Who leaves first
A Socratic walk-through of who leaves first — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why are the first people to leave a struggling organisation the ones it can least afford to lose?
A department starts to go wrong — a merger, a budget cut, a manager nobody can work with — and within a few months a handful of people have gone. Ask anyone who stayed and they will tell you the same thing: they were the good ones.
It sounds like nostalgia, or like a story people tell to explain their own decline. But it is reported far too consistently, across sectors that share nothing else, to be sentiment. So take it as a fact needing an explanation, and note that it does not require anyone to behave badly. Nobody sabotages anything. Everyone acts reasonably. Why does the outcome still sort itself that way?
Reasoning it through
REASONING #The trap is treating departures as a random sample — as though decline shakes a certain percentage loose and which people they are is luck. Ask instead what determines whether any particular person leaves, and the randomness disappears.
Leaving is not one decision but two: whether to go, and whether you can. Deterioration raises the first for everyone at roughly the same rate — the extra workload, the frozen pay, the meetings that go nowhere are shared. The second varies enormously, and along a very specific dimension: how good your alternatives are, which is to say how readily another employer would take you.
And there the mechanism is already finished, because the trait that gives you good outside options is largely the trait your current employer values. Marketable skill, a portable record, a network, recent relevant experience. These do not merely correlate with being useful here; they are substantially the same asset seen from outside. So a uniform push applied to a workforce with unequal exits does not remove a cross-section. It removes people in order of employability, best first.
That is the same shape as adverse selection in a market for goods, where sellers who know their item is good withdraw from a price that averages everyone — and it is worth stating in those terms, because it names the employer's error. An organisation offering the same deteriorating deal to everyone is pricing uniformly against a population that is not uniform. Those worth most take the outside offer, and the ones left are those for whom the deteriorating deal is still the best available. Average quality falls without any individual changing.
Push one step further and it stops being an event and becomes a loop. The first departures do not only remove capability; they redistribute the work onto those still there, and they signal that leaving is normal and recovery is not coming. Both raise the pressure on the remainder — and the same sort is applied again to a smaller pool. Capability falls faster than headcount, which is why organisations are so often surprised: the numbers look survivable and the work does not get done.
Albert Hirschman saw the sharpest part of this in Exit, Voice and Loyalty. The people most sensitive to quality — who notice decline first and care most about it — are also the likeliest to leave, and they are precisely the people who would otherwise have argued and pushed for a fix. Exit removes voice, selectively, from those whose voice was worth most. Decline quietly loses its own correction mechanism.
There is a timing problem too. The levers that would retain the marketable — targeted pay rises, changed roles, credible commitments about the future — are the levers a struggling employer has least access to. Uniform pay scales, common in public services, make it structurally worse: a flat award is worth most to those with the weakest alternatives.
The analogy
THE ANALOGY #It is like a queue that people can leave at any time. Nobody is expelled and nothing is aimed at anyone in particular. But the people who leave first are those with somewhere else to be, and if you look at who is still standing there an hour later, you have not got a random sample of the original queue — you have got the people with nowhere to go, and you never selected for that at all.
a queue is indifferent to who waits in it, whereas an employer actively wants some people to stay and can, in principle, offer them something — so the sorting is a failure of response as much as a fact of mobility.
Clarifying the model
THE MODEL #Two corrections keep this honest.
First, employability and value to this organisation are not the same thing. Much of what makes someone useful here is firm-specific — knowing the legacy system, the customers, why a rule exists — and that knowledge is worth a great deal internally and little on the open market, so its holders often cannot leave. The exodus is therefore biased toward portable ability rather than toward the best people flatly, and an employer losing its most mobile staff may be losing less than it fears, or more, depending on which kind of capability its work rests on.
Second, the sorting is a tendency, not a law. Mission attachment, seniority-linked pensions, family ties to a place and plain aversion to disruption all keep people who could leave. And in a slack labour market almost nobody has an outside option, so a badly run organisation can look admirably stable for years — then lose a great many people at once when hiring picks up, which is when the accumulated damage becomes visible.
The practical upshot is uncomfortable: because the loss is selective, an organisation cannot read its own condition from turnover rates. It has to look at who left.
A picture of it
THE PICTURE #How to readStart at the rounded node — one deterioration, applied equally to everybody. The first diamond is the whole argument: the same pressure produces two different answers depending on a trait the employer never chose to select on. Follow the left branch for who is removed and the right for who stays, and notice both feed the same redistribution box. The second diamond is the only genuine off-ramp, and the one a struggling employer can least often take. The long arrow back up is the loop — the question asked again of a workforce that has already had its most mobile members removed.
What became clearer
WHAT CLEARED #Nothing here is about loyalty or character. A uniform worsening applied to people with unequal alternatives produces a sorted exit, and the sorting runs along employability, which is largely the same asset the organisation was relying on. Worse, it takes with it the people likeliest to have argued for a fix, and it repeats on the smaller group left behind. The organisation experiences it as bad luck and it is nothing of the kind — it is the predictable consequence of offering one deal to a workforce whose options were never the same.
Where to go next
ONWARD #- Why targeted retention offers, though obviously effective, are often politically impossible in the organisations that need them most.
- What the same mechanism does to neighbourhoods, schools and professions rather than firms.
Key terms
TERMS #| Term | What it means |
|---|---|
| Adverse selection | the sorting that occurs when the party with better information about their own quality chooses whether to participate, leaving a worse-than-average pool behind. |
| Outside option | the best alternative available to a worker elsewhere, which sets the point at which staying stops being worthwhile. |
| Exit and voice | Hirschman's pair of responses to decline: leaving, or staying and pressing for change, where the first tends to remove the people best placed to do the second. |
| Firm-specific human capital | skill and knowledge valuable only inside one organisation, which raises internal value while lowering mobility. |
Every term the collection defines is gathered in the glossary.