Monday, December 8, 2014

Economics and Ideology, Terribly Muddled

With great resolve and all the best intentions, I began reading “Political Language in Economics” by Zubin Jelveh, Bruce Kogut, and Suresh Naidu.  This is the research version of the journalism version published by 538, both referenced by Tyler Cowen.

But I stopped at p. 8, wondering how it is possible for three obviously intelligent people to jointly pen such an arbitrary, implausible and internally inconsistent theory of the role of ideology in economic research.

Here’s what they assume:

1. Personal ideology is a scalar on a left-right continuum.  This corresponds to the direction and degree of political partisanship in the two-party electoral system of the US.

2. Economists choose the ideological content of their research to maximize their utility.

3. Their personal ideology is pre-given and unaffected by the results of their research (or anyone else’s).

4. Their professional identity, from which they also get utility, depends on how close their research is to the political center.  This would show they are neutral technocrats.

5. Thus the first choice they have to make is how fully to reflect their personal ideology versus their professional identity in their research.  This is summarized in a parameter.

6. But to achieve their career goals their research needs to be published, and they have to be hired and promoted by their academic departments.  This interest depends on the distance between their chosen research ideology and the ideology of editors, senior professors in their department, etc.  The tradeoff between “personal” (ideology and identity) and “market” (conformism) interests is summed up in another parameter.

All of these assertions are pre-empirical, and no argument is made for any of them.  They are simply assumed.  In fairness to the authors, they probably didn't think this part of their paper was very important.  It was just a standard-issue u-max model inserted to “ground” the empirical work in.....something.  It was their measurement and data analysis that they expected to garner attention, and so they have.

But the theory still matters, because you can’t separate conception and measurement.  For instance, the first assumption, that political ideology is a scalar rather than a vector, is at the heart of their empirical methods.  If they’re wrong about the first, the second doesn't hold up.  And so on.

My view is that assumptions 1-5 are indefensible.  Only #6 is plausible to me, but even here I suspect they are grossly oversimplifying.  Take the role of journal editors.  Editors will vary in the extent to which they will favor submissions based on their personal ideology, for at least two reasons.  First, some ideological positions are more supportive of suppressing alternative views than others.  (I make no presumption as to what those positions are, but you might have a hunch.)  Second, editors too face constraints, and imposing some ideological biases will be less costly for them than others.  Imagine a committed egalitarian who is the editor for a finance journal.  An article is submitted which argues for higher CEO salaries.  The editor would like to discriminate against it but worries about repercussions if a pattern of such discrimination is observed.  Now imagine a committed “market-rewardist” finance editor irritated by a submission that argues for narrower pay differentials.  The symmetrical constraint would be that the editor might worry that his or her finance professor peers may detect too much bias in favor of market-driven pay outcomes.  But are these two scenarios likely to be symmetrical?

Finally, the claim that professional identity is best served by centrism is baldly ideological.  There is an amusing contretemps on p. 8.  The main text speaks of a “preference for being neutral or centrist”, which seems to imply that being in the political center means you are neutral in the sense of unbiased.  Attached to this is a footnote, however, which reads in its entirety, “We do not interpret “centrist" as “unbiased" or more accurate, however, as being non-partisan or centrist could in fact be another form of bias.”  Well, yes, but that’s exactly what they do, since they identify centrism with professional self-respect and “being non-partisan experts”.  Strangely, they add that researchers might derive utility from centrism because it makes them “difficult to pigeonhole politically”, as if centrists don’t sit in a little cognitive box like most everyone else.  My guess: a reviewer made them add that footnote, although the authors were too inside their own centrism-seeking framework to pursue its consequences.

Like I said, I stopped here, and the good stuff is apparently later on.  According to Cowen, however, the authors ultimately suggest “re-centering” research to replace published results by what the results would be, according to the empirical model, if the researchers were perfectly centrist.  Personally, I’m going to adjust their findings to approximate what they would have been if Jelveh, Kogut and Naidu were fire-breathing radicals, in favor of maximum political and economic equality, democracy and freedom.  No bias in that.

UPDATE: Kevin Drum goes after the "good stuff" here.

No comments: