That I'm much more skeptical of, especially seeing how things have transformed recently. I think a lot of it boils down towards a principal/agent problem, where top executive ranks have been captured by an in-group that hires each other. This is true even in universities; when a university is looking for a President, the search committee, made up of high-up university execs, looks at other university execs. This leads to an artificially restricted supply pool as execs hire each other (often, literally, since the execs are on each others' boards). Sometimes the pay is even formalized; they want to pay "competitive market rate", often defined as something like 10% above average, which leads inevitably to an upward wage spiral, since by definition everyone can't pay 10% above average. When political backlash results, they just quote the policy.

I don't think it's necessarily conscious nefariousness, either, but more of a cultural thing: they really do think that their social peers are the only people who could possibly do the job, regardless of whether they have evidence for that, and they really do think that they and their peers are worth the money that they're voting themselves.

At the very least, the opposite hypothesis---that they earn their pay by providing shareholder value---has been extremely hard to show empirically. There is, if anything, a slight negative correlation between CEO pay and shareholder returns. Tellingly, at closely held companies where a family or relatively few people personally control the company rather than diffuse set of shareholders, CEO pay tends to be much lower, after adjusting for the company's size. The hypothesis of what to make of that correlation is that when owners are vigilant, as in closely held firms, they don't pay CEOs as much, because they don't believe that the CEOs aren't worth that much. When owners are asleep at the wheel, though, you get the HP case--- pissing away tens of millions on ineffective CEOs' golden parachutes.