« But Can He Raise Taxes? | Main | On Saddam »

December 30, 2006

Stop. Digging.

Being this wrong, this often, must be genuinely difficult. Cato is touting a coming event on inequality featuring the esteemed Alan Reynolds, even though he apparently remains incapable of understanding how his core data set measures inequality. I guess the event -- "Has U.S Income Inequality Increased?" -- is for his own edification.

December 30, 2006 | Permalink


Mr. Krugman wrote that Census Bureau data was inferior to Statistics of Income tax data used by Piketty and Saez and the CBO because the Census sample is limited. Unless he imagined the SOI sample included every tax return (was unlimited), the complaint about samples being limited made no sense at all.

He also wrote that because of top-coding the Census Bureau records all income above a million dollars as if it was a million dollars. The CBO source he found on Google does not say that, of course, because it is not true.

Top-coding is a problem for researchers who use the March current population survey, because they cannot peer inside the top 5% to see, for example, how much is earned by the top one-hundreth of one percent.
Because a very high income might be used to identify an individual, the Census strips out such information (as well as addresses) before releasing public use files to the rest of us. But the Census Bureau uses confidential, restricted-access microdata, not public use files, when calculating income shares of households and families by quintile and smaller fractiles.

A December 18 post from Herderman (whose name I once misspelled) said, “Whoever wrote this is absolutely wrong on how the Census handles topcoding information. If you are working with aggregate data such as the total amount of income above the topcoded amount, then the Census does report the aggregate amount as accurately as they can. The Census questionaire is NOT topcoded and neither is the data in the CPS. Topcoding is a construct for confidentiality but does not lessen the aggregate amount of reported income.”

Nobody answered that Herderman was wrong about that, because Anne thought Krugman’s statement about top-coding came from me. Yet I am now said to be wrong from saying the same thing simply because the comment Herderman correctly debunked came from Krugman.

Researchers with rare “special sworn status” used their access to confidential Census data in a November 20 paper, “Using the P90/P10 Ratio to Measure Inequality Trends with the Current Population Survey: A View from Inside the Census Bureau Vaults” (November 20, 2006) by Richard V. Burkhauser of Cornell University, Shuaizhang Feng of the Shanghai University of Finance and Economics and Stephen P. Jenkins of the University of Essex.

Using Gini coefficients, as well as narrower ratios of income in the top 10% to income in the bottom 10% (which exclude 80% of households), they found “there is no overall trend in the P90/P10 ratios during the period 1975-2004,” and “there is no change in levels for either the Gini or P90/P10 ratio in the post-1992 period.”

Using a variety of Census, Fed and CBO data, I find no “significant and sustained” (post-2000 boom) change in inequality since 1988. Burhkauser, Feng and Jenkins find no significant change in the inequality of household income of wages of full-time employees since 1975 by one measure, or since 1992 by another.

Like my still-unanwered critique of the Piketty-Saez estimates, the findings of Burkhauser et. al. may irritate people who prefer to put their faith in someone else’s credentials rather than to even bother figuring out what limited samples and top-coding mean.

Posted by: Alan Reynolds | Dec 31, 2006 9:50:56 AM

From what I read, it's not credentials, personality, or any of the other things that you claim is at issue. It's the fact that you were factually wrong, and rather than saying "I erred" you instead keep deflecting blame off of your fuck up (I am not as nice as others). If you were truly interested in honest debate, the first step, the absolutely first, is to admit the mistake, and stop bullshiting. Peace.

Posted by: akaison | Dec 31, 2006 12:46:28 PM

Akaison, to whom is your post addressed? I don't see anything in Alan's post about credentials or personality, though I did see such things made issue of in the posts Ezra linked to, a fair amount of which are just hateful bullshit. Since you apparently understand all this, would you mind pointing out the factual errors in what Alan said? And please look up the word "peace."

Posted by: Sanpete | Dec 31, 2006 2:12:25 PM

Its on the other website. Look it up yourself.

Posted by: akaison | Dec 31, 2006 2:51:15 PM

As my comment implies, I've already looked. You're the one bullshitting here. Going to follow your own advice?

Posted by: Sanpete | Dec 31, 2006 3:04:08 PM

Not really because I think that article, links, subsequent comments, etc do it for me. Your calling it bullshit and then not elaborating on how those people who you made their points over there, with subsequent links, etc are full of shit isn't worth my time. It's also like the whole thing about your calling my use of peace inappropriate. Peace.

Posted by: akaison | Dec 31, 2006 4:34:07 PM

Akaison, perhaps you could point out where in the thread Ezra bases his comment on it is shown that Alan was wrong about the points in Thoma's topic post. In fact that thread shows plainly that Alan was right about the points at issue there, and that Krugman was wrong, but Thoma and others, rather than admitting their mistake, have shifted the topic to other things they feel they can attack him on, while with apparently unintentional irony accusing him of misdirection and obfuscation. Pretty sorry stuff, as is so often true when intellectual debates turn into personal attacks. You, in particular, have shown no error on Alan's part in regard to these points, and all you can do is point to others who haven't either. I think you should take you own advice.

If you think "peace" means "I kick" or something like that, then your use of it is entirely appropriate to your understanding.

Posted by: Sanpete | Dec 31, 2006 5:31:42 PM

okay whatever

Posted by: akaison | Dec 31, 2006 7:44:50 PM

Ezra, you might consider closing comments when you're merely pointing at another thread. These responses belong over there. Mr. Reynolds simply cut-and-pasted his response. I'll link.

Here is my detailed takedown of his comment (can you taste the cherrypicking?), and here to a refutation of Sp's defense (scan past my pleading with "Anne" to stop cluttering threads with NYT articles).

Posted by: wcw | Jan 1, 2007 5:38:46 PM

I see your point, wcw, but I think it's good that Alan was able to post his response here, since the charges were repeated here.

Keep in mind that your "takedown" is in regard to his use of a paper, not his response to Krugman's top-coding comment, the actual topic at hand, which only a few people seem to have managed to keep track of in all the noise about other things.

Posted by: Sanpete | Jan 1, 2007 7:02:14 PM

Cf my response to your more detailed comment, over where it belongs.

As I said, I believe the appropriate policy in pure 'pointer' posts like this is to have comments closed entirely.

Posted by: wcw | Jan 1, 2007 10:09:00 PM

The link at the top, about my being "wrong this often" (twice since 1992 according to Messrs. Krugman and DeLong) is answered in my latest column:


Unlike "wcw" who prefers that comments be closed, I always welcome informed comments. Sanpete's discussions (mostly on other blogs) have linked to informative papers, which is a good technique. If I knew how to put links into comments, I'd do that too.
My book, Income and Wealth, includes URLs where avaialable so readers can check to see if I've made mistakes.

Even if someone reads my new column and still thinks I made some inexcusable mistake in 1992, I don't see how that allows anyone to assume I must have made a mistake in The Wall Street Journal on Dec. 14.

Is it really too much trouble for Brad DeLong to walk down the hall and ask Emmanuel Saez if I am wrong, for example about the top 1 percent receiving 10.6% of personal income? Was I wrong about business accounting for 28.4% of the top 1 percent's income in 2004 -- up from 7.7% in 1981? And if none of that was wrong, can't we all agree to use some less problematic data to demonstrate the supposedly large and continuous increase in income inequality over the past 20 years?

Posted by: Alan Reynolds | Jan 4, 2007 1:59:21 PM

The comments to this entry are closed.