Police leaders, local officials, and analysts have been sounding the alarm about “de-policing,” including mass resignations and retirements in America’s police departments, driven by the increasing challenges of the job in the wake of last summer’s anti-cop protests. Now the Marshall Project, a media organization devoted to criminal-justice reform, has some news: data say that the de-policing trend is all made up.
Worries about de-policing are “unfounded,” claim the Marshall Project’s Weihua Li and Ilica Mahajan. They cite federal labor-force data showing, they say, a 1 percent decline in local police employment in 2020, compared with an overall 6 percent drop across the rest of the economy. The analysis has attracted the attention of the usual cadre of police critics eager to insist that de-policing is just a union talking point.
But the evidence for de-policing is not just a survey or exaggerated claims from unions. It’s the individual reports of large-scale retirements, resignations, and other departures from multiple big-city departments. Last November, I documented that at least half of America’s 50 biggest cities had seen staffing drops; in some cities, like Minneapolis, the situation appears to have worsened since then. And recent research based on data from a “large police department in the western United States” explicitly links a large drop in staffing to the post–George Floyd protests. And as the Associated Press just reported, the Baltimore Police Department cannot comply with a federal consent decree meant to reduce police misconduct because it is dramatically below staffing levels.
So what to make of the Marshall Project’s data, which seem to show something different? The short answer: these data are a poor measure of actual beat-cop staffing levels.
To get their numbers, the article’s authors use data from the Quarterly Census of Employment and Wages (QCEW), a workforce survey conducted regularly by the Bureau of Labor Statistics. In particular, they looked at reported numbers of local, state, and federal employees classified as working in police protection (NAICS code 92212).
That classification covers a swath of law-enforcement-related activity. This includes sworn officers, but it also counts various federal law-enforcement officers (who account for about 21 percent of the total) and sometimes fire departments. It also does not distinguish sworn officers from civilian personnel, though the latter accounted for about a quarter of state and local police employees, as of the most recent Census data.
Because the QCEW data are not appropriately disaggregated, they cannot give us a true sense of the change in employment levels of actual officers on the beat. There are too many other contributors: the magnitude of the sworn-officer decrease could be masked by the larger denominator of all non-sworn-officers counted, and increases in other categories would offset any decrease in the total.
Aggregate data also ignore the mechanics of the de-policing trend. Many officers have left big-city forces not for other jobs, but to go work in the suburbs. That’s not surprising—rather than trying to switch careers altogether, cops are leaving jurisdictions that don’t want them for jurisdictions that do. Such outflows would produce the dramatic decline in force readiness experienced by cities like New York and Minneapolis without changing the overall number of cops.
This explains why de-policing could drive large increases in violent crime—what I and others have called the Minneapolis effect—even if the total number of cops remains constant. Because the large majority of violent crime takes place (almost necessarily) in big cities, a shift of manpower away from them and into the relatively more peaceful suburbs and rural regions will lead to an overall increase in violence.
Data journalism often carries the air of unimpeachability, as though a selection of numbers tells the whole story. But in this case, it doesn’t. In their rush to disprove a hypothesis that challenges their priors, the Marshall Project gets the facts wrong.
Editor’s note: The Marshall Project has submitted a response disputing certain statements in this article.
Photo by Spencer Platt/Getty Images