The Warning Signs Before Someone Quits

Exit interviews are the standard tool for understanding why people leave. They are also, according to more than fifty years of research, one of the least reliable.

• 69% of resignations happen inside five years. Retention is mostly an early-career problem.
• Pay is the reason everyone reaches for. Against actual departures it is a middling predictor at best.
• A signal measured on a survey predicted who resigned over the next three years. Almost no agency collects it.

They leave earlier than you think

Nearly three in ten resignations happen in the first year. Seven in ten happen inside five.¹

That reframes the problem. Most retention effort is aimed at keeping veterans a little longer. The volume is somewhere else entirely — among people you finished training eighteen months ago, who cost you six figures to produce and have not yet returned any of it.

Agency size changes the shape of it. In nationally representative federal data, voluntary resignations were 64% of separations at agencies under 25 sworn, and 30% at agencies of 500 or more, where retirements dominate instead.² A sergeant in a 40-person department and one in a 2,000-person department are not managing the same problem.

What they say is not what predicts it

The largest analysis of this question pooled 316 studies and roughly 1,800 effect sizes to test which factors actually predict voluntary turnover — not intention to leave, but people who left.³

Pay came in around 0.17 — weaker than roughly half the 57 factors tested. The authors flag it themselves: the direction is intuitive, but they write that “challenges remain to explain why this effect is not stronger.”³

Be careful with that. It does not mean pay is irrelevant, and it does not transfer automatically to policing. In one survey of 1,823 officers in a very large department, compensation did predict turnover intention.⁴ The honest reading is that pay behaves like a threshold rather than a lever: badly underpaid people leave, and paying an adequately-paid officer more does not buy proportionate loyalty.

Look at what does sit near the top. Someone already thinking about leaving. Someone actively searching. Falling commitment. Falling satisfaction. Absence creeping up. Every one of those is observable months before a resignation letter.

Why the exit interview misleads you

The problem was documented in 1969 and corroborated since: researchers compared what departing employees said in exit interviews with what the same people reported later, and found only weak agreement. People withhold, and they give socially acceptable reasons.⁵

“Better pay” is a safe thing to say to a supervisor you may need a reference from. “My sergeant never once asked what I wanted to do with my career” is not.

Policing has its own version. When one association asked agencies why officers resigned, the top answer was accepting a job at another local law enforcement agency.¹ That is a destination, not a cause. It tells you where they went and nothing about why they were looking.

There is also a structural problem. In a study of officers who had actually resigned, only 5 of 27 were even offered an exit interview. Participants called it a tick-box exercise. One said he “never really felt I got the opportunity to say my true feelings about it when I left.”⁶

The signal that showed up three years early

Almost everything published on police retention measures intention to leave. Intention is not departure, and the gap between them is where most retention research quietly fails.

One 2026 study did it properly. Researchers took survey responses from officers at a state law enforcement agency and linked them to actual resignation records over the following three years.⁷

The number itself is modest. What matters is that it existed at all, on a questionnaire, up to three years ahead of the resignation. Somebody could have seen it.

What nobody is measuring

A 2025 review of the police turnover literature reached a blunt conclusion. Research on retention strategies is “fragmented and inconclusive.” Stay interviews: “Few studies have examined the implementation.” Exit interviews: “impact remains largely unevaluated.”⁸

Consider what that means next to what agencies already collect. One study analysed 513,666 shifts and 4,868 absence events from 21 years of one department’s payroll records to study fatigue and unplanned absence.⁹ The data was there. No published study has linked police absence records to who later resigned — even though absenteeism is one of the established predictors in the chart above.

Agencies routinely run early intervention systems that flag officers at risk of a complaint. We could find nothing equivalent for attrition. We watch for the officer who might cost us a lawsuit and not for the one who is about to walk out.

What a supervisor can actually watch

The research does not hand you a validated checklist — be suspicious of anyone who says otherwise. What it does support is a shift in where you look.

1. Weight the first five years. That is where the volume is. A structured conversation at six, twelve and twenty-four months costs you an hour and lands where the risk is.
2. Ask before they are leaving, not while. Stay conversations are barely studied, which is a reason to run them carefully rather than a reason to skip them. An exit interview is a post-mortem.
3. Track disengagement as data, not mood. Absence, declining discretionary activity, withdrawn transfer or specialty interest. These are recorded already, in systems you have.
4. Treat burnout as a workforce metric. The one prospective study that linked a survey to real departures found it there. If you never measure it you will never see it coming.
5. Ask about development, specifically. Officers who resigned described lack of voice, unrecognised skills and blocked development.⁶ None of those appear on an exit form as “pay.”
6. Watch organizational stress, not just the calls. In a study of 2,669 officers, stress from the organization predicted wanting out more strongly than either trauma exposure or operational stress.¹⁰ The job is not the only thing grinding people down.

One honest caveat on all of it. These are associations, from cross-sectional research, in a field with almost no evaluated interventions. Nobody has proven that acting on any of these signals keeps a specific officer. What is established is that the signals exist and that they appear well before the resignation does.