Thomas Noe, Ernest Butten professor of management studies at the University of Oxford, has an interesting and analytical look at the economics and arguments surrounding CEO pay, available in its entirety here:

The attack on CEO compensation comes from two directions. Some argue that CEO compensation is too high, exceeding the level CEOs would obtain in arm’s length transactions; others argue that CEO compensation is too low-powered, that is, too insensitive to firm performance.

The problem with the “too high” argument is that it is not easy to find an absolute threshold beyond which CEO compensation becomes unreasonable.
 
At a company such as Walt Disney Co. (NYSE: DIS), with around $3 billion in after-tax profits, a CEO capable of boosting profits by a modest 5% could—even if we capitalize earnings at a modest price-earnings multiple of 5—raise corporate value by $750 million. In this context, the “outrageous” salary earned by Disney’s ex-CEO, Michael Eisner, of around $100 million does not seem so outrageous.

• Because it is difficult for critics of CEO compensation to measure either the scope for CEO value creation or the diffusion of star talent across the population of CEOs, critics for the most part attack the level of CEO compensation through comparisons, either by comparing the CEO of today with the CEO of yesteryear or by making cross-country comparisons.

Lucian Bebchuk and Yaniv Grinstein (2005), for example, show that US CEOs’ compensation relative to corporate profits has grown substantially between 1993 and 2003. Martin J. Conyon and Kevin J. Murphy show that in 1997, CEO compensation in the US was more than double CEO compensation in the UK.

• At first glance, these observations seem to support the idea that, at least in the US, lowering CEO compensation would be in shareholders’ interest. However, upon further inspection, the case is less clear. Much depends on how compensation is measured. Carola Frydman and Raven Saks (2007) show that relative to assets, the increase in US CEOs’ compensation is very modest. Moreover, the increase over the 1990s compensated for a decrease in normalized compensation in the preceding 10 years.

• The indictment of CEO compensation that is based on “low performance sensitivity” is also less than air-tight. The argument is that “high-powered” CEO compensation, which deals out very high rewards on average but concentrates these rewards at the top end of firm performance, is the ideal way to incentivize managers. This rests on the assumption that a board’s problem of designing incentives for the CEO is qualitatively similar to the standard problem of inducing effort from an employee.
 
However, the CEO’s position relative to shareholders is fundamentally different from a worker’s position relative to his boss. The CEO has a huge information advantage over the board and enormous discretion. In this environment, CEOs need incentives to do the right thing, even when their firm is sailing through rough seas and very ambitious performance targets are out of sight.

• Moreover, a weak relation between the CEO’s current performance and current compensation does not imply a weak relation between long-run performance and the CEO’s long-run compensation.

In a dynamic world, rewards for performance need not be meted out at the same time as performance. My research with Rebello shows that optimal CEO compensation can lead to a very weak relation of current CEO pay and current firm performance and yet produce a very strong link between the long-term value of the CEO’s position and firm performance. The empirical research of John F. Boschen and Kimberly Smith (1994) shows that a weak current but strong long-term link is typical of US firms.