The Marginal Net Taxation Of Americans’ Labor Supply

>> John Taylor: Hey, we're very happy to have Alan Auerbach speak to us about marginal net taxation of Americans labor supply. Anyway, this is a joint paper. Anybody else? Anyway, it's up-to-date, January 22nd, 2025. We're very happy to have you come across the bay. Anytime you want, you're here. Alan, go ahead.

>> Alan Auerbach: Thank you very much, John, and great to be here. So as John noted, this is a joint product of efforts of five of us and it's been going on for a long time and part of a larger research agenda as well in terms of the modeling. So there couldn't be a more central issue in taxation than the effect, and as Michael Boskin certainly knows, as the effects of marginal tax rates on labor supply and both employment hours.

And it's long been understood that you don't simply want to look at say a particular tax like the income tax, you want to look at all taxes, payroll taxes. And indeed consumption taxes because to the extent you earn anything, you have to do something with it in order to make it worthwhile.

But there are also implicit taxes lost through benefit, the way benefit programs work. And this too is not something that is a new thought. In fact, I can remember decades ago hearing about concerns about a so called poverty trap where people who get public assistance of various kinds.

If to the extent that they reach a certain income threshold, they may begin to lose benefits. And to the extent that these are not carefully designed programs, their interaction can really lead to high marginal tax rates. Implicit in this case because they're not working through the tax system, but presumably to the extent that people understand how they work should have the same impact on their behavior.

Somewhat less considered in the past is the fact that future taxes ought to be affecting these things as well. I mean, there have been some important papers on this subject. Looking for example at the extent to which people's saving for retirement might be discouraged by the fact that there are various benefit programs they might get in retirement.

That they won't qualify for if they earn too much and save too much. But it doesn't just affect people in retirement, it affects people throughout their lifetimes. And these can have effects as well, not just future taxes. There are consequences of current decisions to work, but also the receipt of future benefits.

Certainly during retirement years is an important time of benefit receipt, but it's not the exclusive one. So what do we do here? We take these things under consideration, the effects on labor supply of taxes and benefit programs both in the year of the labor supply decision as well as in the future.

And we do it in a very detailed manner, modeling a vast array of federal, state, local taxes, benefit programs. And in order to do this, because we're looking into the future, we also have to have projected passive income and spending. And then we just do an experiment. We say suppose you throw a person another thousand dollars of labor income on a person in a current year.

What does that do to the present value of consumable resources that that person has? And if it's the extent that it's less than $1,000, that's a tax. And we estimate that, and we estimate a tax rate by just dividing the taxes by the thousand dollars assumed. We start, this is based, I mean, the starting sample, the sample of households we start with is based on the 2019 Survey of Consumer Finances from the Fed. And then in a moment I'll explain how we bring in other data sets to get all this to work. So, but before. Yeah, John.

>> John Cochrane: It's going to depend on when the household chooses to spend the money, though.

>> Alan Auerbach: That's right. And as I'll say in just a moment, we are making an assumption, and we're making the same assumption for every household which is basically consumption smoothing or consumption moving subject to borrowing constraints. So it actually leads to upward sloping consumption profiles. And we've experimented a little bit with different rates of time preference and it doesn't make that much difference anyway.

>> John Cochrane: Tax on rates of return.

>> Alan Auerbach: Yeah, taxes on rates of return will be in here and they'd be in here even more if we, you know, had a lower rate of time for, you know, higher value on the future.

>> John Cochrane: Also, corporate taxes matter because-

>> Alan Auerbach: They're in here.

>> John Cochrane: They're in there too.

>> Alan Auerbach: They're in here, yeah. As you'll see in a moment. Sorry, so let me just give you a little teaser of some of our key findings. First of all, over half of US adult households have all in marginal tax rates of over 40%.

So to start with, it includes federal taxes, state taxes, so you can see right away how you could get up there. But then, it also includes any additional taxes, payroll taxes, and as well as any effects on benefit program, receipts of benefit programs. The pattern of these marginal tax rates is progressive, increasing with respect to the magnitude of a household's lifetime resources.

Since we're looking at this over lifetimes, we sort households by their lifetime resources, that is their wealth plus the present value of their future are projections of their future life or income. For those in the bottom quintile there's a lot of dispersion there's much more dispersion of people in terms of the tax rates of people at the bottom.

The intuition for that is just these are the people who are subject to all kinds of possible phase outs and phase ends. So they can have very negative marginal tax rates they can have very positive marginal tax rates and they're all over the place. But getting back to the point about the poverty trap that I mentioned before one in four of these households in the bottom quintile face marginal tax rates above 50%.

And you know that's not coming from their marginal income tax federal income tax rate for 1 in 10 it's over 70% and these are would be higher if we had had full program take up. In fact in our original draft of this paper we assumed that households took advantage of whatever benefit programs they qualified for on paper and we were criticized for doing that because in many cases there.

A lot of households don't either because there's rationing such as in section 8 housing or they just there's stigma or lack of information or other reasons. And so, for some programs there's very high participation but for other programs there's lower participation. And so, we now actually build that in.

And so, we target getting the participation rates in our sample being consistent with what aggregate participation rates are. These marginal tax rates would be even higher if we assume full participation because you get a lot more people being hit by phase outs and loss of benefits.

>> Speaker 4: Alan, this is a great exercise. Thank you for doing this. I wanted to ask you, some of the variation in these marginal tax rates is coming from variation in take up. So, conceptually what are you thinking about the, the correlation between variation and take up and sort of people's expected lifetime marginal tax rates? Cuz this is a choice whether they take it up or not.

>> Alan Auerbach: We are not something we, it's actually an interesting point. Something we don't do is, we are matching. It's quite complicated but we are trying to match households using their propensity scores in terms of who gets, who's in and who's out.

And we sort of add households to the take up until we get to the aggregate amount where we don't have information on which ones take it, which ones don't. We don't make it dependent on marginal tax rates which we could. I mean I should say that would be interactive of course because as, as we put them in then their marginal tax rates can change. But that's something that we could take into account, yeah, John.

>> John: How does it compare with the top quintile?

>> Alan Auerbach: The top quintile has higher median marginal tax rates but far less dispersion. And so, I'll actually show you a table, where you can see.

>> John Cochrane: Casey Mulligan and Phil Glam and company get numbers more like 100%. I'm just curious, what's the difference? Is that in kind benefits or

>> Alan Auerbach: We have numbers, some of the people are above 100%. It's just that the top 10% are not 70%-

>> John Cochrane: Quintile within 100%.

>> Alan Auerbach: Yeah, I don't know. It could be that they're assuming full participation. Although even there we wouldn't get that high. Don't know, yes.

>> Speaker 6: How much of the dispersion depends on the size of the family.

>> Alan Auerbach: We were asked that question and we did some calculations of people with and without children. And I believe it's more dispersion with children than without. But I can't remember the quantitative reports.

>> Speaker 6: Is that because there's more programs with children?

>> Alan Auerbach: Yeah, that's right.

>> Speaker 7: Easy to do this for different assumptions about the tax code because 40% is quite high. I think it's interesting to sort of look, say, at the tax reductions in the Trump administration, the first-

>> Alan Auerbach: That would be easy to do. We have not done that. Maybe when the law changes again, we will.

>> Speaker 7: No, I just think it's very interesting to go back to the 1980s. Presumably, then over half of US households had much higher marginal rates of tax. Given the-

>> Alan Auerbach: I don't know about that because the programs have gotten.

I mean, one thing I should mention here is, we are counting benefits as being valued at their cost. And so this comes up. I was just gonna say, programs like Medicare are much more expensive or Medicaid or much more expensive. And that shows up in terms of causing more variation, more movement in the marginal tax rates than it would have in the 1980s.

You might take issue with that and say, well, but if people only value health care at a third of what it costs to give it to them, then maybe we're overstating some of this stuff.

>> Michael Boskin: Yeah, that's the same point I was gonna make. All this evidence that, revealed evidence from discernments, medicated and so on, amongst other things, because they have other options like county hospitals, they value in kind, benefits much less.

>> Alan Auerbach: Right, it is.

>> Michael Boskin: So, you're saying basically, we're gonna sign the cost, the average cost per person or something. That's the value you get if you.

>> Alan Auerbach: That's true and we're not-.

>> Michael Boskin: Above the level.

>> Alan Auerbach: We are not, and for example, for healthcare, we're not building in charity care like that, that's right.

>> John: Excuse me, I'm sorry. There's one other thing.

>> Alan Auerbach: Sure.

>> John: You said consumer finance. Right, are these longitudinal data?

>> Alan Auerbach: No.

>> John: So you're not following the same people over time?

>> Alan Auerbach: No, no, we have a-

>> John: Let me assume, the cross section.

>> Alan Auerbach: We have a cross section-

>> John: Is correspondence to lifetime.

>> Alan Auerbach: I'll say a little bit more about how we do all this.

>> Michael Boskin: So, Bruce Miner has this huge project, right, at Chicago Booth, assembling all these various programs and the tax rates implied by that. And so, looks similar, at least in spirit to what you're doing here.

Do you know about that. And can you talk?

>> Alan Auerbach: Bruce has seen this work. I've talked to him about it. I honestly can't give you a characterization of what the overlap is and what's different.

>> Michael Boskin: Okay.

>> Alan Auerbach: Sure.

>> Michael Boskin: Conceptually appropriate. How do you treat Social Security? And it's hiding benefits. Well, or do you.

>> Alan Auerbach: Sure, in fact, I'll give you an example, case study where someone's Social Security benefits go up because they work more.

>> Michael Boskin: Yeah.

>> Alan Auerbach: So, we do have the benefit.

>> Alan Auerbach: Yes, we have the Social Security rules built in, so that your benefit are based on your average index monthly earnings and which go up a little bit if you-.

>> Michael Boskin: Does the same people know that.

>> Alan Auerbach: Yeah, I mean, all of the other assumption we're making here is that we're calculating your marginal tax rate and we are not estimating labor supply responses. And if we were, then we'd have to really confront the question of what people know.

And that can be really important in some cases. For example, the earnings test on Social Security. Do people really understand that they get more benefits in the future even though they're losing benefits now? It's a question.

>> Michael Boskin: And whether 8% is something.

>> Alan Auerbach: Yeah.

>> Michael Boskin: Corresponds to how they value the future.

>> Alan Auerbach: Anyway, okay, there are big differences across states. We actually do some calculations across states too. So, controlling for income and also controlling for income age, there's still-.

>> Michael Boskin: Control for price level differences across. I don't think so. This is all spatially not.

>> Alan Auerbach: Yep, okay. And current year marginal tax rates understate the full impacts.

Okay, and finally, at the end, we do what I would call a simplified dead weight loss calculation because it involves various assumptions. But we just say, suppose you just consider the conceptual experiment. Suppose you had a way of just making everybody's in any given income group made their marginal tax rates.

It's the same, equal to the mean of the group. So basically removed all dispersion in the group. That could lead to substantial reductions in dead weight loss, improvements in economic efficiency. Basically, because of the nonlinearity of deadweight loss, you get marginal tax rates up to a very, very high number.

It makes the distortions pretty bad or potentially pretty bad.

>> Michael Boskin: You take our consumption taxes and make them the labor tax equivalent.

>> Alan Auerbach: Yes, yes, so, well, we're doing these on a lifetime basis. So everything is converted into a labor tax equivalent. Okay, let me just give you some illustrative cases.

These are three actual households. So here's a household in Arizona that this is just an illustration. This is an affluent household, a single individual. And it's just to show, it shows you the taxes are 360 out of the additional thousand dollars earned in the current year, $360 in taxes.

This is a household's not getting any benefits because it's too affluent, the individual's too affluent to get any of the programs, pre retirement programs. And then that individual, the benefits that individual gets Social Security and Medicaid aren't affected by earning an extra in the future. That's lifetime. L is at the top is lifetime versus current.

But the difference, which is the change in taxes is higher over a lifetime than it is in the current year simply because of the double taxation of savings. And so you have labor income taxes now, and you have some consumption taxes now. But you have some consumption taxes in the future and you have some taxes on capital income in the future as well.

So that's just a simple illustration. Here's a household. Where does the poverty trap come from? Well, here's an actual household that according to our calculations, we're running these households through our calculations. So we're not necessarily seeing all the benefits they're getting or what they would get, but we're doing the calculation of what they would get.

And this is a household that if it works, earns an extra thousand dollars, would lose $6,500 of SNAP benefits in the current. So that gives us a huge marginal tax rate. And then finally, this is a household which the current effects are very different from the longer run effects.

So this is a household that earns an extra thousand dollars, pays an extra $146 in taxes, loses $223 in SNAP benefits. So they still get SNAP benefits that just get less because they're in the phase out range. Okay, it shows how the marginal tax rate has to take account of implicit taxes from benefit loss, not just the actual tax increase from explicit taxes.

But then if you look in the future, this household actually gets a big increase in benefits, Supplemental Security Income benefits, which is a program for the indigent elderly. Because this is a household that is not well to do and receives these benefits in the future by earning more in the present.

This individual pays more taxes and you know, given the way the programs work, this individual is actually going to qualify for additional SSI benefits in the future. And it happens that the individual is going to therefore have negative taxes in the future. So-

>> Speaker 9: Why did earning more money make them get more SSI benefits in the future?

>> Alan Auerbach: The couple loses current year benefits and then they save less and makes them eligible for more SSI.

>> Michael Boskin: What, more net?

>> Alan Auerbach: Yeah, it's complicated. I honestly I was looking at this morning and I was thinking I need to go back and look at that again.

>> Speaker 10: Do you have an idea? Just basic logic. They're still getting lifts 600 something more currently, right?

>> Alan Auerbach: The answer must be that they're losing some time between now and then and when they retire. They're losing some additional benefits because they have additional income. That income, by our assumption that income is spread over their lifetime, which means they're going to have additional consumption in the future.

And so it must be that they're getting hit with a loss of snap benefits or other benefits in the intervening years. This is the explanation we have in the paper, which I think is incomplete.

>> John Taylor: What is the current income?

>> Alan Auerbach: Where is current income? It's not listed here. Just taxes and benefits.

>> Speaker 11: How is the discounting being done? Cuz it's got a, vary by a marginal product of capital, net marginal product of capital.

>> Alan Auerbach: So it's like four.

>> Speaker 11: If you give $1,000 to somebody in the 10% level, I think it's going to be gone in the first year, right? They're so constrained.

>> And that would assume we do have borrowing constraints in here. So if this individual, I don't know if this individual is liquidity constrained in the current year. But if the individual is, then all of it will go into.

>> Speaker 12: Can I ask the discounted question slightly differently?

>> Alan Auerbach: Yeah.

>> Speaker 12: Because it is the case that the life expectancy differs with income.

>> Alan Auerbach: I'll get to that.

>> Speaker 12: Do you take in-

>> Alan Auerbach: Okay, so I'll get to that.

>> Speaker 12: Thank you.

>> Alan Auerbach: Yes, I'll get to that.

>> John Cochrane: The assumption is the time preference is the same across all these people.

>> Alan Auerbach: Well, in a sense, yes, because we're assuming everybody is attempting to smooth consumption uniformly. And so you could say they all have Leontief preferences or something like that, but we're not, we're not really talking about it in terms of preferences. We're just sort of passing everybody through this filter which causes spreads out their resources. You're not.

>> Steven Davis: Okay, so these borrowing constraints don't arise because people differ in their preference?

>> Alan Auerbach: No, everybody has the same objectives, but their earnings patterns are different and so their family size is different. Those can all lead to differences in whether they're subject to borrowing, whether they have binding borrowing constraints.

>> Michael Boskin: When you said the first household, the well off household, are they affecting taxation?

>> Alan Auerbach: Yeah.

>> Michael Boskin: So you're assuming the marginal savings done outside of tax deferred accounts or?

>> Alan Auerbach: We are building in what they qualify for.

>> Michael Boskin: And you assume they use the max?

>> Alan Auerbach: I think so, yeah.

>> Michael Boskin: We know what fraction people actually do.

>> Alan Auerbach: Or perhaps we don't, or maybe we assume that they use the same fraction that they already, they currently have. If they're not subject to. You would think I'd remember it, but I don't.

>> Michael Boskin: Well, that could be important.

>> Alan Auerbach: No, no, no, I agree, I agree with you, I agree with you.

>> Speaker 10: The family size came up before, but now understand more clearly what you're doing. So if you have a family and you have two kids, you're still assuming that the parents are gonna consume the equivalence.

>> Alan Auerbach: We have equivalent scales for the kids. And so as long as the kids are minors, we allocate consumption, smooth consumption per effective adult.

>> Speaker 10: I see, okay? Do we think that people-

>> Alan Auerbach: It's in the paper, but I can't remember what the number is. This is not a detailed question. This is more big.

>> Speaker 14: Do we think that people respond in their labor supply decision more to a lifetime marginal tax rate than just to a periodic marginal tax rate?

>> Alan Auerbach: That is an empirical question that I don't think has been effectively evaluated.

>> Speaker 14: Essentially what you're doing is you're saying we need to know this marginal tax rate so that we can evaluate that question empirically.

>> Alan Auerbach: Yeah, yeah.

>> John Cochrane: Suggestion. Since the question of when you earn the money and when you consume it is so different.

>> Alan Auerbach: Yeah.

>> John Cochrane: It might be worth producing different margins for if you consume it today, if you save it, if you pass it on to your kids.

>> Alan Auerbach: Yeah.

>> John Cochrane: You're just gonna get very different margins and trying to put that all into. Well, if I give you money, then I'm going to model.

I see what you wanna do. You wanna produce one number.

>> Alan Auerbach: Yeah.

>> John Cochrane: But it might be useful to have-

>> Alan Auerbach: We've done other experiments where we've said suppose the discount rate was higher or other things like that where we sort of down-weighed future consumption and future taxes.

>> John Cochrane: A way to put it would be we're measuring not so much a marginal tax ship at a wedge. What's the wedge between wager and consumption? Well, it's a different wedge between labor and consumption today and labor and consumption 20 years from now.

>> Alan Auerbach: It is. I'm a little hesitant just because it would explode the number of numbers.

>> John Cochrane: Yes, I understand that.

>> Alan Auerbach: But we could produce both, produce the disaggregated numbers, and then the bottom line.

>> Speaker 14: It is hard to imagine that somebody's sitting there thinking okay, I got an extra thousand dollars today, my snap got cut, but I'm gonna get this SSI in 30 years.

>> Alan Auerbach: I agree with you, I agree with you. And the question is how? You might argue it's realistic for a 55-year-old to think about retirement income, but not a 30-year-old. Yeah.

>> Michael Boskin: Even if it's not plausible to think that, and maybe it's a big stretch. But you think policymakers might want to think about these things, accounting for that, right?

>> Alan Auerbach: That is our argument. We wanna say, this is what happens and how much you should worry about it is a question that you sort of can't answer until you know what. What it is you're worrying about. Josh's comment, I would believe in economists may, then the non-economists would say you really think the system is going to look like that in next years, right?

>> Michael Boskin: That seems the reason why.

>> Alan Auerbach: Well, yeah, that's true. Things can, yeah.

>> Speaker 14: You should ask that question as well.

>> Alan Auerbach: Yeah, things can change.

>> Speaker 14: We should ask that question.

>> Speaker 12: So you can have a higher discount factor, or sorry, lower discount factor on anything that's policy related future.

>> Alan Auerbach: Well, it depends on what you think is gonna happen. I mean, if there's uncertainty, you might-

>> Speaker 12: Have changes in that.

>> Alan Auerbach: I mean, you might work harder. I mean, if you're worried about your Social Security benefits going away-

>> Michael Boskin: Or save more.

>> Alan Auerbach: Yeah.

>> Michael Boskin: Or both.

>> Speaker 14: Well, but that does actually raise another question which is even within the current system, I mean, are you risk adjusting the discount rates for Social Security benefits?

>> Alan Auerbach: We're not applying different discount rates to different programs or anything like that.

>> Speaker 14: Cuz some are more just to the economy.

>> Michael Boskin: So in some sense, if we're really modeling behavior, we start thinking about all these things, consumption, labor supply, or a function of all the net returns rate, net wages, etc. And maybe we do it all in forward prices or something and wash out, right?. That's kind of the concept you have in mind.

>> Alan Auerbach: Yeah. All right, so let me, let me try to move ahead and talk about what we actually did. We start with a sample, as I mentioned, survey of consumer finances. We have this complicated program and in order to sort of have enough information on households, we bring in other surveys and sort of match households in the other surveys to households in the Survey of Consumer Finances.

So for example, we calculate for different cells based on characteristics, including income, age, family structure, and so forth. We use the Current Population Survey to get labor supply trajectories. We have retirement based on information from the American Community Survey. We incorporate this question came up differential mortality, which is obviously important because you have all these annuity based payments when you get old.

And these are the life expectancies we build in. This comes from a National Academy of Sciences panel on aging and benefits and retirement, which I was a part of. And these were the life expectancy at age 50 by people in different. By males in different quantiles based on birth cohort.

And it highlights something that is, I think, by now well known, which is that the income gradient of life expectancy has increased over time. It was always there and it's gotten bigger. And that's quite evident here.

>> John Taylor: I think, Mervyn has a question. Mervyn, you have a question? Guess not.

>> Mervyn: Okay, I do, I do. I had to mute myself, so I'm muted. So it goes back to what you were saying before, Alan. The traditional method of looking at marginal tax rates on labor income looks at what's the value of resources that you earn through an extra unit of work versus the amount of consumption that you could have today.

And obviously even that represents preferences if there are different effective tax rates or different kinds of consumption goods. But it's a simple number, what you're doing. In a way, you could do the same thing by saying, I think someone mentioned this. What's the trade off between work today and consumption 10 years from now, or consumption 20 years from now? And what you're trying to do is to look at the tangent, the slope of the relationship at the optimal path. Is that right?

>> Alan Auerbach: Well, it's not optimal in the sense that we are assuming that households smooth consumption. And so we're not positing a explicit utility function.

>> Mervyn: No, but it is meant to be in some sense representative of what you think people will want to do.

>> Alan Auerbach: We think it's a good, simple default to be using, at least initially.

>> Mervyn: Yeah, so I think it might be interesting just to calculate the marginal tax rates on what happens if all the expenditure is 10 years from now, just to see what difference that makes.

>> Alan Auerbach: Sure, and if we make differences extreme enough, for example, if we assumed all consumption is in the current year, then we'd. Get essentially the current year marginal tax rate that we get.

>> Mervyn: Exactly-

>> Alan Auerbach: And it would vary across that, I agree. It's something that we can do, and we perhaps could do more experiments like that.

>> Mervyn: It's a very interesting way that you've got of extending the norm analysis.

>> Alan Auerbach: Thank you. Okay.

>> Speaker 10: That's a general answer that will be somewhere between the current and the lifetime that you have here if you're making different assumptions about your-

>> Alan Auerbach: We could do that.

>> Speaker 10: I mean, you could argue that for a wide range of-

>> Alan Auerbach: We could do that, we could do that.

And it's a bit frustrating to think of all the things that we could do that would make it-

>> Speaker 10: No, I know.

>> Alan Auerbach: No, no, but I mean, we talked about valuation of in-kind benefits, finite horizons. There are all kinds of things one could do and some of these things we can do.

Some of the things we have done. Some of the things might be a little bit more complicated such as taking uncertainty into account.

>> Michael Boskin: For Mervyn's sake, you could add what country you earned the income.

>> Speaker 10: Yeah.

>> Alan Auerbach: Yeah, well, we're doing 50 states, I think we'll stop-

>> John Taylor: We'll continue on.

>> Alan Auerbach: Yes, thank you. So here are the programs that we use. So the taxes someone asked about corporate income taxes. We have that at the federal and state level. Medicare Part B premiums, including the progressivity to the extent that you have higher income if you're receiving Medicare and various other things.

All the transfer programs that I've been mentioning.

>> Speaker 10: So what would be left out of here? I mean, obviously-

>> Alan Auerbach: Nothing big. We basically added these things over time, so-

>> John Cochrane: So for the corporate tax-

>> Speaker 16: So UI and the corporate tax are the ones that come to mind.

>> Alan Auerbach: Well, corporate tax is here, second line.

>> Speaker 16: Sorry. How do you allocate that?

>> Alan Auerbach: How do we allocate the corporate tax? I think we're just assuming the corporate tax. I think we're just using a Harper assumption.

>> Nick: I'd be curious to see college financial aid phase out.

>> Alan Auerbach: Bet you would, Nick.

>> Alan Auerbach: Yeah, I mean we don't have any private things like that, like financial aid, yeah.

>> Speaker 18: Those cohort differences a slide or two earlier, aren't you concerned that they indicate that things with the given cohort, which are you using, may be very-

>> Alan Auerbach: We're not using it.

I didn't mean to say we're using a different cohort. We divide people into 10 year age ranges and we do all our calculations. We aggregate in a lot of what we do we aggregate? But everything, for example, when we classify household by quintile, the quintile of resources is by age group.

So for example, when we talk about the bottom quintile, we're talking about say 30 to 39 year olds in the bottom quintile, 40 to 49 year olds. And you'll actually see in one of the pictures I show you there's sort of overlapping because, say, people in their 20s who are in the, say, second quintile might be, if they had the same resources in their 40s, would be in the bottom quintile.

So a lot of the analysis we do, we're doing by age cohort. So-

>> John Taylor: Property taxes-

>> Speaker 10: You have transition probabilities movement in some way, either implicit or explicit?

>> Alan Auerbach: You mean do people? No, basically, at a given age, from the CPS we have a projected age earnings profile.

And we only classify them in the current year. So we're only looking at each household once, but we're looking at them over their lifetime.

>> John Cochrane: Property taxes, I think.

>> Alan Auerbach: Property taxes are bundled in with housing, and we have a special treatment of housing.

>> John Cochrane: Okay.

>> Speaker 20: Michael may have mentioned it, but unemployment insurance.

>> Alan Auerbach: Yeah, I thought we were dealing with unemployment insurance, but it's not in here, so it's possible we've left it out. Okay, we take account of, I mentioned incomplete, this just gives you the take-up rates that we actually observe in the last column. So for things like chip, children's Medicaid, the take up rate is very high.

But for other programs like the ACA subsidy, you might say, well, how can it be only 8.5%? Well, this is basically everybody who's in the income range where they would qualify for an ACA subsidy. They may have the employer provided insurance or they may not be buying insurance at all.

There's no penalty on that anymore. And so, that's how you get a number like that. And so this is all built into our calculation.

>> Speaker 10: So the data for the IT exists?

>> Alan Auerbach: Sorry?

>> Speaker 10: Yeah, it's not available for EITC?

>> Alan Auerbach: I don't know what that means, honestly.

We have 78%. I don't know why that is there. Okay, we impute state residency, basically, to match the distribution. We know we have information from the American Community Survey of the populations by state, cuz that's not available in the Survey of Consumer Finances. And we allocate people, basically fractions of the observations in the Survey of Consumer Finances to different states.

So that we get the various things we're targeting in terms of sort of income shares and stuff. So that California looks like California in terms of the income distribution, Mississippi looks like Mississippi, and so forth.

>> Speaker 10: And you're using current rules.

>> Alan Auerbach: Everything is based on current law.

>> Michael Boskin: Law.

>> Alan Auerbach: Well, sorry, current rules, not current rules.

>> Speaker 10: Right, current rules, yeah. So current law, for example.

>> Alan Auerbach: Right, exactly.

>> Speaker 10: Okay, so it's current rules.

>> Alan Auerbach: Current rules.

>> Speaker 10: So do you close the government's budget constraints-

>> Alan Auerbach: No, no, no.

>> Speaker 10: So either margin-.

>> Alan Auerbach: No, no-

>> Michael Boskin: Future generations-

>> Alan Auerbach: No, we behave just like the government.

>> Speaker 10: That's been gradually getting worse over the last administration.

>> Alan Auerbach: It has, I have a paper on it.

>> Speaker 14: So some of the low take-up of something like TANF. I mean, couldn't that be because TANF income could count in your eligibility against some of these other programs?

I mean, is it like program-

>> Alan Auerbach: Yeah, this gets back to the question you asked before, which is it would be interactive.

>> Speaker 14: Yeah, yeah.

>> Alan Auerbach: If you had people optimizing over their choice of benefit programs for precisely that reason. And also, TANF has a lifetime limit. So is this the time to take TANF for some time in the future?

It's at least something we haven't attempted. Okay, let me now talk about my results. I've already previewed some of these. Median marginal tax rates increase with lifetime resources. And the lifetime medians are higher than the current medians throughout the income distribution. And then here's aggregating everybody ages 20 to 69.

And again, when we classify households by quintile and then by percentile, when we're looking within the top quintile, look at the. Also look at the top 5%, the top 1%. These categories are all based on. Age cohorts. So if we say you're in the lowest quintile, you're in the lowest quintile among those who are where the head of household is age 30 to 39.

If you're 30 to 39, and likewise with whatever age you are. And so we're sorting people that way. So it's not the lowest quintile of those resources of those aged 20 to 69, okay? So you can see both that the last, of course, two bars are for the aggregate.

So you wanna look at that when looking at trajectories over the income resource distribution. So it's upward sloping, it's monotonic for both current tax rates and lifetime tax rates. It's higher for lifetime tax rates, especially when you get into the highest income groups. Those are the groups that are more likely to be saving, less likely to be resource constrained, and therefore subject to double taxation.

You've got higher tax rates at the bottom too, again because of not only taxes, future taxes, but also losses of benefits. For example, loss of Medicaid can be a big thing for households at the lower part of the resource distribution.

>> Speaker 10: So two questions about this in the examples that you showed was often that some benefit would remain constant even as you get higher income.

>> Alan Auerbach: Yeah.

>> Speaker 10: Like Social Security, for example. But the top 1%, I mean, that's just minuscule amount at that.

>> Alan Auerbach: It's not benefits for them, it's not a loss of benefits for them, no.

>> Speaker 10: So-

>> Alan Auerbach: It's taxes.

>> Speaker 10: It's tax.

>> Alan Auerbach: Yeah, it's. Yeah, and future consumption tax.

Because when we're looking at the current year marginal tax rate, we're looking at, under our assumption of consumption smoothing, what additional taxes do you pay this year? So it's-

>> Speaker 10: So clearly an example.

>> Alan Auerbach: Okay.

>> Speaker 10: And the other question is, I mean is Zuckman and Zayz and Piketty, they've been looking at these tax rates for the top earners.

>> Alan Auerbach: They have.

>> Speaker 10: How do these numbers stack up?

>> Alan Auerbach: Well, first of all, they are only looking at current years.

>> Speaker 10: Right.

>> Alan Auerbach: Second, they have, I would say unusual incidence assumptions and maybe leave it at that.

>> Speaker 10: Okay.

>> Speaker 21: Just to clarify as well, the resource quintile, is that based on current resources or lifetime resources?

>> Alan Auerbach: It's based on-

>> Speaker 21: Matters, right?

>> Alan Auerbach: Tangible wealth, basically, asset wealth and human wealth.

>> Speaker 10: Into account my whole trajectory that I'm forecasting, where I'll be in the future.

>> Alan Auerbach: That's right, plus your wealth, your tip, which is pretty small at that age group. Which is one of the reasons why we're looking at segregating people by income groups because the life cycle, you know, components of resources are so different.

>> Speaker 22: Somebody with a very high starting wealth doesn't have that extra wage. Your taxation is already taken.

>> Alan Auerbach: Yeah, if you're in your 20s and you have a huge amount of wealth because of an inheritance, presumably. Well, your profile taxes is gonna look very different than a person with a lot of resources in their 50s or something. But you're being compared to people in their 20s.

>> Speaker 23: So what about people with intermittent labor force participation? If you're ex offender or.

>> Alan Auerbach: We have projections, but I don't think we project transitions into and out of the labor force. We look separately at the question of people who are not below retirement age, who are not working and we ask what the marginal tax rate would be on their decision to participate.

That's a separate kind of marginal tax rate we look at. Okay, now if we had ignored incomplete take up, the marginal tax rates wouldn't be monotonic with respect to resources. They'd actually be U shaped. And this is what. So, the dark series is the one I already showed you as in the medians by income group, by resource group.

It would be the pink one if it actually, if we had complete take up. And it only really matters at the bottom cuz that's where the take up, especially in the bottom quintile and to some extent in the second quintile because of the incomplete take up. Because the way we're handling it is, if I'm not taking up a benefit, we're assuming you're just not gonna take that benefit up, and therefore you can't lose it.

If we assumed you did take it up and then your income drove you out of the benefit program, there's considerable dispersion in lifetime marginal tax rates, particularly at the bottom of the resource dist. So here's a scatter of the total resources is on a log scale on the horizontal axis and then the lifetime marginal tax rates.

And this is a scatter of the distribution and you can see all the way to the right is the top quintile. And yeah, there's some variation but as you get toward the bottom it gets really big. And the reason why these color distributions overlap is because this is point about we're classifying people within their age group.

And so you'll have some people from the second quintile, showing up over, in the same resources as some people in the lowest quintile because they're from different age groups. And so they get classified, given the same resources, they would get classified in say in the first versus the second quartile quintile depending on. Because if they're at say age 60, they might have lower resources than if they're age 50 or something like that.

>> Speaker 23: Even, even in the right most quintile, there's still a lot of variation.

>> Alan Auerbach: Yes, there is.

>> Speaker 23: And in terms of potential distortions on output and economy.

>> Alan Auerbach: Yeah.

>> Speaker 23: Those people should get weighted a lot.

>> Alan Auerbach: Yes, yes.

>> Speaker 23: That seemed like at least as big a headline as the purple dots here.

>> Alan Auerbach: It's a matter of perception, but it certainly matters.

>> Speaker 23: Labor supply elasticities, lifetime measures are reasonably high. Then there's a lot of lost welfare.

>> Alan Auerbach: Yeah, there is-

>> Speaker 23: Associated with it.

>> Alan Auerbach: Yeah, there is. I mean I'll show you the simplified calculations when I get John.

>> John Cochrane: And I'm surprised the estate tax doesn't bump it up even more because all of a sudden you're facing a 40%.

>> Alan Auerbach: You got to be really wealthy to be able to pay estate tax. It's a top roughly now the top 1/10 of 1%.

>> John Cochrane: Yeah.

>> Alan Auerbach: They probably earn 20% of the income in econ.

>> John Cochrane: Yeah.

>> Speaker 23: 1% earn well top 1% or 26%. So 1, maybe 15.

>> Speaker 24: Remember in Greg Menkew's favorite calculation of his marginal tax rate with 90%, but he said I'm gona pay on the margin the estate tax and then give it, give it to charity or to his children, my kids. And then I'm gonna lose 40 some percent of my wealth.

>> Alan Auerbach: So this is the distribute. This is just a table version of the picture you just saw. These are distributions of lifetime marginal tax rates. And the 50% and 70% are there. In the red box for the bottom quintile.

So it's basically showing that a quarter of that bottom quintile is subject to tax rates of roughly 50% or more and a tenth roughly 70% or more. And you can look at the standard deviations in the last column, and you can see, yes, it's definitely positive in the top one in 5%, but you can see how much higher it is for or the bottom.

Of course, part of this is because if we, you might ask, well, suppose you cut out people with marginal tax with ridiculous marginal tax rates. Cuz you can have a marginal tax rate close to infinity or minus infinity. And we haven't done that calculation.

>> Speaker 25: You take some of those people with really high tax rates, effectively you're sitting on these money pumps, right? They get around a little less and get $1 less and get $3.

>> Alan Auerbach: Right.

>> Speaker 25: Look at the way you're forecasting their behavior. Does it make sense or do you see something like, obviously incorrect that they would never have done if they were thinking at all?

>> Alan Auerbach: Yeah, that, that is a good question. But, yes, you could say, well, could you deviate from the assumption of consumption smoothing or could you use a different labor supply profile? And we don't do that.

>> John Cochrane: The whole marginal question stops really making sense if there's a single cliff, like $1 and you lose.

>> Alan Auerbach: Okay, so we do look at different.

One of the things we look at in the paper is we say, okay, given you have all these cliffs and things like that, it's not a smooth function, then it matters whether you're talking about an additional thousand dollars of income or $10,000. We do that in the paper.And at least for the medians, it doesn't make that much difference. But that doesn't mean for the distribution, it could for the distributions.

>> Speaker 26: Can you speculate on what this would look like if you did it for 1950?

>> Alan Auerbach: Yes, well, I think this question came up before.

If you had really go back to 1964, before the Kennedy Johnson tax cuts, you had marginal tax rates above 90%, right? But I think at the bottom, I think it might be rather difficult because you really, I mean, the safety net just wasn't there. And so everything that's coming from the benefits side I think would be just much, much less significant.

>> Speaker 25: Could you say a bit more about what's driving the variation within the high group and within the low group.

>> Alan Auerbach: Well, within the high group, I think it's just we have people of different ages and so I honestly can't tell you we haven't focused that much on the high group.

But it could just be that people of different ages have different, different marginal tax rates. We certainly have different states, different family structure which would then lead to different consumption profiles over time. For the people at the bottom, it's just whether you're near the margin of losing benefits, and some of these programs are really big and you can lose everything.

You get thrown off Medicaid or something. It's just an enormous difference in your resources.

>> Speaker 27: I mean, the variation is really striking.

>> Alan Auerbach: Well, I already said this, current year's marginal tax rates aren't a good indicator of lifetime marginal tax rates. Here's a scatter of one against the other.

So the 45 degree line is shown there the most of the distribute. The large part of the distribution is to the right of the 45 degree line. And consistent with what I said before, the lifetime marginal tax rates tend to be higher than current year marginal tax rates. That's especially true for the yellow and light green dots which are the highest groups because those are the people where the double taxation takes off. But also you can also see much more, a lot of variation in the bottom in terms of whether it's higher or lower. Again because of the kind of examples I gave, yes?

>> John Cochrane: I'm surprised on the highest group.

>> Alan Auerbach: Yeah.

>> John Cochrane: Certainly understand how their lifetime is high because you got to invest. But in your highest income we got 40% federal, 13 and a half state, 10% sales taxes. And yet they have a big whole bunch of people that are under, most of them are under 50% in the current year.

>> Alan Auerbach: Well, remember, some of them live in Texas.

>> John Cochrane: That still would put a boundary of 40. You had a whole bunch of people under 40. So 40 is the federal marginal rate.

>> Alan Auerbach: Well, 37 yeah.

>> John Cochrane: Yeah, 37 to federal. But look at how many of the of the yellows are below 37.

And there's sales taxes even in Texas. So, I'm surprised that so many of the yellow ones are under the federal marginal tax rate, which I would have thought of would be the lowest possible of all these things.

>> Michael Boskin: The top bracket doesn't apply to the top quintile. It applies to a smaller slice of people.

>> Alan Auerbach: Yeah, it's basically the top, a few percent.

>> John Cochrane: Thank you, okay.

>> Speaker 28: What is the top quint income in your sample?

>> Alan Auerbach: I don't remember.

>> Speaker 28: Because I presume billionaires aren't in there. The SCF.

>> Alan Auerbach: It's the SCF over samples.

>> Michael Boskin: High income individuals, but not spectacularly high individuals.

>> Speaker 29: Think about something that Hoyt was saying before. We think about people engaging in behavior that based on this might look hard to explain. So there are certainly a lot of low income people out there where if somebody explained to me, you could just quit your job and just be on Medicaid and TANF and you're basically gonna have more resources than you do now.

I mean there's certainly people who are. Now what's the selection of people who are choosing to work despite maybe knowing that it's going to be people who think they have a lifetime income trajectory that's going to put them in a different category. That's why I'm thinking about these interactions, could be important. What's the selection of people who choose to work even though their marginal tax rate's probably around 100%. I agree.

>> Michael Boskin: So one thing that would be interesting, not in this paper, but in the future this may make a lot more sense for people who are older. You get people that are 20 that just can't expect these programs to be funded the way they are without.

>> Alan Auerbach: Well, or you can't expect them to be even thinking about it.

>> Michael Boskin: Yeah, yeah, I agree with that. But I'm just saying if you wanna get some idea what's likely to happen, you might.

>> Alan Auerbach: Yeah, we could. Sure, we could look at people in their 50s. Sure, we could look at pull out people in their 50s and say this is a key group in terms of thinking about how hard to work and how long to work. Sure.

>> Speaker 27: A lot of of them are retiring early.

>> Alan Auerbach: True, okay, just to reinforce benefit loss is an important component of marginal tax rates among the poor. And so this is the breakdown for the aggregate for the lowest resource quintile. Now in order to do a breakdown of the components, we have to look at means and not medians because you can't.

You can't add up medians to get a median. And so we did that just for the lowest resource quintile. And so for the lowest resource quintile, you give them $1,000 over their lifetime, they're gonna pay $300 more in taxes, but they're gonna lose $134 in transfer payments. And so these numbers are smoother because of course this is the mean, not an individual household.

So you don't get ridiculous numbers for the mean. But the point is, so in the aggregate for the wellness resource quintile, a big part of their marginal tax rate comes from their loss of transfer payments. We also did this question about labor force participation. So we look at people who aren't in the labor force but who are below retirement age, so below 62.

And we say, you know what, let's give them $30,000. Let's suppose they decide to work and earn $30,000 this year and every year until they retire, until retirement age, or 15,000, we treat that as half time. What does that do to their taxes? So it's a bigger experiment in terms of the amount of money and also it's for a particular group and for them you actually do get a U-shaped distribution.

So the light, the turquoise on the left are full time, on the right, part time marginal tax rates. And toward the bottom you do get a bigger marginal tax rates than you do say for the second, third or fourth quintile. And I guess that's because you just have a bigger jump in income and so you really are more likely to be losing benefit programs.

Okay, now this is the differences across states. So if you think about what's going on, if you look at people across states, and, of course, important to stress here that when we're defining people by what quintile they're in, it's on a national basis. So we're not looking at people in the bottom quintile among those in Mississippi as opposed to the bottom quintile among those in California.

Cuz obviously the latter group are going to have a lot more resources on average. So we're sorting people by where they are in the national distribution within their age group. And so here is for people who are in their 30s. So that's when we currently observe them. And they're also in the lowest national resource quintile.

So it's a roughly fairly homogeneous group, at least those two filters. Now we look at among those people in those group, what are their marginal tax rates? What's the median marginal tax rate within each state? And they are in states like Texas or Florida as opposed to states like California or New York.

And there are basically two things going on here. One is the light green states have lower taxes, lack of an income tax, for example. The other is they have less generous benefits. And so, if you lose Medicaid in, I don't know what the Medicaid rules are in Texas, but I bet they're not as generous as they are in California.

And so, if you lose benefits in California, you're losing more. And, of course, you're also subject to higher taxes if you're in California and-

>> Speaker 30: One could show here, I mean, take all the Republican and take all the Democratic-led states and seem to be a pattern.

>> Alan Auerbach: Yes, there is a pattern, although Washington State is lower, considerably lower.

They don't have an income tax. So it's not shocking to see the pattern of variation here. But the point is that these are, these are not, this is not rounding error. There are differences. Now these aren't the calculations you'd want to do if you were thinking about mobility across states because there, you wouldn't want to think about marginal tax rates.

You'd wanna think about what happens to the tax burden net of benefits if they move from one state to another. One could do that calculation as well, we haven't done that.

>> Michael Boskin: Also say that if you look at that-

>> Speaker 30: How about the marriage? What's the marriage tax for low income people?

>> Alan Auerbach: We have not calculated that.

>> Speaker 30: You're gonna have more income.

>> Alan Auerbach: We have not calculated that. We have not assumed changes in marital status. We've assumed-

>> Speaker 30: So can you add that?

>> Alan Auerbach: We certainly can. I'll just add it to the list.

>> Alan Auerbach: Yeah, and the marriage penalties and bonuses can be quite significant. And they matter a lot at the bottom because of benefit programs. And different rules for single head of household versus married couples.

>> Speaker 30: Yep. So if you're sorting people by their national standing and national income, these. So would you say our low tax states are also gonna have disproportionate share of those people.

>> Alan Auerbach: Yeah.

>> Speaker 30: Who are higher up the income scale and real income?

>> Alan Auerbach: Within that state, within the state, yes.

>> Speaker 29: People in Texas.

>> Alan Auerbach: That's right. That's right. That's right.

>> Speaker 30: And that, that can vary by 30%.

>> Alan Auerbach: Yeah, no, that's right, yeah, or Mississippi to take the poorest.

>> Speaker 30: Mississippi, Arkansas.

>> Alan Auerbach: Right fight for that title. Okay, so the final thing we do, and it requires a lot of assumptions which are laid out in the paper. And we just say, okay, suppose that you could magically keep the revenue the same from the households in any cell, but just set their lifetime marginal tax rates the same.

So it's not a constructive calculation in the sense that we're not saying here. Suppose you reform Social Security this way, Medicaid this way, and so forth. We haven't laid out a way of accomplishing this. But we're just saying, suppose, I mean, although you could, I mean you could have a very, very simple.

Linear income tax or something like that where you just have universal benefits for households as opposed to program specific benefits. So there are ways you could do this, but you wouldn't necessarily accomplish things that you want to accomplish with the various transfer programs that we have. But suppose you did that.

Well, given the non-linearity of deadweight loss, that's gonna give you an efficiency gain. How big an efficiency gain is depends on, on how, how big the elasticity. We assume different values for the friction elasticity of labor supply. And this is what we get. And this is relative to income.

And we did it two ways, income weighting and population weighting. Population weighting is probably better. Income weighting is sort of based on their observed income as opposed to what it would be if we do. We do a Taylor approximation, a second-order Taylor approximation, starting from the assumption of uniform.

Taxes and then saying how much of a deadweight loss is there if people face different marginal tax rates. So if you wait by their observed income you're doing it ex-post and so if people have high marginal tax rates you'll see them with lower income and so you'll be underweighting if you wait by income.

So that's why you get lower deadweight loss. Sorry?

>> Alan Auerbach: 3 the low mid high I think are 0.27, 0.4 and 0.53. And so if you look at the population weighting which is I think probably closer to the right way of doing it and this is as a share of income.

So if you look all and I think all is weighted by getting back to the issue of these people being the one Steve's district point of these people being the one that matters. I think that's weighted by income or weighted by resources. So that's what the all is but you can see that for the bottom resource group it's just huge compared to all the others because you saw the standard deviations.

It's enormous. And so it's not that you don't care about the other groups but it's really clear that these very big differences have the potential to lead to pretty significant labor supply distortions. And to the question of suppose people have a limited horizon and they're not thinking about what happens in retirement.

There's still even the current year marginal tax rates for lowest quintile have a lot of dispersion. So that issue would not go away even if we weren't doing this on a lifetime basis. But of course it gets bigger if we do.

>> Speaker 32: So that's all coming from the non-linearity of the labor response.

>> Alan Auerbach: Deadweight loss goes up with the square of the tax that's all it is. And we normally think about little perturbations but if you're giving somebody a marginal tax rate that's way up there then, it could cause them not to work at all. And one of the problems we have when we're looking at very high marginal tax rates is we're going to we might get corner solutions.

And so our second large Taylor approximation is not gonna be correct. So the exact numbers are just I think illustratively.

>> Speaker 14: Is the dead we lost the entire income tax.

>> Alan Auerbach: Of the entire marginal basically taking the marginal tax rates that we measure.

>> Speaker 14: Yeah they just seem incredibly low.

I mean, Marty has a paper where he says the deadweight loss of every dollar of revenue raised is $2. I mean, you could take issue with that. But I'm just-

>> Alan Auerbach: This is pretty high. I mean, if you look at population weighting, bottom resource, this is 18% of their income.

>> Speaker 14: Of their income, yeah.

>> Alan Auerbach: And of course, their taxes are much lower than that.

>> Speaker 14: I mean, I think it goes to the elasticity of taxable income, which was, which was your question. I mean, the high

>> Alan Auerbach: Which is another thing we didn't do. You were gonna say high income people have different channels. Low income people do too. There's the informal economy.

>> Speaker 14: Well, I think the elasticities are highest at the bottom. And at the top, I mean, so I guess I'm sort of surprised the highest. But these are quintiles, so maybe.

>> Alan Auerbach: Yeah, these are quintiles, so

>> Speaker 30: 18.2% income, and what's their cumulative lifetime average tax rate? It's got to be, so this is a number probably not far from Martin's.

>> Alan Auerbach: I think it's probably higher than that. Their lifetime tax. I think their lifetime average tax rates are negative.

>> Michael Boskin: Yeah

>> Alan Auerbach: Cuz remember, we're including all the benefits. So 18 over a negative number is higher than Martin.

>> Speaker 33: So if we think about how to do this with something like you have to deal with the transfer payment side, but on the tax payment positive tax side. If you had some, and we spoke, some income taxes, something like a Bradford X tax with victory cumulative lifetime averaging, would that head you in this direction, right?

>> Alan Auerbach: Yeah, sure. I mean, we were thinking about something like a linear income tax without the lifetime embellishment, but we could. Yeah, we could certainly do that.

>> Michael Boskin: Yeah I mean, of course, you wrote that before we had computers, but withholding and requirements to report income.

>> Alan Auerbach: Yeah like Carols.

>> Speaker 32: Carlos, within these quintiles, you would see a massive amount of redistribution of young people who are getting the programs that are targeted. I think it would be interesting to think about a ratio of how much money is being redistributed relative to that weight loss that we're getting from this distribution.

We should think there's a, there's some kind of a trade off between efficiency and redistribution. And this is actually a really nice place where you can actually quantify that.

>> Alan Auerbach: I mean, we would find that. I mean, and then one would want to look at who these people are. I agree.

>> Michael Boskin: We're also doing a lot of redistribution within quintiles as opposed to redistribution of cross.

>> Alan Auerbach: Yeah, no, I think that's what Juan Carlos was talking about. It's the people who knows what that causes them to be subject to different marginal tax rates, Nick.

>> Nick: These are just the contributions from the variation within each group.

>> Alan Auerbach: That's right, we're assuming the same revenue collected in each quintile.

>> Nick: I think a sufficient statistic to measure these is just the variance of the tax rates or so, right?

>> Alan Auerbach: It ends up being that weighted variance, yeah.

>> Nick: Okay, so that-.

>> Alan Auerbach: We have two pages proving that under the, and, and going through the assumptions that you need.

I mean, without certain assumptions it's not true.

>> Nick: Okay.

>> Alan Auerbach: It's in the paper. I mean that where you end up with is weighted by unobserved initial resources, the variance. Yes. Look, I think as many of the comments brought out, there are a lot of assumptions that go into this and this is a first step.

Which is saying here are the incentives that the tax and transfer system potentially present people with and how they respond to them is not something that we've considered at all. And nor have we gotten into the more into more sophisticated questions such as in terms of take up, to what extent does is take up and the timing of take up influenced by tax rates.

So the calculations can only get more interesting and more complicated. But to us, what it highlights, and I take the point about redistribution occurring if one were to have a tax reform. But what I take from this is that number one, the lack of coordination of different programs.

And this is what in the discussion decades ago about poverty traps, this was what it was all about. You get just hit by a barrage of these things and it's not as if there was sort of one unified transfer program that you were losing when your income went up.

Sure, that would give you a marginal tax rate, but you're not going to, you know, get hit by. You earn an extra dollar, you get slammed or you get, get multiple different programs hitting you at the same time. So the lack of coordination and thought and design of these programs, sort of designing them as if they were sort of there in isolation and nothing else was going on at the same time.

I think is something that really where reform really could be helpful. And beyond that, I appreciate the comments that you've given.

>> Michael Boskin: That was sort of the original idea of Friedman and Tobin separately having a negative income tax replacing all these programs. And of course what happened is we added them and got the Cozen area tax credit.

>> Alan Auerbach: More recently I think this was one of the recommendations in the Murlies. Coordinate things and make it a sort of comprehensive.

>> John Cochrane: Maybe the chaos is part of the design. That way people ignore marginal tax rates.

>> Alan Auerbach: Honestly I just give you one example. Think of the Medicare Part B premiums which go up with income don't go up continuously or smoothly, they go up in jumps.

So you earn an extra dollar, you pay $100 more a month or something like that. I can't think of any coherent reason for that.

>> Alan Auerbach: So it was originally to make numbers fit in some calculation of what? Well, yeah, I'm sure there was a revenue estimate, but you can do revenue estimates with smooth distributions.

>> Michael Boskin: I think not everybody is comfortable with that. And a lot of people that do this just prefer tables with sections. Department of.

>> Alan Auerbach: And they will prefer not to.

>> Speaker 34: Could I put in a plug for your NBR Macro annual paper? I'm a co-organizer of it and it uses the same kind of framework and looks at the effect of inflation on people by all these things and it's just fascinating.

So I recommend reading it, you did? It's the same team

>> Alan Auerbach: Overlapping group.

>> Speaker 34: Yeah, it's very, very interesting. It's astounding how much you get just out of the government programs.

>> John Taylor: Thank you.