Warren Buffett is a famous critic of executive compensation, which he suggests is, as John Kenneth Galbraith once put it, "Frequently in the nature of a warm personal gesture from the CEO to himself."  A shareholder from the Phillipines (people travel a long way to ask Warren and Charlie questions) asks him how he thinks one should structure a good compensation package for managers at a capital-intensive subsidiary.  The implication is that if the banks hadn't had such poorly structured compensation, they wouldn't have taken such outsized risks.

The responses from Warren Buffet and Charlie Munger are worth repeating at length.  They don't actually explain whatever magic metric Berkshire Hathaway has found for aligning the incentives of managers and shareholders, but they're a nice statement of their views on the Principal-Agent problem:

 We think we have a good system.

Your question implies that the board sets these things.  In the recent forty years, basically the board has had little effect on these things.  The CEO has had an important role determining their compensation.  These people pick their own compensation committee.  I've been on one compensation committee out of nineteen boards because these people aren't looking for Dobermans; they're looking for cocker spaniels.  It's been a system that the Ceo has dominated.  In my experience, boards have done little in the way of thinking through as an owner what they ought to pay these people. 

Here in town, Pete Kiewit figured out a very logical way to pay people in his business.  It's not rocket science--you would be able to figure it out, I can figure it out, but you have to understand that not every CEO wants a rational compensation committee.

I don't think there should be a compensation committee. . .

It can be done.  It's very difficult to have a system where the board, thinking as owners, care as much as the guy on the other side of the negotiating table.  But it's very important how you compensate the CEO, and it can be done.

Charlie Munger added:

Liberally paid boards of directors can be counterproductive.  There's a sort of reciprocation--you keep raising me, and I keep raising you, and it's very clublike.

This brought on applause from the value-minded audience.

But as Ed Carr argued in the pages of my former employer a few years ago, there's a problem with the "cosy boards" story.  It's widely agreed upon by almost everyone, including yours truly until Carr got to me.  Here's the problem, though:  private companies pay their managers even more.  It's riskier compensation, of course, and it's plausible to argue that all of the extra pay is a risk premium.  But if CEOs were only getting paid so much because they'd captured the boards, you'd expect to see private companies paying less.  You'd also expect to see incumbents making more than CEOs recruited from outside, but you don't.

It's hard--nay, impossible--to believe that cosy board relationships don't inflate CEO pay.  But the effect doesn't actually seem to be that large.