Cafe Classic: Numbers That Lie

Editor’s Note:  Today we bring you Jim Brennan’s classic cautionary tale for those of us prone to cling too tightly to our numbers at the expense of judgment, an understanding of context and even common sense.  Read on!

Hard numbers can turn out to be really soft and squishy upon closer examination.  Perhaps that is why predictive foresight is rarely fully accurate, even when based on the Best Available Data.  (Note the nifty acronym, proving that even an optimal decision is still doomed to be a BAD decision.)

Reliance on data1 does increase the certainty people feel when making their mistakes.  Numbers are therefore fondly embraced by most compensation people, because numeric data seem to confirm the accuracy of measurements.  But those very numbers can cause us to stumble because they promise a degree of precision that can be quite misleading, if not downright false. 

Last year, the initial Apple effect greatly skewed exec comp survey results (seeThe Apple Effect note at the bottom here).  Some frowned at the $376,000,000 in stock options granted to the new Apple CEO Tim Cook (which temporarily ballooned in value to over $600 million shortly thereafter) but little attention was given to his subsequent 99% pay cut this year… perhaps because it’s all RELATIVE.  Funny, how a relative increase gets headline attention but a relative decrease is ignored.  Surveys that did not make allowances for massive swingsin isolated observations like that would leave their followers spinning in confusion.

Maybe it all depends on whether you want to find what is normal (the median), discover the most popular (the mode) or simply compute the average.  Those arcane terms are really important in compensation work, because each has a different precise meaning

When Bill Gates enters a Starbucks, the average customer in the shop becomes a millionaire.   Averages are directly impacted by outliers, while medians are the middle figure no matter how high the high or how low the low observation.  Conventional society tends to favor AVERAGES even though they are more problematic in statistical terms.  There are all sorts of fancy tricks statisticians use to manipulate the figures they process into “hard numbers.”  They include winsorizingtrimminghot decking (where a missing value is replaced by a similar but different observation) and others .  Excuse me, but some of those gimmicks strike me as flat-out falsehoods; yet they are enshrined in statistical lore and thus may be applied without your knowing.

Likewise, anyone who finesses you with a claim that pay falls in a bell-shaped curve is yanking your chain (or pulling your thumb) in a big way. Wages are never symmetrical nor do salaries have a normal distribution.  That means some of the standard statistical techniques just don’t fit the world of total rewards where the observations are living, breathing human beings.  Whenever the science of number-crunching is invoked, the art of judgment must be added to assure its proper application.

These are just a few of the common compensation conundrums of lying numbers.  What are your favorites?

1Mark Twain – There is something fascinating about science. One gets such wholesale returns of conjecture out of such a trifling investment of fact. 

E. James (Jim) Brennan has spent over 40 years in HR corporate and consulting roles throughout the U.S. and Canada, he’s pretty much been there done that (articles, books, speeches, seminars, radio/TV, advisory posts, in-trial expert witness stuff, etc.), and will express his opinion on almost anything.

Creative Commons image “Bell Curve” by hardeep singh

This post originally appeared on Compensation Cafe
Author: Ann Bares