*Economist’s View* on Alan Reynolds

by on December 17, 2006 at 8:01 am in Data Source | Permalink

Here is a further exploration of data issues.

Another post worth reading is The Conspiracy Against Cuckolds.

8 comments

Edgardo December 17, 2006 at 8:36 am

I have not read Alan Reynolds’ new book, INCOME AND WEALTH, so I cannot comment on his evidence to support the ideas presented in the WSJ article. Apparently none of Reynolds’ commentators have read the book either. I’d appreciate greatly if you can refer to reviews of Reynolds’ books.

Cyril Morong December 17, 2006 at 12:26 pm

Can someone tell me if this issue of making money under the table is related to what the IRS calls the “tax gap?” In 2001, $290 billion in taxes was not collected that was supposed to be collected. Is this due to unreported income? I can’t tell for sure from the article (link below). How much income would have to be unreported to leave $290 billion uncollected in taxes? If the government takes, say, 20% of income in taxes, then we multiply 290 times 5 to get 1450. That would be over a trillion dollars in unreported income. Then the question is which quinitles are earning that income?

Here is the link to the IRS press release

http://www.irs.gov/newsroom/article/0,,id=154496,00.html

I just can’t tell for sure why or what the source is of the shortfall in collections. Is it under reporting income or just not paying taxes? How much unreported income is there estimated to be?

John Thacker December 17, 2006 at 2:22 pm

Hmm. In several places Mark Thoma seems to claim “yes, everyone’s aware of that objection” as though that explains away the objection in a way that I’m not really comfortable with.

But inequality is complicated because plenty of things are not time-homogenous. As all parties concede, changes like the percentage of women working, immigration, and people staying in school longer (trading years of poverty-level student income for higher incomes once they get out) can increase the inequality statistics without necessarily meaning anything real, or at least inequality in the standard sense.

Robb Lutton December 18, 2006 at 5:18 am

Tyler,

A few posts ago you were saying Reynold’s stuff could be “very very important”. Now that Thoma and Delong say that he is essentially a charlatan, you simply point to the posts with no comment.

Cat got your tongue?

Ban Cheah December 18, 2006 at 12:37 pm

There were two points that I took away from the discussion so far:
1) Publicly available Census Population Survey Data estimates of income distribution are biased because of top coding of high incomes. This is the source of the Reynolds data.
2) Piketty Saez’s use of tax returns to compute income distribution may be biased because the sample of tax returns may not be representative of the population.

Here is one approach: Sign a confidentiality agreement with Census to use the non public version of the CPS data. (This may be easier said than done.) This gives the analyst access to tabulate the household income distribution from the CPS which is a representative sample of the US Population.

I’m avoiding the discussion on tax units versus households for now but it is also an important distinction. From my perspective I would be more interested in inter household inequality but using the non public version of the CPS data would give an anchor to an estimate which I would find more believeable. We can then see if there is any bias in the Piketty Saez data.

hederman December 19, 2006 at 10:50 am

” Publicly available Census Population Survey Data estimates of income distribution are biased because of top coding of high incomes. This is the source of the Reynolds data.”

The topcoding of Census Data will only effect distribution analysis that looks at the tiptop. Maybe the 99-10 income inequality ratios are effected but that’s about it. Krugman is absolutely wrong with his understanding of topcoding methodology of the Census Bureau. I believe that Krugman or his assistant read the questions of the march survey and that led to his understanding of a topcoding problem in the survey questions.

Also, the Census Bureau had a more hard topcode prior to ’95 so you have serious comparison problems due to topcoding before ’95.

For topcoding individuals, the Census bureau assembles a matrix based mostly on demographics. Everyone in that cell is given the same level of income that adds up to the aggregate. For instance, if one person makes 800 million and the other makes 200 million, if they’re both in the same demographic matric, the Census will report their income as $500 million.

So if you’re looking at aggregate levels of the top 5 or 10 percnet, you’re fine. If you’re lookign at the top 1 percent you’ll have some problems. But Census data does contain the aggregate level of income as reported to the Census Bureau.

There are other arguments against Census Bureau data such as the fact that it is consistently short of NIPA data. But that’s a different argument.

Alan Reynolds December 29, 2006 at 11:42 am

An early post here asked about reviews of my textbook, Income and Wealth.
It’s fairly new, so I have only seen a few:

http://www.nypost.com/seven/11122006/postopinion/postopbooks/money_myths_postopbooks_diana_furchtgott_roth.htm

http://www.townhall.com/columnists/RichLowry/2006/12/07/america_gets_richer

http://www.amazon.com/Income-Wealth-Greenwood-Business-Economics/dp/0313336881

The topic of my Wall Street Journal piece is covered in much more detail in Chapter 5 of the book and the end of chapter 4, and also in the forthcoming paper Chris Edwards mentioned at cato.org. There will be an academic version published later, with a co-author, but such publications always take a long time.

Most posts at Brad DeLong’s blog and Mark Thoma’s do not address the only topic I covered in my Journal op ed — namely, the misue of tax data to estimate changes in income distribution. Most comments instead try to anticipate the Cato paper, such as assuming I ignore the CBO or rely entirely on Census data. It is rarely prudent to comment on a purely statistical issue without having first looked at the statistics.

Alan Reynolds January 9, 2007 at 11:46 am

Thacker’s comment about family size is very relevant. In replying to me, Piketty and Saez use the word “families” to describe tax units. That’s worse than misleading. There are two workers per household in the top quintile, which usually means one tax unit (a joint return). There is one-half worker per household in the bottom quntile, plus a lot of people (especially seniors) who earn too little to file. There are 5.5 times as many full-time workers in the top fifth as in the bottom fifth, and the middle groups are full of households with two or more single people filing two or more tax returns per household. With two salaries called a tax unit at the top and few or many salaries called a tax unit in other quintiles, there is no way to infer family income from tax unit income.

The CBO tries to adjust for family size, to their credit, but that doesn’t quite work either.

Comments on this entry are closed.

Previous post:

Next post: