Comment on Hacker News.
The range of potential CEO productivity effect is much greater than that of most employees and I suspect is a multiplier effect on the whole company rather than a simple addition. The worst case is a negative effect on a scale that bankrupts the company. On the positive side if a great CEO gets 1% more productivity out of their staff than a good one that difference in a company employing thousands is worth paying upto 1% of the total salary budget for a great rather than good CEO (if that is what they cost in the market and you can identify them).
There are big issues about how you identify great CEO's but there are real reasons to pay massively for the best.
The other factor is that people (CEO's, traders, salesmen) who can directly point to profit/income that they are responsible for can more easily show their value and in many cases claim a portion of that rather than a wage based on more normal market competition.