PATTERNS AND DRIVERS OF CO-PRODUCTION IN NEIGHBOURHOOD WATCH IN ENGLAND AND WALES : FROM NEO-LIBERALISM TO NEW LOCALISM

Justice Studies (ISTD). This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. PATTERNS AND DRIVERS OF CO-PRODUCTION IN NEIGHBOURHOOD WATCH IN ENGLAND AND WALES: FROM NEO-LIBERALISM TO NEW LOCALISM


Introduction
The 'co-production' of public services 'Co-production' is a cornerstone of public policy reform in England and Wales as it is across the globe (Osborne et al. 2016: 639).This provides a model for the agents of the state and citizens to come together to contribute to the design and/or delivery of a public service or public good (Mattson 1986;Ostrom 1996;Pestoff 2006;Bovaird 2007;Parrado et al. 2013;Brandsen and Honingh 2015;Voorberg et al. 2015;Alford and Yates 2016;Van Eijk and Steen 2016;Van Eijk et al. 2017).Co-production may take multiple forms involving a wide range of actors and activities (Brandsen and Honingh 2015;Alford and Yates 2016;Loeffler and Bovaird 2016;Van Eijk and Steen 2016).But however it is configured, it represents a 'process through which inputs from individuals who are not "in" the same organization are transformed into goods and services' (Ostrom 1996(Ostrom : 1037)).As such, it implies an interaction between public agents and citizens that goes beyond consultation or listening to citizens or responding to citizens' concerns towards some kind of active, collective production of tangible outputs or outcomes (Mattson 1986;Ostrom 1996;Alford and Yates 2016).Active voluntary citizen involvement appears to be the critical dimension of all co-production research (Mattson 1986: 52).
The concept of co-production, central to many areas of contemporary public administration, purports to offer a cost-effective way to both increase efficiency and enhance the effectiveness of public services (Mattson 1986;Ostrom 1996;Pestoff 2009;Voorberg et al. 2015;Osborne et al. 2016;Van Eijk et al. 2017).It gained traction during the New Public Management reforms of the 1980s (Pestoff 2009;Osborne et al. 2016) and continued to be highlighted as a means of solving the major demographical, political and economic challenges facing Western democratic governments in the 21st century (Pestoff 2009).It has also been identified as essential for sustaining provision in European welfare systems in the face of budget constraints (Pestoff 2006).However, since co-production differs from the 'traditional model' of public service-where public officials are exclusively charged with responsibility for designing and providing services to citizens (Pestoff 2006)-it has also been seen to offer a means of democratizing public services through promoting citizen empowerment and participation (Ostrom 1996;Pestoff 2009) (see also Lowndes and Pratchett 2011;Stoker 2006).

Citizens, crime control and Neighbourhood Watch
Citizens have long played a role in the enactment of crime control and the ways that they have done so are diverse.In days gone by, communities were expected to join the 'hue and cry', and householders were expected take turns volunteering as watchmen and presenting offenders at court (Rawlings 2002).Today, citizens in England and Wales may volunteer as special constables, may form street patrols and may, through collaborating with the police to identify and resolve crime problems, contribute to community policing (Skogan 2008;Tilley 2008;Bullock 2014).Yet while there clearly is a long history of citizen involvement in crime control and the roles played are wide ranging, contemporary calls for citizens in England and Wales to co-produce can be situated within the neo-liberal Conservative administrations of the 1980s and 1990s, which stressed the desirability of rolling back the state, improving the efficiency of public services and promoting volunteering and community/citizen action (Crawford 1999;Bullock 2014).
It was into this political and economic context that Neighbourhood Watch (NW) became established in England and Wales.NW brings neighbours together to act as the 'eyes and ears' of the police through watching out for suspicious behaviours in their neighbourhoods and reporting them, and to protect their properties and those of their neighbours (Rosenbaum 1987;1988;Bennett 1990;Laycock and Tilley 1995;Gresham et al. 2004;Bullock 2014).NW appealed at once to the politically Conservative governments of the 1980s, concerned about maintaining standards of behaviour, and to libertarian instincts regarding the importance of voluntarism, the role of the 'active citizen' and the minimal role of the state (Hope 1995: 43).The NW movement grew very quickly in this climate (Rosenbaum 1987;1988;Skogan 1988;1989;Bennett 1990;Hope 1995;Bullock 2014) and was to become one of the largest voluntary organizations in England and Wales and one of the largest single organized crime prevention activities in the world (Bennett et al. 2008).Co-production has been promoted as a way of increasing the effectiveness of public services (Mattson 1986;Ostrom 1996;Pestoff 2009;Voorberg et al. 2015;Osborne et al. 2016;Van Eijk et al. 2017), and reflecting these wider trends, citizen involvement in policing was promoted as cost-effective response to the problem of crime (Skogan 2008;Tilley 2008;Bullock 2014).Equally, citizen involvement in the governance and delivery of crime control from the 1970s onwards was perceived to be an effective response to various crises.Community policing styles were endorsed in the context of contested relationships between the police and communities-characterized by a disconnection between agents of the public police and citizens-and by troubled relationships between the police and minority ethnic communities (Scarman 1981;Skogan 2008;Tilley 2008;Bullock 2014).And NW itself was 'promoted as evidence of a break with a past which had been marked by fractured relations with the community, a remote police service and styles of policing which had contributed to major confrontations with the public, including street disorders' (McConville and Shepherd 1992: 1).
The promotion of the citizen as a co-producer of public services continued unabated into the 1990s and beyond.New Labour, through their 'New Localism' agenda, sought to embed the citizen ever more firmly within the oversight and delivery of public services and to redistribute power from central bureaucrats to local structures and, through the processes of double devolution, to citizens themselves (Lowndes and Pratchett 2011;Stoker 2006).Central to this narrative were 'active citizens' (Lowndes and Pratchett 2011;Stoker 2006).Devolving power to 'active citizens' would improve effectiveness and generate new democratic accountabilities and scrutiny (Home Office 2004; see also Bullock 2014).In respect to policing governance this was usually linked to Neighbourhood Policing, a form of community policing, which aimed to engender greater democratic accountability through facilitating citizens involvement in determining how their communities are policed and ensuring that police listen and act on community concerns (Home Office 2004;Quinton and Morris 2008;Bullock 2014).Themes of devolving power to the responsible and 'active citizen' and imploring them to come together with state agents to address crime problems and to engender broader cultures of responsibility and mutuality have continued in the post-New Labour era into the Conservative-Liberal Coalition (2010-15) and the Conservative administration which followed it (2015+) (see e.g.Cabinet Office 2010).In respect to crime control, community policing continued to play a central role in the pursuit of direct citizen oversight of policing.And democratic accountability took a new twist with the introduction of elected officials (police and crime commissioners) who serve to orient police services around the needs of citizens and communities and offer the prospect of indirectly holding the service to account (see e.g.Bullock 2014;Lister and Rowe 2015).

Aims and contribution
While there has been much theoretical interest in the value of co-production, empirical studies of the key drivers of co-production are comparatively rare, with a general reliance on small-scale case studies (Parrado et al. 2013;Van Eijk and Steen 2016;Brandsen and Honingh 2015).And despite its world-wide influence and position as a primary way through which the state and citizens may co-produce crime control, important questions remain about the drivers of participation in NW and the ways that participation is distributed on the basis of household and neighbourhood factors.Existing studies have indicated that it is difficult to both establish and sustain NW activities (Mukherjee and Wilson 1987;Rosenbaum 1988;Laycock and Tilley 1995) and have painted a mixed picture of the drivers of participation.But it has become conventionally understood that NW is more easily implemented in low crime rate, relatively affluent areas where citizens own their own homes (McConville and Shepherd 1992;Laycock and Tilley 1995;Hope 2000).NW then, may actually be regressive in nature, further concentrating crime control in the hands of those that need it least.This raises important questions about the distribution of security, the use of state resources and social justice (Rosenbaum 1987;Skogan 1988;Laycock and Tilley 1995;Hope 2000;Bullock 2014).Similarly, the 'club goods' associated with NW-the security marking, the watchful neighbours-may provide powerful assurances about potential risks, addressing particularly the needs of riskaverse property-owners (Hope 2000: 160).
A number of empirical studies documented the proliferation of NW in the early 1980s (e.g.Husain 1988;Bennett 1989;Laycock and Tilley 1995;Yarwood and Edwards 1995), but few studies of NW have been conducted since the 1990s (Bennett et al. 2008).However, understanding the nature of the relationship between citizens and the structures of the state is arguably of greater importance than ever before.Since the establishment of the first English NW scheme in the 1980s, the relationship between public policing and citizens has been transformed.And the recent promotion of citizen involvement goes hand in hand with the political focus of successive British governments on 'austerity' following the global economic crisis of 2008, which has led to significant cuts to police funding (HM Treasury 2010; HMIC 2017) and the retrenchment of the public police, reducing the visibility of uniformed officers in local communities (HMIC 2017).But the efficacy of such approaches to crime control obviously relies on the nature of citizens' relationship with the state, or at least with the police, and their ability and willingness to organize themselves at the local level.
We use data from the Crime Survey for England and Wales (CSEW) to examine patterns of participation in NW in England and Wales since its inception in the neoliberal governments of the early 1980s through to the end of the New Labour Era in 2010. 1 Combining CSEW data with information from the census, we then use multilevel models to explore the extent that household drivers of participation in NW have changed since its introduction, and the role that area crime and disadvantage may play.In so doing, we consider the vexed issue of whether NW takes root in areas where it is least needed, i.e., in low crime and affluent areas.We find falling rates of participation since the early 1990s, with one-in-ten households in England and Wales part of NW by 2010-11.Participation is less likely in disadvantaged areas, but more likely in high-crime areas.We also find that crime rates moderate the effect of disadvantage on participation in NW.Disadvantaged areas with a higher level of crime are more likely to operate an NW scheme than similar areas with a low level of crime, but both are less likely to have a scheme in operation than areas with a low level of disadvantage.Thus, our findings challenge any assumptions about there being a straightforward linear relationship between crime rates and NW or disadvantage and NW.Instead, we conclude that citizens will participate in NW where the 'conditions are right'.

Literature review
Though arrangements may vary in practice, the term co-production suggests that outcomes relevant to the provision of public services are being produced by the activity of public service agents together with the voluntary activities of citizens or groups of citizens (Mattson 1986;Pestoff 2006;Parrado et al. 2013;Alford and Yates 2016).In respect of NW, co-production primarily involves agents of the public police coming together with citizens or groups of citizens.The public police provide practical support to shore up the operation of NW (usually their time or in-kind support) and provide symbolic support as citizen groups seem to demand the legitimacy that police sponsorship confers (Gresham et al. 2004;Bullock 2014).The formal organization of citizens is not necessary for co-production, but organizations may facilitate co-ordination between citizens and public agencies and so enhance the levels of co-production (Pestoff 2006).Indeed, NW is supported by national infrastructure (see e.g.Gresham et al. 2004).However, coproductive relationships and activities are various and complex, and their contribution may be more or less allied to the core task of a public service (Brandsen and Honingh 2015).NW incorporates diverse activities more or less clearly linked to the core police task (Laycock and Tilley 1995;Gresham et al. 2004).Co-productive relationships and activities also generally create a mixture of public and private value (Alford and Yates 2016).Similarly, an aim of NW is to protect individual private properties and people (e.g. through provision of property marking or personal alarms) while reducing overall public risk (e.g. through reducing the opportunities for crime within a neighbourhood) (Rosenbaum 1987;1988;Laycock and Tilley 1995).
The reasons why citizens co-produce are varied (Alford 2009;Pestoff 2012;Verschuere et al. 2012;Alford and Yates 2016).While it is often held that citizens co-produce to receive something tangible or meaningful in return, citizens actually co-produce for a variety of extrinsic and intrinsic reasons (Alford 2009;Verschuere et al. 2012).Citizens will not, however, spontaneously co-produce simply because benefits could be achieved (Ostrom 1996(Ostrom : 1082)).If co-producing is straightforward (Alford 2009;Pestoff 2012;Verschuere et al. 2012;Van Eijk and Steen 2016), or where an issue is salient or meaningful (Loeffler and Bovaird 2016;Van Eijk and Steen 2016), citizens are more likely to act.There also needs to be attention to organizational issues such as training public officials and establishing appropriate institutional arrangements (e.g.management, oversight and processes) (Ostrom 1996;Alford 2009;Pestoff 2012;Verschuere et al. 2012;Loeffler and Bovaird 2016;Voorberg et al. 2015).Understanding citizen's needs and motivations is essential, as is the provision of information and incentives, and the clarification of expectations and outcomes (Ostrom 1996;Alford 2009).Neighbourhood-level factors also play a role, with the higher levels of social capital, denser networks and the presence of voluntary opportunities all increasing co-production (Voorberg et al. 2015;Van Eijk and Steen 2016; see also Putnam 2000).The presence of self-efficacy-a belief that one can make a meaningful difference-is also often identified as a stimulator of co-production (Parrado et al. 2013;Alford and Yates 2016;Loeffler and Bovaird 2016), as are high levels of trust and satisfaction with government (Loeffler and Bovaird 2016;Van Eijk and Steen 2016).
Studies of the drivers of participation in NW reveal mixed results.While some studies demonstrate that crime and disorder may motivate citizens to participate (Lavrakas and Herz 1982;Skogan 1989;Pattavina et al. 2006), others find the opposite.High rates of crime may deter participation in NW for a number of reasons.They may generate feelings of powerlessness, increase social isolation and undermine any belief that residents have that they are able to collectively control crime (Skogan 1988).Or high-crime rates may generate suspicion among residents, weakening the trust needed to motivate them to work together to implement crime prevention interventions (Henig 1984;Hourihan 1987;Rosenbaum 1987;Skogan 1988;Hope 1995;Laycock and Tilley 1995).
Especially where members are asked to exchange information about themselves, keep surveillance of and/or report on strangers (or neighbours) (Henig 1984;Hope 1995).The relationship between crime and participation in NW may not be linear (Husain 1988;Skogan 1989).For example, Skogan (1989) outlines a situation where residents in both high and low crime rate areas may be deterred from participating-in the former, the intervention may seem too mild to deal with the problem and with the latter there is no need-but in areas of average crime rates motivation may be high.
Existing research tends to identify a negative relationship between neighbourhood disadvantage and membership of NW (Shernock 1986;Hope 1988;Husain 1988;Yarwood and Edwards 1995).This is believed to reflect both reduced opportunities for participation in more disadvantaged communities and that residents of poorer areas are less likely to be involved even when the opportunity is available (Hope 1995).However, Skogan (1989) argues that while there may be an effect of disadvantage when it comes to the opportunity to participate-there are more opportunities in better-off areas-this effect is diminished because people from better-off areas have fewer reasons to participate.By this reasoning, neighbourhood disadvantage must be considered in conjunction with the objective need for NW in the local area, as reflected, for example, by the crime rate.Pattavina et al. (2006) make a similar observation in explaining their finding that participation in community crime prevention was not highest in the wealthiest, lowest crime rate areas.Instead, the factors affecting citizen participation in crime prevention varied by neighbourhood crime risk levels.In high-crime neighbourhoods, residents who felt like part of the neighbourhood, minority residents and residents who believed that the police get to know residents were more likely to participate.In low-to moderate-risk neighbourhoods, a more complex web of predictors fuelled participation, with additional differences based on home ownership and previous victimization (Pattavina et al. 2006: 228).
It is often assumed that citizens' beliefs about the police are related to their willingness to engage in anti-crime measures at the neighbourhood level (Frank et al. 1996).Studies have suggested that participation in anti-crime initiatives, including NW, is facilitated where residents have favourable opinions towards the police and believe that the police do a good job (Shernock 1986;Bennett 1989;Yarwood and Edwards 1995).However, other studies find no evidence of differential participation once other factors were controlled for (e.g.Frank et al. 1996).Bennett (1989) found that participants in NW were more likely than non-participants to view the police as helpful, believed that the police were doing a good job, and were pleased with police performance.But participation was not associated with knowing a police officer by name, feeling the police understood the problems of residents living in their area or among those who had made proactive contact with the police over the last year.This suggests that the nature of the relationship with the police may be playing a role in shaping participation.Indeed, Reisig (2007) found that citizens who judged police practices to be fair and respectful were more willing to participate in community crime prevention initiatives.This finding held across areas with low, moderate and high levels of property crime.Pattavina et al. (2006) also found that participation in crime prevention activities was less to do with perceptions of the effectiveness of the police and more to do with the development of personal relationships between the police and residents in these areas.
North American research tends to find that participation in organized anti-crime initiatives can be understood as an extension of 'civic mindedness' and wider participation in community organizations (Rosenbaum 1988;Skogan 1988;Bennett 1989, and see studies by DuBow and Podolefsky 1982;Lavrakas and Herz 1982;Shernock 1986).Bennett (1989) also found participants in NW were more involved in their community than non-participants, leading him to conclude that participation might be a function of both crime and wider community participation.So while the salience of crime provides the external, environmental impetus for the development of community crime prevention programs, this may be more a necessary rather than a sufficient condition for their origin (Lavrakas and Herz 1982).
The degree of social cohesion-measured by factors such as whether citizens have confidence in the ability of others to combat crime, felt they were friends with their neighbours and reported territorial attitudes or supportive environments-has also been linked with participation in NW (Lavrakas and Herz 1982;Hope 1988;Frank et al. 1996).Skogan (1989) found a tendency for less cohesive communities to employ more formal responses to crime, in part because in these areas informal capacity to problem-solve and organize against crime was weak.Others have reported mixed evidence.Bennett (1989), for example, reported significant differences in one of his research sites but not in the other.In addition, Pattavina et al. (2006) revealed a link between one indicator of social cohesion (residents feel like they are part of the neighbourhood) and participation but not another (rely on neighbours for help).
While studies that have examined the relationship between 'attachment' to the community and participation in NW reveal mixed findings, it has often been shown that members of NW have a 'stake in the community' (Shernock 1986;Frank et al. 1996;Pattavina et al. 2006).Especially important has been long-term residency in an area, home ownership and housing type-with those living in flats often less likely to participate (Lavrakas and Herz 1982;Henig 1984;Hope 1988;Husain 1988;Bennett 1989;Webb 1994;Frank et al. 1996).This has generally been explained by the belief that homeowners have greater commitment to maintaining standards within an area because of incentives to maintain house prices and their greater freedom and responsibility for taking prevention measures (Hope 1995: 45).
There are also socio-demographic differences in participation in NW.But while most research points to NW members being older (Shernock 1986;Webb 1994;Yarwood and Edwards 1995), even this is not universally agreed.For example, Lavrakas and Herz (1982) find participants tended to be aged less than 50 years, and Bennett (1989) reported no differences in participation on the basis of age.The evidence regarding the role played by ethnicity and gender is also mixed.Reviews of US-based studies have suggested that Black residents are more likely to participate (e.g.Bennett 1989), while other studies demonstrate that NW members are more commonly White (Henig 1984;Shernock 1986;Webb 1994;Pattavina et al. 2006) and some find no differences between ethnic groups.Similarly, while some studies have shown that NW members are more likely to be men (Hope 1988;Husain 1988;Yarwood and Edwards 1995), others have found participants are more likely to be women (Webb 1994).

Data
To assess the changing levels of NW in England and Wales and explore the key household and neighbourhood factors that remain important for participation in the co-production of NW, we use data from the CSEW.This is a nationally representative survey of residents' experiences of crime and victimization.It has fielded questions about NW to a random subsample of respondents since 1988, allowing us to provide an overall picture of the changing prevalence of NW over the last three decades.The questions were omitted from the survey in 1998, 2002/03-2003/04 and 2007/08-2008/09, which limits our ability to provide a complete time trend.The survey has consistently maintained a response rate of more than 70%, achieving a response of 75% in 2010-11 (ONS 2012).

Long-term trends in Neighbourhood Watch proliferation in England and Wales
Changing levels of NW proliferation since 1988 are examined using annual weighted estimates of three survey items: 1. Is there a Neighbourhood Watch Scheme currently operating in this area that covers your address?Yes, No, Don't know 2 2. Is your household currently a member of the scheme?Yes, No, Don't know 3. Would your household join a scheme if there were one in this area?Yes, No, Don't know For each survey year, the percentage of households reporting that the area is part of an NW scheme, levels of household participation (overall, and restricted to areas where a scheme was in operation) and the percentage of residents reporting that they would like to belong to an NW scheme if it were available, are collated.We then use time-series cross-sectional models (Gelman and Hill 2007), with observations grouped in survey year, to examine whether overall trends in the prevalence of NW are consistent across household groups.Comparatively, few measures in the CSEW have been included in a consistent way across all survey waves, so we restrict our focus to differences based on accommodation type (detached, semi-detached, terraced, flats or other), household income, housing tenure (owner occupied, private rented, social rented) and reported levels of safety at home at night.We also include controls for respondent gender, age and length of residence in the local area.All household characteristics are interacted with survey year to enable assessment of whether observed differences in levels of participation have changed over time.Models also include a quadratic function of year to account for potential non-linearities.

Results
Throughout the 1990s, there was a general increase in the number of NW schemes in operation in England and Wales, with more than 30% of areas identified as part of a scheme by 2000, up from less than 20% of areas in 1988 (Figure 1, top panel).The number of areas operating NW then remained stable throughout the 2000s and no visible increase following the 2008 financial crisis. 3However, while the number of areas with a scheme appears to have remained comparatively stable, the number of households 2 Despite asking about the presence of an NW scheme in 'this area', respondents were not required to make reference to a specific geographical boundary when answering.Respondents may therefore be reporting on different spatial geographies, leading to response variation within each area.As a result, this is treated as a household-level measure in the analysis. 3The apparent fall in 2004-05 is likely to reflect a change in question wording.Prior to 2004-05, the survey asked whether an NW scheme had ever been in operation in the area.In 2004-05, this was changed to 'in the last year'.
opting to be part of a scheme has fallen steadily (Figure 1, middle panel).Approximately 40% of households in eligible areas were part of NW by 2010-11, down from more than 80% in 1988.The main reason for not being a part of a scheme (reported by more than one quarter of respondents) was that they had not been asked to join.Other frequently noted reasons were being too busy, or not yet getting round to it, and not knowing how to join or having insufficient information about it.Only 6% said they were not interested, and just 1% did not think they were effective (for full results, see Appendix Table A1).The comparatively low levels of participation do not, however, appear to be reflective of a lack of demand, with the number of people indicating that they would join a scheme if one was available remaining relatively stable and more than 70% since 1988 (Figure 1, bottom panel).Of course, the high number of people indicating that they would like to join a scheme may in part be reflective of a social desirability bias, with respondents using this item to offset their existing lack of involvement (Nunnally 1978).
Table 1 includes results from the repeated cross-sectional models, confirming the initial growth of NW and subsequent declines.Residents of semi-detached properties, terraces and flats are less likely to report being in an NW area than those living in detached accommodation (Model 1).Wealthier households, owner-occupiers and longterm residents are also more likely to report being in an NW area.These differences are generally stable over time, although there is moderate evidence that the income gap has reduced, and there has also been a small increase in reported NW prevalence among those living in social housing.A similar picture is evident when household membership of NW is considered (Model 2), although there is evidence of further declines in membership among semi-detached and terraced accommodation.Finally, we note lower odds of future NW enrolment among resident living in flats, those living in rental accommodation and those who report feeling safe at home after dark (Model 3).Conversely, women are more likely to indicate they would join a scheme, as are those with higher levels of income.

Explaining membership in Neighbourhood Watch
To provide a more detailed assessment of the household and neighbourhood characteristics that remain influential for participation in NW, we estimate multilevel logistic regression models (Goldstein 2011) for each of our three measures of NW participation in 2010-11.

Household characteristics
To account for differences in involvement in NW between households, we include details of the household reference person's gender, age, ethnicity and socio-economic status.
We also record the number of children in the household, how long they have been resident in the area and whether it is a multiple adult household.Details of the type of accommodation, housing tenure, whether there is a visible burglar alarm, whether the house is left unoccupied during the day and whether the house is surrounded by visible signs of disorder 4 are also included.Finally, we also identify those respondents who reported being a victim of crime in the previous 12 months, whether they report worrying about crime in general and whether they believe the local police do a good job (all binary coded). 5Full descriptive details are included in Appendix Table A2.

Area characteristics
To examine differences in NW participation between different local communities, we link the household-level data from the CSEW to Middle layer Super Output Areas (MSOA)-a census geography with each area comprising approximately 2,000 households that were grouped together based on similarity of tenure and accommodation type.Respondents are clustered in a total of 2,230 MSOA.For each area, we include a number of area measures derived using data from the 2001 census. 6A total of 21 different census variables were combined using a factorial ecology approach (Rees 1971) to produce five distinct indices measuring differences between each area (full factor loadings in Appendix Table A3).These cover: • Concentrated disadvantage-with more disadvantaged areas typically having higher numbers of single-parent families, households on income support and unemployed people, but fewer residents identified as in managerial and professional occupations, and with fewer owner-occupiers.• Urbanicity-higher scores characterize areas with higher population density and more domestic properties, but relatively little green space.• Population mobility-areas score higher if there are higher levels of in-and out-migration, as well as more single-person households. 4 To measure signs of disorder, we used factor analysis to combine the scores from three interviewer ratings of the immediate area around each household covering (factor scores in parentheses): levels of litter/rubbish (0.78), vandalism/graffiti (0.83) and housing condition (0.84). 5 The CSEW selects one individual from each household to complete the survey; therefore, it is not possible to further unpick the relative influence of households and individuals.We therefore assume that these individual measures provide a reasonable approximation for the shared views and experiences of the household.
6 Boundary changes precluded us from incorporating data from the 2011 census.
• Age profile-higher scores associated with areas that have a younger population.
• Housing structure-higher scores for areas with more terraced and vacant properties, but fewer flats.
We also include a general measure of the level of crime in each area, calculated as the population weighted average score for the crime component of the 2010 Index of Deprivation.This is the most temporally proximate year of the Index of Deprivation that precedes our survey data, and is based on police recorded crime in 2008.In addition to main effects for all neighbourhood characteristics, we also include the interaction between the neighbourhood level of concentrated disadvantage and the crime rate to see whether objective crime risks moderate the link from deprivation to participation in NW (Skogan 1989;Pattavina et al. 2006).

Results
By 2010-11, NW was operating in one third of areas in England and Wales, although only 40% of households in those areas were active members of the scheme.Table 1 includes results from three multilevel models examining in more detail which areas and households were most likely to be part of NW.7 Consistent with expectations, we find that NW is significantly less likely to be in operation in areas where there are greater levels of concentrated disadvantage, with the odds of a scheme being in operation approximately 24% lower for every 1 SD increase in the level of disadvantage (Model 4).NW schemes are also significantly less likely to be found in more urban areas (with an estimated odds ratio of 0.85) and areas with a younger age profile (0.91).We also find NW is less likely in areas that are characterized by more terraced accommodation and vacant properties (0.87).These areas also tend to have slightly lower than average numbers of flats, suggesting a more complex picture of the role that housing structure may play in the presence of NW schemes (Table 2).
Areas with a higher crime rate are significantly more likely to be part of NW, although this effect is dependent on the level of neighbourhood disadvantage.Figure 2 demonstrates this moderating effect for a 'typical' area (when all other individual and neighbourhood characteristics are held at their mean).In areas where there is a low level of neighbourhood disadvantage (bottom 25% of the distribution), the probability of an NW scheme being in operation is comparatively high and is not influenced by the surrounding levels of crime.Conversely, in areas with higher levels of disadvantage (top 25% of the distribution), NW is less likely to be in operation and is more closely dependent on the levels of crime.So disadvantaged areas with a higher level of crime are more likely to operate an NW scheme than similar areas with a low level of crime, but both are less likely to have a scheme in operation than areas with a low level of disadvantage.Additional sensitivity checks restricting the analysis to the most disadvantaged neighbourhoods confirmed this non-linear relationship with crime, but that all other effects operated as shown in the main models.Significant differences between neighbourhoods are still evident when these area characteristics have been accounted for, with a residual neighbourhood variance of 0.52.Differences are also evident between particular types of households.Households that hold more positive views of the local police have almost 50% higher odds of living in NW areas.Social housing and households that have a visible burglar alarm are also Fig. 2 Predicted probability that area is part of a Neighbourhood Watch scheme.Predicted probabilities are calculated for the 'typical' area, with all other variables held at their mean value more likely to be in NW areas.In contrast, households surrounded by more signs of low-level disorder are significantly less likely to be in NW areas (0.83), as are households containing more children (0.91).
Restricting the focus to those areas where a scheme is known to be operating (Model 5), households are more likely to belong to NW if they have been resident in the area for longer and when the head of the household is older.People living in terraced housing, semi-detached accommodation and flats are less likely to participate, as are those who rent their property.People living in houses surrounded by greater levels of disorder are also significantly less likely to participate.The included area characteristics are generally not influential in determining whether particular households opt to be part of an NW scheme, although we do find that houses in more disadvantaged areas are less likely to participate while those in high-crime areas are more likely to be involved.Significant differences between neighbourhoods remain, with a residual neighbourhood variance of 0.81 suggesting that other features of the local neighbourhood may be influential in determining whether particular households choose to opt into an NW scheme.
Despite the steady fall in the proportion of households opting to be part of NW since 1988, the levels of interest from non-member households have remained relatively stable.Model 6 shows that Black, Asian and Other Minority ethnic households have approximately 50% higher odds of indicating that they would be willing to join a scheme.People who worry about crime and those who have more favourable attitudes towards the police are also more likely to want to be part of NW.In contrast, those living in flats, as well as those in rented accommodation, have significantly lower odds of wanting to participate.We also find that people living in houses that are surrounded by more visible signs of disorder are also less likely to want to be part of NW.However, we find no systematic differences between areas.

Conclusion
This article has examined patterns of participation in NW in England and Wales since its inception in the early 1980s to the end of the New Labour era .This issue is an important one since the co-production of services has been a major theme within the narratives of successive governments in England and Wales, as elsewhere.The influence of co-production is also contested contest as it is based on an assumption that citizens will work with service providers to construct public goods (Mattson 1986;Pestoff 2009).We add to the theoretical and empirical debates regarding who 'produces' crime control in two key ways.
First, we add to the body of literature that suggests that it should not be assumed that citizens will co-produce (Osborne and Strokosch 2013).We find only moderate uptake of NW in England and Wales and a general decline in individual membership.While the number of areas with NW in England and Wales rose quickly following their initial inception in the early 1980s, they have remained stable, at about 30%, since the early 2000s.And at the same time, household membership has been falling-from 80% in 1988 to about 40% by 2010-11.As a result, only one-in-ten households in England and Wales was part of NW by this time.NW rose to prominence at the height of the neo-liberal governments of the 1980s-1990s.But declined during or shortly after New Labour politics came to the fore.The reasons for this cannot be revealed by the present analysis.However, our findings undermine any assumption that citizens are routinely prepared to accept the burden of crime control that current political discourses promote and economic necessity may demand.They also question the long-term viability of approaches to crime control which seek to co-opt the initiative of citizens (see also Mukherjee and Wilson 1987;Rosenbaum 1988).
Second, we add to the body of literature regarding who co-produces, and in particular the vexed question of whether NW has regressive effects (Rosenbaum 1987;Skogan 1988;Laycock and Tilley 1995;Hope 2000;Bullock 2014).NW has often been associated with wealthier, long-term residents who own their own homes and so have attachments to the security of place and property (e.g.Lavrakas and Herz 1982;Henig 1984;Shernock 1986;Hope 1988;McConville and Shepherd 1992;Frank et al. 1996;Pattavina et al. 2006).We make similar observations though we report a small increase in reported NW membership among those in social housing.Indeed, consistent with expectations, we found that NW is significantly less likely to be in operation in areas where there are greater levels of concentrated disadvantage.However, we have also demonstrated that crime rates moderate the effect of disadvantage.And that disadvantaged areas with a higher level of crime are more likely to operate an NW scheme than similar areas with a low level of crime.There are several possible explanations for this.
First, in areas of relative affluence, the costs of establishing and maintaining NW, which may not ask much of its members, may be low.Hence, they can be established even where crime risks are low and action again crime not a necessity.The situation is different in less affluent areas.In these cases, there are hurdles-such as fewer opportunities to volunteer, lower levels of trust between citizens and between citizens and public officials-that may need to be overcome to seed and sustain activity are more extensive.As such, reflecting studies that suggest that propensity to co-produce is related to the salience of a social problem (Loeffler and Bovaird 2016;Van Eijk and Steen 2016) efforts may be warranted only where the crime risk is particularly elevated (see also Skogan 1989).Second, it is plausible that practitioners exert greater efforts to establish NW in disadvantaged areas with higher crime risk (Gresham et al. 2004).However, research has failed to demonstrate that increased efforts by practitioners in areas of disadvantage lead to the establishment of more NWs (Rosenbaum 1987).Third, our observations may reflect variation in the characteristics of those individuals who live in high-risk areas, for example better-off individuals in high-risk areas may be driving participation.To assess this possibility, we conducted additional analyses looking specifically at the most disadvantaged neighbourhoods (available on request).However, this did not reveal further area or household influences on the proliferation of NW.
Whatever the explanation, our finding that citizens will respond where objective risks are high in areas where conditions are unfavourable is an important one.It has implications for how we think about the concept of co-production and more specifically how we think about citizen participation in crime control at the local level.Our observations call into question the oft-held view that neighbourhood disadvantage undermines co-production.Suggesting instead that citizens will co-produce in unfavourable circumstances-but the conditions have to be right.

Fig. 1
Fig. 1 Trends in participation in Neighbourhood Watch (NW).Respondent reports that area is part of an NW scheme (top panel); household a member of NW (middle panel); would join an NW scheme if one was available (bottom panel)