Loading...
By 1984, New York City's subway system had become a rolling philosophy experiment about whether environments shape moral behavior. Every one of the system's 6,200 cars was covered in graffiti — not patches, but total coverage, floor to ceiling, inside and out. Riders sat in dim, stinking cars, averting eyes from broken seats and pools of unknown liquid. Fare evasion ran at roughly 250,000 turnstile jumps per day. Felonies on the subway had risen steadily for a decade, reaching 15,000 per year by 1990. Sociologist George Kelling and political scientist James Q. Wilson had published 'Broken Windows' in The Atlantic in March 1982, arguing that visible disorder — a single unrepaired window, one abandoned car — signals to a community that no one is responsible, no one is watching, and that t...
Popular framing: Tough-on-crime policing finally cleaned up New York.
Structural analysis: Visible disorder broadcasts a social-proof signal that norms have been suspended, licensing escalation through a feedback loop where each broken window invites the next. The Clean Car Program inverted the signal — graffiti immediately erased, fare-beating immediately arrested — re-establishing the perception of consensus before the underlying behavior changed. The mechanism isn't punishment; it's environmental argument. Goodhart risks lurk on the other side: optimizing the visible signal can drift from the goal it was meant to proxy.
The gap matters because broken windows theory is highly actionable (clean cars, arrest fare-evaders) while structural accounts are diffuse and politically costly (rezone, invest, remediate). Policymakers face strong incentives to adopt the legible intervention and claim credit for correlated outcomes, even when causation is unclear. This selection bias means the dominant policy frame will systematically favor visible, short-cycle interventions over structural ones — not because evidence favors them, but because feedback loops in political systems reward them.
