Pay is not the independent variable here. I'm not arguing that if you double someone's pay, you'll double their performance. I'm saying that if there's zero correlation between pay and performance, and that pay to some extent predicts pre-pay performance, then one is forced to argue that well-paid CEOs are superior to poorly-paid CEOs, on average, but that they capture the benefit they create.

And that's not hard to believe. If someone had just a 10% chance of running Exxon 1% more profitably, their market value would be $40 million per year.