Applying Sports Performance Analytics to CEO Compensation

While many people complain about athletes being overpaid – and certainly the make a lot of money – at least in the case of baseball it would be hard to find too many professions where players’ net value was more carefully scrutinized with data. Given the tremendous amount of performance data and analysis, we would have to imagine that baseball players, at least on a relative basis, are fairly compensated – at least to the extent we can use their past performance to predict future performance.

One of my favorite concepts in baseball is VORP – Value over Replacement Player. The idea behind is it pretty simple and powerful. It asks the question, how much value are we getting from our current starting player compared to a “replacement” level player –e.g. minor leaguer at the same position. The point of the analysis is we should make our compensation decisions on the output gap between two players rather than the overall output of the players.

Ever since I have come across this concept, I have been trying to dream up ways you could apply it outside of sports. What if we could somehow measure the contribution a CEO’s or executive team made to an organization and then could compare that to the contribution of say a mid-sized company (500-1000 employees)? There were 100 CEOs with total annual compensation over $15 million in 2009/10. So my question is, if I took the top-100 managers I could find willing to work for $1 million/year, would I significantly degrade the average performance of these 100 companies? Does the VORP of these CEO’s justify an aggregate compensation of $2.25 billion (these top-100 earned $2.35 billion) over the $100 million I would pay the hundred best qualified “replacement level” managers?

I’m asking this question not because I know the answer, but because I don’t. It’s undoubtedly complex to manage corporations that are global, have billions of dollars of revenue and in some cases 100s of thousands of employees. It’s easy to argue that in a company with billions of dollars of revenue, making a mistake and underpaying a CEO to save money would be foolish and would make saving $15 or $20 million seem insane.

And that, in the end, is exactly why executive salaries are as high as they are. It’s the same principal that firms like Goldman use in the IPO market – who wants to get petty over a few million in a billion dollar transaction when the downside to not going with the best is so large? Many-a-house in the Hamptons were built on this principal, to be sure.

I would argue that low wage workers are subject to the same types of data driven scrutiny that baseball players are. The simpler the task, the more easily we can measure performance for it. Data-driven measures (customer satisfaction, absenteeism, productivity) are increasingly used on the low-end of the wage scale in promotion and firing decision, which directly impact compensation (going up upon promotion, going to zero upon firing). To be sure for line working and middle management positions, there is a huge supply of potential “replacement players,” to replace workers who don’t meet performance criteria.

What could be keeping the wages as high as they are at the CEO level is likely viewing the supply of those capable of being a Fortune-500 CEO very narrowly; Boards of Directors are typically very risk averse, and would be quite unlikely to hire a person who is not proven at the highest levels of corporate management. This assessment of the supply of capable executives, may or may not be correct.
What we would need to accurately assess a VORP for CEO’s are two things: an accurate assessment of top CEO’s current “value” to their company and a way to identify and predict the performance of the “replacement level CEO.”

I think to assess a CEO’s value the most broad measure we could construct is some measure of industry adjusted long-term total return to shareholder. This study here (http://hbr.org/2010/01/the-best-performing-ceos-in-the-world/ar/1) does exactly look at this measure of industry adjusted TRS. We could get one clue about market efficiency of compensation just by looking at excess TRS vs. total compensation in a scatter plot. If they seemed to have a strong positive correlation then we would know the market had some efficiency to it. That wouldn’t answer our question totally, but it would be a start. In fact, there have been a multitude of studies looking at CEO performance vs. compensation and there is only weak evidence for positive performance (see here for example http://www.gsb.stanford.edu/news/research/compensation_daines_ceopay.shtml). What I have yet to find is one that tries to calculate the excess pay being wasted – it would likely be an underestimate anyways since we haven’t even considered the wider market for “replacement level” CEOs.

To do a good job of understanding the potential to fill roles with replacement value players we could also find the industry adjusted performance of small company CEO’s (I bet company size and CEO compensation are quite correlated) and then see how much industry adjusted TRS our lower group is contributing vs. their compensation.

Of course then we would need “league adjustment factors” to understand how much performance is degraded by moving up from the “minor leagues” (e.g. small companies) to the majors (“large companies”).

So perhaps this is a start as to how to look at VORP for CEO’s and the efficiency of the compensation market. I’m not totally satisfied with this as a methodology because I think there could be some interim/more-bottom up measures that you might be able to collect that would be predictive of success/high industry adjusted TRS for a CEO that would allow us to better understand future potential of replacement level CEOs. I’m going to do a bit more digging and see what’s out there on correlating personal characteristics and actions taken by a CEO and their correlation with some measure of CEO success.