Posted 2:35 PM by Robert

Lies, Damn Lies and Representative Statistics In The Grand Duchy of Krugmania

Don Luskin points out this beaut of a letter of rebuke to Paul Krugman from Andrew B. Lyon, a deputy assistant secretary at the Treasury Department:

As deputy assistant secretary responsible for tax analysis at the Treasury Department from April 2001 to July 2003, I must respond to Paul Krugman's Aug. 5 column accusing it of political bias.

Mr. Krugman objects to an analysis of the 2001 and 2003 tax cuts for six representative families, including a married couple, both 65, with $40,000 in income, including $2,000 in dividend income. He says that such a family is not representative of elderly taxpayers.

Obviously, no single example can be representative of all elderly taxpayers, but this couple has income quite close to the median for filers their age, and dividends quite close to the average for elderly filers at that income level, about half of whom receive dividends. The couple in the example would see its income tax liability decline from $1,396 to $675 as a result of the tax cuts.

Mr. Krugman is certainly free to oppose the tax relief, but he should be more circumspect in asserting political bias in the underlying analyses.

ANDREW B. LYON

The New York Times actually published this letter. Ouch!

Herr Doktorprofessor's column which the letter addresses was previously discussed by the Man Without Qualities.

Herr Doktorprofessor responds:

Did you get that? I'm more accustomed to this sort of math than 99 percent of readers, and it took me about 6 reads. The average dividend of elderly families with the median income, half of whom receive dividends ... it sounds as if this refutes what I said . But you'll notice that he never takes on my assertion that only about 1/4 of elderly families receive any dividends, which comes from Tax Policy Center ; nor does he challenge my statement that only 1 in 8 elderly households receive as much as $2000 in dividends, which comes from CBPP . Here's CBPP's picture:

[Here Herr Doktorprofessor pointlessly inserts a pie chart showing nothing but the fact that 88% of the "elderly" have less than $2,000 in dividend income, apparently to impress people who are impressed by pie charts with only one slice.]

I leave it as an exercise for readers to figure out how Lyon's discussion is consistent with a real world in which the Treasury's "low-income" elderly household has more dividend income than 7 out of 8 elderly households.

The point, however, is that the confusing discussion shows the Treasury strategy at work. 1. Never answer the question directly. 2. Offer an elaborate calculation that sounds as if it refutes the accusation that the tax cut mainly favors the wealthy, when in fact it does no such thing.

For afficionados: notice how they're still doing the average/median thing. It's clear from what he says that the median dividend for families with the median income is ... 0.

Update: I couldn't provide a direct link to the piece from Tax Policy Center, because their web site appears to be down today. But here is a good summary table from Citizens for Tax Justice. By the way, both CTJ and TPC basically have the Treasury model of tax impacts, so their numbers must correspond very closely to what Treasury knows but won't tell us. For more about all this, read Sullivan's angry article in Tax Notes.

As a preliminary matter, it is worth considering the rhetoric of Herr Doktorprofessor's "money line": It's clear from what [Mr. Lyon] says that the median dividend for families with the median income is ... 0. That way of putting it makes it sound as if the number of median income families with dividend income must be insignificant (even zero) - and that's the impression Herr Doktorprofessor wants to make with this flourish. But this money line means exactly the same as: more than half of all families with median income have no dividend income. The money line would still be true if, say, 49% of all median income families owned stock - and that all of it paid substantial dividends. His rhetorical flourish (his "money line") articulates a standard that doesn't really address the issue it seems to address at all. He is here committing exactly the sin of which he accuses Mr. Lyon.

As a "reader" I take Herr Doktorprofessor up on his challenge: "I leave it as an exercise for readers to figure out how Lyon's discussion is consistent with a real world in which the Treasury's "low-income" elderly household has more dividend income than 7 out of 8 elderly households."

Mr. Lyon writes: [T]his couple has income quite close to the median for filers their age, and dividends quite close to the average for elderly filers at that income level, about half of whom receive dividends.

Mr. Lyon does not use the term "low income" to describe this family, so why does Herr Doktorprofessor use it and put it in quotes, as if he is quoting and debunking Mr. Lyon? Mr. Lyon expressly states that they are a married couple, both 65, with $40,000 in income, including $2,000 in dividend income, ... income quite close to the median for filers their age. Repeating: Mr. Lyon says they are a median income family - not a low income family.

Mr. Lyon also expressly states that the family's dividends [are] quite close to the average for elderly filers at that income level. So if you happen to be a 65 year-old couple with about $40,000 in income, this example will likely be of personal interest.

Herr Doktorprofessor is right when he says of Mr. Lyon's letter: you'll notice that he never takes on my assertion that only about 1/4 of elderly families receive any dividends. Instead, what Mr. Lyon does is point out that this family is representative of a different, well-defined group of the elderly - a group more representative of elderly people who are, in fact, affected by a change in the income tax than are the groups on which Herr Doktorprofessor wants to focus exclusively. Herr Doktorprofessor is entitled to present his own "representative" families. Mr. Lyon hasn't suggested that Herr Doktorprofessor's "representative" families are somehow fraudulent, as Herr Doktorprofessor has suggested of Mr. Lyon's. But Herr Doktorprofessor's "representative" families don't indicate the effect of tax changes on elderly families that have dividend income. Further, now that the tax on dividends has been reduced, more elderly families may in the future want to acquire dividend-paying stocks - and Mr. Lyon's numbers will be helpful to such families in making that decision, but Herr Doktorprofessor's could not be helpful. In important respects Herr Doktorprofessor is like someone insisting that one cannot look for a lost quarter except in those parts of the room with average lighting, regardless of where the quarter fell. [And, if that's the game, you'll notice that Herr Doktorprofessor never, here or elsewhere, takes on Don Luskin's point (also pointed out by those Luskin cites and links) that Herr Doktorprofessor's separate California growth-and-taxes column is full of basic math errors that "somehow" come out favoring Herr Doktorprofessor's biases. A little sauce-for-the goose, Herr Doktorprofessor? Eh?]

In some respects Herr Doktorprofessor's "answer" to Mr. Lyon seems to be little more than a wan, special-purpose reprise of the anti-tax-cut crowd's complaint that the recent public income tax-cut discussion should never have focused on the laws effects on people who pay income taxes. That crowd rants that the public discussion of a law reducing income taxes must always be couched in terms of people who pay any federal taxes. To do anything else, they and Herr Doktorprofessor maintain, is a lie - even if you disclose what you are doing.

I have pointed out in prior posts that the huge expenses imposed on the economy by federal regulation, as well as the prevailing anti-business litigation climate which the Commerce Clause of Constitution gives Congress the power to address, are economically equivalent to "hidden" federal taxes and expenditures. But the anti-tax-cut crowd never want those hidden costs and taxes included in tax cut debates. In fact, all efforts even to formulate and impose a comprehensive federal "regulatory budget" accounting for the costs imposed on the economy by federal regulation have been savagely opposed by those Herr Doktorprofessor supports in his anti-tax-cut fervor. That doesn't really bother me - intellectual constructs have to stop somewhere. But I just don't understand why people like Herr Doktorprofessor keep insisting that constructs that don't stop exactly where they want them to stop are fraudulent.

Why can't he just give his own examples, let the reader determine how good they are, and then shut the heck up?

Mr. Lyon uses the median as his basis of determining this "representative" couple (which he never maintains tells the whole story, contrary to Herr Doktorprofessor's whole rant), while Herr Doktorprofessor seems to fume that Mr. Lyon just must (must, Must!, MUST! MUST!!!! - goddamnit) stick to "averages" - at least to the point of not choosing a family of median income. But which is more "representative" of the effect of Warren Buffet walking into a bar: The resulting change in the average or median income of the people in the bar? Yet, as with Herr Doktorprofessor's "money line" above, comparisons based entirely on medians can omit a great deal of important information - he would say such comparisons were fraudulent if they were made by a Republican, especially a member of the Administration. But the correct response is not to foam at the mouth, but to point out the limited sweep of the example one doesn't like and give examples of one's own. Yes, not all Americans are 65 making $40,000 and receiving dividends, Herr Doktorprofessor. But that's hardly a weird, fraudulent not of the "real world" example - even if the couple is not a completely average elderly couple either.

"Medians" are often used to express (imperfectly) generic economic developments, especially when addressing non-specialists. For example, many newspapers, including the Los Angeles Times normally report fluctuations of local home prices by reference to median sales prices within given zip codes - and not average sales prices. Do those reports tell the whole story of how the housing market is fluctuating? No, of course not. Are such medians worth publishing? Yes, of course they are. But Herr Doktorprofessor seems to be implying that those papers are all but perpetrating some kind of fraud on the public. All illuminating comparisons based on "medians" and "averages" (like the Times home sales reports and Mr. Lyon's family example) are to some extent unrepresentative - and so are such comparisons such as Mr. Lyon's which are based on combinations of "median" and "average." Indeed, Herr Doktorprofessor frequently rails against the use of averages as being unrepresentative in economic matters - as, by way of many examples, his rants against what he considers the national over-concentration of income and wealth - including, as an astute reader points out, the case of Bush administration citing for its tax cut an "average" benefit of $1,600 per family, which elicited howls of protest from Paul Krugman's crowd that this was a misleading number because a few super-affluent families can skew averages substantially.