The Peter Principle Is Necessary

Author’s note: not sure why I decided to let this languish in drafts. It seems both basically true and basically complete. Perhaps I had more thoughts, now forgotten, that I hadn’t managed to fit in? Maybe it felt too obvious?

The Peter principle states that employees are promoted to the level of their incompetence. I suspect that this is actually a necessary consequence of the best way to run an organization, given some starting assumptions that are not particularly out there.


  1. Your organization is arranged such that some positions (classically managers) require fewer employees than some other positions (classically ICs),
  2. The positions with fewer people in them are rewarded more highly than the ones with more in them, in terms of money, or money-proxies such as vacation days or travel budget, or even just in terms of respect and esteem afforded to them,
  3. Employee promotions occur with some regularity,
  4. Maximizing employee productivity is an important goal, and
  5. Employees can be assumed to behave at least somewhat rationally with respect to their economic needs.

The Argument

Before I begin, I’m going to make some simplifying assumptions that aren’t required for the argument but save me some caveats. First, let’s assume that, for any position that is open, there is at least one employee eligible for it for whom the added marginal utility of the higher salary or other perks outweighs the costs of actually performing the job. Second, let’s assume that we’re only talking about employees who actually have some need for money and aren’t just working to scratch an itch.

Let’s go!

We have, by point 2, that there are certain obvious benefits of being in a higher position: more money, more respect, etc. All else equal, your employees would like more of all these things (5). Therefore, employees are incentivized to “go for” promotions (which, by point 3, must exist).

By point 4, the organization wants to maximize the output of its employees. The way to do this is to set up incentive structures that encourage said employees to work hard. The primary form of incentive employers can leverage over employees is money. The way for an employee to get more money from the employer is by getting promoted. Therefore the incentive we control is eligibility for promotion, and what we want is for our employees to do good work.

The obvious solution here is to promote the candidate who has done the best work at the time the promotion is going to occur. Any other solution will fail to incentivize workers the way we want.

And there we have it: whoever’s best at job level N gets promoted to job level N+1 in order to continue encouraging the efforts of the rest of the level-N employees. We never make any effort to evaluate a person’s (possibly latent) skill at job N+1, so we are selecting effectively randomly amongst the pool of candidates for job N+1 (colloquially known as managers).

Is This Universal?

No. There are many ways to break the givens above, although they may have their own problems. For example, assumption 1 can be broken with a sufficiently-flat organization. Assumption 2 can break under a commission or performance-bonus model — I would be surprised if there were many organizations where the worst-paid sales manager made more money than their best-performing salesperson with any regularity. Assumption 3 can be broken in a slow-growing organization with low turnover, or at an organization that mainly hires outside people for managerial positions. Assumption 4 can broken in some industries that just need warm bodies and performance is a commodity; the best fast-food cashier and the worst are so close in terms of actual impact to the business that incentivizing for great performance is of limited utility itself. Assumption 5 can be broken if you assume your employees are human. These are merely examples.

But can we escape the conclusion consciously, in all (or most) cases? I don’t think so, at least not without significant cost. Let’s say you want to avoid the Peter principle by always promoting the person who will be best at the job they’re up for promotion to. Now your employees are incentivized to improve their signaling for the N+1 job, which costs them effort that you would rather they put into their current job. You end up with a system of highly-skilled managers, none of whom manage anyone who’s trying to get any non-managerial work done! Say you want to avoid it by paying people based on their performance, so the promotion isn’t going to come with a hefty pay bump. You end up with people turning down promotions because they can earn more at their current job, and effectively promote the employee who is worst at their current job, because that’s the only one who will accept. And so on.

Utopian Arrangements

Some people think the future of work, that might mostly solve this problem, is extremely flat hierarchies — say one person at top who gets final veto and makes no other managerial decisions, all of which are totally self-organized. This seems to me to be mostly fantasy, or else to require a fast-increasing amount of fuzzy bonding time at the expense of productivity until everyone has to spend all day keeping up their bonds of trust with everyone else and no one produces anything. A flat organizational structure works, and solves the Peter principle problem, for 5 people, but it can’t possibly for 5000.

Author’s note 2021: the previous paragraph is probably true, but without an argument for *why* it’s true it’s pretty unconvincing, I realize.

Maybe there’s a way to make a show of promoting the best performer, and then somehow quietly moving them around until they’re back at level N somewhere else if they were bad at N+1, and finding someone new to go there, but that seems likely to be found out very quickly and thus the trust of employees lost. It also sounds like a great way to get an awful lot of people in the pattern of “work hard, get promoted, flail for a while, get demoted before finding footing, quit because of morale hit of demotion,” which isn’t great.

You could try an option where you can continue doing the level N work but make more money in line with what you’d get at level N+1, but that only goes so far; a manager can legitimately increase the output of 10 or 100 people, whereas an IC cannot, so you eventually can’t pay them more without moving into management because they only control their own output. This also leads to the problem that the only people who move into management are the people who most want to move into management, which runs into the old problem of politics: the quality of wanting power does not correlate well with the quality of being responsible with it.

Concluding Thoughts

It’s not really so bad, I think. We are not exactly living in a Malthusian dystopia, so there’s some room for slack when people don’t perform well. Managers only get promoted up to their level of incompetence, which means that managers who are bad at their jobs are generally going to manage small groups, while the managers in charge of large groups may well not be great at their jobs but have grown up through managing small and medium sized groups particularly well and thus are unlikely to be truly disastrous. And besides, the higher up the hierarchy you go, the fewer positions there are, and thus the less frequently new people get promoted. Someone becomes a VP in charge of a large organization, doesn’t do very well, and gets fired or moves on; a few more tries and you end up getting lucky and randomly getting someone who actually does have the ability to succeed. Remember, it’s not that we are selecting people because they’re not going to succeed, we’re merely not selecting them on likeliness-to-succeed.

This entry was posted in sociology. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s