Teachers, doctors and pay for performance

In one paragraph, Gail Collins summarizes an important issue with basing teacher compensation on student performance:

Gail Collins - Teachers Always Show Up - NYTimes.com

... while it’s important to make teachers accountable, telling them their jobs could hinge on their students’ grades on one test is a terrible idea ... The women and men who go into teaching tend, as a group, to be both extremely dedicated and extremely risk-averse. The stability of their profession is a very important part of its draw. You do not want to make this an anything-can-happen occupation, unless you are prepared to compensate them like hedge fund traders...

The same is true of physicians by the way.

The real problems aren't simply incompatible personality traits however. The real problem is that these systems are dominated by Goodhart's law.

As Texas demonstrated many times, the easiest way to improve outcomes is to game the system. With both students and patients this is done by changing the denominator -- either by reclassification (change who takes tests) or by purging the problems (zero tolerance discipline) and filtering the candidates (programs appealing to elite students, wealthier families).

The same things happens in health insurance (risk assumption). Rescission is one way to change the denominator, another one is to promote (inexpensive!) alternative therapies that appeal primarily to health people. Putting the enrollment office on the second floor of an elevator free building is the classic approach to denominator bias.

These effective strategies don't have to be consciously applied. The "invisible hand" will reinvent them time and again.

There are many ways to improve the performance of teachers and physicians that reduce the appeal of system gaming solutions. They don't have the simplistic appeal of "pay for performance" however.

So we'll suffer through a lost decade before, once gain, letting these mistaken policies quietly die.