What’s Impacting American Workers?

If you want to be an economics professor at a top university like Harvard or M.I.T., there’s essentially one path. You major in economics at an elite university, and then you go directly to get a Ph.D. from a top five school, maybe pausing for a year or two to do research with a big name economist. Almost every top economist I know followed that formula. And then there’s the M.I.T. economist David Autor. I call him “the accidental economist.”

AUTOR: I may sound like a happy-go-lucky, it all goes well, just put myself in the right place at the right time and boom, there I am. But it was really challenging.

Welcome to People I (Mostly) Admire, with Steve Levitt.

David Autor is one of my favorite economists. He has an incredible knack for seeing things other people can’t see. But as soon as he points them out, you kick yourself because it seems so obvious once he explains it to you. If I have a question about the labor market, David Autor’s who I turn to. He seems to know the answer to every important question in that area.

*      *      *

LEVITT: David, it’s a sign of how much you’ve accomplished as an academic economist that the popular press has bestowed monikers, or nicknames upon you. And that’s a level of notoriety that’s rare in our field. The Economist magazine christened you “The Academic Voice of the American Worker.” And John Oliver, he called you, quote, “a twerpy economist.” Do you wear those two monikers with equal pride?

AUTOR: Oh, yeah, “twerpy M.I.T. economist” is my favorite of all. My kids always tease me about the first one. They go, “There goes ‘the Academic Voice of the American Worker,’ shopping for shoes online.”  

LEVITT: That’s awesome. It’s not completely by chance that you became the academic voice of the American worker because you are one of the few top economists that I know who actually spent some time working at a real job. Most economists follow a very straight and narrow path. You started on the typical path, growing up in a Boston suburb, you had two highly educated parents, you got into Columbia University. But then you jumped off that path. You dropped out of Columbia. That must have taken a lot of courage.

AUTOR: I don’t know if it took courage or just desperation. I was so immature. I knew I wasn’t making it.

LEVITT: What was going wrong? I’ve been around you a lot — you’re completely functional. What didn’t you know how to do?

AUTOR: I was completely distracted by, you know, one, non-academic interests, two, misbehavior, three, girlfriend, four, crippling insecurity. So, I dropped out, and I have to say, it was a great decision. But the look on my parents’ face when I got home from New York was something I’ll never forget. Their look just said, “Well, it’s over. We failed.” I started just doing temp work and then I got a temp job at a hospital doing basically clerical work. They had just got a computer for the first time, a desktop computer, and I started programming it. And I ended up writing software for their nursing department. So I did that for — I guess it was really two years. . And then I went back to college—  

LEVITT: Why did you go back to college? Your parents were still glaring at you after two years?

AUTOR: No, no, no, no, no. You know, this actually happened a number of times. I sort of reached the end of what I was doing. Somehow I knew I had done what I was going to do there, it was great, and I knew that I needed to do the next thing. So it made sense to go to college. And I applied to a number of universities. I chose Tufts and that turned out to be a fabulous decision. I got so much individual mentoring there. And I had complete focus at that point. I was actually almost mature for my age. Or actually no, for the age of college kids of whom I was older.

LEVITT: So you graduated from Tufts, but from what I know, the sum total of your post-college career planning consisted of buying a beat-up old car and starting to drive westward without any idea where you were going. Is that a legend, or is that actually a true story?

AUTOR: That’s actually a true story. That’s far truer than my being the economic voice of the working class, which I think is a lovely moniker that I wish I lived up to. But, no, it’s true. I studied clinical psychology in college and I was very serious about it. I even published as an undergraduate with a faculty member and another student. But by the time I was done I realized it didn’t feel right to me. I almost worked in software. Even when I was in college, a friend of mine had a computer consulting business. I worked there over the summers and I liked the puzzles, but I didn’t feel like I was doing something useful. With psychology, I like the questions, but I didn’t feel like I had good answers. So again, I knew I needed to reset. I didn’t know what it was. So yeah, my girlfriend of the time and I, we got in this little roller skate of a car, and we drove for seven weeks, like zigzagging around the country, up and down. And then we were at the Big Top Chautauqua in Minnesota, and I heard an NPR story about this computer learning center for poor kids and adults opening up in San Francisco, and I thought, ‘Well, I’m going to San Francisco’ — that’s where her family was — so I applied to volunteer there. And that turned into a job for three years. I was the education director there. I did teaching, but also organized volunteers and classes and also did all the technical work and fundraising. And that was the first time I felt like I was using all my skills. You know, working on problems that I would think about as a psychologist, but at a much larger level — social and economic challenges, though I didn’t use the word “economic” very much back then. And also working in the technical capacity that I found so interesting. So, that felt right. Then I did related work in South Africa as a volunteer for the Congress of South African Trade Unions, a non-racial trade union. But again, after three years of doing that, I also knew I had reached the end of that career. And that’s why I applied for graduate studies.

LEVITT: So you ended up going to the Kennedy School, Harvard’s public policy school, and you got a master’s and a Ph.D. I’m surprised you thought that was a sensible path.

AUTOR: It was not a plan. I realized I was reaching the end of the second career, and I thought, “Well, I should go to graduate school. I should do something at another level.” And I didn’t know what to do. I said, “Well, maybe I should go for a post-baccalaureate pre-med, you know, for a medical degree.” And my girlfriend at the time, now my wife Marika, said, “Wait a minute, you don’t want to go to med school. You won’t even cut a chicken.” And I said, “Well, I’m not going to veterinary school. I’m just going to medical school here.” But she was right. And I decided to apply to a public policy program, and it wasn’t until I got into that program — I said, “Well, I’m really interested in this question of sort of technology and power and incomes and inequality.” That’s what I’d seen doing this work in San Francisco, and how this sort of digital revolution was making some people far more productive and far more affluent and leaving others with this big skills gap. And I thought I could do a thesis about that. So I signed up for the upper level statistics and economics classes, and that was my first economics class. It was called Advanced Welfare Economics.

LEVITT: And how old were you by this time?

AUTOR: I was 29 when I took my first economics class, and I was 30 when I took my first calculus class, sitting in Harvard Calc 1B with 18-year-olds.

LEVITT: I went to Harvard and I never made it to 1B. I stopped after 1A. So you’re way ahead of me math wise.

AUTOR: And that graduate economics class just blew my mind. I thought, “Oh, this actually works on the questions I care about, which are questions of well-being and opportunity, but uses tools that I relate to, which are formal analytical tools. This is a way to combine the way I like to think with the things I like to think about.” And so I was enraptured — so enraptured that a couple years in, I thought, “Man, I should have become an economist. That’s what I should have done in life. I guess it’s too late. I’m going to do this policy degree, and I’m always going to feel like a wannabe, but too late to turn the page now.”

LEVITT: I remember when I first heard about you, you had done this Ph.D. in public policy and M.I.T. Economics had decided to hire you. And that made no sense. Places like M.I.T. simply didn’t hire people from outside of economics. And I remember my very first instinct, the first three seconds that I heard about this, I was outraged, like, “What’s going on? What does M.I.T. think they’re doing?” And I was smart enough and reflective enough to pause and say, “Wait a second. Whoever this guy David Autor is, if he managed to get a job at M.I.T. with a Ph.D. in public policy, he must be something special.”

AUTOR: As shocked as you were by my appointment at M.I.T., I was more startled and much more convinced than you that they had made a terrible mistake. I knew very little economics compared to people around me. And I remember my first year of teaching undergraduate micro theory, I would be in my office late at night, literally crying because I didn’t feel I understood the material well enough to teach it. And it was hard on me. It was hard on my spouse. We had two little kids. And it took me quite a while to find my bearings. And I will say, I benefited enormously from the generosity of people around me, particularly my M.I.T. colleagues Daron Acemoglu and Josh Angrist and Esther Duflo and Amy Finkelstein. Some of them have appeared on your program. They gave me a lot of time, and allowed me to get stuff wrong and, rather than judge me, help me get it right.

LEVITT: So you got into economics. And you immediately started looking at computers and how that had affected the labor force, just as you had promised yourself when you went to the Kennedy School. Can you talk about that first research?

 AUTOR: My Ph.D. started in 1994, and there was a lot of talk about personal computers and their effect on the labor market, and that’s because there had been this explosion of inequality. But back then, it took economists seven to 10 years to notice what was happening because the data came in so slowly. And so people started realizing in the early 1990s that inequality had been growing since the early 1980s, growing very fast. And they were looking around for possible explanations. Could it be labor unions? Could it be minimum wages? Could it be trade? Could it be technology? And people started saying, “Maybe it’s the computer.” And then the most influential work at that time was by Alan Kruger, who is a personal friend and also a co-author eventually, someone who passed away tragically very young. But his work was about: have computers changed the labor market? And he was focused on computer skills, people who did word processing and programming, and how they seemed to make more than others. And I thought about this for a while. And I was talking often with Dick Murnane, who was one of my advisors, and Frank Levy, who was a sort of external advisor at M.I.T. in the Urban Studies and Economics Department. And I said, “This just doesn’t seem the right way to think about it.” Like, the biggest effect of computers is not on the people who use them, but the people who don’t use them. And it’s not the computer skills that — you know, anyone can use a word processor; it’s what other skills that they substitute for or compliment.” What’s difficult for computers is easy for people, and what’s difficult for people is easy for computers. You know, computers can calculate Pi to a million decimal places, but they can’t empty a wastebasket. They follow routine, codifiable tasks, and procedures. So they’re going to be really applied to those types of tasks — you know, clerical tasks, administrative tasks, routine production tasks done in a very consistent environment where you can just follow rules. And what they’re not going to do is either the manual, dexterous, common-sense work that people do in services and in blue-collar trades. And they’re also not going to do a lot of the abstract reasoning, creative thinking that professionals do. But they are going to be really valuable to professionals because we need analytic inputs. We need information to make good decisions. So they could be really polarizing. They could sort of hollow out this middle and leave the upper stratum of highly educated professionals, and then another set of people who are doing personal services. And that’s socially valuable work, let me be clear. But because it’s not expert, because many people can do it productively without a lot of training and certification, it doesn’t tend to be highly paid.

LEVITT: Right, there’s lots of supply.

AUTOR: Exactly.

LEVITT: So it drives down the wages.

AUTOR: So that’s what we wrote down in a paper that we published in 2003. And when we published it, I thought, “Well, this is kind of the last nail in the coffin of this literature. I guess we’re done. No one’s going to care in a few years.” But it turned out to be relevant.

LEVITT: As you describe that research now with 20-plus years of hindsight, it feels obvious. It seems like the only way to look at the world. But I can attest that it was really novel, really eye opening at the time. I remember when it came out, I didn’t know what to think of it. I was ready to dismiss it. But the two economists at the University of Chicago that I respected the most, Gary Becker and Kevin Murphy, they loved that paper. And they didn’t like anything. So when I knew that they loved what you were doing, I knew that it would turn out to be important.

AUTOR: I did not know that. That’s awesome to hear. I mean, I visited the University of Chicago in 2006, 2007. I got to know Gary Becker a bit. He was such a brilliant man, but also such a stellar human being in the way he treated others. And then Kevin Murphy, of course, is one of the sharpest intellects of many generations. Yeah, so that’s really — thank you for telling me that.

LEVITT: As I look at my own life, it has been completely and utterly transformed by advances in computing. I could give you a thousand examples. So the U.S. government routinely tries to calculate measures of worker productivity, the value of worker’s output. And my instinct is that advances in computing should have hugely transformed output — should have been a period of incredible productivity. And yet the data don’t show that. What do you make of that? Do you think the data are wrong or do you think that my experiences are not capturing what’s really happening in the economy?

AUTOR: I do not have a definitive answer. I’m going to say a few things. One, the data are somewhat wrong. I think most people would agree that we undercount productivity growth because we don’t know how to measure it well, right? If your computer gets 20 percent faster, is that a 20-percent improvement? Is it a 10 percent improvement? Is it a 0 percent improvement because it still does the same things? So it’s hard to quantify that well, but this has always been true. So it’s not clear the mismeasurement is any more extreme in recent years than it has been in the past. I think a second reason is: computers were very good at a comparatively narrow set of activities. Now I know that sounds crazy, right? We use them in everything. They’re in our televisions, our phones, in our cars and in our blenders and in our toys. But if you compare them to the technologies that preceded them — which would include like penicillin, indoor plumbing, electrification, air conditioning, flight — those things were even more transformative. So I think we have experienced real productivity growth. And no one thinks it’s zero; it just may not have been as fast as the roller coaster of productivity growth we’ve been experiencing really for two centuries, but in particular, after the Second World War. You know, most of human history is a flat line. There is no productivity growth for a century at a time. Essentially, people are doing the same stuff, and then they get ravaged by disease or pestilence, and they’re sent back further. They actually lose things they knew, writings are lost, scientific knowledge is lost. And only in the last 200 years have been one of accumulative, accelerating progress. You know, if we’re growing at 2 percent a year, that means we’re growing faster and faster in level terms because 2 percent of a larger number gets larger and larger as that number grows. So we haven’t hit the breaks. We just may have decelerated a bit, but we might be accelerating again. I think there’s reason to think that artificial intelligence will certainly speed up frontier scientific progress that will improve medicine, that will improve energy generation, that will improve, hopefully, the way we educate, and the way we design things.

We’ll be right back with more of my conversation with economist David Autor after this short break.

*      *      *

LEVITT: I want to take advantage of having you here to talk about the broad forces that have been working in the U.S. labor market over the last 50 years and to use your research to put some of that into context. And I can think of four or five key trends that are extremely important, and I think in most cases came as a big surprise to economists when they appeared. And one of those is the rising level of income inequality. You had mentioned that before. But if my memory serves me correctly, from 1930 to 1970, I think the income distribution in the U.S. had actually been becoming more equal.

AUTOR: The Second World War was a great equalizer.

LEVITT: And then things just totally flipped.

AUTOR: That’s absolutely right. There was a period from basically the end of the Second World War through the mid 1970s, incomes were growing rapidly and evenly. And then comes the first oil crisis in 1973, they stagnated quickly and evenly. And then there’s an explosion of inequality starting right around 1979, 1980.

LEVITT: So it’s not that easy to measure, but when people like Thomas Piketty try to measure it, it looks like income inequality today is roughly on par with the most unequal distribution we’ve had since we started measuring it back right before the Great Depression.

AUTOR: In 1915. Yep, that’s correct.

LEVITT: Okay, so we’ve talked about your work on computers. What share of that increase in income inequality would you attribute to the rise of computers? Big or small?

AUTOR: I would say it’s probably one of the two most important forces. Actually the force that gets the least credit but probably was the most consequential is the slowdown in educational attainment of college going after the Vietnam War. If you look at the U.S. data from cohorts born from like 1910 to 1950s, there was year on year a rise of like a half year of education — maybe a half year is overstating, maybe a quarter of a year, but even so this kind of steady rise. We’re producing more and more educated adults. And the labor market was demanding them, was absorbing them, because we’re growing the professions, we’re growing medicine, we’re growing law, we’re growing engineering, and we’re growing even computer and information science. And then after the Vietnam War, men and women not only slow down, but actually reverse the trends of greater college attainment. And so around 1980, that starts to bite. Those people who would have been exiting college and joining the labor force all of a sudden were very scarce. But demand hadn’t slowed down. And so it was the kind of steadily moving outward demand, probably accelerated somewhat by computerization and then this lack of supply. The tap was running dry and that contributed a lot to rising earnings for highly educated adults.

LEVITT: So, essentially what you’re saying is there was a huge increase in income inequality heavily tied to the amount of educational attainment that people had. So what we observe is that people who, say, only had a high school education, their real incomes, I think, are lower today than they would have been —

AUTOR: They fell rapidly, especially among men in the 1980s. Then they have come back somewhat. The last few years have been relatively good, but they’re certainly not much higher by most measures, especially among men, than they were 40 years ago.

LEVITT: And let’s say for men who have not only gone to college, but gotten some kind of a graduate degree, how do wages compare today to, say, the 1970s?

AUTOR: The wage gap between college and non-college is more than twice as high. And then when you’re talking about people with college and especially graduate degrees, they could be three or four times on average what they would have been in 1980. There’s been a lot of growth at the top. And it really has fanned out from kind of the 80th percentile on up. And of course, way, way, way, way up. But that’s not the only factor. Technology played a key role. It really contributed to the sort of taking special talents — whether they are in finance, whether in medicine, whether in law, whether in engineering or design — and really magnifying their value because those people could accomplish more, so much better. They had better tools. They could produce more. So that contributed as well to this kind of — you might call that a bit of a “superstar effect,” to use the term of your former colleague Sherwin Rosen.  It’s also important to recognize the decline in the real value of the U.S. minimum wage, which fell in real terms by 30 to 40 percent during the Reagan administration because of inflation, right? They didn’t want to raise the minimum wage and its real value fell. That made a big difference at the bottom, more than I think people recognized for a long time, even more than I was willing to recognize for a long time. And the decline of labor unions — you know, there are many things you can like or dislike about labor unions, but they did raise wages for blue collar workers. And then even a change in tax policy and norms about, you know, “greed is good.” The big declines in marginal tax rates on the affluent during the Reagan years made it more worthwhile to go out and really fight hard for that last million because you got to keep it.

LEVITT: You sounded like a good Republican economist there for a moment, implying that lowering marginal tax rates actually led people to work harder.

AUTOR: Well, I just meant that makes them fight harder to get that slice of the pie — whether it produces more output or they get it from someone else in the company. I’m just saying it really matters if it’s not being taxed at 70 or 80 percent.

LEVITT: Okay so the payoff to getting education has gone way up. Isn’t it puzzling to you and to me as economists why, in response to this enormous growth in the return to schooling, we didn’t see people go get more schooling? That flies in the face of all of our models, right?

AUTOR: We did see a supply response, but it took a while. So in 1980, men were about 8 percentage points more likely to have a bachelor’s degree among people 25 to 34 than women. By 2018, women were 8 percentage points more likely to have a bachelor’s degree, but both of them had increased. However, there are many caveats to that. One is: part of the expansion was in for-profit colleges, which don’t have a very high payoff. Two, the financial cost of college has risen a lot. And it’s risky because about 40 percent of people who start a bachelor’s degree do not complete it, but they do come out with a lot of debt. Who completes bachelor’s degrees? They’re kids from relatively affluent families. They go to schools that are highly resourced. Like, it’s really hard to get into M.I.T., where I teach. Once you’re in, it’s really hard to get out without a degree. We’re going to make that happen. And so there is a real, significant financial risk at this point for people who, you know, it’s just not as clear that they’re as well prepared, that they’re going to a place that’s going to help them push through. And additionally, if their loans go into arrears, their families can’t step in and say, “Oh, I’ll cover that for you for a couple of years,” because their families aren’t going to have those resources. So it’s a risky bet, more so than it used to be when college was much cheaper. It is also the case, by the way, that the measured earnings differential between college and non-college has plateaued over the last decade, and by some measures has come down a bit. It’s still quite high, but we are actually in a period where inequality is moderating, which is stunning to say. My entire career, there have been a bunch of tectonic trends that have been always running in the same direction, and all of a sudden they are slowing or changing direction. It’s actually quite exciting. We’re really in a different era.

LEVITT: You’re lucky to be active long enough to be able to see things change, and also to be open minded enough both to accept the change and to be trying to explain it. It’s difficult not to get locked into the views you held when you were 30 and to let them go. Okay, so another huge change in the U.S. labor market over time has been driven, I think, by cheap imports and that impact on U.S. workers. China’s growing role in the world economy is just completely stunning, right?

AUTOR: Yeah! You know, when I was learning about the rise of inequality as a graduate student, there was this kind of raging debate. Is it technology or is it trade? Is it trade or is it technology? Which is it? And labor economists and trade economists got together and they all collectively agreed it was technology. It wasn’t trade, because there just hadn’t been that much change in trading patterns through the 1970s and 1980s. And then the question got put down. This is something I’ve worked on with Gordon Hanson of the Harvard Kennedy School and David Dorn of the University of Zurich — Gordon Hanson said, “Well, this China trade change, the magnitude is just unprecedented and this is potentially quite different.”

LEVITT: So what are the numbers on China? What is their share of trade? How has that changed?

AUTOR: China went from making under 2 percent of the world’s manufacturing goods back in the late ’80s to more than a quarter now. And it has unseated everybody in every potential category. It’s amazing but not mind blowing that China is such a large part of the world manufacturing economy because China is an enormous country. What’s remarkable is how fast it went from being a technologically backward, politically unstable, non-market economy to a premier competitor across every sector. That change has been, in many ways, invaluable. It’s brought hundreds of millions of Chinese out of abject poverty. It’s also created prosperity in Central and South America, in Sub-Saharan Africa. It’s done lots of good things. But it was incredibly destabilizing for the countries that were used to being those producers. And no country was more destabilized by China’s rise than the United States. We were most directly in their kind of — I don’t want to say line of fire, but in their path they were traveling. Because the U.S., unlike many wealthy industrialized countries in Europe, was still doing a lot of low-tech, labor-intensive production in textiles, in shoes, in commodity furniture, like you’d buy at Walmart or Target or Amazon. Many other countries had moved out of that, but because we have a real low-wage economy for such a rich country, especially in the U.S. South, we were still doing a lot of that. And so China’s incredible acceleration into the world manufacturing system in the 1990s, and even more so after 2000 when it joined the World Trade Organization in 2001, displaced more than a million American manufacturing workers in a very short period of time, with spillovers that could easily reach 2 million workers.

LEVITT: I’m surprised your numbers are so small. The number of workers in manufacturing in general must be down 6 or 7 million from the peak, but you think the Chinese only account for a third of that?

AUTOR: Well, okay, so good question. Of the almost 4 million manufacturing jobs that the U.S. lost between 1999 and 2007, which is unprecedented as a share of all manufacturing employment, we say conservatively a million. Our estimates suggest much more, but we were trying to be conservative. But I think it’s possible to say, “Oh, a million, two million, what’s the big deal? We have a labor market of 150 million workers. Come on, that’s a drop in the ocean.” But of course, manufacturing is extremely concentrated. Where it occurs, it occurs in a very intense way. And not just manufacturing generically, but you know, “This is where we make textiles. This is the sweater capital of the world. This is the furniture capital of the world.” And so when those jobs went, it was like a thermonuclear device going off in the middle of Main Street. It really decimated those places. They are now recovering, but it’s a different set of people who are doing the jobs that are present there now.

LEVITT: So I can attest as an economist that this research that you did was extremely influential. But I always try to put myself in the shoes of people listening to this podcast who are outside of economics. And I suspect their reaction to your research about the effect of Chinese competition devastating chunks of American industry and society is, “Well, I didn’t need a twerpy economist to tell me that. It’s obvious.” But somehow it wasn’t obvious to economists, maybe because we’re brainwashed from the first day we show up in an economics class to believe that free trade is the most perfect and amazing thing that exists.

AUTOR: Yeah, no, I think economists are often some of the most surprised people in the world because things that are self-evident to everyone else, we’ve unlearned. I have in my office at M.I.T. a letter that someone wrote to me. There was a headline in The Wall Street Journal, “Economists find that Chinese imports reduce U.S. manufacturing.” He says, “Wow, M.I.T. economists have made a great discovery. If you had ever been in any store anywhere in the country, you might have noticed this. I welcome you to come to mine and I will show you.” And it was brilliant. I wrote back to him and thanked him and I framed it. Because it’s true. It should have been self evident. And economic thinking was like, “Well, labor markets are frictionless. People will find new jobs. We have a reemployment program. Sure, people will lose manufacturing jobs, but they’ll take something equally good right nearby.” And they completely overestimated how readily people adjust. So the shocker is not that some manufacturing jobs were lost. It’s that the scars that that left were so deep and the healing process so slow. I think that’s what scholars misjudged.

LEVITT: So what’s your policy advice? When American manufacturing is no longer competitive in the sector, the typical reaction is tariffs, which most economists hate. I suspect you probably hate them too, right? Because they’re really inefficient.

AUTOR: So if we could roll back the clock and say, “If we had to do over, what would we do differently?” It wouldn’t be keep China down. But we should have made this unfold more slowly. We should have said, “Look, we’re going to adjust, but we can’t do it in the course of seven years because it takes a generation for people to move out of one sector into another.” And that was feasible, by the way. That’s not just a fantasy. China’s accession to the World Trade Organization was negotiated. But in terms of looking forward, at this point, manufacturing is not just a question of who’s making textiles, who’s making shoes. It’s really a strategic and even a military, a geopolitical question about who has control of artificial intelligence, who is going to make the drones, where will power generation come from, who has control of rare earth elements? The telecoms equipments that we’re making, who’s spying on them? And so it’s a much more complicated question than just: who has the jobs? And so I don’t think simple economic models of “how much is it worth to save this manufacturing job versus those wages?” is really going to do it. But I would say if the goal is seriously to protect or revitalize certain capacities — like electric vehicles, like semiconductors, like wind turbines, like aviation — then it’s going to take not just building walls, where tariffs are a form of walls, but also investment. You cannot win a race just by hobbling your competitors. That will work for a little while, but then they’re just going to get stronger and faster. You have to invest in yourself. You’ve got to bulk up. 

LEVITT: The economist Kevin Murphy, who you and I both admire very deeply, he once said to me, “If you own a factory that’s making trinkets, it’s not because you want to make trinkets. It’s much more valuable to make semiconductors. It’s just that trinkets are easy to make and semiconductors are hard to make. So when you’re not that good at making stuff, you make trinkets.” And that’s really always stuck with me.

AUTOR: And that’s why the trade topic, especially vis a vis China, is so sensitive and kind of consequential at this point because China is no longer just making trinkets. And it has in its sights a lot of the sectors that are, you know, both critical for innovation and the products that we’ll be spending large shares of national incomes on, but also have huge implications for military, for how we will compete with one another, not just in the economic realm. So if the U.S. wants to rebuild these capacities that have been so critical to innovation and to profits — you know, everyone uses iPhones, but we’re happy that Apple is here, right? We all fly. But I’m glad — well, until recently — that Boeing was here. I think we need to make ourselves good. And part of that, by the way, we may put tariffs on certain companies or competitors, especially when they’re being hugely subsidized, but we should be welcoming foreigners. Part of the great strength of the United States has been our ability to attract incredibly talented people from around the world. I mean, that’s one of the secrets of our great universities. They don’t just produce good people; they attract good people. They’re a magnet for capability, and those individuals are at the heart of our innovation system. You just look at the people who won Nobel Prizes and formed companies, and they do a lot of good. So, the last thing we want to do is put tariffs on people.

LEVITT: Okay I can think of a couple other important trends in the labor market we haven’t talked about yet. Economists measure the share of income in the economy as a whole that goes to labor, as opposed to the amount of income that goes to people because they own capital or things like machines or buildings or land. And my impression is that after being really steady for a very long time, that share of the total income that’s going to labor has just plunged.

AUTOR: That’s right. Let me say why this matters. You might think that in the wealthiest countries in the world that have the most machines, the most capital, the most infrastructure, that capital would play a very large role in the economy and labor would be kind of a small deal, because it’s so much of what we’ve built and, you know, so much of what we automated. But in fact, labor share of national income is higher in the rich world than elsewhere. One of the remarkable achievements of modernity is that even as we’ve advanced technology, we’ve continued to make human expertise really valuable, really necessary. And that’s critical because that’s kind of where the labor market lives. People are needed to make the technology work. Labor share of national income is higher in the rich world than elsewhere. It’s, you know, 55 to 60 percent in the United States and Europe, meaning 55 or 60 out of every $100 is first paid to workers, rather than owners of stocks or machines. And that fell by 5 to 7 percentage points, depending on what measure you use, in the United States from 2000 to 2020. And there’s a raging debate about what that was caused by. For a while, people thought it was trade directly. I don’t think people think that now. Some, including I, have argued that it probably has to do with the growth of very large corporations and their growing market size. They use much less labor. Other people think it has to do with automation and that we’re just kicking out workers and replacing them with machines. It’s probably some of both. And if people are increasingly superfluous, that the machines tend themselves and there isn’t that much work to do, that’s a problem. It’s not a problem because we’re getting poor. It means we’re getting rich, in fact. But it’s a problem because most people get most of their income from their labor. And if they didn’t get it from their labor, we would be dependent upon transfers and taxation and common ownership of capital to redistribute that income. And people don’t like that. People want to make a living based on their own skills and work, rather than be in a system where they’re sort of wards of the state or where we rely on Sam Altman and Marc Andreessen to write us a check.

LEVITT: So what you just described is, for some people, a nightmare scenario associated with A.I. where, as A.I. gets better and better, workers are superfluous and A.I. does everything. Now, what might surprise people, given what you just said about A.I., is that you’re actually, among the people I know, about as optimistic as anyone about the possible impact that A.I. will have on workers. You’ve thought about it a lot more than I have, but I have to say I don’t share your optimism. So tell me your vision for this great future in which A.I. reinvigorates the American middle class.

AUTOR: It’s not A.I. that’s going to reinvigorate it; it’s how we could use it to reinvigorate it. You know, we’ve been talking earlier about expertise and how traditional computerization hollowed out the middle — either you moved up into law and medicine, or many other people are doing services where their expertise wasn’t as valuable.

LEVITT: So the middle before that was people who were doing jobs that required a lot of routine calculations that were not easy to do, but you could learn to do them. And computers were just way better and they wiped out that part of the market.

AUTOR: So production, office, clerical, administrative support, operative positions, those were skilled jobs. People needed a high school degree, some had a college degree, and they were doing skilled work. It just turned out to be work that was well described by rules and procedures that could then be executed by machines once they were codified. And by codified, I mean written as computer code. But A.I. is completely different from that traditional computing that follows rules and procedures. A.I. actually can’t follow rules. It can’t keep facts straight and it doesn’t do math well.

LEVITT: Not yet.

AUTOR: What it does do well is it supports decision making. It can draw inferences and recognize patterns in data that you might or might not perceive yourself, whether you’re doing medical diagnosis, whether you’re doing writing, even doing artistic creation or engineering or design. And so it can support people to use judgment more effectively. Let me give an example why this matters. Most of the valuable jobs in the economy are decision-making jobs. They’re jobs where the stakes are high and the answer is uncertain. And that would be true if you’re diagnosing a patient, if you are architecting a piece of software, if you’re designing a building where you want it not only to stand, but people to want to stand in it. And those jobs are completely dominated by guilds of highly educated professionals — lawyers, doctors, professors, architects. And what’s holding people back from doing them is developing the expertise to do that work. So a good example is, in the field of medicine, there’s an occupation called the nurse practitioner. Nurse practitioners are registered nurses who have additional training that allows them to diagnose, to treat, and to prescribe pharmaceuticals, something that previously only medical doctors could do. This is an example of people who are not at the absolute elite of their profession — they have five fewer years of education than a medical doctor — and yet they’re doing valuable work that previously was controlled by a guild. By the way, they fought like hell to do this over decades.

LEVITT: I’m sure, yeah.

AUTOR: Doctors did everything they could to prevent it, and they still do. But now at this point, they’re heavily supported by technology — electronic medical records, software that looks for adverse drug interactions, diagnostic software. And so technology enables their expertise, their knowledge of the human body, their knowledge of treatment, to go further. They can take that set of skills and apply it to a broader set of problems. And I think A.I. is an enabling tool that can now allow more people who are not at the frontier of computer coding or of writing or of law or of healthcare to do expert work with appropriate training.

LEVITT: So you have this vision that there’s a whole set of people who maybe don’t have the skills or the financial resources or the patience to go become doctors or lawyers. But armed with A.I., many of the tasks that the current guilds, the doctors and lawyers do, will fall to them and they will become the mainstay of the economy. Now, what you haven’t talked about is this doesn’t bode well for the guilds, right? The doctors and the lawyers are going to take a big hit in that world, right?

AUTOR: That’s fine. Yeah, they’ve had a good five decades. They’ve saved enough. They’ll be okay.

LEVITT: Professors too, right? We’ll be taken down as well, right?

AUTOR: And that’s fine. I don’t mean to say that everybody will be able to do all those jobs with these tools. What I mean to suggest is they’ll be able to, with the right foundation, do more with less. So the analogy I like to use is YouTube. If I want to go do some big electrical project in my house, let’s say I want to replace my fuse box with a breaker box, I could go to YouTube and I could watch a video about how to do it. There are literally hundreds of these. But if I didn’t know what I was doing, I’m surely going to electrocute myself or set my house on fire, right? It’s a terrible idea. If I’m a master electrician, I’m not going to watch that video. What would I do that for? But if I’m a person who’s a hobbyist, I actually know how to use an Ohmmeter, and insulated gloves, and I understand principles of electricity, I could take this tool and I could do a project that I couldn’t do on my own. And in fact, 60 percent of Americans report they use YouTube to do things they don’t know how to do. I think A.I. is an enabler, or can be used as an enabler, to allow people to go further — you know, paralegals to do more valuable legal work, people with understanding of software to write harder code, people who are doing house remodeling to do more design work. We talked about how the people in the middle have been pushed downward into inexpert services that are socially valuable but pay poorly. The opportunity here is that they could move back towards the middle in a different set of activities. So much of our economy now is these expensive services, whether they’re education, health, law, design, software, and they’re dominated by highly educated people who pay high prices for them but then earn high salaries. And then not highly educated people who have to pay for them as well and don’t earn high salaries. And if more people could move into that type of work using better tools, I agree, lawyers, doctors, professors, and so on wouldn’t be quite as far out of reach financially. But that’s a price I think we should all be willing to pay.

LEVITT: It’s hard not to like that vision and to want that vision. But somehow it feels to me like what you described is not a permanent equilibrium, but more like a way station on the path to somewhere else. What I see happening with A.I. relatively quickly is that we’ll have personal A.I. devices. Let’s say David Autor is a fantastic teacher of labor economics. You’re going to have a little A.I. device that rides around with you and you talk to it and it sees you teach and you explain to it why you teach what you teach. And in the end, it will be a supercharged version of you. It will be better than you because it knows all the things that you’ve taught it, plus it has the kind of memory and learning that it knows every statistic and it knows what every other good teacher can do as well. And that supercharged A.I. is your intellectual property, and people are willing to pay you a lot of money for that thing because it’s amazing. It actually supports and crystallizes a superstar economy, which is exactly the opposite of what you described where everybody gets a boost from A.I. It seems more likely to me that the very, very best will be able to create bespoke A.I. with their own human capital, which will make everybody else irrelevant. What do you think of that scenario?

AUTOR: It’s an interesting one and I’m going to make the claim that these are not as far apart as you think. The thing that unites them actually is capacity. So let’s say A.I. made all the restaurants in the world 20 percent better, but it made the best ones 75, or 100, or 200 percent better. If those restaurants are so good, does that put all the other ones out of business? And the answer is no, I’m never going to go to the best restaurant. I can’t afford it. And the waiting list is two years. I care about the 75th-percentile restaurant. When I go to a doctor, I don’t get to see the best heart surgeon or the best neurologist, but I want the one I see to be good. It’s quite possible for technology to make a lot of people better and even make the superstars that much more super. But if they can’t serve the whole market, then it’s possible to both have the upper tail pull away and have a lot of improvement in the realm of services that most of us will participate in.

LEVITT: So the big question there is who delivers that last mile of service. So in your world, it’s a reasonably skilled person who gets paid pretty well for doing it. And that’s my sense of what doesn’t feel like an equilibrium, because it feels to me like if we hit this moment where A.I. is better than humans at all kind of thought, then these tools are going to figure out better ways of delivering the service that don’t require humans to have any skill at all. I mean, take the extreme — you could imagine in the long run that humans would take on A.I. partners as opposed to human partners. I mean, you could actually specify every element of your partner and have some control over it, as opposed to the whim of what real humans are like. So, I don’t know, somehow for me, I quickly devolve past this point where you’re at, where humans still kind of run the show. I had Max Tegmark, one of your colleagues at M.I.T. was my guest, and maybe he put the idea in my head that that’s not the stopping point. The stopping point is when A.I. takes over.

AUTOR: Well, I don’t like to forecast out beyond 20 years. Because I just don’t feel I have enough certainty. So I feel like my scenario is much more plausible for the next 20 years, but maybe after that it’s different. But the scenario you’re describing really relies on dexterous robotics, not just cognition. So, you know, imagine you come to your medical office and you say, “Professor Autor, I didn’t realize you’re a doctor. I thought I was going to see my doctor.” I was like, “Well, I’m not a doctor, but I have this A.I. here and it’s going to tell me what to do.” That would not be a good idea. It’s like, what if I made a mistake? I nicked you with a scalpel and then you start to bleed out. When I say, “Oh, A.I., tell me what to do now,” right? This would be a bad scenario. So for machines to do all those things you’re imagining, it’s not sufficient that they have the cognition. They also need the dexterity, and that’s actually proved to be a much harder problem. Robotics is moving much more slowly than A.I. because the physical world doesn’t really tolerate hallucinations. You’ve got to get it right every time. And that is actually something that computer scientists understand less well than cognition. I will also say, I do not think the current path of A.I. converges to that level of judgment or expertise. Why are self-driving cars so bad after all the money in the world has been spent on them, and all the data in the universe has been fed to them? And the answer is, because they’re pattern-recognition engines; they are not thinking and extrapolating the way we do. So people say, “Well, the cars can’t handle edge cases. They haven’t ever seen a cow fall from a bridge. So how would they know what to do?” Well, you would know what to do if a cow fell from a bridge, right? You would swerve around it. Because you have a model of the world that goes beyond patterns. It understands causality, the arrow of time, the motivations that cause objects like cows or children or balls to do what they do. And so I don’t think A.I. in its current kind of dustbowl empiricism without models converges to the point that people are forecasting. That doesn’t mean it’s not an incredibly powerful, useful tool. I just don’t think it’s where people imagine it to be. It’s improved so fast that the next leap is, “Well, if it got this good in two years, imagine what we’ll do in 10 years.” And at the current rate of progress, I think it’ll do nothing much more spectacular in 10 years than it does in two years, unless there’s another paradigm shift.

LEVITT: Wait, you’re pretty hard on autonomous vehicles. They’re way better drivers than humans already. 

AUTOR: On average.

LEVITT: But humans crash into stuff constantly. And we just say, “Oh, well, that, you know.”

AUTOR: No, no, okay, so you’re absolutely right that on average, they’re actually better because they never stop paying attention. They don’t fall asleep at the wheel and they have good reflexes. What they’re bad at is the extrapolation from unfamiliar circumstances. And that was the point I was making — not about autonomous cars, but trying to use that to illustrate: what is the limits of cognition of A.I.? Because it doesn’t have a model, it can’t think beyond the data. And so it can’t make the judgment of: if I see this unfamiliar animal running along the side of the road, I ought to think to myself, “This could run into the road.” Something that’s outside the data, but inside the realm of reasoning, that’s where they fall short. And of course, so much of what we do, especially when we’re making high-stakes decisions, involves extrapolating beyond the data and making an educated guess.  

LEVITT: Okay, let me take a different track of skepticism and see how we handle this one. The invention of machines devalued manual labor. The invention of computers devalued rote calculations. Why wouldn’t the invention of A.I., which stores huge amounts of knowledge and thinks deeply and creatively, why wouldn’t that devalue human thinking? Having taken out the three pillars of labor and left not much behind.

AUTOR: Okay, well, a lot is left behind. So we have a full employment economy in the industrialized world — very low unemployment rates. Most people who want to work can work. And we have higher standards of living than we’ve had coming from labor. And so part of the reason “Why haven’t we just run out of stuff?” is because we’re constantly creating new work, new forms of expertise that require human skills, human learning, human creativity, human judgment. At the turn of the 20th century, 39 percent of the U.S. labor force was in agriculture. Four in 10 people.

LEVITT: That is incredible. And what is it today? Two percent or something?

AUTOR: Under 2 percent. And it’s not because we’re eating less. Not only are we feeding a nation of more than 350-million people with a couple of million farmers, but we’re the world’s largest food exporter, and we eat a lot. And that’s more than a century of improvements in irrigation, fertilization, mechanization like tractors, and also genetics. So it is an extraordinary accomplishment, and it is a case where we basically have produced ourselves out of a job. We need relatively few people doing that work. Doesn’t mean we’ve created mass unemployment, because we have so many new things that we do — many of them safer, indoor work. And our ability to do so many other things, to focus on medicine and entertainment and education, comes from the fact that we have nutritional superabundance.

LEVITT: I think if you had been around a bunch of economists in the year 1900, and you told them, “Hey, what do you think would happen to the U.S. economy if we went from having 40 percent farmers to 2 percent?” They probably would have thought it would be the end of the world, right? They would have had no working model to understand how that would happen.

AUTOR: Right. At the time when 40 percent of people were in agriculture, there were no pediatric oncologists, there were no botnet herders, there were no systems architects and virus writers and virus defenders. So much of what we do now didn’t really exist a hundred years ago. In recent work we try to estimate what share of the work people do in 2018 didn’t exist in 1940. And we come up with something like 62 percent. So we have managed to not just automate, but expand our capabilities. Most technologies are transformative, not primarily because they allow us to do what we did in the past faster and cheaper, but because they allow us to do things we couldn’t do in the past, right? When we learned mechanical flight, it didn’t change the way we flew. We didn’t fly before we had airplanes. And we didn’t have antibiotics until we had penicillin. So technologies change our world, not just because they allow us to do what we did in the past. If we automated all of ancient Greece 3,000 years ago, we wouldn’t have modern America. We would have ancient Greece without horses. And so I think A.I. will be enabling for people to do a different set of things. A lot of things that people get paid to do today, you would have thought “why would anyone pay for that?” a hundred years ago. That seems like a luxury, right? Luxury travel, massage, diet coaches. Come on. What is that about? But those are a living now. Those are a living. There’s room for pessimism, let me be clear. I don’t want to say there’s nothing to worry about. I’m sketching a scenario where I think we could use it well. How certain am I that that scenario will come to pass? Fifty-one percent.

You’re listening to People I (Mostly) Admire with Steve Levitt and his conversation with economist David Autor. After this short break, they’ll return to talk about how David stands out as an economist.

*      *      *

David Autor seems like exactly the kind of economist we need working in Washington, D.C., making public policy. But he’s never done a stint there. I want to ask him why not, especially because he was a guy who jumped from one thing to another every year or two before he became an academic. He’s about my age. It’s an age at which many academics get stir crazy. I wonder if he’s got something new in his sights.

LEVITT: Your insights have never been more valued than they are today, which is rare. Most economists have a career in which they start out not very highly regarded and then they peak and then they fall off. But one thing that’s interesting about you right now is that there’s a lot of attention being given to predictions that you’re making when the nature of academic research is almost always backwards looking. Does that unsettle you at all?

AUTOR: Yeah, yeah. It takes a leap of faith and you know you could be wrong. I mean, I wrote an article on Labor Day in The New York Times years ago saying, “Good news, there’s a labor shortage.” And I talked about how inequality was going to fall or was falling and labor markets would be tight. And I thought, ‘Man, I hope I got the sign right on this one.’

LEVITT: And you did.

AUTOR: I did. But I am trying very hard to provide a framework that can predict the past. Because if it can predict the past, it has some hope of getting the future, right? And I feel like most frameworks that people use to understand how technology interacts with labor can’t even explain the things we know, let alone the things we don’t yet know about. They can’t explain: why has labor become more valuable rather than being automated away? Why are some forms of expertise so scarce now and why are others so abundant? And that’s sort of where I started with Frank Levy and Dick Murnane 20 years ago, and I’ve been building on that — most recently in my lecture I gave at the European Economic Association meeting in Rotterdam in August called, “Will automation replace experts or augment expertise? The answer is yes.” And so I’m trying to provide an interpretation that both can accommodate what we know and explain how that’s going to change when the nature of what is susceptible to technology automation augmentation changes. And that’s where I’m going. But yes, I am definitely out on a limb a lot. And I try at the minimum to be honest about what I don’t know and that I’m speculating.

LEVITT: Before you got into academics, you did a lot of different things. You have been laser focused on labor markets and technology for 25 years now. Do you think that’s it for you? Can you imagine doing something different?

AUTOR: You know, that’s actually funny. I do a lot of different things that aren’t that, but they’re not academic. So, I play ice hockey, I sail, I have three kids who I spend time with, I love building and fixing things. The truth is, it is actually very hard for me to stay focused. And I reward myself for work by, like, doing chores, like actually doing electrical work, for example. I’ve never doubted the conviction the labor market is the central institution of human welfare in democratic countries. That it distributes income, but it also gives people meaning and purpose and structure, but it also is foundational to democracy because when most people are producing things of value, so they’re not just seen as vassals of the country, then it’s natural for them to have voice. In America, you know, we famously said, “No taxation without representation.” Well, think of Dubai or Kuwait. They’re like, “Okay, we’ll take that deal. No representation, no taxation. We’ll just give you some money, but you don’t have a voice.” And so I do feel like the labor market provides the fulcrum on which everything else balances. And so I’ve never felt like the most central questions to ask weren’t right there. But I think about them as they interact with trade and with technology and with institutions. I’ve got the opportunity to come at this from many angles. So It hasn’t felt constricting. It felt ever expanding.

LEVITT: You have spent no time in Washington working with an administration. In every respect, except for one, it seems like the absolute natural thing for you to do now. And the one reason I think you shouldn’t go to Washington is because you’re too nice. And I don’t think nice people do well in Washington. But are you open to public service?

AUTOR: I serve on the State Department’s Foreign Affairs Policy Board, so I meet with Secretary Blinken a couple times a year along with people from other sectors. I’ve had the pleasure of meeting with President Obama, then Vice President Biden and so on. But I’ll tell you, the thing that keeps me out of Washington is not that I don’t think it’s valuable or that I don’t think people are nice, but I don’t get a contact high from interacting with power and making lightning fast decisions. I find that setting exhausting. So I just don’t think I would thrive there. For me, it’s digging deep into a problem. I want to grapple with ideas and I can’t do it in that environment.

I never get tired of hearing David Autor talk. If you feel the same way, check out the recent Freakonomics Radio episode number 605 called “What Do People Do All Day?”

LEVEY: Hi, listeners, Morgan here. I am the producer of People I (Mostly) Admire, and I want to ask Steve a question today. He and Stephen Dubner were recently in Nashville to accept an award. Steve, what was the award you were getting?

LEVITT: It’s called the Adam Smith Award, and it’s given out by the National Association for Business Economics. And, we join a pretty impressive list of people to win that. People who have been on the show, like Robert Solow, my mentor Gary Becker, Milton Friedman, Paul Krugman. The list is full of really big names.

LEVEY: Does that matter to you who else has received an award before you have?

LEVITT: You know, to be totally honest, what I do is when I hear of some award, I go and look who won it, and if it’s people who I admire, then I always go do it. It’s a lot of work to win an award, you gotta fly someplace and actually give a speech and spend all day shaking hands with people. So I’m mostly allergic to getting awards, but I feel like if Gary Becker and Robert Solow thought was worth their time to do it, I better do it too.

LEVEY: So, was it a positive experience?

LEVITT: It was almost all positive, except for the fact we had to give speeches, and Dubner gave a speech that was so much better than mine. I had forgotten that Dubner had done this three part series on Adam Smith. And to win the Adam Smith Award, Dubner crafted an absolutely brilliant speech in which he talked about Adam Smith and Adam Smith’s life and details, and it was really masterful. And the only good thing was that I got to speak first. My God, if Dubner had spoken first and I had to follow him, I think I would have just said, forget it.

LEVEY: What was your speech about?

LEVITT: I was talking about something that puzzles me, which is, why is it that natural experiment thinking is almost completely absent in the business world? Natural experiments or accidental experiments, the kind of research that I do, where you can’t run a randomized experiment, so you’ve got to go out and find other ways to get at causality. It’s really just an approach. It doesn’t take a lot of knowledge of statistics, it’s an attitude for finding answers to problems. And I think it would be really useful to businesses if they adopted it. But, I think in part because in high school or college or even in the M.B.A. programs, I don’t think we teach about the idea of what a natural experiment is and how to run it. It just doesn’t happen. So it was sort of a polemic about how business economists should latch onto the idea of analyzing their data through the lens of natural experiments.

LEVEY: Do you think you convinced anybody?

LEVITT: I’ve been trying to do this for 20 years and I failed in every instance. So I kind of doubt I changed everyone’s mind. And I think as soon as Dubner got up there and started talking so beautifully about Adam Smith, everybody completely forgot about natural experiments.

LEVEY: Listeners, if you have a question for us, our email is PIMA@Freakonomics.com. That’s P-I-M-A@Freakonomics.com. It’s an acronym for our show, People I (Mostly) Admire. If you have a question for Autor we can send that to him and, possibly cover his response in an upcoming listener question segment. Again, our email is PIMA@Freakonomics.com. We read every email that’s sent and we look forward to reading yours.

Next week, we’re back with an encore presentation with Pete Docter. He’s the chief creative officer of Pixar and the director of films including Monsters, Inc. and Inside Out. And in two weeks, we’ll have a brand new episode featuring Richard Reeves. He argues that we have a problem right now with gender disparities. Believe it or not, the data suggest that men are getting trounced by women. As always, thanks for listening and we’ll see you back soon.

*      *      *

People I (Mostly) Admire is part of the Freakonomics Radio Network, which also includes Freakonomics Radio, No Stupid Questions, and The Economics of Everyday Things. All our shows are produced by Stitcher and Renbud Radio. This episode was produced by Morgan Levey with help from Lyric Bowditch, and mixed by Jasmin Klinger.  We had research assistance from Daniel Moritz-Rabson. Our theme music was composed by Luis Guerra. We can be reached at pima@freakonomics.com, that’s P-I-M-A@freakonomics.com. Thanks for listening.

AUTOR: We’re standing on uh, moving, uh, on a, on a continent that is, uh, that is shifting. Sorry, let me skip that analogy.