American police departments spend $115 billion a year. They solve fewer than half of violent crimes. The question is not whether we need public safety — it is whether what we have built actually delivers it.
The United States has approximately 19,000 law enforcement agencies employing over 800,000 officers. Annual spending exceeds $135 billion — more than the entire GDP of many countries. Compared to peer nations, American policing is uniquely decentralized, uniquely armed, and uniquely lethal. That combination has consequences the data makes visible.
Qualified immunity is a judicial doctrine created by the Supreme Court in 1967 and dramatically expanded in 1982 — with no basis in the text of federal civil rights law. It protects officers from civil liability unless a prior court case established nearly identical facts. In practice, it makes it nearly impossible to sue a police officer for misconduct, no matter how severe. It did not come from Congress. It came from nine justices deciding to write it into existence.
Officers have been granted immunity for: shooting a 10-year-old while trying to shoot a dog, using a police dog to maul a surrendered suspect, stealing $225,000 while executing a search warrant, and shooting an unarmed man 4 times in the back. In each case, the court found no prior case with sufficiently similar facts to put the right "beyond debate."
Racial disparities in policing exist at every point of contact: traffic stops, searches, use of force, and fatal encounters. These disparities persist after controlling for neighborhood crime rates, socioeconomic conditions, and officer characteristics. They are not fully explained by differential behavior. They are substantially explained by differential treatment — a finding replicated across dozens of peer-reviewed studies using national data.
The "veil of darkness" finding is among the most controlled in policing research: during daylight hours, when officers can see race before stopping a driver, racial disparities in traffic stops are significantly larger. At night — when race is less visible — the gap narrows. This is not about crime rates. It is about perception at the moment of decision.
Civil asset forfeiture allows law enforcement to seize property — cash, cars, homes — without charging the owner with a crime. The owner must then prove their innocence in civil court to get their property back. In 2019, the federal government and states collectively seized more through civil forfeiture than all U.S. burglars stole combined. This is not a bug in the system. For departments that keep a portion of what they seize, it is the incentive structure.
Since 1997, the Department of Defense 1033 Program has transferred over $7.4 billion in military equipment to local law enforcement agencies — including armored vehicles, grenade launchers, aircraft, and rifles designed for warfare. The research on whether militarized policing reduces crime is clear: it does not. The research on whether it increases use of force complaints and civilian distrust is equally clear: it does.
The evidence on public safety is broader than the policing debate. Most of the factors that predict crime — unemployment, housing instability, childhood poverty, mental health access, and neighborhood disinvestment — are not addressed by patrol officers. Cities that have shifted resources toward these upstream conditions show measurable crime reductions. The choice is not police vs. no police. It is how we build communities where fewer people are desperate enough to harm each other.
| Approach | Evidence Strength | Crime Impact | Community Trust Impact |
|---|---|---|---|
| Community violence intervention programs | Strong (multiple RCTs) | -30 to -60% shootings in target areas | Positive — community-led |
| Early childhood investment (pre-K) | Very Strong (50yr data) | -40% adult crime for participants | Positive — broad community benefit |
| Mental health co-responders (CAHOOTS model) | Moderate-Strong | Reduces unnecessary police contact | Strongly positive for served communities |
| Hot-spots policing (targeted, procedurally just) | Moderate | -20% crime in targeted micro-areas | Neutral to negative without community buy-in |
| Increased patrol saturation | Weak | Minimal measurable effect | Negative in high-distrust areas |
| Military equipment transfer (1033) | Negative evidence | No crime reduction found | Strongly negative |