throw tasks to it a little bit like I would say a pretty decent but slightly temperamental graduate student. So sometimes it just disappears for six or seven hours at a time. And one of the differences between using an agent like that and using Claude is that R. Mini Arnold has a lot of my life’s context. It knows the music I like, it knows the book I’m working on, it knows the investments I’m making, it knows the essays I’m doing, it’s got the calendar of speeches that I’m about to give. And so when it goes off and does a task, it tries to figure out what in my world is this going to be relevant to and where can I draw threads from? And when it works, it is really sublime and it does feel a little bit like science fiction. But I would say it’s incredibly brittle. I mean there’s breaks every four or five days. A specific example was, I was thinking about the Paul David’s research about why electrification took the time it took. And I wanted to understand what were the determinations of determinants of that thirty-five year lag from Pearl Street generation to, you know, productivity growth. What could the levers be? And so I threw that into R. Mini Arnold and it set up a team of sub agents which had personalities of key economists and was able to go off and do research the way the AIs do, but also research on all the academic papers that I have downloaded in the past. I have access to JSTOR, I’m allowed to download a hundred PDFs a month. It can look at all of those and start to compile an answer in a way that perhaps a Chat GPT can’t. And it knows the context of my book and it knows the context of the essay I wrote. So what then comes back is something a little bit more structured that I can then play with. It’s a marginal improvement on doing this on Chat GPT. I’m sure you could probably figure out how to do it. But it’s quick. I use it on my iPhone. I often do this when I’m walking through the airport and I want to solve this and have this result when I’m sitting on the plane. I’ll fire that query out and it goes back and goes out and sorts that out for me. Krugman: Okay, I guess I’m getting it. But obviously you and I are not typical. The people who are using AI the most are going to be middle managers, business people, etc. And I find myself thinking about what I think of as the homemade pasta problem. Azhar: Mm. Krugman: You’re probably too young for this, but there was a time when I when young and we were using stone axes for computing, and there was a big fad of making your own pasta. Little pasta machines were everywhere. And then at a certain point there was kind of a collective, “What the hell are we doing? You know, store bought pasta is actually better. The Italians don’t do this.” And I have to think that for most tasks, the range of agents can’t be that wide. But why wouldn’t they sell that kind of thing off-the-shelf, as it were? Azhar: Yeah, well I think it’s different for an independent person or a small business or a middle manager in a big company. I would imagine that you will start to see people selling specific agents that solve your marketing problem. If you have a barber’s shop and you’ve got four chairs and maybe 30 people a day coming through. Right now what you do is, you go to ChatGPT and you help it write your collateral for your website. That feels like it’s an interim step to somebody delivering the actual finished product. Why haven’t we seen it? I think we haven’t seen it yet because the terrain is still big enough. Beyond Anthropic and OpenAI, there’s a lot of other companies building agents that are these end-to-end workflows for businesses. They still believe that the prize for them is to build the generic platform that is the tool for all tools. Because if you get that right, you have a much, much bigger business than if you’re just a vertical application. And I think we’re only a year or two into these entrepreneurs building such businesses. I think as some of them succeed and some fail, the ones that are not able to succeed in the general space will start to verticalize, which is what we saw in the advent of the internet. We saw it in software as well. But I think within a big company it’s a different set of questions because you have far fewer degrees of freedom as a marketing manager in a large company than you do if you own your own barbershop. You have all these rules, you have all these other teams you have to interface with, you are held to the priorities and the plans of the company as a whole. And in that instance, I think, it’s much harder to see how you use AI to really change the way you work. Krugman: Yeah, I mean, again we’re talking about ancient history here, but you know, everybody still uses Excel, even though it has always been horrible. But the constraints of corporate life mean that everybody has to use Excel. So that means maybe we’ll see quite a lot less coding a few years down the pike because the people will just be able to purchase whatever it is they need. I don’t know. Azhar: I think there’s a balance. You hear people proselytizing heavily, saying, “I think this technology is going to be impressive and have a significant impact.” But when people pitch this, they forget that there are other actors in the market who might respond to what’s going on. Right now, if you’re a large company, you want to be building as much as you can because what you can buy isn’t right for the market. If you think about Henry Ford putting together the Highland Park plant, he couldn’t go to a supply chain and buy what he needed because nobody was thinking in those terms. I think we are slightly at that stage for large corporates now. Whether we’ll be there in five years, I don’t know. The question we have to consider is where the value will reside: between having your own capabilities to design software for your processes, or handing that over to another company designing software for a hundred businesses like yours. Historically, it has made more sense to hand it over to another company, but the cost curves may have changed sufficiently that you’d rather have the nuance and control to do whatever ‘vibe coding’ becomes in 2030. Krugman: I know with healthcare software, organizations like the VA that built their own have done much better than the ones who tried to buy it from Microsoft. So yeah, it might be a story that makes sense. And actually, since we’re talking about going for the models versus something much more specific, how do you think about the Chinese versus the big US AI firms? Azhar: I’ve just spent eight days in China and I was really fortunate. I got to speak to developers and engineers and management from about a dozen of the Chinese labs. In many cases they hosted us in their offices. The main thing the Chinese companies say about the US firms, is that Claude code is brilliant and Claude is the best model that is out there and they really couldn’t get enough of it. The term is, they’re Claude-pilled. They talk about the constraints on getting access to computational power but just in a way that’s a fact of life. I mean there’s no sort of commentary on it other than it’s hard. They have to figure out how to get around that and how to build a culture of efficiency when you don’t have as much [computational power] and I think they have built a culture of efficiency really, really well. I think it’s going to help them over the longer term. They don’t really talk about competition with US labs the way the US talks about competition with China. But they do see themselves competing with each other. And as you know, that’s what the Chinese economy is. It’s mayors in different cities who almost act as venture capitalists who compete tooth and nail with each other to become the electric vehicle hub or the solar hub or the AI hub of the nation. And what I would say is, the models are really, really capable. They’re very efficient, which is why they’re so cheap to run, which makes them very competitive for a whole range of tasks. But at the margin, it’s instructive to note that everyone was using Claude for coding as opposed to the cheaper Chinese version. Krugman: That’s interesting. So you can imagine a future where a lot of businesses are actually using these less comprehensive but much cheaper models. I think what I’m gathering from you and from other people is that a lot of entrepreneurs in the US are still dreaming of the uber-model that solves all problems but that probably is not going the way it all goes. That in the end we’re gonna end up with a lot of specialized models, but also the uber-models will still have a role. Azhar: Yeah, it never made sense to me that you’d have a single model that would do everything because if the single model is going to solve the Riemann hypothesis, it’s gonna require a lot of resources. And if all you need to do is get it to root a bill to the finance department, it seems a bit silly to ask Einstein to come and do that for you. We’ve had segmentation of markets for a long time and it’s like with airlines. There’s a reason why not every seat on an airline is first class. Some passengers don’t want it, don’t need it, won’t want to pay for it. So I do think that the ecology looks like a whole array of much, much cheaper models that are serving by volume lots of corporate needs, and then having more sophisticated, complex models for the more complex tasks. I think you’re already starting to see this. I don’t see it, by the way, as a shock to the industry. I just think this is what happens as an industry matures. You know, you start with one size fits all, then you start to segment your customer needs and you start to serve them in the most profitable way you possibly can. And that just feels to me like the way that the markets have matured. Krugman: Okay. Let’s move to more macro considerations. People have been worrying about a bubble. A lot of us still remember the nineties quite vividly and think about all of that. But you just aren’t seeing the bubble. You wanna talk about that? Azhar: I remember what it was like in the nineties. I lived through that one and also the housing bubble, which frankly was far, far worse and much more terrifying. I have a really simple mantra here, which is that honest customer revenues tend to be the engine that gets you through this, right? You know, what caused the problems with the US railroads in the 1870s and 1880s? It was that the revenues didn’t materialize because the tracks were being laid in places where there were no towns. That was a problem. The same was true in the dot-com era. My team and I realized last year that it’s very hard to get good quality data on how much was actually being spent by American businesses and consumers on AI. So we’ve spent several months building systems and gathering data to give a deduplicated view of what that number is. And just to give you a sneak preview, is $150 billion per annum, annualized at the end of May 2026, and about 90 billion dollars in the previous 12 months, from May ‘25 to May ‘26. So you can see it’s growing, and those are deduplicated numbers. So if you spend a dollar with OpenAI, and they have to pay Microsoft 60 cents to run the servers, we only count that as a dollar. We don’t count it as, you know, $1.60. It’s a much faster revenue growth rate than mobile or the internet. It’s also a small number because the US is a $32 trillion economy. And I think the thing is that at that level of spend, you are able to roughly cover the depreciation on the enormous capital expenditures that have gone into AI just this past year. But next year or the year after, you have to double your revenues again and again in order to cover these increasing commitments. The thing that often pricks a bubble is when financing starts to get a bit smelly. That was clearly the case in the global financial crisis, where synthetic collateralized obligations were magnifying the risk on subprime mortgages—it was all “smelly finance.” In the dot-com bubble, the dot-coms themselves didn’t really have much smell about them. There was a lot of disbelief, but the telecoms clearly had issues with their internal revenue generation. So the other thing that we look at is how bad, poor, or strong or robust is the funding quality. And that funding quality measure is definitely getting worse. It’s worse now than it was nine months ago. But it doesn’t seem from the numbers to be at the level that it has been historically when these things have imploded. Nor does it seem to be the type of exposure that is really systemic, which is what we saw in the global financial crisis. There are companies like Oracle and Coreweave whose debt looks very risky, and it’s harder and harder for them perhaps to raise money, although Oracle just did. But it doesn’t feel like it’s systemic. You know, when the the global financial crisis popped, no one knew who was in trouble, whereas now you’d be able to isolate it with a single company or a single firm. So at the moment we feel that this is still a demand-led boom, that funding quality has definitely gotten worse, but not so bad that I would say that there is an imminent problem on the horizon. Krugman: So at this point, you’re saying that roughly speaking, final demand for this is about half a percent of GDP. What share are AI-related stocks in market value? It has to be substantially larger than that. Azhar: They’re about forty percent of the S&P 500 right now. Krugman: That’s a huge mismatch. Revenues are not the same as profits, but you’re talking about what is still a relatively small business relative to this immense economy, yet it dominates the financial markets. That would be at least a possible source of alarm. Azhar: Let’s dig into that, because a stock price is a reflection of the expected future value aggregated across the market. Forty percent feels high, but if you look at the measure of earnings, these companies actually have a much higher proportion of earnings and earnings growth. If you look at the US stock market in 1900, after the railway calamities of the mid-to-late 19th century, railroad stocks were sixty percent of the capitalization of the US market. We had worked our way through the busts by that point. There’s a fantastic piece of academic work by an American finance professor named Bessenbinder. He looked at the stock returns of 23,000 US stocks from the 1900s through 2022. Those returns are highly concentrated. About two-thirds are concentrated in roughly 30 companies. Those companies are oil, electricity, or car companies—the general-purpose technologies at the start of the 20th century—or they are the IT companies like Apple and Nvidia. The only exceptions were Walmart, a couple of healthcare businesses like Pfizer, and JP Morgan. Historically, you get this concentration of a number of winners when you have a new general-purpose technology, and that is showing up today. I don’t feel we’re overly concentrated from the perspective of risk, and the price does not feel totally out of whack compared to where we were during the dot-com era. Krugman: One last devil’s advocate question. I keep thinking of the California gold rush. If you had looked at the revenue and spending on gold-rush-related businesses as a whole, it probably looked solid. But the trouble is it wasn’t the gold; it was the picks, shovels, blue jeans, women and whiskey that were the revenue streams. Is that a fair question to ask about AI right now? Azhar: It’s a great question to ask. The question is what determines that $150 billion annualized demand? We see that just under 30% of the S&P 500 have pointed to a generative AI project with a quantifiable result in their earnings calls. They are under pressure to say they do this, so maybe that’s what’s going on. But when I talk to executives, like 30 finance businesses in New York, they all plan to spend more next year, even though not a single one could point to even a 10 basis point improvement in their business from the investments made so far. Krugman: Right. Azhar: When we break out that $90 billion, $60 billion of it is in the US. That’s a lot of money for a single company, but spread across thousands of firms, it’s still at the experimental stage. We should consider whether these executives are learning by doing. The messages I get vary from those having success in the tens of millions who want to reach hundreds of millions, to those finding it harder but persisting. We’re slightly beyond pure picks and shovels, but in Paul David’s work, it took 50% of American companies getting electrified before the productivity rise. We’re a long way from that. Krugman: Headlines flashed about a KPMG study with case studies on the usefulness of AI [https://smallpdf.com/file#s=a35e268b-87cf-4380-b3f2-77b78c9e2274&r=read] that turned out to be AI hallucinations. It’s a wonderful thing. Azhar: It is brilliant. One thing that is quite challenging is that the market has talked a lot about bottlenecks. We saw this with railroads when the US couldn’t make enough steel. There are these bottlenecks, and there’s a lot of emphasis on power and getting electricity to the system. There’s more demand than supply capacity for AI right now, but there’s a question of whether there’s enough capital. We may see another few trillion dollars of intention from tech companies to build infrastructure to 2030, which starts to rival the new issuance of the US Treasury at $2 trillion a year. I’m wondering if this capital constraint is going to be an issue or if the market knows how to clear it. Krugman: Ordinarily, we’d expect to see that in prices. Real interest rates are well off their pre-COVID lows. They are higher now, but still substantially lower than at the peak of the nineties tech boom, when they were around four percent. They’re more like two now. It’s surprising, given the AI boom and massive budget deficits, that rates aren’t even higher. Whether this is an actual constraint, Nvidia is not the US Treasury. They need risk-tolerant capital. The possibility that these firms may not be able to raise enough money is something we need to think about. Azhar: Yeah. On that Nvidia point, I saw that credit default swaps on five-year Nvidia bonds—the cost of insurance against default—are currently lower than US Treasuries. Krugman: I saw that, and it strikes me as completely crazy. If you think the US government is not reliable, you shouldn’t be investing in chip stocks; you should be investing in canned goods for your bomb shelter. But anyway. Azhar: Are you telling me that markets aren’t perfectly rational, Paul? Krugman: Good heavens, I can’t say that; they’d take away my economist card. We’re recording this on SpaceX Day, and I’ve been wondering if there are limited pools of capital for cutting-edge investments. I wonder whether Elon Musk is diverting capital that AI might need. A whole lot of meme money is pouring into SpaceX right now. Is that something I should be thinking about? I mean, he’s got what everybody tells me is a crud AI product in Grok, and yet… Azeem: Musk showed his willingness to adapt; his AI product is now being subsidiarized using his capacity to serve customers like Anthropic. He has an incredible following, but people who have worked with him say his ability to relentlessly focus and optimize sets him apart. His first-principles thinking has brought down the cost of space launches faster than anyone in history. He pushes the rate of learning aggressively. For all the challenges and his mercurial behavior elsewhere, that’s generally a good thing because technology has brought down the cost of inputs significantly. We’re going to be much further ahead in space than we would have been if SpaceX had not been successful. It raises questions about how to govern what used to be a commons, but there is a definite benefit from coming down that learning curve so quickly. Krugman: That’s fair. The one time I looked at Musk’s activities and thought he was really onto something was when I realized he diagnosed that the cost of space launches is really the rocket, not the fuel, and recovering it makes all the difference. Being able to make it happen is a real productivity thing. This is all moving so fast that we don’t have time for the technical productivity issues we had in the past. It’s feeling like a Solow moment where people say, “I see the technology everywhere but in the productivity statistics.” Do you want to talk about that? Azhar: It comes up all the time. I wonder if we need things to happen more quickly than we used to. We aren’t seeing it in the numbers yet. Erik Brynjolfsson at Stanford says he thinks it is showing up in the aggregate numbers. How quickly should we expect a technology like this to show up? At $90 billion a year, that’s not much of US GDP. These are early stages where companies are learning. The first $100 million you might spend on AI is about learning, and we’re in that mistake-making phase. The model Paul David and William Devine talked about in electricity is helpful. In the first phases, you’re retrofitting your capital stock and processes with the new technology. It’s not until you depreciate existing capital and change processes—like Ford did at Highland Park—that you see productivity benefits. To put numbers to that, what would we expect to see in the Ford equivalent of Highland Park in terms of output? Krugman: Yeah. Azhar: I thought we might see what happens to revenues per employee in an AI-native firm. Across high-end companies like McKinsey, it’s about $400,000. For Meta or Google, it’s about two to two and a half million dollars. In AI-native firms like Mercor, that number is closer to seven million dollars per employee. For Anthropic, it’s close to ten million. You can measure the enormous commercial productivity of a single employee if a firm is AI-native. We’re talking about a handful of firms, but we can pick up the shape of what’s possible for the productivity of a single employee. It may be hard, it may take time, but it’s possible. Krugman: What would those numbers look like per dollar of invested capital? One worry is that this is an enormously capital-intensive business that replaces labor. The oil refineries of New Jersey have enormous revenue per employee because there are no workers, just monstrous capital installations. Is that a factor? Azhar: Anthropic has raised in the tens of billions rather than hundreds of billions and had a profitable quarter ahead of schedule. What we don’t know is how much of that capital goes into developing the next model versus monetizing previous generations. Their IPO in the next six to nine months will tell us. Chinese companies are using much less capital to build models that are nearly as good. So I think the harder part of your question is that if every model that OpenAI or Anthropic costs ten times as much to deploy and develop, but lasts only a couple of years before it’s defunct because of competition, what needs to be true for that to be sustainable for more than a year or two? To me, that is a really tricky question as well. Krugman: You’ve cited intermediate measures. Rather than revenue, we look at generated lines of code, which has exploded, versus actual usable applications, which hasn’t. Does that tell us anything? Azhar: Lines of code is an odd measure. We’ve made it much cheaper to write code, so less determined people are writing it now. It’s unsurprising the increase hasn’t been met by proportional productivity. Data suggests we’re getting more high-quality code, but also a lot of useless waste. This isn’t the first time a useful input in the economy generated waste. Think of a barrel of oil: we count the whole value in GDP, but two-thirds is thrown away as waste heat. Only one-third is useful energy. Sloppy lines of code are a similar form of waste we’ve been happy to tolerate in other sectors for a century. Krugman: A weird analogy is when widespread word processing came in. Books started getting longer. It was so easy for authors to turn out hundreds of pages. What might have been a two-volume series became five. Azhar: On that front, we’re at an enlightenment moment. In 18th-century France, the battle was over who gets to write and express their story. Men and women produced remarkable works with quill pens that encapsulated a world. Krugman: Right. Azhar: Is it worse that we allow for more expression? We are worse off when that connects to an algorithmic recommendation system that drives constant slop at us. But we aren’t inevitably worse off because we’re giving access to many more people. In reducing costs of access, we might find amazing people. In breaking down silos of knowledge, we might find connections—perhaps something in battery chemistry that is useful in cardiology. We don’t know because we’ve never been able to get those experts to talk. I look at each opportunity discreetly. Krugman: There is a potential book here: The Upside of Slop. This is an unrecognizable scene from eighteen months ago. Wow. Azhar: We could get ChatGPT to write it. Krugman: I started my career writing papers longhand on yellow legal pads. Amazing change. Azhar: I still write everything with a fountain pen. I’m writing my new book longhand and most of my research is too. The computer is turned off because AI does all the boring stuff like PowerPoint and emails, giving me time to apply my brain to things I want to think about. I’d be happy to continue this conversation in a few months. Thank you for inviting me. Krugman: Thanks so much. Take care. Get full access to Paul Krugman at paulkrugman.substack.com/subscribe [https://paulkrugman.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4]

A Gesture of Contempt

A quick video, thankfully not from Midtown Manhattan Hi there. Paul Krugman with a very quick update. I haven’t done a regular post today because I’m jet-lagged out of my mind, but I just wanted to weigh in on something that will be happening a few minutes after I record this. Which is that a significant piece of Midtown Manhattan — the area surrounding Madison Square Garden — is about to be closed to all pedestrians. This is because of the Knicks game which is in Madison Square Garden. And Donald Trump is attending the Knicks game. Which means that the game entry itself is going to require enormously strict security. People are forbidden from bringing any kind of bag in there. It means that what should be an exciting joyous occasion is going to become quite hellish with long lines and who knows what else. But what really may not be obvious to many people — you might not know if you’re not a New Yorker — is that Madison Square Garden sits on top of Penn Station. That’s a story in itself, but there it is. And Penn Station is the busiest transit hub in America. It is where 600,000 or so people pass through on their way to and from New York by way of the Long Island Railroad and New Jersey Transit. I’ve spent a lot of my life waiting for trains at Penn Station. And it’s completely insane to ruin people’s day like that. You could say, well, what else are you going to do if you’re going to have to provide security for the President of the United States? And the answer is, Why does he have to go to this thing? The simple way to make several hundred thousand people’s lives noticeably better, at least for today, would be to just not go to the damn game. He can watch it on TV. He can go have a cage match in the ripped up White House lawn, if he likes. It’s not such a small thing. It shows a kind of contempt for ordinary people and a kind of self-aggrandizement — I want this so I’m going to make other people’s lives miserable just to indulge my whim — that is part and parcel of everything else that’s going on. It’s a small thing but my god I would actually have had a problem if I went into my office today because my office is not that far from Penn Station. It’s not in the banned zone but it’s going to be nightmares all around. All right, just another message that the people in charge do not care about people like you. Get full access to Paul Krugman at paulkrugman.substack.com/subscribe [https://paulkrugman.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_4]