Here's more from the paper by Robert Gordon and Ian Dew-Becker, "Unresolved Issues in the Rise of American Inequality." This section addresses whether the principle agent model of CEO pay is valid, i.e. whether the best interests of boards of directors, CEOs, and shareholders are in alignment and finds "a substantial amount of evidence that the principal-agent setting cannot explain the salient facts about CEO pay." This important because, according to this evidence, the manner in which CEO pay is set is inconsistent with the best interests of shareholders:
6.4 Firm‐Level Models of CEO Pay The Gabaix and Landier model can be thought of as a general equilibrium model of CEO pay. It involves firms that are differentiated only by size and does not consider any sort of bargaining. However, there is an extensive literature that studies exactly how CEO compensation is set, in particular, the interactions between CEOs, boards of directors, and shareholders (see Murphy (1999), for a survey of this literature). The classic principal-agent model treats directors as chosen by shareholders, and then studies the optimal contracts they set up for CEOs. This framework requires that boards have only the shareholders' interests in mind. As we have discussed above, this assumption seems at best implausible.
Bebchuk and Fried (2004) provide a substantial amount of evidence that the principal-agent setting cannot explain the salient facts about CEO pay. They propose an alternative model in which CEOs have control over boards of directors and are mainly restricted by an "outrage constraint" where shareholders retaliate if they perceive executive compensation to be excessive. Weisbach's (2007) review notes a number of especially convincing pieces of evidence that the Bebchuk-Fried model is superior to the principal-agent model. Specifically, they provide evidence that CEO contracts are far from optimal, that CEOs control directors, and that directors put substantial effort into disguising the size of CEO compensation packages. Their proposal obviously has been met with some criticism, notably in the Chicago Law Review with articles by Bebchuk, Fried, Walker (2002), and Murphy (2002).
The key assumption that directors are independent turns out to be highly questionable. To start, their pay is far from negligible-an average of $152,000 per year in the top 200 firms. While directors also usually own stock in the companies they oversee, presumably the amount they stand to gain from good governance is smaller than the salary they would lose if they were not renominated. Moreover, directors receive substantial non-salary benefits in the form of perks, or in business directed to their own firms. Also, as we have noted before, if a CEO is also on the boards of any of his directors, there are ample opportunities for tit-for-tat relationships.
Bebchuk and Fried also provide compelling evidence that CEO contracts are in no sense optimal.
They begin with the example of stock options. First off, CEOs often exercise and sell their options as soon as they are vested, long before they leave the firm, entirely undoing the incentive effects. Second, adept use of derivatives, known as "collars", allows CEOs to undo the incentive effect of their options and lock in a set value. Third, options are not indexed to the market as a whole. This means that when the market goes up, CEOs earn more even if their company performs below average. On the other hand, when the market goes down, boards of directors reindex out of the money options in order to insure that they still have an incentive effect. Bebchuk and Fried argue that this asymmetry is simply a way of claiming that high pay is actually incentive pay. If CEO contracts were optimal, stock options would be indexed to the market instead of rewarding them for factors completely out of their control.
Finally, Bebchuk and Fried provide ample evidence that firms work to disguise the magnitude of CEO pay. If contracts are optimal, there is no reason to try to hide what CEOs earn. On the other hand, if there is an outrage constraint, then firms have every reason to hide what their CEOs earn. This type of camouflage is available because the financial press generally only quotes annual compensation, ignoring deferred compensation and benefits. Executives often receive lifetime healthcare, company vehicles, offices, and consulting contracts. Firms often also give loans to their CEOs at well below market rates. Similarly, they will pay interest far above market rates when CEOs defer their compensation. It is difficult to imagine that these sorts of compensation could actually be more efficient than simple cash payouts. What is clear, however, is that they are very effective ways of reducing the visibility of compensation. ...
Weisbach (2007) questions whether there is any way to improve CEO compensation. He notes that the corporation has been a stunningly successful organization, and that even going back Adam Smith there have been complaints similar to those of Bebchuk and Fried. We take the view that sunlight is the best disinfectant. Simply airing these sorts of problems makes them more salient in shareholders minds, as we have seen in the past few years in the popular press. The fact that there were complaints 230 years ago about executive compensation does not mean that shareholders today should ignore what their CEOs are paid. Rather, research like that of Gabaix and Landier, Hall and Liebman, and Bebchuk and Fried is important because it tells shareholders what to expect and where their outrage constraint should be set. If we were to take a very strong view of the efficient markets hypothesis, we might think that shareholders are already taking into account every available piece of information on CEO pay. Given the ample evidence that in fact shareholders do not take into account all available information, we believe that the simple act of making certain irregularities and disparities more salient can improve outcomes.
. This section is currently based entirely on Weisbach's 2007 review of Bebchuk and Fried. This will be rectified in future versions.