The massive feedback continues on my blog post yesterday, “Obama’s inequality argument just utterly collapsed.”
To recap: The core thesis of Obamanomics, as I see it, is that the past 30 years of pro-growth, pro-market policies – lower marginal tax rates, less regulation — in America have failed. Of course, that perspective fits nicely with Democratic Party ideology. But liberals say the economic literature support their view, particularly the work of economists Thomas Piketty and Emmanuel Saez. Their analysis of IRS tax data suggests income inequality has skyrocketed in recent decades while middle-class incomes have gone nowhere. Perhaps you’ve seen some version of this chart:
And from that chart, we get Obama’s call to raise tax rates on the rich, create a new healthcare entitlement, and institute the Buffett Rule.
My post highlights the work of economist Richard Burkhauser and his team of researchers at Cornell University. Their work suggest the middle class has done just fine — medium income up 37% vs. 3%, according to Piketty and Saez — thank you very much, and that inequality has increased only modestly, with the Gini index increasing from 0.33 in 1979 to 0.372 in 1989, then falling to 0.364 in 2000 and again to 0.362 in 2007. So the rise was pretty much an ’80s phenomenon.
What’s more, all income levels did pretty well:
So which group of economists is correct, especially since one is using IRS data and the other Census Bureau data? As it turns out, Burkhauser just gave an hour-long, must-listen interview to my pal Russ Roberts at EconTalk. It addresses many of the questions my readers have about the study. So I have excerpted some of the podcast, though I urge anyone really interested in the topic to have a listen. I am going to do this in pretend Q&A format, substituting my question for Russ’s, just to be more concise.
PETHOKOUKIS: So who’s right, you or them?
BURKHAUSER: The puzzle to me is how can you get such apparently wildly different visions with the same data? And to bring us up to the paper we talk about today, it comes from a quest on our part to figure out how two extraordinarily smart people–Piketty and Saez–could get such wildly different perspectives on what’s happening in the United States using their data than I could using these traditional measures that I’ve been looking at for the last 30 years.
So, I was getting some results that were much different from the vision you get looking at Piketty and Saez. And I knew these guys were really smart. So, either there was something wrong with the data they were using versus the data that we were using; or they were using different assumptions in the way they were measuring this data than we were; or one or the other of us had made mathematical errors. And it took us 4 years to actually figure this out. But the answer is: It wasn’t the data. It wasn’t that we were making mathematical errors. It was that we were fundamentally asking different questions, but using almost the same words to describe our findings.
They are asking a perfectly valid question: What’s been happening to the median market income of tax units? That is, when you add up the returns to land, labor, and capital, what’s been happening to median income? They don’t adjust for tax-unit size, and as you said, they don’t adjust for government transfers. They are just looking at the returns to land, labor, and capital. And when we use their measure in the CPS data from 1979-2007, we get a pretty discouraging result. The median income of these tax units has only increased by 3.2% over the entire 30 years.
PETHOKOUKIS: 3.2% over 30 years is bupkis.
BURKHAUSER: That’s stagnation. And that’s what they call it: the middle class has stagnated over the last 30 years. And that statement is correct. But what does that statement mean, and how do you put it in the context of how the typical American has done in the last 30 years? Is it true that we haven’t increased at all in the last 30 years in our ability to consume things?
Okay, so let’s–and this is the fun part of the puzzle. Let’s no longer take the tax unit as the unit of analysis. Now let’s look at the household. And you go from a gain of 3.2% over that period to a gain of 15.2%. And that happens because the number of tax units per household has been rising over that 30-year period. There are a lot more people living together and sharing everything except a marriage certificate.
PETHOKOUKIS: Just changing that measurement alone increases income by five times. Wow! Why so much and why is that a better measure that using “tax units” as P&S do?
BURKHAUSER: Strictly going from a tax unit to a household unit changes that because in these tax units, you have people–their assumptions about tax units are: they care very much about what the top 1% or 5% of tax units are. But they also want to include not only the people who pay taxes, but the people who don’t file taxes. So they have to make some assumptions about who is in these tax units. And their assumptions are that anyone over the age of 20 who is not married is an individual tax unit. So, they get a lot more individuals, low numbers than you would when you recognize that most of these people who are above the age of 20 but aren’t married are connected in some way with other individuals in households. So you get fewer.
Now, I want to be clear on this. They can’t do this. That’s why they didn’t. The advantage of what we can do is we can actually get a median because we know what the distribution is below these top units. They don’t know that. They make assumptions about that. Piketty and Saez can’t tell you what the median tax unit’s pre-tax, pre-transfer income is: they just don’t have that data. We do because we have the CPS and we can put it into smaller tax units. And the important thing about the recent paper that we’ve done is we can virtually get their results using the CPS. That’s what is really cool about that paper.
PETHOKOUKIS: People have argued that the Census Bureau’s Current Population Survey is an inferior way of looking at these issues.
BURKHAUSER: Here is the really important intellectual point here: No one has been able to do this before. So, people have argued that you can’t actually use the CPS to look at the tails of the distribution, and therefore, Piketty and Saez, even though their data has problems, those problems are less than the problems in the CPS. The thing that took us 4 years to do was to show people: No, that’s not true. We can actually get very close to the levels and trends that they get in the IRS data using the CPS data. That’s what took a long time. Once that is established. Once you grant me that the CPS can get approximately what the IRS is getting, then you can write a National Tax Journal article that can actually show how sensitive the assumptions that they are forced to take, because of the limits of their data, matter if you go beyond what it is that they are asking. That’s what this paper is I think interesting about, and is so profound–this one table, as we go through it, is, I think, so shocking that you can get such dramatically different numbers for a simple concept, like median income. So, we’ve gone from tax units, where there’s an increase between 1979 and 2007 of 3.2%, simply from going from the tax unit to the household unit, we get 15.2%.
PETHOKOUKIS: And then you start adding in income sources that P&S left out, yes?
BURKHAUSER: So now, when you go to household, size-adjusted, pre-tax, post-transfer income–so now we are going to the usual measure that I’ve been using, the Census Bureau’s measure: pre-tax, post-transfer income and the unit is the household; and we are adjusting by the number of people in the household to the 0.5. You go from 15.2% to 23.6%. So, now, instead of 3.2% for a tax unit, pre-tax, pre-transfer, you now have 23.6% for a household sharing unit, adjusted by the number of people in it; and we are including government transfers.
Not only does the government transfer money from high-income to low-income people through government transfers like Social Security benefits, disability benefits, Temporary Assistance for Needy Families (TANF) benefits, and those sorts of things, but it also does it through a progressive income tax. So, when we then use the National Bureau of Economic Research’s tax simulator, with the CPS data, we can simulate the amount of income taxes, state taxes, and payroll taxes that people pay. So, then we go from 23.6%, which is the median income without taxes, to 29.3% growth.
PETHOKOUKIS: Anything else?
BURKHAUSER: Yes, and then the last change is to talk about this growing share of earnings that comes in non-wage compensation, and also the fact that Medicare and Medicaid have been growing also. So, as an example of how important it is to think about in-kind transfers as well as in-cash transfers, and to think about the value of health insurance for employees as well as their wages, we are able to estimate the employer share of employer-provided health insurance to workers and the insurance value of Medicare and Medicaid to lower-income people who are getting those benefits. And when you do that, it goes from 29.3% to 36.7%. So, we are really talking about a difference from 3.2% in the way Piketty and Saez are thinking about things in tax units, to 36.7% if we simply include the value of health insurance as well as do the other things that we’ve talked about.
PETHOKOUKIS: Again, so who’s right, you or them?
BURKHAUSER: Well, I think the right answer depends on what the right question is. So, if you are asking what’s been happening to market income, I think there’s no question that wage rates have become more unequal. But once you adjust for health insurance and other things, it’s less unequal. And that in terms of real compensation, it’s clearly rising. So, I think that even there, returns to work have been rising. The notion that we as a society are not doing as well as we were 30 years ago, I think by virtually any reasonable measure, is just false. And I think that’s the main statement. The issue of distribution is a little harder. It’s a little harder to argue. We have become somewhat more unequal. But even there, what we find, in the CPS data, is between 1992 and 1993, there was a change in the ability of the Census to capture exotic incomes of the top 1% of the income distribution. So, if people look at the CPS data and don’t recognize that in 1993 we suddenly were able to better capture income, they will get the false notion that income inequality increased between 1992 and 1993. It didn’t. It just was in our ability to capture that income. And we actually adjust for that here. But what it says is that when you get people telling you things that just don’t seem to be consistent with reality, you really do need to look very carefully at the assumptions they are making.