  |
|||||
 |
 |
 |
 |
||
 |
|||||
 |
March 2008 | Issue 32
The "Aha!" Report What's the "AHA!" REPORT all about?
This series of newsletters contains AHA! information to help people and organizations hire the best employees, make the best promotion decisions, retain the most qualified people, maintain the widest applicant pool, follow best practices, and (if you are subject to US law) remain aware of EEOC hot-spots.
A Rogue Psychologist Among the Cubicle-dwellers In a recent edition of the comic strip Dilbert, the pointy-haired boss inflicted something called the Dogbert Personality Predictor Index on his team. The test was, of course, a no-win proposition for the hapless employees. A sample question, “How would other people describe you?” offered only three options “A) Angry loner; B) Embezzler; C) Lazy.” True, the aim of the test was to help management “leave you in the dead-end job that most closely matches your lack of potential.” Unfortunately, the Dogbert instrument shares a disturbing similarity with many real-world personality assessments: the test, and presumably the scoring, have no obvious predictive relationship to job performance. Which characteristic is more closely correlated with success as an engineer: “unidentified hominid” or “inappropriate toucher?” How about “INTP” or “ESTJ”? Get it? Personality Testing, Right or Wrong It’s true that engineering, legal, IT, or actuarial jobs largely require technical skills. It’s relatively easy to test for these – and rigorous professional credentialing often helps do the job for us. However, it also takes certain personality factors to make a good job fit. If being an “angry loner” isn’t a useful indicator of job performance (actually, it probably is, but I can think of no legally defensible way to measure it), how can we tell which personality factors are important? And how do we measure them? The Wrong Way(s) Well, there's a right way and a wrong way to link scores to future job performance. Unfortunately, the wrong way is the norm. It comes in two varieties:
Both varieties are filled with mistakes. The first is common among most organizations I know. The second is common among folks who might have taken a class in statistics, but skipped the class in measuring human performance. Either way, they tend to turn away good employees and hire bad employees. What is This Thing Called Performance? Some people think they can measure "performance" by looking at performance appraisals or supervisor ratings. However, we all know performance ratings are often primarily personal opinion shaped by office relationships and politics. As a result, most performance review ratings mean that an employee is either:
“Harder” performance data is often not much better. In general, any data that can only be evaluated at the end of a long performance period is error-prone. Sales volume, for example, is fuzzy because it's a function of persistence, fact-finding, learning, strategizing, presentations, adapting to buyer personalities, economic conditions, market conditions, and so forth. To use an analogy, if you want to measure the quality of grapes in a fruit salad, you cannot put all the fruits into a blender, press “liquefy,” and come back six months later with your grape-o-meter. Performance data must be easy to see and easy to measure. Big Nets Often Contain Big Holes Statistics is deaf, dumb, and blind. Only a human being can decide whether one factor causes the other. For example, organization or teamwork factors may correlate with performance, but if organization and teamwork don't cause performance, then hiring people based on these scores will reject qualified candidates and hire unqualified ones. How Many Factors? Take another example, with just one variable. I once visited a company where the HR guru only hired people who excelled at teamwork. After a few years, it had 300 employees who would not schedule a meeting unless everyone could attend, could make a decision unless everyone agreed, and dared not confront a production problem because it might hurt someone's feelings. Did they get the results they expected? No. They got the results they measured. University research shows that it takes about nine to ten factors to predict job performance. Six or seven are based on job fit and three are based on job attitude. It is probably a good idea to check your test vendor to see if the test you are using was developed based on this research. If not, it probably won't either predict the performance or attitudes you need. Statistics What about the size of correlation? Should we celebrate a correlation of .30? Does it mean we have a 30% relationship? No. It means we have a 9% relationship. The technical details can be used as a non-habit-forming remedy for insomnia, but in short, correlations have to be squared to learn how much performance they predict. People who do not understand statistics are easy to fool. Wrap Up Otherwise, your organization would probably do just as well to have an evil cartoon dog handle your pre-hire assessment program.
|
 |
|||
 |
|||||
Join Email List