08 July 2009

Merit Promotion From Within Fails

In 1969, a Canadian psychologist named Laurence Peter encapsulated this behavior in a rule that has since become known as Peter's Principle. Here it is:

"All new members in a hierarchical organization climb the hierarchy until they reach their level of maximum incompetence." . . .

[C]ommon sense tells us that a member who is competent at a given level will also be competent at a higher level of the hierarchy. So it may well seem a good idea to promote such an individual to the next level.

The problem is that common sense often fools us. It's not so hard to see that a new position in an organization requires different skills, so the competent performance of one task may not correlate well with the ability to perform another task well.

Peter pointed out that in large organizations where these practices are used, it is inevitable that individuals will be promoted until they reach their level of maximum incompetence. The unavoidable result is the runaway spread of incompetence throughout an organization. . . .

[Researchers] simulated this practice with an agent-based model for the first time. . . they find that it leads to a significant reduction in the efficiency of an organization, as incompetency spreads through it. . . .

But is there a better way of choosing individuals for promotion? . . . [The] model shows that two other strategies outperform the conventional method of promotion.

The first is to alternately promote first the most competent and then the least competent individuals. And the second is to promote individuals at random. Both of these methods improve, or at least do not diminish, the efficiency of an organization.

From here.

Not promoting from within is actually fairly common.

The military, for example, drops from one pool for its enlisted ranks, another for its officer corps, and others to fill the ranks of its civil service and political appointee civilian bosses. Traditionally, the Secretary of Defense is not appointed from the ranks of current active duty military personnel. More officers come from war colleges or ROTC, than officer candidate school for enlisted soldiers. The military's elite special forces recently began a campaign to encourage non-traditional candidates to apply.

Similarly, hospitals have highly stratified labor pools in which promotion from one job to another higher tier job, e.g. from nurse to physician, is rare. This is true even within professions. Few primary care physicians, no matter how successful, are promoted to the higher paying and often more prestigious ranks of specialist physicians.

In education, elementary school teachers are rarely promoted to teach high school students, and high school teachers are rarely promoted to teacher in higher educational institutions.

In law, paralegals are generally not promoted to become lawyers, no matter how competent they may be.

Most businesses draw their management personnel and their rank and file workers from different labor pools. Academic study of business management and large organization management suggests that direct management of people, and indirect management of groups of people through their managers involve different skills and methods.

This kind of class stratification of employment prospects was the main response that Peter himself advocated.

Random promotion, while not common, is also not as unusual as one might expect in civic organizations. Family reunions, Rotary clubs, book clubs, and the Bahai religious denomination often employ either random promotion, or the similar technique of annual leadership rotation regardless of performance from within a small membership. My son's elementary school class rotated the job of "class president" and many other elementary school tasks are rotated in the same way. Juries are lead by foremen and women selected from randomly drawn pools, and elected from within their own ranks despite the fact that newly seated juries barely know each other or their qualifications. Contested elections for low profile political posts (e.g. state treasurer or local judge) are similarly nearly random if media coverage is only slight, as is often the case.


Michael Malak said...

Some comments:

1. The original paper seems to have disappered from the Cornell University archive, arxiv.org

2. This "experiment" was done with "agents", which means computer simulation. The assumptions that went into the computer simulation control everything. E.g., organizations today that promote from within are self-conscious about the Peter Principle and probably have devised means to determine competency and/or provide training for higher positions. So a crucial question was whether the computer simulation modeled parallel competencies: for current job, and for potential job (where the competency for the potential job is somewhat unknown and should be modeled by a random variable)? This cannot be known without access to the original paper.

3. In my field of software development, the Peter Principle comes into effect just as much or more when a technical person in the trenches moves from an old technology to a new technology than it does when such a person is identified as having "soft" (i.e. social) skills and moves from the trenches to management. Then once in management, technical skills fall even further behind, and the trench developers lampoon the manager. Said manager then often moves into a non-software field or becomes a consultant in the old technology. But again, the interesting thing is the Peter Principle in effect for within the same position of trench developer, and that is due to the rapid pace of technology improvement in the field -- involving complete and deep changes in philosophy as new and different ways to conceptualize larger, more complex, more robust/redundant, more well-tested software systems are invented.

bikebali said...
This comment has been removed by a blog administrator.
Andrew Oh-Willeke said...

arxiv.org is just being cranky on a technical level at the moment. None of the recent articles (i.e. pre-October 31, 2008) on anything are coming up in their search engine, even though an author search returns the correct citation.