|
One of the most common errors made regarding performance appraisal rating systems occurs when the individual ratings are added to together, or an average (the mean) taken to describe overall performance. In fact, for statistical and mathmatical reasons, you cannot perform these functions on this information and get meaningful and useful results.
Why?
Here's the completely non-statistical explanation, stripped of jargon. First, you have to understand that numbers (1,2,3..) can be used in two different ways. They can be used to designate a specific and exact quantity of something (e.g. $27.10). The second use is as a label. A label is simply a shorthand description for something. For example, the number on a basketball player's uniform is a label. It has no quantity to it. A license plate on a car (whether it has only numbers or letters and numbers) is a label.
Now, think of a performance appraisal rating scale, which has a short description, and five points on a line. At the left hand side point is written "poor". Moving to the right it says "needs improvement", and in the middle, "adequate". Again moving to the right, there's "very good", and then at the far right, "excellent". The actual words can vary. Now, you're probably familiar with the ratings appraisal process where one of those points on the line is chosen to reflect (supposedly), the employee's performance.
Each of those phrases (poor, excellent) is a label, in fact a non-numeric label. It's clear you can't add them up, or multiply poor x very poor, or even take a mathematical average, because ther aren't any numbers there. There's been no measuring. And the labels do not represent quantity.
What some companies do is replace the text labels with numbers, or use both. So, poor is 1, and excellent is 5. Then they treat those numbers as if they are real quantities based on counting something. In fact, all they have done is replace text labels with numeric labels, that have identical mathematical qualities -- i.e. none. All they've done is make a superficial labelling change. So, here's the question. How come we can replace the word "poor" with the number "1" and then magically feel that we have made the rating into a counted "thing", a quantity?
So, whether a scale uses numbers or text labels, neither represents counting, or quantity. And in the same way as you can't average (take the mean) of poor and adequate, the averaging of numeric labels is meaningless.
In order to be able to add things together, or multiply them, or take an average (the mean) the numbers MUST represent a quantity. If they do not, you can't do it and end up with meaningful results.
Also, in order to be able to add, subtract, etc, the numbers must reflect a particular relationships amongst themselves. That is, a rating of 5 must mean that the person is five times better than the person that was rated as a "1". Because adding, subtracting, and using the mean demands that the distance between each rating point is equal. If not, you can't make sense of it.
Finally, rating scale results can be analysed and sorted, but to be meaningful, one must use the median as the way to determine the "average", and not the mean.
Is all this important in the real world? Yes. Let's say you want to use (incorrectly) the mean computed from rating scales to make personnel decisions. In fact, what you end up doing is making decisions based on numbers that have no actual meaning (although you can pretend). So, the result may be that you will make the wrong decisions, because the numbers are misleading, and may end up erroneously telling you one employee is "better" than another when he is not.
It would be very similar to deciding to buy a football team because the sum of the numbers on the backs of the uniforms is the highest in the league. In other words, making decisions based on mistaking labels for real quantities.
|
Performance Management & Appraisal 
