In 1982, a social scientist named James Q. Wilson and a criminologist named George Kelling published an article in The Atlantic that would become one of the most influential and contested ideas in American urban policy for the next four decades. It was called "Broken Windows."
The article's central metaphor was simple. Imagine a building with a few broken windows. If the windows are not repaired, people who pass by conclude that no one cares about the building, and no one is in charge. Someone breaks more windows. The signal of disorder invites further disorder. Given enough time, the building may become a target for occupation by squatters, and the surrounding block may descend into a spiral of neglect, petty crime, and then serious crime.
The metaphor extended beyond buildings to neighborhoods. Wilson and Kelling argued that visible signs of social disorder -- not just broken windows, but graffiti, litter, public drunkenness, aggressive panhandling, rowdy teenagers on corners -- send the same signal. They tell residents that the informal social controls that normally maintain order have broken down. They invite further disorder from the outside. And they may ultimately invite serious crime by communicating to predatory individuals that a place is easy prey.
Few ideas in criminology have been so widely adopted, so aggressively applied, and so thoroughly disputed.
The Original Argument
Wilson and Kelling's 1982 article was not a research paper presenting new data. It was an essay drawing on existing research, personal observation, and theoretical reasoning to argue for a different approach to policing and urban order.
Their central claims:
Fear and disorder are related, but not in the way most people assume. The relationship between crime and disorder is not just that crime causes fear; disorder itself causes fear and social withdrawal, which then reduces the informal social controls that communities use to maintain order. Residents who fear disorder stop walking the streets, stop intervening in minor problems, stop knowing their neighbors. This withdrawal weakens community.
Minor disorder signals that serious disorder is permissible. When visible signs of disorder go unaddressed, they communicate to potential offenders and troublemakers that the area has low informal social control, and that their behavior will not be challenged or sanctioned. This communication can be self-fulfilling.
The foot patrol officer is not just a crime fighter; he is a norm maintainer. Wilson and Kelling described the work of foot patrol officers in Newark, who were not primarily arresting people but were maintaining a kind of social contract -- establishing who the regulars were, setting informal rules for public behavior, enforcing a shared understanding of order. This norm maintenance, they argued, was valuable independent of its effect on crime statistics.
The decline of this norm-maintenance function has social costs. As policing became increasingly focused on rapid response to serious crime (the patrol car model), the function of maintaining social order in public spaces atrophied. The broken windows argument was partly a call to restore that function.
The article explicitly did not argue for aggressive crackdowns on petty crime. Wilson and Kelling warned against harassment and excessive enforcement and emphasized that foot officers maintaining order needed to operate with community consent and within legal constraints. This nuance was substantially lost in subsequent policy applications.
The New York City Application
The most prominent and contentious application of broken windows theory was the New York City policing strategy adopted under Mayor Rudy Giuliani and Police Commissioner William Bratton beginning in 1994.
Bratton had developed his order-maintenance philosophy during earlier work with the New York City Transit Authority, where he focused on fare evasion in the subway -- a broken windows type of disorder in an enclosed public system. He argued that addressing minor violations sent a signal of order and also produced operational benefits: people arrested for fare evasion were sometimes found to be carrying weapons or wanted on other charges.
The citywide strategy extended this logic. Officers targeted quality-of-life offenses: fare beating, graffiti, squeegee operators who cleaned windshields at intersections and expected payment, public drinking, panhandling, and minor drug offenses. Simultaneously, Bratton introduced CompStat -- a data-driven system for mapping crime and holding precinct commanders accountable for crime rates in their areas.
New York's crime rate fell dramatically through the 1990s. Homicides fell from roughly 2,400 in 1990 to fewer than 700 by the end of the decade. Major crime categories across the board declined sharply. Proponents of broken windows policing pointed to this as vindication.
"The theory was that if you dealt with the low-level stuff -- the fare beating, the graffiti, the drinking in public -- you sent a message to the entire criminal element that the city was not going to tolerate disorder. And then the bigger crimes didn't happen." -- William Bratton, interview in Governing magazine
The Problem With the NYC Evidence
New York's crime decline was real and substantial. The broken windows attribution is contested because crime fell everywhere in the United States during the same period -- in cities that adopted order-maintenance policing and cities that did not, in cities with falling unemployment and cities with stable unemployment, in states that built more prisons and states that did not.
Possible alternative explanations for the national crime decline include:
- Demographic change: The crack epidemic's peak in the late 1980s was followed by stabilization; the population of young men aged 15-24 (the highest-crime demographic) was relatively stable or declining
- Incarceration increases: The US incarceration rate approximately tripled from 1980 to 2000; incapacitation effects are plausible but disputed
- Economic improvement: Unemployment fell substantially through the 1990s
- Lead abatement: Researcher Rick Nevin and others have documented a strong correlation between childhood lead exposure (associated with gasoline and paint) and crime rates approximately 20 years later; the removal of lead from gasoline in the 1970s-1980s may have contributed to crime declines in the 1990s-2000s
- Policing strategies: Beyond broken windows specifically, various operational and strategic changes across departments may have contributed
The problem for broken windows advocates is isolating New York's specific policing strategy as the cause when so many things changed simultaneously, and when similar crime declines occurred elsewhere without similar strategies.
Bernard Harcourt's Critique
The most rigorous and comprehensive academic critique of broken windows theory was developed by legal scholar Bernard Harcourt in his 2001 book Illusion of Order: The False Promise of Broken Windows Policing.
The Empirical Critique
Harcourt examined the empirical studies cited in support of broken windows theory and found them methodologically weak. The most frequently cited study, by Wesley Skogan (1990), found correlations between disorder and robbery after controlling for other variables. Harcourt's reanalysis found that when one outlier neighborhood (the data from which Skogan reported was apparently anomalous) was removed from the analysis, the relationship disappeared. Other studies that found disorder-crime links used cross-sectional designs that cannot establish causal direction.
The fundamental problem: disorder and crime are correlated. That correlation does not tell us whether disorder causes crime (the broken windows claim), whether crime causes disorder (places with high crime attract less investment, more transience, less social control), or whether both are caused by a third factor (poverty, segregation, housing decay, institutional disinvestment). Most natural correlational studies cannot distinguish these possibilities.
Harcourt argued that the evidence base for broken windows as a policy guide was far weaker than its policy advocates acknowledged.
The Definitional Critique
More fundamentally, Harcourt argued that the category of "disorder" used in broken windows theory is not objective. What counts as disorder -- panhandling, loitering, drinking in public, "aggressive" behavior -- is socially defined and selectively enforced.
The enforcement of order-maintenance laws is not neutral. It tends to fall disproportionately on the poor, on the homeless, on Black and Hispanic residents, and on others who use public space because they lack access to private space. The "disorder" that the broken windows approach targets is substantially the public presence of poor people -- a presence that more affluent residents find uncomfortable and that their political power can mobilize police resources against.
Harcourt documented the racial and socioeconomic disparities in quality-of-life enforcement and argued that what is described as crime prevention functions primarily as social order enforcement -- enforcing the invisibility of poverty and marginalized communities, not actually preventing crime.
The Cost Accounting Critique
Even if order maintenance reduced some crime, Harcourt argued, the policy question requires asking what the costs are and who bears them. The costs include:
- Direct harms to those targeted: Arrest records, fines, loss of employment, disruption of lives for minor violations
- Damaged police-community relations: Aggressive order maintenance erodes trust between police and the communities being policed, which itself impedes crime solving (witnesses and victims cooperate less)
- Opportunity costs: Police resources devoted to order maintenance cannot simultaneously be devoted to serious crime investigation
Whether the benefits (if any) exceed these costs for the specific populations bearing them is a question broken windows theory does not address.
What the Evidence Actually Shows
The empirical literature on broken windows theory is large, methodologically varied, and genuinely contested. Several strands are worth examining.
Experimental Evidence on Disorder Spread
Kees Keizer and colleagues' 2008 study published in Science provided some of the strongest controlled experimental evidence that disorder can spread disorder. In a series of field experiments in Groningen, the Netherlands, they manipulated the presence of visible disorder and measured subsequent rule-breaking behavior:
- A bike locked to a fence with a "no bikes" sign: researchers put irrelevant flyers on all bikes in the area. When an adjacent wall had graffiti on it (disorder condition), 69% of cyclists took the flyer and dropped it on the ground (littering). Without the graffiti, 33% littered.
- Similarly, fireworks sounds suggesting that someone had set off illegal fireworks increased subsequent envelope theft from an unlocked mailbox
These experiments support a behavioral contagion mechanism: the presence of one norm violation reduces the perceived social cost of committing another. People use environmental cues about norm enforcement to calibrate their own behavior.
This is interesting and credible evidence for a psychological mechanism. But it does not establish that order-maintenance policing is an effective or equitable crime prevention strategy. The experimental context is deliberately controlled in ways that strip out most of the complicating factors of actual urban neighborhoods.
Quasi-Experimental Studies
Several studies have tried to use natural variation in policing intensity to estimate effects:
- A study by Braga, Welsh, and Schnell (2015) meta-analyzed 30 studies of "problem-oriented policing" (which overlaps with but is not identical to broken windows policing) and found generally positive effects on crime reduction. However, problem-oriented policing is a broader, more tailored approach than pure order maintenance.
- Weisburd and colleagues have found consistent effects of focused deterrence strategies in hot spots policing -- concentrating police presence in small, high-crime micro-places -- with less dependence on the order-maintenance logic.
- Studies specifically examining the effects of misdemeanor and quality-of-life enforcement have generally found weak or inconsistent effects on felony crime rates.
The Fundamental Counterfactual Problem
The core methodological problem with all broken windows research is the difficulty of constructing an adequate counterfactual: what would have happened in New York in the 1990s without the policing changes? Given that crime fell everywhere, including places without those changes, the estimated contribution of the broken windows strategy to New York's specific decline is genuinely uncertain.
A 2019 study by MacDonald, Fagan, and Geller that used a natural experiment created by a US District Court ruling limiting stop-and-frisk found that reducing intensive order-maintenance stops did not increase crime in New York. This is evidence against a strong broken windows effect, though stop-and-frisk is a specific tactic rather than the entirety of the strategy.
Order Maintenance vs. Zero Tolerance
A critical distinction frequently collapsed in public debate is between order maintenance policing (as Wilson and Kelling described it -- community-embedded, norm-focused, relationship-based) and zero tolerance policing (aggressive misdemeanor enforcement, as actually implemented in many cities).
Wilson and Kelling's original article emphasized the role of the foot patrol officer who knew the neighborhood, negotiated with residents about community norms, and operated with community consent. This is very different from a policy of systematically arresting everyone who jumps a turnstile or sits on a sidewalk, regardless of context or community norms.
Zero tolerance policing is both more operational (quantifiable targets) and less contextual than the original theory. It is also more prone to the disparate enforcement Harcourt documented. Many of the harms attributed to broken windows policing are more specifically the harms of zero tolerance policing implemented in the name of broken windows theory.
The Racial Justice Dimension
The intersection of order-maintenance policing with racial justice is impossible to separate from the empirical debate. The documented disparities in stop-and-frisk, quality-of-life arrests, and misdemeanor enforcement show consistent racial disproportionality that cannot be fully explained by differential crime rates.
New York City's stop-and-frisk program, which peaked at over 685,000 stops in 2011, was approximately 84% stops of Black and Hispanic individuals. The federal court found in Floyd v. City of New York (2013) that the program violated the Fourth and Fourteenth Amendments due to both unreasonable searches and discriminatory application.
The broader point is about how the definition of "disorder" operates in practice. The behaviors targeted by order-maintenance policing -- loitering, public space use, street vending, visible poverty -- are not race-neutral behaviors. They are behaviors that vary by socioeconomic status, by neighborhood context, and by cultural practice, and their selective enforcement reflects existing social hierarchies rather than objective assessments of threat.
Broader Applications Beyond Policing
The broken windows logic has been applied well beyond its original criminological context, with varying degrees of theoretical justification:
Organizational Culture
The claim that small cultural violations predict larger ethical failures has intuitive appeal in organizational management. Research on ethical cultures in organizations does suggest that norms operate at multiple levels and that tolerance for small violations can signal that larger violations will also be tolerated. However, the causal mechanism is different from urban crime: organizational norm violations are more clearly visible to the relevant norm enforcers, the social dynamics are more transparent, and the feedback loops are faster.
Software Development
The software development community uses "broken windows" as a metaphor for technical debt: the idea that leaving code messy, tests failing, or documentation outdated signals that quality is not a priority and invites further degradation. The metaphor is descriptively useful. The causal mechanism -- visible disorder invites further disorder -- has some support in how development culture shapes individual behavior. The strength of the effect and the optimal policy response are less well established than the metaphor suggests.
Public Health
Environmental disorder (lack of maintained green space, broken equipment, poor lighting) has been associated in epidemiological studies with worse health outcomes, mental health, and community cohesion. The causal direction here is genuinely complex -- unmaintained environments reflect disinvestment that itself causes health disparities through multiple pathways -- and "clean up the environment" as a health intervention has mixed evidence.
What the Evidence Supports
A fair summary of the research:
- There is credible experimental evidence that visible disorder increases norm-violating behavior through behavioral contagion mechanisms
- The empirical evidence that order-maintenance policing reduces serious crime is weak, contested, and methodologically difficult to establish
- The racial and socioeconomic disparities in how order-maintenance policies are applied are well-documented and substantial
- Crime in New York City fell dramatically in the 1990s, but similar declines occurred nationally, making the broken windows attribution uncertain
- Wilson and Kelling's original argument was more nuanced than its policy applications, emphasizing community consent and relationship-based policing that was largely abandoned in practice
- The "disorder causes crime" versus "crime causes disorder" versus "both are caused by underlying social conditions" question is unresolved by current evidence
The broken windows idea captured something real: social environments shape behavior, norms are communicated through visible signals, and disorder can be self-reinforcing. The policy conclusions drawn from those observations -- aggressive enforcement of minor offenses as crime prevention -- went well beyond what the evidence supported and imposed serious costs on already-disadvantaged communities.
That gap between an interesting theoretical insight and a specific policy application is perhaps the most important lesson of the broken windows story.
| Claim | Evidence status |
|---|---|
| Visible disorder spreads norm violations (experimental) | Moderate support (Keizer et al. 2008) |
| Order maintenance policing reduces serious crime | Weak and contested |
| Broken windows explains NYC 1990s crime decline | Unproven; alternative explanations plausible |
| Order maintenance is applied equitably | Clearly false; documented racial disparities |
| The broken windows mechanism is useful in non-policing contexts | Plausible but weakly tested |
Frequently Asked Questions
What is the broken windows theory?
Broken windows theory is a criminological hypothesis proposed by James Q. Wilson and George Kelling in a 1982 Atlantic article arguing that visible signs of social disorder -- graffiti, broken windows, litter, public drinking, loitering -- create conditions that signal to potential offenders that an area is uncontrolled, invite further disorder, and eventually lead to more serious crime. The theory suggests that addressing minor disorder aggressively can prevent the escalation to serious crime by maintaining social order signals.
How was broken windows theory applied in New York City?
Under Mayor Rudy Giuliani and Police Commissioner William Bratton beginning in 1994, New York City adopted a policing strategy explicitly based on broken windows theory. Officers targeted minor quality-of-life offenses -- fare evasion, graffiti, public drinking, panhandling, squeegee men -- alongside statistical crime mapping (CompStat). New York's crime rate dropped dramatically through the 1990s. Proponents credited the policing strategy; critics pointed out that crime fell everywhere in the US during this period, driven by demographic and economic factors, not specifically New York's tactics.
What is Bernard Harcourt's critique of broken windows theory?
Bernard Harcourt, in Illusion of Order (2001), argued that broken windows theory lacks rigorous empirical support, that the studies often cited in its favor suffer from methodological problems, and that the policy implications are discriminatory and harmful. He showed that the categories of 'disorder' used to justify enforcement are not objective but are socially constructed and disproportionately applied to poor and minority communities. The result, he argued, is a form of order maintenance that functions as social control of marginalized groups rather than genuine crime prevention.
What does the research evidence actually show about broken windows theory?
The evidence is mixed and contested. Kees Keizer's Netherlands study (2008) found experimental support for disorder spreading disorder in controlled settings. George Kelling and others cite several city studies suggesting that order maintenance reduces crime. However, rigorous studies with appropriate controls have generally found weaker effects than advocates claim. A key problem is reverse causation: disorder and crime are correlated, but it is unclear which causes which, and both may be driven by underlying social conditions (poverty, segregation, disinvestment) that order maintenance does nothing to address.
What are the broader applications of broken windows thinking beyond policing?
The broken windows logic -- that small signs of decay or norm violation signal permission for further violation and decay -- has been applied to organizational management (small protocol violations predict larger ethical failures), software development (technical debt and messy codebases invite further degradation), community management, school discipline, and public health. Whether the theory holds in these contexts with the same strength as in its urban criminology application is an open empirical question, though the psychological mechanism -- that visible norms shape behavior -- has support in social norm research independent of the contested crime causation claims.