, , , , , , , , ,

From Dilbert’s PHB to The Office, the incompetent manager is such a popular trope that it’s in danger of becoming a tired cliché. Is this just a sign of the times or might it reflect some deeper truth? Could this be because of how we decide who to promote? A recent study has confirmed a counter-intuitive idea proposed over 30 years ago — that promoting those who perform best may, in fact, be a losing strategy.

Promoting someone who has done her job well would seem to be a pretty reasonable way to run an organisation. Perhaps not, acording to Canadian psychologist Laurence J Peter.  In 1969, he proposed the “Peter principle”, claiming that this seemingly sensible and fair approach actually results in each person being promoted until she reaches her level of “maximal incompetence”.  The reasoning behind this is relatively straightforward: an employee who does her job will will be regularly promoted until she is doing a job at which she performs poorly; once she’s reached a job at which she is incompetent, the promotions will stop. In 2009, a group of scientists in Italy tested out this idea using computer simulations; they not only confirmed the Peter principle, but also identified an unexpected optimal strategy.

The researchers built a computational model with individual agents which were each given a random competence level and placed in a hierarchical organisation.  Each year, the agents would age and those above a certain age would retire; others were fired for being below a certain competence level.  The vacancies were filled by promoting an individual from the lower level using one of three strategies: promote the best (ie, most competent), the worst, or a random individual.  The simulations confirmed the Peter principle: promotion of the best individuals led to each agent reaching the level at which they were “maximally incompetent”; in other words, individuals were promoted until they reached the level at which they were least competent.  Since the efficiency of an organisation depends on the competence of the individuals which constitute it, promoting the best individuals also led to a drop in overall efficiency as each employee settled into a job at which they were least competent — and so least useful to the organisation as a whole.  Surprisingly, the opposite strategy leads to the opposite result.  Organisations which promoted the worst performing individuals had an overall increase in efficiency with individual agents settling at the level at which they were most competent!  This seems counter-intuitive at first but makes sense on further reflection.  Incompetent individuals will keep getting promoted out of their position until they reach a level at which they are competent; since everyone ends up in a position where they are as competent as possible, the organisation also becomes more efficient.

Crucially, these different outcomes depend on how an agent’s competence is determined after a promotion.  Following the “common sense” view, an individual’s competence doesn’t change much with their promotion to a new level; alternately, the “Peter hypothesis” suggests that competence might not transfer from one level to the next and should be randomly determined after promotion.  The common sense view might hold when someone is promoted within the same domain, such as a junior salesperson moving up in the ranks, while the Peter hyptohesis might be more applicable to promotions that transfer someone to a new kind of work, such as a salesperson becoming an executive.  The research team ran simulations using both methods to determine competence and found that they led to opposite outcomes.  The Peter principle only holds under the Peter hypothesis; when the “common sense” approach was followed, promoting the best individual led to efficient organisations with individuals at the level where they were most competent.  While this may seem reassuring, there’s no obvious reason to prefer the common sense view.  People who are promoted may be quite competent at their new job, but they may not, particularly if the new position requires a different skill set than their previous work.

So what’s the best way to promote individuals if we can’t know how well their competence will transfer?  It turns out that randomly selecting someone for promotion means that individuals don’t get trapped in positions based on their competence; since those below a certain competence are always fired, the overall efficieny of the organisation also increases.  Of course, the efficiency gains in this strategy are modest compared to the successful combinations but in cases where it’s unclear how well competence will transfer, promoting people randomly seems to be the safest bet.

While it’s tempting to try applying these results to real organisations, they shouldn’t be taken at face value.  Like all models, the system used in this study is a caricature designed to be simple enough to understand and to provide insights. Our decisions should also take into account the many factors that are left out, such as the importance of factors such as experience, which may give someone a good overall view of how an organsiation works, making them likely to be competent at higher levels in the hierarchy.  In addition, my experience with small organisations suggests that promotions often aren’t simply based on who’s best at a job, but also on how well they might do in the new position — so-and-so would make a good leader, for example.  This study is a good example of the value of computational models, which can often present us with unexpected predictions and deepen our understanding of how a system works.  By highlighting the importance of how competence is transferred,  this study provokes us to be aware of the different skill sets required at different levels of a hierarchy and to consider the best way to evaluate people for promotions which might require a different skill set.  While promoting the worst performer would certainly provide the wrong kind of incentive, it seems somewhat unfair to ignore job performance and promote randomly.  Personal evaluation can play a large role in small organisations, but it opens the door to cronyism and may not be feasible in larger organisations.  What do you think would be best?

Alessandro Pluchino, Andrea Rapisarda, & Cesare Garofalo (2009). The Peter Principle Revisited: A Computational Study Physica A 389 (2010) 467-472 arXiv: 0907.0455v3