Austerity policing: Is visibility more important than absolute numbers in determining public confidence in the police?

Abstract

The recent deep cuts to police force budgets in the United Kingdom have reawakened longstanding debates about the effect of police numbers and organization on the crime rate and public confidence in the police. Whereas some claim that a reduction in numbers is likely to have a deleterious effect, others argue that raw numbers are less important than how the police are organized and deployed. By cutting red tape and focusing staffing reductions on ‘back-office’ functions, the argument goes, it should be possible to maintain a consistent ‘frontline’ presence, which is a key aspect of policing for maintaining citizen confidence in the service. In this article we use administrative data linked to the British Crime Survey in order to assess the relative importance of police numbers and police visibility in determining public confidence in the police. We find, as expected, that visibility has a significant and positive effect on confidence. However, we also find a significant and positive effect of police numbers over and above the effect of visibility. Moreover, because the extent to which police are visible in local areas is itself a function of the number of police employed, we find that the number of police has an additional indirect influence on public confidence through its direct effect on visibility. By implication, reducing police numbers is likely to erode public confidence in the police, even if frontline visibility is maintained through organizational efficiency.

Keywords

British Crime Survey confidence in policing multilevel modelling police numbers police visibility

Introduction

Policy-makers and members of the public have long been concerned about the number of police officers who are employed and working to reduce crime. In many countries, this has led to steady increases in police numbers over the past 20 years, a trend that is only now beginning to falter in the face of deep and widespread contractions in public sector expenditure. In the United States, for example, the total number of officers increased steadily from the early 1990s (Reaves, 2011), and in Canada police numbers increased from 184 to 201 police officers per 100,000 people between 2001 and 2011 (Statistics Canada, 2011). Increases in police numbers over the last 10 years have also been evident in Australia and countries across Europe, including Sweden, Spain and Belgium (UNODC, 2010), and the same trend was evident in England and Wales (Dhani and Kaiza, 2011). Policies to support increasing police numbers are founded on the intuitive belief that more police will result in fewer crimes being committed, more crimes being solved and heightened confidence among the general public in the job that the police are doing (Australian Federal Police, 2008; Home Office, 2001). However, in response to the unsustainable burden on public expenditure of continued expansion in police numbers, questions have recently been raised about whether the achievement of these goals is really a matter of absolute numbers (also referred to as ‘police strength’), or whether it is more a function of how the police are structured, managed and deployed between ‘frontline’ and ‘back-office’ roles (Cameron, 2011). Although there has been considerable research into the relationship between police numbers and crime, comparatively little is known about how public confidence in the police is affected by police numbers, and whether it is possible to reduce police strength without harming public confidence in the service, by maintaining the visibility of frontline officers. It is to this debate that we seek to contribute in this article.

Background: Police numbers, visibility and public confidence

There is a large body of research that has sought to evaluate the effect of police numbers on the incidence of crime. Systematic reviews of this literature have revealed a rather inconsistent mix of both null and positive effects (Eck and Maguire, 2000; Marvell and Moody, 1996). The inconsistency in findings across studies has been attributed to the difficulty in untangling the direction of causality between police numbers and crime in these early investigations, with the obvious problem that police deployment is not random but concentrated in areas with higher levels of crime (Cameron, 1988; Marvell and Moody, 1996). Thus, without appropriate adjustments, it is common to observe a counterintuitive negative association between police strength in an area and the local crime rate. More recently, researchers have applied more sophisticated research designs such as instrumental variables and experimental and quasi-experimental methods, which are better suited to addressing questions of a causal nature than standard observational study designs. These have generally concluded that increasing police numbers does in fact reduce at least some types of crime in certain contexts (Di Tella and Schargrodsky, 2004; Draca et al., 2011; Machin and Marie, 2011).

Insofar, then, as confidence in the police is responsive to variation in the objective incidence of crime (Sampson and Jeglum-Bartusch, 1998; Schafer et al., 2003), we should expect reductions in police strength to have a negative effect on confidence. But public confidence in the police also appears to be strongly affected by the way in which the police organize and execute their functions. Jackson and colleagues (Jackson and Bradford, 2009; Jackson and Sunshine, 2007; Jackson et al., 2009), for instance, have argued that the police are perceived as prototypical representatives of the community and should, therefore, demonstrably represent and protect community values and maintain control and civility, engaging as active members of the community. It is this symbolic function, they argue, rather than more instrumental concerns regarding crime rates and fear of crime, that drives confidence in policing. Similarly, Bahn (1974) has pointed to the reassuring presence of the ‘bobby on the beat’ – ready to apprehend criminals, prevent crime and protect local people – as a crucial means of mollifying public concerns. This ideal notion of the local ‘bobby’ has come to embody ‘consensus, community and order’ (Loader, 1997: 15).

Indeed, the 1990s and 2000s saw an increasing emphasis on foot patrols (often as a part of wider community policing strategies) in response to suggestions that placing police officers in patrol cars and focusing on crime reduction and targeted resources had led to an increase in fear of crime, a decline in confidence in the police and a feeling that the police had lost touch with local communities (Herrington and Millie, 2006; Newburn, 2003; Povey, 2001). What evidence, though, is there of a direct effect of police visibility on public confidence? Most studies that have investigated this question have considered the effect of police tactics, rather than raw numbers. A number of evaluations of community policing trials concluded that confidence in the police rose as a result of an increased police presence on the streets (Bennett, 1991; Pate et al., 1986; Skogan and Hartnett, 1997; Tuffin et al., 2006). For example, in 1982 a variety of policing strategies were implemented in Houston and Newark in the USA, including the increased use of foot patrols. Analysing pre- and post-intervention survey data, Pate et al. (1986) showed that, net of the effects of demographic characteristics and perceptions of the police pre-intervention, confidence in the police was higher in areas where the police were regularly patrolling on foot, engaging with and familiarizing themselves with local residents. These findings have also been corroborated by cross-sectional studies, including that by Hawdon and Ryan (2003) who found that the more frequently residents believed that the local police patrolled their neighbourhood, the more effective they thought the police were at controlling crime. Similarly, Skogan (2009) found that individuals who had seen a police officer patrolling their neighbourhood over the previous week were more likely to report that the police were effective at controlling crime, helping victims of crime and keeping order on the streets. Finally, in a study of London residents, Bradford et al. (2009) also found that perceptions of low levels of police visibility were associated with lower confidence in police effectiveness, fairness and engagement with the community.

Although the body of existing research, then, generally supports the idea that both police strength and visibility influence public confidence in the police, it provides no real insight into their relative importance. This is because no empirical study to date has simultaneously included measures of both police numbers and visibility as predictors of public confidence. Our objective in this article is to address this lacuna by linking administrative data to the British Crime Survey at the level of police Basic Command Units (BCU). We proceed to estimate multilevel regression models, in which we include measures of police numbers per 1000 of the population alongside residents’ direct assessments of the extent to which they have noticed police patrolling the streets in the local area.

A good deal of research has explored the correlates of confidence in policing, identifying the importance of demographic factors, attitudinal variables and contact with the police, be that direct or vicarious (for reviews of this literature, please see Bradford et al., 2008; Brown and Benedict, 2002; Decker, 1981). Others have examined changes in confidence over time (Sindall et al., 2012; Bradford, 2011; Jackson et al., 2009), or highlighted the role that the structural characteristics of a neighbourhood can play in the formation of residents’ attitudes, pointing to the prominence of residential instability, socioeconomic disadvantage and crime in the erosion of confidence (Dai and Johnson, 2009; Reisig and Parks, 2000; Sampson and Jeglum-Bartusch, 1998; Schuck et al., 2008). Undoubtedly, such characteristics are crucial to a complete understanding of the ways in which police numbers and behaviour influence public confidence in the way the service operates at both a national and a local level. However, our intention here is not to develop and test a complete or comprehensive causal model of public confidence in policing. Instead our objective is to address the more limited but nonetheless crucially important question of whether, by careful and well-managed deployment of resources, a police service can mitigate the deleterious effects of reductions in police numbers on public confidence in their work.

To foreshadow our later results, we find that both visibility and police strength have a direct effect on public confidence. However, because the visibility of the police is itself a function of strength, police strength also influences confidence indirectly through its effect on visibility. The remainder of the article is set out as follows. First we describe our data, key measures and analysis strategy. We then present our empirical results, before concluding with a consideration of the implications of our findings for how public confidence in policing is likely to be affected by reductions in police numbers.

Data and measures

Data for the analysis are taken from the English component of the 2006/7 and 2007/8 British Crime Survey (BCS). The Welsh part of the survey is excluded because it was not possible to obtain comparable administrative data to derive a measure of recorded crime at the appropriate level for our analysis. The survey employs a multistage sample design: a random sample of postcode sectors is drawn, stratified by police force area, population density and the proportion of adults in non-manual occupations, with selection probability proportional to the size of the sector; 32 addresses are then randomly selected within each sector and one adult aged 16 or over is randomly selected at each address (Bolling et al., 2009). The BCS achieved a response rate of 75 percent in 2006/7 and 77 percent in 2007/8, yielding data from 93,682 respondents for use in our analysis.

Our dependent variable, confidence in the police, was constructed from eight indicators included in the BCS relating to crime fighting, procedural justice and community engagement. Existing research has shown that these represent the key dimensions underpinning what the public understand as the police doing a ‘good job’ (Bradford et al., 2009; Jackson and Bradford, 2009; Stanko and Bradford, 2009). The BCS includes two ‘global’ evaluative questions regarding the performance of the police overall, as well as six questions about specific areas of police activity, conduct and performance (shown in Table 1). These are all measured on a five-point scale ranging from 1 = ‘strongly disagree’ to 5 = ‘strongly agree’, apart from the second global measure, which is measured on a five-point scale from 1 = ‘very poor’ to 5 = ‘excellent’. To create a single measure of confidence, we combined these items using principal components analysis. This produced a single component solution using the criterion of retaining components with an eigenvalue greater than 1. Factor loadings for this model are shown in Table 1.

Table 1.

Principal components analysis: Individual-level factor loadings

Item	Factor loading
Confidence in the police (number of cases = 85,465)
How good a job are the police doing in this area?	.777
The police in this area can be relied on to be there when you need them	.779
The police in this area would treat you with respect if you had contact with them for any reason	.676
The police in this area treat everyone fairly regardless of who they are	.669
The police in this area can be relied on to deal with minor crimes	.739
The police in this area understand the issues that affect this community	.764
The police in this area are dealing with the things that matter to people in this community	.824
Taking everything into account I have confidence in the police in this area	.880
Interviewer assessments of disorder (number of cases = 85,466)
Rubbish or litter lying around	.877
Vandalism, graffiti or deliberate damage to property	.910
Homes in poor condition	.898

For our measures of police strength and police visibility we use survey and administrative data aggregated to the level of BCUs. In England and Wales the police are organized into 43 police forces, each of which is split into a smaller number of BCUs. In 2007/8 there were 228 BCUs in England and Wales. Some BCUs are staffed by over 1000 police officers, while others have fewer than 100 (HMIC, 2011b).¹ The Home Office (2008b) publishes data that give the number of police officers per 100,000 of the population within each BCU between April 2007 and March 2008 and we use this as our measure of police strength.² This measure includes only the number of police officers; it excludes Police Community Support Officers, Special Constables and civilian police staff. There are no administrative sources of information about police visibility across BCUs, so a measure was created from responses to a question in the BCS that asks respondents ‘On average, how often do you see police officers or Police Community Support Officers on foot patrol in your local area?’. Responses to this item were aggregated across respondents within a BCU to produce an estimate of the average level of police visibility within each BCU. Because the visibility question was administered to a random quarter of the full BCS sample, the estimate of the true mean is based on a sample of respondents within each BCU (average size = 202 respondents). The implication of this is that we are likely to underestimate the true effect of visibility on confidence as a result of random error attenuating the coefficient estimate towards zero (Griliches and Hausman, 1986).³ There was a high level of agreement between respondents within BCUs about the level of visibility of the police, with an inter-assessor reliability coefficient of 0.85.⁴ Histograms showing the distribution of police strength and visibility across the 2007/8 BCUs are included in the appendix. They show that there is considerable variability in police numbers (mean = 2.41; standard deviation = 0.44) and police visibility (mean = 182.37; standard deviation = 70.2) across BCUs. These histograms and the analysis presented below do not include data from ‘Bishopsgate’ (an area in the centre of London) and ‘Central Birmingham’ BCUs, because they both have a very large ratio of police officers to residents (10,986 and 2466 per 100,000 residents, respectively). These outliers emerge because there are very few residents in these inner-city areas and not because there is a substantially greater police presence. Consequently, we excluded the 26 respondents living in these BCUs to prevent them from having a disproportionate effect on the results of our analyses.

Of course, residents are not randomly assigned to BCUs and it is likely that some of any difference we observe in confidence in the police will be a function of self-selection into and out of areas. We therefore control for compositional differences between BCUs by including a range of individual- and area-level covariates in our models. At the individual level we control for gender, age, ethnicity, socioeconomic status, marital status and whether or not the individual had been a victim of crime. We also incorporate a measure of visible signs of physical disorder in the immediate area of the respondent’s home (see Table 1 for factors loadings), which is taken from observations recorded by survey interviewers prior to making contact with the household (see Brunton-Smith and Sturgis, 2011). At the BCU level we control for population density, the recorded crime rate and the crime clear-up rate (which the Home Office calls ‘sanction-detections’⁵).⁶ The crime clear-up rate was derived from Home Office statistics (2008a), and the remaining measures are aggregated up from the Lower Layer Super Output Area (LSOA) geography (Martin, 2002; ONS, 2005) to BCU level. The measure of recorded crime was taken from the indices of multiple deprivation (Noble et al., 2008) and the measure of population density from data published by the Office for National Statistics (ONS, 2001). Because there is currently no way of linking the full set of LSOA for England to BCUs, these aggregations are based on the random sample of LSOA (comprising 49 percent of the total number of LSOAs) that were included in the 2006–8 BCS.

Analysis

To evaluate whether variation in police strength and visibility across BCUs is predictive of reported levels of confidence in the police, we use a multilevel model with a random intercept (for a more detailed explanation of multilevel modelling, see Goldstein, 2003; Hox, 2002; Snijders and Bosker, 1999). A multilevel model enables us to account for the non-independence between individuals living within the same BCU and to partition the variability in confidence in the police between areas and individuals. The model has the following form:

\begin{array}{l} y i j = β_{0 j} + β_{1} x_{1 i j} + β_{2} w_{1 j} + β_{3} w_{2 j} + e_{i j} \\ w 1 j = α_{0 j} + α_{1} w_{2 j} + v_{j} \\ β_{0 j} = β_{0} + u_{0 j} \end{array}

(1)

where confidence, y, for the i’th individual living in the j’th BCU is a function of an intercept (β _0j), a regression coefficient (β ₁) for individual i in BCU j for the individual-level covariate x _1ij, a regression coefficient (β ₂) for the BCU-level measure of police visibility, w _1j, a regression coefficient (β ₃) for the BCU-level measure of police strength, w _2j, and an individual-specific error term (e_ij), which is assumed to be normally distributed and with a mean of zero.

The second line of equation 1 shows that the model also includes the regression of police visibility on police strength, with α ₁ denoting the unstandardized regression coefficient and α _0j the BCU-level intercept. This path in the model is included in order to estimate the indirect effect of police numbers on confidence, via its effect on visibility. If police visibility is, as we might assume, a function of numbers, then it is necessary to consider the possibility of such mediated effects (Bollen, 1989). The last line of equation 1 shows that the intercept is allowed to vary across BCUs, with β ₀ denoting the mean level of confidence across all BCUs and u0_j denoting the BCU-specific deviation from the overall mean. Models are estimated using MPlus 5.2 (Muthen and Muthen, 1998–2007).

Results

In a model with no explanatory variables, the variance of the BCU-level random intercept is significantly different from zero (σ_u² = .025; standard error = .003; p < .001), which indicates that just 2.5 percent of the total variability in confidence in the police is attributable to the BCU level.⁷ Thus, it is clear that differences in the way that police forces structure themselves and carry out their work across these organizational units has only a limited effect on the level of confidence expressed by citizens. Nonetheless, it is common in multilevel analyses to find that the large majority of variance in attitudinal and other psychological outcomes is partitioned at the individual level and, although the proportion of the total variance is certainly small, it is nonetheless non-zero and, therefore, merits further analysis to understand its determinants (Hox, 2002).

Table 2 presents the parameter estimates for the full models.⁸ In line with previous research, we find that confidence in the police is significantly higher among older people, women, those in professional/managerial occupations, those who are not married and those who have not been a victim of crime. We also find that the longer a resident has lived in their local area, the less confident they are in the police, as are those who live in streets with evidence of physical disorder. In contrast to the majority of existing evidence, however, we find that ethnic minority groups are more confident in the police than are white people (although see Jackson et al., 2009, for complementary findings in the UK).

Table 2.

The effects of police numbers and police visibility on individual confidence in the police (unstandardized)

	Model 1: The direct effect of police numbers	Model 2: The direct effect of police visibility	Model 3: The direct effect of police numbers and visibility	Model 4: The indirect effect of police numbers
Individual-level effects
Intercept	−.01 (.06)	−.11 (.08)	−.13 (.08)	−.13 (.08)
Age (Ref: 65 years plus)
Aged 15–34	−.07 (.01)*	−.07 (.01)*	−.07 (.01)*	−.07 (.01)*
Aged 35–64	−.12 (.01)*	−.12 (.01)*	−.12 (.01)*	−.12 (.01)*
Gender (Ref: Male)
Female	.13 (.01)*	.12 (.01)*	.13 (.01)*	.13 (.01)*
Ethnicity (Ref: White)
Asian	.32 (.02)*	.33 (.02)*	.32 (.02)*	.32 (.02)*
Black	.21 (.03)*	.20 (.03)*	.21 (.03)*	.21 (.03)*
Mixed/other	.12 (.03)*	.11 (.03)*	.11 (.03)*	.11 (.03)*
NS-SEC (Ref: Professional/managerial)
Intermediate	−.06 (.01)*	−.06 (.01)*	−.06 (.01)*	−.06 (.01)*
Routine and manual	−.03 (.01)*	−.03 (.01)*	−.03 (.01)*	−.03 (.01)*
Unemployed	−.02 (.03)	−.02 (.03)	−.02 (.03)	−.02 (.03)
Student	.03 (.03)	.04 (.03)	.03 (.03)	.03 (.03)
Marital status (Ref: Married)
Single	.02 (.01)*	.02 (.01)*	.02 (.01)*	.02 (.01)*
Widowed	.16 (.01)*	.16 (.01)*	.16 (.01)*	.16 (.01)*
Separated/divorced	.03 (.01)*	.03 (.01)*	.03 (.01)*	.03 (.01)*
Length of residence	−.03 (.002)*	−.03 (.002)*	−.03 (.002)*	−.03 (.002)*
Year (Ref: 2006/7)
Survey sweep	.05 (.01)*	.05 (.01)*	.05 (.01)*	.05 (.01)*
Victimization (Ref: Non-victim)
Victim of crime	−.24 (.01)*	−.24 (.01)*	−.24 (.01)*	−.24 (.01)*
Disorder	−.07 (.004)*	−.07 (.004)*	−.07 (.004)*	−.07 (.004)*
Basic Command Unit effects
Population density	.002 (.001)*	.003 (.001)*	.002 (.001)*	.002 (.001)*
Recorded crime	−.21 (.03)*	−.19 (.03)*	−.21 (.03)*	−.21 (.03)*
Sanction-detections	.001 (.002)	.002 (.002)	.001 (.002)	.001 (.002)
Perceptions of police visibility		.06 (.03)*	.06 (.03)*	.06 (.03)*
Police officers (per 1000)	.05 (.02)*		.04 (.02)*	.04 (.02)*
Indirect effect: Police officers through police visibility				.02 (.01)*
Variance components
BCU level	.014 (.002)*	.014 (.002)*	.014 (.002)*	.014 (.002)*
Individual level	.929 (.01)*	.929 (.01)*	.929 (.01)*	.929 (.01)*
Number of cases	71,736	72,309	71,736	76,995

p = <.05

Our primary interest, however, is not in the individual-level coefficients but in those relating to the characteristics of the BCU. In Model 1 we see that confidence in the police is higher in BCUs with a greater number of police officers per 1000 of the population, an effect that is only slightly attenuated by the addition of police visibility in Model 3. Similarly, Model 2 shows that police visibility is positively and significantly related to confidence: in BCUs where police officers are more frequently seen patrolling the streets, residents report significantly higher levels of confidence in the police. Moreover, the effect of police visibility is robust to the inclusion of police strength in Model 3. Importantly, the effects of police numbers and visibility are apparent when controlling for the level of recorded crime and the sanction-detection rate in the BCU. This excludes the possibility that people are more confident in the police in these areas, not because of the numbers and visibility of officers per se, but as a result of police effectiveness in preventing and solving crime. These results suggest, then, that both the volume of officers employed and the number who are seen by residents patrolling the streets in the area inspire confidence amongst local residents. Model 4 shows that the effect of police strength on confidence is also partially mediated by visibility, as the indirect effect of numbers on confidence is positive and significant. In other words, visibility is itself a function of numbers (β = .32; p = <.001), so that the higher the ratio of police officers to residents in a BCU, the more frequently police officers are seen patrolling in the area. Thus, because visibility promotes confidence, police strength exerts an indirect (as well as a direct) effect on public confidence. The total effect of police strength (the sum of direct and indirect effects) in Model 4 is, therefore, 0.06, which is statistically significant at the 99 percent level of confidence.

With regard to the measures of police effectiveness, perhaps unsurprisingly we find that confidence is lower in BCUs with higher levels of recorded crime. However, variation in the sanction-detection rate across BCUs is not significantly related to confidence in the police. Confidence in the police is higher in BCUs with greater population density, echoing existing research that finds that residents of urban areas are more confident in the police than those living in rural areas (see Myhill and Beak, 2008). Because there is no clear reason to suppose that higher population densities should cause people to have more confidence in the police, we believe this is likely to reflect differences in the demographic composition of areas that are not fully adjusted out via inclusion of covariates.

Discussion

The recent deep cuts to police force budgets in the UK have reawakened longstanding debates about the effect of police numbers and organization on the crime rate and public confidence in the police. Her Majesty’s Inspectorate of Constabularies (HMIC) forecasts that, between 2010 and 2015, police officer numbers will have declined by approximately 16,000 posts (HMIC, 2011a). The government has argued that raw numbers are less important than how the police are organized and deployed; by cutting red tape and focusing staffing reductions on ‘back-office’ functions, the argument goes, it should be possible to maintain a consistent ‘frontline’ presence, reducing crime and maintaining citizen confidence in the service (Cameron, 2011; Home Office, 2010). Although existing evidence in the criminological literature suggests that a reduction in police numbers is likely to lead to an increase in crime (Di Tella and Schargrodsky, 2004; Draca et al., 2011; Machin and Marie, 2011), much less is known about how such reductions are likely to affect public evaluations of the police. But there are good theoretical reasons to believe that a less visible police service will lead to declining public confidence, because confidence emanates from feelings that the police are watching over communities and acting as a symbolic figurehead of the community (Jackson and Bradford, 2009; Jackson and Sunshine, 2007; Jackson et al., 2009). Our objective in this article has been to address this empirical gap in understanding by investigating whether variation in public confidence in policing across police jurisdictions can be explained by variation in police strength and the frequency with which police are observed on foot patrol by local residents. Our analysis provides, for the first time, a direct comparison of the effect of police strength with the effect of police visibility, as assessed by local residents.

An important initial outcome of our analysis was to show that differences in police force organization and practice at the BCU level have only a modest influence on public confidence, with the vast majority of variation in confidence residing in individuals rather than the BCU. This is perhaps unsurprising given the importance that existing studies have attributed to the effect of individual-level factors on confidence, such as perceptions of crime and cohesion and the quality of police contacts with the public (see, for example, Bradford et al., 2009; Jackson and Bradford, 2009; Jackson and Sunshine, 2007; Jackson et al., 2009; Reisig and Parks, 2000; Skogan, 2005; Weitzer and Tuch, 2005). A practical implication of this finding is that, however the police are organized and deployed at this administrative level, public confidence in the police service will not be much affected. Of course, this should not be taken to imply that organizational differences at lower levels of geography will not have a larger effect, only that differences across the current BCU organizational structure are likely to be minimal. For example, it is likely that there is far greater variation in confidence at the level of ‘communities’, areas that will be far more meaningful to residents than much larger units of police organization. Nevertheless, even in the absence of administrative data that would allow analysis at a lower level of aggregation, differences in confidence were evident across BCUs and our models were able to explain some of this variance. For example, we show that confidence is lower in areas where the crime rate is high and that the crime rate affects confidence over and above the sanction-detection rate, implying that it is the prevention, rather than the solving of crimes, that builds confidence. We also show that confidence is higher in areas where there is a higher ratio of police officers to residents and where the police are visible.

Having demonstrated a significant association between police visibility and public confidence, it is important to be clear about exactly what we mean by visibility in this instance. Our measure of visibility is based on a survey question that asks respondents to report how frequently they have seen police and Police Community Support Officers patrolling on foot in their local area. It relates, then, only to instances in which police are seen patrolling on foot and excludes other aspects of visible policing, such as patrolling on bicycles or in cars, which may also have an independent effect on confidence. Be that as it may, our specific focus on foot patrols dovetails felicitously with recent trends in police strategy. Following the example of community policing in the US, foot patrols were increasingly emphasized in the UK during the 1990s and 2000s as being key to the effective implementation of what was termed ‘reassurance policing’ (Innes, 2007; Povey, 2001; Skogan and Hartnett, 1997). This policy shift was introduced, at least in part, as a response to public concerns about safety, which themselves elided with a growing perception that the police had lost touch with local communities and become overly focused on targets, crime reduction and reactive response (Herrington and Millie, 2006; Home Office, 2005; Newburn, 2003). It was envisioned that a shift towards neighbourhood policing would foster stronger relationships between citizens and the police, enhancing public confidence in the service and allowing the police to address problems that were prioritized by local people (Bullock, 2010; HMIC, 2008; Quinton and Morris, 2008). Our findings here lend support to these ideas and confirm the findings of previous studies that have found a positive effect of foot patrol frequency on public confidence (Bennett, 1991; Pate et al., 1986; Skogan and Hartnett, 1997; Tuffin et al., 2006). Our aggregate measure of confidence does not, however, allow us to address whether the effect of visibility is general in nature, or whether its influence is exerted primarily on procedural or competence aspects of confidence evaluations (see Bradford et al., 2009, for evidence of this). The way in which the relationship between visibility and confidence might vary according to how these concepts are defined and operationalized represents a potentially fruitful avenue for future research.

We have also shown that the effect of police numbers on confidence is partially mediated through visibility, because the extent to which the police are seen patrolling the streets on foot is, historically at least, a function of the number of police officers employed in that area. Consequently, our results suggest that substantial cuts in police numbers are likely to reduce public confidence in policing, at least to some extent, even if reductions are concentrated within ‘back-office’ functions. This is because police strength has an independent effect on confidence, over and above the effect of visibility. Although our data provide no leverage on the mechanism(s) underlying the direct effect of numbers that remains after controlling for visibility, we speculate that it arises because, no matter how well implemented, reductions in the number of ‘back-office’ staff are likely to result in frontline officers undertaking important or mandatory administrative functions by default.

But exactly how significant, in substantive rather than statistical terms, is the relationship we have observed between visibility and confidence? Given our very large sample size, it is important not to place too much weight on simple rejection of the null hypothesis. To provide some intuition as to the substantive importance of the effect size, we can compare the magnitude of the visibility coefficient with one with a more direct and tangible interpretation. For example, the estimated difference in confidence between victims and non-victims of crime (in the 12 months prior to interview) in our models is 0.24. This compares with the coefficient of 0.06 for the total effect of one additional police officer per 1000 of the population. Thus, we can infer that it would require an increase of four officers per 1000 of the population in England and Wales to induce a change in confidence that is equal to the difference between victims and non-victims of a criminal offence. Put somewhat differently, if we take the Southampton BCU as an example (2007 population = 228,635), an additional 914 officers would need to be employed to induce an increase in confidence that is equal in magnitude to the difference in confidence between victims and non-victims. To contextualize this still further, between March 2010 and March 2012, the number of police officers in England and Wales fell by 16 officers per 100,000 of the population (Dhani, 2012; Sigurdsson and Dhani, 2010).

Our findings have implications for policing policy because the police are currently substantially reducing the numbers of officers employed in England and Wales in order to meet the current government’s reductions to the police budget. Although the government holds that these reductions can be implemented without damaging the service, by reducing paperwork and bureaucracy and those in back-office roles rather than the number of police patrolling the streets, a recent report found that police forces already had plans to cut police numbers, including frontline officers (HMIC, 2011a). Existing research suggests that reductions in police numbers are likely to lead to an increase in crime. Our findings complement and extend this by showing that reductions in police strength are also likely to erode public confidence in policing. Reduced public confidence in the police would be consequential not only for political reasons but also because, when citizens believe that the police carry out their role effectively and treat people with respect, they are more likely to perceive police power as legitimate and to obey the law and cooperate with the police (Hough et al., 2010; Jackson et al., 2012; Sunshine and Tyler, 2003; Tyler, 2004).

Our conclusions are, as ever, founded on an important ceteris paribus assumption: that the future will be like the past. It is unlikely, but perhaps not impossible, that organizational efficiencies introduced as part of the cost-reduction strategy in UK policing will serve to weaken or even break the independent association we have observed between police strength and public confidence. Unless this is achieved, however, our findings suggest that the proposed reductions in police officer numbers, regardless of whether they are in frontline or back-office roles, will have an adverse effect on the public’s confidence in the police.

Footnotes

Appendix

Notes

References

Australian Federal Police (2008) Portfolio budget statement. URL (accessed 30 January 2012): http://www.ag.gov.au/Publications/Budgets/Budget2008/Pages/Budgetstatements/PortfolioBudgetStatements20082009.aspx.

Bahn

(1974) The reassurance factor in police patrol. Criminology, 12: 338–345.

Bennett

(1991) The effectiveness of a police-initiated fear-reducing strategy. British Journal of Criminology 31(1): 1–14.

Bollen

(1989) Structural Equations with Latent Variables. New York: Wiley.

Bolling

Grant

Donovan

(2009) 2008–09 British Crime Survey Technical Report Volume I. London: Home Office.

Bradford

Jackson

Hough

Farrall

(2008) Trust and confidence in criminal justice: A review of the British research literature. URL (accessed 16 September 2009): http://ssrn.com/abstract=1303567 or https://dx-doi-org.web.bisu.edu.cn/10.2139/ssrn.1303567.

Bradford

Jackson

Stanko

(2009) Contact and confidence: Revisiting the impact of public encounters with the police. Policing and Society 19(1): 20–46.

Bradford

(2011) Convergence, not divergence?: Trends and trajectories in public contact and confidence in the police. British Journal of Criminology 51(1): 179–200.

Brown

Benedict

(2002) Perceptions of the police: Past findings, methodological issues, conceptual issues and police implications. Policing: An International Journal of Police Strategies and Management 23(3): 543–580.

10.

Brunton-Smith

Sturgis

(2011) Do neighbourhoods generate fear of crime? An empirical test using the British Crime Survey. Criminology 49(2): 331–369.

11.

Bullock

(2010) Generating and using community intelligence: The case of neighbourhood policing. International Journal of Police Science and Management 12(1): 1–11.

12.

Cameron

(2011) PM’s speech on the fightback after the riots. URL (accessed 22 August 2011): http://www.number10.gov.uk/news/pms-speech-on-the-fightback-after-the-riots/.

13.

Cameron

(1988) The economics of crime deterrence: A survey of theory and evidence. Kyklos 41: 301–323.

14.

Dai

Johnson

(2009) Is neighbourhood context a confounder? Exploring the effects of citizen race and neighbourhood context on satisfaction with the police. Policing: An International Journal of Police Strategies and Management 32(4): 595–612.

15.

Decker

(1981) Citizen attitudes toward the police: A review of past findings and suggestions for future policy. Journal of Science and Administration 9(1): 80–87.

16.

Dhani

(2012) Police Service Strength: England and Wales, 31 March 2012. London: Home Office.

17.

Dhani

Kaiza

(2011) Police Service Strength: England and Wales, 31 March 2011. London: Home Office.

18.

Di Tella

Schargrodsky

(2004) Do police reduce crime? Estimates using the allocation of police forces after a terrorist attack. American Economic Review 94(1): 115–133.

19.

Draca

Machin

Witt

(2011) Panic on the Streets of London: Police, Crime and the July 2005 Terror Attacks. London: Centre for Economic Performance.

20.

Eck

Maguire

(2000) Have changes in policing reduced violent crime? An assessment of the evidence. In: Blumstein

Wallman

(eds) The Crime Drop in America. Cambridge: Cambridge University Press.

21.

Goldstein

(2003) Multilevel Statistical Models. London: Arnold.

22.

Goldstein

Kounali

Robinson

(2008) Modelling measurement errors and category misclassifications in multilevel models. Statistical Modelling 8(3): 243–261.

23.

Griliches

Hausman

(1986) Errors in variables in panel data. Journal of Econometrics 31(1): 93–118.

24.

Hawdon

Ryan

(2003) Police–resident interactions and satisfaction with police: An empirical test of community policing assertions. Criminal Justice Policy Review 14(1): 55–74.

25.

Herrington

Millie

(2006) Applying reassurance policing: Is it ‘business as usual’? Policing & Society 16(2): 146–163.

26.

HMIC [Her Majesty’s Inspectorate of Constabulary] (2008) Her Majesty’s Inspectorate of Constabulary – Serving Neighbourhoods and Individuals. London: HMIC.

27.

Hox

(2002). Multilevel analysis: Techniques and applications. Mahwah, NJ: Lawrence Erlbaum

28.

HMIC (2011a) Adapting to Austerity. London: HMIC.

29.

HMIC (2011b) Basic Command Unit (BCU) reports. URL (accessed 17 August 2011): http://www.hmic.gov.uk/inspections/basic-command-unit-bcu-reports/.

30.

Home Office (2001) Policing a New Century: A Blueprint for Reform. London: Home Office.

31.

Home Office (2005) Neighbourhood Policing – Your Police; Your Community; Our Commitment. London: Home Office.

32.

Home Office (2008a) Basic Command Unit – Sanction detections for key offences 2006/07 to 2007/08. URL (accessed 22 December 2009): http://www.homeoffice.gov.uk/rds/pdfs08/bcu3-0708.xls.

33.

Home Office (2008b) Police service strength: Supplementary tables. URL (accessed 22 December 2009): http://www.homeoffice.gov.uk/rds/pdfs08/hosb0808add_tab.xls.

34.

Home Office (2010) Policing in the 21st Century: Reconnecting Police and the People. London: Home Office.

35.

Hough

Jackson

Bradford

Myhill

Quinton

(2010) Procedural justice, trust, and institutional legitimacy. Policing: A Journal of Policy and Practice 4(3): 203–210.

36.

Hox

(2002) Multilevel Analysis: Techniques and Applications. London: Lawrence Erlbaum.

37.

Innes

(2007) The reassurance function. Policing: A Journal of Policy and Practice 1(2): 132–141.

38.

Jackson

Bradford

(2009) Crime, policing and social order: On the expressive nature of public confidence in policing. British Journal of Sociology 60(3): 493–521.

39.

Jackson

Sunshine

(2007) Public confidence in policing: A neo-Durkheimian perspective. British Journal of Criminology 47(2): 214–233.

40.

Jackson

Bradford

Hohl

Farrall

(2009) Does the fear of crime erode public confidence in policing? Policing: A Journal of Policy and Practice 3(1): 100–111.

41.

Jackson

Bradford

Hough

Myhill

Quinton

Tyler

(2012) Why do people comply with the law? Legitimacy and the influence of legal institutions. British Journal of Criminology 52(6): 1051–1071.

42.

Kuha

(2011) Group means as explanatory variables in multilevel models. Conference on Social Statistics Honouring the Scientific Contributions of Professor emeritus David Bartholomew, London School of Economics, 12–13 December.

43.

Loader

(1997) Policing and the social: Questions of symbolic power. British Journal of Sociology 48(1): 1–18.

44.

Machin

Marie

(2011) Crime and police resources: The Street Crime Initiative. Journal of the European Economic Association 9(4): 678–701.

45.

Martin

(2002) Geography for the 2001 census in England and Wales. Population Trends 108: 7–15.

46.

Marvell

Moody

(1996) Specification problems, police levels and crime rates. Criminology 34(4): 609–646.

47.

Muthen

(1998–2007) Mplus User’s Guide (5th ed.). Los Angeles, CA: Muthen and Muthen.

48.

Myhill

Beak

(2008) Public Confidence in the Police. London: NPIA.

49.

Newburn

(2003) Policing since 1945. In: Newburn

(ed.) Handbook of Policing. Cullompton: Willan Publishing.

50.

Noble

McLennan

Wilkinson

Whitworth

Barnes

(2008) The English Indices of Deprivation 2007. London: Department for Communities and Local Government.

51.

ONS [Office for National Statistics] (2001) Population density (UV02) NeSS geography hierarchy (United Kingdom). URL (accessed 10 August 2011): http://neighbourhood.statistics.gov.uk/dissemination/instanceSelection.do?JSAllowed=true&Function=&%24ph=60_61&CurrentPageId=61&step=2&datasetFamilyId=789&instanceSelection=13267&Next.x=10&Next.y=11.

52.

ONS (2005) Neighbourhood statistics: Geography policy. URL (accessed 18 June 2011): http://www.neighbourhood.statistics.gov.uk/HTMLDocs/images/GeographyPolicy_tcm97-51009.pdf.

53.

Pate

Wycoff

Skogan

Sherman

(1986) Reducing Fear of Crime in Houston and Newark: A Summary Report. Washington, DC: Police Foundation.

54.

Povey

(2001) Open All Hours: A Thematic Inspection Report on the Role of Police Visibility and Accessibility in Public Reassurance. London: HMIC.

55.

Quinton

Morris

(2008) Neighbourhood Policing: The Impact of Piloting and Early National Implementation. London: Home Office.

56.

Reaves

(2011) Census of State and Local Law Enforcement Agencies, 2008. Washington, DC: US Department of Justice.

57.

Reisig

Parks

(2000) Experience, quality of life, and neighbourhood context: A hierarchical analysis of satisfaction with police. Justice Quarterly 17(3): 607–630.

58.

Sampson

Jeglum-Bartusch

(1998) Legal cynicism and (subcultural?) tolerance of deviance: The neighbourhood context of racial differences. Law & Society Review 32(4): 777–804.

59.

Schafer

Huebner

Bynum

(2003) Citizen perceptions of police services: Race, neighbourhood context, and community policing. Police Quarterly 6(4): 440–468.

60.

Schuck

Rosenbaum

Hawkins

(2008) The influence of race/ethnicity, social class, and neighbourhood context on residents’ attitudes toward the police. Police Quarterly 11(4): 496–519.

61.

Sigurdsson

Dhani

(2010) Police Service Strength: England and Wales, 31 March 2011. London: Home Office.

62.

Sindall

Sturgis

Jennings

(2012) Public confidence in the police: A time-series analysis. British Journal of Criminology 52(4): 744–764.

63.

Skogan

(2005) Citizen satisfaction with police encounters. Police Quarterly 8: 298–321.

64.

Skogan

(2009) Concern about crime and confidence in the police. Police Quarterly 12(3): 301–318.

65.

Skogan

Hartnett

(1997) Community Policing, Chicago Style. Oxford: Oxford University Press.

66.

Snijders

Bosker

(1999) Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling. London: Sage.

67.

Stanko

Bradford

(2009) Beyond measuring ‘how good a job’ police are doing: The MPS model of confidence in policing. Policing: A Journal of Policy and Practice 3(4): 322–330.

68.

Statistics Canada (2011) Police Resources. Ottawa: Statistics Canada.

69.

Sunshine

Tyler

(2003) The role of procedural justice and legitimacy in shaping public support for policing. Law & Society Review 37(3): 513–548.

70.

Tuffin

Morris

Poole

(2006) An Evaluation of the Impact of the National Reassurance Policing Programme. London: Home Office.

71.

Tyler

(2004) Enhancing police legitimacy. Annals of the American Academy of Political and Social Science 593: 84–99.

72.

UNODC [United Nations Office on Drugs and Crime] (2010) Total police personnel at the national level. URL (accessed 26 January 2012): http://www.unodc.org/unodc/en/data-and-analysis/statistics/data.html.

73.

Weitzer

Tuch

(2005) Determinants of public satisfaction with the police. Police Quarterly 8(3): 279–297.