Matt Yglesias thinks it's hopeless to argue about whether American CEOs are overpaid:
Take CBS, which I write about in the column. They had almost $15 billion in revenue in 2011. So the value to the company of a CEO who can boost revenue 1 percent higher than a replacement-level CEO would to the company is about $150 million. So if you have a 50 percent confidence level that CEO Leslie Moonves is 1 percent better for the company than a replacement-level CEO, then you'd be justified in paying him as much as $75 million a year—making him "underpaid" with 2012 compensation of around $60 million.
As a practical matter, I agree. There are certain metrics that suggest CEOs are overpaid, and there are others that suggest they're just earning fair market value. We liberals can point to gigantic yachts and gold-plated toilet seats as ways of swaying public opinion on this subject, but we'll never win on empirical grounds.
Still, I want to use this as an opportunity bring up one of my longtime hobbyhorses. Ask yourself: What would we see if American companies were really paying their CEOs gigantic salaries based on a belief that a great CEO has a huge impact on earnings1 compared to the 2nd best CEO? Answer: we'd see genuine pay-for-performance packages. That is, Les Moonves's compensation would be set at a fairly modest level unless CBS performed objectively better than its peers, at which point his compensation would go up quickly.
Now ask yourself another question: how many Fortune 5000 CEOs are actually paid this way? Answer: not many. This suggests pretty strongly that neither companies nor CEOs are truly confident in their ability to do better than the 2nd best guy out there. And that in turn suggests that fat CEO pay packages are based on something else entirely.
1This is usually the metric companies actually care about. Or maybe stock price, or return on invested capital, or something like that. But probably not revenue.