What happens when policing pulls back: the evidence on crime
Police presence, particularly proactive, targeted policing, reduces crime, and reductions in that presence, whether through budget cuts, withdrawal from proactive work, or officer demoralisation, are consistently followed by increases in crime, especially violent crime. The people most harmed by red
The debate over police funding tends to run on instinct rather than evidence, from both directions. The actual research on policing and crime is more settled than the political argument suggests, and it points somewhere specific.
What the research says about policing that works
The most rigorous modern synthesis of policing research, from the US National Academies of Sciences, found that several specific proactive strategies consistently reduce crime. Hot spots policing, concentrating resources on the locations where crime clusters, produces short-term reductions without simply displacing crime to surrounding areas, contrary to the "balloon effect" critics feared. Problem-oriented policing, addressing the underlying conditions that generate crime at a location rather than just responding after the fact, produces measurable short-term reductions. Focused deterrence, targeting high-risk individuals and groups with clear, communicated consequences for continued offending, consistently reduces gang violence, drug market violence and repeat offending. Increased visible patrol reduces motor vehicle theft, property crime, violent crime and gun crime, and also reduces the volume of calls for service in the areas it covers.
What happened when proactive policing retreated
The period following the 2020 death of George Floyd produced an unplanned natural experiment in what happens when police pull back from proactive engagement. The United States recorded 4,901 more murders in 2020 than in 2019, the largest single-year increase in murder since the FBI began keeping national statistics in 1960. Eight of 28 major cities examined experienced homicide increases of 100 percent or more in 2021 compared to their pre-pandemic five-year average, concentrated in cities where police visibly pulled back from proactive engagement. Researchers have termed this the Minneapolis Effect: a documented pattern where public anti-police sentiment causes officers to reduce proactive engagement even without any formal budget cut, and the deterrent effect that engagement previously provided simply disappears along with it.
The defunding movement that mostly didn't defund
A detail regularly missing from the debate is that actual budget defunding largely did not happen, and where it did, it was reversed quickly. Across more than 100 US cities and counties examined, 83 percent were spending at least 2 percent more on police in 2022 than in 2019, despite the prominence of defunding rhetoric. Cities that made token cuts, including Minneapolis and Portland, reversed them within one to two years as crime rose. Portland's police bureau remained fully funded in every year except a brief exception in 2020. The political movement to formally defund largely failed at the level of actual budgets, but the rhetoric itself, and the resulting demoralisation that reduced officer willingness to engage proactively, produced real crime effects independent of any formal funding change.
Who actually bears the cost
This is the part of the debate most consistently left unstated. The communities with the highest violent crime rates are overwhelmingly lower-income and disproportionately minority, and residents of those communities consistently report in surveys that they want more police presence, not less. The most vocal advocates for defunding have typically lived in low-crime, high-income areas with minimal direct exposure to the consequences of reduced policing. Reduced policing is, in practice, a policy whose costs fall on the most vulnerable residents while its advocates are frequently among the best protected from those costs.
The genuine trade-off, and Australia's version of it
None of this means every policing practice is effective or that documented problems, excessive force, racial disparity in stops, don't exist; both are real and the evidence on proactive policing's effectiveness doesn't erase them. The honest trade-off sits between the harms of aggressive policing and the harms of under-policing, and a defunding argument that acknowledges only the first while ignoring the second is not a complete argument. Australia's version of this debate has been less acute than the American one, but it surfaces in police resourcing debates across every state, particularly around First Nations communities, where over-policing, reflected in incarceration rates, and under-policing, reflected in victimisation rates, can be simultaneous problems in the same community, a nuance the evidence on aggregate policing effectiveness doesn't neatly resolve.
Frequently asked questions
Does more policing always mean less crime?
The strongest evidence is for specific, proactive strategies, hot spots policing, focused deterrence, targeted patrol, rather than police numbers in the abstract. What the evidence clearly shows is that withdrawal from these strategies is followed by crime increases.
Did American cities actually cut police budgets during the defunding movement?
Mostly not, in aggregate. The large majority of cities examined were spending more on police in 2022 than in 2019. The crime effects that followed 2020 are attributed more to reduced proactive engagement by demoralised officers than to formal budget cuts.
Who is most affected by reduced policing?
Residents of higher-crime, lower-income communities, who are also the residents most likely to report in surveys that they want more police presence rather than less.
Does this evidence apply directly to Australian policing debates?
Partly. The core finding, that proactive policing reduces crime and withdrawal increases it, generalises, but Australia's specific debates, especially around First Nations communities, involve simultaneous over-policing and under-policing that the general evidence doesn't fully address.