Friday, December 10, 2010

Huh ? Ryan Avent is stumped by tricky math. He doesn't seem to be able to handle complex calculations which involve more than two numbers as in E = A - B - C + D. Either that or he is one of the many people driven to shrill unholly innummeracy by Paul Krugman.

I surfed over to this horror honestly thinking that Krugman must have been unreasonably hard on Avent. I was appalled that anyone who could make the two gross gigantic errors that Avent made could be published by The Economist. I will not elide anything I leave off Avent's final argument (the year before the election includes December 2011 and most of November 2011) and conclusion just to avoid being sued, but I don't skip any of his text in order to make his argument seem more stupid than it really is. I don't think I am personally capable of generating a more idiotic argument if I tried.

Ryan Avent can't handle the concept of a difference in differences. He can't recognise the idea. Arithmetic inolving more than two numbers appears to be totally outside of his ken.

Dec 9th 2010, 16:02 by R.A. | WASHINGTON
THIS is one thing Paul Krugman has gotten very right about the dynamics of recent economic policymaking:

[T]he history of the past two years drives home, if anyone doubted it, that economic policy must be considered from a political economy point of view; that you have to think ahead to how current policies affect the environment in which future policies will be decided.
But then there's this:

Put these two observations together — and what you get is that the tax-cut deal makes Obama’s reelection less likely. Let me repeat: the tax cut deal makes Obama less likely to win in 2012.

One gets the feeling from reading these lines that it's the point Mr Krugman wanted to make before he ever started looking at the data. That would help explain why he, rather sloppily, makes a big analytical error.

Note how high Avent sets the stakes. He doesn't say that he disagrees with Krugman. He "gets the feeling" that Krugman didn't approach the issue rationally. Because of his inability to handle simple arithmetic, he chooses to accuse Krugman of intellectual dishonesty. I will link what Krugman actually said with what Avent idiotically thinks he wrote with asterixes, double asterixes etc. Emphasis mine.

The observations in the above quote are as follows:

Look at the Zandi estimates: they show a boost to the economy in 2011, which is then given back in 2012. So growth is actually slower* in 2012 than it would be without the deal**.

Now, what we know from lots of political economy research — Larry Bartels is my guru on this — is that presidential elections depend, not on the state *** of the economy, but on whether [Krugman's error here*****] things are getting better or worse in the year or so before the election. The unemployment rate in October 1984 was almost the same as the rate in October 1980**** — but Carter was thrown out by voters who saw things getting worse, while for Reagan it was morning in America.

Let's do look at the Zandi estimates. Mr Krugman is right that the deal provides a big boost to output in 2011 and then a drag to growth in 2012. But growth is still strongly positive* in 2012, according to Mr Zandi's estimates. Things won't be getting worse** in 2012. They'll be getting better, to the tune of 3.4% real GDP growth. And they'll be*** a lot better than would otherwise be the case. Absent the deal, says Mr Zandi, the level of employment in 2012 would be significantly lower than with it, and the unemployment rate would be around 8.7%, as opposed to 8.4%****.

*Avent attempts to refute a claim about the difference between two predicted growth rates (with and without the tax cuts) by reporting only the growth rate with the tax cut.

** Again Avent refers to the rate of growth with the tax cuts not the difference in rates of growth with and without the tax cuts.

*** "They'll be" is a statement about levels. Krugman's whole point is that, according to Bartels, elections depend on recent trends not levels.

**** Avent attempts to refute a claim about the effect of the tax cuts on the rate of improvement from 2011 to 2012 of the unemployment rate by reporting Zandi's estimates about the effect on the level of unemploment in 2012.

***** Krugman may have confused things by trying to make them simple. As written, he asserts that only the sign of GDP growth or the change in unemployment matters. Ooops. This helps explain Avent's weirdness marked **. English is not suited to writing about numbers. "whether things are getting better or worse" should be "whether things are getting better and, if so, how fast they are getting better or things are geting worse worse and, if so, how fast they are getting worse." It is clear what he means from "growth is actually slower in 2012 than it would be without the deal." I think it is clear what he means anyway.

Krugman says elections depend on the rate of change of GNP and of the unemployment rate in the year or so before the election. I will try to make things simple enough that even Ryan Avent can understand by talking first about GNP. Krugman says that the election depends (very roughly) on log(GNP2012)-Log(GNP2011) so the effect of the tax cut on the election depends (very very roughly) on
logGNPwithtaxcut2012)-Log(GNPwithtaxcut2011)-log(GNPwithouttaxcut2012) +
log(GNPwithouttaxcut2011). He notes that this quantity is negative and concludes that the tax cut will hurt Obama's chances in 2012.

Avent thinks he can refute a claim about the difference between two differences, a calculation made of four numbers, by presenting claims about two numbers at a time.

So he discusses the sign of logGNPwithtaxcut2012)-Log(GNP2011withtaxcut) which has nothing to do with Krugman's calculation. If Avent were to argue that Obama will be OK even though the tax cut will hurt his chances, then the calculation would be relevant. But he claims he can refute a claim about the effect of the tax cut on something whithout any consideration at all of what would happen without the tax cut.

This is profound innumeracy. It is also plain idiocy. I would have thought that people who can't stand words like "difference" or "sum" or "rate of growth" could understand that statements about causation are statements about how something changes something else.

Then Avent makes an equally gross error when discussing unemployment. Krugman claims that the variable of interest is

Unemwithtaxcut2012-unemwithtaxcut2011-unemwithouttaxcut2012) +
unemwithouttaxcut2011). Avent discusses unemwithtaxcut2012-unemwithtaxcut2011.

This is even more extremely idiotic as Krugman specifically noted that the level of unemployment in 1980 and 1984 was similar, but the electoral experience of Reagan and Carter were very different. Avent has completey forgotten Krugman's claim about political science *and* the overwhelmingly dramatic example which supports the conclusion of political scientists. He isn't talking about changes in unemployment at all but about the level of unemployment.

The appalling thing is that I suspect that Avent is sincere. I fear the following might have happened.

Avent may have decided that Krugman is being dishonest and has resorted to special pleading. The reason would be that Krugman might appear to Avent to have chosen not to peform a simple natural calculation, but rather to have found a strange bizarre calculation that appears to support his conclusion. Avent's idea might be that fancy intimidating mathematics can be nonsense, since the poor reader can't understand what is going on and can't see it is invalid. I think that Avent thinks (as I do) that honest people don't obscure things with incomprehensible math. But I fear that Avent considers taking a difference in differences to be incomprehensible math.

No comments: