Last Updated Mar 8, 2010 9:05 PM EST
Two facts about this issue stand out. First, the relationship between pay and performance is astonishingly small. One meta-analysis found that firm performance accounted for less than 5% of the variation in CEO pay, while company size explained about 40% of the variation. Second, there is no evidence that attempts at reform, such as more disclosure or ensuring that the compensation committees of publicly traded companies are comprised solely of independent directors, has had any effect. The miniscule connection between executive pay and company performance occurs in other countries as well, so the problem is not something specific to certain companies, countries, or governance arrangements. It must run deeper than that.
I recently served on a compensation committee, and saw the source of the problem-and a solution-firsthand. The problem: nothing in the process of setting CEO compensation produces a pay-performance link. Companies and their compensation consultants choose a set of similar comparison firms, often the basis of size and industry, and then compute the median pay for CEOs of these comparable companies.
Setting aside the tendency to game the choice of comparisons, which is plentiful, what is most visible in the discussion of compensation is the median figure. Precious few companies take variables like company performance into account when determining compensation. So, what the compensation committee sees is not an equation relating pay to company-specific factors, but just the median. The tendency to pay at least at the median is overwhelming. And because the comparable firms have been picked partly on the basis of their similarity in size, it's not surprising that size looms large as a determinant of compensation. Moreover, because firms rely on comparables, if none of the comparison companies base pay on performance, neither will the focal company.
The solution is both straightforward and reasonably simple. Get rid of the idea of trying to find comparisons similar in size or other attributes-statistical techniques can take these differences into account. Draw a large sample of companies, and compute the effect of performance on pay for those organizations, measured in shareholder return or accounting measures and similar factors. To produce a suggested pay figure for a given company, plug that company's values for each factor into the equation derived from the large sample.
A procedure like this diminishes the focus on the median in the pay determination process and gives performance a role. Over time, as pressures for a performance component to pay determination persist, the effect of performance will become stronger, captured in the statistical estimation process and then transferred to the pay inside individual companies.
I heard some objections to this process, of course-that it removes "judgment" was one. But that judgment, heavily influenced by the fact that compensation consultants work for the very CEOs whose pay they are affecting, is far from unbiased. Also, because performance hasn't been used much to set pay, an equation derived from current statistics would likely show weak performance effects and won't solve the problem. That's true, but it would do a better job than the current procedures and provide some hope of introducing performance into the pay determination process over time.
The bottom line is this: the very process of pay determination virtually dictates that performance will have little to no effect on CEO pay. Unless that process gets changed, neither public protest nor regulatory reform will have much effect.