Searching for Guns, Finding Only Your Own Bias

A man enters a New York subway station through an emergency gate, no payment. Police find a loaded gun on him.

Another man at a different stop avoids the turnstile with forty-eight decks of heroin and a stolen, loaded firearm in his pocket. A third fare evader carries the same.

The pattern feels clean, almost inevitable. Catch the small crime and you catch the big one. Minor infractions become a proxy for serious violence — stop enough people for turnstile jumps and you prevent murders.

William Bratton and George Kelling sold this exact mechanism to New York City in the early 1990s. The theory was structural — disorder signals abandonment, and abandonment invites predation. Clean up the graffiti, arrest the squegee men, and the causality felt tight, almost mathematical. But across 25 years of NYPD stop-and-frisk records, researchers found that officers were stopping Black and Latino New Yorkers at rates that bore no relationship to crime occurrence in their neighborhoods. The stops proliferated most aggressively in places where crime was dropping, not rising.

You will find more guns if you search more people.

What looked like detection was actually selection bias dressed in methodology. You find guns in the pockets of men you choose to search when you search men in certain neighborhoods with certain resources at a certain rate. The fare evasion theory contains the same invisible hinge. It assumes that randomness, who evades and when, correlates with criminal intent. But policing doesn't work randomly, it concentrates. And concentration, when it's racial or economic, produces a perfect illusion of causation.

The outcome in the 1990s was measurable harm that exceeded any crime prevented. The question is not whether the correlation exists. It's whether we learned which direction it flows.