Keep in mind that we are already juggling a few margins here, including "getting this CEO to work harder or better" and "this CEO vs. another."
Read the whole thing, as well as his previous post. Some remarks:
1. A corporate CEO's pay is determined by how well he or she manages two entities: the company; and the board. If you manage your board well, you can manage your company poorly and still make a lot of money. (Alex Tabarrok agrees.) Conversely, I thought that Leland Brendsel managed Freddie Mac pretty well, but several years after I left he got ousted by the board in an accounting "scandal" that consisted of holding down accounting earnings to more closely reflect economic earnings. His successor, Richard Syron, managed the board well, but he was a disaster for the company--and for the whole financial system.
2. I think that Tyler's reasons for skepticism about measuring the marginal product of a CEO are applicable to many, if not most, employees in a modern business. Recall the Garett Jones characterization of workers as producing organizational capital, not widgets. CEO's have a lot of influence over the organizational capital development of their firms. When you have an entity as big as a Fortune 500 company, slightly better approaches to developing organizational capital can make billion-dollar differences in value.
The concept of marginal product assumes a sort of mathematical continuity that just isn't there in the real world. The textbook representation of business decisions as solutions to calculus problems is grossly misleading. In an actual business, you don't have an equation for your demand curve. You don't have a cost function.
3. I think of CEOs as players in a tournament. Their performance reflects both skill and luck factors. Their pay reflects both their performance and still other luck factors.
4. I suspect that the relationship between pay and value for CEO's falls far short of the standards of cosmic justice. But I suspect that even coming up with those standards is far beyond anyone's capability.