4th Health Center Conference: The Rise of Healthcare Jobs ? Keynote

[music] >> Actually, I have a >> Okay, good morning everybody. >> budget constraint on hours. >> I'm really winning in Toulouse at the Toulouse School of Economics. So, on behalf of Angie Aquatella and Jean Tirole as the co-organizers of this conference with me, I would like to welcome you for this fourth edition. Um I think your presence today reflects the vitality and growing importance of our work in the addressing key questions and challenges in the health economics. So, over the next 2 days, we are going to have a very rich program structured about around scientific session that cover many field, pharmaceutical innovation, health system organization, access to care regulation, health data, financing of innovation. And at the heart of the of this edition, well, we have some highlights with two keynote lectures delivered by leading experts. Uh Neil Mahoney this morning and Jonathan Gruber tomorrow. As well as round table that are going to gather distinguished guests that you will see today just after lunch and that will shed light on the today's strategic challenges about the most favored nation policy in the US. And this will complement a broader set of scientific sessions of outstanding quality. So, um a few practical points. So, at 11:45, we will have a group photo just outside the university at the entrance. So, uh just go there at the end of the of that session quickly so that we can have the efficiently a group photo before we we we go on. Uh and and tonight for all those that are invited, we have this cocktail dinner at the Musée des Abattoirs at 7:00 p.m. And we we This is walking distance. We finish at 6:10, so you should have enough time. I would like also to thank the CPR with sponsoring selection of young researchers for attending this and presenting at this conference. And I also would like to thank the the funding of the European Research Council and the partners of the of the TSE Health Center. So the partners of TSE Health Center are Biomérieux, Clinique Pasteur, GIE Data, LEM, EUCT Oncopole as well as the BPI France through the framework project of Arpège. And I also would like to to thank very warmly our assistants Valérie Novakhine and Marie-Hélène Dufour who has done a amazing job in preparing this this conference. Finally, do not hesitate to learn about our activities at the Health Center visiting the website. We have a lot of things including of course scientific publication, but also a lot of policy focused papers. And I would like to mention that next year we'll have the fifth conference which will be on the 16th 17th of June of 27. And hope many of you will come back. So now, thanks a lot for your engagement, your high-quality discussions and I wish you a successful conference and will give the the mic to Jean Tirole who's going to share the first session. >> [music] >> And good morning and thanks so much for being with us today and tomorrow. It's a great pleasure to welcome you and to welcome Neil Mahoney who is a well-known. I'm not going to describe his work. You all know of it and and it's a great, great pleasure to have him today. So, let's without without further ado, let's let's start this session. And Neil, it's the floor is yours. >> Uh >> Um >> I'm Michael Powell. >> We have If you can leave a few minutes for the for discussions. >> Yeah. And Sorry. I don't need both of these. Let me put them down. It's enough. Um Thank you, John. Thank you, Pierre. It is a delight to be here. Um So, I'm going to Guess as is the norm present a paper in in broader context and seeing some good friends here as well. So, uh extra delight. Um so, this is joint with Josh Gottlieb, uh Kevin Renz, Victoria. Um Josh has been studying health care labor markets for a long time. And when he originally came to me, I said, "Well, the the point of the health care sector is not to create jobs. Uh we should study health outcomes and to the extent employment is an input, then we can study it. But let's not focus on on studying jobs." But over time he convinced me two things. One, that health care employment in the US was so large that if you want to understand the US labor market, you need to understand health care employment. And second, that for political economy reasons, also for health care reasons, that understanding the health care labor market was crucial to studying uh the sector and I think the more welfare relevant outcomes. So, in terms of table setting, here's a chart of employment in selected sectors in the United States uh including the largest ones over time. Uh you know, when I was born, uh you know, when I was a kid, uh manufacturing was still the largest uh sector by employment. Uh and I think everybody knows the story. There's been many papers, newspaper articles, political campaign the decline in manufacturing in the United States. There was a short window of time when retail was the largest employer in the United States. Uh but healthcare employment, we don't see lines like this uh in economic data. Healthcare employment has grown remarkably linearly through a number of recessions, right? This was a tremendously deep uh recession in the United States with a well-known jobless uh recovery, has grown uh inexorably uh for for generations. Uh there was, like the rest of the economy, there were furloughs uh during COVID, but has come back to trend. In fact, has shot through trend. And so, I do a lot of media work on how healthcare has been contributing to more than 100% of job growth uh over recent years. That is, if you took out healthcare, the job growth in the United States would be negative. Now, when I talk about uh a rebound from the COVID shortfall, I talk about a rotation away from um nursing homes towards home health coming out of COVID. People have shown less preference to be in institutional facilities and more preference to age in their homes. There is good work not yet uh circulating on the role of the 85 plus year old population in driving these trends. But this is just all to say that health care employment is hugely employment important in the United States, uh but our view was had received uh relatively less attention and so we wanted to contribute to a literature that uh sought to understand it. Uh And you know, I think the a through line through this presentation will be asking the question of is health care different? Uh and maybe a little bit on why, but mostly just on facts. Uh so the the main story lines for the US labor market are well known. Manufacturing decline, automation, uh a polarizing wage distribution, you know, associated with the work uh of Autor and others. Uh health care could be different for lots of reasons. Um it's been growing quickly, it's heavily regulated, uh it's geographically dispersed uh in in ways that we can talk about more because of the funding mechanisms. Uh so, you know, this is a sort of who, what, where, when, why type paper. Um what types of jobs are emerging, who gets them, how do they pay, and where and where are they created? Uh and so hopefully I will whet your interest in these topics um as we go. Here, do I take questions as they come at the end. I'm happy to take I I start feeling a little awkward when there are no questions after a few minutes. Uh it's probably because I spent too much time in the University of Chicago. Um so, give you an overview talk a little bit about who I'm going to talk about the rise. I will unpack these terms uh but I'll tell you about the rise of a occupation called mid-levels or advanced practice professionals and then uh tell you a little bit about where. Uh uh uh uh All of what I'm going to show you I I think is supported by these data sets. The main data source uh Kevin was uh and Victoria is still is an employee of the census, so we're using restricted access census ACS data and that gives us to access to non-top coded 1 second uh wages, which is important for physicians, and more geographic granularity. Yes. Uh I don't know or and and I certainly don't know whether these patterns are apparent elsewhere. Um I think that's probably 70% my ignorance uh but but might be because we are still developing our our understanding uh of healthcare labor markets. So, these data we bunch of other ancillary data sets uh that are I think fairly common or use training program graduates just to paint a more holistic picture. Healthcare jobs, the We spent a bunch of time thinking about what is the best way to categorize the data, and we landed on occupations of because we think it's natural and it's informative. Uh so, I think everybody is familiar with the occupation of a doctor. Uh the other occupations may be less familiar. I'll just quickly Hopefully, there isn't too much interference. I'll quickly review them now and and give you a couple data points on how they have evolved. Aids are the lowest training occupation uh which you can perform with a high school uh degree and a certificate, which takes weeks to months. Uh they're low paid. Um you know, people churn in and out of them. Uh the next level is a licensed practical nurse or licensed vocational nurse. Uh that is a position you can take with a 1-year certificate, often at a um an associate's degree. Um The next in the vast majority of employment is uh registered nurses. Uh so, that is an associate's degree or a bachelor's degree, a two- or four-year degree uh after high school. That role has evolved tremendously over time. Right? At a point in time, nurses were providing bedside comfort, maybe checking in on you, and providing medications. Uh today, a nurse, if you think about an ICU nurse, is using advanced equipment in a high-stakes setting, uh you know, involved at the cutting edge of patient care. And accordingly, uh the you know, skills requirement uh uh, nurses increasingly have a four-year degree and a nursing degree is a stepping stone to even uh, more advanced uh, mid-level occupations. So, what are these mid-level or advanced practice professional occupation there is a alphabet soup of them. Uh, but uh, you know, the main ones are nurse practitioners and physician assistants. Uh, and those are earned with a master's degree on top of an undergrad or sometimes even uh, a D on top of an undergrad degree and then they are there CRNAs and certified nurse midwives which are also earned with an advanced degree. Um, so these are positions that take uh, 6 to 8 years in training after high school. So, 4 years of an undergrad degree, maybe 2 years of a master's or or 4 years of a PhD. That is less than the 11 to 15 years that a physician needs uh, for their training. 4 years of undergrad, 4 years of med school uh, and a fellowship possibly a second fellowship. Uh, but still substantial uh, training. So, that's sort of some table setting. What's happened to employment across these different groups? Uh, I'm going to show you a bunch of charts like this which are designed to convey a lot of information though. X-axis is employment on a I guess a log base two scale. Uh, here I have annual growth and these lines show the starting value and the ending value. And so, the length of them is the proportional growth given the scale. Um, so you already know about growth in health care relative to the rest of the economy, uh it's been growing at about uh twice the rate as the rest of the economy. The fastest growth Mid-levels, they were so small of a category that they weren't even included in as a category in the data before 2010 or not all the categories were included. Over the time period we have visibility, they've been growing at almost 7% per year. Uh So, the fastest of them all uh with you know, about 2% growth for the other clinical occupations. Um What about earnings? Um Growth in earnings has been almost twice as fast as the total economy, 1.1 versus 7. Uh Across occupations, we see this inverse U-shape growth with the fastest growth of wages for nurses and for mid-levels, slower growth for doctors and for aids. Um So, here's another way of looking at Uh I think most of you are familiar with these sort of charts from David Autor and colleagues showing wage growth by percentile of the income distribution. Uh And this is between uh 1980 and 2022, so over a uh 42-year period, roughly my lifetime. Uh And what I think is is well known is that uh this is real growth. That outside of the top, the real growth has been anemic. Uh What's interesting is health care has been an exception to this trend. There's actually the largest gap between non-health care and health care in the 50th to the 95th percentile. Uh and so healthcare has been the sector where you know, the middle class, upper middle class wage growth has been strongest. You know, doctors have seen strong wage growth, not as strong as some sectors, so that's why there is But but really strong wage growth. And what are the occupations in which this is occurring? Well, if you look at the you know, fifth to the 90th percentile of the wage distribution, this is nurses. All right, so so nurses are uh you know, a great middle class, upper middle class job in the United States. They have emerged as such. Uh and you know, that conflicts with some of this sort of nostalgia view of the manufacturing male breadwinner in the United States. Um but but these are the facts. Like them or not. I see a hand. Yes. Uh I'm going to show you in two slides. Uh His voice is hoarse cuz he's been celebrating the Knicks victory. Uh so he's a good excuse. So, one slide. Sorry. Um So, healthcare across this 42-year time period has stayed at 80% female, unlike the total economy which has gone from a little bit over under 50 to a little bit over 50% of female. Uh but what's happened in healthcare is we've seen this uh I don't know if you want to call it massive, but material convergence in female shares across occupations. At the beginning of this time period, very few female physicians and virtually 98% of nurses were female. Uh and over time you've seen a huge increase in female physicians. Right, the change in the flows you need to get the stocks to get close to 50% is massive. And you've seen a modest decrease in female nurses and aides such that the aggregate has stayed exactly the same. No deep reason, uh but because of the higher share in these groups, uh we've seen convergence or a flattening sort of of the gender occupation gradient. Uh Other facts, uh foreign-born population, that's the uh best metric we have for for immigrants in these data that we can measure over time. Uh and health care and non-health care have seen, you know, similar increases. In fact, it's been less pronounced in health care. Uh Where we've seen it in health care, now this is a U-shape pattern, uh a growing proportion of foreign-born doctors and aides and nurses, uh but less in the middle. Right, so these uh where you've seen the strong and wage growth is actually where you've seen the lowest foreign foreign share. And that's going to connect with some of the geographic patterns I'm going to show you as well is that a lot of these jobs are in Appalachia and in the Midwest. Yes. >> For physicians, do they have to recertify in the in the US? Do they have to basically do their studies twice? >> Uh so, not all foreign-born, but uh people who come to the United States uh after med school do have to go through much of their training again. And I I would need to go through exactly uh one of the different pathways. But yes, there are uh both sort of high bar in terms of exams and then also uh yeah, significant time required. Uh So so Another fact, I think like the economy is a There's been increasing education across the board, uh you know, except for doctors, which already sort of max out on education. Uh and the the wage trends I show you are not being affected by that. But this is just something that's going on underneath the hood. Uh So those were some some basic facts. Uh they may be I'm sure there are is together sort of holistically. What I want to spend some time on is the rise of mid-levels. And uh uh something that undergirds all of this is, you know, the United States, I think like many countries, has uh a variety of binding caps on the production training of doctors. Uh we have restrictions on the slots in medical schools, uh slots in residency programs, uh and you know, various ones are buying different binding constraints over time. Uh but given that, I this is a sort of long-standing uh concern, then uh the market evolves. And in particular, it's evolved through the rising number of mid-levels, the rising sophistication of the tasks they do, increasing pay in this group, where mid-levels now I'll show you some stats, but I have read they provide more than 50% of pediatric so of care for kids in the United States. And if you're not someone like me who's like paying attention and reading the initials on someone's badge when they see your kids at the doctors, you may not even know. But they are they're 50% to understand the sector without understanding that I want to dig in a little bit more. So, here's you know, lots of ways of thinking about economists don't like to think about you know, supply not keeping up with demand. It's a weird concept for us, but all non-economists think about the world that way. And so, you know, I am going to use that language here where the number of physicians so these are normalized to zero in the base year 2005 has grown only slightly since this time period. Healthcare use, you know, is driven by people over 65 and people over 85 and that population has almost two x over that time period. Number of doctors has increased Doctors per capita has only increased 20% and so there's just increasing demand which cannot be fulfilled positions unless they were working much longer days or having much shorter appointments. Uh What you've seen at the same time, and you know, I I previewed this as well, is the, let's see, the red line is the flow rate of new doctors. And the blue and the gray line are the flow rates of um nurse practitioners and physician physician assistants, who have not perfect, but have large overlap in the tasks they can provide. Uh under Medicare, they bill 80 cents on the dollar of doctors. And you can see the combined flows of the blue and the gray line are now significantly higher than the red line. Yes. >> Um I think but I got the side of the coordinated that they're being a supply response to the educational sector in saving all the time but this is still education based occupations. >> Yeah, so a supply response and we spent a lot of time figuring out whether there was a, I guess we were looking for an instrument for supply. Uh but, you know, increases in these programs, increases in enrollment programs over time, and just everything was increasing so fast everywhere, sort of all at once, so we didn't have an instrument. But, you know, the Yeah, the underlying these charts are huge increases in uh the supply of education. Yes. >> Yes, uh I think we have a similar finding friends or several things, but uh so you think there's a family parallel that can be done? We see also uh higher turnover rates because the more you get into education, the less it is about a calling at times if you want to be simplistic in the approach from nurses. So, do you see also a higher turnover rate for those new uh like registered nurses or mid-level type of uh level people? >> Uh I don't know. It's a great question. I think we we've been looking at of cross sections over time rather than following people and looking hazards sort of out of the profession. Uh, but I will make a mental note cuz uh I think doctor burnout is been a issue that has received a lot of attention and understand uh it's well known that there's a lot of churn among aids in the US that that is a uh stressful low paid occupation and people are turning in and out. Uh but I haven't looked at it across different occupations. Uh it's a interesting question. >> I'm mentioning that in the same phrase used to say that the uh the average uh time for a nurse in the in the career is about like before used to be 10 15 years. Now it's roughly more 6 years. >> Mhm. >> You require higher education to get there and you want to stay in job 6 years. So in terms of the stability of the overall system you have to be thinking that. >> Yeah. No, I I think it's certainly something that we should look into and uh Yeah, I'm sort of embarrassed for having not looked into it and now that I realize it it's interesting. So thank you for that. Um Yes. >> What explains the giant drop in the new doctors? >> What explains the giant drop line? My guess it's actually a data issue, but I would need to go back. Uh that would be the most likely explanation for something like that. Thank you. Um So here's a figure I told you about the pediatric visits. Here's some other cuts uh using the market scan data. This is the share of mid levels providing uh anesthesia services. Uh here's the shares providing primary uh care and overall encounters. You see these increasing for these categories it's not 50% now but with the ratio of flows it is if they persist it's only a question of time until the sort of stocks of care start catching up. So we did a deep dive I don't know if this will ever end up in a paper. We did a deep dive into CRNAs there was a lot of conversation around CRNAs during COVID because of their roles in COVID care. Um Let me actually show you this one. This one is showing um parity care provided by CRNAs over time. The solid lines are high severity care provided over time and as you see there's both a an increase CRNAs are more important in rural areas. I'll preview something um doctors like to live on the coasts in the United States and so there tends to be a larger portion of supply in rural from from mid-levels so they're providing an increasing fraction of care. They're working up the severity distribution. Uh not the most elegant way of showing that but this is exactly sort of what you would imagine I think based on first principles but it's also born out starkly in the data. Uh so I guess here's what I was previewing um physicians per 100,000 residents. You see a lot of doctors like to stay where they were trained you know in the Boston New York BosWash corridor it's all a corridor area on the coast there's more of them in Hawaii understandably fewer of them in Alaska uh and this is NPs where we have more visibility uh but much higher fraction of them in Appalachia, in Maine, you know, in the uh sort of upper Midwest. Like so, again, uh not the market is working like this. So, one factor that, you know, we dug into for a variety of reasons, we thought it was an interesting policy lever, we thought it was an interesting form of variation which could teach us about the substitution patterns uh and, you know, I had a student who did a lot of the spade work here was to understand scope of practice laws. So, there's a set of laws at the state level which determines the scope, the set of activities that mid-levels can do. And in particular, sort of it provides a an upper constraint on how much supervision, how advanced uh the uh types of care they can provide. Uh and over time, in particular in places, you know, with less physician supply, so not in Massachusetts but in Maine, you know, in the upper Midwest, in Appalachia, you've seen a broadening of the scope of practice. Uh and so, we wanted to see how much of what of the trends is driven by these policy changes. And so, what you do when you have state variation over time, you run uh sort of year state diff-in-diff regressions uh with uh some measure of the regulations on the right-hand side uh and some measure of the number of uh mid-levels on the left-hand side. So, that is what we did uh uh you know, lots of people have done this in various ways. Uh but without even showing you the regression first, here the trends for places that, you know, were never treated, always treated, and had changes in scope of practice over a time period. And this illustrates what's happening here is the share of mid-levels is increasing rapidly everywhere. Uh and so there's some diff-in-diff effects if you squint. Uh and you know, people use different research designs and different controls, and you can show impacts of these laws, but it's in the context of this. Right? Where where everybody is vastly, rapidly increasing uh mid-levels. And so, like it's just second, third order. Right? And so, we we do the calculation uh you know, with our estimates in treated states, scope of practice explained 5% of the growth. And because treated states are only a small fraction of states, overall, they're a tiny fraction of the growth. So, it is real. You know, I do policy work, we talk about, you know, uh providing federal incentives for states to expand their scope of practice. Uh I don't see this as a huge binding constraint. Yeah. >> So, uh the quality of care provided by the mid-levels versus other >> I I'm I'm not in this project. There is There's research out there which finds uh conflicting effects, and there's probably file drawer uh consequences for all the no-effect papers in that space. Uh I don't think we So, so one one result by, I guess, my former colleague, David Chan, is mid-levels order more tests. You know, whether that's because they feel less confident about their abilities, because they feel like they have to prove it in a workplace where they get less respect, but they order more tests. Tests are more expensive, so that generates higher costs for similar outcomes. There's other research which shows actually no difference. Uh, I don't think there's one answer. Like there's operating lots of environments. John, you may know more than I do. >> [snorts] >> Uh, I haven't seen any research. It's I think a very plausible theory. Um, yeah, thinking about the just production function with telehealth with different occupations is a like good area for research sort of combines personnel economics and uh, I like it. Are there grad students here? Yeah. That's what That's what I was thinking about is there Yeah, I I thought I could leave Silicon Valley and escape the conversations about AI, but apparently it's not true. So, I have one bonus slide at the end that speaks to this, and so maybe I'll I'll try and wrap up, you know, so we can just then we can have a conversation. I think that's that will ground that conversation in one fact uh, that I saw. So, where are the jobs created? So, we this is actually the heart of of a research paper we have in this space. Um And this was sort of fascinating to us because we think there's a there's a sort of a powerful null hypothesis here. Um I think we're all aware as I I spoke about the the narrative of declining industrial towns in the United States. Uh and the fact that some of them have successfully pivoted from manufacturing to meds or maybe to meds and and eds to medicine and education. You know, prominent examples Pittsburgh, Pennsylvania with their healthcare systems, Cleveland, Ohio with the Cleveland Clinic, Rochester, New York, former home of Kodak. Well, and uh you know, we've had this back and forth and honor on um whether and to what extent uh a gains in education and healthcare and I can sort of uh I think we've together jointly figured out uh sort of come to the same view on what's going on. Uh but I can speak to this. Um But you know, the question we we asked is have places hit hardest by manufacturing decline seen a disproportionate growth in healthcare jobs? And why could be that be the case? Well, one there's these salient examples, right? And and mayors and planning commissions saying we need to invest in technology of the future. Uh we as economists know it's hard to do. Uh The second is you know, funding and demographics. Uh when manufacturing jobs decline, uh often sort of young people leave There's a story that young people leave those communities. And who's left? Older people who consume a lot of health care, which is funded by the federal government. So, you can easily imagine situation where there's been an increase in services demand towards health care, and the uh reduction in the local economy has not carried over to health care because the funding there comes from the federal government. And so, that's a that's a story where these are then the driver of local economies. So, so have the places that have been hardest hit hardest by manufacturing decline seen disproportionate growth in health care jobs. Uh and I think like many things, you know, the answer depends upon your benchmark. Uh and so, we can quibble about benchmarks. Uh but we thought the reasonable benchmark here was uh has health care grown faster uh as a share of its level that of its national level in those places than in other places, right? Has it grown disproportionately quickly? Uh so, we think we also did the the China shock. Uh but again, I guess we weren't playing the identification in our paper. So, we wanted to use sort of all of the variation in the data. We think that allows us to speak, you know, to the broadest uh descriptive fact. So, this is a chart. Uh actually, it's a manufacturing share of in and the change in manufacturing employment between 1980 and uh 19 or 2022. And what you see, this is a slope of almost 2/3. Basically every place lost roughly pretty close to 2/3 of their manufacturing jobs. And that if you were at 30%, you lost 20%. If you were at 10%, you lost sorry, if you were in the baseline manufacturing share of 20%, you lost 10%, 12%. Like that sort of proportional everywhere about 2/3. So let's look at those places as a function of their baseline share and see what their growth is in health care employment. And the here which is of every 100 manufacturing jobs, 11 jobs are gained in health care. Now, is that big or small? Today health care is 10.8% of employment. So what does it tell us? It tells us that health care is growing but in these places at or is growing is capturing the same fraction of employment as it is in the overall economy. So so our view is it's not growing disproportionately which is growing at the rate you might expect. Here's another version of this. It's it's capturing jobs, but it is not it's not disproportionately offsetting manufacturing jobs more than retail, more than financial services, more than anything else. Yes. >> Uh what are the jobs, right? Else yeah, jobs. So, you know, no more jobs. >> Yeah. >> So, you would not expect them to do well monthly and have the back like green mosses, right? >> Still >> exported to other regions. >> Exactly. So, I I think that's where we ended up. Uh and the reason why we thought the alternative hypothesis was credible was one, the specific examples of Pittsburgh and Cleveland that were heralded in the press of cities reinventing themselves around industries of the future. And you know, those are outliers. Right? And you know, you might have known that all along. Uh it took us doing the analysis to figure that out. Uh the second was, you know, this shift in demographics where you know, you see a more elderly population where that money for health care is coming in externally. So, you can easily tell a story where the young people leave, the old people have federal money to buy health care. And so, the whole the remaining cities are old people and the young people provide them health care services. And there are certainly examples of that. You know, a city here, a city there, but comprehensively, that doesn't seem to be the story. Uh so, yes, maybe ex post obvious, but we felt like there was this not a straw man, but a steel man argument that it could have gone in the other direction. So, some other trends, well, I think we'll have time to talk about AI or or whatever you want to. Uh You know, despite you know, the sort of at baseline uh aggregate trends, you see very different patterns across different demographic groups in ways which you know, conform probably with stereotypes that uh So, this is the increase in health care employment over the decline in manufacturing. And health care, you know, for for males when you think about uh white and and black males working in manufacturing uh those groups are not seeing large increases in in health care employment. There are very few manufacturing workers that are retraining, male manufacturing workers that are retraining as as nurses. Again, right, that's consistent with, you know, everything we've seen uh on the non-economic outcomes for for these groups, from deaths of despair to divorce to uh other um non-economic outcomes here. Uh you know, where you have seen uh where there was relatively less manufacturing employment and very strong health care employment is uh women, black and white women. Right, and so uh while there health care has not been disproportionately offsetting manufacturing jobs uh in these communities, you have seen this pivot from manufacturing breadwinners to nurses and physician assistant and uh are a nurse and other mid-levels being the breadwinners in those communities. Right, and if you've, you know, read uh J.D. Vance or others in narratives on the Midwest, right, that story uh is is evident there as well where the shifting and the shifting I think gender dynamics are also important for thinking about the sociology in these communities as well. So limited evidence of disproportionate shift from manufacturing to health care. Health care employment growth is widespread nationally slightly faster faster in manufacturing areas but offsets only 11% manufacturing job loss. Stronger offsets for for women. So the anchor institutions the Cleveland clinics Pittsburgh other places are not anchoring abroad Rust Belt employment revival. So Maybe I went faster and have fewer questions up front than I anticipated. So let me conclude and then say a couple words on AI with the slide I promised Jean. Health care is a exception. So I still think the the point of our health care sector is not to provide jobs but from a political economy perspective health care is an important employer in virtually every House members district. I probably every House member has at least one hospital in their district. So they think about it as a source of jobs despite us telling otherwise. So that's important. It is hugely important for the US labor market. It's where a lot of the growth is. We'll talk about AI. And you know it's been a driver of a lot of these big structural trends we've seen. Health care is you know it's stayed 80% female but we've seen this flattening of the gender gradient across occupations. Uh more than 50% of uh female or of med school graduates are female now, and that's doing a lot of work on on the stocks of female doctors. Uh see, some more female nurses, but that's slow moving. It's foreign-born, it's educated, it's where you see middle-class job growth. Uh mid-levels are the tip of the spear here, uh and the flow is faster than the flow of doctors. Uh so, increasingly, when you get primary care in the United States, you're being seen by a mid-level. Uh nationally, though, uh it is not offsetting the declines we've seen uh in in former uh steel towns uh and other industrial sectors. Yes. >> Obviously, Medicare is a big share of the federal budget, um and an expensive but what I wonder is, you know, you've shown us uh that there's some redistribution going on uh through the health care job sector. What I wonder if is, does this come up in the in the congressional debates, like in CBO reports, that there is some redistribution happening uh through this federally funded federally funded created uh health care >> Yeah. So, I don't know in CBO reports, sort of breakdowns of inequality, uh we hadn't seen a chart like the sort of percentile chart we produced before we produced it. Uh but I mean, certainly, it's front-of-mind, right, in the policy process, that, you know, all of the discussions that we like to focus on health outcomes, on federal spending, uh where the rubber hits the road for uh a member of Congress is what is this going to do to jobs in my district. Um and you know, you can try and steer them away from that, but that's often how they think about it and and partly it's because the unions are in their ear, but it partly it's because they see this as sort of driving good employment uh and and given the setup in the US with disproportionate representation from the you know, upper middle S and other places with lower population, I think these issues punch above their weight. Um So on the AI point, uh so this was just uh put out by folks at Goldman Sachs. So everybody's hunting for There's a lot of talk and and a lot of evidence about AI job displacement. Right? And so there's probably half dozen uh AI displacement risk scores that people make uh and they produce plots of, you know, employment and youth employment and wages by these risk scores. You've probably seen lots of them. Uh the Anthropic folks put them out, Goldman puts them out. Um Let's see, Erik Brynjolfsson, my colleague at Stanford, puts them out. Uh This is one of the first that actually shows a slope. And what is it looking at? It's looking at an AI displacement risk, right? You think this one is coded up based on uh the occupations then mapped to industries, uh but it then it's connecting it to uh data on uh the degrees or majors that college graduates had. And it's looking at the change uh the shift in majors uh between this graduating class and the prior graduating class and it is showing uh you know I'm seeing we're seeing this at Stanford which is uh CS was uh the fastest growing major uh eating our lunch in economics in terms of enrollment uh until the last couple years when we've seen uh um actually economics and and other fields uh seeing probably not higher majors but we're seeing uh larger enrollments right due to the concern that uh you know there's you've seen this uh in uh Eric's Canaries in the Coal Mine paper we can quibble about how much of that is retrenching so he's shown graphs of declines in hiring of you know like 25 to 29 year old uh CS grads some of it it is probably AI some of it is higher interest rates some of it is retrenching in the tech firms after their hiring so post-COVID hiring boom uh but you see lowering enrollments in these and where you're seeing people uh increasing enrollments you're seeing them in health care. So nothing causal here uh but but certainly I think a narrative emerging right and and students voting with their feet that uh um nursing and and other health care jobs are likely to be uh a safe harbor from from AI but there's a remarkable story I see Institute for Progress I think has a has a really nice write-up of it about the resilience of radiology jobs. Basically there were two bold claims 10 years ago on technology one it was going to put all truck drivers out of business and the second it was going to put out all all radiologists out of business, right? We were able to read scans with computers as well as with a person 10 years ago and we haven't put truck drivers out of business and we haven't put radiologists out of business yet. And interesting to understand why but there is still there's still concern and so people are you know flowing into what Alex calls I think the relational professions and maybe that's not so bad. Maybe we weren't designed to sit behind a computer. We were designed to interact with other humans and these are jobs where interacting with other humans is is crucial and perhaps rewarding so maybe it's not a bad thing but but is obviously disruptive. So I will finish there. Welcome questions now during the break. Thank you so much for inviting me. Yeah, a real pleasure. Ah, okay. >> So thank you so much Neil for those this wonderful lecture. I'm sure I mean we have had a lot of questions but we still have time for a couple of them. Yeah? >> Look at the level of mid-level providers versus physicians. >> Physicians activity have changed if you know over the last 40 years as you show the rise of mid-level but much less of physician and whether it's they have they have become more research oriented or not. And that could explain partially the location choice of physicians like in the Boston area etc. And I've never studied that so I don't know if you have an answer. >> So, most of what I know here is through research by by Josh, my co-author, Maria. Uh so, apologies if I don't get it exactly correct. Uh among physicians, we've seen this huge shift uh towards the specialties. Um they didn't want so moving uh up market, maybe more up market than the zip planner would want. There's for institutional reasons, uh the specialties are paid much better due to something called the RUC, sort of the uh reimbursement, much better than uh general practitioners. So, we've seen that. Um there is a strong correlation in terms of where physicians work, and so, research may be part of it. Uh the strongest predictor is that that area, so either a puma or a commuting zones, uh share of college graduates. And so, what could be going on there? It could be amenities, you know, that doctors who train at Harvard or at Stanford want to stay sort of in the coasts. They may have partners who work in those superstar cities. Um could be they wanted Some of them are at research universities, but that can't explain all of it. Uh when they do move to places like North Dakota, they are paid a fortune. Uh and so, they there is a compensating differential for them to be outside of sort of the research hubs and the coasts. Uh North Dakota has the highest paid doctors in the country, it being like the you know, one of three uh uh you know, cardiac thoracic surgeons in that Helena might be in South Dakota and in Pierre, uh North Dakota, uh is very well remunerated. >> Thank you for your talk. Um I'm Susan Mendes, University of Melbourne. And just following up on this question, uh one of the things that we observe is that especially for physicians, there's has been a massive growth in specialists. So, not not the get what we call general practitioner or family doctor. And this has been going on for the past 10 20 years. So, these are doctors that are um the top 10 earners across Australia are always going to be specialists. And I think it's very similar to the US. And I just wanted to hear your thoughts on that, which you have just partially done. Um and moving forward on that, the fact that well, it takes 16 years to train a doctor and plan healthcare workforce. But so, this movement towards how how do you think of reconciling the movement the movement towards more individualized medicine with the development of uh maybe what we see in the in the community as increasing chronic disease and multimorbidity that is treated in silos by different specialists. So, this this very niche and very strong specialization of the workforce. Well, trying to combine those those two things. >> Uh they're great questions. Uh I don't have great answers. You know, I do think the the time to build the the length it takes uh to change the the stocks in these different occupations is and the prominent role of government in training just makes these really important areas for scholarship. I thought some places market forces will take care of things and we can observe. Uh, but but here we probably have an active role to play in understanding how the sector is evolving and you know how the workforce can meet meet the future moment skate to where the puck is is going. So I don't I don't know what the answers are but I do think it's exactly what we should be thinking about. >> Two short questions. So you pointed out that this is driven by the increase in elderly people in the population. Now the baby boomers eventually are getting older and older and then passing away. So how do you see those trends in the age age distribution over the next decades affecting the facts that you now showed us? >> Yeah, yeah. So I'm not a demographer if you go to the census and you look at their demographic progression uh projections see the fastest growth uh, in the 85 plus year old population second only to the 65 to 84 year old population. Uh, so and the 85 plus year old population particular uh, is a huge driver you know many folks need daily if not 24 hour nursing care. Uh, and so as that population grows there's going to be huge demand at least until the robots get good enough for health care workforce probably nurses and aid. but yeah, I think that's in some sense that's an obvious trend. It's easier to predict, you know, even with there will be some shocks, hopefully positive to longevity. I think that is pretty much baked in. >> Basically, you showed us that you through the shift of in the healthcare jobs, you can make an argument that you increase access to care in especially in some of the states where you don't see a lot of doctors. What about the two other goals that you can maximize or try to optimize which is like either cost or quality? Do you see like an impact or maybe some research that can give us some ideas on where it's going? >> So, Sir John has important work on immigration and healthcare outcomes. So, that's a drawing a direct line from um so, the supply of physicians to mortality, maybe morbidity. I don't remember all the outcomes you look at. So, thinking can be certainly important there. Um the paper that I was hoping to present and then getting data out of the census takes longer than expected was on looking at something called Baumol's cost disease uh in healthcare and how much of the driving wages and expenditure is driven by sort of productivity growth elsewhere, which then has the unintended but real consequence of increasing spending uh in the healthcare labor market. I think the the preview there is right, those forces are alive and well. They're most important in the lower skilled occupations, which are relatively less of the wage bill, and least important for doctors as you can imagine because uh you know once you've sunk your human capital into being a doctor increases in wages in Wall Street might make you envious but are not going to make you switch your profession and so like those forces are there they're not not a huge drivers of what we see in the US but important but look this is once I started thinking about these issues I think there's a lot of fertile ground for more research I'll try to be back to you with more answers and more questions as we move forward. >> Okay well thanks thanks a lot Neil thanks again. >> [music] [music]