If you manage a large organisation, then people will come and go. There are always decisions to make about promoting people, promising newcomers versus experienced middle managers, all of whom are aspiring to move up the corporate ladder. Given two candidates, one very successful in their current job, the other less successful and erratic, most people would regard it as a no-brainer that you promote the first one in preference to the second. However, in 1969, the Canadian psychologist Laurence Peter introduced his famous Peter Principle, which states that every new member in a hierarchical organisation climbs the hierarchy until he/she reaches his/her level of maximum incompetence. Thus, the principle argues against the common sense approach.

Climbing up the corporate ladder.

One obvious result of the common sense promotion strategy is that everyone might be promoted upwards until they find themselves in a job they are not good at. They won't be promoted again and the whole organisation will tend towards having incompetents in every position! The overall effectiveness of the organisation then falls. That is what Peter meant when he argued that every new member of an organisation climbs upwards until they reach their level of maximum incompetence.

Peter's Principle is based on the assumption that there is little or no correlation between competences at different levels in an organisation. The opposite, or "common sense", view that we set against it assumes that there is. The validity of the common sense view is not so clear. There is a pattern of promotion in organisations where people who are good at doing certain things get promoted to positions in which they no longer do those things — they merely administrate other people doing them. Good class teachers don't necessarily become good head teachers, very successful footballers don't often become good managers, good academic researchers may not be good university vice-chancellors, good athletes may not be good coaches, and good medical students may not become good doctors.

Alessandro Pluchino, Andrea Rapisarda and Cesare Garofalo at the University of Catania have studied the results of different promotion policies using computer models of organisations with different levels for promotion. Strikingly, they found that the policy of always promoting the least competent person can maximise the overall competence of the organisation, while promoting the most competent candidate can significantly reduce it!

In the table we show the results of their computer experiments. The researchers compared the change in overall organisational effectiveness of three policies — promote the best, promote the worst, or promote at random — under the different assumptions that there is, or isn't, correlation between the abilities needed to perform your existing job and the one you are promoted to do, respectively.

We see that if that correlation doesn't exist between new and old jobs, as Peter assumes, then "promote the best" leads to a 10% fall in effectiveness across the organisation, while "promote the worst" leads to a 12% gain! Even if there is a correlation, "promote the best" only leads to a 9% gain. In the last column are the results of choosing at random between the best and the worst candidates. There are of course other possibilities like promoting the best for half of the positions and the worst for the others, or some other mix of the best, worst, and random strategies. All this is very worrying for chief executives of all sorts and quite contrary to their intuitions. The fall in overall effectiveness comes about because of people who are far less competent in their new jobs than they were in their old ones. The performance of the person doing a bad job, on the other hand, can only go up.

Of course, activities with human actors can behave unpredictably when new policies are introduced. If the worst candidates get promoted there will be a race to the bottom to convince your boss that you are the most incompetent and most deserving of promotion. This can be avoided by promoting at random but then there is not much incentive for anyone to work hard. Mathematical modelling can inform decisions like this but it can't take them for you.