In Which I Fall Down on the Job...

A correspondent writes, asking where is my quarterly post reminding the internet that Donald Luskin--National Review's contribution to the grand coordinated right-wing Paul Krugman-trashing enterprise ably reported by Nicholas Confessore--more often than not simply doesn't know what he is talking about.

Now it is true that the right-wing campaign has collapsed--even its two original leaders, Mickey Kaus and Andrew Sullivan, now admit that Paul Krugman's batting average since he started at the New York Times has been above 90%. But Luskin is still out there: now he is trashing Alan Blinder in a piece that could not have been written in good faith if Luskin had actually read Blinder's stuff on outsourcing/offshoring.

You do have to wonder just what National Review thought and thinks it is doing by sponsoring Luskin. The overall effect is loony. After all, we have seen Luskin's ignorance of what a real exchange rate is, his claim that David Brooks is both 100% correct and is a traitor to his party for saying that the Bush administration routinely lies, his denial that the yield on a bond is an interest rate, his accusation that Gretchen Morgenstern is a plagiarist, his claim that the Bushies--Armitage, Libby, Rove, Cheney, Fleischer, and company--could never have revealed the identity of CIA undercover operative Valerie Plame because that would mean that they were guilty of something tantamount to high treason, his claim that George Soros would try to crash the market on October 31, 2004 to elect Democrats, his being off by a factor of five in calculating the liabilities of the Social Security system, his erroneous claims that faster productivity growth doesn't help the Social Security system--you get the picture. Is National Review deliberately trying to further damage the intellectual reputation of all its contributors?

But I can't do it.

I don't have the heart to surf on over to Luskin's web site and point out stupid errors, and egregious tendentious deliberate misinterpretations.

It's just too depressing.

I can, however, grab two truly unbelievable--but, alas! not atypical--examples of Donald Luskin's work from my weblog archives:

For my first example, let's recall Luskins demonstration that he cannot calculate a real exchange rate [1]:

Back in late 1982, the young Larry Summers and Paul Krugman--working for Marty Feldstein in the Reagan administration--wrote a memo in which they warned of the (very real) danger that large budget deficits and a falling dollar might reignite domestic inflation:

As the figure above, which plots the real exchange rate--the real value of the dollar--shows, they were right: the dollar was then overvalued (and was about to become much more overvalued in the next two years before returning to equilibrium). Luskin, trying to trash them, presented this graph:

and wrote:

The chart... shows the real exchange rate of the US dollar for a decade before the 1982 memo, and then through the end of the Reagan presidency. It did drift slightly lower for the first couple of years after the memo. But then it took off to new highs -- nothing resembling anything like a "return to approximately their historical levels."

It is clear that one thing wrong with Luskin's graph is that it is upside down--the dollar falls in value from early 1985 on, it doesn't rise. The dollar doesn't "take off to new highs" after 1985--as anybody with any knowledge of the historical configuation of asset prices would know. But it is also clear that there are other things wrong with it: the real exchange rate in 1989 was within 15 percent of its value in 1973, not different by a factor of four.

We still do not know how Luskin generated this graph.

For my second example, there is Luskin's single-handed denial of the existence of the entire intellectual discipline of statistics:

The Times and the Journal cite many authoritative-sounding studies [of income inequality].... But to get an accurate picture... you'd have to track hundreds of millions of individuals.... [N]one of this is reliable... the Panel Study of Income Dynamics... tracks only 8,000 families out of a U.S. population of 295 million individuals.

The whole purpose of the science of statistics is to tell us that this is simply not true. As long as you can take a random sample of your population, you can find out an enormous amount about the population from a relatively small number of observations. You can find out what proportion of rich people had poor paretns, or what proportion of twenty year olds think they will graduate from college, or pretty much any other average proportion that you want.

Now the "random sample" part of this is very important. But if your sample is random--if the fact that the yes-no pattern of observations so far makes it no more (or less) likely that you next observation will be a "yes"--then the law of large numbers tells us that the sample average you compute will converge to the true population average at a frighteningly rapid speed.

The standard demonstration of this is to repeatedly flip a coin and count the excess proportion of heads over tails. We know that--with a coin flipped and caught in the air by a human being at least--the population average taking all coins that have ever been flipped of the excess proportion of heads is zero. How many observations do we have to take--how many coin flips--before the sample average converges to this population average of 0% excess heads?

Let's see. Here's one run of 1,000 "flips" from Excel's internal random number generator:

Here are ten more:

Impressive, no?

Try some yourself.

You could have a population of 295 million flipped coins. Yet you don't need to look at "hundreds of millions" of them to determine what is going on. Looking at 1,000 will do.

This is the principal insight of the science of statistics. it is an important insight. It is a powerful insight. It is also not an obvious insight--that's what makes it powerful and important.

Yet because statistical studies sometimes produce results ideologically inconvenient for the Republican Party, Donald Luskin either feels he has to pretend that this insight doesn't exist, or is so stupid that he doesn't understand the first thing about the Law of Large Numbers.

That's really sad.

---

[1] Daniel Davies affects to feel sympathy for Luskin:

It is precisely this [same] mental block which led to my being forbidden by colleagues from ever mentioning exchange rates in public meetings. Oddly enough, clients tended to complain about the fees when the team's "technical expert" was seemingly unable to distinguish between division and multiplication :-)...

Comments

[MaryLou]

But didn't you whack Mr. Luskin in your March 23, 2007, post about social security and productivity growth?

[Barry Ritholtz]

Brilliant defense; Here's the dirt on Luskin when he was at TheStreet.com:

Check out this:

http://www.thestreet.com/p/_rms/dps/cc/20010607/columnistconversation13.html#entryId108136

http://www.thestreet.com/p/_rms/dps/cc/20010607/columnistconversation14.html#entryId108138

http://www.thestreet.com/p/_rms/dps/cc/20010614/columnistconversation7.html#entryId108357

[Bruce Wilder]

"As long as you can take a random sample of your population, you can find out an enormous amount about the population from a relatively small number of observations. You can find out what proportion of rich people had poor parents, or what proportion of twenty year olds think they will graduate from college, or pretty much any other average proportion that you want."

This is really just a quibble, but it should be pointed out that the size of the proportion matters. If the proportion you are estimating is one, whose true value in the population is close to .5, then only a very small random sample is necessary to estimate it accurately, as is demonstrated by the case of a coin-flip in the post. Thus, the sample-size needed to estimate the vote in a Presidential election is pretty small, in part because the true proportion is near .5

But, the examples given in the quotation above are actually proportions of proportions -- proportions of subpopulations, which are themselves only small proportions of the total population. Rich people or twenty-year-olds form only a small proportion of the total population. You would need quite a large random sample from the whole population, before you could accurately estimate a proportion of a small subpopulation.

If you want to know what proportion of twenty-year-olds favor a Presidential candidate, you basically need a random sample of twenty-year-olds. And, the size of that random sample would be very close to the same as the sample needed to estimate the proportion of the whole population, who favored a Presidential candidate.

[MattF]

Hey Brad-- it's Gretchen MORGENSON. MORGENSON.

[Zeno]

A few months ago I wrote a post (A primer on polling) about some of the math behind political polling. I gave some concrete examples demonstrating how quickly a poll's accuracy increases with growing sample size. The sample size does not have to be particularly large before the results start becoming quite meaningful. This surprises a whole lot of people.

[Barry]

Bruce, (a) some statistics will have lower confidence intervals than others; (b) stratified sampling can be used.

[bakho]

Maybe Luskin gets his info from "Conservapedia"?

http://www.conservapedia.com/Main_Page

"Tired of the LIBERAL BIAS every time you search on Google and a Wikipedia page appears? Our study suggests that Wikipedia is 6 times more liberal than the American public. Now it's time for the Conservatives to get our voice out on the internet!
Conservapedia began in November 2006, as the class project for a World History class of 58 advanced homeschooled and college-bound students meeting in New Jersey.
Conservapedia has since grown enormously, including contributors worldwide. Conservapedia already exceeds the number of entries in the Oxford Dictionary of World History. Conservapedia is rapidly becoming one of the largest and most reliable online educational resources of its kind."

[Julian Elson]

I have a hard time believing that, even from a narrowly partisan, NRO perspective, without regard for the quality of public discourse, Luskin's positives in spreading fake conservative economics outweighs the damage he does to NRO's credibility among independents, intelligent conservatives, and generally everyone except exceptionally dogmatic and unintelligent conservative readers. Is he really worth hiring from the National Review's perspective? Why do they keep him? Is it just wingnut welfare? Or do people actually read Luskin and take him seriously?

[Kate G.]

Brad,
You are doing g-d's work but I really don't envy you.

Kate G.

[W.C. Varones]

Since when are Mickey Kaus or Andrew Sullivan "right-wing?" You'd have to be a truly delusional left-wing academic to believe that.

[Matt]

The New Accounting view of interest rates.

Interest payments are an approximation, with decreasing error, of the power spectra removed from input volatility by business process; in a system undisturbed.

The equation is: productivity = interest + estimation noise. (For each increment of volatility removed)

I left out entropy inefficiency and bundled it into estimation error.

Noise is the base noise level below which any information about the system is unreliable. (This is related to Delongs law of large numbers)
Volatility decreases in theory, but entropy is still increasing, so the noise level always remains. (The central limit can never be reached)

Note: By virtue of the min energy assumption, some triangle relationships appear, so there is a central limit theorem.

So, as the P(z) equilibriate by removing volatility from the input and creating a better estimate of the true P(z), the certainty of P(z) increases and volatility decreases. As volatility decreases the opportunity of removing volatility approaches the noise level, the system approaches a liquidity trap and current interest increment goes to zero. [An informal description of the central limit theorem]

There is another theorem, which I present as a conjecture. The interest paid is most efficient when distributed on the yield curve in proportion to P(z). Which is to say, a business wants its next amortization to be funded by borrowing in a frequency spectra that matches its P(z).

The yield curve, therefore represents that aggregate recent estimate of the aggregate
P(z) of the whole economy. It is maximum probability next best estimate of
aggregate P(z).

The yield curve will be less a "min energy" chunk of volatility as time increase. But each increment of the yield curve in a sample period always makes the P(z) in aggregate and individually more "min energy"

So, finally we have some norms to apply. To the extant that all the P(z) in the economy differ dramatically, interest payments are used less efficiently, and in particular, large P(z) groups that dominate the economy tend to get the most efficient use of interest payments. As the economy approaches a liquidity trap, all the P(z) tend to allign in relative spectra, the people-property curve goes to normal, and paradoxically, the bank starts to own the economy. (New Libertarian accounting predicts Marxist Nirvana in the limit)

In probability based linear prediction, Brad's law of large numbers implies that only a small sample estimate of P(z) gives the the realistic number of terms in P(z), leaving the remaining error as white noise. The number of sample estimates to achieve the best estimate is the same as the sample size that Brad talks about.

The P(z) that is reported under current NASB rules is a single order estimate of the real P(z). (An average; rate of return). Most shrewd investors invest on inside knowledge that gives them a 2 or 3 order estimate
of P(z) which gives something like a parabolic estimate. Utilities in the past used to have very stable P(z); probably known to three terms. Social Security has a very stable P(z) (rest assured) but its sample period (16 years) implies it carrys a huge base noise relative to the P(z) that update yearly.

All P(z) approach a gaussian spectral function at infinity (closed system), and in Marxist Nirvana, the inputs to P(z) are white noise and the outputs are white noise.

For the middle class, this theory predicts they will work their asses off to match their P(z), in spectrum, to that of the big corporations, otherwise they get less efficient use of interest payments.

The min energy assumption also implies standard laws of supply and demand, so it is merely the frequency transform version of the complete economy as we know it.

Nth order estimates of stock P(z) are easily done over past samples, and investor do this in various forms.

Because we are linear predictive, and assume business process to be linear predictive, we often see volatility rise in anticipation of an under sampled, large P(z). So, an economist must look aty volatility as a predictor of events to come. Hence the need to operate in the frequency domain.

I'm tired or writing and you are tired of reading.

[Chris]

Slamming Luskin is easy. Offering a sweeping claim that Krugman is nearly always right isn't.

Consider his March 26 column.

BEGIN EXCERPT
[P]olling data on the issues, from Pew and elsewhere, suggest that the G.O.P.'s problems lie as much with its ideology as with one man's disastrous reign.

For the conservatives who run today's Republican Party are devoted, above all, to the proposition that government is always the problem, never the solution. For a while the American people seemed to agree; but lately they've concluded that sometimes government is the solution, after all, and they'd like to see more of it.

Consider, for example, the question of whether the government should provide fewer services in order to cut spending, or provide more services even if this requires higher spending. According to the American National Election Studies, in 1994, the year the Republicans began their 12-year control of Congress, those who favored smaller government had the edge, by 36 to 27. By 2004, however, those in favor of bigger government had a 43-to-20 lead.

END EXCERPT

Uh, Paul, for your thesis to be true, Republicans have to acted as if they truly believe government is always the problem. The exact opposite is true. Federal spending exploded from 2001-2006, when Republicans held both Congress and the White House. And, no, this wasn't just because of increased national security spending. As has been widely documented, we've been on a spending binge not seen since the days of LBJ, vastly increasing federal involvement in education, setting new records in the adoption of worthless earmarks and -- worst of all -- adding a staggeringly costly new entitlement in the prescription drug benefit.

Krugman is living in a fantasy world in which Republicans are parsimonious spenders. Krugman peddles this fantasy because it fits his facile thesis that Americans don't like this parsimony because they've become more liberal.

[mik]

"Since when are Mickey Kaus or Andrew Sullivan "right-wing?" You'd have to be a truly delusional left-wing academic to believe that."

But DeLong IS a delusional left-wing academic thug.

And he is sooo obese.
He should reduce his food consumption and donate to starving kids in Africa.