First off, take an imaginary league table of surgeons. The further up or down surgeons are in that list, the more they get paid. The goal of the table is to make sure that the best surgeons are given the most difficult and dangerous jobs, and so, hopefully, to optimise the system. Brain surgeons at the top, and, er, earwax removers at the bottom.
How do you measure the surgeons? One way would be to score them on results: some fiendishly complicated, combination of number of malpractice suits, of numbers of ensuing complications, of “deaths at the hands of”, etc.
Sign off this wonder-system and see how the surgeons do…
Enter the brain surgeon, at the top of the table, with a scalpel as sharp as his mind. He has two types of surgery open to him: brain and non-brain. Brain surgery is high-risk and as well-paid as you can get, but why would anyone do it? High-risk means that there’s a high chance that you will be scored down, which means there’s a high chance that s/he’ll have fewer first choices on surgery, which means that it’s more likely that you’ll get given higher-risk jobs (because nobody really wants them).
Effectively, if the personal goal of the brain surgeon is to do the best surgery he can and at the same time get as well paid for it as possible, he will sub-optimise the system. In fact, won’t there be a whole talent scramble to avoid risk and so sub-optimise the system?