Using a Rural Index to Assess Crime Risk and Crime Prevention Behavior Across the Urban

Abstract

The study proposes a new method of crime analysis combining data from multiple secondary data sources (census, open crime data, and social survey) to assess the risk of victimization and crime prevention behavior in resource-limited settings. Principal component analysis was performed on municipal-level census data (n = 1,883) to generate a rural index that represents the ecological characteristics of each municipality across the urban–rural continuum. Multilevel logistic analyses were then applied to crime incident data (n = 207,771) to assess the municipal-level effects on victims’ use of locks in motor vehicle and bicycle thefts. A linear pattern of victimization was found for bicycle theft (the risk was about one-thirtieth in the most rural municipalities than that in the most urban municipalities), while the pattern found was nonlinear for motor vehicle thefts. The analysis also revealed that victims in rural areas were less likely to have locked their vehicles before they were stolen than those living in urban areas. Using the rural index developed in this study, police forces can have a better understanding of crime problems in their jurisdiction across the urban–rural continuum. The study discusses the implications of the results for crime prevention and problem-solving policymaking in the urban–rural continuum.

Keywords

rural index rural areas factorial ecology multilevel analysis prevention behavior vehicle theft bicycle theft

Introduction

Investigating patterns of crime victimization across areas is important not only for understanding the determinants of crime but also for the development of crime prevention policies. Although many studies have provided insights into crime victimization in nonurban settings (Giblin et al., 2012; Jobes et al., 2004; Spano & Nagy, 2005), they failed in distinguishing the terms “urban” and “rural” so that the two can empirically be differentiated. Therefore, the notion of rural–urban continuum can be useful in this context because it takes distance from the dichotomy of urban and rural and offers instead a scale representing a rich variety of environments (Planning Tank, 2017). We believe that the term urban–rural continuum is especially important in Japan because the country has experienced both population concentration in metropolitan areas and depopulation in rural areas.

Two issues are raised regarding public safety in rural areas. First, rural municipalities have limited resources available for crime prevention interventions. In the United States and New Zealand, research shows that the police in rural municipalities are underresourced, especially in terms of personnel (Buttle et al., 2010). In Japan, Takahashi (2016) argues that rural areas consist of an ageing population which exacerbates the problem of lack of available workforce and resources to undertake community policing. Therefore, it is necessary to consider how standardized data, such as open crime data and the census, can be utilized in crime prevention initiatives in resource-limited municipalities. Second, due to poor public transport, people in rural areas rely on vehicles more than in urban areas. In Japan, 46.5% of the population uses cars to commute, although this percentage varies from prefecture to prefecture (9.4%–77.6%) (Statistics Bureau of Japan, 2010). In the most remote municipalities where railroad secondary transportation is underdeveloped, bicycles are an important means of transportation for commuters who do not have a driver's license. It is therefore essential to consider factors related to the risk of vehicle-related crimes, especially in the context of the urban–rural continuum.

The present study explores ways to assess the risk of motor vehicle (car and motorcycle) and bicycle theft across the urban–rural continuum as well as their prevention behavior for these particular offences. We first generate a composite index of the rurality of municipalities using census data—a rural index. Using crime data and social survey data, we then discuss variations in both the risk of motor vehicle and bicycle theft and behavior to prevent these crimes across Japan.

Theoretical Background

Theories of Crime and Crime Prevention

Since rural area policing has received less attention from researchers (Mawby, 2004; Schafer et al., 2009), many crime prevention measures have been adopted from the policing experiences in urban areas (Buttle et al., 2010). These urban-based policies are sometimes considered equally compatible with the rural context (Schafer et al., 2009). However, since policing in rural areas can be quite different from that in metropolitan areas due to their different geographical characteristics and the limited number of police officers (Buttle et al., 2010), rural municipalities must plan and implement crime prevention initiatives which fit their contexts.

Problem-oriented policing (POP), a type of policing in which a wide variety of countermeasures are introduced in a narrowly targeted manner (Clarke & Eck, 2003) has been implemented in suburbs and rural settings (Carson & Wellman, 2018). Japan has been using POP since early 2010 for training its police officers in crime prevention in all prefectural police headquarters, including rural areas (Shimada et al., 2015).

In POP, routine activities and prevention behavior are considered important to solve problems. Routine activity theory (RAT) suggested by Cohen and Felson (1979) argues that a crime occurs when three elements converge in time and space: motivated offenders, suitable targets, and the absence of capable guardians. Based on the RAT framework, it can be argued that crime prevention behavior reduce crime opportunities for both motivated offenders and suitable targets. In the United States, the introduction of security precautions (extra locks on doors and a dog in the household) has been found to reduce the risk of vandalism (Tewksbury & Mustaine, 2000). RAT has also been applied to non-Western settings, including East Asia; the type of housing (Sidebottom, 2012) and the presence of a household guardian (Zhang et al., 2007a, 2007b) are associated with property crime risks. However, not many studies have employed RAT yet to detect patterns of criminal victimization in rural areas.

Previous studies have also argued that the levels of crime prevention behavior can differ across the urban–rural continuum; those who live in urban or densely populated areas are more likely to adopt protective behavior than those in nonurban or sparsely populated areas (Giblin et al., 2012; Roth, 2018; Tewksbury & Mustaine, 2003). The low likelihood of prevention behavior in rural settings may be due to their low levels of the perceived risk of victimization (Brunton-Smith & Sturgis, 2011; Franklin et al., 2008). Additionally, high levels of informal social control among neighbors in geographically isolated rural settings can weaken prevention behavior compared to those in urban settings (Giblin et al., 2012). Therefore, crime prevention behavior needs to be discussed across the urban–rural continuum.

The Urban–Rural Continuum

Although research has focused on the difference in crime levels across urban–rural settings, they have used different criteria to define rural communities. Since rural areas can be diverse and have different socioeconomic characteristics (Jobes et al., 2004), it can be problematic to simply categorize areas into “urban” or “rural” based on census classification, or by depending solely on one variable such as population size or density. To solve the problem of defining rurality, we introduce an empirical approach using factorial ecological analysis to arrange municipalities in scale: the urban–rural continuum. The use of principal component analysis, which extracts a small number of factors representing an area's characteristics from a large number of variables, enables us to ensure discriminant validity among multiple factors, as well as reliability within the same factor. The factorial ecological approach has been used to generate an index of urban and rural areas in criminological research in the United Kingdom (Brunton-Smith & Sturgis, 2011) and the United States (Congdon, 2011).

The Current Study

Pre-analysis

First, we employ a factorial ecology approach to the census to generate a municipal-level composite rural index. We then conduct multilevel analyses where the rural index is integrated into individual-level open crime data and the social surveys to examine variations in victimization and prevention behavior across urban and rural communities. The first research question concerns urban–rural differences in the victimization rates of vehicle theft. In the preliminary analysis, we construct an index from census data and combine it with open crime data to examine the following two hypotheses: (1)-1. There is a municipal-level relationship between the risk of motor vehicle theft and the urban–rural continuum. (1)-2. There is a municipal-level relationship between the risk of bicycle theft and the urban–rural continuum. The second research question is whether urban–rural factors are related to the adoption of crime prevention behavior against vehicle theft.

Study 1 and Study 2

Study 1 focuses on crime prevention behavior among the victims of vehicle theft. Study 2 is concerned with crime prevention behavior among the general public, regardless of whether they are victims of crime. By comparing the results of these two studies, we can obtain evidence on how differences in crime prevention behavior in the urban–rural continuum affect potential vehicle theft offenders’ choices.

In Study 1, we investigate the following two hypotheses based on open crime data, in which crime prevention behavior at the time of victimization is recorded. (2)-1. Victims of motor vehicle theft living in rural settings are less likely to adopt prevention behavior. (2)-2. Victims of bicycle theft living in rural settings are less likely to adopt prevention behavior.

The third research question concerns whether urban–rural differences exist in crime prevention behavior against vehicle theft among the general public. Using a nationwide survey, Study 2 investigates the predictors of lock use among the general public to test the following hypotheses: (3)-1. People living in rural settings are less likely to adopt prevention behavior against motor vehicle theft. (3)-2. People living in rural settings are less likely to adopt prevention behavior against bicycle theft.

Study Area

Japan is known for having lower crime rates than other advanced countries (Roberts & Lafree, 2004; Takahashi, 2010). Various explanations have been presented to explain the consistently low crime rate in Japan, including a thriving economy (Komiya, 1999), high levels of informal social control (Bayley, 1978), and police–public cooperation (Tsushima & Hamai, 2015). In Japan, each prefecture has its local government and police department; however, the number of prefectural police department personnel differs from 43,000 in the most urban area to 1,200 in the most rural area (National Police Agency, 2013). Therefore, in Japan, rural police are unable to devote enough human resources or budget for crime analysis, though urban police have established specialized crime analysis units for investigation and policymaking.

The number of police-recorded crimes in Japan started to increase sharply in 1995 (Shimada et al., 2004), followed by a “crime drop” in the early 2000s (Sidebottom et al., 2018; Tseloni et al., 2010). Although the crime drop in Japan started later than it did in Western countries, the crime pattern showed a similar reduction; the number of motor vehicle thefts in Japan decreased by about 70% from 2006 to 2016 (Sidebottom et al., 2018).

Several studies have investigated the patterns of crime in Japan, with a spatial focus on urban–rural differences. Using prefectural-level crime data, Ladbrook (1988) revealed that higher proportions of young people in urban areas were found to increase property crime risk among prefectures. Through a citizen survey and interviews with community leaders in one rural prefecture, Takahashi (2010, 2016) argued that staffing challenges and fewer resources in rural police departments affected the differences in crime rates. These studies have provided the possibility of rural criminology in Japan, a highly populated and industrialized nation, and underlined some similarities with studies on crime and policing in Western countries.

The spatial unit of analysis is the municipality (substate level). Japan consists of 47 prefectures. Each is comprised of municipalities (cities, towns villages, and the special wards of Tokyo). In total, there are 790 cities, 745 towns, and 183 villages in Japan as of 1st January 2015 (Tokyo Metropolitan Government, 2018). Although the municipality is considered macro-level (Battin & Crowl, 2017), it is generally appropriate to be used as the unit of analysis to examine variations in victimization and prevention behavior across the country (Estévez-Soto et al., 2020; Gerell et al., 2020). Since municipal governments and police stations are responsible for controlling crimes, using the municipality as a unit of analysis helps authorities to understand their position in the urban–rural continuum, which is useful for crime prevention planning.

Data and Methods

The present study primarily utilized three nationwide data sets: the 2010 and 2015 census data provided by the Statistics Bureau of Japan, the 2018 open crime data provided by prefectural police departments, and the 2018 community safety survey data conducted by the Nikkoso Research Foundation for Safe Society. The average temperature from the Japan Meteorological Agency is used in the study as a proxy variable that reflects the level of people's outdoor activity (Ratcliffe et al., 2009; Shimada & Motoyama, 2016).

The first dataset, the census, is gathered every 5 years in Japan. It includes variables such as age group, family structure, employment, housing type, and mode of commute. The spatial aggregation units include prefectures, municipalities, and census tracts. In this study, municipally aggregated data were used to construct a rural index.

The second dataset, “open crime data,” is the incident-based record of property crimes, which has been open to the public since 2019 as part of the Japanese government's introduction of evidence-based policymaking. The variables in the dataset include the address, date, and time of the incident, location type (e.g., street, detached house) and age of victims. The use of locking devices for vehicles is also recorded to illustrate the prevention behavior of vehicle-related crime victims. In the open crime data, “unlocked” is defined as the nonuse of present locking devices in vehicles, while the engine key is left either inserted into the slot or on/around the driver's seat, leading to motor vehicle theft.

The last dataset used is the nationwide repeated community safety household survey, the purpose of which is to understand the fear of crime trends and its correlates over time since 2002. The survey taps the use of locks in motor vehicles (while leaving the car or motorcycle with the engine key still attached to the dashboard) and bicycles (not locking them when parked), adopting two-stage probability proportional sampling nationwide. In the 2018 survey, 150 municipalities were sampled in the first stage, followed by 30 households that were randomly sampled from each municipality, using the Basic Resident Registration Network System (response rate: 52.1%, final sample: 1,718).

Results

Preliminary Analysis: The Rural Index at the Municipal Level

Before investigating the urban–rural patterns of crime and prevention behavior, we conducted principal component analysis to generate a municipal-level index that represents the ecological characteristics of rural areas using census data. Of the 1,896 municipalities across Japan, five were excluded from analysis because of the Fukushima Daiichi nuclear power plant accident and seven were excluded due to the 2016 Kumamoto earthquakes. One metropolitan municipality located in central Tokyo was also excluded from the analysis because the principal component score was an outlier. Finally, data from 1,883 municipalities were included in our analysis.

In the first stage, the following seven variables were considered in the analysis: population density, average age, percentage of households in detached houses, housing tenure, the number of shopping centers per 100,000 people, and the ratio of daytime and night-time population. Two factors were extracted with eigenvalues greater than 1.0 (Kaiser–Guttman criterion). Six out of seven variables were loaded on the first principal component (57.1% of the variance), and only the ratio of daytime to night-time was loaded on the second component (16.1% of the variance). Therefore, we reconducted the principal component analysis using six variables in the first component to obtain a composite index of rural areas (rural index). The factor loadings are displayed in Table 1. Figure 1, a choropleth map based on the rural index, shows that highly urbanized areas are concentrated in coastal cities, such as Tokyo and Osaka, while depopulation is progressing in other regions.

Figure 1.

Population density and rural index by municipal.

Table 1.

Principal Component Loadings of Rural Index.

Variables	Loadings
% of households in detached houses	0.886
Housing tenure (having lived in for more than 5 years)	0.862
Population density (log)	−0.835
Average age	0.827
Income per capita	−0.799
No. of shopping centers per 100,000 population	−0.613

The municipalities were listed in the order of the rural index and then classified into 10 strata such that each stratum had the same total population. The 10-point scale was created for the distribution of the population by deciles (Table 2). As the rural index increases, the number of municipalities belonging to this stratum increases, and the population size of municipalities falls. Stratum 10 shows that about 10% of the Japanese population lives within 40% of municipalities, which are relatively small and located in nonurban areas. Figure 2 shows the ratio of daytime to night-time population and the percentage of commuters who use motor vehicles or bicycles. Regarding the daytime to night-time population ratio, only Stratum 1 registered a value of over 100%. This means that in Japan, business and commercial activities are concentrated in municipalities with populations in the top 10%. As the rural index increases, so does the proportion of commuters using motor vehicles, while the proportion of those using bicycles decreases. It can be interpreted that public transport services are poor in municipalities with low population density, so people there travel long distances and have to rely on using motor vehicles and are less likely to use bicycles.

Figure 2.

Ratio of daytime and night-time population and percentage of commuters using motor vehicles and bicycles by rural index stratum.

Table 2.

Descriptive Statistics of Municipal-Level Population by Rural Index.

Rural index	N. of municipalities	N. of population	Median	Min	Max
1. most urban	52	12,647,787	243,227	57,598	903,346
2	61	12,761,935	209,212	18,952	681,298
3	65	12,583,485	193,592	3,022	599,814
4	73	12,526,908	171,601	14,752	535,664
5	95	12,932,190	136,128	178	422,542
6	106	12,545,273	118,352	6,347	406,586
7	118	12,456,952	105,567	335	429,508
8	184	12,622,841	68,602	629	350,237
9	310	12,613,733	40,689	407	275,133
10. most rural	819	12,603,379	15,389	370	129,652
Total	1,883	126,294,483	67,071	178	903,346

The preliminary analysis demonstrated that municipalities can be placed on the urban–rural continuum using indices derived from the census data. Using crime data and social survey data, Study 1 examines the impact of rurality on the relationship between vehicle theft and the use of locks.

Study 1. Crime Prevention Behavior for Vehicles: Open Crime Data Analysis

Study 1 aims to determine if the likelihood of victims using locks in vehicles is different across urban–rural settings. We used the open crime data (23,907 motor vehicle thefts and 183,864 bicycle thefts) in 2018 provided by the police. The dependent variable for Study 1 is whether the stolen motor vehicles and bicycles had been locked by the victims (1 = yes, 0 = no). Our multilevel logistic models include six independent variables (Table 3).

Table 3.

Descriptive Statistics for Study 1.

			Variables	Mean	SD	Min	Max
Dependent
		Bicycle lock use (n = 180,578)		0.40	0.49	0.00	1.00
		Motor vehicle lock use (n = 23,846)		0.44	0.44	0.00	1.00
Independent
	Level 1
		Individual (bicycle only: n = 180,578)
			Age of victim	25.24	17.22	10.00	70.00
		Place (n = 204,424)
			Street	0.14	0.35	0.00	1.00
			Car park	0.42	0.49	0.00	1.00
			Detached house	0.10	0.30	0.00	1.00
			Low-rise multifamily housing	0.10	0.30	0.00	1.00
			High-rise multifamily housing	0.12	0.33	0.00	1.00
	Level 2
		Municipal (n = 1,883)
			Rural index	0.00	1.00	-5.87	2.15
			Ratio of daytime and night-time population	0.98	0.06	0.84	1.69
			Temperature	14.67	2.88	8.90	23.10
			% of commuters using motor vehicle	0.29	0.09	0.01	0.52
			% of commuters using bicycles	0.03	0.02	0.00	0.16

We calculated the number of motor vehicle and bicycle thefts in each stratum, as shown by the vertical bars in Figure 3. Because the population of each stratum is almost the same, the number of crimes is equivalent to the risk of victimization in that stratum. A nonlinear relationship was found between the rural index and the number of motor vehicle thefts. In most densely populated metropolitan areas (Strata 1–3), an increase in the rural index was positively associated with the occurrence of motor vehicle theft. The number of motor vehicle thefts peaked in municipalities belonging to Strata 3–5, which have medium urban intensity. In Strata 6–10, the rural areas where half of Japan's population lives, the increased municipal rurality was associated with a decrease in motor vehicle thefts.

Figure 3.

Number of motor vehicle thefts and ratio of locking device use when stolen, by rural index stratum.

The dark lines in Figure 3 show the percentage of victims’ locking device usage, by stratum, based on open crime data. The proportion of motor vehicle thefts in which the vehicles were locked increased from 71% in Stratum 1 to 79% in Stratum 5 and decreased to 68% in Stratum 9. The percentage of motor vehicle theft in which the doors were locked fell sharply to 51% in most rural municipalities where 10% of the population lives.

On the other hand, bicycle theft exhibited a linear relationship with the rural index. The risk of bicycle theft in the most rural municipalities—Stratum 10 (599 cases)—was about one-thirtieth than that in the most urban municipalities—Stratum 1 (18,132 cases). Bicycles with locking devices accounted for 21% of the bicycle theft cases in Stratum 10, compared to 48% in Stratum 1, and decreased linearly with the increase in the rurality of the municipalities. These findings suggest that in the rural municipalities, where half of the population lives (Strata 6–10), the increase in rurality is linked to a decrease in both motor vehicle and bicycle thefts; however, at the same time, rurality is associated with a weakening of crime prevention behavior, at least among the reported vehicle thefts.

The open crime data also records the type of location (e.g., car park, street, detached house) where the vehicle was stolen. Vehicle thefts occurred in different types of locations, and the ratio of each location varied within the different rural indices (Figure 4). This means that prevention behavior (the use of locks) differ depending on the location, both its type and classification—rural or urban.

Figure 4.

Location of vehicles when stolen by rural index stratum.

Since crime location is nested within municipalities, Study 1 conducted a multilevel logistic regression analysis to examine the factors associated with victims’ use of locking devices for motor vehicles and bicycles when these were stolen. We compared three models to investigate the impact of the type of location and socioeconomic characteristics in municipalities (rural index) as well as the interaction effects of these two factors, separately. Model 1 includes the type of location (car park excluded as reference category), age of victims (only for bicycle theft) as incident-level explanatory variables, and the rural index and temperature as municipal-level explanatory variables. In addition to the variables used in Model 1, Model 2 includes the percentage of commuters using motor vehicles and bicycles, to determine if the results of Model 1 are still robust after controlling for municipal-level motor vehicle and bicycle use. Model 3 includes the variables used in Model 2 and the cross-level interaction variable between the type of location dummy variable and the rural index.

Table 4 shows the results of multilevel logistic regression. Victims’ use of locking devices for both motor vehicles and bicycles are significantly associated with municipal-level explanatory variables including the rural index. Compared to motor vehicle theft in car parks, those which occurred on roads or other locations showed a low likelihood of the vehicles being locked, while those at detached houses showed a high likelihood of locking device use. Additionally, higher rural index levels were associated with a lower likelihood of the vehicles being locked. These patterns are still observed in Model 2, which controlled the municipal-level ease of motor vehicle and bicycle use. Model 3 showed cross-level interaction effects between the rural index and detached houses as well as other places such as commercial facilities.

Table 6.

Multilevel Logistic Regression Models Predicting Use of Locking Devices in Case of Motor Vehicle Theft Among the General Public.

	Motor vehicle theft				Bicycle theft
	Model 1		Model 2		Model 1		Model 2
	OR	SE	OR	SE	OR	SE	OR	SE
Level 1: Individual
Gender (female = 1)	1.34	0.23^†	1.33	0.23^†	1.16	0.20	1.17	0.20
Age	1.17	0.07^*	1.17	0.07^*	0.97	0.06	0.97	0.06
Previous victimization: motor vehicle theft	2.01	2.17	2.03	2.19	0.50	0.20^†	0.49	0.20^†
Single-person household	13.84	14.07^*	13.92	14.16^*	0.87	0.28	0.83	0.27
Income	0.98	0.06	0.98	0.06	0.99	0.06	0.97	0.06
Employment status: full-time	0.60	0.14^*	0.60	0.14^*	1.25	0.27	1.25	0.27
Employment status: part-time and student	0.94	0.28	0.94	0.28	1.44	0.39	1.46	0.39
Level 2: Municipality
Rural index	0.71	0.07^^	0.74	0.12^†	0.97	0.10	0.90	0.10
Ratio of daytime and night-time population	0.17	0.30	0.22	0.43	0.24	0.32	0.19	0.25
Temperature	1.01	0.04	1.01	0.04	1.01	0.05	1.03	0.05
% of commuters using the vehicle			0.56	0.75			0.00	0.01^†
Variance components
Municipality	0.00	0.00	0.00	0.00	0.21	0.14	0.18	0.14
Log likelihood	472.27		472.17		50.02		498.40

Note. ^*p < .05, ^*^*p < .01, ^†p < .10.

Model 1 for bicycles shows that higher levels of the rural index were correlated with lower levels of locking device use. Victims who experienced bicycle theft on streets, from detached houses, and other places showed low levels of locking device use, while those in high-rise multifamily housing demonstrated high levels of locking device use. These patterns were observed even after controlling for the municipal-level target availability. Moreover, the cross-level interaction effect on the usage of locking devices among victims was statistically significant between the rural index and the type of parking locations such as multi-family housing and commerial facilities.

Study 2. Prevention Behavior: Survey Data Analysis

Study 2 used nationally representative survey data to investigate prevention behavior with respect to motor vehicle and bicycle theft among the general public, regardless of their victimization. The survey asked the respondents to report whether they intentionally left their vehicles unlocked in the past month. The answers were reverse-coded to be used as the dependent variable for Study 2; those who locked their vehicles were coded as 1 and those who did not lock their vehicles intentionally were coded as 0. The independent variables include gender, age, number of family members, household income, employment status, previous victimization, temperature, rural index, and the percentage of commuters using bicycles (Table 5). As with Study 1, the percentage of commuters using motor vehicles and bicycles was added to the analysis in Model 2. The number of respondents for motor vehicle theft was 1,018 and for bicycle theft was 858, who own motor vehicles and/or bicycles.

Table 5.

Descriptive Statistics for Study 2.

		Variables	Mean	SD	Min	Max
Dependent
		Motor vehicle lock use (n = 1,018)	0.81	0.40	0.00	1.00
		Bicycle lock use(n = 858)	0.72	0.45	0.00	1.00
Independent
	Level 1 (n = 1,152)
		Gender (female = 1)	0.50	0.50	0.00	1.00
		Age	3.90	1.54	1.00	6.00
		Previous victimization: bicycle theft	0.01	0.09	0.00	1.00
		Previous victimization: motor vehicle theft	0.03	0.17	0.00	1.00
		Single-person household	0.07	0.26	0.00	1.00
		Income	2.97	1.35	1.00	6.00
		Employment status: full-time	0.53	0.50	0.00	1.00
		Employment status: part-time and student	0.16	0.37	0.00	1.00
	Level 2 (n = 149)
		Rural index	−0.79	1.00	−5.87	1.48
		Ratio of daytime and night-time population	1.99	0.07	1.88	2.59
		Temperature	15.26	2.18	8.90	23.10
		% of commuters using	0.05	0.03	0.01	0.15

Multilevel logistic regression was conducted to predict the use of locking devices by the general public in cases of stolen vehicles (Table 4). For motor vehicles, a higher rural index at the municipal level reduced the likelihood of using locking devices among the respondents. Age increased the likelihood of using locking devices, and full-time workers were less likely to use locks than part-time workers and students. The significant impact of these variables on prevention behavior was confirmed even after controlling for the percentage of commuters using bicycles. For bicycles, only previous victimization was significantly associated with the use of locking devices, as shown in Model 1. Model 2 demonstrated that the percentage of commuters using bicycles was correlated with low levels of lock use at the municipal level.

Discussion

Although less attention has been paid to crime trends in nonurban areas, it is important to understand the factors associated with the risk of victimization and prevention behavior across the urban–rural continuum, for crime prevention and policymaking. The present study employed open crime data and a nationwide survey to empirically investigate variations in the risk of vehicle thefts as well as prevention behavior between urban and rural settings.

To generate a municipal-level index that represents the ecological characteristics of rural areas, we conducted a principal component analysis. Rural areas were more likely to show lower levels of bicycle theft than urban areas (Figure 3). Regarding the first research question on the relationship between the victimization of vehicle theft and the municipal-level rural index, Hypothesis (1)-2 was supported. This finding is consistent with a country-wide study reporting that the urban–rural dichotomized variable predicted the levels of crime (Ceccato & Dolmen, 2011).

On the other hand, the relationship between the rural index and motor vehicle theft was nonlinear, with the highest risk of motor vehicle theft occurring in moderately rural municipalities. Therefore, Hypothesis (1)-1 was not supported. Two different causal mechanisms can be considered to explain this nonlinear relationship: the surveillance of parking locations and the cost of target searching, as RAT argues. In densely populated urban areas, it is more difficult to commit vehicle theft because of the high levels of surveillance at car parks. In addition, people in urban areas are less dependent on cars for transport, which, as a result, makes it more difficult for potential offenders to search for suitable targets (Figure 2). On the other hand, in rural areas, offenders’ risk of being detected during identification and breaking of vehicle locks is considered lower than in urban areas because the parking places are normally away from facilities. Meanwhile, finding suitable targets—parked motor vehicles—is costlier in rural areas, owing to the sparse population density. Thus, because of these two compelling mechanisms, the risk of motor vehicle theft is the highest in Strata 3–5 (i.e., moderately rural municipalities).

Using the open crime data, Study 1 determined the predictors of the use of locking devices among vehicle theft victims. The results of the multilevel logistic regression analysis showed that the victims of vehicle theft in rural areas were less likely to have kept their vehicle locked than those in urban areas (Table 6). Therefore, Hypotheses (2)-1 and (2)-2 regarding the relationship between rurality and prevention behavior for vehicle theft were supported. The theft of locked vehicles can be interpreted as a consequence of the offender's two different choices: to search for a vehicle without a lock or to break the lock. This contextual effect on offenders’ choices was statistically significant even after controlling for target availability and the cross-level interaction between the rural index and the parking locations in Models 2 and 3. It is noteworthy that there was a robust relationship between the urban–rural continuum and the likelihood of locking devices for vehicles among victims.

Table 4.

Multilevel Logistic Regression Models Predicting Victims’ Use of Locking Devices in Case of Vehicle Thefts.

		Motor vehicle theft						Bicycle theft
		Model 1		Model 2		Model 3		Model 1		Model 2		Model 3
		OR	SE	OR	SE	OR	SE	OR	SE	OR	SE	OR	SE
Level 1: Incident
	Age of victim							1.01	0.00^^^*	1.01	0.00	1.01	0.00^^^*
	Place: (reference: car park)
	Street	0.54	0.03^^^*	0.54	0.03^^^*	0.54	0.06^^^*	0.88	0.01^^^*	0.88	0.01^^^*	0.90	0.03^^
	Detached house	1.14	0.06^*	1.14	0.06^*	0.92	0.07	0.37	0.01^^^*	0.37	0.01^^^*	0.41	0.02^^^*
	Low-rise multifamily housing	2.11	0.13^^^*	2.11	0.13^^^*	2.29	0.27^^^*	1.01	0.02	1.01	0.02^^^*	1.17	0.04^^^*
	High-rise multifamily housing	2.06	0.13^^^*	2.07	0.13^^^*	2.55	0.37^^^*	1.15	0.02^^^*	1.15	0.02	1.44	0.06
	Other	0.62	0.03^^^*	0.62	0.03^^^*	0.84	0.06^*	0.71	0.01^^^*	0.71	0.01^^^*	0.73	0.02^^^*
Level 2: Municipality
	Rural index	0.69	0.03^^^*	0.62	0.04^^^*	0.59	0.04^^^*	0.66	0.01	0.67	0.01^^^*	0.65	0.01^^^*
	Ratio of daytime andnight-time population	0.13	0.06^^^*	0.10	0.05^^^*	0.10	0.05^^^*	0.55	^* ^* ^*	0.55	0.11^^^*	0.58	0.11^^
	Temperature	1.02	0.02	1.02	0.02	1.02	0.02	0.98	0.11^^	0.98	0.01^†	0.98	0.01^*
	% of commuters using motor vehicle			2.84	1.45^*	2.82	1.43^*
	% of commuters using bicycles									3.15	1.85^^	2.77	1.62^†
Cross-level interaction (rural index × )
	Place: (reference: car park)
	Street					1.01	0.07
	Detached house					0.77	0.05^^^*					1.02	0.02
	Low-rise multifamily housing					1.07	0.10					1.08	0.03^^
	High-rise multifamily housing					1.16	0.10 ^†					1.11	0.02^^^*
	Other					1.43	0.09^^^*					1.14	0.02^^^*
Variance components
	Municipality	0.64	0.06	0.64	0.06	0.62	0.05	0.14	0.01	0.14	0.01	0.13	0.01
	Log likelihood	−12,820.94		−12,818.84		−12,780.89		−115,746.00		−115,744.41		−115,713.51

Note. ^***p < .001, ^*^*p < .01, ^*p < .05, ^†p < .10.

Study 1 also revealed the cross-level interactions between the location of the incident and the degree of municipal-level rurality, as shown in Model 3 in Table 6. The cross-level interactions show that the ratio of locked vehicles in cases of vehicle thefts occurring at detached houses in rural areas shows a higher decrease than the cases in urban areas. Further, the ratio of detached houses to the total number of motor vehicle thefts increases as the rural index increases (Figure 4). This suggests that not using locking devices for motor vehicles at detached houses in rural areas is a problem in POP, as this behavior can elevate victimization risks. Similarly, another cross-level interaction between rurality and motor vehicle theft at commercial properties was observed; more victims of motor vehicle theft at commercial properties in rural areas had kept their vehicles locked when they were stolen than those in urban areas. In addition, in rural areas, commercial properties comprise a higher ratio of motor vehicle theft than in urban areas. This result indicates the risk of motor vehicle theft at commercial properties in rural areas, regardless of the use of locking devices. From the POP approach, an intervention to increase the use of locking devices for motor vehicles, especially when parked at detached houses, and the introduction of CCTV to parking at commercial properties can be suggested to solve these problems.

Bicycle theft occurring at bicycle parks in rural settings is a problem both in terms of their high ratio of victimization locations and the high rate of unlocked bicycles. It is noteworthy that these cross-level effects were observed even after controlling for rurality and theft location; the low likelihood of the use of locks in certain types of locations can be identified as a “problem” in POP in rural areas.

Using data from a nationwide survey, Study 2 aimed to examine crime prevention behavior among the general public. Municipality-level rurality had a negative contextual impact on the locking of motor vehicles, but not a significant one on the locking of bicycles (Table 4). Consequently, with regard to the third research question, only Hypothesis (3)-1 was supported. Rurality has been found to have a negative impact on crime prevention behavior against vehicle theft (Giblin et al., 2012), residential burglary (Roth, 2018), and assault (Tewksbury & Mustaine, 2003). This mixed result can be attributed to the difference in the monetary value of motor vehicle and bicycle. The greater the loss due to victimization, the more likely people are to adopt prevention behavior in response to the perceived risk of victimization. Therefore, a systematic association between crime prevention behavior and rurality is expected to emerge. Indeed, the difference in the levels of the perceived risk of victimization in urban and rural areas has been confirmed in previous research (Brunton-Smith & Sturgis, 2011; Franklin et al., 2008).

With regard to bicycle theft, only previous victimization was found to be negatively associated with using locking devices. Since the current study adopted a cross-sectional survey, we were unable to infer a causal relationship between previous victimization and not using a locking device; however, we can at least argue from the result that prevention behavior for bicycles depends on individual-level factors rather than municipal-level factors, including rurality.

The results of Study 1 and Study 2 showed that the use of locks for motor vehicles was lower in rural areas than in urban areas, both among victims and the general public. By contrast, regarding bicycle theft, the use of locks was lower in rural areas than in urban areas, even though there was no significant difference in its usage among the general public. This might be influenced by the offenders’ motivations and the availability of suitable targets. Offenders of motor vehicle theft selectively search for specific models of cars, as they steal certain popular models to sell (General Insurance Association of Japan, 2020). For this reason, this is consistent with the problem found in Study 1: offenders of motor vehicle theft discretely search for their suitable targets at detached houses in rural areas. Conversely, bicycle theft offenders are likely to steal unlocked bicycles as a means of transport. From the POP perspective, the results of Study 2 provide additional support for the problem of unlocked bicycles in bicycle parks, which we pointed out in Study 1.

Conclusions

The study empirically demonstrated the following three points. First, high levels of rurality (based on rural index) were associated with the risk of motor vehicle and bicycle thefts at the municipal level. Second, municipal-level rurality had a negative contextual effect on the use of locking devices among victims of motor vehicle and bicycle theft, interacting with the type of parking location. Third, municipal-level rurality also had a negative contextual effect on the use of locking devices for motor vehicles among the general public.

Findings have important implications for crime prevention in rural settings, in both Japan and elsewhere. First, the present study validated the usefulness of open crime data as an important resource for crime research and policymaking in both urban and rural areas. As far as the authors are aware, the United Kingdom (England and Wales) is the only country where unified nationwide crime-incident data have been made open to the public. The newly released open crime data across Japan enable researchers and policymakers to analyze crime across the nations’ urban–rural continuum. Japan has experienced both population concentration in metropolitan areas and depopulation in rural areas. Even in rural areas with limited resources, policymakers can now access crime data for crime prevention planning, allowing policies to reflect the crime patterns and trends of each area. Our findings demonstrated that a composite index of rural areas combined with analysis of census data and open crime data are all useful for crime prevention interventions in rural settings. In the scanning process of POP, Japanese police have used crime and other population-specific statistics to examine the risk of victimization by jurisdiction, failing to fully understand the jurisdictional context, which may affect spatio-temporal crime trends. Using the rural index developed in this study, police forces can have a better understanding of crime problems in their jurisdiction across the urban–rural continuum.

Second, the present study marks the first step towards enhancing our understanding of variations in motor vehicle and bicycle theft across the urban–rural continuum. Studies 1 and 2 showed that motor vehicle theft victims and the general public in rural areas were less likely to engage in prevention behavior (use of locking devices) than those in urban areas. This leads to our conclusion that although in rural areas the risk of motor vehicle theft itself is lower than that in urban areas, not using locking devices for motor vehicles in rural areas elevates the relative risk of victimization more than in urban areas.

Similarly, this study examined the interaction effect between urban and rural areas and the victims’ use of locks at the locations where thefts occurred. This interaction effect provides indications for possible countermeasures for vehicle theft prevention in rural areas. Specifically, since unlocked vehicles account for a high ratio of stolen motor vehicles in detached houses, educational interventions to promote the use of locks are required for those who park their vehicles at their homes. In contrast, in commercial facilities, where a majority of thefts occurred by breaking of locks, managers should be recommended to install CCTV to prevent motor vehicle theft, rather than educating vehicle users about the facilities. In addition, since a majority of thefts in rural areas occurred at bicycle parking lots, and no use of locks statistically increased the risk of victimization, awareness campaigns should be conducted to encourage the users of bicycle parking lots to lock their bicycles.

The present study has some limitations. First, we focused only on motor vehicle and bicycle thefts recorded using open crime data. Much research on the victimization patterns of other types of crime in other countries has been carried out, such as larceny (Mustaine & Tewksbury, 1998; Spano & Nagy, 2005), burglary (Jobes et al., 2004), violent crime (Spano & Nagy, 2005), or vandalism (Tewksbury & Mustaine, 2000). Future work should examine prevention behavior for other types of crime across the urban–rural continuum. The second limitation relates to the sampling design of the social survey used in this study. Since the nationwide survey used in Study 2 adopted two-stage probability proportional sampling, those in low-density areas might be undersampled. Additionally, the sample size was not large enough to determine respondents’ risks of motor vehicle and bicycle theft. Conducting a large-scale victimization survey is therefore recommended as it will enable to examine the interaction effects between individual-level prevention behavior and area-level ecological characteristics.

The most remote areas in many countries are facing demographic challenges problems due to population ageing and people moving to urban centers. Municipalities with scarce human and financial resources need to develop crime and safety methods and policies that suit their conditions and contexts. Despite the limitations, the proposed method in this study is a contribution to the toolbox that helps identify places on the urban–rural continuum that demand more attention by the police forces and make crime prevention more efficient in resource-limited settings.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Takahito Shimada

Ai Suzuki

Author Biograpies

Takahito Shimada is a managing social scientist of Crime Prevention Section, Department of Criminology and Behavioral Sciences at National Research Institute of Police Science of Japan. His research interests involve studies on community crime prevention, risk communication, intimate partner violence, fear of crime, and spatio-temporal crime analysis.

Ai Suzuki is a PhD candidate in the Department of Security and Crime Science at University College London. Her current research examines the patterns and predictors of crime and fear of crime in Japan.

References

Battin

J. R.

Crowl

J. N

. (2017). Urban sprawl, population density, and crime: An examination of contemporary migration trends and crime in suburban and rural neighborhoods. Crime Prevention and Community Safety, 19(2), 136–150. https://doi.org/10.1057/s41300-017-0020-9

Bayley

D. H

. (1978). The future of social control in Japan. Pacific Affairs, 51(1), 24. https://doi.org/10.2307/2757007

Brunton-Smith

Sturgis

. (2011). Do neighborhoods generate fear of crime? An empirical test using The British Crime Survey. Criminology, 49(2), 331–369. https://doi.org/10.1111/j.1745-9125.2011.00228.x

Buttle

Fowler

Williams

M. W

. (2010). The impact of rural policing on the private lives of New Zealand police officers. International Journal of Police Science & Management, 12(4), 596–606. https://doi.org/10.1350/ijps.2010.12.4.202

Carson

J. V.

Wellman

A. P

. (2018). Problem-oriented policing in suburban low-income housing: A quasi-experiment. Police Quarterly, 21(2), 139–170. https://doi.org/10.1177/1098611117744005

Ceccato

Dolmen

. (2011). Crime in rural Sweden. Applied Geography, 31(1), 119–135. https://doi.org/10.1016/j.apgeog.2010.03.002

Clarke

R. V

Eck

J. E.

(2003). Become a problem-solving crime analyst in 55 small steps. Jill Dando Institute of Crime Science, University College London

Cohen

L. E.

Felson

. (1979). Social change and crime rate trends: A routine activity. American Sociological Review, 44(4), 588–608. https://doi.org/10.2307/2094589

Congdon

. (2011). The spatial pattern of suicide in the US in relation to deprivation, fragmentation and rurality. Urban Studies, 48(10), 2101–2122. https://doi.org/10.1177/0042098010380961

10.

Estévez-Soto

P. R.

Johnson

S. D.

Tilley

(2020). Are repeatedly extorted businesses different? A multilevel hurdle model of extortion victimization. Journal of Quantitative Criminology. https://doi.org/10.1007/s10940-020-09480-8

11.

Franklin

T. W.

Franklin

C. A.

Fearn

N. E

. (2008). A multilevel analysis of the vulnerability, disorder, and social integration models of fear of crime. Social Justice Research, 21(2), 204–227. https://doi.org/10.1007/s11211-008-0069-9

12.

General Insurance Association of Japan. (2020). Survey on motor vehicle theft. Retrieved from July 14, 2021, https://www.sonpo.or.jp/news/release/2020/2005_01.html

13.

Gerell

Kardell

Kindgren

. (2020). Minor Covid-19 association with crime in Sweden. Crime Science, 9(19). https://doi.org/10.1186/s40163-020-00128-3

14.

Giblin

M. J.

Burruss

G. W.

Corsaro

Schafer

J. A

. (2012). Self-protection in rural America: A risk interpretation model of household protective measures. Criminal Justice Policy Review, 23(4), 493–517. https://doi.org/10.1177/0887403411421215

15.

Jobes

P. C.

Barclay

Weinand

Donnermeyer

J. F

. (2004). A structural analysis of social disorganisation and crime in rural communities in Australia. Australian and New Zealand Journal of Criminology, 37(1), 114–140. https://doi.org/10.1375/acri.37.1.114

16.

Komiya

. (1999). A cultural study of the low crime rate in Japan. British Journal of Criminology, 39(3), 369–390. https://doi.org/10.1093/bjc/39.3.369

17.

Ladbrook

D. A

. (1988). Why are crime rates higher in urban than in rural areas?: evidence from Japan. Australian & New Zealand Journal of Criminology, 21(2), 81–103. https://doi.org/10.1177/000486588802100203

18.

Mawby

R. I

. (2004). Myth and reality in rural policing: Perceptions of the police in a rural county of England. Policing, 27(3), 431–446. https://doi.org/10.1108/13639510410553158

19.

Mustaine

E. E.

Tewksbury

. (1998). Predicting risks of larceny theft victimization: A routine activity analysis using refined lifestyle measures. Criminology, 36(4), 829–858. https://doi.org/10.1111/j.1745-9125.1998.tb01267.x

20.

National Police Agency. (2013). The white paper on police 2013. Retrieved from July 14, 2021, https://www.npa.go.jp/hakusyo/h25/index.html

21.

Planning Tank. (2017). Rural-urban continuum and causes of rural-urban continuum. Retrieved from July 14, 2021, https://planningtank.com/settlement-geography/rural-urban-continuum.

22.

Ratcliffe

J. H.

Taniguchi

Taylor

R. B

. (2009). The crime reduction effects of public CCTV cameras: A multi-method spatial approach. Justice Quarterly, 26(4), 746–770. https://doi.org/10.1080/07418820902873852

23.

Roberts

Lafree

. (2004). Explaining Japan’s postwar violent Crime trends. Criminology, 42(1), 179–209. https://doi.org/10.1111/j.1745-9125.2004.tb00517.x

24.

Roth

J. J

. (2018). The role of perceived effectiveness in home security choices. Security Journal, 31(3), 708–725. https://doi.org/10.1057/s41284-017-0125-y

25.

Schafer

J. A.

Burruss

G. W.

Giblin

M. J

. (2009). Measuring homeland Security innovation in small municipal agencies: Policing in a post-9/11 world. Police Quarterly, 12(3), 263–288. https://doi.org/10.1177/1098611109339891

26.

Shimada

Motoyama

. (2016). Leaving home without locking the front door and windows among young-single-women living alone in apartment buildings. MERA Journal, 19(1), 11–20. (In Japanese with English abstract).

27.

Shimada

Saito

Harada

(2015). An attempt of strategic crime analysis training for police officer. Reports of National Research Institute of Police Science, 64(1), 24–34. (In Japanese with English abstract).

28.

Shimada

Suzuki

Harada

. (2004). Fear of crime and perceived risk: The causes and structures. Japanese Association of Sociological Criminology, 29, 51–64. (In Japanese with English abstract).

29.

Sidebottom

. (2012). Repeat burglary victimization in Malawi and the influence of housing type and area-level affluence. Security Journal, 25(3), 265–281. https://doi.org/10.1057/sj.2011.22

30.

Sidebottom

Kuo

Mori

Farrell

. (2018). The East Asian crime drop? Crime Science, 7(1). https://doi.org/10.1186/s40163-018-0080-x

31.

Spano

Nagy

. (2005). Social guardianship and social isolation: An application and extension of lifestyle/routine activities theory to rural adolescents. Rural Sociology, 70(3), 414–437. https://doi.org/10.1526/0036011054831189

32.

Statistics Bureau of Japan. (2010). The 2010 census. Retrieved from July 14, 2021, https://www.stat.go.jp/data/kokusei/2010/.

33.

Takahashi

(2010). Creating a crime-resistant society in Japan: Maintaining public safety with community resources. Japan Studies Association Journal, 179–196.

34.

Takahashi

(2016). Crime and response in rural Japan. In Donnermeyer

J. F.

(Ed.), The Routledge International Handbook of Rural Criminology (pp. 45–54). Routledge.

35.

Tewksbury

Mustaine

E. E

. (2000). Routine activities and vandalism: A theoretical and empirical study. Journal of Crime and Justice, 23(1), 81–110. https://doi.org/10.1080/0735648X.2000.9721111

36.

Tewksbury

Mustaine

E. E

. (2003). College students’ lifestyles and self-protective behaviors: Further considerations of the guardianship concept in routine activity theory. Criminal Justice and Behavior, 30(3), 302–327. https://doi.org/10.1177/0093854803030003003

37.

Tokyo Metropolitan Government. (2018). The structure of the Tolyo metropolitan government (TMG). Retrieved from July 14, 2021, https://www.metro.tokyo.lg.jp/ENGLISH/ABOUT/STRUCTURE/structure01.htm.

38.

Tseloni

Mailley

Farrell

Tilley

. (2010). Exploring the international decline in crime rates. European Journal of Criminology, 7(5), 375–394. https://doi.org/10.1177/1477370810367014

39.

Tsushima

Hamai

. (2015). Public cooperation with the police in Japan: Testing the legitimacy model. Journal of Contemporary Criminal Justice, 31(2), 212–228. https://doi.org/10.1177/1043986214568836

40.

Zhang

Messner

S. F.

Liu

. (2007a). A multilevel analysis of the risk of household Burglary in the city of Tianjin, China. British Journal of Criminology, 47(6), 918–937. https://doi.org/10.1093/bjc/azm026

41.

Zhang

Messner

S. F.

Liu

. (2007b). Bicycle-theft victimization in contemporary urban China. Journal of Research in Crime and Delinquency, 44(4), 406–426. https://doi.org/10.1177/0022427807305852n

Using a Rural Index to Assess Crime Risk and Crime Prevention Behavior Across the Urban–Rural Continuum: A Japanese Case Study

Abstract

Keywords

Introduction

Theoretical Background

Theories of Crime and Crime Prevention

The Urban–Rural Continuum

The Current Study

Pre-analysis

Study 1 and Study 2

Study Area

Data and Methods

Results

Preliminary Analysis: The Rural Index at the Municipal Level

Study 1. Crime Prevention Behavior for Vehicles: Open Crime Data Analysis

Study 2. Prevention Behavior: Survey Data Analysis

Discussion

Conclusions

Footnotes

Declaration of Conflicting Interests

Funding

ORCID iDs

Author Biograpies

References