Wednesday, February 25, 2009

Reference letters

Andrew Gelman offers some thoughts on how long to spend writing letters of reference.

I would add four bits to this:

1. First, it is important in economics to distinguish between letters to employers, operators of summer internship programs or graduate schools on behalf of undergraduates and job market letters for graduate students. Job market letters consume more time in part because the norm is to describe the student's research in some detail, in part because you usually know graduate students much better than undergraduates and so have more useful information to provide, and in part because you know that your colleagues are really going to pay attention to those letters, and you will hear about it if they interview or fly out someone based in part on your letter and the person crashes and burns.

2. Most students do not appreciate that letters are essentially all fixed cost. Once you write one, the rest are easy.

3. Many undergrads do not appreciate that if you have never interacted with them, there is not much you can say in the letter that is not obvious from the transcript and their resume. In these cases, I encourage the student to get a letter from someone who knows them better. If there is no such person (and may undergraduates never really interact with any faculty in a meaningful way outside of class, so this is often the case) then I try in the letter to give a sense of what goes on in my class, of how students select into my class, and of what the grade distribution is, so that they can better interpret the transcript (what does an "A-" mean) and the resume (how much Stata does this person really know).

4. To me, the costly part of writing letters has to do with the fact that the equilibrium is that everyone lies. All letters oversell. So one faces both a moral cost and a cost associated with trying to get the level of overselling just right, which is more difficult than just being honest. As Gelman notes, the latter issue is partially mitigated by the development of a reputation. My sense is that I have a reputation for writing informative and relatively honest letters. Once the reputation is there my student bear no cost from my relative honesty, but I suspect that some early students may have paid a price for my establishing this reputation.