Discriminant Validity Of Disorder And Crime Research Paper

This sample Discriminant Validity Of Disorder And Crime Research Paper is published for educational and informational purposes only. If you need help writing your assignment, please use our research paper writing service and buy a paper on any topic at affordable price. Also check our tips on how to write a research paper, see the lists of criminal justice research paper topics, and browse research paper examples.


The broken windows thesis posits a causal relationship between disorder and crime whereby disorder occurs initially and, if left unchecked, ultimately produces serious crime. This causal sequence requires that disorder and crime be empirically separable phenomena; that is, they must possess discriminant validity. If discriminant validity is absent, then the broken windows thesis is fatally flawed because disorder cannot be said to cause crime if disorder and crime are two facets of the same underlying problem. This research paper reviews three studies that have spoken to the issue of discriminant validity and have concluded that disorder and crime are not, in fact, unambiguously distinct factors. The future of the broken windows thesis is thus uncertain, and if order maintenance policing does reduce some types of crime, this might be accomplished through a net-widening effect rather than through the informal social control mechanisms that are keys to the thesis.


For policy analysts, the question “Where does crime come from?” is subordinate to “What institution of control can be utilized to reduce crime?” In their 1982 Atlantic Monthly article that laid out the central tenets of what would become known as the “broken windows thesis,” policy analysts James Q. Wilson and George Kelling selected the police as the institution that would get the limelight. Wilson and Kelling argued that the police were the keys to crime reduction and prevention at the community level and that the best way for them to accomplish this would be to proactively target disorderly people and conditions.

Today, the idea of proactive, order maintenance policing has become part of the common vernacular in discussions of police goals and activities. In 1982, however, Wilson and Kelling faced a serious obstacle that impeded their proposed line of reasoning: The professional model of policing and several constitutional rulings issued by the US Supreme Court had made it unacceptable and legally questionable for police officers to interfere in matters not pertaining directly to serious crime. Wilson and Kelling, then, could not merely assert that police should start concentrating their efforts on reducing annoying yet nondangerous, noncriminal behaviors like loitering and panhandling – police leaders and policymakers would have pointed to the various industry standards and court rulings and thrown their hands up at the suggestion. Wilson and Kelling therefore had to be creative.

And they were. They laid out a formula wherein (a) disorder was uncoupled from serious crime and (b) disorder led to serious crime in a causal fashion. The reasonableness of the claim that disorder is well within the legitimate purview of police enforcement activities was thus made obvious. It boiled down to a simple syllogism: Police are responsible for controlling crime, and disorder causes crime if nobody does anything about it, so police are responsible for controlling disorder.

The problem is that Wilson and Kelling offered no empirical evidence for the unlinking of disorder and crime; in fact, there really was no reason to think that Wilson and Kelling’s formulation of the disorder-crime relationship was any more valid than, say, Garofalo and Laub (1979) contention that people’s fear of “crime” is actually a very general feeling of anxiety and apprehension about neighborhood-level conditions of sociostructural malaise wherein crime and disorder are part of overarching judgments about quality of life. Wilson and Kelling, then, did not prove that disorder and crime are distinct from one another and that disorder causes crime – they speculated that such was the case.

Wilson and Kelling’s oversight may not have been problematic had others stepped up to fill in the gaps, but this did not happen. Instead, the disordercrime distinction and supposed causal sequencing was accepted as fact, and efforts were launched to put broken windows policing into action in the streets. Doing so violated the basic principle of scientific inquiry that requires theories and their constituent predictions (i.e., hypotheses) to be empirically tested repeatedly before being taken as valid. Scientific testing is hardly a minor detail. The broken windows thesis has assumed a position of preeminence in policing without any strong evidence that Wilson and Kelling were right. Worrall (2006a, p. 379) put it well when he argued that “just because theorists have pushed for the separation of crime and incivility indicators does not mean both concepts are distinct.. . the incivilities [broken windows] thesis has evolved with little attention to the validity of its measures.”

This research paper assesses the evidence pertaining to the discriminant validity (i.e., the empirical uniqueness) of disorder and crime. Discriminant validity is said to be present when two factors that are theorized to be separate and unique from one another actually do display these qualities in empirical tests. The question of discriminant validity is vital to the broken windows thesis (BWT) because this framework posits that disorder is an independent variable and crime is a dependent variable. This means that these two phenomena must be related but distinct and separable from one another. If disorder and crime possess discriminant validity, then this portion of the broken windows thesis holds up; if they do not, then the BWT argument is seriously weakened. Disorder can only cause crime if disorder is not crime, and crime, likewise, can only be caused by disorder if crime is not disorder.

Disorder And Crime Under The Broken Windows Thesis

There is no shortage of practitioners and researchers who claim that the BWT has been firmly empirically supported (e.g., Kelling and Coles 1996; Kelling and Sousa 2001; Skogan 1990). Most of the studies claiming to find support for the thesis, however, have not actually tested the thesis itself but, rather, have assessed the crime-reduction impact of police efforts to crack down on disorder. As will be discussed later, some evidence does suggest that order maintenance policing may be promising under certain conditions; however, it is not at all clear that the mechanism by which order maintenance policing may effectively reduce disorder and crime is actually that specified by the broken windows thesis. Researchers who have studied the thesis itself and the hypotheses derived from it have found little empirical support for it and have offered several reasons why the disorder-crime causal link posited by the BWT may be flawed (Harcourt 2001; Harcourt and Ludwig 2006; Sampson and Raudenbush 1999, 2004; Taylor 2001).

Perhaps the biggest challenge to most of the existing broken windows studies is researchers’ tendency to rely on cross-sectional data. Cross-sectional data are those that are gathered at a single time point, such as when a survey asks respondents to report their perceptions about various neighborhood and community characteristics. Cross-sectional research designs make the establishment of causation difficult because there is no time lapse between the measurement of the independent variable and that of the dependent variable, and thus, it is very easy to mistake correlation (i.e., simple, noncausal coincidence) for causation. To the extent that areas plagued by serious crime also suffer high levels of social and physical disorder (Sampson and Raudenbush 2004), cross-sectional research may merely capture the convergence of these two phenomena in space and time and erroneously portray this contemporaneousness as evidence of causality. Alternatively, Taylor’s (2001) longitudinal study showed that disorder measured at one time point did not consistently or strongly impact crime measured at a later date. This severely undermined the broken windows thesis and called its core tenet into serious question.

In the search for reasons why disorder may not, in fact, cause crime in the manner advocated by Wilson and Kelling (1982), the issue of discriminant validity stands prominent. As explained in the preceding paragraphs, disorder can only be a predictor of crime if disorder and crime are empirically distinct phenomena. Content overlap between these two factors would violate the principles of scientific inquiry because the independent and dependent variables would essentially be the same thing.

Some proponents of the broken windows thesis might argue that discriminant validity is not a prerequisite for the veracity of the claim that disorder causes crime. Two lines of argument might be advanced in this regard. First, it might be proposed that an absence of discriminant validity would merely signal that the broken windows process has occurred – disorder went unchecked, disorder caused crime, and now crime and disorder are both prevalent in the area. This line of reasoning, however, tautologically begs the question of whether disorder causes crime, as merely pointing to the co-occurrence of the two problems proves neither that disorder temporally and causally precedes crime nor that the two problem types are empirically distinguishable.

Second, it could be contended that disorder and crime are more like a continuum of problems than two separate issues, with crime being on the more serious side of the continuum and disorder on the less serious side. The idea of such a continuum is, in and of itself, quite plausible; however, it contradicts the logic of the BWT. The continuum model would render broken windows unfalsifiable by making it impossible to figure out exactly what the independent and dependent variables are. An accurate empirical test could never be conducted because there would always be doubt as to whether something that was measured as crime should have actually been considered disorder or vice versa. The crucial assumption made by Wilson and Kelling is that disorder causes crime. This premise requires that disorder and crime be two different problems; that is, it necessitates discriminant validity.

It is immediately apparent that this requirement is violated to at least some extent in the BWT, because many of the behaviors the thesis identifies as disorder (e.g., prostitution, tagging, vandalism, minor drug dealing) are crimes by statute or ordinance. There is, therefore, overlap between disorder and crime from a legal or definitional standpoint. This fact provides preliminary evidence against the scientific validity of the broken windows thesis.

The only way that the broken windows thesis can survive this dilemma is if disorder and crime are perceived as being unambiguously different in the public’s view. Broken windows is grounded in social psychology and hinges on subjective perceptions and on the emotions that people feel and the inferences they draw when they see either the presence or absence of disorder in their neighborhoods of residence. In other words, disorder does not exist in some kind of objective “reality” or “truth” (Piquero 1999; Sampson and Raudenbush 2004), but, rather, disorder takes on whatever meaning is ascribed to it by local residents. If people see a difference between disorder and crime, then broken windows might be a viable theory of crime production despite the aforementioned statutory overlap.

One of the most significant demonstrations of the similarity between disorder and crime comes from Sampson and Raudenbush (1999), who relied on systematic social observations of disorder rather than on survey-based perceptual measures. Their results clearly and consistently showed a weak relationship between neighborhood-level disorder and crime once structural covariates were controlled. The only exception to this pattern was with respect to the crime of robbery, which was, interestingly, the only crime type linked to disorder in Skogan’s study published in 1990 (Harcourt 2001). Moreover, it appeared that disorder and crime both sprung from the common origin of weak collective efficacy – one of the most consistently robust community-level predictors of such problems (Pratt and Cullen 2005). Far from the causal sequence predicted by the BWT, then, it appeared that disorder and crime were both “symptoms” of the same “disease”: the breakdown of a community’s ability to regulate its group members and enforce standards of civility and legal conformity.

Testing for discriminant validity between disorder and crime requires the use of a statistical modeling procedure that was created specifically for the purpose of testing proposed factor structures (e.g., Gau 2010). This analytical technique is called confirmatory factor analysis (CFA). Confirmatory factor analysis has allowed tests of discriminant validity within the BWT framework; without this modeling tool, tests of this sort would be much more difficult and the results would likely be harder to interpret. Because CFA is so central to the discriminant validity debate, the next section will briefly outline this analytical strategy and will be followed by summaries of the studies that have employed CFA to test for discriminant validity.

Confirmatory Factor Analysis

Confirmatory factor analysis is a statistical procedure that permits extensive examination of factor structure. A factor is something that is not observed directly but accounts for the similarity between a set of objects. For instance, a fire truck, a stop sign, and the pen a professor uses to grade student exams may all be objects representative of the factor “red in color.” Certain variables or objects can be indicators of different factors depending on the full item set. If the fire truck were put into a group with a police car and an ambulance, then the factor would be “emergency vehicles.” Confirmatory factor analysis permits researchers to propose hypothesized models by saying, “I think this variable (e.g., vandalism) is part of this factor (e.g., disorder) and I think that that variable (e.g., robbery) is part of that factor (e.g., crime)” and to then test these predictions.

The quality of a hypothesized model is assessed using fit indices and factor loadings. Fit indices are established criteria that have certain maximum or minimum values – depending on the index – that are indicative of good model fit. Factor loadings are similar to regression coefficients and measure the strength of the relationship between a variable and the factor it is predicted to represent. In the context of tests for discriminant validity in broken windows research, this type of validity would be present when all of the fit indices met their respective thresholds for good fit and the items thought to measure disorder all loaded highly on the disorder factor and not on the crime factor, while the crime items loaded well on crime and not on disorder. If the fit indices or factor loadings do not meet these requirements, then there is a problem with the model and discriminant validity might be absent.

Confirmatory factor analysis is a popular statistical procedure in psychological research and has gained recognition and use in criminal justice and criminology, as well. The software required to run these analyses is widely available, reasonably priced (as far as software goes), and sports user-friendly, point-and-click interfaces (Gau 2010). There is, in short, no good reason for CFA to not be a staple of broken windows research, which makes it surprising that only a handful of studies have used this method to assess disorder and crime factors for measurement validity (Armstrong and Katz 2010; Gau and Pratt 2008; Piquero 1999; Worrall 2006a). The next section details the studies that have used CFA to test for discriminant validity in perceptual measures of disorder and crime.

Empirical Inquiries Into The Discriminant Validity Of Disorder And Crime

Worrall (2006a) and Armstrong and Katz (2010) addressed the issue of discriminant validity by entering variables tapping survey respondents’ perceptions of disorder, crime, and personal victimization experiences into a series of CFA models. Worrall employed data from 12 cities in 12 different states, and Armstrong and Katz used a sample of survey respondents in a single city in Arizona. The results of both studies painted a murky, inconclusive picture. Survey respondents did not make clear, consistent distinctions between the three phenomena, but neither did they unambiguously see them as one and the same. One-factor models positing disorder as being a single factor along with crime or victimization did not display good fit, yet neither did two-factor models allowing disorder to function as a factor apart from the other two. Overall, then, the authors of both studies concluded that discriminant validity had not been established and that this key premise of the broken windows thesis was not empirically supported.

Gau and Pratt (2008) added to the debate by examining survey data from respondents in neighborhoods in 21 municipalities in Washington State. The variables of interest came from 17 crime and disorder items that asked respondents to indicate whether each issue was 1 = no problem, 2 = uncertain, 3 = a problem, or 4 = a serious problem in their neighborhoods. While some of the items were obvious indicators of crime (e.g., gun violence) and others were clearly measures of disorder (e.g., noise), many others (e.g., vandalism, youth gangs) were ambiguous and did not lend themselves clearly to either category. The authors therefore relied on prior literature (Giacopassi and Forde 2000; Reisig and Parks 2004; Sampson and Raudenbush 1999; Taylor 2001) to construct the disorder and crime factors. The disorder factor contained the following items: dogs running at large, drunk drivers on the road, people drinking to excess in public, groups of teens or others hanging out and harassing people, youth gangs, people using illegal drugs, vandalism, noise, traffic problems, and garbage/litter. The crime indicators were people’s homes being broken into and things being stolen, people being robbed or having their purses/wallets taken, domestic/ intimate partner violence, rape/sex crimes, child abuse, violent crime, and gun violence.

Gau and Pratt (2008) constructed two different models so that each one could be tested against the other to determine which one seemed superior to the other. This strategy is necessary for valid hypothesis testing; it is not a good idea to run only one model because then it is impossible to determine whether the tested model truly is the best one or whether some other model that was not tested might fit better. The first model that Gau and Pratt specified predicted that all of the disorder and crime variables were part of a single factor. This model represented the prediction that people see no difference between disorder and crime and that they instead tend to lump them together as part of the general condition of the neighborhood. Support for this model would undermine the broken windows thesis. The second model predicted that disorder and crime were two separate, distinct factors, each measured by the variables listed above. This model represented the original broken windows framework and predicted that disorder and crime possess discriminant validity. Support for this model would strengthen the broken windows thesis.

Both models displayed similar fit indices that offered lukewarm validation of each of them. Some indices lent the impression that the models were good, while others fell short of their ideal values. On the basis of these alone, then, it was not clear whether the one-factor model or the two-factor model was the better representation of crime and disorder.

There is, however, another way to evaluate model fit above and beyond fit indices: the between-factor correlation. A between-factor correlation of 0.85 or greater indicates that the factors cannot be justifiably separated (see Gomez et al. 2005) because discriminant validity is unacceptably low. It becomes impossible to tell whether items that are purported to be indicators of one factor are not, in fact, equally plausible indicators of the other factor. Gau and Pratt (2008) discovered that the correlation between the crime and disorder factors was 0.92. The two-factor model, therefore, had to be rejected because the correlation between the crime and disorder factors made factor separation indefensible. Perceptions of crime and perceptions of disorder appeared to constitute a single factor; that is, discriminant validity was not present.

It is possible that discriminant validity in individual-level perceptions of crime and disorder is only part of the story; that is, the separability of disorder and crime might be influenced by neighborhood – or community-level conditions. Wilson and Kelling (1982) hinted that there might be a tipping point whereupon a neighborhood becomes so overrun by disorder and crime that police cannot do much other than try to keep pace with calls for service. The tipping point concept may apply to resident perceptions, too. People living in areas where disorder has been present for some time may hardly notice these incivilities because they are simply part of the neighborhood landscape, whereas people residing in areas that have been disorder-free in the past may react very strongly when even the slightest sign of incivility springs up. In other words, the broken windows thesis might be more applicable to some neighborhoods or communities than to others, depending upon extant levels of disorder in those places.

Gau and Pratt (2010) investigated the possibility of spatially patterned discriminant validity by creating a difference score that captured the extent to which people saw disorder levels as being higher, lower, or similar to crime levels in their neighborhoods. They regressed the difference score on several variables, including respondents’ perceptions of disorder and crime. The results indicated that the difference between crime and disorder increased as perceived neighborhood-level disorder increased. The authors then split the sample into low-disorder and high-disorder subsamples to test for interaction effects. Among the low-disorder group, increases in disorder caused a reduction in the difference score, while among the high-disorder group, increases in disorder were associated with a closing of the gap. Evidence for a perceptual tipping point was thus uncovered: People living in areas where disorder is mild or absent are more sensitive, and when they see visible signs of incivility, they are likely to interpret these signs as ominous harbingers portending the arrival of serious criminals. People accustomed to seeing disorder in their neighborhoods, on the other hand, were better able to make a distinction between disorder and crime and to more clearly see them as two separate phenomena.

Implications For The Broken Windows Thesis And Order Maintenance Policing

The absence of clear, consistent discriminant validity in the three CFA-based studies described above renders indefensible the claim that disorder is an independent variable and crime is a dependent variable. These two phenomena are no doubt interrelated; however, this relationship is likely correlational rather than causational (see also Sampson and Raudenbush 1999). The possibility that people’s ability to discriminate between disorder and crime is dependent upon the amount of disorder present in their neighborhoods of residence (Gau and Pratt 2010) adds another layer of complexity to the matter. These findings have important implications for criminological theory, the measurement of concepts in broken windows research, and crime control policy.

Theoretical And Methodological Implications

The broken windows thesis insists that people see disorder as a visible indicator that the community is spiraling out of control. If, however, people do not differentiate between crime and disorder, then crime itself could serve as the visible indicator of the lack of informal social control in a community. If this is so, then broken windows theory is untenable because it is tautological to claim that crime causes itself.

The only way for the BWT to survive is for there to be a careful refinement of the measures of disorder and crime (Kubrin 2008). Wilson and Kelling (1982) outlined an interesting theoretical framework but they failed to take the quintessential step of defining their terms. It is perhaps not surprising, then, that there has been wide variation in the items used as indicators of disorder and of crime (see Armstrong and Katz 2010; Gau and Pratt 2008, 2010; Giacopassi and Forde 2000; Reisig and Parks 2004; Sampson and Raudenbush 1999; Taylor 2001; Worrall 2006a). It is difficult to test for a relationship between an independent variable and a dependent variable when one is not sure what the independent variable actually is or how it should be measured.

It also might simply be time to retire the BWT from active duty on the criminological stage. There is at this point no solid reason to believe – and plenty of reason not to believe – that disorder causes crime in the manner stated by Wilson and Kelling (1982). In addition to the apparent absence of discriminant validity between the BWT’s core theoretical constructs, there are identifiable conceptual problems with the thesis. The broken windows idea was based on the notion that if a rock is thrown through a window and the damage done to the window is not repaired immediately, then pretty soon all the windows will be shattered. This seems logical on its surface, but there is a troubling question embedded in this scenario: Who threw the rock? If people are prowling about a neighborhood and maliciously throwing rocks at things, then it seems that a problem is already in place and that the newly smashed window is incidental to deeper, more pressing issues. Misplacing attention on the window diverts effort and resources away from identifying and addressing the genuine, underlying problems in a community that allow both disorder and crime to take root and flourish (Sampson and Raudenbush 1999).

Policy Implications

As stated at the outset of this research paper, the broken windows thesis is a policy-based proposal (hence the limited attention paid to theoretical and measurement issues). Wilson and Kelling (1982, p. 36) made police the centerpiece of disorder and crime reduction by arguing that “Though citizens can do a great deal, the police are plainly key to order-maintenance.” Broken windows theory has had a profound effect on the field of policing and has revolutionized some departments’ missions and functions.

There is evidence that broken windows policing (often also called order maintenance policing, quality-of-life policing, proactive policing, and zero tolerance policing, though some broken windows advocates take issue with this last label due to its negative connotation) can reduce some types of crime, some of the time, in some jurisdictions. Sampson and Cohen (1988) discovered a relationship between order maintenance policing and reductions in robbery rates, though they were not entirely sure that this effect was attributable to deterrence or to broken windows with respect to the underlying theoretical mechanism. Braga et al. (1999) found that a disorder crackdown seemed to reduce crime, yet paradoxically did not reduce disorder or nuisance offending. Worrall’s (2006b) analysis indicated that arrest rates for public order offenses had an immediate suppressive impact on assault, a delayed impact on burglary, and no effect on larceny. Kubrin et al. (2010) discovered a relationship between proactive policing and robbery rates. On the other hand, Novak et al. (1999) studied a disorder crackdown and found no evidence that the crackdown reduced any type of serious crime.

An important aspect of existing evaluations of disorder-based policing is the absence of attention paid to the mediating variables that were central to Wilson and Kelling’s (1982) original specification of the BWT. True order maintenance policing that reduced crime in a way consistent with Wilson and Kelling’s predictions would do so by enhancing social cohesion among area residents and stimulating social interaction and the frequenting of public spaces by law-abiding persons whose collective informal surveillance provided guardianship for all those present. Absent evidence that cohesion and informal social control emerged as a result of a concerted disorder-reduction effort by police, then order maintenance policing is empirically indistinguishable from crackdowns and hotspots policing (e.g., see Sherman and Weisburd (1995), where a randomized crackdown-backoff tactic in crime hot spots reduced crime and disorder simply through sporadic, unpredictable increases in police presence).

Another alternative theoretical mechanism by which order maintenance policing might reduce crime in a manner inconsistent with the BWT was hinted at by Kelling and Coles (1996). They noted that nearly one quarter of the people arrested by police for disorderly behaviors were either carrying illegal weapons or had outstanding arrest warrants in connection to serious crimes. Order maintenance policing, then, might simply be a form of net widening that legitimizes police interference in people’s freedom of movement for myriad reasons and thus permits them greater power to investigate people they feel as somehow suspicious or worthy of scrutiny. Net widening might accomplish the goals laid out by the BWT, but not via the same means, and therefore, it is not proper to call this “broken windows policing,” strictly speaking.

The evidence to date that disorder and crime lack discriminant validity results also bears on the issue of citizen satisfaction with the police. If citizens do not see disorder as a problem apart from crime, then they are unlikely to understand the logic behind order maintenance policing. Victims of crimes like robbery and residential burglary may be unlikely to fully endorse police crackdowns on jaywalking and littering. Indeed, though citizens rate order maintenance as an important endeavor (Skogan and Hartnett 1997), they still see traditional law enforcement against serious crimes as the primary role of the police (Skogan 1990; Webb and Katz 1997). A tunnel-vision focus on disorder could discredit police in the eyes of community residents.

Finally, the possibility that the discriminant validity of disorder and crime is dependent upon neighborhood context (specifically, the amount of disorder already present in the area) implies that order maintenance policing may be appropriate for better-off areas and inappropriate for more disadvantaged neighborhoods. Bucolic neighborhoods luxuriously free of dire problems might be an optimal place for police to adopt a quality-of-life approach in order to maintain the status quo and prevent disorder from seeping in. These are the places, too, where residents are most likely to hold positive attitudes toward police and to be willing to help police keep the social order in check.

By contrast, areas wherein disorder has already left its mark may, perhaps ironically, be the worst candidates for order maintenance policing simply because of the volume and severity of incivilities present in the area. Order maintenance policing places vast discretion in the hands of police and charges them with a mandate to intervene in the lives of people who do not pose an immediate or obvious threat to public safety. Exacerbating the danger inherent in the combination of discretion and low visibility is that order maintenance has been interpreted by many police departments as being best accomplished via the widespread use of Terry-type stop and frisks for suspicious behavior.

The convergence of these conditions – discretion, a disorder mandate, and reliance on Terry stops – has implications for police treatment of citizens in disadvantaged areas. Neighborhood-level disadvantage may encourage police misconduct (Kane 2002) and the use of more severe forms of coercive force (Terrill and Reisig 2003) by giving police the impression that the entire area has descended into lawlessness (Klinger 1997) and by offering the promise of impunity for bad deeds, given the social and political powerlessness of the people who reside there. In some disadvantaged, high-crime areas, the use of Terry stops has been racially discriminatory (Fagan and Davies 2000). Already-rocky relationships between the police and community can be further strained when officers stop people for what the people themselves see as “nothing” and, especially, when officers are rude and disrespectful during these encounters. This can worsen existing tensions and lead to even more crime and disorder as police systematically undercut their own legitimacy and authority (Gau and Brunson 2010).


When Wilson and Kelling wrote their Atlantic Monthly magazine article in 1982, they wanted the police to do something that the police could not do at that time: focus their efforts on the eradication of disorder. The professional model of policing and several US Supreme Court rulings had limited police duties to reactive response to calls for service and to the identification and arrest of people suspected of having committed serious crimes. In order to legitimize a broader conceptualization of the police role and to credibly argue that police can and should intervene in disorderly activities, Wilson and Kelling posited a restructuring of the disorder-crime relationship such that these two problems were linked sequentially and causally with disorder being a precursor to crime. The broken windows thesis thus casts disorder as an independent variable and crime as a dependent variable.

In order for the BWT’s formulation of disorder and crime to be accurate, though, disorder and crime must be empirically separate phenomena; that is, they must possess discriminant validity. Preliminary evidence that discriminant validity is not present arises from the fact that many activities technically classified as disorder are, in fact, criminal by statute or ordinance. The broken windows thesis, however, is grounded in subjective, individual perceptions of disorder, so the thesis could survive the legal criticism if people see a difference irrespective of legal overlap.

Studies employing confirmatory factor analysis – a statistical modeling procedure that permits tests of factor structure and examinations of discriminant validity – have found either mixed support or no support for the central BWT proposition that people discriminate between disorder and crime. Gau and Pratt (2008) discovered that the disorder and crime factors were too highly correlated to justify separation and that disorder and crime were, therefore, both part of a single factor. Worrall (2006a); Armstrong and Katz (2010) reported that one – and two-factor models alike fell short of the criteria necessary to establish good fit. Discriminant validity was either weak or was absent entirely across the different sets of models the authors ran. Finally, Gau and Pratt (2010) constructed a score measuring the difference between people’s perceptions of disorder and those of crime and found that neighborhood disorder was a significant predictor of that difference score such that people living in low-disorder areas had trouble distinguishing between the two problem types, while those in high-disorder areas did seem to draw a line between them.

The apparent absence of discriminant validity between disorder and crime, and the evidence suggesting that perceptual discrimination may vary by neighborhood type, has several implications for the broken windows thesis and the policing strategy arising therefrom. The thesis itself needs to be refined and its measures more thoroughly defined. More radically but nonetheless on the table as an option, the thesis needs to be abandoned in favor of alternative explanations for the covariance between disorder and crime that are better conceptualized and, therefore, hold more promise for theoretical validity and useful policy. Order maintenance policing might, moreover, hold more promise in areas wherein disorder is not currently present and can be kept out of the neighborhood in a preventive fashion. Once disorder has taken hold, however, a new strategy might be required.

In sum, the broken windows thesis has focused national attention on disorder and crime. It is clear from a long line of research that both matter greatly to people’s fearfulness and their satisfaction with their neighborhoods of residence; however, it is time to question the causal, disorder-to-crime sequence posited by Wilson and Kelling’s (1982) broken windows thesis. It simply might not be true.

Researchers and practitioners should consider alternative theoretical models that hold promise for reducing disorder and crime and improving people’s quality of life.


  1. Armstrong T, Katz C (2010) Further evidence on the discriminant validity of perceptual incivilities measures. Justice Q 27:280–304
  2. Braga AA, Weisburd DL, Waring EL, Mazerolle LG, Spelman W, Gajewski F (1999) Problem-oriented policing in violent crime places: a randomized controlled experiment. Criminology 37:541–580
  3. Fagan J, Davies G (2000) Street cops and broken windows: Terry, race, and disorder in New York City. Fordham Urban Law J 28:457–504
  4. Garofalo J, Laub J (1979) The fear of crime: broadening our perspective. Victimology 3:242–253
  5. Gau JM (2010) Basic principles and practices of structural equation modeling in criminal justice and criminology research. J Crim Justice Educ 21:136–151
  6. Gau JM, Brunson RK (2010) Procedural justice and order maintenance policing: a study of inner-city young men’s perceptions of police legitimacy. Justice Q 27:255–279
  7. Gau JM, Pratt TC (2008) Broken windows or window dressing? Citizens’ (in)ability to tell the difference between disorder and crime. Criminol Public Policy 7:163–194
  8. Gau JM, Pratt TC (2010) Revisiting broken windows theory: examining the sources of the discriminant validity of perceived disorder and crime. J Crim Justice 38:758–766
  9. Giacopassi D, Forde DR (2000) Broken windows, crumpled fenders, and crime. J Crim Justice 28:397–405
  10. Gomez R, Burns LG, Walsh JA, Hafetz N (2005) A multitrait-multisource confirmatory factor analytic approach to the construct validity of ADHD and ODD rating scales with Malaysian children. J Abnorm Child Psychol 33:241–254
  11. Harcourt BE (2001) Illusion of order: the false promise of broken windows policing. Harvard University Press, Cambridge
  12. Harcourt BE, Ludwig J (2006) Broken windows: new evidence from New York City and a five-city social experiment. Univ Chicago Law Rev 73:271–320
  13. Kane RJ (2002) The social ecology of police misconduct. Criminology 40:867–896
  14. Kelling GL, Coles CM (1996) Fixing broken windows. Simon & Schuster, New York
  15. Kelling GL, Sousa WH Jr (2001) Do police matter? An analysis of the impact of New York City’s police reforms. The Manhattan Institute, New York
  16. Klinger DA (1997) Negotiating order in patrol work: an ecological theory of police response to deviance. 35:277–306
  17. Kubrin CE (2008) Making order of disorder: a call for conceptual clarity. Criminol Public Policy 7: 203–214
  18. Kubrin CE, Messner SF, Deane G, McGeever K, Stucky TD (2010) Proactive policing and robbery rates across U.S. cities. Criminology 48:57–97
  19. Novak KJ, Hartman JL, Holsinger AM, Turner MG (1999) The effects of aggressive policing of disorder on serious crime. Polic An Int J Police Strateg Manag 22:171–189
  20. Piquero A (1999) The validity of incivility measures in public housing. Justice Q 16:793–818
  21. Pratt TC, Cullen FT (2005) Assessing macro-level predictors and theories of crime: a meta-analysis. In: Tonry M (ed) Crime and justice, vol 32. University of Chicago Press, Chicago, pp 373–450
  22. Reisig MD, Parks RB (2004) Can community policing help the truly disadvantaged? Crime Delinq 50:139–167
  23. Sampson RJ, Cohen J (1988) Deterrent effects of the police on crime: a replication and theoretical extension. Law Soc Rev 22:163–189
  24. Sampson RJ, Raudenbush SW (1999) Systematic social observation of public spaces: a new look at disorder in urban neighborhoods. Am J Sociol 105:603–651
  25. Sampson RJ, Raudenbush SW (2004) Seeing disorder: neighborhood stigma and the social construction of broken windows. Soc Psychol Q 67:319–342
  26. Sherman LW, Weisburd D (1995) General deterrent effects of police patrol in crime “hot spots”: a randomized, controlled trial. Justice Q 12:625–648
  27. Skogan WG (1990) Disorder and decline: crime and the spiral of decay in American neighborhoods. University of California Press, Berkeley
  28. Skogan WG, Hartnett SM (1997) Community policing, Chicago style. Oxford University Press, New York
  29. Taylor RB (2001) Breaking away from broken windows. Westview Press, Boulder
  30. Terrill W, Reisig MD (2003) Neighborhood context and police use of force. J Res Crime Delinq 40:291–321
  31. Webb VJ, Katz CM (1997) Citizen ratings of the importance of community policing activities. Polic An Int J Police Strateg Manag 20:7–23
  32. Wilson JQ, Kelling GL (1982) Broken windows: the police and neighborhood safety. The Atlantic Monthly, March 29-38
  33. Worrall JL (2006a) The discriminant validity of perceptual incivility measures. Justice Q 23:360–383
  34. Worrall JL (2006b) Does targeting minor offenses reduce serious crime? A provisional, affirmative answer based on an analysis of county-level data. Police Q 9:47–72
  35. Xu Y, Fiedler ML, Flaming KH (2005) Discovering the impact of community policing: the broken windows thesis, collective efficacy, and citizens’ judgment. J Res Crime Delinq 42:147–186

See also:

Free research papers are not written to satisfy your specific instructions. You can use our professional writing services to buy a custom research paper on any topic and get your high quality paper at affordable price.


Always on-time


100% Confidentiality
Special offer! Get discount 10% for the first order. Promo code: cd1a428655