A saying from the 1970s claims that first-rate people hire first-rate people and second-rate people hire third-rate people. This chapter takes it apart: what it asserts, why a sensible person does exactly that, how the habit compounds through a hierarchy, what it looks like from the inside, and what the evidence says about whether it is getting worse.
The sentence has been in circulation for more than fifty years and it has several fathers. In April 1970 a Canadian business paper printed it as Rosten's Law, after the American writer Leo Rosten: first-rate men hire first-rate men, second-rate men hire third-rate men. David Ogilvy sent every new head of an office a Russian nesting doll; inside the smallest one lay a note: if each of us hires people who are smaller than we are, we shall become a company of dwarfs, but if each of us hires people who are bigger than we are, we shall become a company of giants. Steve Jobs said it as A players hire A players and B players hire C players, and Guy Kawasaki, who worked for him, added the diagnosis in 2011: B players hire C players so they can feel superior to them. What follows, Jobs called the bozo explosion. Fifteen years before Rosten, in 1955, C. Northcote Parkinson had written the sober version in The Economist: an official wants to multiply subordinates, not rivals.
Three claims are folded into one sentence. They are worth separating, because the evidence for each is different.
Hiring quality is inherited downward. Whoever a manager hires will hire in the same spirit one level down, and the effect multiplies through the hierarchy. This is arithmetic, and the next tab makes it visible.
The second-rate manager does not hire down by accident. He hires people who will not overlook him, because his own position, not the product, comes first. This is a claim about incentives, and the tab after next collects what has been measured about it.
The danger, in this view, is not incompetence below but competence: the first-rate employee notices real problems, names them and questions decisions, not out of malice but in order to improve things. To a manager who is measured on appearances, exactly that is the threat.
Both are used, and they lead to different remedies.
The reading about character. The first-rate manager is brave. He wants the best product, seeks out difficult situations, accepts that the best people will contradict him and hires them anyway. The second-rate manager is afraid, and fear hires small. On this reading the remedy is to find brave people and to promote them.
The reading about structure. Both managers respond sensibly to how they are measured. Where a manager is judged by the results of his unit, a strong hire raises his own score. Where he is judged by his standing relative to peers, by how he appears upward, or by whether his unit runs without visible trouble, a strong hire is a risk and a weak hire is insurance. On this reading the remedy is to change the measure, because character is scarce and measures are cheap.
It does not say that every weak hire is self-protection; budgets, availability and the labour market hire weakly too. It does not say that first-rate is one dimension; the best engineer, the best seller and the best manager are three different people more often than not, and the fifth tab returns to that. And it does not say that everyone who questions a decision is first-rate; some are, some are merely loud. The sentence describes a tendency and a direction. The chapter treats it as that.
The saying is a compound-interest statement. Whatever a manager does when hiring is done again one level down by the people he hired, and again below them. A small habit per level becomes the character of the whole firm. The model below makes that arithmetic visible. It is a stylised model with one number for quality, not a measurement of any real company.
Two kinds of managers share the organisation. One kind hires people at least as good as themselves, five per cent better in this model. The other kind keeps a gap: it hires people a fixed distance below itself, so that nobody in the room overlooks it.
Average quality by level for three organisations: one where everyone hires up, yours, and one where everyone keeps the gap. ● = your organisation, level by level.
The older proverb says the same thing from the other end: a fish rots from the head down. It is at least as old as Rumi, who wrote in the thirteenth century that a fish begins to stink at the head, not the tail; the British ambassador Sir James Porter brought it home from Constantinople in 1768 as a Turkish saying about disorderly servants and their disorderly master. Biologically it is wrong, a fish rots from the guts. Organisationally it is the cascade above, read downward: what the top does with its first hire is inherited by every level below, and by the time the smell reaches the working level it has been multiplied once per level. Two additions from the model. First, the head need not be bad: a first-rate top that tolerates a ranking measure below it produces the same cascade through second-rate hires it never made itself. Second, the proverb says where to look, not whom to blame. The tail cannot fix it, and the head can only fix it by changing what it rewards.
Three honest caveats. First, quality is not one number; a person can be first-rate at the work and third-rate at judging people, and the model collapses that into a single scale. Second, real organisations hire from outside and people learn, both of which soften the cascade; the model shows the pull of the habit, not the whole labour market. Third, the share slider is the one that policy moves. In a Harvard experiment published in January 2025, participants who were paid relative to a partner chose the more talented partner 60 per cent of the time; paid on absolute results, 90 per cent. Roughly that difference is what the slider describes, and it is set by how a firm measures its managers, not by who they are.
None of the mechanisms below needs a villain. Each is a person responding sensibly to how they are measured, evaluated and paid. That is the uncomfortable part of the saying: the second-rate manager is often behaving rationally inside a first-rate mistake. The eight cards collect what has been written and measured, from a 1955 essay to a survey of executives published in 2025.
The first of Parkinson's two axioms: an official wants to multiply subordinates, not rivals. A subordinate adds to one's weight, a rival takes from it. His examples were the British Admiralty, whose staff grew while the fleet shrank, and the Colonial Office, which reached its largest size as the colonies disappeared. Written as satire, read for seventy years as description.
Where promotion is a contest among peers, pay depends on rank rather than on output, and a strong subordinate is tomorrow's competitor. Lazear showed in 1989 that the steeper the prize, the more it pays to hold others back. Harbring and Irlenbusch confirmed it in the laboratory in 2011: effort and sabotage both rise with the prize spread, and calling sabotage by its name reduces it.
Where evaluation is subjective, the worker's best strategy is to hold the opinion the boss already holds. The firm then loses exactly the contrary views it needed, and decisions drift toward the centre. Prendergast showed that the loss appears even when nobody is risk-averse; it is built into the way of evaluating, not into the people.
People asked to recommend a candidate avoided the one who would beat them on their own strong dimension. When they were strong on quantity, 31 per cent recommended the quantity leader; when they were strong on quality, 69 per cent did. In a second study 38 per cent against 74 per cent chose the teammate who was stronger where they themselves were strong. The authors called it the social comparison bias: people protect the comparison they win.
Supervisors with a taste for hierarchy treated high-performing subordinates abusively, and the path ran through the perceived threat to their position. Coworkers undermined colleagues whose trajectory pointed upward faster than their own. A 2025 study in the Journal of Business Ethics found the same for overqualified subordinates: the supervisor's perceived status threat predicted undermining, more so when the subordinate was outgoing.
A Harvard survey of 335 executives: 71 per cent had witnessed managers sabotaging talented subordinates, 5 per cent admitted to doing it, 28 per cent said it was happening to them in their current organisation. Where managers were evaluated relative to each other, 38 per cent reported it, against 33 per cent elsewhere. In an experiment, 60 per cent chose the more talented partner under relative pay and 90 per cent under absolute pay.
Managers devalue ideas that come from internal rivals and overvalue the same ideas when they arrive from outside. The consultant is believed where the colleague was not. This is why a first-rate employee's correct objection can lose to an external slide deck saying the same thing a year later, at a fee.
Sales workers at more than 130 firms: the best sellers were the ones promoted, although sales ability did not predict managerial quality, and their teams' performance fell under them. Firms promote for the incentive effect on the rest and pay for it with weaker management. The Peter Principle of 1969, measured. Many second-rate managers were first-rate at something else.
In economic terms this is a principal-agent problem. The firm, the principal, wants the best product; the manager, the agent, wants his position; and the firm can only observe part of what the manager does. Wherever the observable part is appearance, calm and relative standing rather than results, the agent's rational choice is the small hire. The mechanisms above are the ways that choice is made and defended. The last two tabs are about changing what is observed.
From the inside the mechanism is quiet. Nobody announces that they hire down. What can be seen are its by-products, and they are listed below as a checklist. Each single sign has an innocent explanation; the pattern does not. Tick what applies to a team or firm you know; the count is indicative, not a diagnosis.
What people who work in such organisations report most often, ordered from the earliest to the latest sign.
Three or fewer signs describe most organisations on a bad day. Four to six describe a pattern worth watching over a few months, in particular the first and the sixth sign, because they are the ones that compound. Seven or more mean that the mechanism from the previous tab is probably at work, and that the firm's measures, not its people, are the place to look. None of this identifies a person. A single manager under a ranking system will show several signs and change them the moment the system changes; that is the point of the structural reading.
The observation behind this chapter comes from conversations with partners and colleagues: in many firms the point of the game is no longer the product but the appearance, and the people who name real problems have become a danger to the people who manage appearances. What holds, where a caveat belongs, and what the data say about the last few years.
People protect the comparison they win, in the laboratory and in recommendations. Managers under relative evaluation choose weaker partners, abuse high performers when their position feels threatened, and sabotage upward talent often enough that 71 per cent of executives have seen it. Subjective evaluation produces yes men by design. And the arithmetic of the cascade is not in dispute: a habit that is inherited downward compounds.
Would contradict it: a firm under ranking incentives that nevertheless hires up at every level for years.
The best seller promoted into management is the commonest way a second-rate manager comes into being, and no self-protection is needed for it. The talent mindset has its own failure: Enron hired first-rate on paper and Malcolm Gladwell described in 2002 how a firm that worships talent stops checking results. Netflix wrote brilliant jerks into its culture document in 2009 as people it would not keep, because a first-rate skill with third-rate cooperation costs the team more than it brings. And not every objection is optimisation; the loud are not always the right.
Note: nothing here contradicts the saying, it narrows it: the direction is right, the single scale is not.
Gallup measured in 2024 a fall of global engagement from 23 to 21 per cent, and among managers from 30 to 27, five points among managers under 35 and seven among women. Spans widened: in 8,500 small US firms a manager had 3.15 direct reports in 2019 and 5.76 in September 2024; Google cut a tenth of its vice president and manager roles in 2024, Microsoft removed 9,000 positions in 2025 while flattening, and the remaining managers absorbed the scope. Whether self-protection has grown cannot be read from this; that the conditions for it have, can.
Interpretation: wider spans and ranking make the small hire more tempting; flattening removes the layers where the habit lived. Both at once.
One thing is new, and it connects this chapter to the previous one. Producing a plausible document has become almost free (Costs & leverage, the tab on agentic flooding). A manager who could not judge the work used to be protected by the fact that little work reached him; now a flood of polished output reaches him, and the only defence is judgment. The person who can tell substance from filler was always valuable; in a flood, that person is the one the second-rate manager can least afford to hire and least afford to lose. That is why the saying feels more current than it did, even though the mechanism is seventy years old.
If the structural reading is right, the remedy is not to search for brave managers but to change what the firm observes and rewards. Six rules follow from the mechanisms; each one moves the share slider in the cascade model, and none of them needs a new tool.
Ogilvy's criterion, made explicit: whether a manager hires people better than himself becomes a line in his own evaluation, with names. What is written down and asked about gets done; what is admired in speeches does not.
A panel or a committee decides, and the direct superior has one vote. Google has used hiring committees for this reason since its early years, as Laszlo Bock described in 2015: the person who would be outshone must not be the only judge.
Retention, growth and where the team's alumni end up. Skip-level conversations and upward feedback belong to the normal routine, not to the crisis. Google's Project Oxygen found in 2008 that its best managers coach and do not micromanage; both are things only the people below can report.
Rankings among peers invite sabotage, and the Harvard experiment put a number on it: 60 per cent chose the talented partner under relative pay, 90 per cent under absolute pay. Where a ranking cannot be avoided, name what it invites; in the laboratory, calling sabotage by its name reduced it.
Prendergast's yes men are bred by subjective evaluation. Add criteria that can be checked, ask for decisions with reasons that could be wrong, and make sure the person who raises a problem gets the credit for it, not the label. A firm learns exactly as fast as its cheapest disagreement.
Exit interviews by somebody outside the line, and a simple statistic per manager: who left, how good were they, where did they go. The best leaving first is the earliest measurable sign of the mechanism, and the one most often explained away.
Before joining a team, ask two questions that the checklist cannot: who was the last person to leave this team and why, and what happened to the last person who told the manager he was wrong. The answers, or the silence, tell most of what the ten signs tell. And for those who are the first-rate employee in a second-rate manager's team: the mechanism is about his measure, not your worth. Whether staying is worth it depends on whether the measure can change, and that is decided above him.
The terms used in this chapter, each with the year and the person they come from where there is one.
First-rate people hire first-rate people; second-rate people hire third-rate people. Attributed to the writer Leo Rosten, first printed in April 1970. The saying this chapter is built on.
What happens after the first B player is hired: B hires C, C hires D, and within a few hiring generations the organisation is staffed by people chosen for not being a threat.
Work expands to fill the time available, and its first axiom: an official wants to multiply subordinates, not rivals. From an essay in The Economist, later a book.
In a hierarchy every employee rises to his level of incompetence, because promotion rewards performance in the old job, not aptitude for the new one. Measured in sales organisations by Benson, Li and Shue in 2019.
Promotion as a contest in which pay depends on rank. It motivates effort and, at steep prizes, sabotage; a strong colleague becomes a rival rather than a resource.
Workers who report the opinion the evaluator already holds, because subjective evaluation rewards agreement. A cost of the evaluation method, not a character flaw.
The tendency to keep others from surpassing oneself on the dimension one is strong in, shown in recommendations and in choosing teammates.
The perception that another person's rise lowers one's own standing. Predicts undermining and abusive supervision of high performers, more strongly in people who value hierarchy.
One party (the firm) depends on another (the manager) whose actions it can only partly observe and whose interests differ. The gap is filled by whatever the firm can measure, and that decides what the manager does.
A manager keeping a strong employee from a transfer or promotion because the unit, and the manager, would lose. One of the forms of top-down sabotage counted in the 2025 Harvard survey.
The number of people who report directly to one manager. It widened from about three to nearly six in small US firms between 2019 and 2024, which changes how much a manager can judge and how much he must trust.
Charm toward superiors, pressure toward subordinates. The everyday name for a manager whose measure is what the level above him sees.