Broken Windows Theory: The Evidence For and Against the Policing Strategy

A Single Broken Window, a Theory, and a Crime Drop Still Under Debate

In 1969, Stanford psychologist Philip Zimbardo conducted an experiment: he abandoned a car in the Bronx with its hood up and another car in Palo Alto. The Bronx car was stripped within ten minutes; by the end of the day it had been demolished. The Palo Alto car sat untouched for a week—until Zimbardo himself broke a window, after which it was vandalized within hours. This observation shaped James Q. Wilson and George Kelling's landmark 1982 Atlantic article "Broken Windows," which argued that visible signs of disorder signal to potential offenders that an area is unmonitored and socially abandoned, inviting further disorder and eventually serious crime. A decade later, New York City adopted broken windows policing under Mayor Rudy Giuliani and Police Commissioner William Bratton—and the city's crime rate fell by more than 70% over the following decade. Whether these two facts are connected is one of criminology's most vigorously contested questions.

The Theory: How Disorder Escalates

Wilson and Kelling's argument operates through a specific causal chain:

• A neighborhood tolerates visible minor disorder: broken windows, graffiti, public drunkenness, panhandling.
• Residents interpret the disorder as a signal that no one is in control and informal social controls have broken down.
• Law-abiding residents withdraw from public spaces, reducing natural surveillance.
• Criminals interpret the signals the same way: the area is uncontrolled and enforcement is unlikely.
• More serious crime follows as potential offenders act on these signals with reduced fear of intervention.

The policy implication: aggressively address minor disorder before it invites serious crime. This "order maintenance policing" strategy translates into zero-tolerance enforcement of minor violations—subway fare evasion, public drinking, loitering, squeegee men, graffiti removal—to restore visible signals of social control.

New York City: The Grand Experiment

When Bratton took over as NYPD commissioner in 1994, he implemented broken windows policing systematically alongside CompStat, a data-driven management system that tracked crime statistics in real time and held precinct commanders accountable for results. New York's crime outcomes in the 1990s were dramatic:

Bratton and Giuliani credited broken windows policing. But the crime drop was not unique to New York: virtually every major U.S. city saw comparable declines in the 1990s, including cities that did not adopt broken windows strategies.

The Evidence: What Research Actually Shows

The causal relationship between broken windows policing and crime reduction has been tested in multiple academic studies, with mixed results:

• Kees Keizer et al. (2008, Science): A Dutch field experiment found that people were more likely to litter and steal when surrounded by visible disorder—providing experimental support for the disorder-crime link in controlled conditions.
• Harcourt and Ludwig (2006): A reanalysis of New York data argued that the crime drop was concentrated in precinct areas regardless of disorder levels, suggesting CompStat-driven policing rather than disorder reduction drove results.
• MacDonald et al. (2016, Journal of Urban Economics): A quasi-experimental study of cleaning and greening vacant lots in Philadelphia found that addressing physical disorder reduced gun violence in surrounding areas—supporting a causal effect of environmental improvement on crime.
• National Bureau of Economic Research (2001): Economists found that the national crime drop was substantially explained by the legalization of abortion in 1973 (Donohue and Levitt), declining crack cocaine markets, and increased incarceration—factors unrelated to policing strategy.

The Civil Liberties Critique

Broken windows policing has been criticized extensively for its implementation—specifically the stop-and-frisk program that accompanied it in New York. At its peak in 2011, NYPD officers made 685,724 stop-and-frisk encounters. In Floyd v. City of New York (2013), a federal judge found that the NYPD had engaged in a pattern and practice of unconstitutional stops, with approximately 88% of those stopped being Black or Latino men who were innocent of any offense. The subsequent decline in stop-and-frisk following the Floyd ruling was not accompanied by an increase in crime—further complicating causal claims about aggressive order maintenance policing.

The broken windows debate illustrates a persistent problem in social science: major policy interventions are rarely implemented as controlled experiments, making causal inference from observational data extremely difficult. The theory's influence on policing practice has been enormous regardless of its empirical status—and its legacy includes both genuine crime reduction initiatives and policing practices that federal courts found unconstitutional.