Superstar CEOs

SUPERSTAR CEOs….Eduardo Porter writes in the New York Times about today’s insane levels of executive pay and manages to include at least a dozen nuggets worthy of comment. But sort of at random, I’ll choose this one:

Processing reams of data, [Xavier Gabaix and Augustin Landier] estimated that hiring the most effective chief executive in the country would, statistically, increase the stock value of a company by only 0.016 percent, compared with hiring the 250th chief executive. But at a company like General Electric, which is worth about $380 billion, that tiny difference would amount to $60 million.

This, the economists argued, helps explain why that top chief executive earned five times as much as the 250th. “Substantial firm size leads to the economics of superstars, translating small differences in ability to very large deviations in pay,” the economists wrote.

Needless to say, this is ridiculous. For starters, a microscopic number like 0.016% is probably just a statistical artifact. But let’s say it isn’t. Let’s say it’s real. It still doesn’t matter because it’s still plainly too small to be a predictable difference.

So sure: if you interviewed 250 candidates, the top candidate might be 0.016% better than the 250th candidate. But which is which? With a difference that small it’s impossible to say. You have no idea beforehand which one of those candidates is going to earn you that extra $60 million. And since you don’t know, statistically you’d be better off choosing someone from the middle of the pack and saving yourself some money.

But that would bring the whole game to a screeching halt, and we can’t have that, can we? Much better to keep the rubes conned with stories about “superstar CEOs” and “more complex modern environments.” Look! It’s Halley’s comet!

UPDATE: A reader wrote this morning to say that he thought I was unfairly dismissive of Gabaix and Landier in this post. Unfortunately my email responses keep getting bounced back, so I can’t get a fix on exactly what the problem is. But let me clarify something anyway.

Basically, I’m accepting Gabaix and Landier’s results here. If their research says 0.016%, then for now I’m willing to assume that’s correct. However, if that result is right, I think it’s ridiculous to interpret it as explaining skyrocketing CEO pay. If you could systematically predict who was going to deliver a benefit that small, that would be one thing. In that case, maybe the higher pay would be justified. (Maybe.) But the microscopic size of the difference in measured CEO performance makes that vanishingly unlikely. CEO pay has certainly skyrocketed over the past couple of decades, but minuscule and unpredictable differences in performance don’t seem likely to explain why.