NEIGHBOURHOOD CHARACTERISTICS AND ... - Oxford Journals

doi:10.1093/bjc/azi096

BRIT. J. CRIMINOL. (2006) 46, 719–742 Advance Access publication 18 November 2005

NEIGHBOURHOOD CHARACTERISTICS AND REPORTING CRIME Effects of Social Cohesion, Confidence in Police Effectiveness and Socio-Economic Disadvantage1 H E IK E G OU D R IAA N , K A R I N W I T T E B R O O D and P A UL N I E U W B E E R T A * Various scholars have suggested that neighbourhood social cohesion and confidence in police effectiveness influence the probability that victims report crime to the police, but this has never been properly tested. Neighbourhood socio-economic disadvantage is also often assumed to influence reporting, but empirical support is limited. This study examines the effects of these three characteristics on Dutch victims’ reporting decision. Data from a large-scale victimization survey are merged with data on characteristics of neighbourhoods to test the hypotheses. Hierarchical logistic modelling is used to analyse the nested data. The results show that, in addition to crime and victim features, neighbourhood social cohesion and socio-economic disadvantage affect reporting. Neighbourhood confidence in police effectiveness does not have an effect. Approximately 25 per cent of the people in the Western world are crime victims every year and about one in five of them is victimized more than once (Van Kesteren et al. 2000). Many of these crimes are never reported to the police (Goudriaan et al. 2004). In the Netherlands, for example, only slightly more than a third of the crimes were reported to the police by or on behalf of the victims in 2002 (Eggen 2003).2 Since the bulk of information on crime reaches the police via victims and witnesses (Greenberg et al. 1982; Bennett and Wiegand 1994), this means that a sizeable amount of information remains concealed from the police and courts. Even though victim reports are the most important source of information on crime for the police (Hindelang and Gottfredson 1976; Greenberg and Ruback 1992; Warner 1992; Mayhew 1993), the fact that almost two-thirds of the crimes are not reported to the police is not, by definition, a problem. Victims apparently do not always feel a need to report a crime. Some crimes, for instance, are not very serious and victims do not feel they require any further police or court efforts.

* Heike Goudriaan is an assistant professor in the Department of Criminal Law and Criminology at Leiden University, Karin Wittebrood is Senior Researcher at the Social and Cultural Planning Bureau (SCP) in The Hague, and Paul Nieuwbeerta is Senior Researcher at the Netherlands Institute for the Study of Crime and Law Enforcement (NSCR) in Leiden. Correspondence address: Heike Goudriaan, Leiden University, Faculty of Law, Department of Criminal Law and Criminology, PO Box 9520, 2300 RA Leiden, the Netherlands; e-mail: [email protected]. 1 The authors are grateful to Eric Baumer, Wim Bernasco, Ferry Koster, Kim Ménard and the anonymous reviewers for helpful comments and suggestions on a previous draft of this article. Furthermore, they would like to thank the Royal Netherlands Academy of Arts and Sciences (KNAW) for their financial support. 2 In daily usage, we speak of reporting a crime, even if the police have only been informed of a crime and no charges have been filed. In this article, we do not draw a distinction between instances in which the police have been informed of a crime and those in which charges have been filed; both are referred to as reported crimes.

719 © The Author 2005. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected]

GOUDRIAAN, WITTEBROOD AND NIEUWBEERTA

Failure to report crimes can nonetheless have important consequences for the victims, as they deny themselves the option of turning to the criminal justice system or other facilities (Van der Vijver 1993). In addition, if victims’ reporting decision depends in part on their position in society, this causes social inequality in the access to public facilities. For example, neighbourhood differences in willingness to report crimes make efficient police efforts problematic (Baumer 2002), as it is all the more difficult for them to know where to best focus their scarce time and attention. Dozens of studies have been published on the relation between neighbourhood characteristics and delinquent behaviour (Sampson et al. 2002). Surprisingly, only a handful of studies have been conducted on the influence of neighbourhood characteristics on victims’ reporting behaviour. In research on the determinants of the reporting decision, it is often assumed that victims calculate the costs and benefits and that the most important factors in their calculations are related to the nature of the specific crime. Most studies concentrate on the effects of the perceived severity and the financial damages, physical injury and psychological harm caused by the crime (Skogan 1976; 1984; Sparks et al. 1977; Fiselier 1978; Van Dijk and Steinmetz 1980; Kury et al. 1999; Pino 1999). Only a few studies focus on characteristics of the victims themselves or the neighbourhood they live in (Baumer 2002). Moreover, the studies that do examine effects of neighbourhood characteristics predominantly focus on a single characteristic, i.e. neighbourhood socio-economic disadvantage, and hardly give attention to other neighbourhood characteristics that can be assumed relevant, e.g. the degree of social cohesion in the neighbourhood. The lack of interest in neighbourhood characteristics in research on crime reporting is striking, especially since the social sciences have devoted so much attention in recent decades to the influence of features of the environment. In various criminological studies, the effects of the level of social cohesion or informal social control and the socio-economic status of a neighbourhood on victimization and the fear of crime are examined (Sampson et al. 2002; Smith and Jarjoura 1989; Rountree et al. 1994; Rountree and Land 1996; Sampson et al. 1997; Lee 2000; Lee and Earnest 2003). These studies measure the dependent variable at the individual level and the explanatory variables include individual features as well as aggregated measures of neighbourhood characteristics. In recent years, multilevel studies of this kind have also been conducted in the field of victimization research in the Netherlands (e.g. Wittebrood 2000; Maas-de Waal and Wittebrood 2002; Van Wilsem 2003). The aim of our study is to continue to build upon the growing knowledge on the effects of neighbourhood characteristics in the field of criminology and apply this knowledge in research on victim reporting behaviour. Elaborating on theories on the influence of social cohesion in a neighbourhood (Sampson et al. 1997) and social stratification (Black 1976; Anderson 1999), among other things on individual behaviour, we first examine whether the level of social cohesion as well as confidence in police effectiveness and the socio-economic disadvantage in neighbourhoods affect the probability that crime victims report to the police. Secondly, we examine the extent to which the effect of neighbourhood socio-economic disadvantage on the probability that victims contact the police—as found by Baumer (2002)—is mediated by neighbourhood social cohesion and confidence in police effectiveness. These research questions are answered using data from multiple sources. Information on more than 100,000 victims and incidents stems from the nationwide Dutch 720

NEIGHBOURHOOD CHARACTERISTICS AND REPORTING CRIME

Police Population Monitor (Politiemonitor Bevolking), which has been held biannually since 1993. Respondents are questioned about their experiences with different types of crime and their reporting behaviour. For the present purpose, these data are unique, since precise information on the addresses and neighbourhoods in which victims live is available, which could be coupled with adequate information on distinct characteristics for each neighbourhood. Following the example of Raudenbush and Sampson (1999) and using data from large-scale surveys, ecometric analyses have been conducted to give each neighbourhood a score on the degree of social cohesion and confidence in police effectiveness. For the socio-economic disadvantage, an index based on official data has been used. We are unaware of other databases in the Netherlands, Europe or the United States that contain all this information. As a result, our study is the first to separately measure three neighbourhood characteristics that are considered theoretically relevant—social cohesion, confidence in police effectiveness, and socio-economic disadvantage (cf. Baumer 2002)—and the first to examine their effects simultaneously in a multilevel analysis (cf. Gottfredson and Hindelang 1979; Bennett and Wiegand 1994). The Decision to (Not) Report a Crime In theories on the decision-making process of victims on whether or not to report a crime to the police, a victim’s decision is often assumed to be rational. In other words, the lower the costs of reporting a crime and the higher the anticipated results, the more likely victims are to report a crime to the police (Skogan 1984; Gottfredson and Gottfredson 1988; Felson et al. 2002). However, this notion should not be taken too literally. It is not unusual for a decision like this to be made quite impulsively, and a wide range of emotions can play a role. The point of departure, though, is that victims consider the costs and benefits when making a decision (Felson et al. 2002). The nature and severity of crimes play a central role in cost-benefit theories when explaining the willingness of victims to report to the police. In addition to devoting attention to more precise crime features, recent literature on reporting also focuses on the role of the social environment in victims’ decision whether or not to report the crime. Ruback and Greenberg, for example, devote attention to the important role of family and other social relations and networks (Ruback et al. 1984; Greenberg and Ruback 1992). Especially in stressful situations like the one crime victims are in, victims tend to listen to the advice and opinions of the people around them (Greenberg and Ruback 1992). Furthermore, recently, interest in the social and economic characteristics of the neighbourhood people live in has grown. Influence of Neighbourhood Characteristics on Reporting To our best knowledge, eight empirical studies have been published up to now that examine the effects of neighbourhood characteristics on crime reporting (Fishman 1979; Gottfredson and Hindelang 1979; Laub 1981; Warner 1992; Bennett and Wiegand 1994; Avakame et al. 1999; Ruback and Ménard 2001; Baumer 2002). Most of these studies analyse data from national or local victim questionnaires from the United States. Bennett and Wiegand (1994) conducted their study among victims in Belize 721


City—the largest city in Belize, a developing nation in Central America. The study by Fishman (1979) is based on a victim questionnaire in the Israeli city of Haifa. Three neighbourhood characteristics play a central role in these studies: (1) social cohesion; (2) confidence in police effectiveness; and (3) socio-economic disadvantage. Neighbourhood socio-economic disadvantage has been given most attention. Most likely, this is due to the fact that these studies did not use victimization data with direct and reliable measures for social cohesion or confidence in police effectiveness; nor did they use distinct data on these neighbourhood characteristics that could be merged with victimization data (cf. Baumer 2002). Nevertheless, from current theories on reporting, all three characteristics can be assumed to have important and distinct effects on victims reporting decision. Social cohesion The idea that neighbourhood social cohesion is of importance for victims’ reporting behaviour is mainly drawn from the classic social disorganization model (Shaw and McKay [1942] 1969), which assumes that strong informal social control in neighbourhoods is an important mechanism for regulating conduct and mediating interpersonal disputes (Baumer 2002). Social cohesion in neighbourhoods has been proven an important context for the realization of informal social control (Sampson et al. 1997; Morenoff et al. 2001; Silver and Miller 2004). Various scholars have suggested that neighbourhood informal social control, and thus also social cohesion, are important for crime reporting (e.g. Conklin 1975; Black 1976; Gottfredson and Hindelang 1979; Baumer 2002), but this assumption has never been tested properly. Two lines of thought on the nature of the relationship between social cohesion and reporting have appeared in the literature, resulting in two contradictory hypotheses. First, in neighbourhoods with limited social cohesion, there is less of the kind of informal organization that could enforce public order in a neighbourhood, such as collective (informal) social control (Vélez 2001), social capital (Sampson et al. 1999) or collective efficacy (Sampson et al. 1997). Some have argued that this results in difficulties securing an adequate share of various public services, such as formal police protection (Baumer 2002). Therefore, it is often thought that the less social cohesion there is in a neighbourhood, the less easily residents have access to agencies of formal control such as the police (Rose and Clear 1998). This results in our first hypothesis: the lower the social cohesion in a neighbourhood, the lower the probability that crime victims who live there report to the police (H1) (cf. Rose and Clear 1998; Baumer 2002). It is also possible, though, to derive a contradictory hypothesis on the effect of social cohesion in a neighbourhood on the individual reporting decision, i.e. that less social cohesion leads to more reporting. This line of thought is not uncommon in studies on differences in reporting percentages between urban and rural regions in the United States. The residents of urban regions appear to feel more dependent on formal police control than residents of rural regions, who can rely more on the support of their direct environment (Boggs 1971; Laub 1981). In towns and neighbourhoods where social cohesion is limited, residents are thought to feel more of a need for formal social control mechanisms to help solve the problems they are confronted with. According to this line of thinking, residents in neighbourhoods of this kind are more apt to ask the police for help in solving conflicts, preventing repeat victimization and punishing criminals 722


(Conklin 1975; Black 1976; Gottfredson and Hindelang 1979; Laub 1981). It can thus also be hypothesized that the lower the social cohesion in a neighbourhood, the greater the probability that crime victims who live there report to the police (H2). Barely any empirical research has been conducted in which the influence of social cohesion in neighbourhoods on reporting has been tested directly. Four of the six studies that examine this relationship use an extremely rough measure of social cohesion, namely the population density of the victim’s place of residence or the town where the crime was committed Gottfredson and Hindelang 1979; Laub 1981; Avakame et al. 1999; Ruback and Ménard 2001). These studies work from the assumption that there is less social cohesion in more urbanized areas and therefore a greater probability that victims will report. Only Laub (1981) found a significant positive effect. The study by Bennett and Wiegand (1994) does include a direct measure of social cohesion, but the scholars do not observe any significant correlation between cohesion and reporting. Finally, the nationwide study by Baumer (2002) in the United States, which is the most advanced in a methodological sense, is somewhat disappointing because it only includes a single indicator that is supposed to measure social cohesion as well as socio-economic disadvantage. This makes it impossible to disentangle the effects of the two separate factors. Confidence in police effectiveness Another neighbourhood characteristic that is assumed to affect the probability that victims report crimes to the police is the confidence in police effectiveness in a neighbourhood (e.g. Bennett and Wiegand 1994; Baumer 2002; Goudriaan et al. 2004). As is the case with the influence of social cohesion, this assumption has never been properly tested. Many have suggested that victims will not estimate the benefits of reporting crimes as highly if they have low confidence in police effectiveness (Hagan and Albonetti 1982; Sherman 1993; Anderson 1999; Baumer 2002). These estimated benefits of reporting to the police are assumed to depend to a large extent on the victims’ individual judgment of the police. In the event of doubts or in stress situations, however, victims will allow their decision to depend in part on the judgment of their social environment (Ruback et al. 1984; Greenberg and Ruback 1992). The estimated benefits of reporting crimes to the police can therefore generally assumed to be lower (and the costs higher) for residents of neighbourhoods with a lower confidence in police effectiveness than for similar residents of neighbourhoods with a more positive perception of the police. The resulting hypothesis is thus that the less confidence people in a neighbourhood have in police effectiveness, the lower the probability that crime victims who live there report to the police (H3) (cf. Conklin 1975). Of the eight studies on the effects of neighbourhood characteristics on crime reporting, the study by Bennett and Wiegand (1994) is the only one that empirically tested this assumption—but no significant effect of confidence in police effectiveness in the neighbourhood was found. However, Bennett and Wiegand did find a positive relation between reporting and victims’ attitudes toward the police at victim level. Socio-economic disadvantage: direct effect A third neighbourhood characteristic that can be assumed to affect the probability of reporting is the level of socio-economic disadvantage. This idea is not exactly new 723


(e.g. Rousseau 1762) and often empirically tested. Three decades ago, Black summarized these ideas in The Behavior of Law (1976), which has become a classic. He formulated a general sociological theory on the extent to which the use of the law varies in different social contexts such as neighbourhoods. His theory is extremely comprehensive. By use of the law, he not only means reporting a crime to the police, but he also means asking for formal help from the police and how the police themselves deal with victims, witnesses or suspects and the various aspects of a court trial. One of Black’s most important propositions—known as the stratification hypothesis—is that a neighbourhood’s socio-economic disadvantage affects the extent to which use is made of the law. The lower the socio-economic status of a neighbourhood, the less use its residents make of the law in solving their problems and the more frequently they deal with them themselves (for an extensive review of the theory and its testability, see Gottfredson and Hindelang 1979). With respect to reporting, it can thus be hypothesized that the greater the socio-economic disadvantage in a neighbourhood, the lower the probability that crime victims who live there report to the police (H4). In line with Black’s stratification hypothesis, Baumer (2002) also works from the assumption that the extent of socio-economic disadvantage in a neighbourhood affects the probability that its residents report crimes. However, based on the results of an ethnographic study by Anderson (1999) on daily life in the inner city of Philadelphia, he assumes that this relationship is especially strong in extremely socio-economic disadvantaged neighbourhoods. Up to now, there has been little empirical support for this stratification hypothesis. Three of the five studies examining the relationship between neighbourhood socioeconomic disadvantage and reporting do not find a significant relationship (Gottfredson and Hindelang 1979; Warner 1992; Bennett and Wiegand 1994). The two other studies (Fishman 1979; Baumer 2002) note weak correlations. In Haifa, Israel, Fishman (1979) compares the reporting behaviour of victims from five ‘good’ and five ‘bad’ neighbourhoods. He finds a small effect of neighbourhood socio-economic disadvantage on reporting regarding various types of crimes. The effect he finds, however, is an indirect effect via crime severity and victims’ attitude to the police. Baumer (2002) notes a direct negative relationship between a neighbourhood’s socio-economic disadvantage and the reporting behaviour of victims of simple assaults, but only in cases of extreme disadvantage, as he had presumed. Socio-economic disadvantage: indirect effect It is unclear what mechanisms are responsible for the relationship between socio-economic disadvantage and reporting as found by Baumer (2002) and Fishman (1979). It is unlikely that victims from socio-economic disadvantaged neighbourhoods are lacking in money needed to report a crime, since reporting is free of charge. Other explanations are suggested in the literature. What they all amount to is that the relationship between socio-economic disadvantage and reporting is indirect and runs via an intermediary factor. The social cohesion and the confidence in police effectiveness in the neighbourhood are two obvious intermediary factors (Rose and Clear 1998; Baumer 2002). These ideas have not been tested before. It seems plausible that there is a negative relationship between neighbourhood socio-economic disadvantage and neighbourhood social cohesion or social organization. 724


The residents’ limited material and political recourses in disadvantaged neighbourhoods are assumed to lead to an incapacity for social organization. This is the classic thesis on which the social disorganization theory has been developed (Shaw and McKay [1942] 1969; see, for a recent survey, Bursik and Grasmick 1993) and which is confirmed in recent research. Residents of neighbourhoods with extreme socio-economic disadvantage have less social contact with each other (Bellair 1997; Sampson et al. 1997; Morenoff et al. 2001) and participate less in local organizations (Sampson and Groves 1989). Therefore, part of the hypothesized effect of neighbourhood socio-economic disadvantage on the probability that crime victims report to the police might be explained by differences in social cohesion in these neighbourhoods. The same might be the case with confidence in police effectiveness. According to a study by Sampson and Bartusch (1998), neighbourhoods of concentrated disadvantage display elevated levels of legal cynicism, dissatisfaction with police and tolerance of deviance unaccounted for by socio-demographic composition and crime-rate differences. These findings are in keeping with general theories on anomie, strain and subcultures of crime, and with the work of Anderson (1999) and Baumer (2002). Due to the high level of poverty and unemployment and limited labour market opportunities, residents of socio-economic disadvantaged neighbourhoods, especially youth and immigrants, are felt to be alienated from the general norms of society. In neighbourhoods of this kind, own norms and codes of conduct emerge. Anderson refers to them as ‘codes of the street’. One of the things this code tells people is how to deal with and respond to crime. Particularly in extremely socio-economic disadvantaged neighbourhoods, residents are expected to be personally responsible for their own safety and the safety of their property. In these neighbourhoods, it would be weak or even cowardly to go to the police and expect them to help solve any problems (Anderson 1999; Baumer 2002). This kind of subculture with a negative attitude to the police might be reinforced by the residents of socio-economic disadvantaged neighbourhoods being more likely to work off the books, deal drugs, or engage in other criminal activities. After all, they have less access to legitimate economic options. This might also keep them from contacting the police if they themselves are crime victims, since they do not want the authorities to find out about their own activities (Skogan 1984; Anderson 1999). The subculture keeps people who do not engage in criminal activities but who live in disadvantaged neighbourhoods from calling the police, knowing it might get other neighbourhood residents into trouble (Sparks et al. 1977; Wright and Decker 1997). The residents of disadvantaged neighbourhoods can thus be expected to have less confidence in police effectiveness, which results in the hypothesis: If the effects of socio-economic disadvantage, social cohesion and confidence in police effectiveness are examined simultaneously, the net effect of the disadvantage is smaller than if only the disadvantage is examined (H5). This study is the first to separately measure the three neighbourhood characteristics that are considered theoretically relevant—social cohesion, confidence in police effectiveness and socio-economic disadvantage—(cf. Baumer 2002) and the first to examine their effects simultaneously in a multilevel analysis (cf. Gottfredson and Hindelang 1979; Bennett and Wiegand 1994). What is more, this study not only tests hypotheses on the effects of neighbourhood characteristics on the reporting of violent crimes, as is often done, but it also assesses their effects on reporting other types of crime. 725


Data Various data sources have been used in this study to make it possible to test the hypotheses. First, we used the Police Population Monitor, a nationwide Computer Assisted Telephone Interviewing (CATI) survey in the Netherlands. It has been conducted every two years since 1993, with samples of 75,000 respondents on average. Households from which respondents are drawn are selected by randomly picking telephone numbers (within police regions) from the national telephone register. Within the household, the person who is first to celebrate his or her (at least 16th) birthday is selected as respondent. If the selected person is not at home, he or she is interviewed at a different time.3 The Police Population Monitor is the largest questionnaire survey on a subject related to public safety and crime in the Netherlands and also one of the largest in the world. As the survey is conducted on such a large scale, regional comparisons can be made throughout the Netherlands at a detailed level (i.e. neighbourhoods). For this study, the 1995, 1997, 1999 and 2001 data files have been merged into one file with 110,950 victims.4 Overall, in the Netherlands, there are no temporal trends in reporting over these years (Goudriaan et al. 2005). Respondents are asked about their experiences with 12 types of crime in the 12 months preceding the survey: bicycle theft, car theft, theft of items from a car, damage to the car, non-violent robbery, violent robbery, attempted burglary, burglary, other theft, other damage, threats and assaults. If respondents say they were victimized more than once in the previous year, they are asked additional questions about the most recent incident only, including whether or not they reported the crime to the police. Consequently, the data file used for the present study includes one crime incident per victim, even when the victim has been multiply victimized. The survey also includes social and demographic information on the respondents. The Police Population Monitor includes the four-digit zip codes of the respondents’ home addresses so that the neighbourhood they live in is known. There are 3,990 zip code areas in the Netherlands.5 Even though these geographical units have primarily been designed as administrative units by the postal services and they might not always be an optimal way to indicate neighbourhoods (e.g. some zip code areas cover a builtup as well as a rural area), they are the best available nationwide classification of neighbourhoods in the Netherlands (Knol 1998) and have successfully been used in studies examining neighbourhood effects (e.g. Wittebrood 2000; 2004; Van Wilsem 2003; Bernasco and Nieuwbeerta 2005). The average population size in the zip code areas is 4,907 inhabitants and the average number of households is 2,104. For data on respondents’ neighbourhoods, use is made of supplementary sources, the Residential Environment Data Base (Woonmilieudatabase) and the (2002) Residential Needs Survey (Woningbehoefteonderzoek) (VROM 2003). The Residential Environment 3 The average response rate between 1993 and present is 58 per cent. This is relatively high for a Dutch survey—the Netherlands is internationally notorious for its low response rates (Stoop 2005). In the data collection for the Police Population Monitor, a lot of effort has been made to increase response rates. Studies on the reliability and validity of this monitor found no indication that selection mechanisms in non-response are present (Politiemonitor Bevolking 2001). 4 In total, there were 317,954 respondents, of which 35 per cent had been victimized at least once in their own town or city (the population of interest) in the previous 12 months. 5 This is the situation in 2001. The areas are stable over time, but due to the development of new suburbs, 60 new zip code areas have been introduced between 1995 and 2001.

726


Data Base is a compilation of official data on neighbourhoods from various statistical agencies. The Residential Needs Survey is a survey conducted every four years among a representative sample of the population on their residential situation and desires. The neighbourhood data are linked to the data from the Police Population Monitor using the four-digit zip codes. The analyses are conducted on crime victims and neighbourhoods without missing data on the relevant variables. We therefore had to exclude 8.4 per cent of the cases, which resulted in a file with 101,592 victims and the same number of crimes in 3,104 neighbourhoods.6 Operationalization Reporting victimization The response to the question about whether (1) or not (0) the crime was reported to the police serves as the dependent variable in our study.7 Forty-three per cent of the victimizations are reported to the police (Table 1). Only crimes experienced in victims’ own town are taken into account, since neighbourhood characteristics are assumed to especially influence incidents that occur in the victim’s own environment. Features of criminal offences If we are to adequately examine the extent to which differences in reporting percentages are related to neighbourhood characteristics, we need to take into account the types and severity of the crimes. First, we distinguish between the 12 types of crimes the respondents are asked about. Over 90 per cent of the crimes are property crimes (see Table 1). In our analyses, we include all these types together with dummies to control for the type of crime.8 The dummy for car theft, for example, is 1 if the crime is a car theft and 0 if it is some other type of crime. Bicycle theft is the reference category. In addition to the type of crime, we use information on the financial damages and physical injury to the victim. The financial damage categories are: less than fl.100 (reference category), fl.100 to fl.499 and fl.500 or more.9 Dummies are included in the analyses for the last two categories. Victims of threats and assaults are not asked about financial losses, since these crimes often do not lead to financial damages. Therefore, the financial loss for these crimes is set at less than fl.100. As to the extent of physical 6 We excluded 27 victims (0.02 per cent) as we did not know their zip code, 2,271 victims (2.0 per cent) were excluded as the score on the dependent variable was missing, 4,633 victims (4.2 per cent) were excluded as we did not have information on the seriousness of the incident, 554 victims (0.5 per cent) were excluded due to missing data on victim characteristics, and 1,873 victims (1.7 per cent) who lived in 307 different neighbourhoods were excluded because we did not have information on their neighbourhood. Most of the neighbourhoods or zip code areas that could not be included in the analyses are in rural or industrial regions. 7 The crime can either have been reported to the police by victims themselves—as is the case in 77.4 per cent of the cases that are known by the police—by someone else (19.9 per cent), or the police can have been on the scene or discovered the crime themselves (2.8 per cent). If someone else has reported the crime, it is usually a family member who reported a crime that took place in the home (such as burglary). To test for the consequences, the analyses were also performed on (1) a file in which crimes discovered independently by the police were excluded (n = 100,359) and (2) a file in which crimes reported by someone other than the victim were also excluded (n = 91,496). These two extra analyses yielded very similar results. 8 The analyses have also been separately conducted for property crimes and violent crimes. The sample was too small to conduct meaningful analyses as regards violent crimes, and the results for property crimes barely deviate from the analyses of the sample for all the types of crimes collectively. This is why we have decided to address all 12 types of crimes together. 9 fl.100 = €45.38 (about £31 in October 2005).

727


TABLE 1

Descriptives of dependent and control variables (n = 101,592)

Variable

%

Variable

Reporting

43

Crime characteristics Type of crime bicycle theft car theft theft from car car vandalism robbery with violence robbery without violence attempted burglary burglary other theft other vandalism threat assault Financial lossb < fl.100 fl.100 – fl.499 ≥ fl.500 Physical injury none or little average severe

Victim characteristics Agea Male Low education level Paid job >15 hours/week Jobless or unable to work Non-native Household type multiple adults single person single parent (child(ren) < 15)

a b

48 33 55 4 3 80 18 2

%

18 1 6 29 0 4 7 4 11 13 6 1 53 23 24 99 0 0

The mean age is 41.85 years, the S.D. is 15.62, the minimum is 15, and the maximum is 98. fl.100 = €45.38 (about £31 in October 2005).

injury, also three categories are distinguished: no injury to slight injury not requiring medical treatment (reference category), medium-level injury requiring one-time medical treatment, and severe injury that required or still requires hospitalization or medical treatment more than once. Victims are only asked about physical injury in instances of violent crimes (violent robberies, threats and assaults). It is assumed that property crimes do not involve physical injury. Table 1 shows that more than half (53 per cent) of the crimes result in slight financial loss (less than fl.100). In approximately a quarter of the crimes, the damages amount to more than fl.500. Physical injury is less common, and only 0.5 per cent of the crimes result in substantial physical injury. This is largely because the majority of the offences are property crimes and they are assumed not to involve physical injury. Victim features In our analyses we also include a number of victim features that have proven in previous studies to have an important effect on victims’ reporting decision: sex (0 = female, 1 = male), age (in years), family composition (with multi-person households as reference category and dummies for the categories one-person households and single parents with one or more children under the age of 15), educational level (0 = middle-level vocational school or five or six-year high school or college/university, 1 = elementary school/lower-level vocational school or four-year high school), employment status (a dummy for unemployment and a dummy for a minimum of 15 hours of paid employment a week), and ethnic background (0 = native Dutch and 1 = non-Western immigrants). The non-Dutch respondents of Western descent (239 Belgians, 429 Germans 728


and 183 Brits) are categorized as native Dutch. Table 1 shows the distribution of the victim features, which largely coincides with data from previous studies on victims in the Netherlands (Wittebrood 1997) and other countries (Van Kesteren et al. 2000). Neighbourhood characteristics We have formulated hypotheses on the effects of three neighbourhood characteristics on crime victims’ reporting decision: the neighbourhood’s social cohesion, confidence in police effectiveness and socio-economic disadvantage. Several methods and data sources are used to measure the neighbourhood characteristics. To measure the level of social cohesion, data from the Residential Needs Survey are used. In this survey, respondents are asked to indicate the extent to which they agree with the following statements: (a) I feel an attachment to this neighbourhood, (b) I feel at home in this neighbourhood, (c) I have a lot of contact with the people who live next door, (d) I have a lot of contact with other neighbourhood residents, (e) I feel responsible in part for the neighbourhood being a pleasant place to live, (f) people are nice to each other in this neighbourhood, (g) I live in a pleasant neighbourhood with a sense of solidarity, (h) people in this neighbourhood hardly know each other and (i) I am satisfied with the composition of the population in this neighbourhood. Following the example of Raudenbush and Sampson (1999), we have conducted an ecometric analysis to give each neighbourhood in our file a score on social cohesion (see also Van Wilsem 2003). The aim of this type of analysis is to measure a characteristic of ecological units, in this case neighbourhoods, on the basis of survey data and to aggregate the data over responses to multiple items given by various respondents within each ecological unit. We assume that the internal consistency of an area-level scale not only depends on the correlation between the items, the number of items and their extent of difficulty, as with an individual-level scale, but also on the agreement between the respondents within the area and the size of the sample for each area. In practice, scale values for the neighbourhoods can be calculated by means of a multilevel analysis, with three levels: items, respondents and neighbourhoods. The predicted values on the neighbourhood level are then the new scale values. The constructed measure of social cohesion has a reliability (Cronbach’s alpha) of 0.80. The scale is centred on the entire sample (the average is 0) and has a minimum value in our data of –0.56 and a maximum value of 0.49 (see Table 2). To measure the neighbourhood residents’ confidence in police effectiveness, survey questions from the Police Population Monitor are used. Each respondent (n = 317,954) is asked T A B L E 2 Descriptives of neighbourhood characteristics (centred at individual-level means) (n = 101,592)

Socio-economic disadvantagea Social cohesion Confidence in police Percentage non-nativesa Mobilitya a

Mean

SD

Minimum

Maximum

0 0 0 0 0

1.44 0.18 0.09 1.26 0.52

−2.77 −0.56 −0.33 −1.31 −1.22

5.40 0.49 0.48 6.69 2.27

Centred and divided by ten.

729


to assess the extent to which they agree or disagree with the following 12 statements on the functioning, conduct and availability of the police: (a) the police protect the people of this neighbourhood, (b) the police have good contact with the residents of this neighbourhood, (c) the police respond to problems in this neighbourhood, (d) the police have an efficient approach here, (e) the police in this neighbourhood are doing their best, (f) the police are not tough enough here, (g) the police do not intervene here, (h) you don’t see the police often enough here, (i) they don’t get out of their patrol cars often enough here, (j) they are not easy to approach here, (k) the police in this neighbourhood don’t have enough time for all kinds of problems and (l) they don’t come quickly when you call them. To construct a measure in each neighbourhood for the confidence in police effectiveness, again an ecometric analysis has been conducted that is comparable to the one for the extent of social cohesion. The reliability of the scale is 0.83. In our data, the scale is centred on the entire sample and has a minimum of –0.33 and a maximum of 0.48 (see Table 2). The neighbourhoods’ socio-economic disadvantage is measured using four indicators from the 1998 Residential Environment Data Base: (a) the percentage of households with an income under the minimum of fl.14,000, (b) the percentage of households headed by an unemployed person, (c) the percentage of households whose head receives a benefit from the Welfare Department and (d) the percentage of one-parent families with minor children. To determine a socio-economic disadvantage score for each neighbourhood, the scores on the four indicators are summed and weighed for their factor load.10 The resulting scale scores are centred on the entire sample and divided by ten, so that the minimum in the data is –2.77 (no socio-economic disadvantage) and the maximum is 5.40 (extreme socio-economic disadvantage) (see Table 2). In addition to the three neighbourhood characteristics our hypotheses are formulated on, another two neighbourhood characteristics have been included in the analyses as control variables, i.e. ethnic heterogeneity and mobility. By including these characteristics as control variables, we hope to get a better picture of the influence of the neighbourhood characteristics we have formulated our hypotheses on. The percentage of non-Western immigrants is used as indicator of the ethnic heterogeneity in a neighbourhood.11 In the present data, this percentage varies from 0 (30 per cent of all neighbourhoods in the Netherlands) to 80 per cent. On average, 13.1 per cent of the neighbourhood residents are non-Western immigrants. For the analyses, the scores are centred and divided by ten, so that the minimum is –1.31 and the maximum 6.69 (see Table 2). The mobility in a neighbourhood is defined as the percentage of the total population who moved into the neighbourhood in a year. Moves within the neighbourhood are not counted. In our data, the scores range from 0 to 34.9 per cent and the average is 12.2 per cent. Again, the scores are centred and divided by ten so that the minimum is –1.22 and the maximum 2.27 (see Table 2). The data for both characteristics are from the 1998 Residential Environment Data Base.

10

The factor loads are respectively 0.76, 0.83, 0.86 and 0.79 (Cronbach’s alpha = 0.76). Van Wilsem (2003) used a similar measure for ethnic heterogeneity in his study on neighbourhood dynamics and criminal victimization. 11

730


Analysis Method Multilevel or hierarchic models are used to test the hypotheses (Goldstein 1995; Snijders and Bosker 1999). Other than traditional techniques, such as OLS regression, multilevel models take the layered (nested) structure in the data into account. In this study, two levels are distinguished: the level of the crime incident (and the victim), and the level of the neighbourhood where the victim lives. Measuring errors are separately specified at each of the two levels. In this way, the possibility is taken into account that individuals within neighbourhoods might be more alike (e.g. might have more similar attitudes toward the police) than individuals between neighbourhoods. Another advantage of multilevel modelling is that in estimating the parameters, the number of individuals in a neighbourhood is taken into consideration. Neighbourhoods with numerous victims weigh more heavily in the assessment than neighbourhoods with only a few victims. Since the dependent variable—whether or not a crime is reported—is a dichotomy, and the assumption is made that the distribution of the measuring errors at the level of the crime and the victim is binomial, logistic multilevel models are estimated with the first level variance fixed to one. The parameters are estimated using MLwiN 2.0, which has been especially developed for models of this kind (Rasbash et al. 2004). For an extensive explanation of the models used here, see Goldstein (1995) or Snijders and Bosker (1999). Results Before turning to the multivariate analyses, the percentage of crimes reported per type of crime is given in Table 3. Overall, 43 per cent of all crimes are reported to the police. As can be seen from the table, the percentage of reported crimes largely depends on the type of crime. The highest reporting percentage pertains to car theft: 97 per cent of the car theft victims report the crime to the police. Most victims of burglary and robbery with violence (over 80 per cent) also report these crimes to the police. (Car) Vandalism and ‘other’ theft are least often reported: fewer than three out of ten of these crimes are reported to the police. TABLE 3

Percentage of crimes reported to the police per type of crime

Type of crime

% reported

Other vandalism Other theft Car vandalism Threat Attempted burglary Bicycle theft Assault Robbery without violence Theft from car Robbery with violence Burglary Car theft Total

21.6 26.8 28.8 33.8 53.9 59.3 64.5 69.2 75.7 81.3 88.6 97.0 43.2

731


To adequately test our hypotheses, parameters for four different logistic multilevel models have been estimated (see Table 4). Crime and victim characteristics are included in all models, as are the two control variables at neighbourhood level. In each of the first three models, the indicator of one of the neighbourhood characteristics is also included. In the fourth model, all three indicators of the neighbourhood T A B L E 4 Effect parameters (log-odds ratios) of crime, individual and neighbourhood characteristics on the probability that crimes are reported (nvictims = 101,592, nneighbourhoods = 3,104) Model 1

Model 2

Model 3

Model 4

Coeff.

(S.E.)

Coeff.

(S.E.)

Coeff.

(S.E.)

Coeff.

(S.E.)

Intercept

−0.65

(0.03)**

−0.65

(0.03)**

−0.62

(0.03)**

−0.62

(0.03)**

Crime characteristics Type of crime (ref.: bicycle theft) car theft theft from car car vandalism robbery with violence robbery without violence attempted burglary burglary other theft other vandalism threat assault Financial loss (ref.: < fl.100)a fl.100 – fl.499 ≥ fl.500 Physical injury (ref.: none or little) average severe

— 2.67 0.51 −1.40 1.43 0.72 0.35 1.25 −1.01 −1.33 −0.11 0.83 — 0.97 2.21 — 1.57 1.41

— (0.20)** (0.04)** (0.02)** (0.16)** (0.04)** (0.03)** (0.05)** (0.03)** (0.03)** (0.03)** (0.08)** — (0.02)** (0.02)** — (0.16)** (0.19)**

— 2.67 0.51 −1.40 1.43 0.72 0.35 1.25 −1.01 −1.33 −0.11 0.83 — 0.97 2.21 — 1.57 1.41

— (0.20)** (0.04)** (0.02)** (0.16)** (0.04)** (0.03)** (0.05)** (0.03)** (0.03)** (0.03)** (0.08)** — (0.02)** (0.02)** — (0.16)** (0.19)**

— 2.66 0.51 −1.40 1.43 0.72 0.35 1.25 −1.01 −1.33 −0.12 0.84 — 0.97 2.21 — 1.57 1.41

— (0.20)** (0.04)** (0.02)** (0.16)** (0.04)** (0.03)** (0.05)** (0.03)** (0.03)** (0.03)** (0.08)** — (0.02)** (0.02)** — (0.16)** (0.19)**

— 2.66 0.51 −1.40 1.43 0.72 0.35 1.25 −1.01 −1.33 −0.12 0.84 — 0.97 2.21 — 1.56 1.41

— (0.20)** (0.04)** (0.02)** (0.16)** (0.04)** (0.03)** (0.05)** (0.03)** (0.03)** (0.04)** (0.08)** — (0.02)** (0.02)** — (0.16)** (0.19)**

Victim characteristics Age Male Low education level Paid job >15 hours/week Jobless or unable to work Non-native Household type (ref.: mult. adults) single person single parent (child(ren)