The latest Review of Economic Studies has a great article (ungated here) by Marko Tervio. I'll summarize.
CEOs, actors, directors, musicians, authors, and athletes make big bucks because:
Desired abilities are rare and lasting.
It is very expensive to try someone new.
Everyone can see which trials worked or not.
Winners are free to demand more money or walk.
Given these conditions, a few proven winners make big bucks, and few new folks get tried. After all, a new trial who wins will soon demand as much as other winners. Here winners avoid retirement to keep milking their gravy train, and small biases in weak signals on new guys to try can magnify into great social injustice.
Condition 4 is crucial. When long term deals are allowed, more folks are tried, because a few successes can pay for lots of failures. Folks being tried get paid more, and there are more better winners who retire earlier and are paid less even when free to walk. Distorted signals about who to try matter less. Such long term deal gains were realized, for example, in the US movie studio system of the 1920-40s, the old US American baseball club system, and even now via exclusive long-term music album deals.
Over the last century, however, legislatures and courts have consistently moved to limit and prohibit such long term contracts, thereby increasing inequality and decreasing productivity. France even forbids artists from selling the full value of their paintings. The key tipping factor here seems to me to be a public displeased by seeing gains by admired musicians, actors, athletes, artists etc. going to less admired others. The word "slavery" is often invoked. For example, music fans can be outraged to see their favorite musicians shackled to ungenerous album deals.
So our vast wage inequality of superstar CEOs, artists, athletes, etc. is caused not by a lack of sensible regulation to limit random cruelties of unfettered markets, but by a public preferring its heroes unshackled, even if those heros had preferred otherwise. Now maybe insuring heroes against financial variations imposes a negative externality on wider admiring publics, one large enough to justify preventing long term deals. But for now count me as skeptical; I'd rather allow CEO and other superhero "slavery," for their good and ours.
+10 karma, Robin. This is fascinating, concise, and above all useful.
This theory seems like it's being used overbroadly.
For CEOs, actors, and directors, their salary is really the least relevant cost factor. If you're running a large corporation, or making an expensive movie or television production, given the number of people involved, the risk is the whole project rather than the individual. Hiring an incompetent CEO (might) harm the entire company, causing losses that make his salary look trivial. Same thing goes for hiring a bad director or an actor for a major role. I doubt actors for minor or really minor roles suffer from this effect. For the big shots, their salary is not the determining cost, so "enslaving" them wouldn't change too much.
Authors and musicians carry signifant brand risk. That is, if I publish too many bad books or release too many bad CD's, my brand could suffer, and individuals who have some sense of their abilities (the model requires individual ignorance of talent, which is not wholly realistic) will look for other record labels / publishers. Consumers will be hesitant to purchase my brand. Thus, the main expense is not (necessarily) their salaries. I have some other specific ideas of where this fails, but I am not sufficiently familiar with the industries to voice them with confidence.
Athletes I know relatively little about the structure of. This may have a more significant effect there.
This is still an interesting and useful theory; I'm just not sure that binding contracts are the correct solution.