Abstract

This paper explores the micro-level consequences for citizens’ private security of the crime governance strategy of ‘responsibilisation’. Specifically, it analyses data from the British Crime Survey and the UK Census to identify the correlates of  the availability of Neighbourhood Watch (NW) schemes and of individual household participation in them. In general, it finds a strong social class bias affecting both availability and membership. In addition, membership  is found to depend upon the mutually-incompatible processes of anxiety about crime risk and community reciprocity. Further, a comparison of the correlates of property crime victimisation and NW availability reveals a bias towards the protection of households at risk only amongst the more affluent sectors of society. An explanation is proposed based upon the idea that  middle-income  members of society are more able than other social groups to create security ‘club goods’ for themselves, of which NW is seen as a prime example.

“The history of risk distribution shows that, like wealth, risks adhere to the class pattern, only inversely; wealth accumulates at the top, risks at the bottom. To that extent, risks seem to strengthen, not to abolish, the class society. Poverty attracts an unfortunate abundance of risks. By contrast, the wealthy (in income, power or education) can purchase safety and freedom from risk” (Beck, 1992: 35)

Burglary has been called the ‘folk crime of the new millennium’ (Mawby, 2001:15). However, while ‘domestic’ burglary is a common-enough offence of everyday life that invokes worry and anxiety, it may well be that it is an offence more characteristic of the past century than it will be of the future. As the new century unfolds, it is possible that burglary will dwindle in its incidence, at least for the better-off sections of the more affluent societies of the developed world. For national governments, this might appear to be a welcome prospect since burglary contributes voluminously to the current stock of offences that they officially record. They may even be inspired to take some credit for bringing down the crime rate by reducing burglary, even if they remain privately perplexed as to precisely how they have managed to achieve this success with the antiquated policy tools at their disposal (see Garland, 2001). Yet, as we shall suggest in this paper, if the incidence of domestic burglary does continue to decline, it will be as a consequence not so much of the renewed vigour of the state in its “crusade against crime” (Home Office 1999: 2) but more a result of the private and collective behavioural adaptations of actors in civil society. Nor will this be necessarily a happy state since, we infer, this absolute decline in traditional volume crime is likely to be accompanied by a number of other social phenomena, including (paradoxically) a greater salience of insecurity about crime and a greater social inequality - both in the risk of household property crime and in access to the goods of private security that afforded protection from it. To adapt a phrase, we may be heading towards a society of private security-affluence and public crime-squalor1. And if the state has played a role, it is as a collaborator in the growth of private security and hence, unwittingly, in its adverse social consequences.

I. ‘RESPONSIBILISATION’ AND THE SOCIAL REACTION TO CRIME

The decline of household property crime in the West?

In The Culture of Control, David Garland (2001) argues that the state’s present stance towards its citizens about crime is one of responsibilization – of shifting the burden of responsibility for crime reduction and security onto its citizens – as an adaptation in the face of its inability to alter crime rates. Citizens should do more to provide for their own private security in their everyday lives, especially by changing their routine practices and ‘lifestyles’. It is assumed that, if this happens, crime rates will diminish. Yet, this has to be done covertly since there is a consensus (to which Garland also subscribes) that both public anxiety and crime rates are continuing to rise together remorselessly. Governments cannot then afford to admit to failures or uncertainties in crime control because they believe this will not only threaten their legitimacy in the eyes of their electorate but also provoke panic in the streets.

Yet it is a recorded fact – surprisingly little remarked upon in criminology – that household property crime – particularly burglary and vehicle crime (both of recorded crimes and victimisation survey-based rates) - has been declining steadily in the United States and England and Wales. In the case of America, the decline has been continuous, long-term and substantial: with few exceptions, the USA has seen a year-on-year decline since 1973, amounting to a reduction in the rate of burglary of around 74 per cent by 20012. The long-term trend in England and Wales has been somewhat different as a whole, since burglary rates rose sharply during the late-1980s and early-1990s – “…the US burglary rate as measured in the victim survey was more than double England’s in 1981, but in 1995 the English burglary rate was nearly double America’s” (Langan and Farrington, 1998: ii.). Even so, since then, the burglary rate has dropped by about 45 percent to the end of 2002 (Povey et al., 2003).

That such a remarkable decline in major volume crime has been overlooked in public discourse is probably due to two factors: first, it is inconsistent with prevailing professional meta-narratives about crime, which predicate high crime rates as a “normal social fact” of modernity (Garland, 2001: 106). Although scholarly attention has been paid to ‘the crime drop in America’ this has been focussed exclusively on the more recent downturn in violent crime and homicide – signifying a continued American preoccupation with the nexus of race, poverty, drugs and violence (Beckett and Sasson, 2000; Zimring and Hawkins, 1997) – but even so, there is little agreement on its causes (Blumstein and Wallman, 2000). Similarly, the other prevailing theory of the growth of crime – ‘routine activity theory’ (Garland, 2000) – has merely relied upon an assumed continuation of the crime trend, and its associated dynamics of social, occupational and spatial mobility (Felson, 1998), even though US burglary rates started to turn-down the year following the endpoint of the data-series first used to substantiate the theory (Cohen and Felson,1979). No simple alternative explanation presents itself.

Second, governments themselves might want to take the credit for the property crime drop (which applies also to vehicle crime) as a result of their crime control initiatives and increasing punitiveness (Garland, 2001); unfortunately, they are denied this by a public that obstinately believes crime still to be increasing. Recent findings from the British Crime Survey suggest that although there has been a 25 per cent fall in the overall crime rate for England and Wales since 19973, in 2002/03, 73 per cent of respondents still believed that the national crime rate had risen over the previous two years – an increase of 14 percentage points since 1998 (Simmons and Dodd, 2003, p. 128); and, though somewhat more optimistic about their personal circumstances, 53 per cent still believed crime to have increased in their local area (ibid, p. 129), and 22 per cent thought it likely that their home would be burgled – despite a survey-estimated risk rate of 3.4 per cent (ibid, 2003, p.131). Finally, although the proportion of respondents who were ‘very worried’ about burglary fell from 26 percent in 1994 to 15 per cent (ibid, 2003, p.135), 50 per cent of respondents were still ‘worried’ about burglary – the crime most worried-about by the population at large (ibid, p. 133). Clearly, there is a mismatch of perception here: government believes it has been doing good, the public believes otherwise; government tries ever harder to convey its optimistic news to the public, the public persists in ‘exaggerating’ its risk. Yet part of the difficulty in explaining this apparent paradox lies, not in the correctness or otherwise of the ‘evidence’, or of the difficulty of getting this message across to a frightened yet ignorant public, but in the shared belief that it is the actions of the state alone that are responsible for changes in the level of crime in society.

Alternatively, we argue that ‘anti-crime activity’ is not exclusively a property or function of the state, undertaken for the benefit of society as a whole – that is, a public good – but also an activity of private citizens – that is, a private good - in which they seek to engage for their own purposes alongside any greater social benefit that might be had from them doing so (Hope, 2000). Even though the permitted scope and framework of their actions is set by the state, within those limits there is substantial scope for the exercise of choice in the amount and type of anti-crime activity in which citizens engage voluntarily. The main form of private citizen anti-crime activity is in acquiring private security, that is, (relative) freedom from the threat of crime in the conduct of their everyday lives. The direct interests of most citizens in acquiring private security lie first and foremost in self-interest, that is, in protecting their own persons, families and property.

Theories, such as they are, about changes in crime rates fail to acknowledge the role of private citizens in civil society. An exception is the simple but compelling hypothesis of van Dijk (1994), who proposes an ‘equilibrium model’ – that as crime rates rise, so citizens take private actions to avoid crime risk, for example, by not visiting city centres in their cars or moving to a safer neighbourhood, which reduces the proximity of potential victims to potential offenders, thus limiting the opportunities for the latter to commit crime against the former. As a result of the avoidance behaviour of those who have the means to remove themselves from risk, following an increase, crime rates will reduce to a new equilibrium level. Developing this argument, we might infer two additional attributes: first, that a key driver in this process is private security consumption, which is stimulated and sustained by the threat of crime, real or imagined, in the minds of private citizens. Rather than see the fear of crime as necessarily destructive (although it may be for the public sphere) ‘fear’ may well also play a ‘productive’ role, functioning as a stimulus for private action and accommodation to risk 4. Second, if avoidance is crucial, then not all citizens have equal access to the means of crime avoidance; and if, as we shall argue, these means – such as purchasing home security devices, joining-in with neighbours, moving to a ‘better’ neighbourhood, or even calling the police – depend upon access to economic and related forms of social capital – then, as crime rates reduce to a new social equilibrium, they are likely to produce or reinforce inequalities in risk and risk-avoidance, correlated closely with inequalities in the distribution of income and wealth.

Responsibilization and collective citizen action against crime

The strategy of responsibilisation (Garland, 2001) – which has been the driving force behind governmental-sponsored efforts at ‘community crime prevention’ over the past thirty years (Hope, 1995) – has almost universally foundered upon a key issue – the implant problem (Hope 1997;1995). Voluntary, collective anti-crime activities are least likely in low-income, heterogeneous, deteriorated, renting, high-turnover, and high crime neighbourhoods (Hope 1995). Mutual suspicion and lack of trust amongst neighbours undermines community anti-crime efforts (Walklate & Evans 1999), especially ‘spontaneous’ community efforts to mobilise against crime. Thus, only where conditions of mutuality and trust already exist – for instance, in the form of pre-existing community groups and organisations – is it likely that the necessary ingredients of communality can be mobilised into co-ordinated action against crime problems (Hancock 2001; Skogan 1988). Yet the fundamental paradox facing governmental efforts at ‘responsibilisation’ is that while mutuality may be a necessary ingredient it is not in itself sufficient to generate collective action against crime. As we argue below, this capacity depends upon the availability of a combination of different kinds of social capital . In Britain, the consequences of this problem can be seen in the social distribution of "Neighbourhood Watch" members and organisations.

"Neighbourhood Watch" (NW) is popularly regarded as an informal grouping of households who agree to watch over each others' properties, and adjacent public spaces, and to become acquainted with each others' habits and activity patterns so that, should they witness any 'suspicious' persons or activity in the vicinity, they will be ready and willing to take action, usually by reporting such persons and activities to the police. Since the 1980s, NW has been a major feature of government crime prevention activity in Britain, and police forces have complied in promoting and servicing local Neighbourhood Watch schemes. Substantial government publicity was devoted to promoting the idea of Neighbourhood Watch - including national media advertising campaigns, booklets offering advice and guidance, and special events and conferences. As NW has evolved, it has come to be seen as a convenient agent for liaison between the police and the public about crime prevention and as a means of ‘delivering’ crime prevention advice and assistance to private households (Laycock and Tilley, 1995). During the 1990s, the British Government appeared to have successfully persuaded non-governmental bodies to take responsibility for NW: many insurance companies offered discounts or favourable premiums on home contents insurance for householders who belonged to a NW scheme registered with the police; and one of the largest British insurers provided substantial financial support to establish the National Neighbourhood Watch Association (NNWA) to disseminate and support NW activities around the country5.

Nevertheless, periodic analyses of British Crime Survey (BCS) data on the pattern and growth of NW show a consistent bias in its adoption to the more affluent suburban (and hence lower crime) areas of England and Wales (Hope, 2001). Prior to its wide-scale introduction, favourable support for the idea of NW appeared highest amongst households living in relatively socially-cohesive areas, and amongst couple-led households, owner occupiers, and those not in the lowest income band (Hope, 1988). This pattern of membership has continued with the establishment of actual NW schemes, with a general bias towards the middle- to better-off sectors and communities (Mayhew and Dowds, 1994; Sims, 2001). Figure 1 charts the growth in NW membership in England and Wales. The biggest growth occurred during the late-1980s reaching a peak of 23 per cent of households by 1992 and then growing more slowly to 27 per cent in the 2000 BCS. Recorded household property crime reached its peak level also in 1992, following a period of sharp increases, and has declined since (Simmons and Dodd, 2003). Although probably a coincidence, it is tempting to speculate that the two trends may have some relation since NW participation may encourage greater reporting of incidents to the police and that its effect is likely to be highest at the outset when members are still full of enthusiasm (Hope, 1995).

Figure 1. Interpolated trend.

Sources: British Crime Survey. Hope (1988); Dowds and Mayhew (1994); Sims (2001).

Social capital, trust relations and Neighbourhood Watch

Why might we expect community anti-crime groups to flourish in middle-class areas? In the first place, this may reflect simply an institutional bias on the part of the police - the chief agency in Britain responsible for initiating and servicing schemes. In as much as police culture may be biased towards upholding perceived 'respectable society' and defending its property interests (Reiner, 2000), we might expect their main effort to go into supporting schemes directed towards ‘deserving’ households - those who perceive crime to be an external threat to their property interests and on whose behalf state power should be invested. Indeed, Laycock and Tilley (1994) suggest that the main aims of NW in low-crime areas should be to keep crime rates low, maintain confidence in the police and "symbolically to recognise a collective commitment to a set of standards (p. 13)". Nevertheless, to accept this as the sole reason for the class-bias in the distribution of NW may be to bestow upon the police a purpose of intent and organisational capacity for social engineering (see McConville and Shepherd, 1992) that is inconceivable that they could possess in practice. Rather than assume that the police have somehow sought out middle-class areas deliberately, equally it may be that it is the middle-class areas that have offered themselves to the police - thereby providing an amenable location in which to implement the NW concept and thus helping the police to demonstrate publicly their apparent competence in crime prevention.

A primary ingredient for the establishment of voluntary community groups such as NW would seem to be a sense of mutual reciprocity amongst neighbours. Prospective support for NW in Britain before its large-scale introduction was highest where people perceived their neighbourhood to be characterised by neighbourly reciprocity 6, who reported friendship with their neighbours, and who already made informal arrangements to watch each others' homes (Hope, 1988). If collective action such as NW depends upon reciprocity, then reciprocity in turn depends upon trust amongst neighbours. Thus, that which generates trust amongst community members – what in recent years has been called social capital – may be one of the sufficient ingredients for the development of neighbourhood watch groups.

‘Social’ capital may be generated as a product of reciprocal relations amongst members of a social network (Coleman, 1990). To the extent that members are tied together in relations of mutuality and reciprocal obligation, they are able to generate the necessary 'capital' which will enable members to call upon each other to participate in collective activity and to overcome the free-riding and individualism which otherwise undermine voluntary activity. Groups who are able to generate sufficient social capital may not only have the capacity to control crime informally - by being able to generate and enforce social norms - but also to create effective and sustainable social institutions in support of those norms (Coleman, 1990).

Social groups may generate social capital where there is 'closure' within the network of members - that is, where ties and expectations are reciprocated through mutual interests and common obligations (Coleman, 1990). One of the main ways in which expectations of mutuality are reinforced is through the trust which is shared between members of the network. In the absence of close, personal ties in modern communities, the primary indicator of trustworthiness may be the perception of others' social similarity (homogeneity). In as much as the more affluent are able to live in 'communities of choice' while the poor cannot avoid living in 'communities of fate', the former can be more confident than the latter that their fellow residents will be "just like themselves" - especially not potentially victimising (Loader et al., 2000). Reciprocity can be taken on trust; and social capital can develop, enabling the community to organise itself against crime, if needs be.

As Susanne Karstedt has argued, the types of social capital that produce community cohesion in modern societies should not be confused with the more traditional strong bonds embedded in families and closely knitted groups (Karstedt, 2003; im Druck). Modern civil society needs a particular distribution of different types of bonds and a stock of ‘weak ties’ (Granovetter 1973) that link individuals and groups to institutions, that build bridges between groups, and that provide generalised trust and tolerance between members. Bonding social capital is the type produced within families and small groups; while Bridging social capital is embedded in the links and relationships between different groups, for example, between different ethnic groups or different classes and status groups often sharing the same social space. Importantly, however, Linking social capital is formed through the relationships that exist between individuals (or groups) and the institutions and organisations of the wider society. Linking social capital not only defines specific, legal types of rights and obligations that link people to organisations, but also their trust in such institutions. Individuals participate in the social capital of their groups and build a stock of their own that integrates them into the social fabric and defines their identities (Karstedt 2003, im Druck). Linking social capital thus embodies the important ‘vertical dimension’ of communities (Hope, 1995) that links them to the wider arena of resources and power in society, especially their capacity to harness public resources.

In particular, it is likely that the more middle class areas have more of the ‘bridging’ type of social capital that allows such associations to develop since they are more likely to be characterised by ‘weak ties’ amongst residents – the kind of social order that Baumgartner (1988) typifies as moral minimalism. Yet the ‘weakness’ of such social ties is also its strength (Granovetter, 1973) in that it encourages the growth of bridging social capital (Putnam, 2000) which may be required as the right social conditions for spreading community-wide organisations (Hope, 1995, p. 70). Conversely, communities lacking in such capacities may also be unable either to generate NW schemes or to enforce collective norms. Despite neighbourliness – that created by bonding social capital - poorer communities may be lacking in ‘neighbourhoodism’ (Hope, 1997) – that created by bridging social capital.

Yet there are also two qualifications to the capacity of bridging social capital to be transformed into an entity like Neighbourhood Watch. First, while lack of income and status may not of themselves inhibit the formation of social networks and the generation of social capital, they may determine, for individuals and groups, access to external, supra-communal sources of power and authority – that is, ‘linking’ social capital. This 'vertical' dimension of community relations acts as a conduit for bringing additional organisational resources into the community which – by virtue of its demonstrable collective efficacy - in turn strengthens the horizontal dimension of intra-community relations, thereby increasing network closure and generating even more social capital (Hope, 1995). To the extent that higher-status groups may be able to generate additional community organisational efficacy by combining bridging and linking social capital, this in turn can be used to generate ‘symbolic capital’ with external agencies of authority, for example, through pressures – of a persuasive or moral nature - from residents associations and NW itself brought to bear upon local state agencies including the public police. Thus economically-advantaged groups may not only be able to trade their economic capital directly into symbolic capital so as to influence the distribution of protective resources – that is, delivered via a symbiotic institutional bias on the part of the police - but can also capitalise upon their social capital to form groups to lobby for more resources. Thus they may be doubly-advantaged in acquiring protection from the public police (Skogan, 1988).

Second, worry about crime risk may also mediate the establishment of NW schemes. Voluntary activity entails costs as well as benefits. In low-risk areas, or for groups who are not anxious about crime, there may be little benefit from investing personal effort in organising and networking anti-crime groups (Skogan, 1990). Even if such groups generate social capital, there may be little need to direct it towards anti-crime defences. In contrast, anxiety about crime can undermine trust in one's neighbours and thus may itself become another risk to participation in voluntary activity. Here though, we encounter a central paradox in thinking about the community response to crime (Hope, 1995): on the one hand, there is the Durkheimian tradition, which sees the communal reaction to crime as essentially integrative; community solidarity will increase in response to an external moral threat. On the other hand, there is the view that crime is disintegrative: residents of high crime areas may be "deeply suspicious of one another, report only a weak sense of community, have low levels of personal influence on neighborhood events...and feel that it is their neighbors whom they must watch carefully" (Skogan 1988: 45). It may not be then the direct experience of crime victimisation itself, which mediates communities' responses towards anti-crime organisation. Not only may the anxiety which comes to be attached to the perceived threat of crime have sources additional to crime risk itself (Jefferson and Hollway, 2000), but interest, commitment and satisfaction with the neighbourhood may also shape both anxiety about crime and the form of response towards it, notwithstanding the threat of victimisation itself (Taub et al., 1984). Where either use-values or exchange values are high, residents may feel more inclined to support preservationist activities to defend them (Skogan, 1988).

The social conditions for community anti-crime groups

In sum, households’ participation in community anti-crime groups – taken here to be in the form of ‘Neighbourhood Watch’ (NW) - may be shaped by a coalescence of a number of (sometimes incommensurate) social conditions: first, the threat of crime (shaped through direct experience or mediated by generalised community concerns) – though this need not to be so low as to remove the incentive to join, nor so high as to undermine trust in one’s neighbours (Hope, 1988). Second, the availability of bridging social capital, generated from weak but extensive social ties amongst neighbours, that mediate the introversion of the bonding social capital formed by close-knit family and friendship groups and allow wider, more efficacious, organisational capacity to develop (Granovetter, 1973). Such social ties need not to be so strong as to inhibit relations forming beyond friendship cliques, nor so weak as to inhibit trust relations forming amongst strangers. Third, the availability of linking social capital, which provides individuals and groups with a sense of trust in wider institutions and a sense of efficacy that their collective efforts will be reciprocated. Although it is difficult to measure social capital directly, we hypothesise further that bridging and linking social capital will be relatively unavailable in economically marginal and socially excluded communities and, in the absence of prior social organisation to implant these forms of capital, anti-crime organisations like Neighbourhood Watch will not arise spontaneously in such communities. Rather it is more likely to be the more affluent, socially-included communities, where bridging and linking forms of social capital may be more abundant, that spawn the most NW groups.

Fourth, however, members will not be able to join unless they are given the opportunity to participate. This is likely not so much to be simply an objective measure of availability – that is, what the police or the state say is being made available to the community – but also what the public perceive to available, which depends both on personal knowledge that NW is available to them and an understanding that they are being invited or permitted to join. In this sense, perceived availability, willingness to participate and actual membership are not independent of each other in the minds of community members and inferences about them may be confounded unless a specific effort is made to disentangle their effects when examining the social distribution of NW membership. Finally, however, as NW groups are simultaneously voluntary but also imply a relationship with the public police, police support for schemes – and hence their actual availability - may also be conditional: the police are unlikely to initiate or service NW schemes in communities that they (the police) perceive cannot generate informal associations; NW groups are unlikely to form or be sustained without police encouragement and support.

Consequently, the social distribution of NW may be the result of the interplay between these often contradictory conditions. In particular, three sets of relations between these conditions are likely to affect an analysis of the social distribution of NW: first, that between NW availability and membership (both actual and perceptual) – where membership is dependent upon availability; second, between the threat of crime (again both actual and perceived) and social reciprocity – where the latter may be related negatively to the former; and third, that between social capital and social class – where there may be a positive relationship between (higher) social class and the forms of social capital necessary to sustain NW. An analysis of citizen participation in NW that fails explicitly to account for these relations amongst the set of explanatory variables may be confounded by them. The next section investigates these issues using data from the British Crime Survey (BCS).

II. NEIGHBOURHOOD WATCH PARTICIPATION – AN EMPIRICAL EXAMINATION

Data, Variables and Method

Data

Data are taken from the 1992 British Crime Survey for England and Wales (BCS), coinciding with the peak of NW membership growth (Figure 1). The BCS is a sample of individual respondents (one person aged 16 years or more per household), with each respondent answering questions about themselves and on behalf of their households. The analysis presented in this paper is thus an analysis at the level of individual respondents. Nevertheless, we also take advantage of the BCS multistage cluster sampling design which allowed us to attach data from the 1991 National Census to represent the social characteristics of the area in which each respondent resides 7. These comprise a set of area contextual variables which were then attached to each respondent’s case-record. The multivariate statistical models estimated below – logistic regression models – are single equation models at the individual-level, which include both individual and area-contextual variables 8.

Modelling NW participation

The principal aim of the analysis was to estimate a multivariate model of the social distribution of NW participation. This would be difficult to estimate reliably in a single model because, as noted above, there are interdependencies amongst the relevant explanatory variables that need to be taken into account so that the findings are not confounded by the relationships amongst the relevant explanatory variables (discussed above). There are also ‘sample selection’ processes that influence the definition of the sub-sample of NW members; for instance, the BCS question which directly reports household membership of NW is itself censored as a result of the (logical) ordering of questions about NW in the BCS self-report questionnaire. Within the structure of the BCS interview procedure, household membership of NW is defined in three successive stages:

1. Knowledge of NW - whether the respondent knows about NW. Of the full sample of 11,713 respondents, 92% (n = 10,776) had heard of NW (the 8 % who had not were excluded from further analysis). Of those who had heard of the NW concept, 89% said they knew whether or not a scheme existed in their area; the latter constituting the sub-sample of those who had knowledge of NW in their area (n = 9,560). 9

2. Perceived NW Availability - whether a NW scheme is believed to be available in the area. Of those who said they had local knowledge of NW (n = 9,560), we excluded cases where there were incomplete responses to one or more of the key variables (n = 1,333) leaving an effective sample of 8,227. Of these cases, 38 per cent said that they believed there was an NW scheme available in the area which covered the household. Our first estimated model - of NW availability - thus estimated the likelihood of the proportion of the NW Knowledge sub-sample (n = 8,227) who said that NW was available (0.38, n = 3,126) as compared to those who said it was not (0.62, n = 5,100).

3. NW Membership - whether the respondent said the household was a member of NW. Of those who said that NW was available (n = 3,126), 72% said that their household belonged to a NW scheme. Our second estimated model – of NW membership – consequently estimated the likelihood of the proportion of the NW Availability sub-sample (n = 3,126) who said that their household was a member (0.72, n = 2,251) compared to those who said they were not (0.28, n = 875).

Using the appropriate sample, then, separate logistic regression models were estimated for perceived NW Availability (base n. = 8,227) and NW Membership (base n. = 3,126).

Explanatory variables

The Appendix lists the ‘explanatory’ variables that were used in the estimation of the models. Broadly, three types of variable were used: firstly, a set of ‘conceptual’ variables, which sought to operationalise the concepts discussed in Part I of this paper. These were:

1. Worry about burglary. – The following question was used “could you tell me how worried you are about having your home broken into and something stolen?” For this analysis, Worry about burglary was coded as separate binary variables: NVWBURG (‘not very worried’ = 1, other = 0); FWBURG (‘fairly worried’ = 1, other = 0); and VWBURG (‘very worried’ = 1, other = 0). ‘Not at all worried’ formed the base (aliased) category. This variable was selected in preference to a more generalised ‘fear of crime’ question since it related specifically to the kind of crime that NW is intended to counter.

2. Community Reciprocity – This was measured from the question "In some neighbourhoods people do things together and try to help each other, while in other areas people mostly go their own way. In general, what kind of neighbourhood would you say you live in?" The following binary coded variables were derived: GOODSOCO (‘people help each other’, yes = 1, other = 0); and MIXSOCO (‘a mixture’ = 1, other = 0). ‘Go their own way’ formed the base response category. This is the closest approximation to the perceptions of ‘social capital’ available in the 1992 BCS and does not distinguish between bonding and bridging forms. However, some further elaboration of the types of social capital can be had when looking at this variable together with the socio-economic explanatory variables also in the models (see below).

3. Area Satisfaction – was measured from the question “Overall, how satisfied or dissatisfied are you with living in this area?”. The following variables were derived from the coded responses: VERYSATA (‘very satisfied’ = 1, other = 0), FAIRSATA (‘fairly satisfied’ = 1, 0 = other). The responses, ‘Fairly’ and ‘Very’ dissatisfied formed the base response category.

4. Household property crime victimisation – whether respondents reported that their household had experienced one or more incidents (including attempts) of burglary, theft from dwelling or criminal damage in the previous twelve-months. These offences were combined to include a comprehensive measure of the constellation of ‘domestic’ property offences of which burglary itself is the archetype.

The second type of explanatory variables used relate to individuals and households and sought to capture the main social and demographic characteristics at the 'individual' level (i.e. of respondents and their households). A third type of variable relates to the area-context in which the household was located, seeking to capture aspects of community status and demographic structure10. Some of the set of these census-derived variables exhibited multicollinearity. Guided by the results of a principal components analysis three portmanteau variables were constructed, which we have labelled DEPRIVED, AFFLUENT, and YOUNG 11. We also included a set of indicators (binary coded) referencing the Standard Regions of England and Wales. These regional indicators have been found to be significant independent predictors of household and area property crime risk (Ellingworth et al., 1997) and may also reflect, in a crude way, regional variation in the dissemination and implementation of the NW concept12.

Additional Models

As noted in Part I, the set of ‘conceptual’ variables used to model perceived NW availability and membership are also likely themselves to be inter-related. Moreover, some indication of the ‘system’ of relationships amongst them is also likely to illuminate further understanding of the various influences on NW participation. Unfortunately, jointly estimating a system of equations expressing relationships amongst the conceptual variables is likely to be complicated in the current situation not only by the recursive or reciprocal relationships amongst the variables but also by the varying sample sizes (between the NW Availability and NW Membership models) and by the logistic regression method used. In preference, a series of separate models were estimated (in addition to the NW models), using the same set of explanatory variables listed in the Appendix. The significance and direction of the coefficients of the explanatory variables were then used as graphical indicators of relationships, which were then schematised into a hypothetical model (see Figure 2 below) and which does not represent a single but a number of models. Logistic regression models were estimated for each of the following dependent variables in turn:

1. Worry about burglary: coded as 1 = (VWBURG + FWBURG), 0 = otherwise;

2. Community reciprocity: coded as 1 = GOODSOCO, 0 = otherwise;

3. Area satisfaction: coded as 1 = (VERYSATA + FAIRSATA), 0 = otherwise

4. Household property crime victimisation: coded as 1 = victimised over previous 12 months; 0 = otherwise.

To ease comparability, each of the above models was estimated using the same sub-sample as that used to estimate perceived NW Availability (n = 8,227).

Modelling strategy

Logistic regression models, using the appropriate sample, were estimated separately for each of the principal dependent variables - that is, perceived NW availability, NW membership – and for each of the additional ‘conceptual variables’ treated here as dependent variables - worry about burglary, community reciprocity, area satisfaction and property crime victimisation. For each model, in turn, our strategy was to enter initially all the explanatory listed in the Appendix13. Our procedure for producing trimmed, parsimonious models was simply to delete non-significant parameters though retaining categorical variables if at least one level of the variable was significant. Thus, although starting off with the same set of explanatory variables for each logistic regression model (i.e. all those listed in the Appendix), the final models differ in their respective sets of variables retained. By comparing the results from each model, we are able to gain some indication, albeit non-quantified, of the inter-relationships amongst the set of ‘conceptual variables’ these are illustrated in Figure 2 below. Every model, however, included initially the same set of individual- and area-level explanatory variables which provided a common ‘baseline’ for comparing the effects of the conceptual variables across each model (see Table 1, below). In the interests of brevity, we do not generally report all the coefficients of each model14.

Results

Reciprocity and worry. The pattern of relationships amongst the set of key variables is illustrated in Figure 2, and the relevant coefficients from the separate models listed in Table 1 15;.

Figure 2. Influences on Neighbourhood Watch: Some relationships amongst the key variables

TABLE 1 INFLUENCES ON NEIGHBOURHOOD WATCH: SOME RELATIONSHIPS AMONGST THE KEY VARIABLES (Selected results from logistic regression models)

1. As explained in the text, when serving as dependent variables the variables ‘Worry about Burglary’, ‘community Reciprocity’ and ‘Area Satisfaction’ were binary coded

2. Each model also includes significant demographic and social structural variables (not shown), listed in the Appendix. Non-significant variables were deleted from final models (see text).

3. Worry about Burglary was not included in the model of property crime victimisation because this question is asked at the end of the reference period for victimisation reported in the survey. If victimisation had taken place, the ‘causal direction’ would be from victimisation to worry not vice versa.

Consistent with the idea that social capital is a necessary ingredient for NW, the perceived availability of NW is influenced by the extent of perceived community reciprocity. Similarly, in areas where schemes are available according to the respondents, members of schemes are also more likely than non-members to think that local people ‘help each other’ even though NW is perceived as available by both. And, it should be noted, these findings hold irrespective of the social characteristics of households and areas. In particular, it is areas where residents are clear that most people "go their own way" which had fewest schemes according to perceptions of residents, and fewer members where schemes were seen as available. Membership also seems to increase with the level of perceived reciprocity. Although those who thought their area had a ‘mixture’ of people who "helped each other" or "went their own way" (MIXSOCO) were 35 per cent more likely to be members than households who thought people ‘mostly went their own way’, those who believed that local people mostly helped each other (GOODSOCO) were 71 per cent more likely to be members.

Yet although a community's reservoir of social capital may be a necessary ingredient for establishing NW schemes and encouraging membership, it would seem that it is specifically worry about victimisation which distinguishes participants from others. Even though schemes were no more likely to be perceived as available by people who worry about burglary victimisation than by others (which shows that perceptions of availability are independent of worries about crime), households who become members of NW when they have the opportunity are significantly more likely to be worried than are those who do not participate; progressively so: those saying they were "not very worried" (NVWBURG) were 56% more likely to be members than those "not at all worried"; the "fairly worried" (FWBURG) were 78% more likely to be members; and the "very worried" (VWBURG) were twice as likely to be members.

Thus, both a high degree of community reciprocity and a higher degree of worry about victimisation seem necessary conditions for maximising NW membership. Yet these are themselves negatively correlated (Figure 2) - households who see their communities as cohesive are less likely than others to worry about crime victimisation. Consequently, there may be limits to NW participation, for it may be only 'exceptional' households - that is, those who think their community is cohesive but who nevertheless worry about crime risk - who are most likely to become members (Hope, 1988).

Comparison between the various models, assisted by reference to Figure 2, reveals different ways in which this paradox emerges. In the first place, the experience of crime victimisation, not surprisingly, increases worry about the risk of crime and reduces peoples' sense of community reciprocity; together, these heighten dissatisfaction with the neighbourhood. Neighbourhood dissatisfaction, in its turn, reduces community cohesion and increases worry about crime. In particular, having been a recent victim of household property crime victimisation had no effect on NW membership, nor vice versa 16. Part of the reason for this may be because although property crime victimisation is positively related to worry about crime - a necessary ingredient of NW participation - it is also negatively related to community reciprocity - which is also a necessary ingredient of participation. In other words, the two conditions of participation together combine to cancel out the direct relationship between crime victimisation and NW membership, irrespective of the direction of the relationship between them.

To establish that a greater worry about burglary victimisation differentiates members from non-members using cross-sectional data does not, of course, establish the direction of causality. Indeed, we cannot tell from these data whether worry about crime heightens people's need to seek reassurance by joining NW, or whether participation in NW itself actually heightens people's sensitivities towards the risk of crime. Probably both processes operate in what Rosenbaum (1988) calls the "fear arousal" hypothesis – that people are motivated by anxiety to overcome the costs of participation and join anti-crime groups but that the ‘talking-up of crime’ in such groups also serves to heighten awareness and thus anxiety about crime (see also Sasson, 1995) 17.

Victimisation risk and perceived NW availability. Household property crime victims were no more likely than non-victims to perceive NW schemes available in their areas; nor were NW members less likely to be victims than non-members (Table 1). Yet, regardless of any causal connection between the two18, it is important to ask whether the social distribution of NW is aligned with the social distribution of household property crime victimisation. For the purposes of this analysis, what we mean by ‘alignment’ is whether in each separate model – Perceived NW Availability and Crime Victimisation – the respective coefficients for the same variable are similar in significance, direction (sign) and approximate order of magnitude; if not, then the two distributions on that variable can be thought to be out-of-alignment. A comparison between the models for victimisation and perceived NW availability may then reveal the extent to which the social influences (mainly upon the police) in making NW available (and communicate this appropriately to residents) are biased against those who have the greatest crime risk. As such, if NW is not targeted where households are most at risk, it is unlikely to have much of an effect on crime reduction overall, even if it nevertheless reinforces the security of those less at risk, thus helping to stop crime spreading to them, thereby containing growth in the crime rate further. The estimated models of property crime risk and perceived NW availability are listed in Table 2.

TABLE 2 ALIGNMENT OF THE LIKELIHOODS OF PROPERTY CRIME VICTIMISATION AND AVAILABILITY OF NEIGHBOURHOOD WATCH (LOGISTIC REGRESSION MODELS)

Taking those variables where victimisation-risk and perceived NW availability are in alignment19 reveals a bias towards responding to the needs of the better-off ‘at risk’ population. Higher perceived availability of NW is aligned with a greater risk of household property crime victimisation for households where: the head of household was in a non-manual occupation, children were of primary school age (5-11), and residence was in a detached house and/or an affluent area20. Taken together, this would seem to conform to the social distribution of the ‘comfortable’, propertied middle-classes. Those households where perceived NW availability was not in alignment with victimisation risk – that is, where risk was equal to or greater than average but where perceived NW availability was less than average 21 - included those who rented their homes (from either private or social landlords) and/or lived in areas with a high proportion of rental accommodation22. Amongst those who were more likely to be victimised, those who had an unclassified (probably casual) type of job, had moved house recently, and lived in areas that had a greater than average proportion of: children aged 5-15 years and/or lone-parent households, were no more likely than average to believe that NW was available to them 23. Conversely, then, these households whose risk was not being addressed by NW being made available (or who at least reported that there was no NW available in the neighbourhood) conform to the social distribution of the poor and economically-marginal.

There were also other apparent biases in the availability of NW to specific social groups. First, towards older people, where the police appeared to have made NW available to a greater than average extent above the level of victimisation risk: reported availability increased with the age of the respondent and with living in a community where the proportion of elderly single households was higher. Second, both main ethnic minority groups – Afro-Caribbean/Black British and Indian Sub-Continent/Asian British - had a similar average risk of crime victimisation to whites but perceived significantly fewer than average NW schemes available to them24.

Third, while households living in deprived (DEPRIVED), youthful (YOUNG) and inner city (INCITY) areas, faced ‘average risks’ of property crime, they perceived relatively more NW schemes available to them. At first glance, these results may seem counter-intuitive since it might be expected that these types of areas, by virtue of their urban nature, would have higher crime risks and lower reservoirs of social capital. Nevertheless, it should be noted that in these models the area-related effects are estimated having controlled for household-related effects. Complementary analysis of this data reveals that, generally, while household-related ‘risk factors’ of property crime victimisation (controlling for area-related variables) work in the direction of affluence, area-related risks (controlling for household-related variables) point in the direction of deprivation. The resulting social distribution of risk – and its avoidance – depends, then, upon the ‘ecological’ distribution of the individual rich and poor allocated, respectively, to rich and poor areas, creating varying socially stratified ‘risk pools’ (Beck, 1992) of victimisation and safety (Hope, 2001). In the present situation, then, we may infer that NW is more likely to be available (and to be perceived as such) to the better-off in poorer, inner city areas, than it is to the poorer residents of those areas (see Hope, 2000).

Finally, if significant regional differences (having accounted for individual household characteristics and area variation) in perceived NW availability can be taken as a crude indicator of variations in the implementation policy of the constituent police forces of each region, then there was not a single region where perceived NW availability was in alignment with regional household property crime risk – in 1992, some regions had a greater level of risk than of NW availability, while other regions had the opposite.

III. DISCUSSION: THE ‘CLUBBING’ OF PRIVATE SECURITY

Over the past twenty years or so, by design or default, governments have pursued a strategy of ‘responsibilisation’ in seeking to shift the burden of crime reduction and community safety onto private citizens and the institutions of civil society (Garland, 2001). Between the mid-1980s and the mid-1990s, Conservative governments in England and Wales encouraged and enabled the growth of resident-based, community groups - ‘Neighbourhood Watch’ (NW) - as a means of providing security goods through collective self-help. This approach was consistent with their prevailing notion of support for the ‘active citizen’. More or less willingly, the public police were persuaded to serve as ‘agents’ for the development and support of NW (Laycock and Tilley, 1994). The analysis presented in this paper examines the social pattern of development of NW in England and Wales at the point at which – circa 1992 - its growth rate had peaked (Figure 1).

The results from the models presented here offer support for the two broad hypotheses of this paper: first, that the state’s ‘responsibilisation’ strategy, c. 1985-1995, required not only fertile social conditions – in the form of social capital based upon a sense of community reciprocity – but also a strong and ever-present threat of crime in the minds of citizens. In this respect, both victimisation experiences and dissatisfaction with one’s neighbourhood fuel worry about crime which, if the social ‘soil’ is right, and if the level of resulting anxiety has not become too debilitating so as to undermine reciprocity, can be useful in galvanising people into ‘doing something’ (albeit in a relatively mild way) about their own risk of crime in their residential locality.

Given these conditions, NW may then go on to serve the useful purpose for the state of responsibilisation, becoming a conduit for delivering and implementing ‘private security’ into the community. A study of the pattern of self-reported, citizen crime prevention activity in the community – using data from the 1994 British Crime Survey – found a distinct cluster of crime prevention activities associated with NW membership, including, security-marking of household property, asking neighbours to watch the home when away, and having household-contents insurance – activities that reinforce private (target-hardening) security with collective and institutional reassurance (Hope and Lab, 2001). These are the ‘club-goods’ associated with NW membership - public police encourage residents to form NW groups, disseminate home security advice and security-consciousness, and introduce them to the idea of marking their property to deter theft and/or aid its recovery; the groups introduce neighbours to each other and encourage them to share their routine activities; and insurance underwriters, accepting their lower risk and confirming their security practices, offer discounted premiums to NW members (Laycock and Tilley, 1995).

While such measures individually may have limited value in reducing risk (Hope, 1995), they provide, nevertheless, a collection of social assurances about risk, presumably addressing the particular needs of property-owners, who may be especially risk-averse (Field and Hope, 1990). Crucially, these assurances, deriving from membership in civil associations and institutions, complement and enhance other private, individualised security actions of a physical, target-hardening nature25 but are only available to those who, in addition to financial capital, also have access to the social capital needed to gain membership. Indeed, the main feature that distinguished participation in the NW cluster of prevention activities from other clusters of physical-based, property prevention activities (all of which were most common amongst the propertied middle-classes), was membership of other community-based groups and associations.

Second, the pattern of the prevalence of NW schemes as reported by respondents in the 1992 British Crime Survey (BCS) supports the notion that there is an ‘elective affinity’ between certain communities’ adoption of NW and the support, sustenance and legitimacy provided to nascent community groups by the public police. This institutional bias revealed itself during a period of rapid growth not only in NW membership but also in recorded crime. Consequently, this was not a bias towards low-risk (affluent) communities per se – police forces could hardly be seen to be putting in effort where there was little crime. However, the pattern of alignment between perceived NW availability and victimisation risk (Table 2) reveals rather a class-cultural bias towards the better-off sections of those at risk rather than to poor, economically-marginal and ethnic minority communities who were equally, if not more, at risk.

As with the problem of institutional bias on the part of the police more generally (Reiner, 2000), it is difficult to ascertain whether these biases result from a socially self-selecting response on the part of the community to invitations from the police to form NW groups, or from biased, selective invitations from the police to the community. Whatever the reasons, a biased pattern in perceived NW availability and membership does seem to have been produced through a symbiosis between, on the one hand, community support and capacity, and, on the other hand, a likely police affinity with such groups and communities. Both the power and limitations of the NW concept are revealed by its capacity to integrate certain selected groups of citizens and the public police into a common cultural understanding that addresses material concerns about residence, property and the risks attaching to them. However, that cultural understanding is, by definition, predicated upon the exclusion of others.

Conclusion: the ‘clubbing’ of private security26

Much of the literature on the emergence and distribution of anti-crime community organisations in the United States during the 1970s and 1980s - the period during which residential property crime rates began to decline - displayed a strong tendency for such groups to be initiated, and to survive, as an adjunct of existing organised community activity, particularly that oriented to the ‘preservation’ of suburban community values (Skogan, 1988). Likewise, anti-crime group activity may fail to become implanted in communities where general organised social activities are low (Hope, 1995). Further, as Lynn Hancock (2001) shows in her study of community activist groups in Merseyside, crime reduction issues are likely to be addressed, if at all, in ways which are consistent with these groups’ wider agenda and purposes – that is, in the politics of urban regeneration - which may not necessarily coincide with the state’s specific crime control agenda. In similar vein, Melossi and Selmini (2000) describe the emergence of comitati di cittadini (Citizens’ Committees) in the older working-class districts of the cities of Emilia-Romagna, which has contributed to an unfortunate coupling of the issues of migration and crime, both perceived as a threat to that civic order their members feel they created through their past political activism. In sum, local collective anti-crime activity is embedded in particular community values - not only is it likely to need group activity to develop but it will also reflect the particular concerns of the activist groups.

The web of group activity not only indicates the availability of social capital in the community but also its capacity to spawn further groups as needed. Their success in sustaining themselves - thus countering free-riding – is a result of continuing access to the benefits of belonging to the ‘security club’, that depends not simply on residing in a conducive area of like-minded people but is also reinforced by the benefits accruing from membership or support of other community groups delivering social and cultural benefits. Thus, members who wish to benefit from security club goods may be unable so readily to free-ride without risking sanction or opprobrium within the other groups in which they also have a continuing interest or need to participate. So, the web of group activity in suburbs may generate not only activists who are able to deal in social capital so as to mobilise community defences in the form of security club goods but also provide them with a means of enforcing the participation of others (see Olson, 1965). Without such social support it is unlikely that the state and its agencies, especially the public police, would be able to sustain these groups over the long term, whatever institutional bias there may be towards them (pace McConville and Shepherd, 1992). And given that membership of such ‘security clubs’ overlaps with membership of local groups generally, it is also likely that collective, neighbourhood norms of property security may strengthen over time, eventually becoming self-sustaining, both organisationally and normatively (Coleman, 1990).

The principal benefit of a club good to its members is that it is "an institutional solution to the collective action problem that internalises an externality through tolls" (Sandler, 1992: 64). With the kinds of social sanctions available, members can trust that their fellow-members will continue to contribute to the collective generation of private security goods and that they will not free-ride, thus undermining and diluting the efforts of those who do participate in private security production. Such social sanctions may become less necessary the more ‘exclusive’ the club since membership exclusivity ensures that the externalities of individual private security efforts will be retained within the club for the benefit of club-members only and will likewise not be seen to suffer from the threat of congestion of the club’s security from external parties wishing to share in the benefit.

Generally, the price-mechanisms of the housing market in Britain tend to ensure that the more affluent suburbs are the most exclusive, usually through increased social and spatial ‘distance’ placed between themselves and the perceived sources of risk (Hope, 1999). Membership exclusivity is preserved by insulating the club’s boundaries. In open, urban environments, insulation may come from the ‘buffering’ effect of being surrounded by similar suburbs. Yet in environments where security is more congested – or where crime is believed to be all pervasive, growing and crowding-in on the suburbs’ social order - more specific means of ‘policing the boundaries’ may be needed to retain security activity within the ‘club’ of the suburb, to preserve the positional value of its private security against encroachment, and to prevent its dilution from within.

In England and Wales, the growth of Neighbourhood Watch may have been more an expression than a particularly effective cause of this re-ordering of suburban social order. Nevertheless, its social distribution reveals how a self-sustaining social anxiety about the risk of crime victimisation combines with the availability of social capital to produce a socio-cultural mechanism that enhances the ‘clubbing’ of private security within increasingly ‘exclusive’ suburbs and captures the symbolic and material resources of the public police for greater protection. Ironically, a means may have been found for solving the responsibilization problem – for getting private citizens to take greater responsibility for crime prevention and, possibly, for helping to bring down national crime rates. Yet the price to be paid may be the creation of a greater and more sustained inequality not only in crime victimisation risk but also in the means available for its alleviation.

The collective consequence of crime risk avoidance in risk society is the creation of risk pools dictated according to social position. In Britain, the collapsing credibility of the State's ‘universal insurance policy’ against crime victimisation risk (Garland, 2001) may have induced a dynamic which encourages the clubbing of private security (Hope, 2000). Here, the logic of collective action - albeit the unintended consequence of individualised strategies of risk-pooling and risk-avoidance - may act to reinforce a form of social exclusion around crime risk. Thus, the threat of crime victimisation becomes not just a consequence of social exclusion but also a contributory cause. The challenge facing the state remains whether and how to intervene; not just to reduce the crime rate but how to distribute fairly the means for dealing with the socially manufactured risks that produce it.

APPENDIX

VARIABLES USED IN THE ANALYSIS

Household-related variables

All household-related variables are categorical-level, binary-coded (1,0) unless indicated.

Age

AGEHH Age head of household (interval-level variable)

Occupation (base: manual occupation)

JOBNONM Non-manual

JOBOTH Other

Children in household (base: none)

CHIL1215 children 12-15 yrs

CHILD511 children 5-11 yrs

CHILDLT5 children less than 5yrs

Tenure (base: owner occupied)

RENTC rent from local council (social housing) 27

RENTO rent from private landlord (incl social housing landlords)

TIED rental tied to occupation

Dwelling type (base: detached)28

SEMI semi-detached

MIDTERR mid-terrace

ENDTERR end-terrace

FLATM flat or maisonette

UCACCOM unclassified accommodation

Ethnic identity (base: white)

BLACK Afro-Caribbean

INDIANSC Indian Sub-Continental

OTHERETH Other ethnicity

ETHREF Refused

Mobility (base: other)

MOVED Household moved in past year

Satisfaction with area (base: not satisfied)

FAIRSATA Fairly satisfied

VERYSATA Very satisfied

Neighbourhood Reciprocity (base: people go own way)

MIXSOCO Mixed

GOODSOCO People help each other

Worry About Burglary (base: not worried)

NVWBURG Not very worried

FWBURG Fairly worried

VWBURG Very worried

Area-level variables

Area-related variables

All variables are at interval-level and standardised to the whole BCS sample mean for each variable, unless indicated otherwise

ZPUN5HH Proportion of households with children under 5

ZPAGE515 Population aged 5-15 yrs (per cent population)

ZPINDIAN Population Indian-Subcontinental identities (per cent population)

ZPOLD1HH Proportion of one-pensioner households (households)

ZPSPARHH Lone-parent households (per cent population)

ZPCHOUSE Households renting from municipal (social housing) landlord (per cent households)29

ZRENTED Households renting from a private sector landlord (per cent households)

ZHASSOC Households renting from a housing association (non-municipal social housing) landlord (per cent households)30

ZPVACANT Proportion of empty dwellings (of dwellings)

ZPFLATS Proportion of apartment dwellings (of dwellings)

ZPNONMAN Households with ‘household head’ in non-manual occupation (per cent households)

INCITY This is a binary coded variable indicating residence in an inner city area, it uses the same classification as that used in the BCS sampling process (see footnote 9 in main text).

Portmanteau Variables

The construction of these variables was guided by a principal components analysis. See footnote 13 in main text.

DEPRIVED The standardised variables added together to produce this variable are: the proportions of households with more than one person per room, households with three or more children, households living in housing association (non-municipal social housing) accommodation, and the percentage of males who were unemployed.

AFFLUENT The standardised variables added together to produce this variable are: the average number of cars per household plus the proportion of households living in detached houses.

YOUNG The standardised variables added together to produce this variable are: the proportions of the population aged 16 to 24 years and of single person non-pensioner households.

Standard Regions of England and Wales

(base: South East, excluding Greater London). These are binary-coded variables.

NORTH North

YORKS Yorkshire and Humberside

NWEST North West

WALES Wales

WMIDS West Midlands

EMIDS East Midlands

EANGLIA East Anglia

SWEST South West

GLC Greater London

References

Baumgartner, M.P., 1988: The Moral Order of a Suburb. New York: Oxford University Press.

Beck, U., 1992: Risk Society. London: Sage.

Beckett, Katherine. and Theodore. Sasson, 2000: The Politics of Injustice: crime and punishment in America. Thousand Oaks, CA: Pine Forge Press.

Blumstein, Alfred. and Joel Wallman, 2000: The Crime Drop in America. Cambridge, UK, New York, NY and Melbourne, Aus.: Cambridge University Press.

Cohen, Lawrence E. and Marcus Felson, 1979: Social change and crime rate trends: a routine activities approach. American Sociological Review 44: 588-608.

Coleman, James S., 1990: Foundations of Social Theory. Cambridge, MA.: Belknap Press.

Dowds, L. and P. Mayhew, 1994: Participation in Neighbourhood Watch: findings from the 1992 British Crime Survey. Research Findings 11. London: Home Office.

Ellingworth, D., T. Hope, D.R. Osborn, A. Trickett and K. Pease, 1997: Prior victimisation and crime risk. Journal of Crime Prevention and Risk Management 2: 201-214.

Ewald, Uwe, 2000: Criminal victimisation and social adaptation in modernity: fear of crime and risk perception in the new Germany. in: T. Hope and R. Sparks (Eds.): Crime, Risk and Insecurity. London: Routledge.

Felson, Marcus, 1998: Crime and Everyday Life (2nd ed.). Thousand Oaks, CA: Pine Forge Press.

Field, S. and T. Hope, 1990: Economics, the consumer and under-provision in crime prevention. in: Morgan, R. (ed.) Policing Organised Crime and Crime Prevention. British Criminology Conference 1989, Volume 4. Bristol: Bristol Centre for Criminal Justice.

Garland, David., 2001: The Culture of Control. Oxford: Oxford University Press.

Garland, David, 2000: Ethical and Social Perspectives on Situational Crime Prevention. pp. 215-224 in: Andrew von Hirsch, David Garland and Alison Wakefield (eds.): Ethical and Social Perspectives on Situational Crime Prevention. Oxford and Portland, OR: Hart Publishing.

Granovetter, M. 1973: The Strength of Weak Ties. American Journal of Sociology, 78: 360-380.

Hancock, Lynn, 2001: Community, Crime and Disorder: safety and regeneration in urban neighbourhoods. Basingstoke, Hants.: Palgrave.

Home Office, 1999: The Government’s Crime Reduction Strategy. London: Home Office.

Hope, Tim, 2000: Inequality and the clubbing of private security. in: T. Hope and R. Sparks (eds.): Crime, Risk and Insecurity. London: Routledge.

Hope, Tim, 2001: Crime victimisation and inequality in risk society. in R. Matthews and J. Pitts (eds.): Crime Prevention, Disorder and Community Safety. London: Routledge.

Hope, Tim, 1999: Privatopia on trial? Property guardianship in the suburbs’. in K. Painter and N. Tilley (eds.): Surveillance of Public Space: CCTV, street lighting and crime prevention. Crime Prevention Studies, Vol. 10. New York: Criminal Justice Press.

Hope, Tim, 1995: Community crime prevention. pp. 21-89 in: M. Tonry and D.P. Farrington (eds): Building a Safer Society: strategic approaches to crime prevention. Crime and Justice, Vol. 19. Chicago: University of Chicago Press.

Hope, Tim, 1997: 'Inequality and the future of community crime prevention'. in S.P. Lab (ed.): Crime Prevention at a Crossroads. American Academy of Criminal Justice Sciences Monograph Series. Cincinnati, OH: Anderson Publishing.

Hope, Tim, 1988: Support for Neighbourhood Watch: a British Crime Survey analysis' in T. Hope and M. Shaw (eds.): Communities and Crime Reduction .London: HMSO.

Hope, Tim and Stephen P. Lab 2001: Variation in crime prevention participation: evidence from the British Crime Survey. Crime Prevention and Community Safety: an International Journal, 3: 7-21.

Jefferson, Tony and Wendy Hollway 2000: The role of anxiety in fear of crime. in T. Hope and R. Sparks (eds.): Crime, Risk and Insecurity. London: Routledge.

Karstedt, Susanne. 2003: Typen der Sozialintegration und Gewalt: Kollektivismus, Individualismus und Sozialkapital. pp. 269-292 in: W. Heitmeyer & G. Söffner (Hrsg.), Gewalt: Neue Entwicklungen und alte Analyseprobleme,. Frankfurt: Suhrkamp.

Karstedt, Susanne. (im Druck): ‚Linking Social Capital‘. Überlegungen zur institutionellen Dimension des Sozialkapitals. in: F. Kessl & H.-U. Otto (Hrsg.), Soziale Arbeit und Soziales Kapital. Zur Kritik lokaler Gemeinschaftlichkeit. Wiesbaden: Westdeutscher Verlag

Langan, Patrick A. and David P. Farrington, 1998: Crime and Justice in the United States and in England and Wales, 1981-96: Executive Summary. Bureau of Justice Statistics. Washington, DC: US Department of Justice.

Laycock, G. and N. Tilley, 1995: Policing and Neighbourhood Watch: strategic issues. Crime Detection and Prevention Series Paper No 60. London: Home Office.

Lewis, D.A., Grant, J.A. and Rosenbaun, D.P., 1988: The Social Construction of Reform. New Brunswick, NJ: Transaction Books.

Loader, I., E. Girling and R. Sparks, 2000: After success? Anxieties of affluence in an English village. In: T. Hope and R. Sparks (Eds.) Crime, Risk and Insecurity. London: Routledge.

McConville, M. and Shepherd, D. 1992: Watching Police, Watching Communities. London: Routledge.

Mawby, R..I., 2001: Burglary. Cullompton, Devon: Willan Publishing.

Melossi, D. and R. Selmini, 2000: ‘social conflict and the microphysics of crime: the experience of the Emelia-Romagna Città sicure project. in: T. Hope and R. Sparks (eds.) Crime, Risk and Insecurity. London: Routledge.

Olson, M., 1965: The Logic of Collective Action: public goods and the theory of groups. Cambridge: Harvard University Press.

Menard, S., 1995: Applied Logistic Regression Analysis. Quantitative Applications in the Social Sciences 07-106. Thousand Oaks, CA: Sage Publications.

Povey, David, Sian Nicholas and Heather Salisbury, 2003: Crime in England and Wales: quarterly update to December 2002. Home Office Statistical Bulletin 05/03. London: Home Office.

Putnam, R.D., 2000: Bowling Alone. New York: Simon and Schuster.

Reiner, R., 2000: The Politics of the Police. Oxford: Oxford University Press.

Rosenbaum, D.P., 1988: A critical eye on neighbourhood watch: does it reduce crime and fear?. In T. Hope and M. Shaw (eds.): Communities and Crime Reduction .London: HMSO.

Rosenbaum, D.P., Lewis, D.A., and Grant, J.A., 1986: "Neighborhood-based crime prevention: assessing the efficacy of community organizing in Chicago". In Rosenbaum, D.P. (ed.): Community Crime Prevention: does it work?. Beverly Hills, CA: Sage.

Sandler, T., 1992: Collective Action. New York: Harvester Wheatsheaf.

Sasson, T., 1995: Crime Talk: how citizens construct a social problem. New York: Aldine de Gruyter.

Simmons, J. and T. Dodd (eds.), 2003: Crime in England and Wales 2002/03. Home Office Statistical Bulletin 07/2003. London: Home Office

Sims, L., 2001: Neighbourhood Watch: findings from the 2000 British Crime Survey. Findings 150. London: Home Office.

Taub, R. P., Taylor, D.G. and Dunham, J.D., 1984: Paths of Neighborhood Change. Chicago: University of Chicago Press.

Skogan, W.G. 1988: Community organisations and crime. in: M. Tonry and N. Morris (eds.): Crime and Justice: a review of research, Vol. 10. Chicago: University of Chicago Press.

Trickett, A, D.R. Osborn and D. Ellingworth, 1995: Property crime victimisation: the roles of individual and area influences. International Review of Victimology, 3: 273-295.

van Dijk, J.J.M., 1994: Understanding crime rates. On the interactions between the rational choices of victims and offenders. The British Journal of Criminology, 34, 626-639.

Walklate, S. and Evans, K., 1999: Zero Tolerance or Community Tolerance: managing crime in high crime areas. Aldershot, Hants.: Ashgate.

Zimring, Frank and Gordon Hawkins, 1997: Crime is Not the Problem: lethal violence in America. New York: Oxford University Press.