In 1982 we argued that police departments should return to their ancient tradition of maintaining public order as well as fighting serious crime. The urban American public was upset by signs of disorder such as graffiti, public drunks, aggressive panhandlers, street-corner drug dealers and hostile gangs of youth. When the police were first formed in American cities they took such matters seriously, but as other agencies developed that supposedly were interested in human distress, and with the rise in serious crime rates, the police increasingly confined themselves to investigating offenses.
We also suggested another possibility: that high levels of public disorder could indirectly increase crime rates. This would happen as disorder discouraged honest people from using the streets, thereby leaving public spaces available for small-scale offenses and then more serious ones. We used the metaphor of a building with a broken window: If it were not promptly fixed, more windows would be broken. And so if public disorder were not eliminated, more disorder and then more serious crime would become commonplace. Unfortunately, we did not label this argument a speculation.
Nearly a quarter century has elapsed since we published “Broken Windows” in the Atlantic Monthly. During that time many big-city police departments have become concerned about public order, even while some critics have argued that increased public order does not lead to less crime. We believe that when the police work to restore order and do so in a decent and lawful fashion, they have produced an important public good. We doubt it is necessary to justify that result; it is, we think, self-evidently good.
The restoration of order in New York City’s subways is perhaps the best illustration of why order is a self-evident good. During the 1970s and 1980s lawlessness in the subway had become routine. Every car on every train was covered with graffiti; 250,000 people annually did not pay their fares; aggressive panhandlers menaced passengers; indigents sprawled on steps and in passageways, blocking access; and young people jammed turnstiles and extorted tokens from would-be passengers. Despite an $8 billion restoration of the subway’s infrastructure, ridership was declining as ever more people refused to suffer these threats and indignities.
In 1990 William Bratton, who later became New York’s police commissioner under Mayor Rudolph Guiliani, was hired as chief of the transit police with a mandate to restore order. Under his direction, the transit police started to pay attention to minor offenses and rule violations. Within months there was a dramatic turnaround. Today, two million more people ride New York City subways every day than in the 1980s. Fare beating is so rare that the Transit Authority no longer even counts it. Hundreds of people arrested for minor rule violations turned out to have warrants against them for major crimes. Taking small matters seriously made a difference in managing large problems. Just as important was this demonstration of the value of public order: The authorities defeated what some thought was an insurmountable problem.
We believe that it is necessary to test the argument that increased public order tends to reduce more serious crime, such as robbery in public places. There have been studies that attempt to see if, other things equal, crime rates are higher in areas with more disorder. These studies come to somewhat different conclusions. One, by Wesley Skogan, a scholar at Northwestern University, summarized in Disorder and Decline (1992) the findings of people who had conducted surveys of public order in forty neighborhoods in several large cities. The surveys involved interviewing residents to measure how they viewed the level of order and disorder. He concluded that as the level of disorder rose, there was an increase in the percentage of people in those neighborhoods who were victims of robbery. He did not claim that disorder alone causes robbery, but that it is one of several factors (the others being poverty and family instability) that independently contribute to crime.
Ralph Taylor, a professor at Temple University, examined the connection between disorder and crime in Baltimore. Beginning in 1981 and ending in 1994, teams of observers evaluated the level of physical order in a random set of neighborhoods, and interviewers questioned residents and community leaders. Taylor described the relationship between what he called “incivilities” and several (but not all) kinds of crime. “Incivilities”, he concluded, “do matter for crime changes”, but not for all such changes.
Another study from 1999, by Robert Sampson at Harvard and Stephen Raudenbush at the University of Michigan, came to a different conclusion. Their project was set in one city, Chicago, and involved driving the city streets during daylight hours in an SUV and videotaping the physical structures of various neighborhoods. After categorizing hundreds of these pictures as showing more or fewer signs of disorder, they concluded that, if you take into account neighborhood social characteristics (such as poverty, unemployment, female-headed families, and the strength of “collective efficacy” or a strong sense of community), there is no relationship between disorder and crime except for robbery, and that connection is relatively weak.
We have several reactions to these studies. First, we think that the methods used by Skogan and Taylor—interviewing people about disorder—will give a more reliable estimate of this condition than will videotaping buildings during daylight hours. Second, the Sampson-Raudenbush measure of “collective efficacy” may itself be caused by public disorder: When disorder is high, people have a weaker sense of community self-control. Third, two of the three studies (Skogan and Sampson-Rausdenbush) confirm that, independently of social conditions, disorder is correlated with robbery rates. This is exactly the crime that we think is the ideal test of the theory, since other offenses (homicide, burglary, theft) are done surreptitiously and (usually) indoors.
But the best test of the theory would not involve measuring existing connections between disorder and crime. It would require instead an effort to change crime rates by increasing public order. Unfortunately, such a test has rarely been done. Doing it would require randomly assigning police precincts either to an order-maintenance program or to conventional policing, closely monitoring the levels of crime and calls for police assistance in these precincts, and testing to see whether any crime that had occurred in an experimental area was merely deflected to control districts.
We know of only one rigorous test of this view, and its results tend to support our claim. Anthony Braga of the Kennedy School of Government at Harvard, working with five coauthors, did such a test in Jersey City in 1999. In their report they wrote that they were evaluating not broken windows but “problem-oriented policing.” POP, to give it its police name, is an effort to make officers identify and respond to the causes of various crime problems that occur in a specific neighborhood. These problems could be drug sales in apartment buildings or on street corners and burglaries in public housing projects. The best response to these problems is not merely to arrest the drug dealers or burglars, but to investigate and alter the behavior of the owner of the apartment building, make it hard for drug dealers to gather on street corners, and change the management of the public housing projects.
In Jersey City, Braga and his associates identified 56 small neighborhoods with high crime rates. Twenty-four of these places were randomly assigned into experimental and control groups. In the experimental areas, the police mounted problem-solving tactics. What is striking about this study is that most of these tactics involved “aggressive order maintenance” coupled with removing the physical signs of disorder such as trash on the streets and adding things that might increase order, such as more street lights and fencing in vacant lots.
In short, much of the problem-solving in Jersey City was actually an effort to fix broken windows. And Braga et al. found that criminal incidents fell in the experimental neighborhoods compared to the control ones. There were fewer robberies, property crimes and street fights. To check their results, they looked to see if the crimes in the experimental areas had just moved over to control areas. If that had happened, then there would have been no reduction in crime, merely a displacement of it. But this did not happen: The majority of crime incidents were not relocated, they simply ended.
There have been other less rigorous tests of the effects of improving public order on crime rates. One of us (Kelling) and William Sousa studied misdemeanor arrests in New York City when William J. Bratton was police commissioner. Misdemeanor arrests went up as a result of a police policy of getting tough on small offenses, in part because some low-level miscreants also were wanted for more serious crimes and in part because Bratton and Mayor Giuliani believed that if the streets were more orderly they would also be safer. Kelling and Sousa found that in those areas where misdemeanor arrests rose the most, felony crimes decreased the most. A similar study with comparable conclusions was undertaken by two scholars at the National Bureau of Economic Research.
Much but not all of the evidence is consistent with the view that increased order is associated with decreased crime. We happily acknowledge that the matter is not settled. The experiment in Jersey City is the best test of the idea, but one experiment in one city cannot and should not end debate. There ought to be several more experiments in other large cities that examine the effect of increased order on crime rates.
Throughout all of this, one voice has steadfastly maintained that the broken windows idea must be wrong. Bernard Harcourt, a professor at the University of Chicago Law School, has argued for years that there is no connection, and indeed there ought to be none, between disorder and crime. Although he agrees that some forms of disorder are undesirable, he argues that what most people regard as disorder is really a projection of personal beliefs onto people unlike themselves. In his view, the police should stop paying attention to minor forms of disorder and instead concentrate on major crimes. That is, the police should retreat to the strategy they followed during much of the last century.
Harcourt is correct to say that a clear, causal link between disorder and serious crime has not been firmly established. But in defending his view, he accepts all studies, no matter how weak, that argue against the link and rejects all those, however strong, that argue in favor of it. He criticizes Skogan’s book, even though he grudgingly admits that Skogan really has found a link between disorder and robbery. He says that Taylor’s study of Baltimore shows no disorder-crime link when in fact Taylor writes that “incivilities do matter for crime changes.” Incivilities (his word for broken windows) are linked to changes in three out of the four crime rates he investigated.
He praises the Sampson-Raudenbush study of Chicago but admits at the end that they agree there may be a connection between disorder and robbery. After having embraced a study that suggests such a link exists and rejected two others that say so in even stronger terms, Harcourt brought forth his own study. Along with Jens Ludwig, he examined data produced by an ongoing housing relocation study begun by the Federal government in 1994. This project, called Moving to Opportunity (MTO), gives housing vouchers to low-income families living in five cities. Some of these families, randomly selected, were given a voucher that could only be used in a census tract that was not poor, while other families were either given no vouchers or ones that could be used any place. Harcourt and Ludwig concluded that the experimental families, the ones that could move to fairly affluent neighborhoods, had the same crime rate after moving as they had before they moved. Therefore, they argued, the broken windows hypothesis had been disproved.
What? We never suggested that moving people from one neighborhood to another would change their behavior. Every bit of social science evidence we have on criminality shows that deep biological and familial factors cause criminality. There may be some neighborhood effects, but it is unlikely that any such effects would reveal themselves in three or four years. Perhaps after a generation or two neighborhood expectations of conformity would begin to reshape the minds of young people, but we cannot imagine that happening immediately.
Moreover, our broken windows hypothesis suggests that disorder affects crime rates in neighborhoods, not in individuals. We don’t want to belabor an old social-science problem here, but it does not make a lot of sense to study behavior in places by studying the behavior of individuals. Suppose, for example, that the crime rates of the MTO folks who moved decreased in their new neighborhoods. Why might that have happened? Because they could find better jobs or get access to better public services? Possibly. But we doubt anyone would seriously argue that their crime rates went down because their neighbors had low crime rates.
We are eager for more and better tests of the possibility that increased disorder leads to increased crime. Both of us have visited Skid Row in Los Angeles, and we find it hard to believe that an army of drunk and drug-happy people sleeping and defecating on the sidewalks has no effect on crime rates there. But this anecdote does not settle the matter. We think the Jersey City study shows we are on the right track, but more work is needed.
For now, what puzzles us most is why Harcourt wants the police to confine themselves to fighting major crime while ignoring disorder. For a century or more, American police chiefs took this view, but we think for very different reasons than the ones that motivate Harcourt. To him, the problem of order is wrong because disorder is not generally a problem. As he writes in Illusion of Order: The False Promise of the Broken Windows Theory (2001), “order” is “an aesthetic preference that is not shared by all.” At this point he launches into a remarkable postmodernist critique of knowledge. He embraces the views of Michel Foucault and quotes him thus: “[T]he powers of modern society are exercised through, on the basis of, and by virtue of, this very heterogeneity between a public right of sovereignty and a polymorphous disciplinary mechanism.” Is that clear to you? Then he continues, this time quoting Pierre Bourdieu: To deal with order, we need a “full sociological objectification of the object and of the subject’s relation to the object.” From this Harcourt concludes that “it is no longer possible today to speak of objectivity in social science.” Well, if there can be no social science knowledge, then proving or disproving the disorder-crime link is impossible. It follows that the studies Harcourt cites must be as worthless as the ones we cite.
But there is more to his view than these rhetorical flourishes. He suggests that we should legalize prostitution, mount work programs for the homeless, not urge juveniles to snitch on one another, install subway turnstiles over which no one can jump, and buy graffiti-proof subway cars. We should do these things because “the concept of ‘disorder’ is not natural.” Even urinating in the street is not wrong unless the public believes it is wrong. Harcourt seems to think that they should not.
As Robert Jackall has pointed out, Harcourt does not want a society where people follow rules of conduct, but one in which they can be prevented (mechanically if at all possible, by police work only if absolutely necessary) from inflicting serious harm on one another. By contrast, we prefer a society that is decent because people follow learned rules of behavior. Some shared rules can be unjust, as is racial discrimination, but those rules have been changed. Other rules are just, and the public wants them enforced against people who have decided to ignore them. Such people might want to legalize prostitution, public drunkenness and indiscriminate urination, but they haven’t been legalized. Harcourt should direct his arguments to the “community members” about whom he so vaguely writes who prefer a very different world than the one he endorses. They have already voted against the world he prefers, many times over.
The broken windows idea does two things, one indisputably good and the other probably effective: It encourages the police to take public order seriously, something that the overwhelming majority of people ardently desire, and it raises the possibility that more order will mean less crime. The first goal requires no evidence. The second does, and so far most studies suggest that more public order (along with other factors) is associated with less predatory street crime. With all this in mind, we believe that it remains a strategy worth pursuing.