How Can We Fix US Healthcare?
Chicago Booth’s Matthew J. Notowidigdo discusses what it would take for the US health system to produce better outcomes.
How Can We Fix US Healthcare?LADYMAYPIX/Shutterstock
For the past 20 years, the growth in US productivity has been sluggish at about 1.2 percent, compared to the 3 percent pace at which it grew from the mid-1990s to the mid-2000s. Chicago Booth’s Chad Syverson says that if US productivity hadn’t slowed, the US economy would be bigger to the tune of $25,000 per person. So what’s happening to productivity, and is it finally about to turn the corner?
Chad Syverson: Had productivity growth not slowed down, GDP per capita would be over 30% higher, which is about probably another $25,000 per person. So we're all $25,000 poorer than we would've been had productivity growth stayed at its sort of IT boom level.
Hal Weitzman: For the past 20 years, the annual growth in US productivity has been sluggish, at about 1.2% compared to the 3% pace at which it grew from the mid-1990s to the mid-2000s. Chicago Booth's Chad Syverson says that if US productivity hadn't slowed, the US economy would be bigger, to the tune of $25,000 per person. So what's happening to productivity and is it finally about to turn the corner?
Welcome to the Chicago Booth Review Podcast, where we bring you groundbreaking academic research in a clear and straightforward way. I'm Hal Weitzman. Chad Syverson, an expert in productivity, wrote an essay for Chicago Booth Review in 2018 noting that US productivity growth was stuck in a slump. Now he thinks that there are some signs of dynamism in the US economy that could mean that higher productivity is on the way. Chad Syverson, welcome to the Chicago Booth Review Podcast.
Chad Syverson: Thank you. It's nice to be here.
Hal Weitzman: Just tell us, you are an expert in productivity. What is productivity? What do we mean when we say workers are productive or less productive?
Chad Syverson: There's a number of different conceptualizations of productivity, a number of different ways to measure it, but in the end it all boils down to it's the ratio of output to inputs. It's how much stuff comes out of a production process divided by how much stuff goes into it. And that stuff can be workers' time, capital, intermediate materials. And then the output can be any set of a variety of products. So you can think of it as a measure of the efficiency of a production process. If you can get more output from the same inputs, you've become more productive. Equivalently, if you can get the same output with fewer inputs, you've become more productive. So productivity is really just the efficiency of the supply side of the economy.
Hal Weitzman: And some of us think of productivity as being a function of how many people have been fired, because you're told to do more with the same or even with less. So at least according to an economist, that makes you more productive, right?
Chad Syverson: Well, yes and no. So the most commonly used measure of productivity is labor productivity. And the one that most people follow is actually, so it's output per hour. So we are trying to adjust for how many hours people are working. So if the boss says, "Sorry, we got to put an extra five hours per week in there," that should be adjusted and you shouldn't be more productive simply because everyone's working longer. However, if everyone's working harder during those hours, it will be measured as productivity. Now, in concept, I'm not sure I would want to call it more productive production as an economist if everyone's working harder. You really kind of want to hold all the inputs constant and effort per hour is an input.
Hal Weitzman: Oh, so that's interesting. So you mean if people work harder, they produce more, that doesn't necessarily mean they're more productive?
Chad Syverson: No, I think most folks who work in productivity would say, yeah, it might show up in the stats, but we wouldn't conceptually want to say you've become more efficient if everyone's just sweating more. Now, it's a little tricky. If working harder means your time used to be wasted because the workflow in your office was screwed up and so you'd sit and twiddle your thumbs a lot. But the managers figure out how to improve workflow so you're spending actual more time on task, okay, maybe that is more productive. But if it's just the boss is cracking the whip harder and everyone's miserable, but producing more, I don't think that most folks would think of that conceptually as a productivity increase. That's just an increase in effort rather than output per unit effort.
Hal Weitzman: Take us back maybe to some of the history of productivity. You wrote this piece for Chicago Review I think back in 2018 saying productivity's been low for a while. At the turn of the century, productivity growth is running about 3%, but then it slowed down about 2004, didn't it, to about 1.2, and it's pretty much been there ever since. Just remind us of the high level of some of that history please.
Chad Syverson: Right. So the United States has been in a productivity growth slow down since the mid-2000s, and really the entire world has since about that time. And if you just take a step back and think about productivity growth in the United States over the past, say 78 years, there's basically been four periods.
One was about the 20 years immediately following World War II where it was fast productivity growth, 2.5-3% per year. So that's fast. And then starting in about 1973-74, there was a slowdown and that lasted about 20 years. During that period, productivity growth averaged about 1.5% per year. It was towards the end of that period actually where Robert Solo said his famous, "I see the computer age everywhere except in the productivity statistics." So it was a well-known phenomenon that we had slowed down since that earlier fast period.
But then in the mid-90s, productivity growth sped up again and people at the time hypothesized it was sort of the arrival of the IT age, that Solo maybe was just a little early and the computer age had arrived, it just arrived a little bit later. And I think as that period went on, people got more and more sure of that. And just about the time, I think, that consensus had solidified that yes, it was IT that had sparked the productivity growth renaissance, productivity growth slowed down again, and that was the mid-2000s. And it's been that low since.
Like you said, it was up towards 3% per year during the sort of IT boomish area. It's fallen down 1.2% since then. So you're talking over one half per cent year growth decline. Now, just to give you a sense of what that is, GDP right now is about $28 trillion. So 1%'s 280 billion. The way to think about productivity is how much a given worker produces over the course of a period. So if productivity growth is 1% faster in a year, that means we get an extra 280 billion of stuff out of the same workers as we would've otherwise. Okay? So 1 is what? $420 billion. And if you have just one year of that, okay, fine. I mean 420 billion is 1.5% is nothing to laugh at, but that can happen. But when it happens year after year after year, and as you mentioned, we're coming up towards 20 years of this slowdown now, it compounds. And now we're probably, well, it's 1.5% per year, 19 years compounded. That's over 30% less output per worker than we would've had or had productivity growth not slowed down in the mid-2000s.
Hal Weitzman: Right. When you say output per worker, but I mean that means we're all poorer.
Chad Syverson: Had productivity growth not slowed down, GDP per capita would be over 30% higher, which is about probably another $25,000 per person. So we're all $25,000 poorer than we would've been had productivity growth stayed at its sort of IT boom level.
Hal Weitzman: Okay. Now you talk about the IT boom, but even within the rest of the economy, not all the sectors are exactly the same. Right? Well, I guess IT has played a huge role in areas like manufacturing where productivity has been on the higher end, but you've written about construction, which have particularly low declining productivity. Tell us what happened there.
Chad Syverson: Yeah, construction's really the bad case in some sense. So measured productivity growth in construction since 1970 has been negative. Not just slow but positive, which is what the slowdown problem is. Productivity growth is still growing for the economy overall, but in construction, it's actually been negative. A given construction worker makes less and less per year than the year before. That's just highly unusual and certainly it's highly unusual to happen for a period of a couple years for any industry. But for it to happen for half a century is unprecedented.
So Austan Goolsbee and I are trying to figure out what's going on, how could it be so bad? And we looked at a number of issues and I think a fair assessment is we could rule some things out, but in the end it was hard to point to just one smoking gun and say, "That's the problem right there." It looks like the problem is more complex and a little more hidden in the data than that, which maybe isn't surprising if you have such bad performance for so long. If it were one simple trick, you would've figured it out. It seems like a real complex issue.
Hal Weitzman: And what about technology in construction? Because clearly technology has made some inroads into how we do construction now compared to 50 years ago. So does that mean that technology has sort of boosted, even though there's been a decline, it would've been more of a decline if we hadn't developed?
Chad Syverson: Yeah, I think that's what everyone likes to talk about. Nail guns. Okay, you can frame out a house a lot faster than you could before, with nail gun than with hammer. That's an example. But more generally, Austan and I found the amount of capital investment in the sector, it has not been any slower than the economy overall. So it's not that the sector has been falling behind because they refuse to invest in new capital, new technologies. To the contrary, it's despite that. So like you say, the performance might've been even worse if not for these changes in technology.
Hal Weitzman: And this wasn't just in the US, right, with the construction data across the world?
Chad Syverson: So if you look at data on, say, the OECD countries, which are kind of the wealthier economies. By far, I think three-fourths of those countries have had negative productivity growth in their construction sectors going back 25 years. We don't have data going back quite as far as for the US. But as far back as we can look, that's the same story. Negative productivity growth on average year over year.
Hal Weitzman: And is construction the worst? Are there other areas where you've seen declining productivity?
Chad Syverson: So there are other sectors that have poor performance for certain periods or even highly laggard performance for decades upon decades. But I think construction takes the prize for the worst.
Hal Weitzman: Okay. And is construction significant enough for the economy, you talked about that money, that 25 grand that we all don't have, is a large part of that because of construction.
Chad Syverson: Yeah. So if you can think of 4% of the overall economy is the construction sector, and we do a calculation in the paper that had the construction sector just achieved modest productivity growth, it would've added up to real measurable improvements in GDP. So it's one of the sources of the decline we were talking about.
Hal Weitzman: Yeah. Okay. Now we've all been through this revolution in the labor market through the pandemic. What happened to productivity during the pandemic?
Chad Syverson: So it's a little tricky when you have business cycle movements. In general, productivity statistics move around a lot. So productivity is kind of a famously difficult thing to measure. And so you get a lot of ups and downs that may or may not reflect true technological progress or the lack of, but rather the mix of workers who are working, which of course can change a lot over the course of a business cycle. Or of course during Covid when there was an initial period, heavy layoffs and so on and so forth. And then that can get reversed as you unwind that, et cetera.
And so if you just look at the actual numbers themselves what happened was, as we went into Covid, productivity actually went up very quickly. Now, most of that was probably a compositional effect, that the least productive workers were disproportionately likely to be laid off from their job. So you're only left with the most productive workers. So output's gone down, you're in a recession, but output per worker has actually gone up quite a bit because the workers who are left are among the most productive.
So you see this spike up and then that does gradually unwind as the economy recovers from the Covid recession. You're bringing more workers back into the labor force. They're less productive on average than the ones who were there when you hit the bottom. And that process took probably almost two years. And then starting almost a year ago, productivity growth turned around and accelerated again. And so right now we are basically back at, above or a little below depending on how you define it, the trend that existed before Covid. So we've gone through a pretty tumultuous up and then down and then back, but we're kind of back to where we would've been had the pre-Covid stuff continued.
Hal Weitzman: Which is around the 1.3%.
Chad Syverson: Which is useful to remember, was a pretty slow productivity growth. So it's not like we're ready to celebrate here. We're just back to the slow trend. Now, here's the thing. To get back to that trend after the sort of cycle I was talking about, you have to be faster than trend. As far as we know right now, we're still faster than trend. And the question I think going forward, is are we going to continue accelerating through that trend and therefore go into a new higher productivity growth regime? Or now that we're back on the old trend, are we going to level out at that level? And I think the next, let's call it year to year and a half are going to be critical in terms of evaluating which one of those two scenarios is taking place,
Hal Weitzman: Right. Because I mean, the big question is we hear so much about AI just like we heard so much about IT. There's more chatter about AI and more people using AI in some sense, but it's not necessarily at the moment leading to sustainably higher productivity. But you have this interesting research where you talk about how does productivity actually increase because of technology? And you talk about the J-curve. So at the bottom of the J, it's like an umbrella handle, right? At the bottom of the handle part, things are not really improving and then suddenly they improve very, very fast. And that happened with steam, it happened with electric motor, happened with combustion engine, it happened with IT spinners. You talk about, so it may be, right? Is it possible, would you say, that we are in the upward part of the J-curve for AI, that people are starting to spend, and those investments in other words, are starting to make us more productive?
Chad Syverson: I might say actually rather than on the upward part, we're still kind of near the bottom. So the J-curve story is initially you're doing all this stuff, building this intangible capital that's going to be really useful, and this ties to AI, that kind of thing, but you're just getting ready to use it to make stuff. And I think that's kind of what it feels like we're in right now. We hear a lot about, "Oh, look at this new gizmo, think about how you might be able to use it." And we hear about companies trying to figure out how they're going to be able to use it. That's actually the process that happens at the front end of the J-curve. And then once it's implemented as actual input making stuff all over the place, that's when you get the acceleration.
Hal Weitzman: So it's not yet driving the increases.
Chad Syverson: Yeah, I think that's right. And other people might disagree with me, but my own opinion is the productivity growth of the past few quarters isn't really AI driven per se. I think that's going on underneath. And in fact, if our J-curve story is right, like I said, if anything it's kind of pulling down the measured productivity growth beyond what it actually would be. So it might even be better than it is. What I think the actual measured productivity growth of the past few quarters has been about, I think largely is a return of a kind of dynamism to the economy, and in particular in the labor market, but not exclusively. A dynamism that, by the way, had been declining for decades before that and indeed was one of the hypothesized sources of the productivity slowdown we were talking about.
So what happened coming out of Covid, if you go look in the data, there were a few things. One, the number of workers who were separating from their jobs, voluntarily quits it's called in the data, was actually an all time high. And I view quits as a sign of people are moving to something they'd rather do instead. We know that's correlated with wages, and we know that wages are correlated with productivity. I view this unprecedentedly high quit rate as being indicative of people who were moving to more productive jobs at a rate that they hadn't before.
Hal Weitzman: So the Great Resignation led to this increase in productivity?
Chad Syverson: Yeah, I think that's right. And you also see sort of confirmatory or maybe complementary evidence to that, not in the labor market, but in the business formation statistics. So another thing that's been going on since we've been emerging from Covid is business formation accounts in the US have gone up 30% relative to their pre-Covid level. So businesses are forming at an unprecedented rate. It's not just people without knowing what to do out of desperation form some consulting company in their spare bedroom, or trying to make do kind of businesses. If you look specifically at the types of businesses we know are the most likely to grow early on, those formations have also increased by 30%.
So there's a couple signs. A good one is if you start a business and you already have a paid employee. Businesses starting with a paid employee, that's gone up 30% as well. There's another set of markers. People run models using past business formations. Here are the attributes that predict future growth. So if you take those models, they also suggest that those types of businesses, the formation counts on those has been going up as well.
Hal Weitzman: Okay, so we've talked about two measures of productivity that are looking promising. One is the quit rate and one is business formations. What else is leading to this?
Chad Syverson: To be clear, those aren't direct measures of productivity, but measures of dynamism that I think are reflected in the productivity statistics.
Hal Weitzman: Got it.
Chad Syverson: Another common measure of dynamism are gross flow rates, especially in the labor market. So to give you an example what I'm talking about. Every month we hear about the job formation statistics at the beginning of the month. Last month, 200,000 jobs, say, were created. That's a typical number in the US. 200,000 jobs created in the past month. Well, that 200,000 jobs created is actually a net number. What that net number reflects is on average 6.1 million jobs were created and 5.9 million jobs were destroyed. Okay? So there is constantly a massive amount of churn going on underneath the surface in the economy. To get that small amount of net growth, there was huge gross growth and huge gross decline. Some businesses grew, some businesses declined. It nets out to 200,000.
Now, that sort of gross churn relative to the net rate is one measure of the dynamism of an economy. Okay? So how quickly, how efficiently the economy takes activity from one location moves it to another location. It could be industry, it could be geography, whatever. How fast that happens is a measure of how good the economy is at responding to changes in people's tastes, or costs of various inputs, things like that. Just like those other metrics we were talking about, like business formation and the quit rate, what had been declining for decades, those gross flows have been declining for decades as well. They also turned around coming out of the Covid recession. So you saw a return to a faster churn rate in the labor market.
Hal Weitzman: So more churn is good.
Chad Syverson: Generally more churn is good because we know on average churn is correlated with productivity growth, that on average workers move from less productive to more productive jobs.
Hal Weitzman: Even if they're fired, regardless of whether they're fired or whether they quit?
Chad Syverson: They're more likely to move to more productive jobs when they quit than when they're fired. That's why I think the quit rate is a good measure of it. But overall, on average, the typical labor market movement is going to be from a less productive to a more productive job.
Now, I actually have some work on that with co-authors in Chile, and we find, like I said, the average movement is from a less productive to a more productive job, but it's just a little bit. So it's about 52% of workers move from a less productive to a more productive company. 48% go the other way. So on net you get gains, and if you could get more churn, you'll get more gain. But there's a whole lot of churn to get those gains. And one lesson I guess is if you can improve, sort of get that 52% to 54%, we could do some calculations, you get huge productivity growth effects from just getting that churn to work a little better. And that's why I think the quit rate is useful because, as I was just saying, quits are associated with more of this positive churn towards more productive growth. So dialing that up a little bit can have real effects on the aggregate productivity numbers.
Hal Weitzman: So if I wanted to track whether we really are heading up into the next up part of the J-curve, I wouldn't look at companies firing people. If I see companies firing people, it doesn't necessarily mean anything, right?
Chad Syverson: No.
Hal Weitzman: They may feel that they're more efficient and productive, but officially they're not more productive, that's what you're saying?
Chad Syverson: Well look, all else equal, if you can make exactly the same stuff with fewer workers, your productivity goes up. But that's an accounting identity. That doesn't mean if you just get rid of your workers, you'll become more productive, because those workers were making something, and if you suffer a decline in output that's as great or greater than the fraction of workers you're firing, your productivity is going to go down. Okay?
Hal Weitzman: So firings are not going to tell us very much.
Chad Syverson: That's right.
Hal Weitzman: Okay. So what is going to tell us is things like quit rate.
Chad Syverson: Yep.
Hal Weitzman: Business formation, and-
Chad Syverson: Labor market churn.
Hal Weitzman: And then churn.
Chad Syverson: Yeah.
Hal Weitzman: So if we track those, what is the point at which we can say, "Yes, now we've reached the promised land." And what will that be? Will it be 3% again? Is that kind of the upper bound?
Chad Syverson: If we get back to 3% per year, that would be fantastic. That's historically speaking. And we've got data going back 125ish years or something like that. Now that's about as good as you can sustain over a five to 10 year period, is 3% per year. So if we get back to that, that would be fantastic. If we get past that, that will be unprecedented in the history. It'd be really fantastic, a fantastic issue to have to face.
I have said elsewhere I think, I would be disappointed if ultimately AI and its associated technologies didn't improve productivity growth by a half a percent per year. I'd be disappointed if it were less. I'd be surprised if it were more than 1.5% percent per year. So that's kind of getting close to that 3% number. So I guess if you want to say, I sort of expect it to be about 1% per year, that's kind of where I'm at, which would be great. To go from 1.2% per year to 2.2 for a decade or more, that would make a real difference. Maybe we'll do better. Maybe we'll be disappointed. We'll see. But that's kind of where I think we're at.
Now, when are we going to know AI is driving things? I think it's going to be a few years yet. The nature of the J-curve is it's really hard to measure anything about it when you're in it, because the core of the J-curve story, not to go too far into the weeds on this, is it's about intangibles. It's about capital that you're making at the beginning, and you should have counted it as an output, but you're not, and therefore you're understating productivity. And then you're using that capital to make stuff later, and now you're undercounting inputs and productivity is not as high as you think it is.
Intangibles by definition aren't very measurable. The trick we did in the paper to look at sort of past J-curves was to take things that folks at one time couldn't measure well, but then we went back and did measure them. We basically said, well, we know these things at one time could have been thought of as intangibles. We see them now, but let's pretend we couldn't then, and then we can actually run out the numbers.
But if you think, how are we going to do that with AI? Well, we don't measure the intangibles yet with AI. We're trying. Some of my co-authors are making real efforts trying to do that, but it's hard to do that, and it's really hard to do it in real time. I'm going to be honest, we're not going to know the J-curve has happened until I think we're well into the process of coming out of it. All that said, if you talk to me three years from now and productivity growth, measured productivity growth, has continued at the pace of the last year, which is like 2.5, 2.6% per year. I'll say, "Okay, this is more than just the return of churn and dynamism. This is something else." And I'd start putting real money on the J-curve emergence story.
Hal Weitzman: Okay. Well, I want to have you come back and talk more about intangibles, but I've got one final question for you, which is when do I get my 25 grand back? Or when do I get my 25 grand? I never got it.
Chad Syverson: Well, yeah, so you can get it as soon as we have 19 years of 3% productivity growth. So if we can turn things around, we've got 18 more years left.
Hal Weitzman: And how long does that part normally last?
Chad Syverson: Well, typically when we get accelerations, they usually last about a decade, maybe a decade and a half, so.
Hal Weitzman: I'm never going to get it.
Chad Syverson: Well, let's not be too pessimistic. Maybe AI will be unprecedented.
Hal Weitzman: I would settle for 10. I would settle for 10.
Chad Syverson: We can get you 10 with a return to 3% per year.
Hal Weitzman: Excellent. All right, Chad Syverson, thank you so much for coming on the Chicago Booth Review Podcast.
Chad Syverson: My pleasure. You're welcome.
Hal Weitzman: That's it for this episode. To learn more, visit our website at chicagobooth.edu/review. When you're there, sign up for our weekly newsletter so you never miss the latest in business-focused academic research. This episode was produced by Josh Stunkel. If you enjoyed it, please subscribe, and please do leave us a five-star review. Until next time, I'm Hal Weitzman. Thanks for listening to the Chicago Booth Review Podcast.
Chicago Booth’s Matthew J. Notowidigdo discusses what it would take for the US health system to produce better outcomes.
How Can We Fix US Healthcare?Chicago Booth’s Sam Peltzman analyzed nearly 50 years of data to identify the factors most associated with happiness.
Infographic: Which Americans Are Happiest?Ralph Nader joins the podcast to discuss ethically profitable business, the political disillusionment of the public, and the state of capitalism today.
Capitalisn’t: Ralph Nader’s CapitalismYour Privacy
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.