Evaluations and zero sum games

This is Joseph

In a discussion of education reform pros and cons, Curmudgucation notes

I'm giving Kraft a bonus point for this one, because too many reformsters refuse to acknowledge that their evaluation systems set up a kind of teacher thunderdome, a system in which I can't collaborate with a colleague because I might just collaborate myself out of a raise or a job. Because a school doesn't make a profit, all teacher merit pay systems must be zero sum, which means in order for you to win, I must lose. This does not build collegiality in a building.

This is a good point and a general problem with "stack ranking" style systems.  They often work well when first deployed, because nobody has actually adapted to them.  But they quickly insert perverse incentives.

Imagine a system that said "the bottom 10% of employees need to be let go each year".  Used for the first time it would remove a lot of dead wood (and maybe some good people as well).  But people would quickly note some obvious downsides -- like who is the comparison group.  If you get promoted could you be in the bottom 10% of the next rank up and thus be promoted into being fired?
Does it not make it rational to hire the weakest person you can sneak through the hiring system?  After all, if cash bonuses and job retention are based on relative ranking (and not overall performance) then is it not best that the competition be as weak as possible?

I wonder how much this counter-balances quality effects due to ranking systems?  Especially given that churn, itself, is costly.