W. Edwards Deming on Performance Appraisal and Reward

Demingís assumptions about performance appraisal differ from the assumptions made by traditional performance appraisal theorists and researchers. Carson, Cardy & Dobbins (1991) have briefly articulated these assumptions as:

Traditional Performance Appraisal Assumptions:

  1. Employees differ significantly in their contribution to an organization
  2. The source of differential performance must be at at east partially due to the employee
  3. Raters must be able and willing to distinguish personal and environmental sources of performance variation and to base ratings only on performance within individual control

Deming Challenges the Above Assumptions:

  1. Individual employees do not differ significantly in their work performance; any observed differences are random and thus a sampling error phenomena
  2. Variation in performance is due predominantly to factors outside an individualís control
  3. Raterís are incapable of distinguishing between person-caused and system-caused variation in performance


In Demingís book Out of the Crisis (1986), he has the following to say about merit ratings and the rewards attached to them:

"Many companies in America have systems by which everyone in management or in research receives from his superiors a rating every year. Some government agencies have a similar system. Management by objectives leads to the same evil. Management by the numbers likewise. Management by fear would be a better name, someone in Germany suggested. The effect is devastating:

It nourishes short-term performance, annihilates long- term planning, builds fear, demolishes teamwork, nourishes rivalry and politics.

It leaves people bitter, crushed, bruised, battered, desolate, despondent, dejected, feeling inferior some even depressed, unfit for work for weeks after receipt of rating, unable to comprehend why they are inferior. It is unfair, as it ascribes to the people in a group differences that may be caused totally by the system that they work in.

The idea of a merit rating is alluring. The sound of the words captivates the imagination: pay for what you get; get what you pay for; motivate people to do their best, for their own good.

The effect is exactly opposite of what the words promise. Everyone propels himself forward, or tries to, for his own good, on his own life preserver. The organization is the loser.

Merit rating rewards people that do well in the system. It does not reward attempts to improve the system. Donít rock the boat.

Fair rating is impossible. A common fallacy is the supposition that it is possible to rate people; to put them in rank order of performance for next year, based on performance last year.

The performance of anybody is the result of a combination of many forces -- the person himself, the people he works with, the job, the material that he works on, his equipment, his management, his supervision, environmental conditions (noise, confusion, poor food in the companyís cafeteria). These forces will produce unbelievably large differences between people. In fact, apparent differences between people arise almost entirely from action of the system that they work in, not from the people themselves. A man not promoted is unable to understand why his performance is lower than someone elseís. No wonder: his performance rating was the result of a lottery. Unfortunately, he takes his rating seriously.

Outstanding performance may be attributed only to someone that by appropriate calculations* falls beyond the limits of variation of the system."

*Deming provides a simple example of these calculations through analysis of a hypothetical distribution of quantitative indicators of the performance of nine persons: 4, 10,10, 11,11, 12, 15, 17, 23. While it looks as though the persons evaluated as "4" and "23" should fall outside the limits of variation in the system, according to Deming they do not. Deming calculates the variation around the mean (12.55) as plus and minus three times the square root of the mean; this yeilds the following limits of variation: 1.9 and 23.2.