Thursday, February 09, 2006

I Heart Data

It’s time to start making decisions about which departments will get to hire new full-timers for the Fall. Per usual, the demand for new people exceeds the supply of funding, so the VP and I have to make some choices.

Listening to my chairs, I get a particular impression of where the needs are. Looking at the data, I get a very different impression.

Seized by inspiration (okay, impatience), I ran the course counts for the last two semesters, and manually counted off how many sections in each area were taught by adjuncts. Through high-level math, like addition and division, I found the percentages in each area. Suffice it to say, the percentages are a pretty bracing reality check.

The relationship between chair complaints and objective reality is almost random. One department is nearly all full-timers, and the chair of that one (wisely) doesn’t complain. Other than that, I could find no relationship between the truth and the complaining. The loudest (and most self-righteous) complainer is actually in the middle, slightly below the mean. The department in the worst shape complains a bit from time to time, but nothing like most others.

Maybe it’s the social scientist in me, but I’m fascinated by the mismatch between perception and reality.

Malcolm Gladwell has a pretty good essay in this week’s New Yorker about empirical studies of the costs of homelessness. He notes that most people think of homelessness as a chronic condition that can be ameliorated through broad provision of social services – soup kitchens, shelters, etc. Drawing on the work of some people who actually went into the trenches and did some counting, he found that the bulk of the social costs of homelessness (medical care, mostly) derives from a very small percentage of the homeless. Narrowly-targeted but dramatic interventions in a few hard cases produce greater results than broadly-provided but shallow instances of help, like soup kitchens. Gladwell notes that narrowly-targeted, dramatic interventions are hard to sell, since they’re morally counter-intuitive (i.e. they seem to be rewarding bad behavior). But from an objective cost perspective, they’re far more effective.

He describes it as the difference between a standard bell curve and something that looks like a hockey stick. Interventions that target the entire length of the curve won’t achieve very much, since most of the curve doesn’t matter much, and the part that does matter needs far stronger interventions than we would ever apply across the board.

I’m quite taken with this idea, since it seems to explain a persistent frustration of mine. In a tenured and unionized setting, interventions are damn near impossible unless they’re across-the-board. But the across-the-board interventions mostly just hassle the majority, who didn’t need it anyway, and are far too weak to affect the really awful few. The perception of what constitutes fairness pretty much forbids anything that might actually be effective.

Over time, various across-the-board policies pile up as the attempts to corral the few loose cannons keep failing.

What’s so lovely about data is that it doesn’t pull punches. It shows quite clearly where ‘reputational knowledge’ is simply wrong.

Now comes the hard part. How does one do surgical interventions in a process-oriented, tenured, unionized setting?