December 28, 2009

The Pink Car Problem: When Does “Average” Not Exist?

Taking it easy this week and re-running some of the best entries recently posted on some of our other blogs. This first one comes from Fred Whittlesey of the Hay Group. It ran on’s “The Advisors’ Blog” earlier this month:

A company’s Compensation Committee decided to provide the CEO with a company car and asked what color car he wanted. The CEO wanted to ensure that his choice of color was consistent with market norms so he asked the HR department to research the car color of the CEOs of its ten peer companies.

The results were presented to the CEO, indicating that, on average, CEOs in the peer group drove a pink car. The CEO, who knew his peer CEOs well, commented that it was hard for him to believe that so many CEOs were driving pink cars (given the absence of any multi-level cosmetics marketing organizations in their peer group.) “Is that the average or the median?” he asked. “Both the average and the median are pink” was the reply. He asked to see the raw data which indicated that five of the CEOs drove a red car and the other five a white car.

Is this a silly parable? Would such a simplistic analytical shortcoming really occur? I recently spoke with the head of compensation for a technology company who was questioning the data for their peer group data cut from a well-known survey. The data indicated that equity grants at the executive level among the peer group companies were averaging a mix of 50% stock options and 50% RSUs. His anecdotal knowledge of the peer practices made him feel that this couldn’t possibly be correct. When the raw data was examined, however, it turned out that only two of their 20 peers had an options/RSU mix near the 50/50 average. Nine were granting all or almost all (80% to 100%) options and the other nine were granting all or almost all (80% to 100%) RSUs. The pink car problem.

At a time when more individuals and organizations than ever are collecting, analyzing, and opining on executive pay levels and practices, it is critical that the underlying data be collected, analyzed, and reported correctly. Could the CEO have identified the pink car problem without access to raw data? Of course. Simply looking at the 10th, 25th, 75th, and 90th percentiles would have told the story. And when n=10 (or any even number), after all, the median is a computed number in a spreadsheet. While “ratcheting” to the median is an oft-mentioned flaw with the benchmarking process, summary statistics can contribute to flawed decision-making even absent any such intent.

The three-legged analytical stool of collection, valuation, and reporting of data needs to receive more integrated attention. Accurate collection has been made more difficult over the past 18 months due to the prevalence of “special actions” taken during the economic crisis. Valuation continues to be a challenge as experts continue to disagree on how to measure pay. The reporting of pay reflects the turmoil in collection and valuation, sometimes exacerbated by the media. Compensation professionals cannot assume that push-button data provides the answer – that data only provides a starting point for questions.

In my next blog posting, I will provide an illustration of another variant of the Pink Car Problem which created a recent headline indicating that a CEO’s pay was “cut by about half.” (the perceived difference was not “about half” and pay was not “cut.”)

SEC’s IM Division Issues FAQs on Effective Date of New Rules

Last week, the SEC’s Division of Investment Management issued their own set of FAQs regarding the effective date of the new proxy disclosure enhancement rules as they apply to registered investment companies (Corp Fin had released their own set that apply to public operating companies).

– Broc Romanek