One of my graduate school professors was well known in our class for what he called his Behaviorists Bowling Team shirt. According to him, it says “You get what you measure” across the front, and “The best predictor of future performance is past performance” on the back. The second saying has only limited application (if it were completely true, there’d be no point in practicing something, as you’d never get any better.) However, the problem of getting what you measure is like the third law of thermodynamics: you can’t get out of the game.
Consider for example this story about police surveillance cameras being installed in Toledo. As the victim of three break-ins (one attempted), I think the cameras are a good idea to counteract the effects of fewer officers patrolling the streets. Data-driven policing, however, is another matter — search for “compstat” on this page and you’ll see what I mean. The short version: if the data show the frequency and severity of crimes going up, the pressure comes on to bring the numbers down. Unfortunately, there’s two ways of doing so: by arresting and prosecuting the criminals (hard), or by under-reporting crimes and/or intentionally mis-classifying them as less severe than they actually are (easy.) Guess what actually happens.