Community Policing and Fear of Crime

Community policing has been introduced in response to debates about the role and operation of police in developed nations and increasing doubts about conventional crime prevention strategies (Adams et al., 2005; Bayley & Shearing, 1996; Kelling et al., 1974; Rosenbaum & Lurigio, 1994; Silverman & Della-Giustina, 2001; Weisburd & Eck, 2004). Community policing aims to improve and increase the levels of public-police interactions (Adams et al., 2005; Innes, 2005; Salmi et al., 2000; Stein & Griffith, 2017; Weisburd & Eck, 2004); this enables police officers to gain knowledge about their citizens and neighbourhoods and insights into their citizens’ perceptions regarding public safety, fear of crime, and perceived risk of victimisation (Hope & Lab, 2001). Over the last few decades, proactive policing strategies have been the focus of many police departments in the US (Crowl, 2017; Mastrofski, 2019; Rosenbaum & Lurigio, 1994; Weisburd & Eck, 2004; Wu & Lum, 2017). The police are now expected to play an important role in organising community-based crime prevention activities because, unlike traditional reactive policing, community policing emphasises proactive approaches (Bayley & Shearing, 1996; Ferguson & Mindel, 2007; Johnston, 2001; Pelfrey, 2004; Scheider et al., 2009; Silverman & Della-Giustina, 2001; Stein & Griffith, 2017) and is based on the belief that the police cannot realise effective services without the public’s support (Schafer et al., 2003).

Fear of crime is defined as ‘an emotional reaction of dread or anxiety to crime or symbols that a person associates with crime’ (Ferraro, 1995, p. 4). Dealing with high levels of fear of crime is a key issue in urban management (Johnston, 2001). Fear of crime has been indeed ‘one of the most researched topics in contemporary criminology’ (Farrall et al., 2000, p. 399), and even ‘a concept of great social and theoretical significance for social scientists to ignore’ (Sutton & Farrall, 2009, p. 117). Increased levels of fear of crime may deteriorate one’s quality of life (Ferraro, 1996; Hale, 1996; Liska et al., 1982; Rader et al., 2007; Skogan, 1986; Skogan & Maxfield, 1981) and be a factor for judging police performance (Jackson et al., 2009). Japan enjoys a relatively low crime rate compared with other developed countries. However, the level of fear of crime among Japanese people has been high and appears to be increasing over time, indicating a negative view of safety (Hamai & Ellis, 2006, 2008). Although speculative, various reasons have been put forward to explain this situation. Media (Hamai & Ellis, 2008), social changes (e.g. population inflow into metropolitan regions and weakened family relationships), and particular incidents which shook the Japanese public, such as the 1995 Tokyo subway sarin attack by the cult movement which killed 12 people and injured more than 5,000 people, and the 2001 Osaka school massacre, in which nearly 30 pupils and teachers were injured or killed at a primary school (Shibata et al., 2017) are proposed as major indicators.

Scholars have conducted empirical studies exploring how individual, household, and neighbourhood characteristics contribute to fear of crime. Theoretical frameworks for fear of crime are roughly divided into three categories: the indirect victimisation model, disorder model, and community concern model (McGarrell et al., 1997; Roh & Oliver, 2005; Taylor & Hale, 1986). The indirect victimisation model considers people’s vulnerability to explain their fear of crime (McGarrell et al., 1997). The disorder and community concern models propose the idea that social and physical environments can affect people’s fear of crime, either positively or negatively. Specifically, the disorder model is concerned with neighbourhood’s disorder as a facilitator of fear, and the community concern model is concerned with neighbourhood’s collective efficacy as an inhibitor of fear, thereby possessing close theoretical links to social disorganisation theory (Farrall et al., 2007).

The effects of community policing on reducing fear of crime can be explained using the community concern model. Dealing with fear of crime is one of the main features of community policing (Adams et al., 2005; Ferguson & Mindel, 2007; Roh & Oliver, 2005; Scheider et al., 2003; Silverman & Della-Giustina, 2001; Stein & Griffith, 2017; Zhao et al., 2002). Community policing is considered effective in reducing the fear of crime because it focuses on increasing police visibility and building strong community partnerships (Scheider et al., 2003). Extant literature on the relationship between community policing and fear of crime has focused on the effects of police officers’ presence on reducing fear of crime, such as increased police deployment, foot patrols, and door-to-door visits (Bennett, 1991; Ferguson & Mindel, 2007; Kelling et al., 1974; Maguire et al., 2019; McGarrell et al., 1997; Montolio & Planells-Struse, 2015; Roh & Oliver, 2005; Skogan, 1994). Moreover, increasing police visibility is considered one of the main aims of community policing (Salmi et al., 2000), and community policing aimed at increasing public-police interactions is especially effective in decreasing fear of crime (Reisig & Parks, 2004; Weisburd & Eck, 2004).

Previous studies of the impact of community policing on fear of crime have yielded mixed results. According to systematic reviews (Crowl, 2017; Gill et al., 2014; Zhao et al., 2002), some studies advocated the use of community policing as it improves the levels of fear of crime, while others found no significant impact of community policing. Community policing has also been reported to increase fear of crime. A possible explanation for this is that because community policing focuses on police-citizen contact, residents encounter information regarding crimes happening in their neighbourhoods, leading to an increased fear of crime (Ferguson & Mindel, 2007). Additionally, as community policing has been defined and implemented in many ways, evaluating its impact can be difficult (Maguire et al., 2019).

Importantly, critical literature exists on the empirical study of fear of crime. Criticism of the quantitative approach to fear of crime has a long history. First, quantitative research on fear of crime can focus on governmental and political realms (Farrall & Lee, 2009; Gadd & Jefferson, 2009; Hollway & Jefferson, 1997; Jackson, 2009; Lee, 2001, 2009). For instance, Lee (2001, 480–481) states that in the US, fear of crime emerged as ‘a legitimate governmental and disciplinary object of calculation, inquiry and regulation’, and quantitative research on fear of crime can be used to ‘justify a tougher approach on crime’ and ‘sensitize citizens to fear once again’. Jackson et al. (2009) argues that politicians use fear of crime to justify the introduction of crime control policies. The second criticism of the quantitative approach concerns the measurement of fear of crime. Several previous studies argue this point (Ditton et al., 2005; Farrall et al., 1997; Garofalo, 1981; Hale, 1996; Jackson, 2005; LaGrange et al., 1992; Lee, 2009; Maguire et al., 2019; Walklate & Mythen, 2007). For instance, fear of crime has been discussed loosely and inconsistently (Hale, 1996; LaGrange et al., 1992). In addition, many quantitative studies on fear of crime have failed to capture this complex phenomenon (Ditton et al., 2005; Farrall & Lee, 2009; Jackson, 2005). Specifically, as fear can be transitory and affected by social conditions or events, respondents find it difficult to define their fear precisely ‘in one neat summary’ (Jackson, 2005, p. 299). Gadd and Jefferson (2009, p. 128) assert that Likert scale responses are not adequate to understand respondents’ fear of crime because their answers are ‘thoroughly decontextualised’. Several empirical studies show that people demonstrate different levels of fear of crime depending on the time of the day, even within the same location (Solymosi et al., 2015; Taylor & Covington, 1993). Therefore, simply asking about fear of crime without distinguishing the time of day may be insufficient to gain a complete picture of respondents’ fear of crime. In addition, although fear of crime and perceived risk of victimisation are different concepts (Ferraro & Grange, 1987; Rountree & Land, 1996), and perceived risk of victimisation is a significant predictor of fear of crime (Ferraro, 1995), some studies do not differentiate between these two concepts. Lee (2009) argues that although reducing fear of crime has been used as a key performance indicator for high-ranking police officers in some countries, it can be problematic because of measurement issues.

Previous studies on police visibility in relation to different policing strategies offer interesting findings. Neighbourhood ministrations and community centres have been introduced to different sites, aiming to decentralise police services and move their officers closer to their citizens (Rosenbaum & Lurigio, 1994). Although our existing knowledge of the impact of neighbourhood ministrations and community centres is limited (Rosenbaum & Lurigio, 1994), some studies in the US have focused on how the introduction of police community stations affects fear of crime. Brown and Wycoff (1987) determined the effects of five community policing strategies (victim recontact, newsletters, citizen contact patrols, police community stations, and community organisation) on peoples’ levels of fear in Houston. They revealed that police community stations and citizen contact patrols were associated with low levels of fear of personal victimisation and worry about property crimes. Furthermore, they concluded that the impact of the police-community station is expected to grow over time as an increasing number of residents will become aware about the station. In extremely high crime-rate Latino and Vietnamese neighbourhoods in the City of Garden Grove, California, a police substation was set up in an apartment complex by the local police department as part of a community policing intervention. Two police officers engaged in foot patrols and meetings with residents. Torres and Vogel (2001) tested the effect of ethnicity on fear of crime. Their study used a resident survey and demonstrated that storefront police substations were effective in reducing residents’ fear of crime, regardless of ethnicity.

Security Box: An Example of Community Policing in Japan

The most common community policing method in Japan is the koban, a type of community police station found throughout the country that is staffed 24 hours a day. The main roles of officers assigned to kobans include responding to incidents and accidents in their jurisdictions, giving directions, handling lost items, giving advice on security and other concerns to those in their jurisdictions, patrolling, and making door-to-door visits. Koban is a core element of the Japanese approach to community policing. Bayley (1978) used the terms ‘formal system’ and ‘informal system’ to explain why and how the koban system works in Japanese society. Specifically, the koban can be regarded as the front line of community policing that plays the role of a formal system. Simultaneously, it realises good public-police relations through extensive day-to-day contact (informal system). Since the introduction of the koban dates to 1874, it played an important role in community policing before it gained popularity in the US and UK.

Chiba Prefecture, located on the eastern side of the Tokyo metropolitan area is the sixth most populous prefecture in Japan. As a unique initiative by the Chiba Prefectural Office, a security box—a community police station similar to the koban—was introduced to Chiba City with a population of approximately 960,000 and Ichikawa City with a population of approximately 470,000 in November 2013. A security box is a community police station staffed by retired police officers (Fig. 1). It is introduced where a new koban is difficult to establish owing to budget and high land prices. The subsidy scheme for security boxes operated by municipalities started in fiscal year 2016. Thus far, 17 security boxes have been established, and 11 security boxes are in operation as of July 2023. The central goal of introducing a security box is to build safer communities through interactions among citizens, municipalities, and police. A security box plays a role in revitalising community-based crime prevention activities (e.g. patrol and neighbourhood watch) through cooperation between security box officers and community volunteers.

Fig. 1
figure 1

Example of a security box. Source: The Mainichi Newspaper <https://mainichi.jp/articles/20161116/k00/00e/040/197000c>

Full size image

Officers stationed in security boxes are referred to as safety advisors. The main roles of safety advisors include advising crime prevention volunteers, patrolling with volunteers, temporal response to various applications and emergency cases, neighbourhood watches, disseminating information regarding crimes and safety (e.g. newsletters), and standing watches (Chiba Prefectural Office, 2019). The differences between the koban and security box are as follows: safety advisors stationed in the security box do not deal with lost property, and they work in shifts from 10:00 to 20:00. Although the ‘Safety Community Zone’, a jurisdiction of the security box, is a 1.5 km radius from each security box, the safety advisors at the security box flexibly participate in the neighbourhood watch outside their jurisdiction, depending on the demand from the communities (Chiba Prefectural Office, 2018).

The impact of security boxes on fear of crime was also tested. Using two waves of a social survey of residents on security boxes, Yamamoto and Shimada (2016) investigated the relationship between the level of knowledge of security boxes and fear of crime. Their multiple logistic regression analysis revealed that higher levels of knowledge of security boxes were associated with lower levels of fear of crime. However, considerable uncertainty remains regarding the long-term impact of introducing a security box. Yamamoto and Shimada (2016) focused on how security boxes affected residents’ fear of crime immediately following their introduction. Therefore, we used three waves of social surveys to investigate the long-term impact of the security box on fear of crime.

The Current Study

Although there have been various studies on the impact of community policing on fear of crime, research examining the effects of community police stations on fear of crime is limited. This study presents an empirical analysis of the differences in fear of crime before, immediately after, and a few years after the implementation of security boxes in Chiba City and Ichikawa City, Chiba Prefecture, Japan. Specifically, using data from three waves of social surveys, this study attempts to bridge this literature gap by assessing the extent to which security box-related factors are related to residents’ fear of crime. Existing studies have examined the impact of community policing in the levels of fear of crime. However, little attention has been paid to how community residents themselves evaluate community policing. Therefore, in addition to the impact of security box on fear of crime, the current study also investigates residents’ attitudes towards security box. This study has the potential to develop new knowledge on how the koban-style box affects residents’ fear of crime and the impact of its existence on the local communities.

Methods

This study investigated the impact of the implementation of security boxes on the levels of fear of crime in one prefecture in Japan using data from a repeated cross-sectional survey—‘community security and safety questionnaire survey’ (SB survey)—collected by the Chiba Prefectural Office, part of which has been used in prior research (Yamamoto & Shimada, 2016). Data were collected in areas where security boxes had been introduced. Self-administered questionnaires were used for data collection. The SB survey was conducted in October 2013 (before introduction), April 2014 (immediately after introduction), and from July to September 2018 (three and a half years after introduction). In all three waves, self-administered surveys were posted to potential respondents. After anonymously completing the survey, the respondents sent it back to the organiser (the Chiba Prefectural Office). In Waves 1 and 2, paper-based questionnaires were mailed to all households in a target area of Chiba City using the last birthday method to randomly select one respondent aged 20 years (the legal age of adulthood in Japan) or older, and 1000 respondents were randomly selected based on the Basic Resident Registration SystemFootnote 1 in Ichikawa City. The same 1000 respondents were sampled in Waves 1 and 2 in Ichikawa City. In Wave 3, paper-based questionnaires were distributed to all households in the target areas of Chiba City and Ichikawa City using the last birthday method. The details of the SB survey are presented in Table 1.

Table 1 Community security and safety questionnaire survey
Full size table

The content of the SB survey varied slightly between each wave. The surveys comprised 48 (Wave 1), 50 (Wave 2), and 20 (Wave 3) questions. These questions included baseline demographic information (e.g. sex and age), housing (e.g. households with children), lifestyle (e.g. frequency of returning home after 11 pm), neighbourhood environment (e.g. perceived neighbourhood disorder), community policing (e.g. frequency of seeing police officers), criminal victimisation, and fear of crime. In the victimisation question, respondents were asked to report their experiences of two broad types of victimisations: direct and indirect. During the three waves, respondents were asked to report their own and their family’s experiences of criminal victimisation in their neighbourhood and whether they saw or heard about criminal victimisation in their neighbourhood in the last 24 months (Wave 1), 6 months (Wave 2), or 12 months (Wave 3). In Waves 2 and 3, there was also a question that inquired about the impact of the security boxes in the respondents’ neighbourhoods (e.g. increased feelings of safety and invasion of privacy).

Variables

To investigate whether the introduction of the security box influenced residents’ fear of crime, this study performed multiple regression analyses. The variables used were operationalised as described below.

Dependent Variable

This study examined the influence of introducing a security box on an outcome variable reflecting citizens’ fear of crime. This ordinal variable measures how often a respondent rates the fear of being a victim of crime on a four-point scale (0 = never; 1 = rarely; 2 = occasionally; 3 = frequently).

Independent Variables

Eight sets of independent variables were included in the analysis: survey wave, demographic characteristics, routine activities, household characteristics, neighbourhood characteristics, victimisation, security boxes, and interaction variables.

Survey Wave

Survey wave was added as a dichotomous variable (Wave 1 was the reference category).

Demographic Characteristics

Various demographic characteristics can help explain the patterns of fear of crime. Female respondents were coded 1 (sex). Age was measured as a series of dummy-coded categorical variables: 20s, 30s, 40s, 50s, 60s, and 70s and over (50s was the reference category).

Routine Activities

Returning home after 11 pm measures how often the respondents return home after 11 pm (0 = never; 1 = occasionally; 2 = frequently). Participation in neighbourhood association measures how often participants participate in neighbourhood associations’ activities (0 = I am not a member of neighbourhood associations; 1 = rarely; 2 = occasionally; 3 = frequently). Going for a walk in the neighbourhood measures how often respondents walk in their neighbourhood (0 = never; 1 = rarely; 2 = occasionally; 3 = frequently).

Household Characteristics

Household members are continuous variables measuring the number of individuals living in a respondent’s household. A dichotomous variable measuring whether the respondent lived with a child under the age of 12 was also included (household with child). The study also assessed whether the respondents lived in their own houses (homeownership). Length of residence, a categorical variable (less than five years, five–nine years, 10–19 years, or > 20 years), was captured as a series of dummy variables with possible values of 1 or 0 (five–nine years was the reference category).

Neighbourhood Characteristics

The SB survey asked the respondents to rate their agreement or disagreement with the following six statements regarding their neighbourhoods: ‘Many abandoned houses and vacant warehouses in my neighbourhood’, ‘Many dark places with few lights in the evening’, ‘Much litter on the roads (e.g. rubbish and cigarettes)’, ‘A lot of graffiti’, ‘Many abandoned motorbikes and bicycles on the streets’, and ‘Many teenagers loitering in the evening’ (0 = strongly disagree; 1 = disagree; 2 = agree; 3 = strongly agree). The items were combined into a mean scale for neighbourhood disorder (Cronbach’s Alpha = 0.669). In the same vein, this study measures collective efficacy, defined as ‘social cohesion among neighbours combined with their willingness to intervene on behalf of the common good’ (Sampson et al., 1997, p. 918). The SB survey asked respondents to indicate their levels of agreement with the statement ‘people in my area can be trusted’ (0 = strongly disagree; 1 = disagree; 2 = agree; 3 = strongly agree).

Victimisation

Two victimisation variables were used in the analysis. The first takes the value of 1 if the respondents themselves, family, and/or acquaintances were victimised in the reference period (direct victimisation). The second takes the value of 1 if respondents saw and/or heard of any crimes happening in their neighbourhoods during the reference period (indirect victimisation).

Security Box

The SB survey asked respondents if they knew about security boxes. Knowledge of security box was measured as a dichotomous variable (yes = 1). The frequency of passing by the security box and seeing safety advisors at the security box was asked in the second and third waves using a four-point scale (0 = never; 1 = rarely; 2 = occasionally; 3 = frequently) (pass by security box and see safety advisor staffed at security box). The SB survey asked respondents regarding the time taken to reach the security box on foot. Proximity to security box was measured on a three-point scale for Waves 2 and 3 (2 = within five minutes; 1 = five–ten minutes; 0 = more than ten minutes). Because the security box had not yet been introduced during Wave 1, all responses in Wave 1 were coded as 0 for this variable.

Interaction Variables

The interaction between Wave 3 survey and the four security box-related variables were added into the regression models to investigate the time-course impact of the security box: wave 3 × knowledge of security box, wave 3 × pass by security box, wave 3 × see safety advisor staffed at security box, wave 3 × proximity to security box.

Analyses

The analyses were performed as follows: First, we provided an overview of the changes in fear of crime and security box-related variables. Second, three regression analyses were performed to analyse the factors associated with fear of crime. Model 1 included survey waves, demographic characteristics, routine activities, household characteristics, neighbourhood characteristics, and victimisation. Model 2 introduced four security box-related variables, and Model 3 considered the interaction variables. Third, the respondents’ perceived impact of introducing the security box was presented. Table 2 presents the descriptive statistics for the variables used in the study. The variance inflation factor was used to examine the possibility of multicollinearity among the independent variables and no signs of multicollinearity were observed.

Table 2 Descriptive statistics
Full size table

Results

Figure 2 shows the changes in the mean levels of fear of crime and the three security box-related variables among respondents over the three waves. Fear of crime showed a slight downward trend between Waves 1 and 2 and a slight increase in Wave 3. Notably, the number of those who knew about the security box increased drastically from Wave 1 to Wave 2. In Wave 3, more than 95% of the respondents knew about security boxes. In addition, those who passed by the security box and saw the safety advisor staff at the security box increased from Waves 2 to 3. The results of one-way ANOVA revealed that there was no statistically significant difference in the mean levels of fear of crime (Welch’ F [2, 3678] = 1.951, p = .142), whereas there was a statistically significant difference in the mean levels of all analysed security box-related variables (knowledge of security box: Welch’ F [2, 3678] = 2563.572, p = .000, pass by security box: Welch’ F [2, 2678] 3914.382, p = .000; see safety advisor staffed at security box: Welch’ F [2, 3678] = 1703.855, p < .001).

Fig. 2
figure 2

Fear of crime, knowledge of security box, pass by security box, and see safety advisor staffed at security box: Wave 1 (2013), Wave 2 (2014), and Wave 3 (2018)

Full size image

Table 3 shows the results of the regression analyses on fear of crime. Model 1 demonstrated that Wave 3 (B = .092 p < .001), being female (B = .178, p < .001), returning home after 11 pm (B = .057, p < .01), length of residence (over twenty years) (B = .076, p < .05), neighbourhood disorder (B = .353, p < .001), and direct victimisation (B = .185, p < .001) were associated with higher levels of fear of crime, whereas collective efficacy (B = −.101, p < .001) was associated with lower levels of fear of crime. Model 2 considered the impact of the introduction of security box; however, it revealed that no security box-related variables were correlated with fear of crime. Specifically, being female (B = .178, p < .001), returning home after 11 pm (B = .057, p < .01), length of residence (over twenty years) (B = .077, p < .05), neighbourhood disorder (B = .353, p < .001), direct victimisation (B = .185, p < .001), and indirect victimisation (B = .173, p < .001) were associated with higher levels of fear of crime, whereas going for a walk in the neighbourhood (B = −.026, p < .05) and collective efficacy (B = −.099, p < .001) were associated with lower levels of fear of crime. Model 3 examined the interaction variables. It showed that neither security box-related variables nor interaction variables were associated with fear of crime. Being female (B = .178, p < .001), returning home after 11 pm (B = .058, p < .010), neighbourhood disorder (B = .353, p < .001), direct victimisation (B = .182, p < .001), and indirect victimisation (B = .171, p < .001) were associated with higher levels of fear of crime, whereas going for a walk in neighbourhood (B = −.026, p < .05) and collective efficacy (B = −.099, p < .001) were associated with lower levels of fear of crime.

Table 3 Regression analyses predicting fear of crime
Full size table

Wave 3 of the SB survey asked respondents about the impact the security box has had on their neighbourhoods. As Fig. 3 shows, about half of the respondents thought that the introduction of the security box improved their perception of safety in the neighbourhood (770 respondents) and safety when walking (663 respondents), whereas a minority of respondents indicated a negative impact of the introduction of the security box.

Fig. 3
figure 3

Perceived impact of security box on neighbourhoods (multiple answers allowed; valid responses = 1596)

Full size image

Discussion

This study sought to address the correlates of fear of crime with a special focus on the impact of the security box. Specifically, it aimed to develop new knowledge on how the koban-style box affects residents’ fear of crime and its raison d’être in local communities. The study was conducted in two phases. First, we examined changes in fear of crime and security box-related variables throughout the three waves of the survey. The levels of fear of crime increased from Wave 1 to Wave 3, whereas the awareness of security box, as seen in the security box-related variables, increased. The increase in fear can be ascribed to the increase in respondents’ experience with security boxes; however, understanding the factors associated with fear of crime is important.

Second, three multiple regression analyses were performed to determine the factors associated with fear of crime. In Model 1, Wave 3 was correlated with higher levels of fear of crime, showing that levels of fear of crime significantly increased from Wave 1 to Wave 3. Model 2 introduced four security box-related variables: knowledge of security box, passing by security box, seeing the safety advisor staffed at the security box, and proximity to the security box. None of these variables were associated with fear of crime. This implies that the increase in fear of crime observed in Wave 3 was not caused by the introduction of the security box but by other variables that were not considered in the current analysis. Model 3 examined the impact of the interaction between Wave 3 and the three security box-related variables on fear of crime. In Wave 3, no interaction variables were associated with fear of crime. This means that there were no time-course changes in the impact of the security box on fear of crime.

Contrary to expectations, the introduction of the security boxes was not associated with fear of crime. There are several possible explanations for this. First, there have been discussions on the measurement of fear of crime (Ditton et al., 2005; Farrall et al., 1997; Garofalo, 1981; Hale, 1996; Jackson, 2005; LaGrange et al., 1992; Lee, 2009; Maguire et al., 2019; Walklate & Mythen, 2007), and the measurements used in the SB survey may not have properly captured respondents’ fear of crime. Second, as high-crime rate neighbourhoods can benefit the most from the introduction of community policing (Skogan, 2019; Stein & Griffith, 2017), low-crime rate neighbourhoods such as Chiba City and Ichikawa City—the research settings for this study—could not have expected a large impact from the introduction of community policing. Third, this study did not examine the indirect effects of security box installations on fear of crime mediated by other variables. Multiple regression analysis revealed that the security box variable had no significant direct effect on fear of crime after controlling for victimisation, perceived disorder, and collective efficacy. However, the implementation of a security box may contribute to the reduction of crime and disorder as well as the enhancement of collective efficacy, followed by a change in fear of crime. Further analysis is required to explore the indirect influence of security boxes on fear of crime through these mediating variables.

In the final model (Model 3), being female, returning home after 11 pm, neighbourhood disorder, direct victimisation, and indirect victimisation were associated with higher levels of fear of crime. As expected from the indirect victimisation model (McGarrell et al., 1997), women feel more fearful than men. This study supports the disorder and indirect victimisation models; neighbourhood disorder, previous victimisation, and indirect victimisation were associated with increased fear of crime. Going for a walk in the neighbourhood and collective efficacy were associated with lower levels of fear of crime. The impact of collective efficacy on reduced levels of fear of crime can be explained by the community concern model (McGarrell et al., 1997; Roh & Oliver, 2005; Taylor & Hale, 1986); those who feel high social trust in their neighbourhoods are likely to show low levels of fear of crime. Interestingly, two routine activity-related variables worked in opposite directions: returning home after 11 pm was associated with increased levels of fear of crime, and going for a walk in the neighbourhood was associated with decreased levels of fear of crime. A possible explanation for this is that although daytime routine activity in neighbourhoods can expose neighbours to the positive aspects of their neighbourhoods, leading to low levels of fear of crime, nighttime routine activity exposes them to situations associated with criminal risk, leading to high levels of fear of crime.

Third, we examined the respondents’ perceived impact of introducing the security box in Wave 3. Many respondents felt that the introduction of the security box improved their feelings about neighbourhood safety and promoted community crime prevention. These outcomes were expected by local municipalities upon the introduction of the security box. Although the results of the regression analyses did not provide empirical support for the effects of the security box on lowering fear, the respondents believed that it was effective in improving neighbourhood safety and community crime prevention. Future studies can use qualitative methodologies; especially, an ethnographic approach can provide further insight into how community members engage with security boxes, and how their perceptions are potentially shaped by wider contextual changes over time.

Conclusion

This study contributes to the literature on the impact of community policing on fear of crime. To the best of our knowledge, this is the first study to empirically investigate the impact of koban-style community policing in Japan on fear of crime. It revealed that demographic and neighbourhood characteristics are predictors of fear of crime; this is consistent with the theoretical frameworks of fear of crime mainly observed in Western countries (McGarrell et al., 1997; Taylor & Hale, 1986). Since the SB survey asked respondents to report fear without distinguishing between crime types; hence, further research should investigate the factors associated with crime by type.

Recommendations for policymaking regarding fear of crime in neighbourhoods are as follows. First, perceived neighbourhood disorder increases fear of crime, thus providing additional support to the disorder model. Therefore, it is suggested that policymakers, police, and communities consider measures to reduce neighbourhood disorder. Second, this study empirically demonstrated that returning home at midnight was associated with higher levels of fear of crime. Therefore, it is important for policymakers to consider how to make sure the safety of those who come home late at night, such as increasing street lightning or public transport service.

Notwithstanding the possible issues related to the measurement of fear of crime discussed above, this study developed a new understanding of the impact of koban-style boxes on residents’ fear of crime. Substantial obstacles remain to the introduction and achievement of community policing. However, communities are expected to play an important role in crime prevention (Schaefer & Mazerolle, 2018). As the security box initiative has been active only in Chiba Prefecture, the results obtained from this study cannot be generalised to other parts of Japan. A recent attempt has empirically demonstrated that crime patterns and crime prevention behaviours vary across the urban-rural continuum (Shimada & Suzuki, 2021). Therefore, future research should investigate the impact of nationwide community policing in Japan.