Background
In the summer of 2020, a narrative took hold: cops were quitting in droves. Op-eds declared a “retention crisis.” The Police Executive Research Forum surveyed its member agencies and reported resignations up 18% and retirements up 45% compared to the prior year — figures that ricocheted through cable news, congressional hearings, and eventually into the Biden administration’s Safer America Plan. The story had real emotional resonance. Anecdotally, officers felt demoralized, departments struggled to fill shifts, and at least in some cities, the numbers were genuinely alarming.
But we now have something we didn’t have in 2021: a body of peer-reviewed research, a new national dataset, and a job market paper with some of the strongest causal evidence yet. Five years later, it’s worth stepping back and asking what the data actually show, and where they still disagree.
For what it’s worth, I have a stake in this question. I co-authored two of the papers I’ll discuss here. I’ll try to be honest about what my own work got right, where it fell short, and where other researchers’ findings complicate the picture.
A Striking Finding in One City
The first rigorous empirical study on this question came from work I did with Scott Mourtgos and Ian Adams, published in Criminology & Public Policy in 2022. We focused on a single large western police department (with roughly 600 sworn officers) and asked whether voluntary resignations changed after George Floyd’s murder. Using a synthetic control design (essentially, we built a statistical counterfactual of what the department’s resignations would have looked like without the post-Floyd environment), we found that voluntary resignations increased by 279% relative to the counterfactual. Retirements and involuntary separations showed no significant change.
That was a striking number. It also came with an obvious limitation we acknowledged upfront: one agency is one agency. Our city experienced particularly intense unrest during the summer of 2020. Whether this pattern would hold elsewhere wasn’t something we could say.
Scaling Up: Fourteen Agencies
The natural next step was to look at more departments. In a follow-up study published in Journal of Criminal Justice, Ian, Scott, and I expanded to 14 large U.S. agencies (all with at least 100 sworn officers) tracking separations through roughly the end of 2022. Eleven of the 14 showed above-counterfactual resignations, retirements, or both.
When we added up the excess losses across all 14 departments, we got approximately 870 surplus resignations plus 559 surplus retirements — roughly 1,430 officers above what the counterfactual would have predicted. That translates to a 5.4% surplus loss of authorized strength across these agencies combined. But that average hid a lot of variation. Seattle and Chicago saw staffing declines of around 16%. Denver, Salt Lake City, and Wichita each lost between 8% and 10%. Two suburban New England agencies (Burlington and Manchester) showed no significant change at all.
The pattern was becoming clearer: this was a large-city phenomenon, concentrated in metros that experienced protests and political debates over policing.
Grunwald’s Important Corrective
Ben Grunwald’s 2024 working paper complicates this picture in ways that matter. Rather than focusing on large urban agencies, Grunwald constructed the Interstate Police Employment Database (IPED) to track nearly a million officers across ~6,800 agencies in 15 states from 2011 to 2021. His headline finding was that separations were nearly stable in 2020 (i.e., less than 1% above 2019 levels) before rising 18% in 2021. Even then, about a third of that increase reflected a pre-existing upward trend. By the end of 2021, the cumulative impact on national law enforcement employment was roughly 1%.
One percent sounds small enough to debunk the crisis narrative entirely. Some commentators undoubtedly read it that way. I don’t think that’s quite right. Grunwald’s sample is dominated by small and mid-size agencies, which is accurate to how U.S. policing is actually organized. The United States has roughly 18,000 law enforcement agencies; most are small. But the 29% of agencies with 500 or more officers that lost at least 5% of their staff by 2021 — those are the departments that police millions of people in major American cities. When one officer leaves a 12-person rural sheriff’s department and >600 leave the Chicago PD, those events are not equivalent in terms of public safety consequences.
Grunwald also finds something that challenges a competing theory: local protest intensity did not predict which agencies lost more officers. That’s a meaningful null finding. Whatever drove elevated separations, it wasn’t simply proportional to the size or duration of local demonstrations.
The National Snapshot: BJS LEMAS 2020
Earlier today, the Bureau of Justice Statistics released a new report using 2020 LEMAS data: Hiring and Retention of State and Local Law Enforcement Officers. The numbers provide useful grounding. In 2020, U.S. law enforcement agencies hired approximately 55,000 officers and separated about 57,400, for a net loss of roughly 2,400 officers nationally. There were about 64,200 sworn vacancies as of 2020 — 39,500 in local police departments alone.
To put 2020 in historical context: the prior net-loss year was 2013. There were net gains in 2003, 2008, and 2016. So 2020 was unusual, but not without precedent.
The LEMAS data also reveal an important size-based pattern in the type of separations. At large agencies (500 or more full-time officers), only 30% of separations were voluntary resignations, while 57% were nonmedical retirements. At the smallest agencies, resignations accounted for 64% of separations. This suggests the retirement vs. resignation story looks quite different depending on which agencies are being studied. This helps explain why our first one-city study (which found only resignations elevated, not retirements) and our follow-up multi-agency study (which found retirements elevated at most large agencies) aren’t actually in conflict. We were looking at different kinds of departments.
Why Officers Left: Filosa’s California Evidence
The most direct evidence on why officers left comes from Neil Filosa’s 2025 job market paper, which linked California POST records to voter registration data for roughly 44,000 officers at about 400 agencies from 2012 to 2022.
For retirement-ineligible officers (those under 50 who had no financial incentive to leave) the probability of exiting the force increased by 1.4 percentage points from 2019 to 2022, a 70% relative increase. These officers had nothing to gain financially from leaving early. The second finding is more striking. Filosa contrasted police officers with firefighters. Both groups faced the same COVID pandemic, the same labor market, and the same California conditions. But firefighter exit rates barely moved (+0.3 pp) while police exit rates diverged sharply (+2.0 pp for retirement-ineligible officers). The only factor that cleanly explains the divergence is the policing environment itself.
Filosa also finds that white and Republican-registered officers showed disproportionately larger increases in exits, which raises important questions about who is leaving, and whether voluntary departures are concentrating among officers most opposed to reform. This finding sits in tension with Grunwald’s national numbers, where women and Black officers showed slightly higher relative separation rates. These results aren’t necessarily contradictory — different states, time periods, methods, and officer populations — but they should make us cautious about any single account of who was walking out the door.
What the Evidence Shows
Nationally, the picture is less dramatic than PERF’s surveys suggested. Those surveys were voluntary, self-reported by member agencies, and almost certainly biased toward departments that were struggling. Grunwald’s 18% increase in 2021 is more credible as a national estimate. But his 1% cumulative employment effect reflects a denominator dominated by small rural and suburban departments where, frankly, not much changed.
For large city departments (the ones that police the majority of the urban American public) the story is different. A 5.4% surplus loss of authorized strength, concentrated in a handful of major cities, is not a trivial staffing problem. Some departments are still recovering. Vacancies in large cities remain elevated.
On the why question, Filosa’s police-versus-firefighter comparison is the strongest causal evidence I’ve seen: something specific to the policing environment — not COVID, not general labor market conditions — explains the divergence. Whether that’s officers responding rationally to increased scrutiny, moral injury from the political climate, or something else remains an open question.
What the data don’t support is the simplest version of either side of the political argument: neither “the crisis was manufactured to oppose reform” nor “the carnage proved reform was a mistake.” Officers left in elevated numbers from large departments because of something specific to how policing changed after 2020. Who they were, and what their departure means for the departments they left behind — that’s what I’m hoping to work on next.
