Broken-windows policing is a style of policing generally associated with the broken-windows theory—namely, the idea advanced by James Q. Wilson and George L. Kelling that tolerating minor physical and social disorder in a neighborhood (such as graffiti, litter, aggressive panhandling, or turnstile jumping) encourages serious violent crime. Although the broken-windows theory itself did not compel a particular policing strategy, most policy makers interpreted the broken-windows theory as implying a form of aggressive disorder policing that has come to be known under several names, including not only broken-windows policing but also ”order-maintenance policing,” ”qual-ity-of-life policing,” and ”zero tolerance policing.” George Kelling, for instance, the coauthor of the original Broken Windows essay, suggests that the most effective way to address disorder and reduce crime is to increase the number of misdemeanor arrests. There has been a lot of social scientific research conducted to test the efficacy of broken-windows policing—including important work by Jeffrey Fagan, Bernard Harcourt, Jens Ludwig, Stephen Rauden-bush, Robert Sampson, and Wesley Sko-gan—but, to date, there is no reliable empirical support for the proposition that disorder causes crime or that broken-windows policing reduces serious crime.
The Emergence of Broken-Windows Policing
The broken-windows theory was first articulated by James Q. Wilson and George L. Kelling in a short, nine-page article titled Broken Windows that appeared in the Atlantic Monthly in 1982. The theory is premised on the idea that ”disorder and crime are usually inextricably linked, in a kind of developmental sequence” (p. 31). According to Wilson and Kelling, minor disorder (such as littering, loitering, public drinking, panhandling, and prostitution), if tolerated in a neighborhood, produce an environment that is likely to attract crime. It signals to potential criminals that delinquent behavior will not be reported or controlled— that no one is in charge. One broken window, left unrepaired, invites other broken windows. These progressively break down community standards and leave the community vulnerable to crime.
From a policy perspective, the broken-windows theory was, in principle, consistent with a variety of potential policy levers, ranging from neighborhood beautification programs to community organizing. Nevertheless, the theory was soon deployed by influential policy makers in support of aggressive misdemeanor policing and became associated with the type of policing that focuses on minor disorder offenses. New York City mayor Rudolph Giuliani was at the forefront of this development. Under the rubric of the ”quality-of-life initiative,” Giuliani and Police Commissioner William Bratton began enforcing municipal ordinances against minor disorder offenses, such as loitering, panhandling, graffiti writing, turnstile jumping, public drinking, and public urination, through arrest, detention, and criminal charges.
The New York City quality-of-life initiative produced what many observers called a revolution in policing and law enforcement. Today, the three most populous cities in the United States—New York, Chicago, and, most recently, Los Angeles—have all adopted at least some aspect of Wilson and Kelling’s broken-windows theory. In addition to the strategies of former New York City Mayor Rudolph Giuliani, the city of Chicago implemented an antigang loitering ordinance in the early 1990s that it vigorously enforced during the period from 1993 to 1995 resulting in misdemeanor arrests of more than 42,000 individuals. In October 2002, Los Angeles Mayor James Hahn appointed William Bratton police commissioner on a platform that promised a broken-windows approach. According to the New York Times, “Mr. Bratton said his first priority after being sworn in on Oct. 28  would be ending the smile-and-wave approach to crime fighting. He said he wanted policing based on the so-called broken-windows theory.” The popularity of broken-windows policing in the United States is matched only by its appeal abroad. In 1998 alone, representatives of more than 150 police departments from foreign countries visited the New York Police Department for briefings and instruction in order-maintenance policing. For the first ten months of 2000, another 235 police departments (85% of them from abroad) sent delegations to the New York City police headquarters.
The Lack of Empirical Evidence
Despite the widespread policy influence of the 1982 Atlantic Monthly essay, remarkably little is known about the effects of broken windows. A number of leading researchers in sociology, law, and police studies have compiled datasets from different urban areas to explore the broken-windows hypothesis, but the evidence remains, at best, mixed. In 2000, John Eck and Edward Maguire reviewed the empirical evidence and studies on broken-windows policing in their contribution to Alfred Blumstein and Joel Wallman’s The Crime Drop in America (2000), and found that there is little evidence to support the claim that broken-windows policing contributed to the sharp decrease in crime during the 1990s.
To date, empirical testing of the broken-windows idea has taken one of two forms. A first approach attempts to measure neighborhood disorder and crime, as well as other correlates of criminality, such as poverty and residential instability, in order to determine whether there are statistically interesting correlations between these variables. A second approach has focused on measures of broken-windows policing—for instance, rates of misdemeanor arrests—and conducts relatively similar statistical analyses on these variables in order, again, to identify significant correlations.
Disorder and Crime
Early on, many proponents of the broken-windows hypothesis pointed to the research of Wesley Skogan, especially his monograph Disorder and Decline: Crime and the Spiral of Decay in American Neighborhoods (1990), and argued that it empirically verified the broken-windows theory. Disorder and Decline addressed the larger question of the impact of neighborhood disorder on urban decline, but in one section, Skogan discussed the broken-windows hypothesis, ran a regression of neighborhood disorder on robbery victimization, and concluded that “‘Broken windows’ do need to be repaired quickly” (p. 75). Many observers interpreted this as an endorsement of the broken-windows theory and accepted this view of the evidence. George Kelling contended that Skogan ”established the causal links between disorder and serious crime—empirically verifying the ‘Broken Windows’ hypotheses” (Kelling and Coles 1996, 24). Subsequent research by Bernard Harcourt (2001), however, cast doubt on Skogan’s findings and raised significant questions as to what conclusions could properly be drawn from Skogan’s analysis.
A few years later, Ralph Taylor (2001) conducted research in sixty-six neighborhoods in Baltimore using longitudinal data. Taylor attempted to determine the relationship between neighborhood crime and what he termed social and physical ”incivilities”—panhandlers, public drunks, trash, graffiti, and vacant lots, among other things. What he found was that, while certain types of incivilities were associated with crime or urban decay, others were not. He concluded from his data that different types of incivilities may require different policy responses: ”Researchers and policy-makers alike need to break away from broken windows per se and widen the models upon which they rely, both to predict and to preserve safe and stable neighborhoods with assured and committed residents” (p. 22).
In a 2005 study, Broken Windows: Evidence from New York City and a Five-City Social Experiment, Bernard Harcourt and Jens Ludwig explore the empirical results from an important social experiment known as Moving to Opportunity (MTO) that is under way in five cities: New York, Chicago, Los Angeles, Baltimore, and Boston. Under the MTO program, approximately 4,600 low-income families living in high-crime public housing communities characterized by high rates of social disorder were randomly assigned housing vouchers to move to less disadvantaged and disorderly communities. Harcourt and Ludwig compare the crime rates among those who moved and those who did not—using official arrests and self-report surveys—and the results are clear, though disappointing: Moving people to communities with less social or physical disorder on balance does not lead to reductions in their criminal behavior. Neighborhood order and disorder do not seem to have a noticeable effect on criminal behavior.
The implications of MTO for the ongoing debates about the broken-windows theory are significant and suggest that moving people to communities with less social or physical disorder—the key intervening factor in the original broken-windows hypothesis—on balance does not lead to reductions in their criminal behavior. Taken together, the data from MTO provide no support for the idea that ”broken-windows” activities, including broken-windows policing or other measures designed to reduce the level of social or physical disorder within a community, represent the optimal use of scarce government resources.
Another comprehensive study of the broken-windows theory is Robert Sampson and Stephen Raudenbush’s 1999 study, Systematic Social Observation of Public Spaces: A New Look at Disorder in Urban Neighborhoods. Their study grew out of the Project on Human Development in Chicago Neighborhoods and was based on systematic social observation: Using trained observers who drove a sports utility vehicle at five miles per hour down every street in 196 Chicago census tracts, and randomly selecting 15,141 street sides, they were able to collect precise data on neighborhood disorder. With regard to the disorder-crime nexus, they found that disorder and predatory crime are only moderately correlated, but that, when antecedent neighborhood characteristics are taken into account, the connection between disorder and crime ”vanished in 4 out of 5 tests— including homicide, arguably our best measure of violence” (p. 637). On the basis of their extensive research, they conclude that ”[a]ttacking public order through tough police tactics may thus be a politically popular but perhaps analytically weak strategy to reduce crime” (p. 638). As an alternative to the broken-windows theory, they suggest that disorder is of the same etiology as crime—being, so often, forms of minor crime—and that both crime and disorder have the same antecedent conditions. ”Rather than conceive of disorder as a direct cause of crime, we view many elements of disorder as part and parcel of crime itself” (p. 608). Thus, ”a reasonable hypothesis is that public disorder and predatory crimes are manifestations of the same explanatory process, albeit at different ends of a ‘seriousness’ continuum” (p. 608).
Studies of Aggressive Misdemeanor Arrest Policing
Another strand of research, focusing on studies of aggressive arrest policies, has also been brought to bear on the broken-windows hypothesis. Here, too, James Q. Wilson sparked the debate, primarily with his 1968 topic on the Varieties of Police Behavior, and his research with Barbara Boland on the effects of police arrests on crime. Wilson and Boland hypothesized that aggressive police patrols, involving increased stops and arrests, have a deterrent effect on crime. A number of contributions ensued, both supporting and criticizing these findings, but, as Robert Sampson and Jacqueline Cohen suggested back in 1988, the results were ”mixed” (p. 166). There have been strong contributions to the literature, such as the 1999 study led by Anthony Braga, titled ”Problem-Oriented Policing in Violent Crime Places: A Randomized Controlled Experiment,” published in Criminology. But still, most of this research is unable to distinguish between the broken-windows hypothesis and more traditional explanations of incapacitation and deterrence associated with increased police arrests, presence, contact, and surveillance. The problem is somewhat endemic to the design of these studies. As Sampson and Cohen conclude with regard to their own work, ”[i]t is true that our analysis was not able to choose definitely between the two alternative scenarios” (p. 185).
The most recent study here, Harcourt and Ludwig’s article Broken Windows: New Evidence from New York City and a Five-City Social Experiment, reanalyzes and assesses the best available evidence from New York City about the effects of broken-windows policing. They demonstrate that the pattern of crime changes across New York precincts during the 1990s that has been attributed to broken-windows policing is more consistent with what statisticians call mean reversion: Those precincts that experienced the largest drop in crime in the 1990s were the ones that experienced the largest increases in crime during the city’s crack epidemic of the mid- to late-1980s. They call this Newton’s law of crime: What goes up, must come down, and what goes up the most, tends to come down the most.
In this vein, Jeffrey Fagan and Garth Davies (2003) also test, in Policing Guns: Order Maintenance and Crime Control in New York, whether quality-of-life policing in New York City contributed to the reduction in lethal violence in the late 1990s. They analyze precinct crime rates from 1999 and try to determine whether these crime rates can be predicted by the amount of stop-and-frisk activity that occurred in the precinct in the preceding year. Based on their research, they find that ”[f]or both violence arrests broadly and homicide arrests specifically, there is no single category of citizen stops by police that predicts where crime will increase or decrease in the following year.” When they examine homicide fatalities, they observe different effects by type of stop and by victim race. ”Stops for violence are significant predictors of reductions in both gun homicide deaths and overall homicide deaths, but only among Hispanics.” In contrast, for African Americans, no type of arrests predicts homicide victimization a year later; and for whites, the results are not reliable because of the low white homicide victimization rate. Why is it that there may be effects for Hispanics, but not for African Americans? Fagan and Davies suggest that it may have to do with what they call ”stigma saturation” in black communities: When stigma is applied in ways that are perceived as too harsh and unfair, it may have reverse effects. They write, ”When legal control engenders resistance, opposition or defiance, the opportunity to leverage formal social control into informal social control is lost. The absence of crime control returns from [order-maintenance policing] may reflect just such a dynamic among African Americans, who shouldered much of the burden of [order-maintenance policing].”
Steve Levitt’s 2004 Journal of Economic Perspectives review essay similarly suggests that policing practices probably do not explain much of the crime drop in the 1990s because crime went down everywhere, even in places where police departments did not implement new policing strategies. Instead, Levitt attributes the massive period effects on crime throughout the United States during the 1990s to some combination of increased imprisonment, increases in the number of police, the ebbing of the crack epidemic that started in many big cities in the mid-1980s, and the legalization of abortion in the United States during the early 1970s.
The criminological evidence surrounding New York City is not any more helpful to the broken-windows theory. A number of large U.S. cities—Boston, Houston, Los Angeles, San Diego, San Francisco, among others—have experienced significant drops in crime since the early 1990s, in some cases even larger proportionally to the drop in crime in New York City. Several of these cities did not implement the type of aggressive order-maintenance policing that New York City did. Applying a variety of different tests, one recent study found that New York City’s drop in homicides, though not very common, is not unprecedented either. Houston’s drop in homicides of 59% between 1991 and 1996 outpaced New York City’s decline over the same period of 51%, and both were surpassed by Pittsburgh’s 61% drop in homicides between 1984 and 1988 (Fagan, Zimring, and Kim 1998). Another study looked at the rates of decline of homicides in the seventeen largest U.S. cities from 1976 to 1998 in comparison to each city’s cyclical peaks and troughs, using a method of indexed cyclical comparison. With regard to the most recent cyclical drop in homicides, New York City’s decline, though above average, was the fifth largest, behind San Diego, Washington, D.C., St. Louis, and Houston (Joanes 1999).
A straight comparison of homicide and robbery rates between 1991 and 1998 reveals that, although New York City is again in the top, with declines in homicide and robbery rates of 70.6% and 60.1%, respectively, San Diego experienced larger declines in homicide and robbery rates (76.4% and 62.6%, respectively), Boston experienced a comparable decline in its homicide rate (69.3%), Los Angeles experienced a greater decline in its robbery rate (60.9%), and San Antonio experienced a comparable decline in its robbery rate (59.1%). Other major cities also experienced impressive declines in their homicide and robbery rates, including Houston (61.3% and 48.5%, respectively) and Dallas (52.4% and 50.7%, respectively) (Butterfield 2000).
What is particularly striking is that many of these cities did not implement New York-style order-maintenance policing. San Diego and San Francisco bear perhaps the greatest contrast. The San Diego police department implemented a radically different model of policing focused on community-police relations. The police began experimenting with problem-oriented policing in the late 1980s and retraining their police force to better respond to community concerns. They implemented a strategy of sharing responsibility with citizens for identifying and solving crime. Overall, while recording remarkable drops in crime, San Diego also posted a 15% drop in total arrests between 1993 and 1996, and an 8% decline in total complaints of police misconduct filed with the police department between 1993 and 1996.
San Francisco also focused on community involvement and experienced decreased arrest and incarceration rates between 1993 and 1998. San Francisco’s felony commitments to the California Department of Corrections dropped from 2,136 in 1993 to 703 in 1998, whereas other California counties either maintained or slightly increased their incarcerations. San Francisco also abandoned a youth curfew in the early 1990s and sharply reduced its commitments to the California Youth Authority from 1994 to 1998. Despite this, San Francisco experienced greater drops in its crime rate for rape, robbery, and aggravated assault than did New York City for the period 1995 through 1998. In addition, San Francisco experienced the sharpest decline in total violent crime—sharper than New York City or Boston—between 1992 and 1998 (Taqi-Eddin and Macallair 1999).
Other cities, including Los Angeles, Houston, Dallas, and San Antonio, also experienced significant drops in crime without as coherent a policing strategy as other large cities. There was a remarkable decline in crime in several major cities in the United States during the 1990s. Depending on the specific time frame of the ”before-and-after” comparison, New York City’s drop in crime can be characterized anywhere from the biggest decline to a very high performer. Time frames can be easily manipulated, but what is clear is that numerous major U.S. cities have experienced remarkable declines in crime and have employed a variety of different policing strategies. It is too simplistic to attribute the rate of the decline in New York City to the quality-of-life initiative.
Moreover, criminologists have suggested a number of factors that have contributed to declining crime rates in New York City, including a shift in drug use patterns from crack cocaine to heroin, favorable economic conditions in the 1990s, new computerized tracking systems that speed up police response to crime, a dip in the number of eighteen- to twenty-four-year-old males, as well as possible changes in adolescent behavior. There have also been important changes at the New York Police Department (NYPD), including a significant increase in the raw number of police officers. Former Mayor Dinkins hired more than two thousand new police officers under the Safe Streets, Safe City program in 1992, and Giuliani hired another four thousand officers and merged about six thousand Transit and Housing Authority officers into the ranks of the NYPD. As a result, from 1991 to 2000, the NYPD force increased almost by half, up by 12,923 police officers (including those transferred from Transit) from a force of 26,856 police officers in 1991 to 39,779 police officers in 2000. Excluding the Transit merger, the police force grew by almost a quarter. As a result, the New York Police Department now has the largest police force in the country with the highest ratio of police officers per civilian of any major metropolitan area.
The bottom line is that the broken-windows theory—the idea that public disorder sends a message that encourages crime—is probably not right, and there is no reliable evidence that broken-windows policing is an effective law enforcement tool. As Sampson and Raudenbush (1999) observe, ”bearing in mind the example of some European and American cities (e.g., Amsterdam, San Francisco) where visible street level activity linked to prostitution, drug use, and panhandling does not necessarily translate into high rates of violence, public disorder may not be so ‘crimino-genic’ after all in certain neighborhood and social contexts” (p. 638).