On The Nature Of Entrepreneurship

>> John B. Taylor: Anyway, we're happy to have Ellen McGratten from the University of Minnesota, Stanford, PhD, I might add, not too long ago.

>> John B. Taylor: That's not long ago.

>> Ellen McGrattan: It's a long, long, long time ago, I hate to tell you.

>> John B. Taylor: Anyway, it's great title for a paper, the Nature of Entrepreneurship.

 

>> Ellen McGrattan: I added the on so it didn't sound so bold, right?

>> Ellen McGrattan: All right, so this is joint work with my colleague Anmal Bhandari, two students. Well, one is now at the OTA, Toby Cass, Thomas May, who's graduating this year, and Evan Schultz, who's at the Internal Revenue Service.

This is just to give a little bit of background. This is a joint project with people, a team of researchers at the IRS in St.Paul. So we're very lucky to have a group there. They're mostly statisticians. And kinda our part is to kinda bring some of the economics into the question about what is the nature of entrepreneurship.

This is kinda a first. This is our first disclosure. There's a lot of work going on, but we wanted to kinda set a reference point. There's a lot of work done on this question, mostly with survey data. So we're trying to kinda get a sense of what we can learn from the IRS data.

And our big vision really is to develop theories. So we wanna know kinda what are the most salient features that we need to bring into our theories. And those would be used, we hope, by the IRS for thinking about tax administration. So that's the end goal, but the immediate goal is to try to bring to life kinda some things that we're learning from using this administrative data.

There's a big long disclosure thing even worse than the Fed. I'm not gonna read it. All right, so let me just, I'm gonna start with what we do, and then I'll explain why we do it. So what we're gonna do is we're gonna assemble a longitudinal database of business owners.

So, I use the word entrepreneur, but I want you really concretely to think about every business owner in the United States that we can track. So these are primarily pass through business owners that either run an S corporation or they're in a partner in a partnership, or they're a schedule C proprietor.

And we can see information both on their business tax forms and on their individual tax forms. For those who think of entrepreneurs as Bill Gates or Elon Musk, if they had one of those businesses, we would see them after. If they're paid employees in that c corporation, when they ipo'd, we would see them as paid employees.

So we're gonna be, yeah, go ahead.

>> John B. Taylor: What are those two guys?

>> Ellen McGrattan: What are they? They're paid employees, yeah. So we're gonna be estimating with that database, we're gonna be estimating lifecycle income profiles and looking at incomes and growth. There's gonna be 35,000 subgroups. So the usual things, but there's gonna be some additional observations we're gonna wanna bring in and talk about those in particular, in order to think about entrepreneurship.

We're gonna take those profiles and we're gonna be, most of the comparisons I'm gonna do are gonna be like twin studies. There are gonna be twins who one chose to be self employed and one chose to be a paid employee. And we're gonna be looking at their growth and volatility patterns.

And we're gonna be looking at the determinants of the entrepreneurial choice.

>> Speaker 3: Is it obvious Anmal, but I don't wanna jump in the intro, why-

>> Ellen McGrattan: Well, I should say Minnesota rules, so jump in anytime. It's kinda like rugby instead of American football.

>> Speaker 3: Why not looking at business owners roles who are also managers?

This is gonna be your primary criterion to go fine end.

>> Ellen McGrattan: These business owners are effectively, like actively managing.

>> Speaker 3: As well, okay.

>> Ellen McGrattan: Yeah, they are the people. This is your tax accountant, this is your lawyer, this is your, the dry clean, the guy running the dry cleaners.

This might be a franchisee, it's everybody kinda running-

>> John B. Taylor: Contractors and gig workers are not in either of these things, yeah?

>> Ellen McGrattan: You mean 1,099 people?

>> John B. Taylor: Yeah.

>> Ellen McGrattan: No, and I'll go through the sample, but these are gonna be the owners of an S Corp partnership or their own proposal.

 

>> John B. Taylor: You just treat contractors as they're neither employed nor in their self-employed group.

>> Ellen McGrattan: Yeah.

>> John B. Taylor: Okay, fine.

>> Speaker 4: But they get plenty of 1,099.

>> Ellen McGrattan: The schedule-

>> John B. Taylor: You're gonna go through this in detail. Why don't we just hold off a minute?

>> Ellen McGrattan: But I will say one thing in response to this.

If I'm working at the Fed, I get a 1,099 from the Fed and I fill a schedule c, then I am counted as a self if I'll go through the criteria. But if that's all I did, that's how I would count it. But it's not like somebody who just does some temporary thing and is not running their own business.

Okay, let me motivate why we're collecting these data and why we're doing this. Basically, what we're trying to do is update answers to does entrepreneurship pay? That's an important question for the IRS. These are the very people that and it gets to the second question, is there scope for shrinking the tax gap?

These are the people that make up much of the tax gap. So does it pay matters? Is there scope for even shrinking the tax gap? So that's kinda the two things I'll update. Like I said at the very beginning, our ultimate goal is to use the results to think about entrepreneurial theories.

And to think about tax administration and policies related to both design of tax policy and tax administration. And I should say one thing. Is there scope for shrinking the tax gap? This is not just about studying some small set of people in the economy. These guys will be generating more than half of business net income in the United States.

So even from the perspective of macro, or people who look at applied micro questions and use macro data, the people we're gonna be studying are a big group of the macro economy. So preview of findings, yes and yes.

>> Speaker 5: Can you define the tax gap?

>> Ellen McGrattan: Yeah, it's the difference between what is voluntarily paid and what the IRS thinks should have been paid.

 

>> Speaker 5: Thank you.

>> Ellen McGrattan: Okay, so I wanna kinda, as a way of kinda giving the findings, I wanna contrast. There's gonna be two slides, previous work and our work. And I've designed it so that I almost wanted to put two side panels side by side. So much of the previous work on the topic of entrepreneurship, we've been kinda forced to use survey data.

Most of the business owners are in the private sector. You don't see much about them. And, of course, with surveys, there's issues with top coding and there's issues with short panels of data. So there's those ever constant problems. And the conclusions that are drawn from those survey based papers would be that self employed, relative to the like, their twins would have flatter life cycle profiles.

They would enter self employment with lower past income. So some people call them the misfits or maybe the people that Eric studies, the guys who play video games. And then eventually they go off and live in mom's basement and call themselves self-employed. And the third thing is, it's viewed that they enter with higher past asset income because to overcome some liquidity constraints.

And these statistics are motivating theories where the entrepreneurs either earn very large non pecuniary benefits. Or they're kind of, in the Evans and latent sense, misfits, or they face very large liquidity constraints. So I wanna contrast everything on this slide.

>> Speaker 6: I thought, my reading of the literature is that there's a tremendous amount of Haiti where there's some like that, and there are, some are different.

And so should I be thinking about means and averages now? Or are you gonna be thinking about the contibutions.

>> Ellen McGrattan: I think that's an awesome question, and I wanna defer you, because I'm gonna show you a picture of some means and mediums just from empirical data that the survey guys would have access to.

And you'll see there's a big difference between the IRS data and just from a quick, easy, empirical check, but I'm gonna defer you. Okay, so I'm gonna contrast this with what we're gonna find using administrative. There's not gonna be any top coding, so we don't have to blur anything.

Everything I show you today, we will lop off 0.01%. I know there's obsession about the 0.01 for people who might have written down billion by accident instead of million. I want to get rid of that kind of stuff. It doesn't make a difference but just as a full disclosure, we did, we dropped the top and bottom 0.01%.

And we have long panels, so we're gonna be able to see data back to 1996.

>> Speaker 7: I can ask you, there's a tail there, too, no?

>> Ellen McGrattan: Yeah, there will be a tail.

>> Speaker 7: At the 0.01, I mean, why you really don't believe it.

>> Ellen McGrattan: No, am saying everything I'm gonna show you today, it didn't matter, I'm just saying.

 

>> Ellen McGrattan: You were looking at just those guys. Yeah, but we're gonna conclude that unlike in the survey data, these guys have very steep lifecycle profiles. That they enter self employment with higher past labor incomes and they enter with lower asset incomes. We're gonna kind of point people to theories where they have to make significant investments cuz the growth profile that we see is gonna look very hump shaped.

We're going to be pointing to Javanovic like models where there's experimentation early on, and we're gonna be kinda pointing away from models where there's these very high liquidity constraints, yes?

>> Speaker 8: Can you tell us a bit about how the previous literature was defining entrepreneurship, and given that definition, how does that overlap with the definition that you're gonna use?

 

>> Ellen McGrattan: Well, let me do the IRS versus CPS in a couple of seconds, and I'll tell you how we do the same definition. But I'm gonna put that off, too, because there's only certain things you see. Well, I'm gonna wait, I'll get derailed. Okay, so today, I'm gonna talk about our data, what we see, what we compute.

Then I'm gonna talk about our estimation procedure, the challenges we face to identify everything. And then I'll talk about the results I just described, what the income and growth profiles look like buy, groups. Then I'll talk about entrepreneurial choice the entry and exit. And then I'll talk about the characteristics.

Again, it's just gonna be fleshing out what I just summarized here. If I have time, I'll show you how if I had time, I'll show you a dynamic program where we match the dynamic program up to the data and look at young entrepreneurs making a decision to stay or to go.

Okay, so here's the data. The primary source of data is obviously the administrative IRS, but it's for those of you who know the work of Chetty and co-authors like John Friedman and others. Well, they have put together a database of every person that has a social security number.

They've linked things from the 1,040. So that's kind of our starting point. But now, of course, since we're studying business people, we're gonna be able to add to that by looking at the business side. And we wanna do balanced panels cuz we wanna construct, a sequence of balanced groups or we're gonna do birth cohorts, 1950 to 1975.

So that we have you, everybody for 2000 to 2015, when the Chetty data end, that's gonna be updated. So basically, everything I show you with a push of a button, we'll be able to update it to the very latest data. But right now it goes out to 2015.

We merge in the schedule C from the 1,040 and we merge in Schedule K-1. So that's if you're an owner of an S Corp or you're an owner of a partnership, you have to file or your business files scheduled K-1s. Which those owners then report on their 1,040 on the front page and in schedule E.

So we're gonna be merging in that information and we're gonna be looking, like I said at the beginning, at the owners of pass through businesses. Because then we can see you, we can see everything on your individual filings, we can see everything on your business filings. And that's available, the k-1s are available since 2000.

 

>> Speaker 9: Sir, I'm thinking about Zuckerberg, the young Silicon Valley, not necessarily of that type of profile. Guides and guys are not gonna be here. I understand you want longitudinally balanced panels. Why don't you think that the way you succeed as an entrepreneur may have changed over time? These are 50th ish, the youngest.

 

>> Ellen McGrattan: Yeah, so you mean like people who are young people born after 1975?

>> Speaker 9: I think entrepreneurial ecosystem, maybe because you're sitting here, may have changed over time.

>> Ellen McGrattan: These are their birth cohorts, they're in our sample from 2000 to present day.

>> Speaker 9: But you don't have anyone younger than 50 ish.

 

>> Ellen McGrattan: 1975 is the young one.

>> Speaker 9: Right.

>> Ellen McGrattan: We can look at anybody in a non-balanced way.

>> Speaker 9: Right.

>> Ellen McGrattan: Yeah, but what I'm gonna show you today, because of the way we design the econometrics, it's gonna help to do it in a balanced panel. And that will go forward in time as they're updating information at the IRS.

You just, like I said, we're trying to design it. So somebody just goes and pushes the button and it moves forward.

>> Speaker 9: I'm just thinking of some trends, if you are looking at inequality in wealth, in particular, have been accelerating lately. And if you think that this is of the origin of it all, maybe you're not capturing some of the forces behind some of the recent developments.

Most of our guys, like I said, they're tax accountants, they're lawyers, they're franchisees, that's the bulk of our people.

>> Speaker 10: But the forms that you get the data from illustrates a very important point, that you don't consider capital gains from selling a business, even though at the upper end that's really a large fraction of the total.

 

>> Ellen McGrattan: It's true, what I'm gonna show you, we haven't incorporated in because a lot of our businesses are ongoing. So all the numbers are going to be lower bounds of what they're making because, as Bob says, when they sell the business, they will sell. They will make a lot of their late the sweat equity that they put into the business they will make when they sell it.

 

>> Bob: Isn't that on the 1,040?

>> Ellen McGrattan: The capital gain, but not when they sell it, these are gonna be, there's-

>> Bob: Yeah, when they sell it over.

>> Ellen McGrattan: When they sell it.

>> Bob: Yeah.

>> Ellen McGrattan: You have to learn that-

>> Bob: I'd say you're doing it now, you're making it later.

You'd have to run out to when they sell it and then go back and say it's implicitly attributing it to now, that's what you mean?

>> Ellen McGrattan: You have to set the project on sales.

>> Bob: Okay, got it.

>> Ellen McGrattan: What Bob's pointing out is you're going to miss that income later on that they'll get when they sell.

 

>> Patrick: Past 2015.

>> Ellen McGrattan: Correct, but a lot of ongoing businesses are gonna sell, say they sell in 2025, we don't have that.

>> Patrick: Yeah.

>> Ellen McGrattan: Yeah.

>> Speaker 6: So all the sales that are subject to step up a basis, you don't have any handle on at all, right, cuz that's not reported anywhere hwre, actually.

 

>> Ellen McGrattan: That's not in here, but suppose you did a step up like father and daughter, okay? I think there'd be a way to, cuz there's gonna be something you do have to file there, but we can talk later about that. I just don't have that because we don't have it for everybody.

 

>> Patrick: But Ellen?

>> Ellen McGrattan: Yes.

>> Speaker 13: I'm sorry, will this affect your point about the steepness of the lifecycle income?

>> Ellen McGrattan: Yeah.

>> Speaker 13: To the extent that you do observe and sell later on, you'll think it's very steep, but-

>> Ellen McGrattan: Even more steep.

>> Speaker 13: Even more steep, right? But actually, that income was generated before.

 

>> Ellen McGrattan: Yes.

>> Speaker 13: So maybe actually, they had a very flat income profile.

>> Ellen McGrattan: Well, that's a good point, and you'll see from the growth profiles. Can I come back to that, because that's a good point, that they are investing all along, they are, and we'll see it. And that's kind of something I'm looking out for.

And that investment is paid off, ultimately. Okay, income measures, so our self-employed, like I said, they're gonna be the net profit. They might own different businesses also, so we'll combine them. It'll be the net profit from the sole proprietors, the ordinary business income of the partners and S corp owners.

And if they are S corp owners, they're forced to declare wages cuz they have to pay FICA on it. So we're gonna add the wages from their business back in. And the paid employees will be their W-2 wages.

>> Speaker 14: And so aren't many people both? I'm just trying to think like for those people in this room, I used to know Steve, did some consulting.

 

>> Ellen McGrattan: So yes, and I'm about to tell you how we divvy you up, okay? And we're gonna, yes?

>> Bob: Can I just make sure 1,099s contractors, they're just not in either group.

>> Ellen McGrattan: No, if you have 1,099 income, that is your schedule C, that's my fed income, then I see it on your schedule C.

So if you report it as business income and you're filing a schedule C, I got you.

>> Bob: Let's be clear, If you receive a 1,099, you must file a schedule C or some other business-

>> Ellen McGrattan: I think there are some noncompliance issues, Bob.

>> Bob: Yeah, but you're supposed to.

 

>> Ellen McGrattan: You're supposed to. Essentially, would get you.

>> Bob: Conceptually, you're gonna treat those people as business owners.

>> Ellen McGrattan: Okay, ready? So each year, now this goes to there's two things going on, right? For every I, I'm an I and a T year, we're going to put you in one of three pots, a self-employed, a paid employed, or a not employed.

So the criteria by which we're doing the self-employed, are you making at least an absolute value of 5,000, because a lot of these guys lose money. In fact, they're big losers at the beginning of their life. And we wanna count you making those losses, and at least you're making more in the self than the paid.

Or you have at least one employee, as you know, for your share of the business or your gross profits are bigger than your PE. Now, we do have a table where we go through and we take all of our sample and we say how much are you getting from the paid?

How much are you getting from the self? It's not like half, there's nobody doing half if they hit that self employment cut criteria. So it's unusual to see such a thing. The paid, if you're not the self-employed, and you have to have earnings over 5,000, and then the not employed if you're neither of these.

Turns out our non employed are a big group.

>> Bob: They just don't make very much.

>> Ellen McGrattan: The non-employed, we have you if you're out for a year or more. These people are wildly interesting in they're a huge group number of people and they look bad.

>> Bob: They're wildly what?

 

>> Ellen McGrattan: Different than the other two.

>> Ellen McGrattan: Okay, so those are how we bin you later. I'm gonna talk about a characteristic that I'm gonna place upon you. If you're mostly in one kind of activity or another, I wanna distinguish people who do a lot of switching back and forth and people who are really truly engaging in this activity or not.

Okay, we like to have measures of skills in education. And so obviously we have limited tools here, but we do know, because you have to write down your occupation on the thing, we have occupation for many. And then what we do is we use AI tools to impute for the missing codes using peers.

Education, we see students or people receiving a 1098-T, which relates to your education. For older folks that never got those because they were a later form. We use CPS data to kind of impute. We design a classifier system, then we fill in, we use IRS data to make a prediction about education.

 

>> Speaker 15: Just understand, the IRS doesn't provide any demographics whatsoever. This is all education.

>> Ellen McGrattan: Yes, we know gender, we know all the SSA stuff.

>> Speaker 15: Okay.

>> Ellen McGrattan: We know what, that is important, right? Because then we see a lot of things. I'll show you the full list and that's gonna be important.

So we had this sample where we had all this long list of criteria to talk about who got binned as a self employed, including like information about your profits, information about your employees. If I talk to the survey guys, they can't possibly know that. What we did was just as a quick check, I want to go back to the question about what are the means and medians of the IRS sample for the C versus the CPS.

I have to get on there. I have to do the same thing head to head with the CPS data as I would with the IRS data. So let's look at just criteria like you make over 5,000 and you're making more in se income relative to PE income. We're gonna find it, I'll show you a couple of pictures.

We're gonna find the PE guys. Not so different in empirical moments between the CPS and the IRS. The SE, quite different, not in medians.

>> Eric: Can I just say if it's coming to somebody, is the shares, are you getting more entrepreneurs than?

>> Ellen McGrattan: I think I have a picture and then I'll do the shares, okay?

Here's the median SE income, by the way, I didn't do all the real dots. These are just, I fit things through it. The CPS is noisy, this is done, but I fit things through it. There is the median IRS, I say population cuz we didn't go in and do all that fancy stuff we just said.

Give me people who fit these criteria, including everybody. Not just the people in that birth court, not people missing, just everybody. It has a slightly different shape, but you got to look at the y-axis, okay? These are numbers, I put them in thousands of dollars cuz I like units that I can actually understand that are a little, between 20 and 30 and relatively flat, okay?

Now I wanna do the means. Here's the means, and this is everybody, so it includes everybody. I'm not gonna say you're in self employment a long time or anything. Just everybody who has one of these, I just take a mean, those will peak out at about a 100, little under, maybe 110.

And you get this profile even here. Then you see the CPS guys flat, now let me answer Eric's question. We wanted to break up group by groups. What is driving those differences just in the empirical moments? At the very young, there's misrepresentation of the sample because there's very few doctors that are 25 in the CPS, for example.

But in the later years, it's not that, it's mostly the income differences. So they're grabbing not the big guys.

>> Eric: Can you mimic the top coating that they did in your better sample, and then see to what extent the top coating per se can account for part of.

 

>> Speaker 9: Age distribution and the margin.

>> Ellen McGrattan: I get you.

>> Eric: Is it hard to do that?

>> Ellen McGrattan: They have this rank proximity where they take, they bin you up and then they move the things. We don't have, they don't publish the details of that. But our guess is even if we got it, we're going to be missing critical people because of the differences in the industry.

 

>> John B. Taylor: Work on that, the CPS, rich people don't respond to the CPS.

>> Ellen McGrattan: Yes, that's my guess.

>> John B. Taylor: There's high quality work on that, I don't remember exactly whether using the IRS data, the Social Security. There's a paper in the JP, you must know it, around 2019 or so, that has very clear evidence of how the non response varies with income.

 

>> Ellen McGrattan: I should look at that, I should tap that because I'm.

>> Patrick: If you go back to the, if you had the truth in one of them and you could have the weights on the truth and you have people who were as percentage wise under reporting based on characteristics, couldn't you go back to the ones where they're not answering, reweight it and then.

I don't have access to CPS.

>> Ellen McGrattan: I can't do anything, got it, I could only mimic best, yeah. By the way, just to give you a sense that LMA is the IRS population, I grabbed everybody. If I had done the same thing with our sample with the balance panel and etc..

It's not that different.

>> Speaker 15: Demographic wise, are they balanced so they look like similar?

>> Ellen McGrattan: Good question, I'd have to look that up. Okay, that was just kind of aside, just FYI, there's going to be-

>> Patrick: Just the one quick, the share of entrepreneurs that you defined and share that is in the CPS are the same number in older households?

Like10% of the population.

>> Ellen McGrattan: They would have the same in terms of the things we can see, like NAICS, gender.

>> Patrick: Just understate the raw data, suppose we take.

>> Ellen McGrattan: Our data, you would blame the differences on the incomes, three quarters of the difference on the incomes, and one quarter on the, I didn't get the right number of female doctors who are 35.

 

>> Speaker 8: The number is 12% or something like that of the whole population that meets your definition of self employed in this, in the census, because that's kind of what it is in the CPS. I mean, I just wanted to get a sense of the tax data. The tax data is about 12% of 35 year olds or entrepreneurs or something like that by your definition, self employed.

 

>> Ellen McGrattan: You mean that population one or my sample?

>> Speaker 8: Your sample that you were gonna have. Article, I just want to see, are we making, are we in the same ballpark? You said we are, but I'm just.

>> Ellen McGrattan: We did it group by group. I have to go back and give you the sample things.

But we did a group by group and said, where are they missing in terms of representation? In the older ages, they're not missing so badly in representation, they're missing badly in incomes. Okay, so I'm gonna now go into kinda the main thing using the IRS data. So this is gonna be to estimate these life cycles.

Let me just motivate one thing central to what we do is always that we're comparing kind of these twins that are in SE and paid employees. Partly because one thing that we're looking for in particular is that they have, there's human investments on the job learning and all that human capital building.

But there's gonna be a big difference between the self-employed and the paid-employed, and it gets to the what gets paid at the end when they sell the business. The business is the firm-specific investments that they're making. And so we're gonna be looking at people that have same demographics, same industry, same education, but very different investment opportunities.

Cuz one's building their business and themselves, doing general human capital building, but the other is just doing human capital. So we're gonna be looking for that in the life-cycle growth process.

>> Elin: But they're not getting the exit values from capital gains is pretty material, right?

>> Ellen McGrattan: I'm gonna show you what they give up initially, to do this.

 

>> Elin: Well, then you're gonna make some kind of an assumption, let's see.

>> Ellen McGrattan: Yeah, let's see.

>> Elin: Okay.

>> Ellen McGrattan: So here's the object of interest, income, conditional and age, and then we're gonna be bringing in individual and aggregate factors, and then I'll get back to Elin's question about what those are that we see.

Okay, so let me just lay out a couple of issues, of course, when anybody's estimating these things, there's selection. These incomes could be just driven by latent factors. So we're gonna allow for an unrestricted intercept, and then we're ultimately gonna be working in growth with growth rates. There's a survival issue.

Are you picking the successful people? You've already preordained the winners? This is where we're gonna bring in this notion that there are people we're gonna be conditioning and those who stick with the activity for a long time, and there won't be all winners. There will be lots of losers in there, and they're persistent losers, but they stick.

For identification, obviously you can't get time, age and cohort. So I'm gonna explain how we use the overlapping cohorts to estimate age and time, but bring in differences in cohorts, and then-

>> Elin: The only thing you're missing there is an exponential trend, the time age cohort setup, it's just an exponential.

 

>> Ellen McGrattan: Well, I'm gonna show you our setup, and then you jump in with that thing.

>> Elin: I've already jumped.

>> Ellen McGrattan: Okay, jump.

>> Ellen McGrattan: But I'm gonna have you point to something.

>> Elin: You can jump in at the appropriate time.

>> Ellen McGrattan: Yeah, let's jump in the appropriate.

>> Ellen McGrattan: Yeah, okay, and then of course, there's the sign issue, right?

In business incomes, there's tons of negatives. You can't take log of a negative. So I'm gonna tell you how we deal with that, mostly it's gonna be working with a flexible error structure. But we do some robustness to check that we're not getting just spurious results with big guys in there, so I'll show you that.

But if you're dealing with business incomes, you have no choice but to deal with lots of negatives. Okay, so we're gonna be estimating what we call the time and age effects, this beta and the gamma, for income. And then we just notation, yit like I said, is gonna be levels of income, so everything is in levels.

And then we have i indicating the individual, t for calendar date, c for their birth year, a for their age. And the g, that's the thing I talked about in the beginning that we have the 35,000 Gs, that we're gonna be dealing with, that will partition our set of individuals.

Okay, so we need two assumptions to separately identify, and once we make those, this is gonna be something we can do on a hand calculator. I love things you can do on a hand calculator, and that's what I'll show you in a sec. The two identifying, okay, so this comes to dealing with age, time, and cohort.

We're going to assume that the age effects are the same across binned cohorts. You need at least two bins, obviously cuz if you had only one, you can't identify. The things I'm gonna show you later, we bin people up into being born in the 50s, 60s, and 70s, but we're playing around with other binning.

And that's a way being born in the 60s is gonna be your group, that's a group for you. So it gives us kind of a flexible way to get the cohort in there. Then we we need to pin down one number for getting our time effects. So we have group level, think of the mu as kind of the trend growth over time.

In everything I'm gonna show you, we're gonna assume that's constant across groups at the aggregate level of a real wage growth. But we've played around with differences across industries, cuz some industries have higher growth, some slower. And nothing I'm gonna show you, doesn't matter, so it's really an appendix thing to vary that.

Okay, so let's talk about our groups, no, sorry, a practical footnote. So this is the back of the envelope, gets us everything, we are gonna do a least-squares approach, and this is literally group by group. So imagine writing a little MATLAB code that has a loop, i goes from 1,385,117.

And we invert this matrix, which is just gonna be a function of some average incomes and population counts at different times and ages. And the only reason I didn't write it out explicitly, there's some adjustments, because there's demographic change over our time. So the group sizes are changing and we have to have adjustments.

 

>> Speaker 15: Sorry, I was lost in the notation, so you have individual time effect, individual effect, individual time effects, can we go back one second?

>> Ellen McGrattan: Yeah, so we have the alpha i. So we're gonna be working with the differences in the levels, so.

>> Speaker 9: My question was very simple, is an individual specific i, it's not a Latin class specific.

 

>> Ellen McGrattan: It's individual.

>> Speaker 9: Are you super flexible, can we go one second back because we have the gamma 2. We have the alpha, I forgot the gamma.

>> Ellen McGrattan: Sorry, yeah.

>> Speaker 9: Okay.

>> Ellen McGrattan: Yeah, it's super flexible.

>> Speaker 9: It's a birth year, okay.

>> Ellen McGrattan: Yeah, and we made the assumption on the cohort and the-

 

>> Speaker 9: And you could have that third term also, i is interacting with times in terms of the c retrancation in a. I see, it's a non-parametric way of describing i and-

>> Ellen McGrattan: Yeah, yeah, exactly, okay, so here are our groups. Let me do the usual things because we have the SSA.

Well, first we have cohort, we know when you're born, gender. We have our imputed education through our classifier system, we have the skills which we map to cognitive interpersonal and manual. We know the industry of your business or your employer's business. We take the primary one, where you get the primary income.

We know you're married or not, and we have a mostly married versus mostly none. And we know if you have children, now the key thing is this additional thing, and this is what we wanted to add because we're doing occupational choice. And people are very different, the switchers and the not employed and the self-employed are gonna look very different.

So we added this group time invariant feature called employment attachment. So you're attached whether yourself or paid if you have the same, remember our sample of 16 years, if you're in the same employment status for 12 plus years and you have fewer than two switches and no intermediate spells of non-employment.

 

>> Speaker 9: It was a long time.

>> Ellen McGrattan: Yeah.

>> John B. Taylor: What does status mean, and just where?

>> Ellen McGrattan: What is where?

>> John B. Taylor: What variables determine status?

>> Ellen McGrattan: That was that, remember when we said are your self-employment income higher than your paid employment, do you have a

>> Patrick: You can be switching industries, you can be switching all over, you can be switching from industry b, et cetera, but if you don't switch to being like self-employed versus, it's switching that category.

 

>> Ellen McGrattan: Yeah, it's switching in that category.

>> Patrick: I wanna ask about industry, you get classified in the industry once as an individual or you move around from group to group when your industry changes? I have that same question.

>> Ellen McGrattan: That's a great question, I think now this is a guess off the top of my head and I have to ask Thomas if we really did this, I think we took your longest NAICS.

 

>> Patrick: Okay.

>> Ellen McGrattan: Yeah, I think we took, cuz we made it a time invariant thing, so we must have taken your longest NAICS. So, that wouldn't be switching, Patrick, now one could do robustness.

>> Patrick: Yeah, in the regression, fine, it wouldn't be in this variable, it would be in that other variable.

 

>> Ellen McGrattan: So we have another category we did almost attached, there's such a small group you could just lump these two together, the key other group is mostly switchers. So they're in SE or PE for 12+ years but they don't have the intermediate non-employed, these intermediate non-employed people look different, so we wanted to put them separately.

Any non-employment, if you're kinda in the middle of like prime age and you're kinda going out of the labor market. Now, there's one group that maybe I'd like to pull them out, but I don't know how to do it, the guys who go do an MBA and come back, those I would like to move out, but-

 

>> Speaker 9: Understand it's the same employment status but not necessarily the same business, I can be a serial business owner, but I enter as long as I'm for 12 years straight in such a status

>> Ellen McGrattan: Well, no, you could be six years, switch to pay, and go back six years, in other words, you have to have, during our sample, not contiguous, continuous, is it continuous, which one is it?

In other words, you could have done one switch, but you're going back and you're doing it.

>> Speaker 9: But if I have no gaps in your sense and I consider

>> Ellen McGrattan: A lot of them have no gaps.

>> Speaker 9: They do, cuz I'm-

>> Ellen McGrattan: Yeah, a lot have no gaps of the ones that are in attachment.

 

>> Speaker 9: 12 years doesn't imply a lot of attrition, natural attrition, anyway, or you see a lot of people.

>> Ellen McGrattan: No, these are balanced, these are in our sample the whole time.

>> Speaker 9: As attached.

>> Ellen McGrattan: As attached.

>> Speaker 9: Okay.

>> Ellen McGrattan: Yeah, okay?

>> Patrick: Yeah.

>> Ellen McGrattan: All right, and just to give you some numbers, by the way, this is, we're updating these numbers because we found a trove from another database of occupational strings.

So we're gonna add 14 million more, but they're mostly two any NE, but I, so you have 36 million, so if you wanna do your counts.

>> Patrick: It's lower than I thought bacause-

>> Ellen McGrattan: You have 2 million of our attached SEs, but the SEs can be in mostly switchers and they can be in NE, there will be a bunch in there, but they only spent some time in there.

So we pulled those guys out, we wanna keep the, you're doing this for real kind of separation as opposed to your-

>> Speaker 18: Any guys, again, sorry.

>> Ellen McGrattan: They are any non-employed, you spent a year out of employment.

>> John B. Taylor: Yeah, that's it, it's a lot of them.

>> Ellen McGrattan: There's a lot of them, there's a lot of them.

Okay, let me show you results, John, am I going to one,

>> John B. Taylor: Ready?

>> Ellen McGrattan: No John, are you ready?

>> No.

>> Speaker 15: 1:30.

>> Ellen McGrattan: 1:30, okay, 1:30, good, okay, so now, I wanna show you the results of the estimation, I'm gonna start with kinda one of our headline pictures.

And this should say the red should say attached, self-employed, switcher and attached paid-employed. Again, these are in thousands of dollars, these are integrated, I should have written integrated profiles, so we took care of the mean and then we integrated those gammas and betas over time. Does that make sense, Patrick's giving me the that doesn't make any sense at all face.

 

>> Patrick: Okay, I'm not good at this, I'm not good at this, but I thought, of the 9 million firms or whatever, like 8,900,000 or one guy and his brother who cuts grass, and I'm thinking if you take means of that, how the hell you getting $200,000 bucks? That's all I'm thinking.

 

>> Ellen McGrattan: I know, your guy who cut the grass, Patrick he's probably in NE.

>> Patrick: Cuz he probably cuts the grass for a few years, did something else got out of the sample, et cetera, he didn't cut grass for 30

>> Ellen McGrattan: That's why we want to have kind of the separation of, hey, you did some self-employment to, no, you do self-employment.

 

>> Patrick: You're a serious guy.

>> Ellen McGrattan: You're a serious guy.

>> Bob: Cuz that guy cuts the grass and he went and worked for a company for a while, he was paid, then he came back and fixed some shoes or something, he's in all in and not any.

>> Ellen McGrattan: He got fired and he's temporarily working.

 

>> Patrick: How's he doing that? Yeah, yeah.

>> Ellen McGrattan: Okay, okay.

>> Eric: Yes but, what about the ones that start and fail, I mean, at some point this, I was gonna ask a question about risk at some point.

>> Ellen McGrattan: Good, good, cuz risk is coming and entry and exit are coming, where's that coming?

 

>> John B. Taylor: A lot of the people who failed will be in-

>> Ellen McGrattan: PE, that's right, or any, yeah.

>> John B. Taylor: So, when we do the calculation then we have to think about

>> Ellen McGrattan: Well, at the end of these slides, the chance, I'm getting it to it, cuz Patrick's in this room is zeal.

But at the end, we do the young entrepreneurs and we look at the stayers and switchers, if I can't get to it, Eric, I wanna show it to you anyway.

>> John B. Taylor: Another way then potentially cut the data is look at people who enter at given point in T relative to a control group, enter self-employment and then figure out what happens to their lifetime income.

 

>> Ellen McGrattan: So, I'm gonna show you, well, I'm gonna show you some That would be another, I mean, a lot of what I remember the survey data stuff. Okay, so I'm gonna talk to you after and I'm gonna write down all the things we should do so that we can talk to them.

 

>> Speaker 15: But aren't you worried this especially business cycle frequencies that you know super well that I am at a corner store, It's a good coffee.

>> Ellen McGrattan: I'm about to show you 2008 and 2009.

>> Speaker 15: Now my question on the data construction. So it's moderately successful, but a bad cycle, the recession hits.

 

>> Ellen McGrattan: Yep.

>> Speaker 15: And I decide to step out, collect my thoughts.

>> Ellen McGrattan: Yeah, I'm gonna show you exactly that. So I'm putting you off.

>> Speaker 15: But if it is one year, you're going to kill me.

>> John B. Taylor: Let's go.

>> Ellen McGrattan: I'm gonna show you that in a second, okay, hold on.

So headline number, just so we're clear, Patrick's grass guy. If you ignore those guys and you really look at the ones who stick, you're seeing a big difference. Now I wanna have one caveat besides the caveat that Bob brought up, which is there's even more income when they sell ultimately.

 

>> Patrick: Yep.

>> Ellen McGrattan: Which is these are the guys, okay, let's go back to the tax gap stuff. These are their favorite people because they run a business. There's a lot of who knows what going on if we look at data off of audits. And here, so let me just, this is this figure that I'm showing you.

The red is reported incomes on their tax filings, and blue and black, those are reported incomes. If you look at aggregates that actually the BEA takes from the IR's audit for aggregated pass throughs. Okay, just the pass throughs. The BEA would report that in 2012, since I'm doing everything in 20, $12, I did 2012, there were reported net income of 1,200 billion.

But the misreported net income is 700 billion, okay? And these guys, the attached SE, are a lot of that income. In other words, even if I'm down to a small number, 1.9, and I drop all the grass guys, they've got a lot of the income.

>> Bob: I got a question on the up top.

 

>> Ellen McGrattan: Yes?

>> Speaker 19: Hi, I'm sorry, get so many people talking all this, it's hard to tell. Listen, I think this is great stuff, but I have a quick question about your sampling, and from my perspective, what would be interesting, I appreciate that most of the people here are doctors, dentists, lawyers and accountants.

And that's gonna make up a huge amount. And yeah, their income can keep going as long as the brain is still working. And they want to because it's the nature of their business. What I'm curious to know is, especially since you're looking at ages, do you or can you answer this question that over the time period that you've looked at that you have seen among younger people distinctive?

Well, two things, one, a distinctive switch or rise, I should say, into entrepreneurship in technology related businesses, software. I mean, you can fill in the blanks as well as I can. And whether or not that's a growing share on the delta, if it's a growing share of entrepreneurship, and whether or not that growing share is evident among younger workers.

 

>> Ellen McGrattan: Those are good questions, I mean, that could be figured out. I have not figured it out or done those calculations.

>> Speaker 19: And the reason I'm asking is because I think that there is, and some of what you're doing here in terms of people, entrepreneurship, higher life cycle work and all that kind of stuff about the, I can't remember the exact phrase.

I know Phelps uses it about, basically, the psychic return to work and all that. And I'm curious to see, I think from, not from the revenue standpoint versus what the IRS could be getting, but just whether there are shifts and changes in the economy that we're not getting where younger people are.

Because the barrier to self entrepreneurship is lower today because of technology, that you see a decided shift in younger people going on their own to start some business or whatever it is. And I think that's a very interesting narrative for the direction of the economy and labor and all that kind of stuff going forward.

So that's why I'm asking to see whether or not you've come across that.

>> Ellen McGrattan: I have not done the specific calculations, but that might be something we should think about including. Okay, so let me just finish up here. So this is just getting back to, is there scope for shrinking the tax cut?

Yes, small number of people, big income, big part of the misreported income, yeah.

>> Bob: So there was a big reform about ten years ago greatly extending the 1,099 reporting requirement. And the 1,099 reporting requirement is a very, very effective tool, because when you get a 1099, you know that the IR's has got your address and they wanna match that with your survey.

And in terms of improving the take of taxes, be a very interesting study. You have the tools to study that.

>> Ellen McGrattan: The impact of that change. Yeah.

>> Bob: That famous number 68, 17, two numbers there, the BEA study was from before the 1,099 reform.

>> Ellen McGrattan: No, that's true, I should note that these are very outdated calculations that people do when they're filling in the misreported, which is why there is interest in the IRS to have a little thinking outside the box.

Because they do not have the same tools as they used to under, when you could actually knock on people's doors. They're very restricted in what they can do. And so we are looking for, and we, economists, are looking for economic tools like writing down sensible theories and doing counterfactuals with the kinds of things that Bob's talking about.

 

>> Bob: So you should try including those results.

>> Ellen McGrattan: Okay, so this result speaks to Elena's question. They were the beta in that equation, the time effects. One thing that was good that we are allowing for those time effects to depend on group and in particular between the paid and the self.

And you can see that during 2008, 2009, the self employed took some hits to growth. They took a little over 10% hits to growth and then come back up after that, whereas, the paid employees are only about down 1%.

>> Eric: Does that differ dramatically by industry?

>> Ellen McGrattan: It's a good question, I should look at that, I'm gonna guess, yes.

 

>> John B. Taylor: I'm gonna guess based on your work, the manufacturing guys took the biggest hit, that's my guess.

>> Eric: Of there as well, and that's about a third of the self employed, yeah.

>> Ellen McGrattan: We should break that down, because it would be just for the narrative. Was there big deviations, industry specific?

Okay, this is the growth profile, so this has, these are like the age effects. Plus we've added in just the trend. But this, again, in thousands of $20, you're seeing that. Think of this as the promotions people are getting, every year and the, the attached self employed are getting, roughly six to $8,000 a year.

But that stays pretty flat.

>> Speaker 9: It's a standardized zero, or is it zero? Are they standardized growth rates or?

>> Ellen McGrattan: No, this is thousands of dollars. It's like you don't need to get out a slide rule for this kind of number, okay? It's just 6,000 to 8,000, but what I wanna point out is the profiles look very similar across our SE people at the top.

They have this kind of hump shape, which is very hard to get, theoretically, and I can talk about that later if we get time, which I won't, but it's very hard to get that unless you have some sort of firm, specific investment of the business. The paid always look like this down, like they have their highest growth rates early and then it goes down from there.

So they have these concave shaped income profiles where the others are more s shaped profiles. Meaning there's a little bit of a delay and then they take off.

>> Speaker 8: Is that just inflation in the later years? They're not getting nominal

>> Ellen McGrattan: Everything's real.

>> Speaker 8: Right, that's what I'm saying so is the negative later on just inflation?

 

>> Ellen McGrattan: No, the negative later on, the 55 after?

>> Speaker 8: Yeah.

>> Ellen McGrattan: These guys, are sailing into retirement.

>> Speaker 8: Actually. If they just cut back on hours wage rates, it would go down.

>> Ellen McGrattan: Yeah, that's right. Okay, so just to give you.

>> Patrick: A brand or something like that, that'll give you the picture.

 

>> Ellen McGrattan: That's what, that's what we're talking about, exactly. Otherwise you would just say, I know my type, I'm gonna. No, you've got to build things up, there's startups. Okay, so just to give you a sense, like, let me just give you a sense of the kinds of things we can look at.

We can disaggregate down to. This is just a poll, this is whatever my RA picked. But there's a thing in here that's funny, so let me try it. So he picked men who are married, married is big, who work in professional services, they're educated, they're interpersonally skilled.

>> Speaker 13: Ouch.

 

>> Ellen McGrattan: They're not manually skilled, they're not cognitively skilled. That's the funny part for me, it's like half my students, they're educated, but are they really cognitively skilled, I don't know.

>> John B. Taylor: Let me guess, they graduated from Harvard.

>> Ellen McGrattan: Yeah, and in this case they're either attached. So these are two groups either attached to paid or attached to self.

And we can draw the same thing for these very micro, but there's a ton of data underneath there. So perfectly disclosable things that we want the IR's to be able to pop up online. And you guys could go and say, let me get that group. And then you could draw these profiles as well.

But this shows you like we're getting again. And it's not, I'm not cherry picking I see this all the time, the same pattern. I'm getting this hump shape and I'm getting the downward slope.

>> Bob: This was all the disaggregation, this is includes the disaggregation you went through includes all these categories.

For example-

>> Ellen McGrattan: That's that group.

>> Patrick: It was two of the 30 some thousand.

>> Ellen McGrattan: Two of the 35,117 two subgroups. But those groups are still, when you have enough data, you can get to that level. Now, just to give you a sense that there are some key people.

Now here again, it goes back to kind of like, who are our key people. We can take the differences in those growth profiles at the max, say when they're 33, 34, and ask which groups make up more than half of the total. So there's going to be seven groups that get me the share of that growth gap being most of it.

And I can do it, tell you by industry, their maleness, their marriedness, etc, okay. And so the way you read this is the first group would give you 15%, the first two groups would give you 27% and so on. So doctors, professional services, doctors, again, finance guys, professional services, construction contractors, retail, they're all married males.

They're all educated and interpersonally skilled. Some of them are cognitively skilled and one of them are manually, but I don't know who is that like people who like.

>> John B. Taylor: The last three columns are just some aggregation of occupation imputations, right?

>> Ellen McGrattan: Yes.

>> John B. Taylor: Okay.

>> Ellen McGrattan: We took the occupation code that they gave us.

Then we apply like the algorithm that Jeremy and Posto Benet, yeah. To map from sock code to skill. Cuz then we can reduce the dimension a lot. One could go in, get these occupations, but you want to group them somehow. Okay, I've got 15 minutes, so I'm gonna kinda I'm gonna.

 

>> Bob: Bellman equations. That's a good way to.

>> Ellen McGrattan: Okay, so I'm going to show you the Bellman equation at the end because I would be interested, okay. Tracking the dollars. So there's tracking the people. That's what you can do in survey data, but there's tracking the dollars, which kind of need, you know, some of our guys.

So what we did was we took for each industry cohort, gender, we ranked individuals by average income. Why are we doing this? We're not want to say, finance people are here and, contractors are here. We wanna take for each group kind of who are, at the top of their group.

And in terms of PE income and SE income, how much are they accounting for? So we ranked every, nurses, we ranked contractors, we ranked guys who cut lawn, everybody, and then we constructed their income shares to see how much is kind of in the, like 75 plus. I give you the all the self income and the paid income.

So here's the key takeaway from this. If you look at the 90% of the typical dollar for self employment. Sorry, the self employment income is with the 90th percentile of, I'll call them high ranked people. And that's across the board I've weighted them up by their count.

>> Bob: Tell me again that last line.

 

>> Ellen McGrattan: So 64% of self employment income is gonna be with people who probably would be top coded in their category.

>> Patrick: I see.

>> Ellen McGrattan: Okay, let me do volatility patterns because I want to talk a little bit about risk, which is a bit of a work in progress, but I want to tell you what we've found so far.

So we're gonna take the-

>> Patrick: Gives you that pattern, the one you had before, if we did that in the CPS, you would get something close like the.

>> Ellen McGrattan: It's a good question, we should do this exact thing. Yeah, now, obviously for our guys, we can't have we do a bunch of different sampling, but we could do this calculation, yeah.

Okay, so let me do the following thing. We're gonna compute dispersion now in the, let's call it the residual growth. So after you've taken off, the kind of time and age effects, you've got the residual income. And we're gonna divide by the absolute value of the lag, obviously, cuz we have some negatives.

And we're gonna do it two ways. We're gonna do like pool everybody, and we're gonna also do average subgrouping, it turns out not to matter. Results are gonna show, and I'll show you one picture, but results are gonna show that the cell replied are about 2 to 3 times more dispersed.

So the standard deviation is about 3, 2 to 3. So the variance would be, say 9. These decrease with age, so they come down as the means go up. And it's almost all within group variation. So if we split it to try to say, is it finance guys over here and construction workers over here?

No, okay, the main picture I'm just doing here, the 9,010. So this is the dispersion by age of that income change. And you can see that they're kind of coming together with age. You can see kind of the rough factor of 3, some ongoing stuff. We're trying to map this into things like, think Lucas and his welfare costs of shocks.

We're using other information and trying to do kind of what about and card, and those guys would be doing, which is looking at autocorrelations and giving interpretable numbers. One thing that's different is cards. When we try to replicate them, there's gonna be an issue about the top coding.

So we're trying to figure that out now. But what we do see is that our self and paid employees have similar auto correlations. So it's really, the big difference between them is on the dispersion.

>> Speaker 9: Just understand, is there a selected percentile or distribution of changes over which time period?

 

>> Ellen McGrattan: So this is over our sample, but we do it by age. Okay, let me talk about entry and exit, yep?

>> Bob: I think Patrick will like this, if you just write down a dynamic program and then solve it out. Then you can get essentially exact results for, the constant utility function.

 

>> Ellen McGrattan: He's saying you could do what he does?

>> Bob: Yes, exactly.

>> Ellen McGrattan: That's what he's trying to say I'll translate for you.

>> John B. Taylor: Okay, so entering.

>> Speaker 13: Theme many times in my life.

>> Ellen McGrattan: Yeah, good.

>> Patrick: Especially since I arrived at this.

>> Ellen McGrattan: Good, okay, so we're gonna compute frequencies of switches in and out of self employment.

I'm gonna skip these results, I'm gonna skip the first set of results, because we get similar magnitudes to the surveys. The entry and exit looks the same.

>> Speaker 15: Wow.

>> Ellen McGrattan: The only thing is there's, I want to come back to something you said, which is you're not seeing big drop offs for the great recession.

So here's entry to an exiting.

>> John B. Taylor: Different concept of self employment now, that's what I was wondering because the previous one was based on what you did over 15 years.

>> Ellen McGrattan: Yeah, sorry, thank you so much, thank you. Now we're talking about people who switch, thank you.

>> John B. Taylor: So this is, this is year by year definition now.

 

>> Ellen McGrattan: Yes, so now we wanna talk about the determinants of the choice. We include all people, but an important group are these guys who switch, because then we can compare and twins who switched and twins who didn't switch, thank you, sorry. I shouldn't have said that.

>> Bob: For a year, they go out the next year, they might come back five years later.

 

>> Ellen McGrattan: Exactly, and we want to kind of compare them as they're going in and I'll show you in a minute, like how do they look if there's another guy who didn't go in?

>> Bob: This was the first part where you said they don't have huge amounts, that first slide is from this part, anyway, go ahead.

 

>> Patrick: You said they don't look like they built up all their income waiting to get in, this is where this part's coming.

>> Ellen McGrattan: This is where this comes now.

>> Patrick: Okay.

>> Ellen McGrattan: Okay.

>> Speaker 8: Is it all people or people who at least enter once?

>> Ellen McGrattan: So we're gonna be computing frequencies of switches into and out of SE for all people in our sample, by age and by year.

Let me just jump to it cause it's taking too long. So this is the switch in the exit and the entry. So if I had gone to, a survey, I would have seen that same picture, even same magnitudes. And the entry in, then if I look by year, so this is where I'm looking for, is there anything funny going on in 2000?

2000, we expected something to be going on and it's not really, you can't really see 2008, 2009. So that's the only new bit there. Let me now talk about the determinants so that gets to what you were saying, Patrick. I want to do kind of like, I hate to do treatment and control in front of people who are actually, like empirical people do it for the right quotes.

But we really wanna do kind of twins and ask of these twins, who are similar in all the characteristics and all the things that we've got. And somebody goes in and somebody doesn't, how do they look? I call it the misfit hypothesis cause I love, that's what Evans and Layton call it.

That you have these people who have low past PE income and they're kind of using self employment as the backup option. So what we did was we computed, and here, let me just fill in, X is PE income before the switch. We looked at you, we calculated yours, your incomes before the switch.

For every person, we're gonna find a bunch of matches. And we can find a bunch of matches because we have a huge amount. I'm gonna give you kind of a visual probe it almost. I'm gonna show you what happens between the person who switched and the person who switched later.

We also did it for the person who never switched. But let's do somebody who's even closer, somebody who switches later, and I'm gonna read the bottom because this is the punchline. The wage income is higher for the switchers. So this is the switcher income relative to past income relative to the mate.

So they had higher past income than the mate. And this gives you the amount. I'm gonna repeat this, for asset income. I'm gonna ask, did you have-

>> Bob: Yep.

>> Ellen McGrattan: Yeah, but here, Patrick, I'm gonna do one thing, I'm gonna also add in that you're similar in past paid income, so, I'm not comparing people who are completely different.

Here we see the switcher has less asset income. And we did it two ways, we did just the flow, like dividends and so on, and then we also added in capital gains to see if. And either way, we always get this picture where you have less.

>> Patrick: Denarius would flunk those two, correct, okay?

 

>> Bob: Now you say asset income, which may not. How closely related is that to asset level?

>> Ellen McGrattan: Great question, because we don't see, you don't write down your house, you don't write down these things.

>> Bob: You don't have data on the asset level.

>> Ellen McGrattan: Right, so.

>> Bob: That's what I would think an entrepreneur would care about, the important.

 

>> Ellen McGrattan: But what we're trying to do is figure out what they have available to say buy a worker, or buy some supplies. So, that's what we're trying for.

>> Bob: Yeah, but that could we.

>> Ellen McGrattan: There's one thing we can do, and I, and maybe I follow up later, but there is some information about their house based on like what they itemize.

But even there, you need a loan. Okay, so I get two minutes. Patrick is the dynamic program button, which I won't press this, it'll be too much. But there's two features in the dynamic program that we're building in and two more that I have. We have more work to do.

Two features is, bringing in the investment. We're seeing that very slow growth. So we obviously need, the only way you're gonna get that is, there's gonna be some difference. And we see differences between even the early entrepreneurs, in those growth profiles. If they leave, they look like paid employees from the start.

And then, obviously we need the exit and entry to look reasonable. And we need some sort of learning, because if you knew right away, you would just jump right away. So you need some sort of learning, to get that. Now there's two pieces missing, that have to come in, which is features related to the risk.

So this gets back to what Bob was talking about, and thinking about, how risky are the incomes, and how much insurance do they have, right? Is the wife providing or husband providing insurance? Do they have a lot of other income to go get through any kind of big shocks?

So those will be coming in to kinda round out the full thing. But this is what we have so far, and definitely happy to. If anybody wants me to unfurl the.

>> Patrick: Yeah, I have a question.

>> Ellen McGrattan: Yeah.

>> Patrick: Pretend we all know Kajeti Dinardi type models where the people are self financing.

They build up the money ahead of time, and they have two characteristics. They can be good at one, good at the other. They go to be entrepreneurs when they're bad at being workers, and they build up money. Yours says the opposite of that. Your theory must deal with that.

Can you give us an intuitive story on how yours works and.

>> Ellen McGrattan: What we're gonna-

>> Patrick: Their forces are simple, right? If I'm bad, if I'm bad at working in the workplace, and I build up some money, and I have some, this other characteristic makes me good. I jumped to being entrepreneurs.

Yours is saying, would have a hard time getting those pictures that you said where the guys are doing well. In your model, you must get that.

>> Ellen McGrattan: So, at the end if you wait around, we were trying to get that hump shaped growth profile. So that's what we were looking for.

 

>> Patrick: The guys who are doing better, in their old job, and they say, I'm doing pretty me and us too. I'm doing better than him, in my paid job. But then I jump to be an entrepreneur. When their models are predicted, he would jump to be the entrepreneur.

That's what I'm confused about.

>> John B. Taylor: Your hump shapes, were for a heavily selected sample. There were attached use those pictures, to make inferences about the decisions that people are making early in their life cycle.

>> Ellen McGrattan: Well, okay.

>> John B. Taylor: Whether to be an occupier. In this case, we started as entrepreneurs and then you could exit out.

And then, we tracked them.

>> Patrick: I thought they did two twins going along.

>> Speaker 15: No, no.

>> Ellen McGrattan: And anybody who wants to fight with Pat, please stay.

>> John B. Taylor: They selected by the fact you were there 12 years. Yes.

>> Ellen McGrattan: Thank you, John, this opportunity.