AI-Powered Business Formation

- Is artificial intelligence spurring a boom in businesses operated by just one man or just one woman? That's the question on today's episode of Economics Applied. Welcome. I'm Steven Davis, senior fellow and director of research at the Hoover Institution. Joining me is Leah Palagashvili. She's a senior research fellow at the Mercatus Center at George Mason University and co-proprietor of Labor Market Matters on Substack. She also writes a column for The Hill, and you may have seen her opinion pieces in the New York Times or the Wall Street Journal. Welcome, Leah.

- Hello. Thank you so much for having me on.

- It's a pleasure. You have a new paper titled Artificial Intelligence and the Rise of Independent Work. Your term for one-person business operations. So in your paper, you argue that generative AI tools are making it easier to operate solo businesses, businesses without any employees or other than the, the, the owner, operator, him or herself. So, and you also present evidence that at least is suggestive that there's been a sharp rise in the formation of solo businesses in the most AI exposed sectors and occupations of the American economy. Is that, is that a fair overall characterization of your paper?

- Yeah, that's exactly right. I think the main thing is most discussions about AI and work are asking whether AI will replace jobs. And then my paper asks a different question. Is AI already changing how work is, is organized? And I do use preliminary evidence showing that AI exposed sectors and occupations are seeing a faster growth in solo business formation or solo self-employment. And I think the paper is suggesting that AI may be lowering the cost of working independently rather than simply eliminating work.

- Okay. So let's break it down. For give me first, before we get to the data and the evidence, there's a line of economic thinking, which you build on, that suggests it's reasonable, it's plausible at least, that some of these new generative AI technologies and tools that have become in widespread use just really quite recently can make it easier to operate a business, a solo business. So explain that argument. Lay out the, lay out the basic, basic economics for us.

- Yeah. So the basic economics always goes back to Ronald Coast on this. And so from Ronald Coast, we learned that firms basically exist because markets are costly. So he asked this question is, which is, why, why don't we just contract everything out, right? Why, why are there firms? Why don't we, instead of hiring someone to be your software developer, just contract everything out? And that was his biggest question that he was asking in nature of the firm. 1937, I believe.

- Very

- Famous

- Article in economics.

- Very famous, yeah, economics article. And the idea there is, and that he was then expanding on and others have expanded on, which is that if technology lowers some of those transaction costs that make it costly to use markets, this boundary between firms and markets shifts. And that's kind of the, what I think AI might be doing there. And I think we can even start back for a little bit, you know, 15 years ago. The Cosian transaction cost framework, I think, helps us understand the rise of the gig economy as well. And everybody has kind of seen that. So let me start there as just an example because we saw all of a sudden Ubers and Airbnbs and all of that kind of take off. And I wrote a, a academic article identifying several transaction costs that digital platforms reduce to allow this growth in the gig economy. So there's things like trust costs, for example, is a big one that many economists have touched on, which is that prior to kind of the rise of the economy, I wouldn't be able, I wouldn't go into the, you know, the car ever stranger to drive me from point A to point B. But Uber and others reduce those kind of trust costs, so to speak, because they have these digital platforms with two way, two-way rating systems. They vet out, right, some of the bad drivers through various different systems. And just on that aspect alone, I think they've done so much to enable the gig economy. There's o- obviously these other aspects, which is prior, how would you like do payments, right? Cash before. Now they have these digital payment systems and a variety of other technological aspects that really lower the transaction costs to to allow for the buyer and the, and the seller to come together to an agreement. Okay. In a way that wouldn't have been possible 20 years ago.

- Okay. So that's all, that all makes sense. But the, the rise of platforms - Yep.

- For - A whole variety of activities, including Uber and Lyft and, and other similar ride-sharing platforms, that precedes, that predates the recent rise and all the -

- Yes.

- Enthusiasm, hysteria, whatever around AI tools. So give us an example of how ChatGPT or Claude or one of these AI tools can make it easier to operate a business at a very small scale where you're basically running your own sh - one, one person show.

- Yep. So that framework for the gig economy, we can actually, as a co-CN, we can apply it to AI. So in many professional services, I'll give you an example. The minimum efficient scale of production usually required a team. So you have a someone doing analysis plus editing, design, presentation, document review coding support, administrative scheduling, you know, client communications, and so forth. So the idea is that AI tools are increasingly provide some of those compliments directly to the worker. And that accessibility lowers the resource threshold for operating independently. So what does that mean? That means that if I'm an economist or a consultant I would've had to rely on my team previously to do PowerPoint presentations. Maybe a junior consultant would be doing my PowerPoint presentations. Someone to do copy editing, someone to do my research support. And now AI tools can increasingly do that for me. And as a result, that enables me as the independent consultant or researcher to leave the firm, go independent, and do all of that by myself, because now AI tools kind of provide that team for me. Instead of physical people, actual people, you know, or physical office support, AI can provide a lot of those things for me. I'm already using it as an economist. I don't know about you, but I, I, I was thinking about it. I was like, I would have an RA do this, and now I have, like, Claude run this analysis.

- Right. So let me, let me just add a few comments here. First, anyone who's actually run a very small business where you have to hire employees knows there's just a lot of paperwork, overhead, and headaches that come along with hiring just one employee. Okay? Or two or three. You can outsource some of that to payroll processing services, but either got to pay for it or there's still some headache involved. So it's, so there are, you're, so there are some difficulties in doing with employees that you can get around if you can turn to the AI tools. They show up on time, they don't talk back, you don't have to do a bunch of paperwork. So there's that, okay? That's worth noting. Your arguments though about the role of AI tools seems to me apply just as well though to contracting with freelancers outside the firm. Okay? And, and many of the same sectors that are AI exposed are also because of AI itself, but also because of prior changes like the rise of work from home, freelance contract work has become more common, more accepted. So I'm partly laying the ground for, for when we get into the empirical evidence, it's going to be a little difficult to disentangle, and I think you recognize this, between the effects of AI and other things that are happening that have been happening at the same time. So the, the, our familiarity with, for example, working from home in an office setting where you're just engaging people remotely and you're drawing on digital tools, that was already greatly boosted by the rise in work from home. There's, there's a fair bit of overlap between the AI exposed sectors you looked at and sectors where there've been, there's been a big rise in work from home. So all of this is to say that it's a little, it's not, it's, I don't think the actual story that's going on, and I want to get your reaction to this, is just about all of a sudden in 2023 ChatGPT came along and other tools suddenly exploded. There were other things that had been happening before that that I think made the environment receptive to a sharp rise in these solo businesses. Do you agree with that?

- I, I agree with that 100%. And I, and I say that in the study too that this is a preliminary evidence that we're seeing increase in independent work, freelancing, solo self-employment, whatever you want to call it, in AI exposed sectors. But the study's very clear that we can't establish that AI was the primary cause - Yeah. Of

- This, of this

- Study. Yeah,

- You're, you're, you're quite careful.

- Yeah. - You're quite careful in the paper to make, so this is suggestive preliminary evidence. Exactly.

- Yeah. I think - But one of the

- Things - Yeah. One of the things that attracted me to your paper is you've got some evidence. Yeah.

- Yeah. And there's an

- Awful lot of speculation around AI that's essentially founded on almost no evidence. So you're one, you're ahead of the crowd here.

- Yeah. And I think that's exactly right because look, when we present descriptive evidence, we can say like, look, the timing make, makes sense. We're seeing an increase in independent work, specifically in AI exposed sectors, in AI exposed occupations, and I test two different data sets, right? One, Census Bureau Business Formation Statistics. The other one is the current population survey for solo self-employment, which is solo self-employment in that data means self-employed workers who say they don't have their own employees, right? So they're, that's our kind of best proxy in the, in the government data set to get at this, you know, independent worker or solo freelancer.

- Okay. But before we turn to the data though, I wanted to ask you about one more conceptual issue.

- Yep.

- Why don't your arguments about the reduction in transactions cost appealing, appealing to the COSI and logic? Why don't they also suggest there would be just an increase in small, small scale businesses that have a few employees? Not,

- Not - I think that they could. They

- Could. Okay.

- I didn't test for that, but I think that - Okay. I think that's totally plausible in the COSIN framework because it's more, it's not necessarily that the firm disappears, right? In the COSIN framework, it's that the firm might get smaller and the workers who are employees might become more contractors. And I think that's to your point that before AI and ChatGPT even happened, we did have COVID, and then we had flexible work arrangements and remote work and kind of work really changed beyond that. And a lot of workers became maybe more freelancers rather than employees. And maybe AI shock is a secondary shock to that, that's causing workers at firms maybe voluntarily or maybe because the company's like, does it make sense to have these workers as employees anymore, move slowly outside of the firm, right? So they become more in the market rather than inside the firm as employees. Okay. But I think that's totally plausible. I didn't test for that. I focused, my research has always been independent workers solo self-employment. And I put on my lens there and I was like, oh, this is fascinating because in theory, we should see a growth in solo self-employment and independent work. But I think it's equally as plausible to see more smaller firms rise up as well.

- Yeah. Okay. All right. So that, that can be explored in another, another day. But so let's get into the evidence now, and you've already previewed it a bit. And you exploit multiple data sources here. So let's take them one at a time. I, we can take them in any order you, you wish, but I think in the paper you talk, you, you turn first to the census data on, on new business formations as, as measured by applications for EIN, employer identification numbers. So explain, so there, there's some groundwork to lay here. Yep. What is a, what is an EIN?. Why would somebody apply for one? And how can we use it to develop evidence on the question that you're after?

- Yeah. So back to the basics, so an EIN, employer identification number, is a number that any person who wants to apply for a business would, would have to ask for. So it's, it's basically your business number. Now you could, it could be like a freelancer business. Let's say you want to go out on your own and start your own marketing, you know, solo freelancer business, or it could be like you want to open up a restaurant in, on your, in your local town. So for either one of those, you would apply for an EIN number. Now, the Census Bureau, Business Formation Statistics, so they're called BFS for short, track those numbers and report them on a monthly basis. And they track them by sectors only. We don't have occupations, but we have sectors. And one of the key kind of things that the census does is they flag the applications, which they believe are high propensity to employ to employ workers in the future. So we're going to call these like high propensity to be employer firms, meaning that if you open up a restaurant, you're likely going to hire employees because those are going to be your staff and your waiters and your chef and so forth. So they flag a subset of those applications as high propensity business applications, which is what they're called. So what I did is I sent - Let

- Me, let me, hang, let me stop you there. I think I, I want to, I wanted to put on the record - Yeah. A little more fully my understanding of what leads businesses to apply for an EIN. So if you're going to hire an employee and you're going to, you know, and you're going to do it above the ground, above board in a legal manner, you must have an EIN number. Okay? Because you, you, that, that will be required for your social security payroll payments, your unemployment insurance benefit payment taxes and all, all that stuff. Exactly. Now, if you don't have employees, you don't necessarily. You can run a business without an EIN. It, there's going to be. So if I just want to.

- What

- I'm trying to say is not everybody who has freelance or gig income

- Has

- EIN. That's

- True. That's true.

- If

- You want

- To run, if you, if you want to run a business, for example, and your, your contract, you're providing services to another firm and they say, "We want some kind of business insurance," in my ex - personal experience, you're not going to be able to get business insurance unless you've got an EIN number. Yeah.

- So

- There, it is kind of a threshold thing. It's not that everybody who has some kind of gig or freelance income has an EIN, but any, but any substantial. If you want to get credit in the name of the business, if you want to get insurance, if you want to set up an LLC.

- Yeah.

- Okay? All of those things are going to require you to get an EIN. So that's, that's the threshold. And then the Census Bureau well, not the Census Bureau, but the officially the, I think the IRS can't remember who runs the EIS program.

- I think it's the IRS.

- They'll ask you a bunch of ques - They'll ask you a few questions. It's a pretty short form. Like, do you intend to hire people in the next, you know, several months or so? I can't remember the exact timeframe. So that's, so, so that's what leads to the I - EIN number. The Census Bureau has some information from these forms and from the past track record of similar businesses to make this distinction into high propensity and other businesses. And the other ones, those are the ones you're going to focus on. Exactly. And

- Those are, - Those are businesses which at least in the near and intermediate term, intend to operate as one person businesses, right? That's correct.

- Yeah. Okay. That's 100% correct. And I took the, I subtracted the total business applications from the high propensity to get this group over here that are unlikely to be employer firms in the near future. Right.

- Now -

- Okay.

- Yeah, I think that's the right way to put it. Yeah.

- Unlikely that I think, because they could also turn out to be employer firms. They could also just close down, right? They could just be like, I opened my business - Well, no, there's a - I didn't do anything. Yeah.

- There's an, there's another point in there that I think is worth highlighting. You know, historically, the, the, some small share of these businesses that initially start out with no intentions to hire employees in the near term do eventually hire employees. And it's a, it's a modest share. I've, I've, I've written on this with John Haltawinger and others. Yeah. It's a modest share, but we're talking about millions of businesses. So, you know, so there is, there is, even though these businesses are starting out without intention to hire employees, some of them will, will transition to employer firms even perhaps over a long period of time, large, large firms.

- Yeah. And I think one thing to point out too, so I didn't do it in this study, but an economist economic innovation group will release a study soon where they track the BFS unlikely to be employer firms with the another Census Bureau data set called the non-employer statistics. Yes. Which is the actual - I've used

- That before.

- Yeah, you've, I've used that one too, but they're lagged, you know, three years behind, so I couldn't use it here. But he actually, he's going to publish it, I think, in the next week or so, but you can see that they move - Okay. Closely together. So, you know, it, it's - All right. It's possible that they can diverge in the future and things like that. But in the past data, they do move closely together when you take that remainder of the BFS with the non-employer statistics.

- Okay. So we already know before your study that really since the, since in, in the aftermath of the pandemic, there's been a rise in new business applications, new business formation as measured by these EINs. Okay? So that's been going on. That, that, that was occurring before AI tools became really they, they've been around for a while, but before they, their usage exploded. Yep.

- Okay,

- In 2024 and, and, and afterwards. So what is it you do to try to develop evidence that it's really, at least in part, likely the increase in the availability and the usage of AI tools that is driving a part of this increase, this in new business formation as measured by EIN applications or the, the solo business EINs? How do you, how do you untangle that or try to untangle it?

- So I, I untangle it. So just back up one second. You mean how do I untangle it from entrepreneurship and business in general? Well,

- There, there, there was a rise in EIN applications that precedes 2024. Yep.

- So the untangling is basically - And it's

- Still, it's, it's going on. So - Yep. But your story, as I understand it, is you think since 2024, the, the usage of these AI tools has contributed a big part of that.

- Yeah. So I untangle it both through timing, both through comparison sectors between AI exposed and I, AI exposed. And then I do one more thing, which is I add the current population survey, which is looking at workers instead of business applications. Okay,

- But just, we'll get there. But just in the EIN data. So - Yep.

- I have

- All these EIN data. I see a rise in EIN applications, including for solo, the solo business type EINs. And that's been going on since at least 2022 or so. Yeah. 2021, 2022. And it's a reversal of a, of a previous trend, which, you know, declines in new business formation. So there are already things happening - Yeah.

- Before

- It's plausible that AI tools are really driving the show. So explain to me how with EIN data you come to the conclusion that it's probably not, and you're careful about this, probably has something to do with the, the spread of AI technologies.

- Yeah, thank you. So one thing I do is I group the business applications that are in AI exposed sectors as one group I'm studying, and then another group as a comparison group that were not exposed as much to AI. And I use the Census Bureau's own data for that. So they have a survey called the BTOS, and they look at adoption of AI usage across different types of firms. And they give us the sector breakdown, basically, which, which sectors are most likely to have high, which are not mostly, which use, have high AI AI adoption, which ones have less AI adoption. So I break it down there between those two groups. So high AI exposed sectors, low AI exposed sectors. And I think the other thing that is somewhat compelling to the AI story, Steve, is that, is the timing of when all is when we see the divergence between AI exposed sectors versus non-AI exposed sectors start to split. So if you look at so in this study, I start looking at 20, from 2015 onwards between the two groups, AI exposed and then the comparison group. And they kind of move in parallel together. We see the booms, by the way, after post-pandemic as well in the business application and so forth. But then they start to diverge, I would say a little, you know, I put a line in, in my graphs in the study at Q1 2024, but you really start to see it more in the second half of 2024. And so you see the AI exposed sector start to take off in the solo EINs, right? Solopreneurs, solo self-employed, and the comparison group is relatively flat. So I can actually tell you the specific number there. So from Q1 2024 to Q1 2026 solo type business applications in AI exposed sectors rose 26.8%. And then in the comparison group, they were essentially flat at negative - Okay.

- Okay. So that's, that's the core of the EIN based evidence.

- Exactly.

- The, the, the, the two groups that you're comparing here had similar trajectories in the - Yeah. Two

- Years

- Before this.

- They moved somewhat in parallel. And I did a what's called like a preliminary event study just to make sure that they moved somewhat parallel together because if they didn't, then we're not comparing apples and oranges, which is how economists are saying it.

- Okay. So I'll summarize it this way. The, the, the two sec - the two groups, the AI exposed sectors and the other sectors are experiencing a divergence in the solo EINs.

- Yep.

- At pretty much about the time that the COSIN logic we went through earlier would lead you to expect. Yeah, post

- 2020. So it's

- Not circumstantial evidence, but it's strongly suggestive. If I get that, that's how I think about it. Yep.

- Yep, that's exactly right.

- So that's the EIN evidence. And what about the CPS evidence? It says, "Tell us about the CPS. That's a household survey. Very different approach." So what, what are you doing with the CPS or current population survey?

- So the current population survey is the monthly survey that goes out to 60,000 households every month. That, that is the number that's quoted in the jobs numbers reports that we hear about in the news and from economists every month. Now the CPS asks workers, you know, "Have you been working?" And then they ask, "If so, what have you been working? Were you a W-2 employee or were you self-employed?" If the worker says they were self-employed, then they ask them, "Were you self-employed with your own paid employees or did you not have any paid employees?" And so the variable that I use in the analysis is the self-employed who said they don't have any paid employees. And that's our closest approximation of the independent solo EIN or - Right. Solo and self-employed worker, what I call as independent worker. And in that data set, I just do the same sort of thing, except I'm, I also split it by first sectors, and then I look o- occupations as well. Okay. So on the sectors, I break it down between same sectors, high sectors that have high AI exposure, sectors that have low AI exposure. And in those, I find, again, a, a divergence after 2024 in m- more growth in solo self-employment in AI exposed sectors and essentially flat are declining in the non-AI exposed sectors. So the specific percentages there post - 2024, and it map, ma- maps out to, matches to the business applications as well data. So, oh, let me tell you the exact percentage. Sorry. I've, I had it right in front of my notes here. So solo self-employment in AI, AI exposed sectors rose 7.9%. And then in the comparison group, it declined 2.1% after 2024. And that's just the sector, Steve. I also do one that's looking at occupations as well. So there's a well-known index in economics called the AI Occupational Exposure Index. And I just took top 10 highly exposed occupations in the index. Bottom 10 least exposed occupations in that index. And I looked at it as, at occupations as well. And also saw that the most exposed. Sorry. Also saw that from the 2022 to 2023 baseline average monthly solo self-employment and the most AI exposed occupations rose 20%. In the least AI exposed occupations, average monthly solo self-employment remained essentially unchanged.

- Okay. So you're getting roughly similar magnitudes.

- In

- A very different data source. Yep. Okay. It's constructed in a completely different way. So you, you have kind of two quite independent sources of information. They're telling the same suggestive story. That's, that's the essence of the evidence as I understand it. That's right. Yep.

- That the essence of the evidence too. And I think if I had only one, the business EIN numbers, I'd be a bit more cautious and I probably would've waited to look at the, look further at other data sources. But the fact that there were two independent data sources kind of giving me the same narrative and same story I thouht, you know, there's, there might be something here, right? Again, we don't know. Well, there,

- There's still

- Potential for

- Confounding - Exactly. Yeah.

- Factors that

- You can measure. But, but so I think, again, the evidence, if you approach this from kind of a neutral perspective to start with, this would probably lead you to think that, yes, there has been a role since 2024 in by which these AI technologies have made it easier, more attractive, more cost-effective, more profitable to pursue your business ambitions in a solo manner. And that means some of these people might have pursued their business anyway, but ended up hiring a small number of people. Yep. So

- There

- Might be some displacement there. Some of them, but I suspect most of them are just, they have these, you Certain ambitions, dreams, and they didn't really have the wherewithal to pursue them now they do because it takes less to get started. That's

- It. And a lot of these are in like the knowledge intensive sectors, right? So this is more of like our professions and economics or consultants - Yeah. Whether they're writers. You get

- Tired of the academic grind and say,

- "You know

- What? I can just do some, you know, a thousand hours a year of consulting and run my own little shop and make a good living."

- Say, "I'm flying solo."

- Right. Okay. So one, one thing that doesn't come through in your study, and maybe you don't feel like you want to speak to it, but I'm going to ask you about it and you can decline, you don't really give us a sense for what all this means at any aggregate level.

- That's a good question. So I might, I, I don't think this is showing up in job numbers. Is that what you mean, Steven, job numbers

- Or? Well, or even forget jobs. Yeah. Just, just like how many, what, what's been the increase, the overall aggregate increase in the number of businesses out there, one person, even one person businesses? Is this, is this caused a material boost? Let me put it this way. Has this caused a material boost in self-employment? That would be question one. And then question two is, well, how big is that material boost in self-employment re- relative to the overall employment? Those,

- Those - Kind of aggregate numbers aren't in your study. But for people who are like, who are like worried about the job apocalypse, then is this going to be some, is this going to be like spitting into the wind or is it going to be a significant countervailing force to the loss of jobs that some people are worried about? And I'm, I'm not endorsing that view. I'm just saying that many people are quite worried about this. Is this a reason to, to say, "Oh no, there's, there's an important source of optimism," or is this just a sideshow of interest mainly to people who care about solo businesses?

- Yeah, so that's a great question. I think if I had the Census Bureau's non-employer statistics data updated dropped with two years tomorrow, we would be able to see whether we would be able to do a, a stronger analysis, I should say, Steve, about whether it's AI that's now pumping out all of these solo businesses. In terms of like overall job numbers so here's the thing, when we unpack the, the data, unpack the research that I, that I haven't done, but I've read through we don't see overall negative effects of AI on employment right now.

- Oh yeah, that's

- For

- Sure. That's the hysteria I talk

- About. Yeah. And I, and I wonder, so this is like a mechanism story potentially. Is it that, you know, maybe there are, maybe there is hiring slowdown, maybe there is some small levels of layoffs going off, but maybe we're not seeing it pop up in, in the data because maybe some of those workers are now becoming their solo self-businesses and solopreneurs and solo self-employment, if that makes sense. So again - Well, that would

- Be one, that would

- Be one

- Way which it might matter at the aggregate level.

- Exactly. Yeah.

- Okay.

- And I think it's too early to tell whether that is definitely happening. I think that there's some preliminary evidence given that there is these growth, this growth in AI exposed sectors and independent work, that that could be the mechanism by which we're not seeing huge levels of drop-offs in, in unemployment and job numbers potentially, because solo self-employment is a measure that goes in this, in this -

- Okay,

- But

- That's, you, okay, I, I'd like to see you try to make that case, the paper. You don't make it now.

- I don't, I don't. It's not at all

- Obvious that the effects you find are big enough to

- Have. Exactly. Exactly. I think it could. Yeah, it could. And I think, again, it's a little uncertain right now. We, we still have. And we're kind of all operating in the, in the realm of uncertainty with AI right now. So this is preliminary evidence of what I, I think is going on. What I show is going on, which is, I can't say it's from AI definitively, right? But we can see that there's an AI, an, an increase in independent work in AI exposed sectors. I would test this in the future, by the way, to see is, is it that, you know, some of this, some of those would-be workers, you know, who couldn't get hired in an entry level job are going into independent work? And that's kind of preventing, you know, the aggregate jobs numbers from sinking. And I'll give you one quick anecdotal story there. So when I used to teach economics and I was a professor of economics, I would have students these are, you know, I guess, I don't know how old. This, this is about four years ago, so I think this trend is still happening now. But they would tell me that they don't want to apply for jobs or that they've tried to apply for jobs and they're not like getting anywhere. So they're like, "I'm going to start my own like YouTube channel or be my own solo business and so forth." And I just wonder to, to what extent that's happening at an entry level for these entry level jobs, because a lot of the debate on AI and AI's potential negative effect has been concentrated on these entry level workers. And again, something I want to, I, I've been thinking about and working on the side on right now, which is to look at entry level and solo self-employment in particular. Because anecdotally, and I don't know if this is your experience as well, I, I, I know that a lot of young people would come up to me and say, much more than probably many years ago and say, "I'm not applying for jobs or I've tried and not getting a job. I'm starting my own business or I'm doing my own thing. I'm going to be a freelancer."

- Well, I'm at Stanford, so every undergraduate

- Is, is a business.That's the

- Idea they're going to start their own business, and some of them do. So that's in the water here in a way much more so than I think in the country as a whole. But let, let, let's step back and, and now and, and look at, end with, with some bigger picture issues. So I'm going to put some fa

- some claims on the table. You can tell me with ide

- whether you disagree. And this, this, I'm going to step away a little bit from the AI part now. Okay. Which is there's a huge, there are millions and millions, tens of millions of Americans who either generate most of their labor income Through contract work, through gig employment, or supplement their wage and salary income with freelance contract gig work in solo work, so to speak. Okay, so we should understand that that's, that's true. Since 21 or 22, there's been an increase in these solo businesses. And you, you talked about the Uber and Lyft type platforms earlier and, and other platforms. The platform economy is just a part of this, but the point I'm trying to make is there are tens of millions of people who derive a good deal of their labor incomes, in some cases all of it, through non-employee work. Okay? And it looks like, based on your paper, and based on, I think also the rise of work from home, which is for, for a different set of reasons, facilitates that, as does the, just the growth of these platforms, various kinds of platforms, including Upwork. It's not one we haven't talked about here. So if anything, the role of these freelance, gig work, solo type non-em - business operations are going to become even more important. So that's, that I think is an important fact to understand. And I think you agree with that. You can tell me if you disagree.

- I, I agree with that. I think independent work is going to continue to rise. And I think it's going to be even more. It's going to be become, it's going to become a, a bigger part of the American labor force.

- Okay. So you have a set of policy ideas that start with that factual basis and say, look, we need to recognize this. We're not going to get into the details here, that would be for another show. But what's your overall take on how this, the importance of non-employ - non- non-wage and salary labor income is in the economy? What's your overall take for how that should inform the policymaking process and the policy institutions that we have?

- I think this is the most important point for policy, which is that our labor laws and institutions and benefits policy all as, all have been based on and assume a W-2 employee-employer relationship. So if you think about health insurance tied to the employer - Right.

- No, just, just get, it's very hard to get tax preferred treatment of your healthcare expenditures unless you are a W-2 employee.

- Exactly. Yeah, there's - And that's

- The point you're making.

- That is. And also it's cheaper and easier to get it through your employer than it is if you're going solo. And we've interviewed, by the way, some freelancers who just talk about insane how insanely high the prices are for them to get health insurance on their own by themselves. Health insurance is not the only one. I think a lot of labor lawyers will point to things like all the labor regulations, overtime, minimum wage maternity, paternity benefits, depending which.

- Well, that, it may be a good thing that they're not subject to those.

- Yeah,

- Exactly. Those, those cut both ways. It's not, it's not

- Clear whether those are - I'm

- Agnostic. Advantage or disadvantaged non-employee labor.

- That's right. I think I'm agnostic as to whether that's good or bad for the purposes of this discussion. But the point is it's a big deal whether you're on one side of the aisle or the other. And the way to think about it is really that once you go from being a W-2 employee to a independent worker, independent contractor, self-employed, whether you're with the same company or not, by the way, or, you know, maybe I went from being an employee at Google as a software developer and now I'm contracting with Google, Google, from an economist point of view, you've got like a covered market over here and an uncovered market over here where in the covered market, everything applies all at once, minimum wage, overtime, health insurance. And the

- Covered market is tax t - is tax preferred -

- Yes.

- Treatment of many health, healthcare expenses. And then we have, you didn't mention it, but the unemployment insurance program is - Yep. Is, you know, completely premised on the idea of W-2 workers.

- Yep. And that's like -

- Go

- Ahead. I was going to say, I'm working on exactly that one right now on how to I'm creating a policy framework on how to rethink unemployment insurance in the age of AI, assuming that there's going to be even growth of independent workers. But so here's the thing, those policies, especially benefits, and I think in particular, have always been tied to a W-2 arrangement. And now what we need to do is rethink that sort of model, given the growth that's already happened in independent work since at least 1995, right? That's predating AI, the gig economy, and so forth. And one of the models that we've been talking about, and I released a policy framework on this 10 years ago now, almost 10 years ago now which is portable benefits, which is thinking about benefits that are tied to the worker and travel to the worker as they move from job to job. And then you go, that make, that's sort of a, like an HSA, but for everything, and tax preferred, right? That's sort of the ideal. So everyone gets these. We can talk about how are they financed, like who's contributing to them and so forth. But that's sort of the vision and the ideal.

- So that's the central point -

- Yeah.

- Of your line of policy proposals, which is to tie benefit, make the benefits follow the worker as they move between employers or as they move between being a wage W-2 employee and an independent contractor and vice

- Versa.

- That's right.

- And I think, and I think this matters even if you're not an independent worker, as you laid out, because again, 80 years ago, we're talking about a worker who stays at one firm for most of their life, so there's not a lot of changes. Now, you know, our economy is a lot more fluid and there's a lot more changes. So even if you're not going - That, that's

- A, I think that was a myth. There was never an economy like that.

- Oh, okay. I think. All right.

- There were some people working for mainly white males, well educated, relatively well-educated white males working for large corporations for whom that was a, a good description. But it's, it's like a simply

- But would you agree that

- It's become more fluid regardless of the economy?

- No, not actually. Okay.

- Okay. All right. Then let's go back to the independent look. The

- Point is it's all, there's always been a lot of fluidity.

- Okay.

- That's the point that's lost. It's not that somehow there's been this. This is, this is a misconception. I've written about this at length over the course of my career. The, the data are pretty clear. But people just, I don't know, don't, don't want to look at it

- In a careful

- Way.

- So there's always

- Been a lot of fluidity.

- Okay. You don't think there's more fluidity now than there was 50 years ago between jobs or between independent workers?

- No.

- No? No,

- Between jobs and independent workers, I, that's a hard thing to track - Okay. Going back very far in time. Look, I'll, I'll give you a, a bunch of things. You can look at the rate at which jobs disappear as measured by the - Okay. The Census Bureau from year to year. Those have trended down largely. There's, there's, I'm simplifying here, since the early '80s. If you look at the new claims for unemployment insurance benefits, those have trended down for almost since the inception of the series that goes back to the '60s. So you don't, you just don't see this, this -

- Increase in.

- This notion that somehow we're all subject to a lot more insecurity in our employment relationships now than decades past. And I think the evidence points in the other direction. If you look at - Okay. If you look at unemployment inflows from, in, in the CPS data, from employment to from people who had a job to people who become unemployed by reason of losing their job, all these things have trend. And again, I'm simplifying. There's, there's, there's big fluctuations in the data, most notably the, the pandemic, but also in the, in the global financial crisis. But the, the trends are, are very much in the downward direction. Okay.

- And there's

- Many things that drive that. So there's always been a lot of fluidity. So

- I'm gonna, I'm gonna, I'm gonna go with your. There's a lot of, regardless of whether there's a lot of fluidity, there's, there is a growth in independent work. And there is, there likely will be more growth in independent work. And probably that matters most for independent workers. I think that's

- Right.

- Yeah.

- I think that's right. Okay. Well, thanks so much for a, a lively discussion and keep up the good work.

- That's great. Thank you so much. And I will not use that example again of -

- No, no problem. Okay. Thanks, Leah. Take care. Thank