Many pundits, especially highly intelligent liberal pundits, often fall into the trap (fatal conceit?) of assuming that because they can't explain why the market would do something, the market must be wrong. But markets are almost infinitely subtle.
Does the union worker generate twice the marginal value of a non-union worker? Have CEO's become insanely more productive in the last 30 years? Do janitors for financial firms produce substantially more value than janitors at restaurants?
My answers are:
Yes, yes, and heck yes.
Let's start with the union workers. Even if nonunion workers are equally skilled, a firm will have an incentive to hire workers up to the point where the marginal revenue product of one extra worker is just equal to their wage (or marginal cost, if the firm has monopsony power.) In that case a union firm will hire fewer workers than otherwise, and use more capital. Thus even if the workers have identical talent, the union worker will be more productive at the margin. This point often confuses Americans who wonder how we could possibly compete with Chinese workers who seem just as productive but earn only 1/5 as much. (And how do Chinese workers compete with Bengali workers?)
CEO pay has been controversial for two reasons. It has risen very rapidly in recent years, and it often seems unlinked to performance. But pay is very closely linked to expected performance, which matters when contracts are signed. A few months ago Steve Ballmer resigned as CEO of Microsoft and the stock rose by billions of dollars. More recently, Larry Ellison (sort of) stepped aside from Oracle, and the stock plunged by billions of dollars. This shows that CEOs have a huge impact of stock valuations. Whether the market is rational in believing that is a trickier question, but it's the job of corporate boards to put people in place that will maximize shareholder value. That means they need to at least try to get the very best, even if it costs a lot of money in terms of higher salary. If they aren't paying obscene salaries then the board of directors isn't doing its job.
Back in the 1960s, corporate decisions were much easier. You allocated capital to new auto factories, steel mills, appliance makers, and churned out product for which you knew consumers were waiting. Even IBM was fairly predictable for a time. In contrast, a modern CEO at a high tech firm might find the company quickly destroyed by new technology if he doesn't keep on his or her toes. Think how much Sony would have benefited in the past 10 years if it had had the Samsung management team. Perhaps an extra $100 billion in shareholder wealth? And that's also why the finance sector is so much more important today, decisions over where to allocate capital are both more difficult and much more important.
[Another reason CEOs used to be paid less was the 90% marginal tax rates. And by the way, saying CEOs should get high salaries does not imply they should not face high MTRs, which is a very different issue.]
You might think that a secretary is a secretary and a janitor is a janitor. Not so, they vary quite a bit in competence. Goldman Sachs has much more to lose from an incompetent secretary than does a small accounting firm in Des Moines. Even the janitors at Goldman Sachs are more important. The executives there are very rich, and the visiting clients are very important. The cost of an overflowing toilet in the men's or lady's room is much greater (in dollar terms; lost consumption to employees or lost business to unimpressed clients) than at a McDonald's rest room. Thus Goldman Sachs should pay more, to ensure they get the best secretaries and the best janitors.