Alan Reynolds continues the discussion he has been having with Paul Krugman here and elsewhere over inequality, this time in an op-ed in the Washington Times. If you have been following along, Reynolds wrote an op-ed in the Wall Street Journal that challenged the idea accepted by almost every reputable analyst that inequality has been increasing in recent years.
That led to a debate over whether Reynolds had repeated errors he has made in the past in collecting and interpreting data and whether his position holds up to closer scrutiny. Krugman argues, and I think he is correct, that it does not. But you should see for yourself.
Here are the posts Reynolds is referring to in today's op-ed. The last two posts listed, which are posts of email from Paul Krugman on this issue, are probably the best summaries (also see the follow-up in the comments to the two posts by Reynolds, Krugman, and others):
- The Top 1% . . . of What? (Reynolds WSJ op-ed)
- Reynolds' Rap on Inequality
- Follow-Up on Inequality
- Poverty Rates in Recent Years
- Intellectual Garbage Collection: The Unreliability of Alan Reynolds (Brad DeLong)
- Settling the Score on Census Top-Coding of Income Data
- "The Chewbacca Defense"
- [Update: Piketty and Saez respond]
Here's Reynold's latest. It starts with a description of the claims that are in dispute, followed by (his version of) the dispute itself. At one point, he mentions some of your comments:
A 15-year debate on wealth, by Alan Reynolds, The Washington Times: I wrote a Wall Street Journal op-ed in December showing the top 1 percent did not receive 16.1 percent of personal income in 2004, but just 10.6 percent. Most of the post-1980 increase in that number happened between 1986 and 1988, when tax rates fell. Any apparent rise since then is due to business income shifted from the corporate tax to the individual tax. and investment income of middle-income taxpayers shifted from taxable to nontaxable savings.
Two "liberal" blogs went predictably apoplectic, but in ways that had absolutely nothing to do with what I had written. Berkeley professor Brad DeLong initially said there is never any point in paying attention to anything on the Wall Street Journal editorial page. Someone balked at that, so he tried describing two clear, consistent and distinctly separate statements from my article as "three-card monte."
Such amusing irrelevancies led to the usual swapping of links between like-minded cheerleaders at another blog run by Mark Thoma of the University of Oregon. He somehow imagined my article relied mainly on Census Data, and therefore borrowed some curious comments from a New York Times column by Paul Krugman.
Mr. Krugman complained that Census estimates are from a "limited sample" (as if IRS estimates are not) and wrote "the questionnaire is 'top-coded': if the individual interviewed has earnings higher than $999,999, those earnings are recorded simply as $999,999." Someone thought those words were mine, and posted an angry but correct rebuttal. Top-coding is almost entirely confined to "public use" files, censored for privacy. Census officials use internal data with the missing details, including incomes far above $1 million.
In the process of changing the subject -- away from his own error and from everyone's inability to find any fault in my article -- Mr. Krugman posted a link back to Mr. DeLong's blog. Mr. DeLong, in turn, had harvested two "demonstrably false accusations" from the hundreds of articles I have written. One dates back to 1992.
My only "false accusation" since 1992 involved a careless but irrelevant mistake last March, which I acknowledged at the time on Mr. DeLong's blog. It concerned a column in The Washington Post that claimed "the Internal Revenue Service" had reported that the top 10 percent of households earned 44 percent of all pretax income in 2003. The source was a short paper by three statisticians, two of whom -- Michael Strudler and Tom Petska -- work for the IRS. But that does not make it official IRS data.
My column showed their estimated 44 percent is much higher than any official source. The authors exclude Social Security and other transfer payments from the denominator (total income), then pump up top incomes with such indefensible items as accelerated business deprecation and nontaxable IRA rollovers.
When I described the much smaller estimates of the top 10 percent's income from Census and the Congressional Budget Office, Mr. DeLong discovered: "Reynolds is wrong. CBO estimated that for all households the income of the top tenth in 2003 was 37.2 percent. ... Reynolds' 38.3 percent came not from Table 1: All Households, but instead from Table 3: Elderly Households." That is true. I missed that fine print when the spreadsheet popped up on my monitor.
How does the fact that the correct figure is even smaller than I said (37.2 percent, not 38.3 percent) debunk my skepticism about that bloated 44 percent figure? It is true the corrected CBO figure did rise significantly at the time of the 1986-88 tax reform. Top incomes and capital gains also boomed with the stock market in 1998-2000. Yet the CBO estimate of the top 10 percent's share of after-tax income ended up unchanged from 1988 (33.1 percent) to 2003 (33 percent). How does that contradict what I wrote in the Wall Street Journal?
The other reason Mr. Krugman and Mr. DeLong advise people to ignore what I wrote last month is because of something I wrote in 1992. At that time, Mr. Krugman objected to what he called my "main" argument. I argued that short-term changes in the estimated wealth of the top 1 percent could be distorted by a small handful of people ("outliers") with huge windfalls. Think of the two founders of Google getting $12 billion apiece when the firm went public. If they show up in a sample, how meaningful is the average going to be?
Mr. Krugman replied that the working paper -- by Arthur Kennickell of the Fed and Louise Woodburn of Ernst & Young -- had oversampled high incomes from tax return data. He said 400 people had been picked to represent the wealthiest 1 percent. A larger sample certainly helps, but it does not solve the "outlier" problem, because a few huge windfalls can still make the statistical average in a single year appear much higher than typical (median) wealth of any small group with no ceiling on wealth.
Mr. Krugman feigned indignation in 1992 that I had not read the relevant footnote in that unpublished working paper. He could be entirely confident about that, because 1992 was long before you could find such obscure papers online. The only way I could possibly have read an unpublished draft would be if the authors had sent me a copy when they sent one to Mr. Krugman. I did eventually obtain the paper by snail mail, too late to meet the deadline. I had no choice but to comment on what Mr. Krugman had written about the report -- in the supposedly disreputable Wall Street Journal.
Fortunately, a much-improved August 1997 revision of the "preliminary" Kennickell-Woodburn paper is now online. The authors revised their weighting scheme and reversed their previous conclusion. Their original 1992 study found no change in overall wealth inequality between 1983 and 1989, or in the wealth share of the top 10 percent.
For the top one-half of one percent, however, the first draft seemed to show "a dramatic increase from 1983 to 1989 under the original weights." In the revised version, "the point estimate of the [wealth] share of the top 0.5 percent in 1989 is lower than that in 1983." The share of wealth held by the very top group declined, from 1983 to 1989. My skepticism in 1992 proved entirely warranted.
We all make mistakes. My book, "Income and Wealth," repairs some innocent mistakes by Brad DeLong and Paul Krugman. If anyone can demonstrate I misquoted such scholars, or am mistaken about any facts in last month's Wall Street Journal piece, I will gladly correct the record just as readily as I have now corrected the record about my 1992 piece.
As this post demonstrates by quoting Martin Feldstein, Glenn Hubbard, Ben Bernanke, and Ed Lazear, all first-rate economists with impeccable conservative credentials, there is wide agreement that inequality has been increasing. Reynolds says otherwise, but few reputable economists agree. But I think rather than getting into the details of this yet again and addressing all of the ways the latest editorial mischaracterizes the debate and the data, I'll just repeat what Paul Krugman said in the last link above:
On the broader issue: Reynolds says that various statistical issues have created a false impression of rising inequality. Now, serious researchers, from CBO to the IRS to Piketty and Saez, have looked at those issues, acknowledged them, but concluded that they don't make enough difference to change the picture in any fundamental way. ...
One last point: we have a number of indicators other than government data on what's happening to very top incomes and wealth - things like estimates of executive compensation. All these indicators point to a continuing rapid rise at the top compared with the middle. So any claim that the rising inequality we see in both Census data and in tax returns is some kind of statistical illusion faces an additional credibility problem.
Update: Email says I should take a much stronger position on this one. For example:
You're way too gentle. Reynolds is still not admitting his two recent errors: his claim that the censored top income data are included in the Census income share, which is flatly false, and his accusation about the CBO data. He's misrepresenting the debate on your blog to make it seem as if my minor misunderstanding about "reporting limits" versus the questionnaire is more important than his fundamental failure to understand what the Census data do and don't show.
But maybe Brad DeLong has it right in his discussion of another columnist who can't get things right:
As one colleague said on Friday apropos of another issue: "David Brooks? You're using a tenured Harvard statistician to refute David Brooks?! You don't use a tenured statistician, you use a fly-swatter!"
Update: Piketty and Saez respond to Reynolds.