Long-hours arms race
A Socratic walk-through of the long-hours arms race — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does everyone stay late when nobody gains from staying later than the rest?
Ask an office at seven in the evening whether the extra two hours are producing much, and almost nobody will say yes. Ask whether they would prefer an office that emptied at five, and almost everybody will. Ask why they are still there, and you get a shrug.
That shrug is the interesting part. A room full of people, each of whom would be better off under a different arrangement, all choosing the arrangement none of them wants. If everyone stayed an hour later tomorrow, nobody's standing would improve — so what keeps the hour there?
Reasoning it through
REASONING #Begin with what the extra hours are actually for. Not output: if they were straightforwardly output, cutting them would show up as work not done, and the honest answer in most offices is that it would not, or not proportionally. So try the other possibility — that the hours are being read rather than used. They are evidence about the person.
Evidence of what? Of the things a manager badly wants to know and cannot observe: commitment, willingness, whether this person will still be here when it gets hard. Effort is hard to see directly. Presence is trivially easy to see. So presence gets used as a proxy for effort, and hours become a signal.
Now ask what makes a signal work. It has to cost something, and it has to cost more for the people you want to screen out. Staying late is genuinely costly — an evening, a meal, a child's bedtime — and plausibly costlier for someone with weaker attachment to the job. So it functions as a signal, however wastefully. Good so far.
But here is the step that produces the puzzle. What the signal buys is not absolute; it is relative. Promotions, the interesting project, the protected headcount in a cut — these are ranked and rationed. There is one of them. So what matters is not how many hours you worked but whether you worked more than the person beside you.
Follow that through for one person. Whatever everyone else does, staying an extra hour moves you up the ranking a little. So the extra hour is worth taking. Now notice that this reasoning is available to everybody, and does not depend on what anybody else chose — it is individually correct for each person, separately, no matter what the others do. Everyone takes the hour. The ranking is exactly what it was before. Everybody has paid the cost and nobody has bought anything.
And it does not stop at one hour, because the argument repeats from the new baseline. The escalation runs until the private cost of the marginal hour finally exceeds the private benefit of the small rank improvement — an equilibrium, but a bad one, sitting well past the hours anyone would have chosen collectively.
Then ask why nobody simply fixes it, since everybody can see it. Because the fix requires simultaneous agreement. If the office agreed to leave at five and you alone leave at five, you are fine; if you alone are believed to have defected upward, you are the one who is seen to be trying. The agreement is unenforceable by the people it would benefit, which is exactly the shape of a collective-action problem: an outcome everyone prefers, that no individual can move toward alone.
One honest caution. Some late hours really are production, not signalling — deadlines are real, and in some trades output genuinely is close to linear in time. The arms-race account applies where hours are observed and output is not, which is why it is so much stronger in white-collar offices than on a piece-rate site where the count of finished pieces is on the bench at the end of the day.
The analogy
THE ANALOGY #Think of a crowd at a standing concert. One person rises onto their toes and sees better. Their neighbours, now blocked, do the same. Within a minute the whole hall is on tiptoe, everyone is uncomfortable, and every single person sees exactly what they saw when all were flat on their feet. Nobody behaved irrationally; each move was correct given the others.
a crowd can be reset by one usher telling everyone to stand normally, because the view is purely relative, whereas some late hours do produce real work — so an office cannot be flattened to a common level without asking which part of the standing was tiptoe and which part was height.
Clarifying the model
THE MODEL #Two corrections, because the usual readings of this go wrong in opposite directions.
The first: this is not a story about a bullying employer. The employer often loses too — tired staff make expensive errors, and the firms that have cut hours without cutting output are evidence that the extra time was not buying much. The equilibrium is produced by the rationing of promotion and the observability of presence, and it can persist with a manager who sincerely wants people to go home. What her wishes cannot do is change what her staff believe she will notice.
The second: it is not straightforwardly irrational either. Calling the late-stayers foolish misses that each of them is responding correctly to a real ranking. The waste is structural. That is why the effective remedies are structural too, and why they all work by breaking the link between presence and the reward: measuring output where output can be measured, making the hours invisible so they cannot be ranked, or imposing a ceiling from outside that removes the individual choice altogether.
Notice that the last of these — a rule everyone must obey — is the standard escape from a collective-action problem, and that its unpopularity is a good sign rather than a bad one. People object because a binding limit takes away a move they would individually like to keep. That objection is the mechanism confessing.
What would refute the account? If long-hours norms were strongest where individual output is easiest to measure, the signalling story is wrong, and something else — workload, or simple custom — is doing the work.
A picture of it
THE PICTURE #How to readRead across for what the rest of the office does and up for what you do. The bottom-left corner is the arrangement everyone says they would prefer — but from there, moving straight up into the keen corner is individually attractive, and moving right is what happens once your colleagues reason the same way. Both single moves lead away from the corner everyone wanted and into the top-right, where the ranking is identical to the bottom-left and the cost is far higher. Nobody can travel back down alone, because leaving early while others stay late puts you in the bottom-right.
What became clearer
WHAT CLEARED #The hours are not buying output and they are not buying rank either — once everyone has bought the same amount of them, rank is unchanged. They are the price of not being the one who declined to pay. That makes the long-hours office a collective-action failure rather than a culture of workaholics, and it explains why exhortation never fixes it while a rule that removes the individual choice usually does.
Where to go next
ONWARD #- What happens to the signal when the office is remote and presence stops being observable — and what gets used as the substitute proxy instead.
- Why firms that shortened the week saw output hold, and what that says about how much of the marginal hour was ever production.
Key terms
TERMS #| Term | What it means |
|---|---|
| Collective-action problem | a situation where an outcome everybody prefers cannot be reached, because no individual can move toward it without being worse off while the others have not. |
| Signalling | taking a costly, visible action to convey something about yourself that cannot be observed directly; it works only if the cost is higher for those you want to distinguish yourself from. |
Every term the collection defines is gathered in the glossary.