Business Cycles, Growth, Misallocation, and AI with Pete Klenow 

- This is the Capitalism and Freedom in the 21st Century Podcast, an official podcast of the Hoover Institutions economic policy working group where we talk about economics, markets, and public policy. I'm Jon Hartley, your host Today my guest is Pete Klenow, who's professor of economics at Stanford University in one of the world's leading authorities on economic growth. He's made really fundamental contributions in, you know, areas like misallocation on s, prices on development, accounting in many other areas. It's a real honor to have you on, Pete. I think economic growth is one of the most important areas in economics and it's amazing to be able to talk to really one of the field's leading scholars. Well, thank you so much for having me on. I'm excited to talk with you. Well, before we get into what I think is economic's most important question, what causes economic growth? I wanna get into your background. So you grew up in the Bay Area and you got your PhD from Stanford. I mean, what influenced you to become an economist in the first place? I would say in the first place, I was totally attracted to its relevance to policy. In particular things like trade policy, but also industrial policy, whether we should favor certain sectors and, and kind of like what causes economic growth. I was already as an undergrad at Berkeley, by the way, although I did get my PhD at Stanford at Berkeley, I was a business major and, and I was very drawn to, you know, technological progress generated a lot of growth and, and in fact, even from high school, there was like a high school project where they wanted us to put like by date major political events on one side of the, the chronology on the other side, like other events that are not necessarily politics. And I filled that up with technological innovations. So I was always kind of taken with technological innovations and their importance for economic growth. And so I was motivated even as an undergrad then to think about the, that as some of the most important questions for economists. But I was just very drawn to economics from its policy relevance, including levers that could affect growth rates, but also, you know, all kinds of policies. And so it seemed like a powerful paradigm in terms of its use of data, but also math and statistics. And then it had like a, it had like a public, you know, debate and element to it that I, I found appealing as well. Very good. And yeah, I mean, it, it, it, I, I think the story is similar for me. I mean, you know, this idea that, you know, people in academia could argue so much over, you know, these ideas I always felt sort of attracted me in a sense that, you know, these, it must be interesting. There must be work to be done if there's so much disagreement over all these issues. Yeah, that's a great way to put it. So, so you did your PhD at, at Stanford and you know, at that time, my sense is, you know, the big macro folks around at that time, you know, were in the department, folks like, you know, Bob Hall, folks like John Taylor, folks like even Joe Sticklet who sort of does both micro and macro theory, I think was around too. Who are your big influences then, and who sort of influenced you to, to choose growth and to choose or to choose macro? You do a lot of business cycles, business cycle work as well. Do you know, going into your Stanford eCom PhD program that you definitely wanted to do macro and, and coming out of it, you, you felt the same way? Or, or what, what sort of led you, I guess, in your undergrad years to, or or in your PhD to doing macro? Yeah, I was always drawn to macro 'cause they seem like the most important questions, but of course if macro didn't exist, I was really drawn to economics. So I think I would've done another field of economics, probably trade, maybe public finance. 'cause those are very related to policy questions too. But I was, had no question coming in that I wanted to do macro. I was drawn to Stanford partly by Hall and Taylor and ended up working with both of them. And both of them were appealing partly because they, you know, brought in new techniques, but at the same time, and, you know, were empirical focused at the same time that they were model focused. But also that, you know, that it was already clear that there was these emerging freshwater saltwater and they were kind of outside that they were kind of doing research that appealed to both the freshwater camp and the saltwater camp, which is more relevant for business cycles. And, and my research is, you know, since graduate school has, has shifted away from business cycles to more growth. But like you said, I I have kept, you know, a hand in, in business cycle literature partly motivated by Holland Taylor Makes perfect sense. And yeah, you know, I, I think about that, you know, Robert Lucas saying, you know, once you start thinking about growth, you can't, you know, think about anything else. I know a lot of your early work, you know, certainly related to, you know, business cycle seeking prices. And is it fair to say that you've moved to, to doing more growth over time? Is that, would you say that that's fair to say about your sort of the trajectory of your research career that, so you, you've shifted from business more less business cycle work to, to more, more growth and you know, you think about the, you know, difference in, you know, GP per capita between say, you know, the US and you know, the Sub-Saharan African country. I mean it's, you know, probably, you know, 20 x or something like that versus, you know, a say a a one, you know, bad business cycle might cause you know, a 5% drop in GDP or something like that. Was that sort of the, the thing that, is that fair to say that you've been increasingly more attracted to growth questions as your career's progressed? I think that's true. Yeah. And I think it's, it's partly what exactly what you said, which is, you know, you mentioned Lucas. Lucas also kind of famously calculated that the welfare costs of business cycles were not that large. And there's been a lot of research since then finding that they're larger than his simple calculation might've suggested, you know, things like workers not being able to perfectly insure some people thrown out of work and recessions and other people keeping their jobs. But even taking in that into account, just like you said, the, the gaps in income between say the US and and, and some countries in Sub-Saharan Africa is enormous compared to business cycle fluctuations. So just the, the magnitude of the economic question involved has, you know, that pull kind of pulled me away from, from business cycles over time. And, and then, and then also just the, the rate at which economic growth occurs in any country, not just the level differences between rich and poor, but just the rate at which growth occurs in any given economy, including the United States. Is it slowing down? Why is it slowing down? Is there something the government can and should do about it? Those are really powerful questions To me, just the miracle of compounding means that, you know, the difference between growing, you know, 2% a year, we double incomes per every 35 years versus 1% a year every 70 years. So the stakes are really high for average growth in rich economies too. So, so I wanna talk about just business cycles for, for a quick sec before we get into growth and, and what causes it and, and your work here on business cycles. You, you've made some really seminal contributions on sick prices and there's a, a famous 2004 paper that you wrote in the JPE with Mark Bills and is titled, you know, some evidence on the importance OFS prices. And I think, you know, up until that time, you know, there hadn't been a whole time empirical work done on ONS prices. And my sense is that you sort of went deeper into, you know, the CPI underlying data and found that, you know, siki prices were, were, or prices were more sticky than previously thought. I think ECMO Sisson came after and and sort of found that they were more sticky than, than even what your evidence suggests. But, you know, I I guess at that time, and, and correct me if I'm wrong, sort of, there's these new Keynesian theorists that sort of came about in the eighties and nineties and they sort of just put sticky prices in the model in the, in real business cycle models because, you know, there wasn't really room for policy. And that, that's my understanding from sort of the founders of, of new canteen modeling. And, and then I guess sort of later they sort of, you know, empirical economists like yourself sort of dug into the, you know, these questions about, well, how sick our prices actually and could they be sticky enough to actually contribute to, you know, business cycles. And, and you know, the, I'm curious what your thoughts are on that and, and like what was going on in the lurch at the time when you're writing this paper and, and, and why was it important or why has it been important? Yeah, I actually wanna step back and, and really key on something you said, which is that, you know, people were theorizing about mechanisms in business cycle models, which is, I feel like there's been over last decades micro data revolution in macroeconomics. And you probably started on the labor side, you know, where people were looking at wage data from household surveys and consumption data from household surveys. Bob Hall did some of that. Mark Bills you mentioned earlier, a frequent collaborative of mine did a bunch of that work. So this was a of that because it said, well, first of all, I'm very sympathetic with the view that that monetary policy has real effects, that it doesn't just flow into inflation immediately. That, that it has impacts on things like real interest rates and unemployment rates and, and real investment consumption for, for several years. And the evidence for that, you know, is, is basically based on time series. So it's hard to, hard for the literature to prove definitively, but the consensus was certainly that, that, that it had real effects. And so that's one of the reasons we, that the literature wanted to move away from the real business cycle models where the business cycles were efficient and prices were perfectly flexible. And there have been studies using micro data, but for select products or firms like, you know, magazine newsstand prices or LL bean catalog prices or saltine prices at, in the grocery store. So what we, my work with Mark Bills did is they were fortunate enough to get access to the micro data, the CPI, the consumer price index, which covers, you know, we use the non housing portion, but that's still 70%. So just imagine what kind of coverage that was compared to the existing papers that were, that were much narrower. And the, and there were surveys like I mentioned earlier, where they ask firms how often they change their prices, but there's no substitute for looking at how much they actually do change their prices by looking at prices. And so Mark and I were able to document for a much broader set of goods, basically all goods and services outside housing, 70% of out of consumer spending, how often prices change for new cars versus used cars versus close versus grocery store items. So it, like you said, it then provided some discipline to theorizing about why monetary policy is powerful and how monetary policy should optimally be set. Because if those frictions, if those lack of pricing flexibility are the key frictions that mean that monetary policy can, can move the economy in real terms, there are also a reason why we might need them to do that. That the market might itself not produce the efficient equilibrium because prices, wages aren't perfectly flexible. So the economy can have things like under utilization of labor where people are unemployed who shouldn't really be unemployed. It's just a market imperfection that creates that. So that's why documenting the extent to which that's true, the nature of it is quite important for our understanding of business cycles and what monetary and fiscal policy should do about business cycles. Yeah, you know, I, I'm sort of, I've become a bit anti sticky prices recently, like in, in the sense that one, the reason why they did sticky pricess instead of sticky wages, if we go back to Kane's in the general theory, he actually talked I think a lot more about sticky wages than he did about sticky prices. He didn't really talk, I'm told I'm not because I've been told that, but yeah. And, and you know, you sort of have to read it and it, it's, it's not an easy book to read. But, but what I think is, is important is that the reason why, you know, macro theorists chose sticky prices instead of sticky wages in their models in the late eighties and nineties was that models with sticky prices are easier to solve than, than sticky wages than than models with sticky wages. And so, so one, I think what what's interesting now is I sort of have a, a paper sort of documenting some of these facts that like since the time that you've written this, you know, your your famous 2004 paper, I think it's fair to say that the prices have become less sticky. Retail prices have become less sticky. And part of this is, you know, the growth of e-commerce, you e-commerce prices, you know, adjust, you know, more frequently, we now, which in the past couple years had the advent of electronic shelf labels, you know, Walmart a very large, almost I think half of Walmart stores have, will have electronic shelf labels by the end of the year. But increasingly, you know, stores and retailers are, are using electronic shelf labels and they can adjust prices more quickly with, with those as well. So, you know, it is, you know, we've now in the US gone from around like almost 0% e-commerce at the beginning of like the 21st century, around 2000 to I think around like 25% of all retail goods or so at least 20% are, are all e-commerce in China. I think it's like 50%. But I mean, is it fair to say that, you know, maybe 60 wages or informational rigidities might might be more important than, than price rigidities? I mean, than, I mean it's, it'd be interesting to, you know, if if it's the case that you know, wage if prices become much more flexible or almost perfectly flexible with online prices and it's not perfectly flexible 'cause there's still some sort of labor decisions involved with changing these prices. But if we still got recessions and prices, you know, were much more flexible. I I guess it might sort of dismiss I guess some of the importance of, of, of, of sticky prices for macro models. If that's, that's true. I don't know if you, you if, if you spend Much About this One, one quick thing about whether prices have become more flexible is that it's, I think one study for example found that it's true within product categories that they become more flexible, just as you've said. And some of this evidence isn't as recent as some of the trends you, you were mentioning, you were saying, you know, that that Walmart becoming, you know, having half of their products be, you know, the, the, the price is electronic label on the, on the shelf. That might be more recent than the study I'm about to refer to. But the study about to refer to found that there was an offsetting effect from shifting toward services that services have become more important and they're stickier. So that's a force making prices more sticky in the economy, just a compositional change in the, in the components of spending that's tended to work against the, within sectors becoming more flexible. So that's relevant, but again, that might not incorporate the most recent evidence that you're referring to. And then the other thing I was gonna say is I've always had this prior that wages were more sticky than prices. That that kind of stylized view that you change price once a year, that's kind of fallen away. That the evidence is that prices change more frequently than once a year. There's still this active debate about how you think about temporary price discounts. I'm mostly persuaded that they're not used very much as a form of price flexibility. But there's, you know, one interesting study about u the UK finding that the frequency of sales went up in the COVID recession in the UK worked by Alexi Kristoff is at the Bank of Canada, the co-author of mine. So, so I don't think it's completely, you know, unrelated that temporary price discounts are purely for price discrimination, have nothing to do with business cycles, but I think that's largely true. And so that, so, but, but that means the prices are are more sticky than just like the posted price that includes the temporary price discounts. When you filter those out, you get some stickiness be somewhere between six months and a year. But for wages they seem to change once a year or sometimes even less frequently than that. And there's been work about whether bonuses are, are a form of flexibility, but it looks like they're largely a cyclical, they're largely for performance reasons and not like very sensitive to the business cycle. Somewhat surprising 'cause you'd think the profitability of the firm would be very cyclical and therefore the bonuses would be very cyclical. But it doesn't appear to be really a genuine form of wage flexibility. So my sense is that wages are stickier than prices, which also matters for things like our wage change. The leading indicator for inflation, should you be looking at wage changes, expecting them to change first and prices to change. Second might be the other way around it might be prices adjust faster and then wages catch up. But again, the part of the literature has grown in understanding not only getting micro data better and better on how much prices wages are changing along with prices, but also just like thinking about the interactions between those two. 'cause the firm's wages haven't changed, then they have less incentive to change their prices. So they, the wage stickiness feeds into price stickiness. Makes perfect sense that, that's a fascinating way, way to think about, I, I wanna just shift into economic growth 'cause you know, this is so important, you know, why are countries rich and poor, you know, the differences in, you know, living standard between you advanced economies versus, you know, developing frontier ones that, you know, it's so stark. I wanna just like, you know, walk through, you know, growth theory and, and you know, get into, you know, misallocation, you know, and I guess a quick, a very short theory or a very short history of, of growth theory, you know, sort of starts with, you know, mouth this, you know, is there arithmetic growth or, or geometric growth and, and are these, are there Malthusian trapps? And I think there, there's evidence that this was sort of true before the, you know, 1,617 hundreds. And then, you know, what happened was sort of modern growth came along and, and we started seeing, you know, geometric rate in, in a very serious, meaningful way, you know, starting in the 16, 17 hundreds in the UK and, and, and, and other countries, you know, thereafter. So, you know, one theory is, you know, there's, you know, when you have, you know, Bob Solo comes around and you have, you know, the solar growth growth model and you have you sort of capital labor and obviously, you know, technology, you know, your TFP and, and then you sort of have the sort of successors of that. You have endogenous growth, you know, this idea that you know, that that ideas, you know, flow between countries and that at some level, you know, the, your stock of ideas or your a your technologies in part depend on, on, you know, the number of people that you have and, and with more people you'll get more, you'll get more, you know, ideas and, and more growth. But then it turns out that, you know, at some level, you know, we've had ideas spreading for long periods of time now, but for, you know, clearly we still have mass, you know, massive living differentials. And I think since endogenous growth theory, you know, a lot of theorists have tried to, and, and growth accounts have tried to explain why is it that some countries are still rich and poor, even though, you know, the, the great ideas coming outta Stanford today are still, you know, accessible, say over the internet in in places like, you know, Africa, which are, you know, still very poor. So here's where we get in, in a lot of the past sort of 20 years, I think in, in growth is really focused on like in, you know, various explanations of, you know, largely political economy related, you've got, you know, institutions, you've got culture, you've got geography, I, I'd like, you know, institutions to some degree or, you know, different ways of putting that, you know, maybe economic freedom. But you, you have a, a really amazing, you know, take on on this. And first I wanna highlight some of your work on development accounting and sort of trying to expl one, I want you to sort of explain what development accounting really is and why that's helpful to understanding this issue of, you know, what's actually causing growth. And then I wanna really talk about misallocation, which is I think your sort of really semi contribution and, and I would argue, you know, a Nobel Prize winning won or, or, or potentially Nobel Prize winning won, you know, what exactly misallocation means and, and, and how this sort of misallocation framework explains why countries are so rich and poor, even though, you know, ideas can, can move around the world in the way that endogenous growth theory says, says it should be, you know, causing growth. But somehow empirically we don't find this. I'm, I'm just curious, so development, accounting, walk us through your contributions there. Okay, so we're very motivated by some calculations. We, meaning me and Andreas Rodriguez, Claire and some of the work and Mark Bills again, and some of the other work. And the, we were very motivated by this seminal paper by Manky Greg Manky, David Romer and David Weill. And they just had this beautiful calculation where they said, we can explain roughly, you know, 80 to a hundred percent of income differences using just differences in physical capital, meaning like equipment and structures plus human capital. Some measure, some proxy for human capital that was based on schooling attainment, in particular secondary schooling attainment. And they said, look, this makes perfect sense because ideas technology, if it's disembodied, it can flow if it's embodied in equipment that can flow that's tradable and can flow across countries. So the thing that really can't flow is people, and if people in a certain country lack human capital, that's gonna lower their, their income. So they, that doesn't explain why poor countries might have less physical capital, but it would explain why they, why they're poor, despite the fact that ideas could flow. So they, you know, kind of a simplistic version of what they said was that ideas and technology are the same everywhere and that's generating growth that's common everywhere. But that all these differences in the levels of development are due, due to differences in mostly human capital. So part of our reaction to that was just, you know, how exactly they used schooling attainment pri, secondary schooling attainment is a proxy. We're like, well what about primary schools and what, you know, some of these countries, people are only going to primary schools or a large fraction of the population depending on what year you're looking at. So that seemed like an omission. So we want, and of course, you know, high school and college, you know, all of the, all three levels. And so we took all those into account, but then the key thing we did is say, rather than just assume human capital is proportional to how much secondary school you get, instead, there was a large literature in labor economics that tried to say what was one year of educational attainment worth in terms of higher wages. And if one year of education is worth about 10% higher wages, that suggests if a whole country got one year more of schooling, average wages and average incomes in the country might grow about 10%. So putting that discipline on it turned out to mean a much smaller number, much smaller contribution for human capital and a much bigger kind of residual that might reflect, you know, other stuff like technology differences or misallocation of, of resources, which we're gonna talk about in a bit. So it opened up the door that that maybe half of income differences between rich and poor countries might reflect something other than physical and human capital. Although having said that, some of the research since we wrote that has said, Hey, that we'd looked at quantity of schooling, let's try to incorporate quality of schooling. So they get a bigger human capital number from that. They also, this is work, a lot of it's done by David Lakos and Todd Schulman, but other people are involved in the work. Nancy Chen, Ben Ma, Tomaso pio. And one of the things they're incorporating is that people don't just learn in school, they also learn on the job. And there might be learning a lot more on the job in rich countries and in developing countries. That one I think is a great insight and great evidence they brought to bear. But I'd also like to see more, more fleshing this out because like how much you learn in the job could also be a function of the technology. Like if the technology's advanced, you got a lot more stuff to learn. If it's simplistic, you can learn it in much faster time. So it's separating out learning on the job from technology is not so obvious to me. But anyway, that's, that's kind of my view on like what development accounting is, what our current thinking is and the relative importance. But let me stress why this matters because one reaction more applied micro types or even, you know, average humans, not just applied micro economist, but any, any average person could say, well what does it matter? You know, what fraction of income differences do are due to technology versus things like education versus experience on the job? What does that matter? What does it, what is this useful for? And I think the, the answer is that it's helps us prioritize more research, including the applied micro studies that look at individual interventions related to agricultural technology versus health versus education versus technology more generally. So how much research effort we should put into those different areas should be a function of how important those proximate factors are for income differences across countries. To me it's the same as as saying we wanna know what people die from and that helps prioritize what we should be doing, you know, you know, making, doing research for health improvements on Absolutely. I mean, and it, it's so, so critical and so important. I mean, you know, I think back to the, you know, the thank room wallpaper and I mean sure, you know, you could, we could argue, you know, it's like we don't quite have the kind of rigor of, you know, causal, causal inference maybe today. But the idea that you can sort of, you know, understand like, well, you know, a country has this many people, you know, this is what the output is, you know, per person, you know, the, you know, country has this much in the stock of sort of machines, machinery, you know, and, and what, you know, what's sort of the what, what's left, what's missing there, you know, the TFP is, you know, the sort of the residual side of, you know, this is the a or you know, and I think one economists famously said, it's sort of a measure of our ignorance that we don't really fully understand what, you know. Yeah. EFP really is. And I think that's a great way of, of putting it. But I think just even in an attempt, like, you know, that paper and success of papers like yours trying to understand, you know, or human capital, you know, education, you know, what are education levels across country, trying, trying to understand across countries, you know, how, you know, what are factor allocations and how does it line up with, with, you know, output. It's, it's really amazing. I think this, you know, it's a nice segue into, you know, talking about misallocation. So, you know, misallocation and manufacturing TFP in China and India as I you're most side of paper. And, and this is where the, the whole shift in the field towards saying misallocation comes, it's really a pathbreaking paper for, for those that aren't aware, you know, the, the idea, you know, and you please explain it better than I'm sure you can explain it far better than I can, you know, I think my understanding is the idea is that, you know, output is lost because, you know, factors aren't being efficiently allocated across firms. And you study China and India and, and, and firms in both these countries and, and essentially argue that, you know, if resources were, you know, allocated across, you know, factors as efficiently as, as they are in the US then you know, your TFP and, and your growth rates would be maybe like, you know, 40% higher in China or 50% higher in India. Obviously there's assumptions, you know, about what the production function is for firms in these countries, but I want you to explain it 'cause I'm sure you can explain it better, better than I can. I know there's a lot about, you know, small firms and their productivity in, in these countries, but really, you know, it's pathbreaking work by yourself and Chiang, she and, and really want you to explain it in, in your words and we can Okay. Talk a little bit about what might be the sources of misallocation and other things like that. Yeah, let, let me say one background thing. Before going into how we thought about misallocation, what motivated us to look at the background thing is that, you know, if you said 50% of differences in income were not due to physical and human capital there, that other 50% is a measure of ignorance. We want to figure out what what's causing that. And one challenge to a technology story is, like I mentioned, you can import the equipment, you can, you can have foreign firms if they, if they know how to use the technology, they can operate in your country. Now there's barriers to all of those things. There's, there's tariffs that, that restrict equipment imports, there's restrictions on FDI, sometimes it's legal restrictions, sometimes it's just the cost of adapting to a different environment. But then there's also potentially complementarity between human capital and technology. Like your willingness to bring in technology might be a function of whether the workers have the skills to use that technology. So you could have this coordination failure where people aren't acquiring skill because there's not a lot of demand for the skill in the local economy. 'cause the technology is not very skill intensive and that there's not an incentive to introduce a lot of skill intensive technology because skill is relatively scarce in a developing country. So anyway, so that's kind of a broad thing about the role of technology and in, in development accounting. But still people have pointed out, like David Weill and other people pointed out, if you, if 50% of income differences were due to technology, then some countries would have to be centuries behind on average. Now you can look at individual, you know, subsistence farms and say they're a century behind or something, but like for the whole economy, it just, you know, casual evidence would say they're not, they're not using technologies from a century ago or, or older. So that kind of left a challenge of like what else could contribute to these differences in income across countries that aren't trivial directly to average human capital and physical capital or technology being used by workers. And there was actually a really nice paper by Diego Ian and Richard Rogerson that did just a calculation, an illustrative calculation that said, imagine there were three types of firms, big efficient firms, medium sized firms, and then inefficient small firms. And they're small because they're inefficient and imagine you tax the big firms and subsidize the small firms. And they had in mind lots of different policies or market failures that might have this flavor. And they calculated that this simple calculation like a a 20% tax on the big big and a 30% subsidy in the small could lower aggregate TFP 20%, 25%, 30%. And I was stunned by this. I was like, really? I, we thought kind of misallocation across firms or even across sectors was relatively small on the order of 1% of GDP loss that you would get from that. So their calculations stunned me. So that made me say, well let's look in the data ang, Tisha, you mentioned the collaborator on this. We said, let's look in the data and see if there's, there's evidence consistent with a lot of misallocation of resources. And we were also mentioned, you, you you mentioned small firms. We were also motivated by the fact that firms were much smaller in India than the United States. We was like, maybe that's part of why India's relatively poor. There's too many small unproductive firms and not enough of these world class larger firms. And the evidence certainly goes that way, that there's a lot of small unproductive firms in India compared to the US but there's still the, the issue of whether those firms are the optimal size, like maybe there's no misallocation, there's just a bunch of workers who lacked the skill to work at bigger firms, but they're efficiently allocated. So it's really still just a human capital question. So where Chang and I came in was like, well, does it look like when they hire a worker of the same skill, does it look like they get less output out of those workers even controlling for skill at smaller firms than bigger firms? Or just more generally, you know, do some firms get a lot more output per per unit of input than other firms? And if so, that could generate lower aggregate output. 'cause you're not efficiently allocating inputs across firms. If the return to, to the same worker is high at a big firm and low at a small firm, then by just reallocating that worker, you don't change inputs at all in the whole economy. The worker is still there before and still there after the reallocation, but aggregate output goes up. And so we were motivated by the rest and rogerson calculation to say, does it look like there's, as if there's taxes on big firms or barriers, big firms face and subsidies on small firms or advantages small firms get in terms of less regulation, less taxes and so on That or, or it could be market failures like, like maybe the, the big firms are starved for credit. You might think big firms are, are well financed and have no problem getting credit, but not the necessarily the fast growing ones. So some of the moderate sized ones might be, you know, have great opportunities to invest, but not a great way to get external financing so that they, so that they actually are starved for capital and, and don't and have a high marginal return to capital, but they can't get access to the capital. Whereas the small firms might not have a lot of great opportunities. So they might actually have good access to capital relative to the opportunities they have. So it could be market failures, like the financial system doesn't perfectly allocate capital. And maybe that's a bigger deal in places like India and China. China you have things like state owned banks that are deliberately steering the allocation of capital. Now that allocation could be efficient. They could be like have this great x-ray vision into what the, the highest rate of return is and they're better than the private sector of the private banks at lending money. But the evidence doesn't support that. Its supports that they tilt resources towards state owned enterprises that might not be using it more efficiently. They might have political motives for some of the allocation of capital that they pressure state owned banks to do. And we were just pointing out that that has efficiency costs. 'cause if you're gonna funnel capital to a lower return investment project, then it's then that's gonna lower output per per unit of capital. And the same thing with human capital. If you encourage state owned enterprises to hire a bunch of workers for political reasons, that's gonna lower the value of what output that they produce relative to the private firms that are more productive, that are starved if you will have to compete too much for those workers because those, those state owned enterprises receive subsidies from the government to continue to employ lots of workers. So that would drag down aggregate productivity in places like China where there's big role for state owned banks and state owned enterprises more generally. And then in India where you do still have a big role for state owned banks, arguably misal allocating capital, the prevalence of low productivity small firms in India opens the door to the idea that that misallocation that they, they have too many resources, too much capital and too much labor is that these unproductive small firms and if it were freed up, they could produce more output with the same workers of the same skill level. That's really the open question. And so what Chang and I calculated using this micro data on manufacturing firms in India and compared to the US was that there was much more dispersion in revenue per unit of input. You can think of it as profit rates across firms that suggested that the small firms especially may have been operating in, in, in persisting because of tax and regulatory benefits or market failures. Maybe the big firms have higher revenue per unit of input 'cause they have higher markups price markups or markdowns. So it could be market power and market failures or it could be government interventions and some of which by the way the go, some of the government interventions may be to reduce inequality. But we were pointing out they, they appeared to have or at least consistent with those having major impact on the ability to even just produce a lot of output per year of input. So something holding back India and China in an important way could be misallocation by state owned banks, tax and regulatory policies that favor small and formal firms against bigger formal firms. And also things like state owned enterprises in China that get special subsidies and so persist and operate with too much capital and too much labor. They'll be better reallocated to the private sector that's more efficient with it. Is it possible that, you know, things like regulation? You know, I think about about India. I have a, a study or a paper that looks across countries and and run. We ran a bunch of surveys across countries trying to figure out what exactly the occupational licensing prevalence was across countries. And we found out that India was actually one of the most licensed countries. About 40% of its over 40% of its labor force as a license compared to say, you know, 20, around 25% or so for the US and developed countries. A lot of that I think has to do with the, you know, so-called licensing RA in India. I'm just curious, you know, what some of these, you know, frictions may be, you know, behind misallocation, you know, why is it that you know that there isn't, you know, a better allocation of of factors, you know, is it, is it possible that labor market regulation, you know, zoning other, you know, is it possible that government policies are, are causing misallocation? I'm, I'm curious, how do you think about sort of the causes of misallocation? I think for sure government policies are contributing to it. So some that I mentioned directly but are state owned enterprises, which are prevalent in China, much less important in India, but still not trivial. Then there's state owned banks that are steering the allocation of physical capital for non-market, not rate of return reasons, which again, could be benevolent and could be wonderful, but the evidence that that exists mostly suggests that it's for political reasons that they, they misallocate it and endure the output consequences of that for maybe for political reasons. But other ways in which the government is involved is, like I mentioned, that, that it's hard for them to enforce taxes and they choose not maybe to enforce some taxes and regulations on small firms. They exempt them from it and they apply things like you mentioned the functioning in the labor market long before our study. Richard Rogerson, again, who I mentioned before, but also Hugo Oppenheim had an important paper describing what firing restrictions do to the allocation of labor across firms. 'cause obviously we're kind of sympathetic, you know, it's tough to be thrown out of work, to be fired at your job, but knowing that the firm, the firm knowing that it has to pay a big severance package when it lays off a worker or it has to get government permission in the case of, of India in certain years in certain sectors discourages 'em from hiring people. And to begin with, because one of the things we've work learned from the work by John Halter Ringer and others that have used census data in the US and other countries, there's just a tremendous amount of churn. The market is constantly reallocating inputs across firms. And so if you throw sand in those gears, then it's just not gonna reallocate as efficiently. And so a well-meaning policy, which is to kind of shield workers for some of this churning and disruption can have efficiency effects because it means firms are that are productive aren't even bothering to hire their workers as a precaution against having to fire them later. And that means they're, they stay behind at another firm that might also already have them and keep them just to avoid the firing costs. So it's, it's generating misallocation this this policy to put limits on firm's ability to, to fire workers. So there's other government policies you could talk about. There's differences in availability, let's say India, differences in availability of electricity. Some people are getting it for free, some people are paying full price and then water. There's water rights and certain agricultural sectors where there's water rights and there's not a, a well-functioning market to reallocate those. Imagine you're a very efficient farmer, you would like to expand, but you can't because you don't have the permits for the water or land. If the land markets, if the government doesn't enforce land contracts very well or puts restrictions on ability to sell land or to, or to even rent out your land in certain sectors, then that's gonna cause the misallocation of labor, I'm sorry, land across farms, across manufacturing firms. You know, when you talk to these manufacturing firms that are very efficient and productive and profitable and say, why don't you expand? They often say things like, well, we'd have to get the permits for acquiring land or we'd have to hire more workers. And that's difficult to do in this environment where there's firing restrictions and you know, or, or access to electricity. So all of these government policies could be contributing to it. And you can step back and more broadly say things like lack of infrastructure, lack of, you know, like if the roads are too crowded or there's, or interstate taxes for that matter, that might discourage trade across states within India, but also underdeveloped infrastructure, which is less of an issue in China, that is in India. That could mean that efficient firms don't expand as much because they can't sell far away, they can't sell the far away markets because it's too costly to ship long distances because of the state of the infrastructure. So government decisions about the quality and quantity of infrastructure can also affect misallocation. And then there's, you, you ask what what causes, there's also market failures, though. It's not just government steering capital, but the fact that the private market might not, you know, if they know there's, it's, it's hard to observe the quality of firms in a certain sector. They're going to naturally wanna lend less to that sector E but the the return might actually be high. They just can't tell which firms in that sector are the best ones to allocate to. So that, that ends up under providing capital to that sector. And there's potentially a role for the government to improve on that. So there's lots of potential market failures. I'm, I personally think there's lots of pricing power out there related to my sticky price research is that firms are setting their prices and keeping them stable for long periods of time. That's not consistent with a flex. You don't see that in commodity markets. You don't see that in well organized markets that prices are stuck at some level. So I think that's a byproduct of market power that prices are sticky and price stickiness to some extent is pervasive. And same thing with wage stickiness. So I, I think there's labor market power and I think there's product market power that's pervasive and, and it looks like it might play an important role in dragging down productivity because again, in the United States, there might be market plenty of market power and plenty of monopsony power, but it seems to be less dispersed across firms. So there's evidence that we have that there's the differences across firms are not as stark and that's what's leading to misallocation. If all the firms are raising their markup themselves, that has distributional implications that might increase the profit share and reduce the share of income going to labor. Similarly with monopsony power, if the firms are getting more and the workers less, that has really important distribution implications. But the efficiency of the allocation of workers across firms is related to whether some firm has more ity power than other, or some firm has more monopoly power than other, then the firms with a lot of monopsony and, and, and monopoly power are producing too little and hiring two fee workers with too little capital. So that's what leads to misallocation. Got it. So I, I guess I wanna talk about just one li you know, potential limitation to, you know, development accounting misallocation in general. And that is, you know, what, what comms would call mis specification. You know, a lot of this, you know, measuring misallocation or development accounting in general assumes that we understand or know what the, the pro production function is, you know, an aggregate production function for a country or you know, with a production function for a firm looks like. And you know, production function for those that aren't aware kind of assumes that, you know, output is a, you know, a function of, you know, various factors like, you know, capital or labor or you know, human capital or, or, or land and, and so forth. And that there's also some like technology augmenting, you know, thing as well, we, you know, usually use a or TFP technological pack of progress, but you know, we're, we're often, you know, economists are, are making assumptions, but what that, you know, function is, and, and, and misallocation and measurements of it really, and all development accounting really rests pretty heavily on this. I'm curious, you know, you know, is that a challenge? I mean, how, how do you understand, you know, what the production function really is in real life? Yeah, it's definitely a challenge. So as you say, if we wanna know whether firms using too little labor, say relative to capital, like mixed distortions, lap capital versus labor versus intermediate goods they buy from other firms. Not enough to say some firms use more capital relative to labor than other firms because that cap firm might be using more capital intensive technology. There might be nothing inefficient about that. I remember Dave Donaldson, who's a great economist at, at MIT kind of throwing down the gauntlet to me and saying, how can you say there's misallocation? I mean, we don't say people in Japan eat more rice than people in the United States, therefore they're eating too much or we are eating too little. There could be just differences in preferences. So similarly, firms just could have different technologies, just like in consumers have different preferences. We don't say you're consuming too little or too much necessarily. We, you know, there's obviously health considerations we could bring in, but in information. But, but he was basically saying, how do you know this isn't all efficient? And I have one kind of broad response and then one more detailed response. The broad response is there is that that matters for these mixed distortions, how much capital relative to other inputs. But a lot of the dispersion in, in a lot of the misallocation looks like it's due to differences in revenue over total costs. So if you can choose your mix of costs, how much is intermediates, how much is capital and labor completely freely and completely efficiently, and there's no other distortion in the economy, then your revenue per unit of input should be the same. It's like the market should equate the rate of return. And so profit rates should be the same. So the fact that revenue per unit of input costs is the same, we don't need to do anything to, to weight them, but just look at your choices. So that's kind of like assuming your choices of your input bundle, your input mix is completely efficient. So even if we assume that we still see lots of evidence for misallocation, in fact the bulk of it seems to be, well, it looks like a scale distortion rather than a mixed distortion that some firms have too little capital labor and intermediates and other firms have too much capital, labor and intermediates. And the more detailed comment to kind of push back a little bit on whether we need to know whether there's heterogeneity in, in, in production technology that could rationalize the input choices firms make. One thing that some studies have found is that there's difference in prices. I mentioned difference in prices for electricity and water in India, but there's also, you know, California for that matter, right? There's a nice paper by Han Na and co-authors about misallocation of water, which looks like it's especially problematic in India and California. And part of that you can just see in the fact that they're, you could just say, oh no, these are just technology differences, that's why they use a lot of water. But they say, look, they have price differences, they get cheap access to water relative to say households or relative to some other firms. I don't wanna make enemies among almond growers in the, in California right at this moment. But the point is, if we see that some firms get preferential access to inputs at cheaper prices, then that's not just, hey, they have a different technology, they face the same prices and made different choices so it's all efficient. Instead it's like, oh, the market isn't equating the cost of this resource that should be the same across these buyers. It's either price discriminating or they have preferential access for some other reason. So that's evidence consistent with input misallocation that doesn't require us to know exactly what the production structure of a firm is. Asin that, that's a really a great point. I wanna just sort of segue into sort of talking about development as a field. I feel like recently there's been a big debate on Twitter and elsewhere about micro development versus macro development. Okay. And I feel like you're a, a a, you know, big representative of macro development. And interestingly, I think over the past, you know, I think fair to say, over the past 20, 25 years, micro development has been very, very dominant. You know, randomized controlled trials, you know, the, the rise of those, you know, several Nobel prizes being given, you know, estro Duflo, Apogee Banerjee and Michael Kramer, you know, the, the them really being the big pioneers of using RCTs and development in various countries, you know, like Kenya and elsewhere. But recently there was some, I, I think a bit of blowback on Twitter that, you know, that really, you know, it's wonderful, you know, that RCTs can get, you know, very good precise answers, but often these are, you know, for smaller sort of questions versus, you know, in, in macro development you're trying to really tackle big questions, you know, why are countries rich and poor and, and trying to get a, you know, a very big understanding of why that's the case. But obviously this comes with a lot less precision sometimes, you know, not as great identification. I'm curious, you know, what is the case for macro development in, in your eyes as a field? Well, it might surprise you by, by defending micro development at the beginning of my answer, which is that I view them as compliments. So for example, some of our evidence is consistent with misallocation, but without knowing exactly what the sources of that misallocation are, there's no clear policy implication in terms of getting rid of particular government regulation or, or, or, you know, in trying to encourage the market to be more competitive without knowing what the distortions are, we can't, we can't actually do that. And so applied micro people tend to study a narrow question that enables 'em to get at particular individual policies and how they're contributing to misallocation. So, and that's the way they're complimentary, that the macro evidence sets priorities in terms of what, what the big inefficiencies are, or the, the big undersupply is in the market is of human capital. Is it physical capital? Is it the allocation of resources? The micro research can compliment that by saying individual policies and what they would, what they would do. Another analogy I like, I'm trying in this one, let me see what you think. The first time I've tried this is you could say, look, you know, I'm not, I'm not a a life expectancy a demographer or, but just a health researcher can tell you decompose, you know, what, how, what's the most important thing for a life expectancy? But imagine nutrition's really important and, and exercise, but that we have a hard time quantifying exactly which exercise, exactly how much there's research that comes out. But it's is, and same thing with food, it's inherently limited because often we don't have random variation and often, you know, the effects are long term and it's gonna take a while to see them, and we're just not tracking enough people for long enough. So macro development kind of has that problem, which is it's working on something that arguably is incredibly important. You know, it led to hundreds of millions of people being lifted out of poverty of these government. I think these government policy reforms in places like China and India, I think that's true. But at the same time, I can't tell them exactly which policies did what. I can tell some estimate of what, what moving away from state owned enterprises did. But there's lots of details in there that I can't address that applied micro people could better address. So I guess I would say, you know, macro is working on these big things like, like nutrition that are hard to get answers on, but might be really important for people's health, you know, quality of life, life expectancy. But at the same time, you still want the people, the dermatologists, the, the people dealing with some narrower issue that's still important to people's quality of life. Still worth investing resources into figuring out are people, you know, should people be, should vaccines be free? Should bed nets be free? Are bed nets effective? Are they cost effective? Should, should the government subsidize kids to, to stay in school, parents to keep their kids in school or to take them to doctor's visits? All of those seem really use worthwhile. All those are really worthwhile. So I'm a huge fan of banerjee to flow creamer and the entire body of work on RCTs and micro development. I just don't see it as, as either or. I see macro development as being really important in its own right. And, and them being complimentary subfield. That's a great point. You know, that how, how they can be complimentary. My last sort of question for you and you know, p is, you know, really talk about ai, you know, I think it's very much, you know, technology is very intimately related to economic growth and it's widely considered to be, you know, the sort of leading driver of, you know, economic growth. Obviously, you know, political economy and o other factors can, you know, cause technology to be less easily adopted or, or less easily used. I'm just curious what your general thoughts are on, on AI and economic growth and, and you know, what some, what we should be thinking about it and, and how growth economists like yourself are thinking about it right now. Okay. I think obviously incredibly topical, incredibly important. So I'm glad you brought it up and it's a great subject to, to end on. So I think, you know, one of the challenges we face in thinking about AI is that, you know, it might be different than other technologies we've seen in the past. We can think about robots, we can think about software IT in general and changes that that's brought to, you know, that might make, may have made bigger firms bigger and so on. So I think one of the challenges with AI is just that it might be different. And so the things we can carefully study is what happens in the past and they may not a perfect guide to what's gonna happen with ai. So that's one comment reaction. Another very broad comment is that, and this I think is getting less common, is that there was a a lot of discussion about whether growth was slowing down and whether that was inevitable. They were just kind of running out of ideas. And ideas are getting harder and harder and more expensive to find, like every doubling of semiconductor microprocessor capacity requires more and more billions of of research. And eventually we're gonna hit some sort of wall or, or we only have so many people to work on these things. And so progress has to slow down technologically just to come diminishing returns. So I think at the same time, some people were saying that, other people were saying, I'm worried technology's gonna take all the jobs away, including ai. And I'm like AI taking jobs from people. And I said, those aren't consistent because if, if we get a lot out of the work done using AI that we used to get done during with people, then income per per productivity will soar. Income per unit of labor input will soar. Now of course there's major distributional questions that we should be really concerned about. Who's gonna get the income from that? We could talk about that at length, but I think you want to know what I think about the growth potential. And I would say that too much of the work has been focused for my assessment. Too much of the work has been focused on what tasks in just the general job market will AI do and what will that, how that'll affect employment and wages and different occupations. Obviously that's incredibly important and I can see why people are focusing on that. But for the long run growth question, I think a critical thing is whether AI can make research more productive and, you know, I'm using it, everybody I know is using it and, but we don't have a, a magnitude yet of how, how much it's gonna contribute. You know, I would love to see things like more therapeutic drugs being discovered with the help of ai. I mean, I think that could also affect our view of it. If we start seeing concrete benefits that affect everybody. If AI helps come up with treatments for Alzheimer's, that would be huge, right? Or if it mitigates, you know, customized drugs to attack certain types of cancers, that would just be huge for human welfare. And my general reaction would be if it can do that, the growth rate can accelerate and we should figure out how to, how to redistribute the, the income from it to make sure most people benefit from it. So I'm kind of a, a a a c cautious optimist because I see, you know, the people who work on it say you, you know, if you think it's impressive now you think what it's done now, the growth rate it's shown, it's still coming. So I don't know enough about it to know whether it's gonna hit a wall, but everybody who works on it seems to be saying, of course, those people are also selling shares in their stock. So you have to be careful, but everybody who's working on it is claiming this is gonna lead to, to rapid productivity growth. Of course, I have to mention the work by my colleague here at Stanford and the business school, Chad Jones, and some of it's with Chris ti, who's also in the business school here, and they're really emphasizing bottlenecks. And Ben Jones at, at northwestern Kellogg is also emphasizing this. There's a subset of activities that we really need, right? That then those will become, get more and more of the income in the economy. So it's kind of like saying if AI makes software free, it's so efficient, we can, it can do anything we want with it. That doesn't mean software takes over the whole economy any more than water or air Being virtually free means those take over the whole economy and those are essential things, water and air, right? But that doesn't mean we can use them to make everything. So if AI gets so good, it might just make software incredibly cheap and, and, and, and we might get 2% higher income outta that. Or we might get, if instead software can be used to do a ton of different things that it's not doing now, then maybe we'll get a 20% burst in income over the next 10 years out of that. That would be great. And again, my instinct is just try really hard to, to deal with the distributional question. So as many people can benefit from that as as possible. But anyway, the, the, the Jones and TenneT view is that we're not gonna get an explosion of growth in our lifetime, which is one of the more pessimistic takes on it. But I, I have to say, and they're, and they're transparent about this, they don't really know how many bottlenecks there are. And I would again, go back to if in research we can really, they're, they're kind of assuming research uses all goods in the economy. So if there's some bottleneck that's a bottleneck to research too that might not be right, might be that AI can make research so productive that we can solve all kinds of human, you know, we can stop global warming, we can, you know, cure Alzheimer's. We can, and dementia more generally, we can, you know, cure a bunch of cancers. That would be incredible. That's what I'm hope, that's the cautious optimist in me is that it's gonna lead to explosion of research, productivity and growth. I said just the other day, some open AI tool I think solves some very old map problem. So yeah, I'm, I'm constantly optimistic like yourself and I, you know, as Bob Solo used to say, you know, AI or, you know, internet was everywhere, but except in the productivity statistics, I think it's fair enough to say that that could be the case with ai. At least for now though we've seen some slightly, you know, above I think average GP growth prints, but it could, you know, certainly get a lot higher. You know, this stuff, you know, gets more fully integrated into the economy. It's a real honor to have you on Pete. I really think that thank you allocations, you know, a Nobel Prize winning idea and it's a real honor to be able to talk to you about it and all your other terrific contributions here. Thanks so much. This was really fun. This is the Capital and Freedom, the 21st Century podcast, an official podcast of the Hoover Institution's economic policy working group where we talk about economics, markets, and public policy. I'm John Hartley, your host. Thank you so much for joining us.