> The only time I have seen a correlation greater than 0 is where bonuses where directly linked to earnings or sales

Maybe I've been lucky to work for moderately reasonable companies, but I've never seen that.

For example in my current company, yearly raises are directly tied to performance rating. And while there is subjectivity in the rating, it is basically impossible to squeeze in a high performer rating for someone who doesn't do anything even if I wanted to (which I don't). It is also impossible to give a very low rating to someone that all peers see doing great work.

The correlation will never be a perfect 1.0, but it is closer to 1 than to 0. Low performers end up getting low to nothing in raises, and the high performers get a multiplier that increases their raises.