« Kenneth Rogoff: The Fed v the Financiers | Main | Nouriel Roubini: Bogeymen of Financial Capitalism »

Tuesday, September 04, 2007

Gordon and Dew-Becker: Unresolved Issues in the Rise of American Inequality

Whenever one discusses inequality and CEO pay, the results from a paper by Gabaix and Landier inevitably come up as rebuttal to the idea that CEO pay is unjustified in the sense of being disconnected from underlying economic fundamentals. Here's a summary of their work from Tyler Cowen (repeats part of this post):

A Contrarian Look at Whether U.S. Chief Executives Are Overpaid, by Tyler Cowen, Economic Scene, NY Times: ...Not surprisingly, many people think ... American executives are overpaid. ...

But in a new paper "Why Has C.E.O. Pay Increased So Much?," the economists Xavier Gabaix of the Massachusetts Institute of Technology and Augustin Landier of the Stern School of Business at New York University ... suggest that the higher salaries for chief executives can largely be explained by increases in the value of the stock market. Viewed as a whole, these salaries are a result of competitive pressures rather than the exploitation of shareholders.

Their core argument is simple. If we look at recent history, compensation for executives has risen with the market capitalization of the largest companies. For instance, from 1980 to 2003, the average value of the top 500 companies rose by a factor of six. Two commonly used indexes of chief executive compensation show close to a proportional sixfold matching increase (the correlation coefficients are 0.93 and 0.97, respectively; 1.0 would be a perfect match). ...

[T]he debate over chief executives' salaries has moved a step forward. Yes, there are numerous examples of corporate malfeasance. But it is not obvious that the American system of executive pay — taken as a whole — is excessive or broken. ...

But a new paper by Robert Gordon and Ian Dew-Becker ("Unresolved Issues  in the Rise of American Inequality") challenges these results:

6.3 The Conflict among Hypotheses ...[I]t is worth considering the simple equilibrium explanation of Gabaix and Landier (2006) that executive pay moves in proportion to market capitalization.

The Gabaix-Landier model is based on a set of theoretical distributions plus the assumption that the best CEOs manage the largest firms. Their stunning result is that the elasticity of CEO pay to aggregate stock market capitalization is predicted to be exactly unity. "The sixfold increase of CEO pay between 1980 and 2003 can be fully attributed to the six-fold increase in market capitalization of large US companies during that period." They also derive a superstar effect, in that "a very small dispersion in CEO talent . . . justifies large pay differences."

The first reason for skepticism is that the authors’ own data do not support their unitary elasticity. Their Figure 1 shows that over the interval 1970–2000, their preferred "JMW" index of executive compensation increased by a factor of 22 while market capitalization increased by a factor of roughly 8. ...

More troublingly, the results that drive the Gabaix-Landier hypothesis do not hold prior to 1970. Frydman and Saks (2007) study executive compensation going back to 1936. They run regressions similar to those of Gabaix and Landier to find the elasticity of CEO compensation to both own-firm and average-firm size. They ... reject the hypothesis that there is a unit elasticity between CEO pay and average firm size. They find an elasticity closer to 0.1 for most of their sample, both including and excluding the Great Depression and World War II. They also find that the Gabaix-Landier result is not robust to measuring firm size in terms of sales rather than market capitalization. When sales are used to proxy for firm size, the coefficient for 1976–2005 rises from 0.94 to 2.65, but the coefficient for 1946–1975 only inches up to 0.16

We also tried to replicate Gabaix and Landier’s result for 1970-2005. Rather than simply taking the entire period as one regression sample, we ran rolling 20-year regressions. The rolling regression is a technique that allows us to see the evolution of the estimate of the compensation-firm size elasticity over time. Figure 5 plots estimates from the rolling regressions. The standard errors are large due to the small sample sizes and the use of robust standard errors. However, it is clear that there is a large change in the point estimate over time. The estimated elasticity rises from approximately 0.5 to 1.5. So a unit elasticity does not seem to be an accurate description of CEO pay even for 1980-2005.


...[I]t is clear that the one thing we can be sure of is that the relationship between firm size and CEO pay has not been stable over time. ...

I chose this particular section because the Gabaix and Landier results have been cited frequently in rebuttal to the suggestion that CEO pay is excessive (e.g. here's a write-up from the WSJ), but there's a lot more to the Gordon and Dew-Becker paper and I hope to have a summary posted soon.

    Posted by on Tuesday, September 4, 2007 at 10:44 AM in Academic Papers, Economics, Income Distribution | Permalink  TrackBack (0)  Comments (23)


    TrackBack URL for this entry:

    Listed below are links to weblogs that reference Gordon and Dew-Becker: Unresolved Issues in the Rise of American Inequality:


    Feed You can follow this conversation by subscribing to the comment feed for this post.