What Promoting Your Best Rep Costs

The eighty-sixth article was about a study nobody had analysed. This one is about a proposition nobody had tested, which turned out to be true, and which may nonetheless describe rational behaviour.

The Verdict, Stated First

Five claims, in descending order of confidence.

One. The proposition was tested and held up. Microdata on sales workers at 131 firms produced evidence consistent with the principle, described by its authors as the first large-scale empirical evidence for it.

Two. The mechanism is specific and checkable. Firms promote high-performing sales workers ahead of lower-performing ones with greater managerial potential, which is a statement about observable characteristics rather than about hidden qualities.

Three. The authors offer two explanations and rule out neither. Either firms are making inefficient decisions, or the motivational value of promotion justifies the cost of a worse manager.

Four. On our own arithmetic the second explanation is not a stretch. In a firm of fifty salespeople, a two percent motivational lift across the floor pays for the whole mismatch.

Five. Team size determines the answer more than ability does. The same person promoted over two reports needs a 50 percent uplift each and over twenty needs 5 percent, which is the calculation nobody runs.

Our Grades For These Claims

Applying the scheme from the first article in this series.

Grade A for the findings, from an abstract obtained verbatim from the publisher and confirmed identically across four independent records.

Grade A for the mechanism and the counterfactual, from body text of a copy hosted by one of the authors at her own site.

Grade C for the effect sizes, because we obtained no numbers at all beyond the count of firms. The costs are described as high and we cannot tell you how high.

Grade D for the satirical origin, which we report as the 2019 paper's own citation of it and not from the 1969 book.

Grade A for our own arithmetic, which is exact on entirely invented figures.

Our position: a well-tested finding whose most interesting sentence is the one offering an alternative to the interpretation everyone draws.

A Note On Method

Everything here is verified to August 2026.

We obtained the abstract verbatim from the publisher[1], and confirmed it word for word against three further independent records[3][4][5], which is unusually strong confirmation for this series.

We obtained body passages from a copy hosted by one of the authors on her own site[2] and from an institutional repository[3].

We obtained no effect sizes, no regression coefficients and no cost estimates, and report the paper's qualitative language rather than inventing precision.

We did not obtain the 1969 book, and describe it only as the 2019 paper characterises it.

All arithmetic is ours and every figure in it is invented. This article discusses research on promotion decisions and is not employment, management or HR advice.

It Started As A Joke

The origin, which matters for how the finding should be read.

The 2019 paper cites the proposition to Peter and Hull (1969)[2], a book that was written as humour rather than as research.

It records that the idea "has come to define the Peter Principle in the popular press and the academic literature that followed Peter and Hull's original work," and quotes two economics papers adopting it: one describing promotion distortion as "promoting employees who would not be promoted for assignment reasons alone, the Peter Principle effect," and another defining it in terms of firms "imposing simple rules of promotions, based on ... past performance."[1]

Four observations, ours.

A satirical proposition became a defined term in the economics literature, cited and formalised in papers, before anybody tested it on data.

That is a different failure from the ones the last three articles documented. Nothing was misreported; a plausible idea was adopted because it explained something people recognised.

The 2019 authors are careful about this and describe their own contribution as the first large-scale empirical evidence[2], which is a claim about a fifty-year gap.

Fifty years is worth pausing on. The proposition was specific, widely believed, consequential and testable throughout that period, and the constraint was presumably data rather than interest.

And the outcome is the interesting part. The joke was right, or at least the data was consistent with it, which is not how these stories usually end in this series.

Two reasons that ending is worth dwelling on, ours. This series has documented enough failures that a reader could reasonably conclude the whole literature is unreliable, and that conclusion would be wrong.

The difference between this case and the last three is not luck. Here somebody obtained firm-level microdata and tested a specific prediction, rather than describing a pattern from a summary, and the method is what separated the outcomes.

The 2019 Paper

The source.

Benson, A., Li, D., and Shue, K. (2019), Promotions and the Peter Principle, The Quarterly Journal of Economics, 134(4), 2085–2134, November, DOI 10.1093/qje/qjz022[1].

Four observations, ours.

The venue is the Quarterly Journal of Economics, which is among the most selective journals in the discipline, and the subject is a proposition from a comic book about office life.

The paper is openly available from one of its authors' own sites and from an institutional repository[2][3], which is better access than most of what this series covers.

An earlier version circulated as a working paper[5], and the keywords assigned to it include Tournament Theory[4], which signals the alternative explanation before the abstract states it.

And it has already generated modelling work, with an agent-based paper on promotion policies citing it[4], so the empirical result is being built on.

One note on why open access matters here specifically, ours. This is a paper about a decision every small firm makes, and the readers who would benefit most are not behind a university library subscription.

What The Abstract Says

The findings, in the paper's own words.

The question: "The best worker is not always the best candidate for manager. In these cases, do firms promote the best potential manager or the best worker in their current job?"[1]

The finding: "Using microdata on the performance of sales workers at 131 firms, we find evidence consistent with the Peter Principle, which proposes that firms prioritize current job performance in promotion decisions at the expense of other observable characteristics that better predict managerial performance."[1]

Four observations, ours.

The opening sentence is the whole problem in eleven words. The best worker is not always the best candidate for manager, which every business owner knows and almost none acts on.

131 firms with worker-level performance data is a serious sample, and sales is the right setting because individual output is measured rather than inferred.

The phrase "evidence consistent with" is appropriately hedged. This is observational data rather than an experiment, and the authors do not claim to have run one.

And the key words are "other observable characteristics." The claim is not that firms cannot see managerial potential; it is that they can see it, in data they hold, and promote on something else anyway.

Two implications of that phrasing, ours, because it determines whether the finding is actionable.

If the predictors were unobservable, nothing could be done and the finding would be an interesting fact about the limits of information.

Because they are observable, the finding is a claim about how firms weight information they already have, which is a decision procedure and can be changed without acquiring anything new.

The Specific Mechanism

What the body says the firms actually did.

The paper records: "we show that firms discriminate in favor of high-performing sales workers by promoting them ahead of lower-performing sales workers with greater managerial potential."[2]

Four observations, ours.

This is a comparison between candidates rather than a claim about outcomes. Two people were available; the firm took the better seller and the worse prospective manager.

Which makes the finding much stronger than the folk version. The folk version says good performers become bad managers, which could just be regression to the mean and would require no error by anyone.

That distinction is worth holding, because the two versions recommend different things. If it were regression to the mean, the answer would be to expect less of every promotion and change nothing about who you choose.

This version says something different and harder to excuse. The better candidate was there and was passed over, on characteristics the firm could observe.

And we did not obtain what those characteristics were, which is a real gap, since a business owner reading this would want to know what predicts managerial performance in the data.

We would rather flag that than fill it, and the temptation is real. Plausible candidates are easy to generate and none of them would be a finding, and the previous three articles have all documented what happens when a plausible reconstruction gets repeated as though it were reported.

The Counterfactual Policy

The analytical move that turns a correlation into a cost.

The paper records: "We then show that firms overweight sales in promotion decisions by constructing a counterfactual promotion policy that improves managerial quality by promoting fewer", at which point our reproduction is cut off.[2]

Three observations, ours.

The method is the right one. Simulate a different promotion rule against the same data and see what managerial quality it would have produced, which converts a preference into a quantified loss.

The sentence plainly continues "promoting fewer high sales performers" or words to that effect, and we will not complete it for them. Our reproduction stops and we report where.

And this is where the paper's cost estimate comes from, which the abstract describes as high. We obtained the word and not the number, which is the largest gap in this article.

Two Explanations, And They Choose Neither

The sentence this article exists for, and the one that never survives retelling.

The abstract states: "We estimate that the costs of promoting workers with lower managerial potential are high, suggesting either that firms are making inefficient promotion decisions or that the benefits of promotion-based incentives are great enough to justify the costs of managerial mismatch."[1]

Four observations, ours.

"Either ... or" is doing enormous work and it is entirely explicit. The authors found the behaviour, priced it, and declined to say it was a mistake.

The second branch is tournament theory, which the paper's own keywords name[4]. The idea is that the prospect of promotion motivates everyone competing for it, and that motivation has value even when the winner turns out to be a mediocre manager.

Which means the popular reading of this paper, that firms are being stupid, is one of two readings its authors offer and not the one they endorse. They endorse neither.

That distinction deserves emphasis, ours. A paper reporting a cost is frequently read as a paper recommending a change, and this one measured a cost and explicitly declined to say it was not worth paying.

Which is a harder thing to publish than either conclusion would have been. An unresolved finding is less quotable than a verdict, and the authors had the data to assert one and did not.

And whether the second branch holds is an arithmetic question with a computable threshold, which is what the second half of this article does.

Firms Already Adapt

The finding that suggests the behaviour is not blind, and it is the most encouraging thing in the paper.

The abstract records: "We find that firms manage the costs of the Peter Principle by placing less weight on sales performance in promotion decisions when managerial roles entail greater responsibility and when frontline workers are incentivized by strong pay for performance."[1]

Four observations, ours.

Both adaptations are exactly what a firm weighing the two explanations would do, which is a strong signal that the second branch is operating.

When the managerial job matters more, the cost of a mismatch rises, so firms weight sales less. That is the efficiency consideration winning where it should.

And when frontline workers already have strong pay for performance, the motivational value of promotion is lower, because the incentive is already being supplied by the commission structure. So firms weight sales less there too.

That second one is the subtler and more useful. A firm that pays commission does not need promotion as a carrot, and can therefore promote on managerial merit without losing anything.

Two things follow for a small business specifically, ours. Most small sales operations do pay commission, which on this logic puts them in the category where promoting on managerial merit costs the least.

And the implication is unusually clean. If your people are already motivated by what they earn per sale, the motivational case for promoting the top seller is weaker, and the efficiency case is unopposed.

What Actually Survives

Our reading, stated directly.

Five statements.

A satirical proposition from 1969 was tested on microdata from 131 firms and the data was consistent with it, fifty years later.

The mechanism is a comparison: better sellers promoted ahead of worse sellers with greater managerial potential, on characteristics the firm could observe.

The cost was quantified against a counterfactual policy and described as high, and we obtained no figure.

The authors offer two explanations and endorse neither, the second being that promotion's motivational value justifies the mismatch.

And firms already vary their behaviour in the directions the second explanation predicts, which is evidence that they are trading these off rather than blundering.

One thing that follows for a reader deciding how much of this to act on, ours. The finding is about firms in aggregate and your decision is about one person, and an aggregate tendency does not tell you that your candidate is the wrong one.

What it does supply is a reason to run the check. If firms systematically overweight current performance, the prior on your own candidate should be adjusted accordingly, which is a smaller and more defensible use of the result.

The Cost Nobody Counts

Our own arithmetic, on entirely invented figures, and it starts with the item missing from every promotion conversation.

Promote your best salesperson and you stop getting their sales. If an average rep contributes 100 and your best contributes more, the surplus is what you gave up.

The uplift their management must produce per direct report, merely to replace it:

A 1.3x rep over 5 reports: 6.0 percent each. A 1.5x rep over 5: 10.0 percent. A 2.0x rep over 5: 20.0 percent. A 2.5x rep over 5: 30.0 percent.

Four observations.

A rep who sells twice the average, managing five people, must lift each of them by 20 percent to break even. That is a large number, and it is the floor rather than the target.

The loss is immediate and certain, and the uplift is delayed and uncertain, which is a trade most owners would refuse if it were stated that way.

It is never stated that way. The promotion is discussed as a reward and a retention move, and the foregone sales appear in no calculation.

And the better the rep, the worse the trade. The person most deserving of promotion is the most expensive to promote, which is the whole difficulty in one line.

Two consequences that follow directly, ours. The cost scales with the gap between your best and your average, so a firm with one outstanding seller and several ordinary ones faces the steepest version of this.

And that firm is precisely the one most likely to promote. A standout performer in a flat team is the obvious candidate, and the same fact that makes them obvious makes them expensive.

One check that costs nothing, ours. Compute the multiple before the conversation starts, because a firm that discovers its candidate sells three times the average has learned something it will not learn afterwards.

The Break-Even Formula

The rule, worth carrying. Ours, and exact given the definitions.

The required uplift per direct report equals the promoted person's multiple minus one, divided by the number of reports.

Four observations.

A 2x rep over 5 reports gives one over five, which is 20 percent. A 1.5x rep over 10 gives half over ten, which is 5 percent.

Both inputs are things a firm already knows. You know what your reps sell and you know how many people the manager will have.

The formula makes the decision comparable across candidates, which is its practical value. Promoting your second-best rep costs less and may buy nearly the same management, and the difference is computable.

And it says nothing about whether the person will be a good manager, which is a separate question. This prices the cost of the promotion; it does not estimate the benefit.

Two ways to use it despite that, ours. Compare candidates rather than evaluating one, since the formula is exact for each and the difference between two candidates is the part you can be confident about.

And use it as a threshold question rather than a forecast. Asking whether this person could plausibly lift five people by a fifth is a more answerable question than asking how good a manager they will be.

And If They Manage Badly

The downside case, which the paper's finding makes relevant. Ours, invented figures: a 2x rep over five reports.

At a per-head effect of minus 5 percent: the team falls 25, and the total change against leaving them selling is minus 125.

At zero: minus 100. At plus 5: minus 75. At plus 10: minus 50. At plus 20: zero. At plus 30 percent: plus 50.

Four observations.

A manager who makes the team five percent worse costs 125 against having left them in post, which is more than the surplus you gave up, because the losses compound rather than cancel.

A merely neutral manager costs the full surplus. Doing no harm is not the break-even point, which is the intuition most promotion decisions carry and it is wrong.

The paper's finding is precisely that firms promote people with lower managerial potential than available alternatives[2], so the negative rows are not hypothetical.

And we have no effect size from the paper to tell you how negative, which is why our table spans a range rather than naming a figure.

One reason to take the negative rows seriously anyway, ours. A new manager's first year is when their reports decide whether to stay, and turnover costs do not appear in a per-head productivity figure at all.

Our table therefore understates the downside. It prices output effects and not replacement costs, and the second is frequently the larger for a small firm.

Team Size Flips The Decision

The variable that matters more than ability. Ours, break-even uplift for a 2x rep:

Over 2 reports: 50.0 percent each. Over 3: 33.3. Over 5: 20.0. Over 8: 12.5. Over 10: 10.0. Over 15: 6.7. Over 20: 5.0. Over 30: 3.3 percent.

Four observations.

The same person, with the same ability, is a bad promotion over two reports and a good one over twenty. Nothing about them changed.

Which means the decision is mostly about organisational structure rather than about the individual, and the conversation is almost always about the individual.

It also explains a difference between large and small firms that has nothing to do with sophistication. A large firm's promotion spreads the cost over a big team, and a small firm's does not.

And it gives a small business a concrete instruction. If the management job is two or three people, promoting your best seller is very hard to justify, and the arithmetic says so before any judgement about the person.

One structural response worth considering, ours and untested. If the arithmetic depends this heavily on team size, the team size is a variable too, and a firm can sometimes widen a management role rather than abandon a promotion.

That is not always available and it has its own limits. A manager with twenty reports and no experience is a different problem, and we would not present widening the span as free.

The Tournament Arithmetic

The paper's second explanation, priced. Ours, same invented figures.

If the prospect of promotion motivates everyone who can see it, the benefit is spread across everyone below, not just the direct reports. The motivational uplift needed to cover a 2x rep's foregone surplus:

Across 5 people: 20.00 percent each. Across 10: 10.00. Across 20: 5.00. Across 50: 2.00. Across 100: 1.00 percent.

Four observations.

In a firm with fifty salespeople, a two percent motivational lift across the floor pays for promoting the wrong person entirely.

Two percent is small enough to be plausible and small enough to be unmeasurable, which is precisely why the paper cannot rule the explanation out and neither can we.

The mechanism does not require anyone to be motivated by the promotion itself. It requires only that visibly promoting the best performer signals what the firm rewards, and that the signal changes behaviour slightly.

And this is why the paper's careful hedge is right rather than evasive. The threshold is low enough that the question is genuinely open, on any figures we could construct.

Two cautions about our own computation here, ours, because it is the one most likely to be over-read. We have shown what the motivational effect would have to be, not that it is that large.

And a two percent lift across fifty people is exactly the size of effect this series has spent eighty-six articles being sceptical about. A threshold small enough to be plausible is also small enough to be undetectable, and we would apply our own standard to it.

Which leaves the question genuinely unresolved rather than resolved in the tournament's favour, ours. An explanation that cannot be measured is not thereby established, and a firm relying on it is relying on something nobody has demonstrated.

Why That Changes The Verdict

What follows for a business owner. Ours.

Four observations.

The popular reading says stop promoting your best performers, and the arithmetic says that is right only if nobody is watching.

In a firm of five, almost nobody is watching, and the motivational threshold is 20 percent, which is not plausible. The efficiency argument wins in a small business.

In a firm of a hundred, the threshold is 1 percent, and the signal sent by passing over the best performer is large. The tournament argument becomes serious at scale.

And that is a genuine divergence in the right answer by firm size, which the popular version of this finding does not contain. The advice inverts somewhere between the two, and we cannot tell you where.

Two features of that divergence make it worth stating despite the uncertainty, ours. Most business advice on this topic is written for large firms and read by small ones, and here the two need opposite conclusions.

And the direction is the one a small firm would not guess. The received wisdom against promoting top performers is most correct where firms are smallest, which is the opposite of how most received wisdom scales.

Why This Is Not A Bias

A classification point, since this series has spent eighty-six articles on biases and this is not one. Ours.

Four observations.

Nothing here requires anyone to reason badly. A firm that promotes its best seller may be maximising something real, and the paper explicitly leaves that open.

It also does not require regression to the mean, which is the explanation most people reach for. The folk version, that strong performers regress to mediocrity in new roles, would happen with no error by anyone, and the paper's finding is stronger and different.

What the paper documents is a choice between two visible candidates, resolved in favour of the one whose observable characteristics predicted managerial performance worse.

And that is either an error or a trade, which is exactly the ambiguity the abstract preserves. Calling it a bias would resolve a question its own authors left open, so we do not.

One consequence for how the series should be read, ours. Not everything in the behavioural literature is a defect to be corrected, and a publication that treats every documented pattern as an error to fix will eventually recommend against something that was working.

The Retention Argument

The reason owners actually give, which the arithmetic has not yet touched. Ours.

Four observations.

The stated reason for promoting a top performer is frequently neither efficiency nor motivation. It is that they will leave otherwise, and losing them entirely is worse than losing their selling time.

That is a real consideration and it changes the comparison. The alternative is not keeping them in post; it is keeping them at all, and if the counterfactual is departure the surplus is lost either way.

But it is testable rather than assumed, and it usually is assumed. Has this person said they want to manage, or has the firm inferred it from their being good at something else?

And the answer matters, because many strong individual performers do not want the job and take it because it is the only route to more money, which is the problem the second ladder exists to solve.

Two ways to find out cheaply, ours. Ask before offering, which sounds obvious and is skipped because an offer is a compliment and a question is not.

And give them a piece of the job first: onboarding one new hire, running one weekly meeting, owning one account handover. That is a real signal about both willingness and aptitude, and it is reversible in a way a promotion is not.

The Small Team Problem

Applying it where this publication's readers actually sit. Ours, and not employment or management advice.

Four points.

A firm with eight people promoting its best salesperson to manage three is in the worst region of the arithmetic: the surplus lost is large, the team over which to recover it is small, and the audience for the motivational signal is seven people.

The break-even on our figures for a 2x rep over three reports is 33 percent per head, and management that lifts three people by a third is exceptional rather than typical.

Which does not mean never promote. It means the promotion has to be justified by something other than reward, and if the justification is that they earned it, the arithmetic says you are paying a great deal for the gesture.

Two justifications that do survive the arithmetic, ours. Succession, where the firm needs somebody who can run it and the cost is an investment rather than a transaction.

And genuine aptitude that has already shown itself, which is different from strong selling and is observable if anyone looks for it, which is precisely what the paper found firms failing to do.

And the alternative most owners reach for is worse than they think, which the next section covers.

The Obvious Answer And Why It Fails

The remedy everyone proposes. Ours, and untested.

Four observations.

The standard answer is a second ladder: a senior individual-contributor track, so a strong seller can be promoted in pay and title without managing anyone.

On the arithmetic it is exactly right. It removes the foregone surplus entirely while retaining most of the reward, and the paper's own finding about pay for performance points the same way[1].

It fails in small firms for a reason that is structural rather than cultural. A senior title with no reports is only credible if the firm has several of them, and in a firm of eight it reads as a consolation prize.

One version that does work at small scale, ours and untested. Give the role a function rather than a rank: responsibility for pricing, for the largest accounts, or for training new hires, which is a real job with a real title and no reports.

And the honest version of the alternative is the unglamorous one. Pay them more and do not promote them, which requires a conversation most owners avoid and is what the arithmetic actually recommends.

Two things make that conversation easier than it looks, ours. The arithmetic is the argument, and showing someone what their selling is worth to the firm is a form of respect rather than a refusal.

And the offer is better than it sounds if it is real. More money for the job they are good at, without the job they have not been trained for, is what a large number of strong performers would take if it were ever put to them.

The Two Jobs Have Almost Nothing In Common

The underlying reason the arithmetic bites. Ours.

Four observations.

Selling and managing sellers share a subject matter and almost nothing else. One is executed personally and measured individually; the other is executed through other people and measured collectively.

Which means being excellent at the first is weak evidence about the second, in the way that being an excellent driver is weak evidence about running a haulage firm.

The paper's finding is consistent with something stronger than irrelevance, though it does not establish it. Some of what makes a great individual seller may work against managing, if the same drive to close personally makes it harder to let a junior person struggle through a call.

And we would flag that last point as ours and unsupported. The paper reports that other characteristics predicted better, not that selling ability predicted negatively, and those are different claims.

A Promotion Decision With The Arithmetic In It

What the conversation would contain. Ours, and not management advice.

Four elements.

The candidate's multiple, being what they contribute relative to an average person in the role, which most firms can compute in an afternoon.

The number of direct reports, which converts that multiple into a required per-head uplift by simple division.

The comparison candidate, since the paper's finding is about choosing between people rather than about one person, and the second-best seller costs less by exactly the difference in multiples.

And an explicit answer on the retention question. Whether the alternative is keeping them in post or losing them entirely changes the comparison completely, and it is usually assumed rather than asked.

Two things that list deliberately excludes, ours. How long they have been with the firm, which is a fact about the past and predicts nothing about managing.

And whose turn it is, which is the criterion most small firms actually apply and which appears in no defensible version of this decision.

What You Would Have To Measure

Turning the paper's finding into something checkable. Ours.

Four points.

The paper's contribution is that observable characteristics predicted managerial performance better than sales did[1], and we did not obtain which characteristics.

That gap is worth stating as sharply as we can, because it determines what a reader can do with this. The finding tells you that you are weighting the wrong thing and not what the right thing is, and only the second would change a decision this week.

Which means a firm cannot simply adopt its answer, and has to find its own. The available method is to record what you observed before each promotion and what happened after, which most firms never connect.

Team turnover, team output relative to trend, and how long the new manager's reports stay are all measurable, and the relevant comparison is against the promoted person's own prior team rather than against the firm.

One refinement on that comparison, ours. The prior team was managed by somebody, and the honest question is how it performed under the previous manager rather than how it performs against the firm average.

And a firm promoting once every two years will not accumulate enough cases to learn this, which is the seventy-fourth article's condition. For most small businesses the arithmetic has to substitute for the experience, because the experience will never arrive.

One partial substitute worth knowing, ours. Other people's promotions are data too, and an owner who has watched a dozen managers appointed across their industry has more cases than their own firm will ever supply.

That evidence is weaker and it is not nothing. Unsystematic observation of many cases beats systematic observation of one, provided the observer remembers the ones that did not work out, which is the fifty-eighth article's warning applied to memory.

The Honest Position

Where we land, and it is less decisive than the popular version. Ours.

Four observations.

The finding is real and well-evidenced. Firms do prioritise current performance over observable predictors of managerial performance.

Whether that is a mistake is unresolved by the paper, deliberately, and our arithmetic says the alternative explanation clears a low bar at scale.

For a small firm the arithmetic points fairly clearly. The motivational audience is too small and the team too small, so the efficiency argument dominates.

And for a larger one it genuinely does not. We would not tell a fifty-person firm that promoting its best performer is a mistake, because on our own figures a two percent floor-wide lift settles it, and nobody knows whether that lift exists.

One thing we would say to a firm of any size, ours. Whichever way you decide, decide it rather than defaulting. The arithmetic takes an afternoon and the promotion lasts years.

Bibliographic Note

The series keeps a count.

The paper's own running header, on two separately hosted copies, reads "The Quarterly Journal of Economics (2020), 2085–2134"[2][3], where the publisher's record gives Volume 134, Issue 4, November 2019[1].

Three observations, ours.

A November issue appearing in the following year's bound volume is ordinary publishing practice, and both dates are defensible.

It nonetheless produces the situation we have flagged before. A reader citing from the PDF header will write 2020 and a reader citing from the publisher will write 2019, and both are citing the same paper.

That brings the running count of bibliographic variants across this series to thirty-two.

What To Do

Price the foregone sales before promoting anyone. On our own arithmetic the required uplift is the promoted person's multiple minus one, divided by the number of direct reports.

Count the reports. A 2x rep over two reports needs a 50 percent uplift each and over twenty needs 5 percent, which is the same person and opposite decisions.

Do not treat neutral as break-even. A manager who changes nothing costs you the entire surplus you gave up.

Ask who is watching. On our own figures a two percent motivational lift across fifty salespeople pays for the whole mismatch, and across five people the threshold is twenty percent.

Consider your second-best rep seriously. The cost of promoting them is lower by exactly the difference in their multiples, and the management may be the same or better.

Weight sales less when the management job is bigger. That is what the firms in the study did, and it is what the cost arithmetic recommends.

Weight sales less when you already pay commission. The study found firms do this too, and the reason is that the incentive is already supplied.

If the answer is pay rather than promotion, say so plainly. The arithmetic recommends it and a title with no reports in a firm of eight will not survive contact with the person receiving it.

The Limits Of This Analysis

Several caveats matter. This article discusses research on promotion decisions and is not employment, management or HR advice; the applications are our own reasoning and untested. Everything is verified to August 2026. We did not obtain the paper's results, only its abstract, which we confirmed identically across four independent records, and body passages from two hosted copies; we obtained no effect sizes, no regression coefficients and no cost estimates whatever, so the paper's description of the costs as "high" is reported as its word and not converted into any figure by us. We did not obtain which observable characteristics predicted managerial performance, which is the single most useful thing in the paper for a business reader and its absence is the largest gap here. Our reproduction of the counterfactual-policy sentence cuts off mid-clause and we have not completed it. We did not obtain the 1969 book and describe it only as the 2019 paper characterises it. The study is observational rather than experimental, covers sales workers specifically, and its authors describe their finding as evidence consistent with the proposition rather than as proof of it; sales is a setting with unusually measurable individual output, and whether the result generalises to roles without it is not something the paper claims. All arithmetic is ours and every figure in it is invented: the performance multiples, the team sizes, the per-head effects and the motivational thresholds are constructions demonstrating a structure, and none is an estimate of anything real. Our arithmetic prices the cost of a promotion and not its benefit, and a decision requires both. And the tournament explanation, which our figures show clearing a low threshold, is not thereby shown to be true: we have computed what it would take, not whether it happens.

Related Reading

Behavioural Finance Entirely Fictional The opposite case: a famous claim that did not survive the data being found. Behavioural Finance Nobody Has To Be Wrong The other case where selecting on one measure gets you something else entirely. Behavioural Finance The Hiring Validity Revision What actually predicts performance, when somebody checks the numbers properly.

Frequently Asked Questions

Is the Peter Principle real?

A 2019 study using microdata on sales workers at 131 firms found evidence consistent with it, describing itself as the first large-scale empirical evidence. Specifically, firms promoted high-performing sales workers ahead of lower-performing ones with greater managerial potential, on characteristics the firm could observe.

So firms are making a mistake?

The paper does not say so. Its abstract offers two explanations and endorses neither: either firms are making inefficient decisions, or the motivational benefits of promotion-based incentives justify the cost of a worse manager. That hedge almost never survives in retellings of this finding.

What does promoting my best rep actually cost?

Their foregone sales. On our own arithmetic the uplift their management must produce, just to break even, is their performance multiple minus one, divided by the number of direct reports. A rep selling twice the average, managing five people, must lift each of them by 20 percent.

Does team size matter?

More than ability does. The same 2x rep needs a 50 percent uplift per head over two reports and 5 percent over twenty. Nothing about the person changes; the decision does, which means it is mostly a question about structure rather than about the individual.

When does the motivational argument work?

At scale. On our own figures, covering a 2x rep's foregone surplus needs a 20 percent motivational lift across five people but only 2 percent across fifty. Two percent is plausible and unmeasurable, which is why the paper cannot rule the explanation out and neither can we.

What should a small firm do differently?

A small firm sits in the worst region of the arithmetic: large surplus lost, few reports to recover it over, and a small audience for the motivational signal. On our own figures the efficiency argument dominates there, and the honest alternative is usually to pay them more without promoting them.

Do firms already account for this?

Yes, in two ways the study identified. They place less weight on sales performance when the managerial role carries greater responsibility, and when frontline workers already have strong pay for performance. Both are what a firm weighing efficiency against motivation would do.

References

1. Publisher record for Benson, A., Li, D., & Shue, K. (2019), Promotions and the Peter Principle, The Quarterly Journal of Economics, 134(4), 2085–2134, November, DOI 10.1093/qje/qjz022, reproducing the abstract: that the best worker is not always the best candidate for manager, and asking whether in these cases firms promote the best potential manager or the best worker in their current job; that using microdata on the performance of sales workers at 131 firms the authors find evidence consistent with the Peter Principle, which proposes that firms prioritise current job performance in promotion decisions at the expense of other observable characteristics that better predict managerial performance; that the authors estimate the costs of promoting workers with lower managerial potential are high, suggesting either that firms are making inefficient promotion decisions or that the benefits of promotion-based incentives are great enough to justify the costs of managerial mismatch; and that firms manage those costs by placing less weight on sales performance in promotion decisions when managerial roles entail greater responsibility and when frontline workers are incentivised by strong pay for performance. The same record reproduces body text on how the idea came to be defined in the popular press and academic literature following Peter and Hull's original work, quoting Fairburn and Malcomson (2001) and Faria (2000). Note: the publisher's record. Our source for the abstract, which we confirmed identically against three further independent records. We obtained no effect sizes, coefficients or cost estimates. academic.oup.com
2. Copy of the same paper hosted by one of its authors on her own site, reproducing body passages: that using detailed microdata on sales workers in United States firms the authors provide the first large-scale empirical evidence of the Peter Principle, a hypothesis that firms prioritise current performance in promotion decisions at the expense of promoting the best potential managers, citing Peter and Hull (1969); that in particular they show firms discriminate in favour of high-performing sales workers by promoting them ahead of lower-performing sales workers with greater managerial potential; and that they then show firms overweight sales in promotion decisions by constructing a counterfactual promotion policy that improves managerial quality by promoting fewer, at which point our reproduction is cut off. The running header gives the year as 2020. Note: a copy hosted by one of the paper's authors at her own site. Our source for the mechanism and the counterfactual method. The counterfactual sentence is truncated in our reproduction and we have not completed it. danielle.li
3. Institutional open-access repository copy of the same paper, reproducing the abstract identically to the publisher's record and the same body passages, and giving the citation as Benson, Alan, Li, Danielle and Shue, Kelly, 2019, Quarterly Journal of Economics, 134(4). Its running header also gives the year as 2020. Note: an institutional repository record. Used to confirm the abstract independently and as a second source for the body passages. Recorded also as a bibliographic variant: the running header year differs from the publisher's stated issue year. dspace.mit.edu
4. Working-paper repository record for the same study, reproducing the abstract identically and listing its keywords as Peter Principle, Incentives, Promotions, Managers and Tournament Theory; together with a reference list carried on a later agent-based modelling paper on promotions, efficiency and mitigation policies, confirming the citation as Benson, Li and Shue, Quarterly Journal of Economics 134 (2019), 2085–2134. Note: a working-paper repository record and a citing paper's reference list. Recorded because the assigned keyword Tournament Theory names the alternative explanation before the abstract states it, and because it establishes the result is being built on. papers.ssrn.com
5. Research bureau record for the earlier working paper version, number 24343, confirming the authors, the published citation and reproducing the same abstract passage on the costs of promoting workers with lower managerial potential and the two explanations offered for it. Note: the issuing research bureau's record. Used as a fourth independent confirmation of the abstract's wording. nber.org

This article discusses research on promotion decisions and is not employment, management or HR advice. The paper's results were not obtained, only its abstract and two body passages; no effect sizes or cost estimates are reported because none were obtained, and the observable characteristics that predicted managerial performance are not identified here because our sources do not name them. The study is observational and covers sales workers specifically. All arithmetic is the authors' own and every figure in it is invented; it prices the cost of a promotion and not its benefit.