Pur Autre Vie

I'm not wrong, I'm just an asshole

Saturday, March 03, 2007

I'm Caught in A Trap, I Can't Get Out

So I've been thinking a bit about Paul Krugman and his influence. In particular, I wonder what will happen if a Democrat wins in 2008. It might be a good time to retire his column in the New York Times and either take an administration position or concentrate on academic work. Alternatively, it would be interesting to see him shift back to what he did with his column in the late 90s, mostly analyzing the economy and defending markets.

It seems to me that he's caught in a classic pundit's trap. The short version is that he's established such a reputation as a Bush-critic that he's sunk into the background in terms of changing people's opinions. It will be notable if he ever praises Bush, but aside from that his columns will not change many minds (it doesn't help that people like me no longer read him because it's too inconvenient to circumvent TimesSelect).

The long version is captured in a thought experiment. Let's say you're going to face a number of yes-no choices about policy matters. Each of these issues has an ideological dimension, but it also has a "right" answer that will be evident after the fact. Now you have to choose an adviser to help guide you through the choices. Each adviser has some expertise, but isn't perfect at predicting outcomes. Imagine that each adviser looks to factual inputs and comes up with a point on a 0-1 scale. If the point is closer to 0, the estimated correct answer is no, and if the point is closer to 1 then the estimated correct answer is yes. Advisers can't (reliably) tell you the underlying number, they only tell you the yes-no outcome of their process. Some advisers have biases that make them more likely, other things equal, to prefer one choice over the other. In other words, in addition to the normal error that shifts estimates, they also include a built-in shift to the left or the right (you can think of the ideology any way you want, "left" and "right" need not correspond to political ideologies).

Now, do you want a neutral adviser or a biased one? The neutral adviser is on average correct, while a biased observer might on average give you bad advice. However, if you can identify the bias, you can ignore him when he conforms to that bias and listen to him when he plays against type and recommends something he's inclined not to support. The nature of the process is such that these recommendations are the ones you can have the most confidence in. Thus, depending on how much you value that confidence, and depending on the size of the errors in question, you might want a biased adviser rather than a neutral one.

Now I think my point about pundits becomes clear. Some pundits spend a lot of time trying to disguise their biases so that people see them as neutral advisers. Some pundits have obvious biases, and they only really make a splash when they say something out of character. This is the trap: the more loyal you are to your side, the less your advice matters except in the rare instances when you criticize your side. You can imagine a sort of time lag, as people figure out your bias, so that your effectiveness declines over time (let's hope this is happening to Fox News).

You can partially overcome this trap by being accurate in the first place. I trust Krugman on economics more than I trust comparable liberals who don't have his background. Still, the public probably can't tell Paul Krugman from Thomas Sowell in terms of economic expertise, and might not even be able to tell Krugman from the WSJ op-ed page. So I think Krugman has fallen into the trap.

Of course pundits aren't just advisers. Krugman provides arguments for liberals and sheds extra attention where it's needed. A Democratic administration may spring Krugman from the trap, in a sense, but maybe he'd be better off focusing his energy elsewhere.

0 Comments:

Post a Comment

<< Home