Alan Reynolds revises some standard views about income distribution in his new book, discussed in today's Wall Street Journal, Op-Ed page. The opening question: is it true that America's top one percent takes in sixteen percent of national income? Maybe not. I don't usually offer such long excerpts, but if Reynolds's skepticism is warranted, it is very, very important...
…The
architects of these estimates, Thomas Piketty of École Normale
Supérieure in Paris and Emmanuel Saez of the University of California
at Berkeley, did not refer to shares of total income but to shares of
income reported on individual income tax returns-a very different
thing. They estimate that the top 1% (1.3 million) of taxpayers
accounted for 16.1% of reported income in 2004. But they explicitly
exclude Social Security and other transfer payments, which make up a
large and growing share of total income: 14.7% of personal income in
2004, up from 9.3% in 1980. Besides, not everyone files a tax return,
not all income is taxable (e.g., municipal bonds), and not every
taxpayer tells the complete truth about his or her income.
For
such reasons, personal income in 2004 was $3.3 trillion, or 34.4%,
larger than the amount included in the denominator of the Piketty-Saez
ratio of top incomes to total incomes. Because that gap has widened
from 30.5% in 1988, the increasingly gigantic understatement of total
income contributes to an illusory increase in the top 1%’s exaggerated
share.The same problems affect Piketty-Saez estimates of share of
the top 5%, which contradict those from the Census Bureau (which also
exclude transfer payments). Piketty and Saez figure the top 5%’s share
rose to 31% in 2004 from 27% in 1993. Census Bureau estimates, by
contrast, show the top 5%’s share of family income fluctuating
insignificantly from 20% to 21% since 1993. The top 5%’s share has been
virtually flat since 1988…
Unlike
the Census Bureau, Messrs. Piketty and Saez measure income per tax unit
rather than per family or household. They maintain that income per tax
unit is 28% smaller than income per household, on average. But because
there are many more two-earner couples sharing a joint tax return among
high-income households, estimating income per tax return exaggerates
inequality per worker.…the
amount of income Messrs. Piketty and Saez attribute to the top 1%
accounted for 10.6% of personal income in 2004. That 10.6% figure looks
much higher than it was in 1980. Yet most of that increase was, as they
explained, "concentrated in two years, 1987 and 1988, just after the
Tax Reform Act of 1986." As Mr. Saez added, "It seems clear that the
sharp, and unprecedented, increase in incomes from 1986 to 1988 is
related to the large decrease in marginal tax rates that happened
exactly during those years."
That 1986-88 surge of reported high
income was no surprise to economists who study taxes. All leading
studies of "taxable income elasticity," including two by Mr. Saez,
agree that the amount of income reported by high-income taxpayers is
extremely sensitive to the marginal tax rate. When the top tax rate
goes way down, the amount reported on tax returns goes way up. Those
capable of earning high incomes had more incentive to do so when the
top U.S. tax rate dropped to 28% in 1988 from 50% in 1986. They also
had less incentive to maximize tax deductions and perks, and more
incentive to arrange to be paid in forms taxed as salary rather than as
capital gains or corporate profits.The top line in the graph shows
how much of the top 1%’s income came from business profits. In 1981,
only 7.8% of the income attributed to the top 1% came from business,
because, as Mr. Saez explained, "the standard C-corporation form was
more advantageous for high-income individual owners because the top
individual tax rate was much higher than the corporate tax rate and
taxes on capital gains were relatively low." More businesses began to
file under the individual tax when individual tax rates came down in
1983. This trend became a stampede in 1987-1988 when the business share
of top percentile income suddenly increased by 10 percentage points.
The business share increased again in recent years, accounting for
28.4% of the top 1%’s income in 2004.As was well-documented years
ago by economists Roger Gordon and Joel Slemrod, a great deal of the
apparent increase in reported high incomes has been due to "tax
shifting." That is, lower individual tax rates induced thousands of
businesses to shift from filing under the corporate tax system to
filing under the individual tax system, often as limited liability
companies or Subchapter S corporations.
IRS economist Kelly
Luttrell explained that, "The long-term growth of S-corporation returns
was encouraged by four legislative acts: the Tax Reform Act of 1986,
the Revenue Reconciliation Act of 1990, the Revenue Reconciliation Act
of 1993, and the Small Business Protection Act of 1996. Filings of
S-corporation returns have increased at an annual rate of nearly 9.0%
since the enactment of the Tax Reform Act of 1986."
Switching income
from corporate tax returns to individual returns did not make the rich
any richer. Yet it caused a growing share of business owners’ income to
be newly recorded as "individual income" in the Piketty-Saez and
Congressional Budget Office studies that rely on a sample individual
income tax returns. Aside from business income, the top 1%’s share of
personal income from 2002 to 2004 was just 7.2%-the same as it was in
1988.
In short, income shifting has exaggerated the growth of top
incomes, while excluding a third of personal income (including transfer
payments) has exaggerated the top groups’ income share. [emphasis added]
There are
other serious problems with comparing income reported on tax returns
before and after the 1986 Tax Reform. When the tax rate on top salaries
came down after 1988, for example, corporate executives switched from
accepting stock or incentive stock options taxed as capital gains
(which are excluded from the basic Piketty-Saez estimates) to
nonqualified stock options reported as W-2 salary income (which are
included in the Piketty-Saez estimates). This largely explains why the
top 1%’s share rises with the stock boom of 1997-2000 then falls with
the stock market in 2001-2003.In recent years, an increasingly
huge share of the investment income of middle-income savers is accruing
inside 401(k), IRA and 529 college-savings plans and is therefore
invisible in tax return data. In the 1970s, by contrast, such
investment income was usually taxable, so it appears in the
Piketty-Saez estimates for those years. Comparing tax returns between
the 1970s and recent years greatly understates the actual gain in
middle incomes, and thereby contributes to the exaggeration of top
income shares.
In a forthcoming Cato Institute paper I survey
a wide range of official and academic statistics, finding no clear
trend toward increased inequality after 1988 in the distribution of
disposable income, consumption, wages or wealth. The incessantly
repeated claim that income inequality has widened dramatically over the
past 20 years is founded entirely on these seriously flawed and greatly
misunderstood estimates of the top 1%’s alleged share of
something-or-other.
Opinions? I am embarrassed to admit I have yet to read Pikaetty and Saez. If you would like an alternative perspective from that offered by Reynolds, here is Paul Krugman.
Addendum: Here is Greg Mankiw on same, with related links.
In addition, inflation hasn’t affected the income distribution evenly. While this doen’t mean there isn’t a problem with increasing inequallity, it reduces the impact. In the five years leading up to 2005, the BLS shows that while incomes had remained flat for the bottom half of earners, expenditures had gone down. For most goods, prices didn’t increase or alternatives were easily substituted. Until the more recent gas price spike, the impact on living hasn’t been too severe. However, flat earnings does mean that mobility decreased. It would mean that more savings are needed for investments to move into the higher income groups.
The main claim seems to be that the TRA86 threw everything off. However, even if you throw out ’85-’87, there was a huge increase in inequality both before and after that period. It’s true that 1986 and 1987 were odd years, but the increase in inequality has been consistent over time.
I think it’s alsoimportant to recognize that there are other data sets that show this same sort of inequality. One is Saez and Kopczuk which uses data on estate tax returns to get a good estimate of the wealth distribution in the top 2%. Another is the CPS which shows us how inequality has been increasing at levels below the top 1%. Tax data certainly has some problems, but the fact that it’s not perfect, such as in terms of measuring household income rather than individual income, does not mean that inequality hasn’t increased.
It is difficult to tell whether his points are legitimate or whether this is just an example of what Leland Yeager calls “Alan Reynolds style argumentation” (i.e. spitting out a bunch of facts that aren’t related to your argument and then restating your argument).
Most of what’s excerpted seems sound, however Reynolds does overstate his case. One example is his point of examining household income vs per capita. Even if household income is stable, additional expenses are incurred inorder to be able to manage a dual income lifestyle. I agree that inequallity is probably exaggerated, but that might not be true in the future. I think that in recent years upward mobility at the lower end likely decreased.
Anytime economists staft talking about the tax system I get nervous.
Economics is a science? Ya boy.
And those little people have too much already, we need to send more to
Wall Street.
Perceived inequality versus measured inequality being conflated in a demagogic fashion
The trouble with the US inequality debate is that commentators like Krugman conflate abstract statistical economic measures of inequality with people’s subjective feelings about inequality. I see no reason why the abstract and the personal should be at-all tightly correlated – indeed, they quite obviously are not correlated since it is well established that the greatest psychological resentment is directed towards similarly-situated people (eg. the guy in the next cubicle who does not much more but gets paid twice as much).
If it is the supposed socially-corrosive subjective effects of inequality that worries liberal commentators, then logically in response policies should focus on these subjectivities. In other words, (logically) the solution to perceived inequality is to make people ‘feel’ more equal, presumably by some kind of propaganda.
However, commentators such as Krugman are doing the opposite – they are trying to stir-up popular resentment about inequality (especially class and race resentments), trying to make people feel subjectively worse by highlighting statistics which portray inequality as bad and getting worse.
The implicit argument is that if you don’t already feel bad about this – then you should.
I don’t find this kind of thing admirable – indeed it strikes me as pretty close to demagogery.
And here’s what happens when you adjust for family size -
http://www.taxpolicycenter.org/taxmodel/tmdb/Content/PDF/T06-0306.pdf
18% goes to 20.6% for the top 1%. Not exactly what was implied by the article you excerpted.
“But they explicitly exclude Social Security and other transfer payments, which make up a large and growing share of total income: 14.7% of personal income in 2004, up from 9.3% in 1980. Besides, not everyone files a tax return, not all income is taxable (e.g., municipal bonds), and not every taxpayer tells the complete truth about his or her income.
For such reasons, personal income in 2004 was $3.3 trillion, or 34.4%, larger than the amount included in the denominator of the Piketty-Saez ratio of top incomes to total incomes.”
The number used for total incomes is 34% larger, and of which we know that 14%/34% is due to transfer payments. So he explains about 40% of the difference between the two numbers. What makes up the other 60%? There aren’t many options. Why doesn’t he tell us? You would think he might want to break out what those are. That he doesn’t makes this CFA immediately suspicious – he doesn’t seem to have a problem with talking about numbers through the rest of the article. Its his very first statistics, his most likely strongest argument, and he doesn’t explain about 1/2 of the effect?
In any case, his major argument seems to be that inequality isn’t increasing, but rather that it has been bad all along.
My textbook, Income and Wealth, is carefully documented and everything I state as a fact is linked to an URL so you can check for yourself. This is a topic about which people form very stong opinions on the basis of very weak facts. After 35 years of debating this issue, I grew tired of all the deception. Reality is not a matter of opinion. The income tax data I refer to in the Journal is not suitable for the purpose to which it is put, to put it as kindly as I possibly can.
“In any case, his major argument seems to be that inequality isn’t increasing, but rather that it has been bad all along.”
Americans are able to tolerate quite a bit of static inequality. That is why the argument is always formed about increasing inequality stifling the American dream by doing things like reducing social movement. As a normative question I’m not troubled at all by extreme wealth on one tail as long as there is what see as a non-awful living at the other tail and a well functioning economy in the middle.
Alan,
It’s been a while since we last spoke. You’re right that the IRS data is far from ideal. However, it’s the best we’ve got. There is simply no other data set that gives us as much data on the very top end of the income distribution. the fact that there are some imperfections, as with any other data set, doesn’t make it useless.
Well, that’s why Piketty and Saez used the data for a number of different calculations. While top income shares were important, top wage shares also have risen a lot too. I think alan’s criticisms of the tax data’s validity for total income are useful. However, I’m much less convinced by his arguments about how well the IRS measures wages. The arguments about family size and hours can be shown in other data sets to cause pretty small problems. ECI data also confirms many of the results on the wage distribution and can extend them to total compensation including benefits like health insurance.
So, Lance, I think you’re right that we can’t use the tax data for just anything we want, but Alan’s comment above made it seem as if the data have almost no use at all. I respectfully disagree.
Paul Botts has it exactly right. I wasn’t saying much about the level of inequality, only that data based on a sample of individual tax returns (including CBO or Tax Policy Center data) cannot be properly used to compare changes in income distribution between greatly different tax regimes. Behavioral elasticities aside, and switching between corporate, capital gains and W2 income, even the accounting changed. Tax-free interest did not have to be reported until 1987, yet CBO includes it in time series going back to 1979. Social Security was not reported until 1984, and only a fraction is reported now because a couple on the average benefit earns too little to meet the requirement to file.
Income shifting was not limited to 1986-88, as the top graph in my WSJ showed. And that is just one form of inccome shifting. I mentioned two others, explained at length in my book.
Ian Dew-Becker replies that that such garbled tax data is all we have to make the points he hopes to make, because Census just covers the top 5%. But the fact that Census figures for top 5% are much lower and flatter than the comparable Piketty-Saez estimates (although Census also leaves out transfer payments) means we have to pick between those time series. This may be a difficult choice for political reasons, but it should not be a difficult choice for apolitical economists who care about being accurate.
I did not intend to accuse anyone, least of all Ian, of bad faith — just bad data.
At least for years after 1989, we don’t have to choose between Census and IRS data because we have the Fed’s Survey of Consumer Finances.
https://www.federalreserve.gov/pubs/oss/oss2/2004/bulletin.tables.pub.xls
This shows real income among the top decile rising by 18.3% from 1989 to 2004, while mean income rose 22.7% in the bottom quintile and 20.8% in the second lowest fifth. CBO came to similar conclusions using Census data this Spring, but they began with the recession year of 1991 which bias the results.
The problem with Census data is not top-coding of the public use dataset but internal limits on specific categories of income (not total income) that were greatly increased in 1993. This makes data before 1993 “not comparable” with data after 1993.
I have a set of five graphs on these and other topics from an 11-4-07 presentation at UNC Chapel Hill which I’d be glad to send to anyone. Just email me areynolds@cato.org
Comments on this entry are closed.