The First Egalitarian Enrichment: Economic Growth and Inequality in America, 1870 to 1945

>> John Taylor: Welcome everybody for this lunchtime extravaganza. We're really looking forward to it. We're having Vince Geloso speak to us. You change your title?

>> Vincent Geloso: I can't seem to commit to the exact wording of different variations of the same thing.

>> John Taylor: The title I have is the first Egalitarian Enrichment: Economic Growth and Inequality in America from 1870 to 1945.

So you may have changed it around a little bit, but we're very welcome to have you here. We are anxious to hear what you have to say. We love economic history.

>> Vincent Geloso: I'm very happy to hear that.

>> John Taylor: Especially Gaban. Anyway, we're very anxious to hear what you have to say, so please go ahead.

>> Vincent Geloso: Okay, so first of all, thank you very much for inviting me. I'm sorry I couldn't be there in person. The negotiations of young children and family. So what essentially I'm pitching to you is the summary of a book that spawns from some of my research on the measurement of inequality in the past.

This started with redoing the very famous piketty and size estimates of top income shares prior to 1960. So I see David Splinter is there. David has done the post 1960. I did the pre1960 and in the process of this I realized that there's a series of things about inequality that we might be getting wrong with the United States.

And what I'm gonna pitch to you is a very different story than what most priors I think have are sharing on this. So just to give you the usual story, take the period 1870 to 1910 regardless of how you try to break it, which measure you want to use.

TFP wages, wages of unkilled workers, compensation of all types of workers. So here what you're seeing is I took every series that existed in economic history that economic historians use, and I just stuck it and showed. Yeah, this is a period of very decent economic growth. We would like to have rates like these.

Again these are very cool rates of economic growth but these rates of growth come with during the Gilded Age with some form of a trade off. Here is, I'm doing the same thing but this is with the inequality measures that are available. The first two columns are, first one is my estimates where I redid the pickety stuff.

The second one is pickets and size columns, threes, and fours. Same thing with the top tens with an additional series that comes from the work of Jacob Madsen and then some estimates afterwards of Gini coefficient 6 is from Madsen 7 is from the work of Gene Smiley. And the part I want you to notice most is that there is clearly some form of an increase for at least the top 1% of the distribution.

So that the distribution, it's not clear if it really increased for the top 10 income share. But at the very top, the very top gets to pull further away from the rest of society and you get thus the language that we frequently use to describe the Gilded Age where the very rich are pulling ahead from from everybody else.

And this is tied with one last point of details. It's that, well, the United States really in that period didn't have a welfare state. And it's kind of a puzzle to historians why does the US So it's not that the US only has a small welfare state. It's growing much slower than the welfare states of Europe and elsewhere.

The United States is a late comer to the game of social spending. Even if we include education, the US is very much a form of a late comer. If you exclude education from spending and you really look at transfers directly, the US is an even clearer a latecomer to that game.

And what I'm gonna argue to you is actually there is no puzzle to be had regarding this because the story I'm gonna tell as these points actually growth from 1870 was entirely egalitarian. So there is as much gain for people below the top 10%. So the bottom 90 had as much improvements in well-being as as people at the top 1%.

The growth in the United States that it enjoyed in that period was an egalitarian form of growth. Not only that, the United States is the only country to get a very strong type of that type of growth. So not only is it egalitarian, it is the largest proportional increase for people at the bottom.

So the magnitude of the absolute increase for people at the bottom is larger than anywhere else amongst the developed countries at that time. And we mean developed by today's standards. So you think every OECD countries, most of these other countries yet, sorry, generational mobility as well.

>> Speaker 3: You've talked about the level of inequality, but I was wondering about evidence on intergenerational mobility.

Is this also an area where there's a lot of mobility so in inequality went hand in hand with the chance of moving up the income distribution very rapidly?

>> Vincent Geloso: The answer is yes, but with some caveats. I don't wanna jump into them right now and I didn't put them in the slides, but they are in the book because I since I have 45 minutes.

I'm trying to get through as much as possible. But in the Q&A you can ask me more questions about the income mobility and I have answers for that. I just didn't put them in the slides. I apologize since I had to make some choice of economy. And the point that I make, and this connects with the welfare state is the US Is a latecomer to the welfare state trend because its market economy was doing what the welfare state was meant to address.

It was generating massive improvement at the bottom. And the United States, unlike European countries, just had a lesser political demand for welfare state because the improvements at the bottom were so substantial and so well divided. It's not that it had no demand for it. It's that it had a much smaller demand.

It's why the United States actually is a late comer to the welfare state gallary. And all the things I'm that I'm saying here, they're all There's a series of adjustments I'm going to do. And I don't think any of them are controversial either because we do them as economists today when we do inequality debates, but we don't or have not yet been able to do it to historical estimates and these corrections matter.

Or there are things that historians know and very much agree with but have never fully incorporated into the data. And so they're not controversial things that I'm doing to the data, but they are generating something that I do believe. I expect some pushback on the conclusions that I'm drawing, but I don't think what I'm doing in of itself methodologically is controversial.

And there are five adjustments I'm going to do. The first one is I'm gonna point out that the period 1870 to 1910 is a period of what I call cost of living egalitarianism. So this is essentially an argument about we need to adjust real incomes for the fact that not everyone has the same price deflator.

And it matters a lot and much more so than you would think. Number two is the issue of the missing poor. And this is the mortality that differs across income class and the missing poor from census documents. So that there is some people that we are not capturing that leads us to underestimate changes over time n levels.

But the level errors vary over time in ways that then affect our understanding of trends. There is then the value of longer lifespans, but also a reduction in the gap in lifespans so that the gap in how many years people got to live reduced from the bottom to the top.

And this kind of matters in our understanding of inequality. There is the issue of composition bias induced by immigration. And the last one, and this is something that is not frequently said today, we tend to think that tax evasion tends to be the rich man's business. But for the period I'm looking, tax evasion actually used to be the poor man's business.

Tax evasion was far more rampant below the top 90%. I'll show you evidence of this, and these are five things that historians. So most of them are things that we would agree as economists. Some of them are things that historians have noted but have not been incorporated into the adjustment.

And I'm gonna do each adjustment one by one. I'm gonna start with the usual estimates, tell you how they're built, and then I'm gonna build on each set of additions where you're gonna see how much change in inequality you get over the period. So the first one, the cost of living egalitarianism.

>> Michael Boskin: Vince, can I ask you a question? So 1873 and 1893, we had depressions. And there's a long period of declining agricultural prices at a time when food was a larger share of people's budgets than it is today. So how much is that that's what's driving this. I mean, we had a bimetallism debate between Brian and McKinley at the end of your period.

What's going on here? Is this.

>> Vincent Geloso: So, to answer that, I'll do a two parter. The first one is tell you how the estimates of inequality are built. Generally what we do, the first estimate we have is 1870 and it comes from Peter Linder and Jeffrey Williamson. And then the next one we have is 1910.

So we have two point estimate for the entire period. We don't have continuous estimates.

>> Michael Boskin: Okay.

>> Vincent Geloso: So that's the first thing, some of the other stuff that we have we can do continuous. And this is for example, the one I think you're seeing on the screen right now.

I'm hoping you're seeing. This is a price index that me and Peter Lindert created where we created an index for people at the bottom of the income distribution. So we found a basket that was representative of the bottom 40% and an index that was representative of the top 1%.

And what you're getting on this graph here is the division of the two price indices for, for the two groups. So poor over rich. So the way, the way you would read this graph is whatever the differences were in 1914, because we indexed everything to 1914 as the base year.

Say in 1870, the differences were 15, or here it looks like more like 17% larger than whatever the cost difference was by then. And the idea is that what you're getting here is that there's a strong egalitarian trend in the cost of living over time. Two factors that play into this.

>> Speaker 5: What's the key driver of that? I mean, is it that agriculture became more productive or why are we seeing what we are seeing, single cause?

>> Vincent Geloso: Part of this is that poor household get lower cost of staples. But the other part is in rich households baskets, there's a great amount of services that come that are being hired.

And so if you have rapid productivity growth in the rest of the economy, but not in services that rich people hire, the wages are being bidden up and that means that the price of services increase in conjunction with wages. So the most of the story here is being driven by the fact that the increase in wages that is driven by manufacturing productivity or other sectors where unskilled workers are hired.

Increases the price of the services in rich household, but not in the same amount as, as the productivity gain and services. And you can see that in census data when we look at occupation, the number of households that higher servants is reducing. So there is a reduction in the number of, in household servants in that period, which is consistent with that price development that I'm telling.

So Mr. Boskin, when you were asking the question, it was a long winded way of getting to this. I hope it answered what you were asking, took a longer route to get to there.

>> Michael Boskin: Yeah, but clearly a big part of it is the larger share of food, in particular in a period of declining farm prices.

>> Vincent Geloso: Yeah, so if you were to divide it by period, it's true to say until the 1890s say most of the leveling comes. Say half of it, I think from what when I did the exercise and I mentioned it in the book, I just forget the exact number.

But until the 1890s, half of the egalitarian trend is the reduction in the staples. The rest is services, and after 1890, it's largely services that are driving the egalitarian trend. Does that answer what you were concerned?

>> Michael Boskin: I think so, the big change is from the 1870s to the 1890s, right?

We're going from over 1.25 down to 1.05. There's another longer tail, but it's flatter, sure. That continues, so that shift makes sense.

>> Vincent Geloso: Okay, perfect, I'm happy I answered, I try my best. So this is what I just said. And now here is something. So I tend to think that the estimates I did for inequality where I corrected the stuff that Piketty did up to 1960, is superior to what Piketty did.

I think my methods are far better. And I won't get into this was a six year adventure that ended up in the Economic Journal. But I don't ask you to take my word for it. I'm gonna show you the results with both the Piketty series and mine. And what I'm going to do is, so on the right side here you're seeing based on my estimates of inequality, where both the Piketty series start with Lindert and Williamson, 1870.

1910 is what Piketty estimates. 1910 is what I estimate from my own series. And regardless of which way you wanna look at it, with the inequality from Piketty. And the way I'm re-expressing this is rather than taking the income shares, I am visualizing for you the ratio of the income of the top 1% over the income of the bottom 90%.

So this is the only way I could use the deflator. So I had to transform the way to express inequality in that particular way. And here what you wanna visualize is the black line is with no corrections for the cost of living egalitarianism and gray lines are with the correction.

And you can see that with the usual estimate, inequality nearly doubled, right? So the 85% increase that you're seeing right here, but with the corrections with the Piketty's estimate, it increased by 54%. Rounding up if with my series it's 52 if you use the price corrections. So we've reduced a large share of the increase in inequality just by accounting for the pro-poor price trend that existed over time.

So the real gains actually make us miss some of the relative differences, actually make us accentuate the relative differences. So adjusting actually captures them.

>> Speaker 6: Are you going to adjust some of these estimates by transfers and taxes, which have been done to some of the Piketty estimates to reduce the variance in income?

>> Vincent Geloso: So pre-1940 there is so very little transfers that it's not gonna change much, right? And since I'm ending most of the focus is pre1910, that issue isn't there. There is obviously, so I do it. Okay, so to get the income denominator, I do remove transfers from total income, but it's such a small share that if I put it on this graph, the two lines will overlap so much that I would just end up cluttering everything and not add much.

However, I say extended this graph to 1950, for example, you would start to see them behave a bit differently.

>> Speaker 6: Sorry, the surge in 1929 looks a lot like stock prices, but this is income inequality.

>> Vincent Geloso: Yes, this is income inequality.

>> Speaker 6: I mean, did wages double at the top and not at the bottom or is that capital gains on stocks?

>> Vincent Geloso: So one reason for this, and it talks to something I'll mention later regarding tax evasion. In that period, we have to remember that pre 1943 there is no withholding in the US which means that tax compliance is voluntary. And if there is huge variations in tax rates and huge variations in the personal exemptions, there is variations in reporting behavior.

 

So years of tax cuts with years of personal exemptions being heightened means more people comply, and more income enter our estimates of inequality. So one of the big issues is that we get false movements in the period pre 1943. And this is something that has been underappreciated, but was noted by guys like Gene smiley and the JH in the 1990s and early 2000 where you have to the movements are not correct from year to year.

So for example, in his correction in 1929, the increase is much smaller because there's coincident around these years tax cuts that cause people to start reporting their income. So some of the variations are not real.

>> John Cochrane: Under Harding and Coolidge, taxes went down from 70% marginal rate, 25% marginal rate.

And even they noticed that they were money was swimming into the treasury as a result of the tax.

>> Vincent Geloso: Yes, so apparently when you make tax reporting voluntary, some people will not report honestly. Sorry, it's my sense of humor. I think this is a funny joke, but whatever.

Okay, so this is for the ratio, sorry, I went too fast. This is the ratio of the top 1% over the bottom 90%. I did the same thing for the ratio of the top 10% over the bottom 90%. And already what's your, sorry, this is not the one yet, there's one after.

Here I also try to adjust for household size differences. So just in case richer household had more people, smaller poorer household tended to have fewer people. What if we adjust, the level falls down a little, but the increase is a bit larger than with only the corrections for cost of living.

So a standard correction says that maybe a bit more of an increase than otherwise is the case, but still, right? This is the usually depicted one. So that's the piketty in black, gray here is mine. This is mine with the cost of living adjustment. This is mine with the cost of living adjustment and a household size adjustment.

And the increase is from a lower base, but it's a larger proportional increase. If we do it for the top 10% income over the bottom 90, then the Piketty series that shows a mild increase with mine a mild decrease, it turns into a much larger decline in inequality between the top 10 and the bottom 90.

So already we're seeing some form of larger egalitarian trend. At the very least with that first correction between the top 10 and the bottom 90. And all the other corrections I'm going to do are gonna show further and further reductions in the increase between the top one and the bottom 90.

>> Michael Boskin: Can I just ask you, the more common, I mean, piketty and size obsessed on the top 1% radically overestimates. That's been kind of demonstrated. But it's been much more common in the last 20 years to look at the difference between the 90:10 ratio and the 50:10 ratio.

So you're getting inside the income distribution, what's going on rather than as a tail. I'm wondering if you do any of that.

>> Vincent Geloso: So this is the 9010. So the one I'm showing here is a 9010. But we can't do 50, 10.

>> Michael Boskin: 50, 10 and 90, 50 would be really interesting.

>> Vincent Geloso: I would love to be able to do this. I've tried to do something, and I was making assumptions that were way too heroic, but I felt like it was an action movie so much it was heroic that I decided not to do the 50,10.

>> Michael Boskin: It can't be done. It can't be done. That's just labor economists tend to focus.

>> Vincent Geloso: It's the only answer I can give. I'm sorry if it's not the greatest answer in the world, but I can do the 90, 10 which I'm showing you here and every other thing I'm about to do is gonna give us even more of a fall in inequality between the top 90 sorry, the top 10 and the bottom 90.

So very clearly, there are huge gangs here that are visible. I will point out that I'm not gonna show all the 90, 10 because already the case of egalitarian growth has been made in the book. I have the graphs for each of them. But what I'm even going to be able to show you is even the top one versus the bottom 90, there is no increase in inequality.

Which is the strong statement that I wanna push and convince you is the correct statement about the period.

>> Michael Boskin: You also must be using some national cost of living effort, do you have any? And of course, you're ranking people based on some in the distribution based on some national data, but-

>> Vincent Geloso: Yes.

>> Michael Boskin: There's huge variation I think even much more in this period geographically within the United States. Anybody is there.

>> Vincent Geloso: So I haven't put it in the book yet because I have a paper that's under revise and resubmit and there is a point of contention with one of the referees.

And I'm waiting for the paper to be finalized to start talking more about it but it's available online already as the working paper version. But I'm presenting stuff that I've published or is I'm willing to discuss with greater certainty. But yes, I'm actually also in the book planning and between now and when.

So I've sent this to University of Chicago Press once they come back to me, hopefully with a positive response. I plan to add an extra chapter for cost regional cost of living adjustments. So it's in the works for putting it in. I just wanna finish get the final convincing answer with referees and then put this in the book. So-

>> Speaker 8: Vincent, before you go on to missing poor, just a question about the inflation inequality. What has the literature concluded about the modern era and the impact of differential changes in cost of living for the rich and the poor on the standard income inequality measures?

>> Vincent Geloso: So there are some works that do this on the top of my head I have them in the references but I can't name the sources per se.

I know Magni Moza, I hope I'm not scrapping the name by saying it, has worked on this. But generally, especially in developing countries, there are for example one of the fate the papers I'm thinking of does it for India after the Singh reforms of the 1990s. In which there is a pro poor price effect of the liberalizations in that period suggesting that the effects are in, when you're taking nominal income you're probably missing some of the gains that the poor had in India. So there's more improvement from the bottom than we realize. So generally-.

>> Speaker 8: I guess also-

>> Vincent Geloso: Yeah, sorry.

>> Speaker 8: Conceptually I'm just trying to understand what are the pros and cons of looking at, by decile inflation adjusted income inequality measurement versus looking at consumption inequality.

>> Vincent Geloso: I mean, I would love to have consumption inequality but I'm doing the past. I don't have tons of this-

>> Speaker 8: The reason is this is a backdoor way to doing that cuz we don't have consumption data so.

>> Vincent Geloso: Exactly, that's a good way to think about it is obviously I'd love to have the modern stuff we can do. The thing is, I'm dealing with the limitations that surviving sources or surveys that were made or questions that were asked that are available and I don't have a ton.

So I have to find reliable proxies and make sure that they're good enough to withstand the discussion or at least do confidence ranges around them. That's the best answer I can give you on this. The next big adjustment, and this one is called the missing pour. And there's two types of missing pour.

The first one is when a census is taken, some people die within the census year. So we say it's the census of 1870, but that means that some people died within the year. But when the enumerator is there and visits, the member who died within the year is not counted as having been there, but from the household standpoint, he is present.

Now, if mortality is well distributed across all groups, it's not an issue of trends, right? We're just gonna have a constant error across all group and it cancels out. But if there is differential mortality across group, this means that there is differential number of missing poor. And if you think about the mortality at all age point and you we have some stuff that allows us to look at it like at least rich versus poor.

There is especially the best way to see it is blacks versus white in the US because we have life tables for them that will speak very well to at least a differential in mortality at each age point. Much more likely to die very young and at each other ages later, the probability is much higher.

So you have more missing poor than missing rich. So that means that you're going to affect your level estimate. So you're gonna underestimate how much people were poor. Because if you had an income of 100, say in the year, but you truly had five members, not four, then the actual income per head is actually lower than we realize than what is gonna be reported.

But if mortality gaps so the num. The ratio of missing poor versus missing rich changes over time, the level error will also change over time. And the way you wanna picture this is this kind of visualization which is the best one I want to picture is the teta term is how much of an error there is.

Because of a gap in the number of missing poor or the proportion of missing poor and the proportion of missing rich. Initially at the bottom there is a larger proportion of missing poor than there are missing rich. That proportion collapses over time such that in the group we realize at any point in time, yes, the correction means that people are much poor.

But the improvement is the trend is gonna be different as a result if the errors collapse over time. The other type of missing poor is from census under enumeration. So this is super well known. It's very well documented that the groups. So every census has a missing rate.

That missing rate is collapsing over time. But the missing rate is not random. The groups that are generally missed are black Americans, traveling populations, foreigners, foreign born people in cities and borders. So people who live as as resident inside another person's house, so they're boarding, essentially. All of these groups are generally poorer than average groups.

So we know the census tends to miss poor people more than they miss rich people. But that proportion falls over time. And these two things together really affect our measurement at the start point and at the end point in an uneven way, such that we need to make corrections for for inequality for the missing poor.

And by the way, this is a visualization for the census problem that I mentioned. Top one is you wanna imagine that I've ranked people as a uniform distribution into ten deciles, and these are the ones that we get from each box would be the average income in each group.

And each group will assume uniformly distributed. This is what I'm measuring. The one at the bottom is the true population distribution. So notice two things. First of all, we're assuming that we're missing a lot of poor for the sake of visualization. So we're missing a lot of income at the bottom.

So we are missing less income than we are missing people. But notice that there's another effect on inequality. The thresholds between the different group change. So there's a re ranking of who is the top 10 and who is not the top 10. So here in reality, notice that your top 10 in reality here is actually more like the true top 7%, say for example.

And at the bottom your 10%, your bottom 10 that you measure is in reality a mixture of the between the 10th and 30th centile of the distribution, if over time you have a reduction in under enumeration. So this would be jumping to the one here is you're capturing only people you used to miss, so you're adding more people than you are income.

But in a situation where every group's income doesn't change, right. The addition of people we were missing is going to give the impression that the bottom 10% is actually growing poorer. Whereas by assumption I said that it's not supposed to change, right? Because in that way of illustrating it, there is no change in groups.

But what you're gonna measure is a decline in inequality, sorry, an increase in inequality and a decline in the living standard of the bottom 10%. But that's only because you're finally measuring the poorest. So, what I go through is, well, here is evidence of things that would speak to the missing poor problem.

So you see that there's falling inequality in life expectancy or health outcomes. And that speaks to the missing poor from mortality. So dying within the year of the census. And this is the evolution of under enumeration rates. So this is only for whites, so for example, southern born whites, the south used to be underestimated.

And you see that the net under enumeration was really high in the census in which we have the first data point. So we're missing 9% of people, most of them concentrated at the bottom. And by the time we get to later censuses, the under enumeration rate has collapsed to smaller proportions. So we know-

>> Speaker 9: How does one know the under enumeration given that they aren't counted in the first place?

>> Vincent Geloso: Generally we use BERT records, that's one method. There's the survival method, you can use also from life tables you can do some form of forcing to test out how much you might be missing out.

There's multiple ways you can try this, and the ones I'm showing is actually the most conservatives. So if you try other methods, they tend to show even less, a more a stronger reduction in under enumeration. Just to give you an idea, for example, amongst black Southerners, the estimate is that we're missing maybe in 1870, 25% at most, at least maybe 15% of Southern blacks.

So, it's a huge share of the population we're missing. Most of them will be clustered at the lower end of the income distribution, causing us, when we estimate inequality and that the reduction in time collapses. So the under enumeration falls to a bit less than 10% by the 1910s.

So we know for sure that we get some form of bias in the estimation. And what I'm doing here is doing all the corrections to the census under enumeration of all the chapters in the book, it's the most laborious. And it's the most boring because I have to go through census procedures, explain everything that's done but the way you want to read this.

So I show the ratio top 10, bottom 90, top 1 bottom 90. Let's just look at the one I wanna emphasize most, the top one to bottom 90. This is the piketty numbers, right? That I showed you initially, this is my own, this is if I add to my own so above plus household size.

This is if I adjust for household size. Now I adjust for the cost of living inequality here I adjust the under enumeration only for I adjust the under enumeration for 1870. So I'm not doing the missing poor babies, okay? I'm not counting people who died. I'm just doing census enumeration and I'm assuming I'm giving a range.

So this is going back to, I forgot who asked me the question about and I answered we. I have to give you range because I can't get precise numbers. I'll assume either 5 or 10% under enumeration for 1870 and same thing for the 1910 census. And here you're getting that the increase of 33% now becomes between 21 and 23 when I adjust for the death within the years further to between 17.

So notice that you can start from piketty and you end up with between 20 and 17 or you can start with mine and you end up with between 20 and 17. So, whichever your start point is, I mean I tend to think mine is large is more relevant but more accurate.

Sorry, but in practice we have reduced massively in terms of proportion the increase just from getting things that we understand are problematic. I don't think they're controversial, I'm gonna keep repeating this. I don't think they're controversial things to that are nest that they're actually necessary to do, but they do affect the increase significantly.

>> Speaker 10: How are you adjusting for immigration? You were concerned about this disparity in death rates, and then within a year, and so immigration which will always be on the bottom end. Well, not always, but majority particularly that period be on the bottom end. And moreover, there are gigantic variations from each one of these decades in the number of people coming into the country and where they're coming from.

They need somehow adjusting for that.

>> Vincent Geloso: Yes, I have a section, I have a one of this second to last adjustment I'm gonna show you. I actually do corrections for the immigration stuff via the composition bias and I make sure to not double count things. So what I'm gonna show you that the immigration correction for composition bias I'm going to point out that in it, I actually make sure not to double count the what's coming in from there.

So I always make sure that I don't that all these adjustments I get their independent portion so that when I add them I don't end up double counting things. So I over correct by doing this, so the next one I have is lifetime earnings.

>> Speaker 11: Just before you move on, I have a clarification question.

So just to clarify, by assumption you're assuming that all of the undercounting occurs at the bottom decile of the, okay? And then, given that you said that the estimates of the undercounting are coming potentially from birth records, is it possible that you're double counting the effect of the missing pore and the undercounting?

If part of the reason that a person might be undercounted is because they've died.

>> Vincent Geloso: Okay, it's really good question. So first of all, the fact that I assume that they're all at the bottom normally it would matter if I was looking at the top 10. Like I'm sorry, the bottom 10, but since I'm taking the bottom 90, it doesn't really matter for me when I calculate the average for the average income of the bottom 90, right?

So this is not particularly problematic for me in terms, of an assumption. However, your second part regarding double counting. When I present them the way you want to. So I don't know if I did it in the tables, no so I should have added this to the table.

I also did them separately, so I only did one alone. So at the very least, cuz I know for this one it's one of those, I can't remove it. So I say very clearly, well, let's say you only can take one with certainty and not the other. I also show that regardless you end up with far less inequality increase.

So I tend to think you should take at least like you should take both. But yes, there is at the very least maybe some double counting problem, which is why I also present them separately in the book and say, well. I'll let you decide whichever one you want because in the end it's actually I'm still gonna be able to substantiate my case of egalitarian growth. Does that answer?

>> Speaker 11: Yeah, thank you.

>> Vincent Geloso: Okay, so the next correction is lifetime earnings. And for this, there is a lot of evidence for this from a lot of demographic work. So demographers have been immensely helpful in this. And here what I'm doing is I'm using lifetime cohort sorry, lifetime period tables, life tables by period.

And what you get is people who the mean lifespan for people who survived above age 10, those who get to live the night. So you distribute the lifespan. So how long people are can expect to live in 1870 at that period, a person below the 90th centile in the lifespan distribution conditional on reaching age 10 is can expect to live roughly 52 years.

And somebody above the 90th percentile in the lifespans okay? It's really important to the lifespan. This is not in the income. I am assuming that these correlate relatively well with income ranks. That doesn't mean they correlate perfectly, but there are substantially large differences in it. But what I want you to notice is that the gap collapses between them over time.

So a person who was there in 1870 and was age 10 in 1870 could expect to work 52 years. So by the time he reaches 1910, which is pretty much close to how much years he would have expected to live anyway since he survived to age 10, he actually gets to live an extra six years.

Which is something we should put a value on in some way. Either true the value of a statistical life or anything else. I'm gonna show you the way I found because I couldn't find a reasonable way to do VSLs. But what I did to simulate what is the values of these extra years is I started from the Historical Labor Statistics project.

I took the best survey that was available and this is the main labor bureau survey of 1892. I fitted an age-earning profile and what I said is we'll assume that a person at the bottom 90 is moving along the curve as he ages and the curve moves year to year as he ages in line with the average growth of the group.

So I'm saying that all the changes between rich and poor are going to be, it's an oversimplification, but it's going to allow me to get an idea of lifespan gains. And the way I do it is once I simulate this lifetime income path, I then do another thing as I'm creating a price theory example where the person who was there in 1870.

We would expect it to live until 1912, then he realizes he can live until 1918. But at 1870 there's an angel that drops from the sky and he says you can live to 58 and keep the income you're earning or I give you the annuity and I guarantee you'll die at 52.

What is the value of that annuity? And that annuity is going to be added to income. So this is the way I have of adjusting the average income to get an idea of what's the value of longer lifespans on net for the poor. And it's a pretty conservative approach cuz I'm assuming no income mobility.

So that goes back to an earlier question. I'm assuming no income mobility upward, no income mobility downwards. So all the changes are group specific changes and that's it. As soon as there's some income mobility upward. These proportions that I'm about to show you. So this increase in the annual income should be even larger than this because there's a chance that you could move up by far more than the average of your group from the bottom.

And once you do the correction, so I'm showing it here, top one, usual stuff, my stuff, my stuff with lifespan gains. So rather than being 51, it's closer to 41. Then I add in the cost of living adjustment, we're down to 17 with the under-enumeration we're down to between 7 and 8.5.

With the household size bumps it up a little, but adding in the missing poor from mortality, you get down to somewhere between 9 and 11. So we've cut down again the increase with this further adjustment. And if you look, by the way, the top 10 to bottom 90, we're in very large proportional declines within that metric of inequality.

The last two are a bit-

>> Speaker 3: Has the focus now shifted? I mean, before I thought was income inequality where compare person that's, getting high, very high salary or very high income in a particular year where somebody was young. Now you're looking at lifetime.

>> Vincent Geloso: No, sorry okay, sorry maybe I didn't explain clearly what I was doing.

So from the lifetime simulations, I got an idea of what was the value of the annuity within that data set. The proportion I got within that data set, I applied it to the income figures. So I really.

>> Speaker 3: I understand that, but it's saying you think you're only earning, I don't know, $10,000 a year.

You're really earning $20,000 a year because you have to work twice as long as you used to before. And is that really an improvement in the income? I mean, improvement in life is great, right? I mean, the value of life is enormous, but seems it's shifting the focus.

>> John Cochrane: To put it another way, you wanna be careful not to confuse income with consumer surplus. Consumer surplus is the right measure of welfare. And so people add all sorts of stuff to GDP in order to try to measure that. But really the value of living longer it's a free good that you're getting.

It's not market income.

>> Vincent Geloso: So, here's the way I wanna picture what I was doing. I'm trying to keep all else being equal in a certain way. So I wanna compare the poor bottom 90% to the poor bottom 90% of 1870. I wanna compare them in a way that I'm not missing out some other stuff.

So the way I'm picturing this is I'm imagining the guy of 1870. Sorry, the guy of 1910 with the same thing as a person with 1870. But I am indeed assuming that there's a compensating variation that is being given for making comparable. Yes, I get your critic.

>> John Cochrane: The critic might say we should count in, well, the poor live in the countryside, so they enjoy the bucolic splendor and the clean air, and that's got to be worth 100 bucks a year to them.

>> Vincent Geloso: Okay, you guys are shaking my privacy.

>> John Cochrane: You're making an important point, people lived longer, and that was good.

>> Vincent Geloso: Yeah, your point is good. What I'll probably end up doing with the book with this is I'll point out that this is worth something, but I'm not gonna work it in into the example.

At least say this is what it was worth at the very least, and maybe not do this table here or put too much too much emphasis on it, where. How much it would look like if you tried to account for this.

>> John Cochrane: You can rightly say lives were getting better for people at the bottom end of the distribution in some ways faster than people at the top end of the distribution.

I mean, how many fancy pianos can you own? And better health, living longer, better working conditions are all part of that. Just perfectly honest.

>> Vincent Geloso: Yeah, no, I'll point it in, but I think I'll retract this table from the book in that case and say we're missing some of the improvements at the bottom.

That one I'm okay with saying. But, yes, maybe correcting the inequality figures might be overdoing it a bit too much for this, but I'll. I'll move on to the next one cuz I don't wanna miss out on time, but you guys have shaken my priors on at least that part.

>> Speaker 6: All right, before you move on.

>> Vincent Geloso: Sorry.

>> Speaker 6: Is there a reason that you're not doing a corresponding correction for the top decile? Who also experienced-

>> Vincent Geloso: I didn't.

>> Vincent Geloso: I did it, what I was showing was the net effect, so I was allowing extra years of life for the top 10 decile.

So that what it generated was that the increase for the poor was larger than for the rich. So yes, I didn't only do the poor. I did rich and poor and not net, they were better off relative to the rich by that 76 to 7% more than we, than we appreciate.

So I brought that in. So yes, it's there. So I don't, yes, sorry. No, no, it's okay. There's a lot to summarize. I realize, and I hope I'm not overloading you with a series of corrections. The next point is the immigration issue. And some of the economic historians who work on immigration point out that one of the critics that was made at the time regarding immigration was the lesser quality of immigrants.

And although, the language is not super great by today's standard, what they meant by that is that the new immigrants from say 1870 to 1910 tended to have much lower levels of human capital than say settlers who used to come from Britain or Ireland. The ones who came, the French Canadians, the Italians, the Southern Italians, the Turks, very different group.

But in practice there is an implication from this. The change in the composition of who immigrates to the United States means we could be getting some form of rising inequality by who is coming in. And the reason why we know this is that, if we look at the inequality between groups inside the US.

So we have work by Robert Higgs, but a few other people as well, that there is no growing differences across groups. So if you take white Americans versus say Irish or Germans versus White Americans, the gaps are not extending. The gaps are actually pretty stable over time. But the total gap between all immigrants and native born actually is increasing.

So here for example, using the data in the Dillingham Commission, in 1870, the average immigrant earned 94% so this is in manufacturing earn 94% as much as a Native American. So someone who was born in the U.S, but by 1910 or 1909, sorry, it's 89%. So there is some form of composition bias that is at play.

So what I'm gonna do is remove the composition bias. So here is what I'm showing you. If given what I think John was asking, let's ignore the lifespan gains. Actually no, I can't because the way I've constructed it, but for sake of argument, just think about the ratio between the previous one and the one above.

We've further reduced the increase by roughly a quarter of close of what it is. And then, when you add on the underdeumeration, the household size and the missing poor, you're getting further and further, you're getting closer and closer to near zero. So stripping the composition bias is relevant.

And the last thing, and this is the one I think is the one I think I had to spend a lot more time through tax history stuff. And there's a lot of tax history in there. And I'm gonna try and try to point out as simply as possible what's in, is that before the 1940s, the IRS, there's not only no withholding, if you look at the labor force of the IRS.

It has less than three hours per auditor to assign to every tax return in the US and that's in the 1930s. After ramp up in the 1920s, they have less than one hour available per tax report in terms of investigative resources. And so the IRS actually develops a rule and they actually say it publicly, they say it in congressional testimony.

When they talk about it, people are aware of it. They simply do not investigate what they call the small peoples, or the small incomes. Every return below 5,000 before nine the night like the late 1930s are accepted as is, no questions asked. And what I'm gonna show you in a second is this.

We're talking more than 95% of tax returns are just accepted as is, no questions asked. And so we know that poor people lied when they reported and that's when they reported. So you would lie two ways. You'd have an extensive lie where you don't report if you're close to the threshold enough, or if you're close to a threshold but above the personal exemption, you would lie about your income, an intensive lie.

You would reduce your income reported to shift to a lower income tax, a lower income rate to know that you wouldn't be investigated. So there's that and obviously there is prohibition where there's a very substantially large underground economy where people will simply not report for a very obvious reason.

It's close to 3% of output. But the weird part is for political reason, the enforcement on the top 1% was incredibly good. So I'm gonna show you evidence in a second showing that for the top 1%, if we use different types of sources regarding reporting, I'm gonna show you what I mean by that is that actually the top 1% was reporting income very, very well.

So that means that in practice we know that the top 1% income is probably well reported, but the bottom 90% is probably underreported because of this. So this is why I say tax evasion was I say the poor man's business, but we think every man's business essentially. And here is to show you this is from the piketty estimates and what you have is from his method and mine as well.

Cuz we're using the same method, we're just doing different assumptions to the data. This is where the different thresholds for the different centile start. This is for the bottom, the top 10, top 5, top 1. This is the one the $5,000 no inspection line that the IRS sets, as you can see, qualifying to the top 10% is far below the rate at which the IRS simply doesn't investigate.

Even the top five is largely not investigated. And in fact in some years some people that would fall in the top 1% are also not investigated. So the vast majority of investigation is concentrated on people above the top 1%. And not necessarily perfectly, and on the right, what you're seeing is the share of tax returns in the US that fall below the IRS enforcement threshold.

And generally, you have more than 92% of tax returns that are simply not considered by the IRS, automatically not considered. That doesn't mean that the other ones are fully investigated. Even the other ones, there's partial enforcement. It's the one that very much the highest for which there's a high probability of being investigated.

>> John Cochrane: In your answer to my question about why there was this spurt of inequality coincident with the 1920s, you said the opposite, that rich people started complying more because the rate was lower, but now you're saying they were complying all along.

>> Vincent Geloso: No, so in the reason why it creates false variations is because the way I estimate the bottom 90%, right.

It's gonna create false movements from year to year as a residual of computing whatever the top 10 is. So in practice, the movements we have at the bottom, we're very uncertain of it. So there's a series of problems that we know with it and we know which ways the errors should go with changes in reporting.

So for example, when the tax rates are reduced, more people would say, yes, I know I'm close to the no inspection line, but there might be a chance I might be inspected eventually and so they'll come back. So some people start complying. So we know which way the error should go, but we have very little ability to determine how big the errors are.

So we know we should have in practice more muted movements is what I was trying to point out that the movements we get are not as strong from year to year and we know which way they should go. But all of these problems come in. But I'm gonna show you in a second actually that there is some way at the very least to strip out some of this problem.

And the first one is we have some partial ramp ups and enforcement pre-withholding. And here what I've added is the three big tax revenue acts where what happens to the number of tax returns below the enforcement threshold? So the enforcement threshold remains like a very well discussed thing, but there is discussion of ramping up enforcement.

So the entry into action of the 1936 Revenue Act, but starts acting in 1937, first big spike in number of people come below 5,000. There's another Revenue Act that is adopted in '38. You see a further increase and then the Revenue Act of 1940 a further increase. So you can see visually that the enforcement is actually eventually capturing, especially in the late 1930s, we start capturing the reporting of people closer to the bottom.

So the data get more accurate with the better enforcement, but it's more of a problem before. And this by the way, is the number of tax returns. The other way I have to show you this is there are some states that have their own income tax. And some of these states, especially Delaware, are very aggressive enforcers in the period where the federal government is not enforcing aggressively.

In fact, Delaware is very aggressive. They coordinate with the Census Bureau to see if what people say on the census matches or plausibly matches other things that are on the tax return. They also have a relatively large auditing force. And here what I'm showing to you is I took the IRS data for Delaware and I took the Delaware data for its own tax system.

And what you have is that there's 20 time, 21 times more people in the Delaware data that report less than $2,000. And then you see that as you move closer and closer to the threshold all it collapses to some degree. But then afterwards it's very much close to one, right?

And for the very riches it's actually pretty much the same number of returns. Like there's only is like there's like 27 guys in one, maybe 26 in the others. But it's nearly always very close to one. So we actually get really high quality before at the top. Sorry?

>> Speaker 12: What are the remedies for tax evasion for both the high and low income?

>> Vincent Geloso: So withholding is the answer for this, the invest.

>> John Cochrane: Wait, wait, wait.

>> Vincent Geloso: Sorry.

>> Speaker 12: Penalties, remedies for people who evade taxes. You have to read Becker and notice that you can have, if the penalty is several times the magnitude of the shortfall, then you get optimum.

>> Vincent Geloso: So the reason why I'm pointing out the withholding as the answer it actually fits with that Becker kind of punishments versus in the calculation is the things that's the easiest to hide in that period is actually wage income. So for example tips, really easy to lie about them.

Income in kind, very easy to lie about a form of income in kind. You can easily lie about how much wages you have since payrolls were not very much digitized. So when we looked at what type of income is missing more in the federal stuff, it's wages and salaries.

Wages and salaries is the one we're missing more. Whereas capital gains actually especially since some. You will often have contracts of sales with receipts that are far more easy to investigate. These ones, you wanna think about it that detection of certain types of evasions. Sorry, the detection of certain types of incomes being evaded is easier.

Withholding, it's done automatically. And this is why by the way, in the discussions over withholding, they're saying that the reason why they're saying they're doing withholding in the 40s. Is that they can capture so much more of something that people really lie a lot about. So, you can get captured more easily in this case.

>> Speaker 10: You're still completely omitting the magnitude of the penalty for underreporting. And that's key as Becker pointed out.

>> Vincent Geloso: So I forget the numbers, I do have in the book I do discuss what the fines were and what the consequences were for evasion. So I just think that most people, what they waited was not.

If you have a 1% chance of a $10,000 fines or a 0.00001% chance of being caught for something that gives you a $10,000 fine. As I'm emphasizing the fact that most people knew that lying on wages was so unlikely to be caught that you'd have to make a very high penalty to encourage actual enforcement.

So detection was so hard that it was really hard to catch it. But the fines in Delaware were actually not particularly crazy high. They were very much in line with those of the IRS. But they did more auditing of people's employers and then checked well, this is what he says as a payroll, you're not getting that you said something else.

So they actually did stuff like that. And there's something else that helps us. The IRS had to share information with states, but states were not mandated to cooperate with the IRS. So we actually get the inability for coordination for Delaware to share information constitutionally with the federal government.

But the federal government has to give information with the Delaware, for example, asked for it so they could do enforcement one way.

>> Speaker 10: According to Becker, if I may, the two things matter exactly the same amount, that is the probability of being caught and the fraction. And you're saying that you're not giving that equal weight that Becker would require. So-

>> Vincent Geloso: So maybe I'm misremembering the, the punishment model so I'd have to go back to it. I don't wanna commit to an answer because you seem more sure than I am about this. So my only answer is really, I'm gonna reiterate that I think detection was so very hard that from the IRS that people knew they could lie and whatever the fines were, it was hard to get away with it.

But I'll reread the model and incorporate this into the discussion. And since I'm presenting you work in progress and it's not like something that's final, I'm more than happy to incorporate that in. So I'm taking this as a helpful suggestion.

>> Kenneth Judd: What this discussion is missing is the fact that withholding happened because of World War II.

And in fact, Milton Friedman was a guy who was sent to Congress to argue for withholding. And the argument for withholding had nothing to do with any of these things. Argument for withholding was that if you didn't withhold it, and you just paid taxes the next April 15, which had been the law.

That people wouldn't save enough because the magnitude that we're talking about in terms of taxes was enormously greater after World War II, when they passed law in 1942. So this discussion with holding and all that penalties, I mean it was World War II that got us withholding for various very different reasons than being discussed here.

>> Vincent Geloso: So in the guy who also worked with Freeman that one I can say Friedman's objection regarding essentially what is essentially stealing some real income. Because people could essentially keep saving within the year to pay for their tax obligations after being reported and thereafter having reported and there was even tax deferred vehicles that you could use to pay your tax obligation for next year.

So there was a series of vehicles that were there, and I'm forgetting his name, the guy who proposed the one year layover to introduce the system. So the tax break, it wasn't a tax break, but the flip over for the next year. He actually said something different than Friedman.

He did say very explicitly and he was a deputy treasury secretary and he was saying, no, what we really want is all these revenues that we know we can't investigate. This is the sum of all small incomes that give so much because they know they can't. They have a hard time getting to it.

And withholding creates enforcement nearly automatic cuz you're enforcing it on via employers. And so you're deferring and you're shifting the cost of enforcement from the taxpayer to essentially the firms. And that's is essentially his argument. Friedman brings in other stuff, but from the vantage point of the treasury, the big argument is we can stop this evasion.

Because this is where most of the largest sum of money's not being reported ends up being lighted. It's on wages and salaries. So yeah, and so what I'm showing you here is, I took all the states for which we had income tax data. And these are states that did their own state income tax and were very aggressive about enforcing.

And what I'm showing you here is either net income based, so this is after reducing, after removing deductions. This is adjusted gross income. This is a result of how the data is reported in the IRS tables. But when you do that adjustment or regardless of which adjustment, what I'm showing here is I re-estimated for each state with the IRS data and with the state income taxes for these states.

What is the ratio of the estimated inequality and the from the IRS versus the state? If it's above the line here, that means that the IRS data says higher inequality than the state income tax data. And if it's below, it's the reverse. And what you get is generally that the IRS data says higher levels of income of higher levels of income inequality for than the than the state do.

And on average it's roughly 20% that is missing. And so what I essentially do is, I said okay, let's take 10% of this, sorry, only 5. We're gonna say that whatever this is as overestimation. There's at least an underestimate, there's an overestimation of inequality of say just only 5%.

So I'm taking a very downwardly biased estimate. I bring it in and now we get once we pile on all the adjustments pretty much around the zero range. So all these five adjustments, when you bring them in, they give a very different story of the period. And now, I'm gonna have to wrap up because I know we're close to the time.

This is the black line here, is the piketty and size increase for the bottom 90%. This is mine. This is mine with the cost of living with the household adjustment census, undermunation of the missing pore and everything else. So we're getting that we're missing a large share of the improvements at the bottom either with the piketty series or with mine if we don't do the corrections that we're getting.

And not only that, when you chart it with other countries, what you get is this picture. So the way you wanna read this picture is the income in 1870 as it start from one country and the arrow indicates where it ended in 1910. You can see that the living standards of all the adjustments that I could keep for all countries equally.

So I can't include the tax evasion effects on this one. The United States gets a really large increase in the living standard at the bottom 90% and very little increase in inequality. If you compare with Canada, with New Zealand, with Australia, with Germany, with the United Kingdom, with France, with Italy, you see that in most cases, actually in every case there is always a horizontal component to the movement.

But in nearly all cases, not all. So for example France and Sweden, most of the time you actually get also an increase in inequality in the period. So the United States has, this is why I say the US has egalitarian growth. And again I'm only looking top one versus bottom 90.

I'm not looking top 10 versus bottom 90. The US has the largest growth possible that is equally shared across people. So yes, it's a bit more egalitarian in say, France, but the increase is far smaller in terms of the living standards of people at the bottom. And this is why the argument I really wanna make to you is this period in America is very much the period of egalitarian growth.

The gains at the bottom are underappreciated to a very high degree, so much so that we miss the fact that this is a period where there is essentially no change in inequality. And I know that's a very controversial claim to make, but it also makes sense with the puzzle I mentioned at the beginning regarding the welfare state.

The big question that a lot of people ask is, I'm gonna skip something because I'm for the interest of time. One of the argument that people have made is that the US had high inequality. So why didn't have the welfare state as early as, say, Britain started developing it and France started developing stuff like.

And other systems of social transfers? Why is the US So late to this game? And there's been tons of effort through multiple works of trying to see what was very different about the United States that wasn't present in any of these countries. That might be another factor moving in the background that would explain why, despite rising inequality, the United States doesn't do a welfare state.

My point actually is say, all of these explanations could be true, but they could also not be relevant because the United States didn't have that increase in inequality. The massive increase in living standards at the bottom and the high level of distribution of it really dampened the demand for a welfare state.

And the only reason that there's a puzzle is actually that the welfare state emerges. So my work, I think, answers, I think says something new, but it also creates a new puzzle which I tend to work on next in my series of work. Is that it's actually the period after 1910 that is puzzling because after 1910 inequality continues to fall.

So it rises a bit in the 1920s, but with the Depression, there's a huge drop in inequality. But the real interest part, if I take whatever trend I have from so here it's an expression as a deviation from the pre 19 trend growth path. In each case you're getting that each measure that we can try to bring in of the income of the bottom 90%, there's a, there's a deviation and it's a deviation that seems to widen over time and it only resorbs after World War II.

Why? I don't have an answer for this, but I think it opens up a puzzle that I think is far more consistent because that puzzle does speak at the very least, if there was that huge deviation. Where people had expectations of really fast growth because of the 1870-1910 period.

Once the deviation started, people started demanding a welfare state and the welfare state eventually arrived in America. Because then people's expectations were disappointed and the demand for redistribution shoots up and the result is you finally get a welfare state. Why is there that slowdown? I cannot answer yet and in the book I trying to present this as this is a new fact about American economic history that we do not fully appreciate.

But which really, really really matters and is far more consistent with other development. So, I hope I didn't bore the crap out of you and that you found this interesting or at least worth critiquing or worth engaging with. So, it feels like I'm ending a bit flat in the way I'm saying it, but that's it, folks.