Sun Aug 24, 2003

Take the Money! Open the Box!

Brad DeLong wonders why Dan Weintraub seems least inclined to support the candidate for Governeror of California about whom most is known. On Dan’s own admission, McClintock and Simon are liars, Schwarzenneger is an unknown quantity and Bustamente has a known program that at least holds together. And yet Dan leans towards McClintock (whom he knows is lying) or Arnie (about whom he knows nothing). Brad says:

A normal person, if offered a choice between candidates (McClintock, Simon) who are lying to you, a candidate (Schwarzenegger) who refuses to say what he would do both because he has no clue and because he thinks “people do not care about the numbers and figures,” and a reasonably-smart guy who understands what the tradeoffs are and has a set of ideas about what to do with them—as I said, a normal guy would choose the clued-in candidate who is not lying to him.

But, as I said, Dan Weintraub is strange. The clued-in candidate who is not telling lies is to be avoided at all costs … Anyone have any idea why Dan Weintraub is such a strange guy?

Well, no. But there’s a neat experiment by Eldar Shafir that I want to tell you about, which may possibly be relevant. I think it appears in Redelmeier, Shafir and Aujla, “The beguiling pursuit of more information,” Medical Decision Making, (2001) 21(5): 376-81, but I’m not sure because Eldar told me about it himself a few years ago, just after he’d run the experiment at Princeton.

You take a bunch of sophomores and tell them they’re going to play the role of Admissions Officer to Princeton. They look at application files, and judge whether an applicant should be admitted. There are several criteria—let’s simplify and say there are only two: GPA and SAT scores. Applicants who score highly on both measures are strong candidates for admission. After having the experimental subjects rate enough files so you know what their standards are, you divide them into two groups and present the first (control) group with a applicant who has a 3.7 GPA (out of 4) and excellent SAT scores. Nearly all the subjects say this student should be admitted. The other (experimental) group gets the same file, except they’re told there seems to be an inconsistency in the application. The school transcript says the student’s GPA is 4.0, but the letter from the Principal says its 3.7. Do you admit, reject or ask for more information to resolve the inconsistency?

Most subjects ask for more information. It turns out that the Principal is right and the applicant’s true GPA is 3.7. With this information in hand nearly everyone votes—surprise!—to reject the applicant. The interpretation is that new information is weighted more heavily than it ought to be (in the light of evidence from previous decisions) simply because we have bothered to go and find it out, and not because it’s useful or ought to change our mind.

The link to Dan Weintraub is tenuous. We might say that Dan thinks he already knows all he needs to know about Cruz Bustamente, whereas he knows very little about Arnie. Therefore, the few bits of information he has about Arnie have been weighted far too heavily (and positively). This is consistent with the psychological mechanism identified in the experiment, but on reflection it doesn’t seem to be what’s happening in Dan’s case. Rather, a more severe version of the pathology may be at work. The mere fact of not knowing anything about Arnie makes him a more attractive candidate. Uncertainty about Schwarzenneger not only makes him look good, but it encourages people to discount what they already know about the other candidates. Better the devil you don’t know than the devil you do.

This state of mind is irrational but not uncommon. For instance, it’s widely known that someone who has been teaching in your department for a year or two is much less likely to be offered a tenure-track job than an external candidate, even when their records are basically the same. Familiarity breeds contempt. More severely, a faculty member once told me he would rather hire a candidate with excellent letters and no publications than one with excellent letters and a publication in a good journal, because the latter candidate had already shown what they could do whereas the former was still full of promise.

All of this may be beside the point, especially seeing as Weintraub is the kind of columnist who can write a sentence containing the phrase “turning California around from the ground up.” I’d like to see that happen.