If your best and brightest 10% are losing 30k, it's worth asking if the effects are worse (even if it's only 3k per average employee, assuming there are 80% of them, that's a lot of 3k), and if it's worth it for the benefit the least capable or experienced 10% will get.

And please, don't latch on to the mathematical definition of average. I'm talking about a roughly bell curved distribution of talent.