The Statistical and Scientific Thinking blog, by John Dowd, has several interesting posts on the Performance of People [the broken link was removed]:
Why can’t performance be numerically rated and ranked? It can’t be defined operationally, it can’t be measured with any degree of precision, it can’t be separated from other effects, and it is destined to vary over time in any case. Any one of these factors present significant (if not insurmountable) problems itself. Combined the problems create an impossible barrier.
Performance of People III [the broken link was removed]:
This essentially ended the practice of raise administration as a ‘zero sum’ game. Many (if not most) companies make a total figure available for raises. That figure is then stated in terms of a percentage of total salaries. Each supervisor is instructed to average that percentage in administering raises among his or her employees. Thus, a given employee can only receive more than the average if another employee within the same supervisory unit receives less. This fosters competition within small units of the company. That is disastrous.