Residential burglary victimization

Abstract

The current study employed the International Crime Victimization Survey for 142,665 subjects in 35 countries to compare residential burglary patterns across countries by their developmental levels. The mixed-level modeling shows that households in less developed countries were more likely to experience residential burglary than those in highly developed countries. In addition, married status and low income levels were positively, while living in a detached house was negatively, related to burglary victimization. These findings are consistent with routine activity theory.

Keywords

Residential burglary routine activity theory rational choice theory cross-national study burglary in developing countries

Introduction

Burglary is commonly defined as a criminal act involving breaking and entering a building or other structures to commit a crime (Pollock et al., 2010). A common subject for studies on burglary is the pattern of victimization. There have been tests on which factors make a household vulnerable to a burglary. Despite the significant number of studies on residential burglary victimization, a few limitations are noticeable. First of all, previous studies have primarily targeted highly developed countries, such as the United States, the United Kingdom, and Australia, with only a few exceptions (e.g. Messner et al., 2007 for China; Tabrizi and Madanipour, 2006 for Iran). However, these studies targeted highly developed countries and included only a small number of countries in their analyses. One exception was Bennett’s (1991) work that employed a pooled cross-sectional and time-series analysis of 52 nations from 1960 to 1984. However, he did not specifically target residential burglary; rather, he studied property crime as a combination of robbery and burglary in general. As a result, many tests have been inadequate as to whether theories developed in highly developed countries can be generalized to less developed countries. For example, Spano and Freilich (2009), in their review of studies conducted between 1995 and 2005, suggested that lifestyle theory and routine activity theory were more applicable in the United States than in other countries.

The current study examined the relationship between the developmental level of a subject’s country and his or her likelihood of residential burglary victimization by relying on prior theories of burglary victimization. To achieve the research goal, the current work utilized the household-and country-level mixed-modeling technique by using 142,665 samples from 35 countries in different regions of the world from the International Crime Victim Survey (ICVS International Working Group, 2005).

The neoclassical school of criminology

Although contemporary criminologists should be well aware of the neoclassical school of criminology it may be beneficial briefly to review it before discussing residential burglary patterns. The fundamental assumption regarding the neoclassical school of criminology is based on rational choice theory. According to this theoretical position, generally speaking, a human being is rational, and calculates the benefits and costs of criminal behavior. When one perceives that the benefits of committing a crime are greater than the risks associated with offending, he or she is more likely to decide to commit a crime. Rational choice theory can be applied to both property crimes and violent crimes, though it has been more commonly applied to property crime victimization than violent crime victimization. A number of theories are similar to rational choice theory. First of all, as an architect, Newman (1972) emphasized the importance of architectural design that enhances surveillance by both residents and bystanders for crime prevention. Newman argued that clearly discerning public, private, and semi-private territories is critical in improving surveillance. Land used for mixed purposes may lead to ambiguity in responsibility for surveillance, which in turn can bring a decrease in territoriality and an increase in burglary victimization. On the other hand, Clarke (1980) introduced several important crime prevention techniques. One particular example is target hardening. Target hardening, in a burglary context, refers to the technique that makes it difficult for a burglar to take an item from property, by employing safes, locks, alarms, and other security devices.

Another well-known theory is Cohen and Felson’s (1979) routine activity theory that focuses on the routine activities of the parties involved in crime, specifically, an offender, a victim, and a third party, to understand the occurrence of a crime. Cohen and Felson proposed that a motivated offender, a target, and the absence of a guardian against crime are required for crime to be committed. Specifically, Cohen and Felson proposed that criminal activity is very closely related to the ordinary activities of perpetrators and victims, such as going to work and shopping. Cohen and Felson attributed the changes in crime rates in the United States between 1947 and 1974 to changes in routine activity, such as an increase in participation in the female labor force and also in single-adult households. Another example is that the most vulnerable time of a day for residential burglary is the early morning, which is the time when a resident(s) usually leaves his or her dwelling for work or going to school. Next, lifestyle theory (Hindelang et al., 1978) is related to routine activity theory; lifestyle theorists particularly view demographic variables of an offender and his or her victim as all being related to the occurrence of crime, because demographic factors are related to their lifestyle; these important demographic variables include gender, age, social class, and marital and employment status.

Factors influencing residential burglary

The developmental level of a country

The social characteristics of less developed countries may be different from those of highly developed countries. An extended family system is more common in less developed countries than in highly developed countries. For example, in China, many adults live together with their parents in the same residence. Cohabitation among close family members and relatives may provide extra guardians (Messner et al., 2007). In their analysis of 2,500 subjects in Tianjin, China, Messner et al. (2007) suggested that routine activities in China are different from those in western societies. Chinese people are much less likely to leave their homes to eat out than their counterparts in Western or other highly industrialized countries. Consequently, a household with only a single adult is not a risk factor for burglary victimization in China as much as it is in highly developed countries.

Another important impact of the developmental level of a country upon burglary victimization may be related to the possibility that the overall poverty level in a country creates highly motivated burglars. In his presidential address to the American Society of Criminology, Richard Rosenfeld (2011) argued that the macro-level study of criminology is important because such study contributes to a better understanding of motivation for crime. The national poverty level may be one of the factors linked to the motivation of property criminals. Given the discussion, it is worth testing whether a different burglary victimization pattern exists according to the developmental level of residents’ countries.

Other factors

Other factors that could be related to residential burglary victimization pattern include individual, household, and city characteristics. First of all, individual characteristics comprise age, employment/school student, and marital status. Residential burglary is usually considered a passive crime because a burglar does not want to be seen by other people. For instance, in their interviews with ex-burglars in Tehran, Iran, Tabrizi and Madanipour (2006) reported that 68 percent of all burglars interviewed never targeted occupied dwellings (see also D’Alessio et al., 2012). Based on this finding, older and retired individuals may be likely to spend more time in their residences than younger ones. Thus, the presence of an older resident may provide a guardian (Capowich, 2003; Cohen and Cantor, 1981; Phillips and Walker, 1997).

In a similar line of reasoning as age group, the unemployed may be more likely to spend longer hours at their homes than the employed, which in turn provides guardians at their residences (Coupe and Blake, 2006). By using state-level pooled cross-sectional and time-series burglary victimization data in the United States, D’Alessio et al. (2012) indicated that weekday residential burglary victimization is inversely related to unemployment rates. Home guardianship may be affected not only by the employment status of a resident, but also by the resident’s other activities outside the home, such as going to school, playing sports, and eating out, because the resident leaves his or her home without a guardian for those activities (Capowich, 2003; Miethe and Meier, 1994; Roundtree and Land, 1996). Based on the same logic, a household with a married or cohabitating couple may be better protected than one with a single adult, though only because one of the married or cohabitating couples is likely to stay at home if they are not employed outside the home (Miethe and Meier, 1994; Phillips and Walker, 1997).

Next, household characteristics include household income, dwelling type, household size, and the use of target-hardening techniques. A dwelling with high-income residents or high property market values may provide a suitable target for a potential burglar because the dwelling is likely to contain valuable items (Bernasco and Luykx, 2003; Ham-Rowbottom et al., 1999; Miethe and Meier, 1994; Wellsmith and Burrell, 2005; Zhang et al., 2007). However, other studies have reported that in reality, lower-class households are more frequently burgled than better-off households (Tilley et al., 2011), perhaps because better-off households are more likely to be able to afford sophisticated security devices, such as wired burglar alarms and surveillance cameras, than their poorer counterparts (Cohen and Cantor, 1981).

Another household characteristic includes dwelling type. Some studies have reported that a detached house is more vulnerable to burglary than an apartment, because a detached house provides easier access while offering less natural surveillance by neighbors and bystanders than does an apartment (e.g. Tabrizi and Madanipour, 2006; Tseloni et al., 2004). Contrary to these studies, however, Bursik and Grasmick (1993) contended that detached houses are less likely to be vulnerable to burglary victimization than an apartment. Thus, further research is warranted on this subject. Another important variable may be household size, because the greater number of household members available at a residence, the more it may provide guardians (Roundtree and Land, 1996).

In addition to regular guardianship by human beings, physical guardianship is called ‘target hardening’ (Wilcox et al., 2007: 782). Employment of target-hardening techniques by an inhabitant may provide protection for a dwelling. Those target-hardening techniques may include burglar alarms, special door locks, door grilles, a watchdog, and a high fence (Wilcox et al., 2007). However, there are some research findings that are inconsistent with such an expectation for target hardening: the dwellings that employed target-hardening techniques are more likely to be subject to a burglary than those that did not do so. This counterintuitive finding may suggest a reverse causal order; burglary victimization may actually lead to a victim’s employment of target-hardening techniques, and not the other way around (Tseloni et al., 2004). In other words, burglary victimization may lead to a homeowner’s fear of further victimization, which is in turn conducive to their installing antitheft systems (van Kesteren et al., 2014). This problem is pronounced because a majority of victimization surveys failed to ask a subject when he or she installed security devices (Phillips and Walker, 1997).

Finally, there is a consensus that urban areas are more prone to crimes and other social problems, including burglary, than rural areas, as there may be more motivated offenders in large cities than small towns (Tseloni et al., 2004). Also, in comparison with rural residents, urban residents are more anonymous and often have a problem with distinguishing strangers from their neighbors. Thus, surveillance in a large city is not very effective, which may lead to a higher burglary rate in an urban area.

Methods

Data sources

The data for the dependent and independent variables were obtained from the International Crime Victim Survey (ICVS), with the only exception being Gross Domestic Product (GDP) per capita. The ICVS contains information on criminal victimization of various types, including both property and violent crimes, with residential burglary one of such crimes. The ICVS information is more accurate than the sources obtained from police agencies, which usually underestimate actual crime occurrence (van Kesteren et al., 2014). The ICVS used a standardized survey format with different languages (Uludag et al., 2009) and interviewed a sample of at least 2,000 respondents from each country using computer-assisted telephone interviews. However, a face-to-face survey supplemented the telephone survey when it was not feasible due to limited telephone availability in some areas in the countries under study (Uludag et al., 2009; van Kesteren et al., 2014). Approximately 70% of all surveys were obtained via telephone. The ICVS has two different types of sampling: one method obtains national samples, while the other sources samples from the major or capital city in a country. The current study selected only the survey collections from national samples. Among a total of 186,790 samples, 142,665 (76.4%) of samples were obtained from national samples, while the remaining 44,125 (23.6%) were obtained from capital city or other samples. As a result, the present work included 142,665 samples in 35 countries for regression analyses (see Appendix A for the list of countries).

The ICVS International Working Group has periodically collected and published the ICVS data since 1989 (van Kesteren et al., 2014). The current study used the fourth (1999–2003) and the fifth waves (2004–2006). The subjects included in the current study comprised those who answered either ‘yes’ or ‘no’ as to whether they had experienced residential burglary victimization for either the surveyed or the previous year. The current study downloaded the SPSS version of the ICVS data from the Data Archive and Networked Services.¹

Dependent variable

The dependent variable was residential burglary victimization. The ICVS poses the following question: ‘Over the past five years, did anyone actually get into your house or flat without permission and steal or try to steal something? Not including thefts from garages, sheds, or lock-ups.’ A burglary is defined as a successful ‘break-and-enter’ into a residence, regardless of whether items were taken from the home. The respondent can then select, ‘yes,’ ‘no’ or ‘don’t know.’

The next question is contingent for those who had experienced burglary victimization during the last 5 years. The question asks: ‘When did this happen? Was this… (1) this year, (2) last year (2004), (3) before that, (9) don’t know/can’t remember.’ In this particular study, an issue was found regarding a subject’s victimization over a long period. For instance, Zhang et al. (2007) used burglary victimization data collected over the last 5 years to include a larger number of victimizations. However, the researchers in that particular study admitted that the use of a longer time-frame hampers the detection of the direction of the relationship between the implementation of the target-hardening technique and burglary victimization. Hence, guardianship activities such as target-hardening techniques may be merely a response to victimization rather than a predictor of victimization. Thus, the present study used a time-frame of 2 years: this year and last year. Another contingent question was: ‘How often did it happen last year?’ The respondents can answer from ‘once’ to ‘five times or more.’ Given such information, the present work created a six-point scale: 0 = no victimization, 1 = once all the way through to 5 = five times or more.

Independent variables

Independent variables included various individual, household, city and country characteristics (see Table 1 for a description of variables). Individual characteristics comprised the ICVS respondents’ age group, gender, marital status and employment/student status. The ICVS classified 12 age groups ranging from 1 = 16–19, 2 = 20–24 through to 12 = 70 or older. The current study treated age groups as an interval level variable. Gender was a dichotomous variable, female (1) and male (0), while marital status indicated whether or not a subject was married (or cohabitating with a partner). Finally, a subject’s employment/school attendance status, which included both full-time and part-time, was introduced.

Table 1.

Variable description.

Variable name	Attributes
Dependent variables
Residential burglary	The frequency of victimization: 0 = no victimization, 1 = one, 2 = two, 3, = three, 4 = four 5 = five or more times
Independent variables
Individual characteristics
Age group	1 = 16–19, 2 = 20–24, 3 = 25–29, through to 12 = 70+
Female gender	1 = female, 0 = male
Marital status	1 = married or cohabitate, 0 = non-married or cohabitate
Employment or attending school	1= yes, 0 = no
Household characteristics
Household income	lower 25%, lower 25–50%, upper 25–50%, upper 25% of national population
Household size	1 through 9, and 10 (for10 or more)
Type of dwellings (House)	fully detached house and all others
Target hardening (5 points scale)	Burglar alarm, special door locks, special window or door grilles, dog that would detect burglar, and high fence (Range from 0 to 5)
City/Country characteristics
City size	1=under 10,000, 2=10,000–50,000, 3=50,000–100,000, 4=100,000–500,000, 5=500,000–1,000,000, 6–1,000,000+
GDP per capita	In US dollars

Household characteristics included household income level, household size, type of dwelling and target-hardening techniques. Household income, which was based on a subject’s own assessment, was divided into four categories according to relative household income level within a subject’s country of residence, i.e. lower 25%, 25–50%, 50–75% and 75–100% of national population. Next, the household size refers to the number of occupants aged 16 or older, which ranged from 1 to 10,² while the type of dwelling was classified into a completely detached house and all others. On the other hand, target-hardening techniques were assessed on a composite scale out of five items of a burglar alarm, special door locks, special window or door grilles, a dog that would detect a burglar, and a high fence. The composite scale ranged from 0 to 5 because one point was given to each of those techniques.

Finally, the city-level indicators included the population size of the city where a respondent lived. The city population ranged from 1 for under 10,000 to 6 for over 1,000,000. On the other hand, the country-level indicator was GDP per capita in US dollars for the year of 2002; a common indicator of economic development level of a country. Unlike other variables, the information on GDP per capita was obtained from The United Nations Development Program’s Human Development Reports (2003).

Analytical strategies

The current dataset required a few remedial measures: data transformation, missing data replacement, mixed-level modeling and negative binomial regression. First of all, as GDP per capita is skewed, it was log transformed to reduce the violation of normality of GDP. Next, there are some missing data. If the current study uses a listwise deletion technique that eliminates values in all other variables from a multiple regression analysis when a case is missing for one variable, it therefore wastes a significant number of data and leaves only 29,075 subjects. In addition, the listwise deletion technique is appropriate when the loss is missing completely at random (MCAR), and the absence of values on the independent variable is not related to a dependent variable or other independent variables. The loss in survey research, however, is not likely to be MCAR (Allison, 2002). Thus, it is necessary to find a sophisticated technique to replace missing data; to do this, the present study employed a multiple imputation technique (three imputed datasets) using the automatic method,³ which is known to be superior to other techniques of missing data replacement. A regression analysis with the data from multiple imputation produces unbiased parameter estimates, and it is also robust to departures from normality assumptions with a large number of missing data (Wayman, 2003). Given the assumption of missing at random, imputed values are predicted and created by considering the relationship of the variable with missing values with those in other variables.⁴ Following these procedures, average or pooled regression coefficients are obtained from multiple data sets with imputed values (for more information on multiple imputation, see Allison, 2002). Missing data replacement with multiple imputation techniques may not significantly alter the original data. For example, although the city size had many missing data (89,987 available out of a total of 142,665), missing data replacement via multiple imputation techniques did not significantly change the mean from the original dataset (3.01) to that of an imputed one (2.97). By using SPSS (version 22), the present examination calculated average regression coefficients over those three complete data sets with imputations.

Next, given the multilevel structure of the current dataset, multilevel modeling is appropriate (Raudenbush and Bryk, 2002). The current study included country-level information on GDP per capita. The present work employed the SPSS generalized linear mixed-level package (GENLIN MIXED). In addition, given the limited range of the dependent variable (from 0 to 5), the ordinary least square regression analyses may be inappropriate. Instead, the dependent variable is considered a count outcome because it does not have negative integers. Also, the count represented a rare event, and 97% of subjects answered that they had not been burgled. Therefore, the present research employed negative binomial (NB) regression with log link function, $η = \log (λ)$ . The NB distribution incorporates overdispersion by estimating an extra parameter in the variance function (Heck et al., 2012). There were 11 fixed effects (intercept, nine within-country predictors and one between-country predictor of GDP per capita) and one random effect (random intercept variance parameter, $u_{0 j}) .$ The current mixed-level NB regression model may be summarized as follows

η_{i j} = γ_{00} + γ_{01} GDP per capita + γ_{10} age + γ_{20} gender + γ_{30} marital status +_{} γ_{40} employment + γ_{50} household income + γ_{60} household size + γ_{70} house + γ_{80} target hardening + γ_{90} city size + u_{0 j}

Two separate models were developed to examine the differences in repeated residential burglary victimization according to different levels of a country’s economic development. The first model included all variables except the country-level variable, GDP per capita. The second model added the GDP per capita to the first model to investigate how a country’s development level is related to repeated burglary victimization.

Results

Description of sample

Table 2 shows descriptive statistics. The dependent variable of burglary victimization was available for 142,665 subjects, and the current work excluded those who did not answer the question on victimization. Only a small percentage of those subjects had experienced a burglary (less than 3%). This small percentage of victimization is partially related to the fact that the current study limited victimization to only within this and the last year. Next, 56.2% of all subjects were females, and 58.4% of them were either married or cohabitated. Some 61% of them were either employed or attended schools. Although the household income levels were somewhat evenly distributed across these four categories, the lower 25% was a little less common (20.5%) than the other three income levels. The most common household size was two (26.5%). Of the subjects, 24% lived in completely detached houses. A majority of subjects used none (33.6%) or only one (32.8%) target-hardening technique out of a total of five. Finally, approximately a half of all subjects had their residences in small towns with populations under 50,000.

Table 2.

Descriptive statistics.

	N	Min.	Max.	Mean	Std. Dev.
Interval variables
Burglary victimization	142,665.0	0.00	5.00	0.03	0.23
Age groups	128,421.0	1.00	12.00	6.68	3.35
Household size	119,791.0	1.00	10.00	3.09	1.65
Target hardening	131,852.0	0.00	5.00	1.20	1.17
City size	89,987.0	1.00	6.00	3.01	1.88
GDP per capita	142,665.0	5,010.00	61,190.00	22,199.76	9,565.22
Categorical variables	N	Percentage
Gender	142,665.0
Male Female	62,455.0 80,210.0	43.8 56.2
Marital status	139,048.0
Married Non-married	81,259.0 57,789.0	58.4 41.6
Employment or attending school	135,335.0
Employed Unemployed	82,648.0 52,687.0	61.1 38.9
Household income	115,600.0
0–25%	23,667.0	20.5
25–50%	30,770.0	26.6
50–75%	29,493.0	25.5
75–100%	31,670.0	27.4
House (fully detached)	85,917.0
House All others	20,672.0 65,245.0	24.1 75.9

Multilevel modeling results

The unconditional or null model (not shown) indicated a significant variability of household burglary victimization across the countries under study: the intercept of 0.662 at the variance component was significant at the 0.001 level (Z = 3.925), thus suggesting that it was worth creating a mixed-level regression model. Table 3 shows the multilevel NB regression results for residential burglary victimization. The first regression model estimated 10 fixed effects (i.e. the intercept and slopes for nine level-1 predictors), except the country-level variable of GDP per capita, and one random effect at level 2 (the intercept). Six variables had a significant relationship with residential burglary victimization. Age and married status were negatively related to burglary victimization. Given the odds ratio, the following interpretation is possible: one unit change of age group was related to a 2% decrease in the predicted count of burglary victimization and its repetition, while married status was linked to a 10% reduction in burglary victimization count when the regression model held other level-1 variables and level-2 variance (country) constant. Unlike these two variables, being employed was associated with a 9% increased count of household burglary victimization. Also, a lower household income status was positively related to residential burglary victimization. In comparison with those in the upper 25% of national income level, being in the lower 25% or in the 25–50% band of national household income level was associated with a 26% and a 9% increase in the burglary victimization count, respectively. Next, one unit change in using target-hardening techniques and city size correlated to approximately 6% and 5% increases in the count of burglary victimization, respectively. Unlike the variables above, some other variables failed to display a significant relationship with burglary victimization, namely: gender, household size and dwelling type (house). The level-2 variance components for the intercept (0.672, Z = 3.925, p < 0.001) indicated that considerable variance still exists to explain at level-2. Therefore, Model 2 added the country-level variable, which was GDP per capita.

Table 3.

Negative binomial regression for burglary victimization (N = 142,665).

Level 1	1	2
Fixed effect
Intercept	–3.759*** [0.023]	0.031 [1.031]
Age group	–0.016** [0.984]	0.001 [1.000]
Female	0.049 [1.050]	–0.002 [0.998]
Married status	–0.110*** [0.896]	–0.023* [0.977]
Employment/attending school	0.089* [1.093]	–0.022 [0.978]
Household income
Lower 25%	0.228*** [1.256]	0.089*** [1.093]
25–50%	0.088* [1.092]	0.064*** [1.066]
50–75%	0.016 [1.016]	0.035** [1.035]
75–100%	reference group	reference group
Household size	–0.010 [0.990]	0.002 [1.002]
House (fully detached)	0.055 [1.053]	–0.023* [1.024]
Target hardening	0.058*** [1.059]	0.007 [1.007]
City size	0.048*** [1.049]	0.003 [1.003]
Level 2 (Country) GDP per capita (ln)		–0.087*** [0.917]
Negative binomial	1.000	1.000
Random Effect Variance components
Intercept	0.672*** Z = 3.925	0.000

Notes: 1. unstandardized regression coefficients are shown; odds ratio in brackets.

2. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

3. 35 countries in Level 2.

GDP per capita in Model 2 displayed a significant and an inverse association with residential burglary victimization: one unit change in GDP per capita (ln) was connected to an 8% decrease in burglary victimization. There were other noticeable changes in Model 2 from Model 1. Age group, employment status, target hardening, and city size of a resident became no longer significant when Model 2 added GDP per capita, the level-2 variable. In addition, dwelling type (house) obtained a statistical significance at 0.05 level in Model 2: Living in a detached house was associated with a 2% reduced odds of burglary victimization. By contrast, other variables in Model 2 retained a similar relationship with the one in Model 1. Although the significance level changed from 0.001 level to 0.05 level, married status was still related to a reduced count of burglary victimization: Being married or cohabitated was related to a reduced burglary victimization count by a factor of 0.977 (3%). By contrast, in comparison with households in the upper 25%, lower income households experienced a high rate of burglary victimization. For instance, households in the lowest 25% were linked to a 9% increased count of burglary victimization. As seen in Model 1, however, gender and household size still maintained a non-significant relationship with burglary victimization.

Discussion and conclusion

Interpretation of the findings

The current study investigated residential burglary victimization in households in 35 countries by using a relatively large number of samples from the ICVS. The multilevel regression results demonstrated that households in less developed countries were much more likely to be susceptible to residential burglary than those in highly developed countries. This finding is interesting because some people may perceive that burglary, a property crime, is more common in highly developed countries, while violent crimes are more frequent in less developed countries. This result is also inconsistent with the expectation that burglary is more common in highly industrialized countries than in less industrialized ones because of the high availability of expensive and portable items in highly industrialized countries. However, this finding is not necessarily inconsistent with routine activity theory, because one of the most important elements of the theory is the presence of a motivated offender. It is an accepted view that developing countries experience high levels of poverty, which in turn may create highly motivated offenders in those countries (Bennett, 1991). In other words, poverty may play a role of ‘encouragement’ for burglary because many individuals in developing countries may struggle to make a living by legitimate means. Instead, some of them may choose burglary as an alternative means of making a living.

Controlling GDP per capita in the regression Model 2 changed the statistical significances of several variables: living in a fully detached house obtained statistical significance at the 0.05 level, while age group, employment status, target hardening, and the city size of residence lost their significance. Such a finding suggests that it is very important to introduce a country-level variable such as GDP per capita when researching an international survey because the regression results may be misleading without a nation’s economic development level.

As well as the findings for the development level of a country, those for other variables were consistent with routine activity theory for both developed and developing countries. The existence of married or cohabitating couples was related to a reduced chance of residential burglary. Given the routine activity theory, the following interpretation is possible: one person in a married or cohabitated couple may be more likely to stay at home during the day and thus play a role of a guardian for residential burglary.

Next, households in the lowest income level were more likely to experience burglary victimization than their high-income counterparts. Once these households were burgled, they were more likely to be revictimized than their high-income counterparts. Such a finding is also consistent with routine activity theory. A household with a low income was likely to be located in a low-income neighborhood where potential offenders are more likely to reside and also have easy access to the household; one of the important elements of the routine activity theory is the existence of motivated offenders. Another factor may be that households with lower incomes are less likely to have sophisticated anti-burglary systems, which in turn makes them vulnerable to victimization. By contrast, residential properties with high-income family members are more likely to be located in a middle- or upper-class neighborhood that is likely to be well defended against intrusion by outsiders. As a result, burglars often have problems gaining access to a property in a middle- or upper-class neighborhood, although the property provides a lucrative target.

In addition to the variables discussed above, households with detached houses were less likely to be burgled than those with other types of dwellings. This outcome refutes the suggestion that a detached house is more vulnerable to a burglary than an apartment, because a detached house provides easier access while offering less natural surveillance by neighbors and bystanders than does an apartment (e.g. Tabrizi and Madanipour, 2006; Tseloni et al., 2004). The current finding, however, is consistent with Bursik and Grasmick (1993).

Unlike those variables discussed earlier, there were non-significant findings for a respondent’s individual characteristics such as age group, gender, and employment status. A respondent’s age, gender, and employment status in the ICVS data does not represent the demographic of a whole household, which may be the reason why a respondent’s individual characteristics were not significantly related to household burglary victimization. Another interesting finding is that household size, target hardening, and city size were not significantly associated with burglary victimization when controlling for GDP per capita. One possible explanation is that those three variables are highly covariate with GDP per capita: the bivariate correlations (r) of GDP per capita with household size, target hardening and city size were –0.18, 0.13, and 0.04, respectively, and they are all significant at the 0.01 level (see Appendix B). Stated differently, these three variables may also be indicators of the development level of a country; thus, this finding may confirm that the developmental level of a subject’s country of residence is a strong predictor of burglary victimization. This discovery also suggests that residents in developed countries use more target-hardening techniques than those in developing countries. It is also inconsistent with some previous work that reported a positive link between target hardening and burglary victimization. That research suggested a reverse causal order between those two variables. In other words, a household member employs target-hardening techniques in response to prior burglary victimization (Phillips and Walker, 1997; Tseloni et al., 2004). The current study, however, found no clear evidence of the relationship between using target-hardening techniques and household burglary victimization.

Limitations of the current work

In spite of such interesting findings highlighted in this study, some limitations of the current study deserve a brief discussion. Again, an individual characteristic of an ICVS respondent does not necessarily provide information on overall household characteristics. For instance, the ICVS respondent’s gender does not provide information on how many males and females are in a given household. That may explain why female gender failed to show any significant association with burglary victimization. Another issue is limited data availability. For example, data on city size were missing for many subjects. As a result, multilevel modeling was impossible to conduct with three levels, such as household, city and country. Moreover, the current study did not include other lifestyle variables of the subjects, which included involvement with sports activities and going out to movies and dinners, among other activities. Next, a temporal pattern of crime is also an important element of routine activity theory, as well as the spatial pattern. As discussed earlier, D’Alessio et al. (2012) indicated that unemployment is more closely related to residential burglary committed during weekdays than during weeknights and weekends, because unemployment more strongly affects the number of occupants at home during weekdays, rather than weeknights and weekends. However, due to the lack of data availability in the ICVS, the current study could not introduce a temporal dimension for occurrences of burglary. Finally, because of questionable compatibility, the present study selected data only from national surveys by excluding regional surveys. As a result, the number of countries in the current examination was reduced from 60 to 35, which eliminates many developing countries from the regression analysis. Future researchers need to include more countries, and they need to test other important country-level variables, as well as GDP per capita, that may be related to burglary victimization, such as the Gini-coefficient of income inequality.

Conclusion

The current study employed a relatively large number of samples from the ICVS, including those in both highly developed and less developed countries. A significant difference was found for burglary victimization according to a country’s economic development level. Households in less developed countries were more likely to experience burglary victimizations than those in highly developed countries. Thus, future studies need to continue to delve into the differences in residential burglary patterns between these two groups of countries.

Footnotes

Notes

Appendix

References

Allison

(2002) Missing Data. Thousand Oaks, CA: Sage Publications.

Bennett

(1991) Routine activities: A cross-national assessment of a criminological perspective. Social Forces 70: 147–163.

Bernasco

Luykx

(2003) Effects of attractiveness, opportunity and accessibility to burglars on residential burglary rates of urban neighborhoods. Criminology 41: 981–1002.

Bursik

Grasmick

(1993) Economic deprivation and neighborhood crime rates, 1960-1980. Law and Society Review 17: 263–268.

Capowich

(2003) The conditioning effects of neighborhood ecology on burglary victimization. Criminal Justice and Behavior 30: 39–61.

Clarke

RVG

(1980) ‘Situational’ crime prevention: Theory and practice. British Journal of Criminology 20: 136–147.

Cohen

Cantor

(1981) Residential burglary in the United States: Life-style and demographic factors associated with the probability of victimization. Journal of Research in Crime and Delinquency 18: 113–127.

Cohen

Felson

(1979) Social change and crime rate trends: A routine activity approach. American Sociological Review 44: 588–608.

Coupe

Blake

(2006). Daylight and darkness targeting strategies and the risks of being seen at residential burglaries. Criminology 44: 431–464.

10.

D’Alessio

Eitle

Stolzenberg

(2012) Unemployment, guardianship and weekday residential burglary. Justice Quarterly 29: 919–932.

11.

Ham-Rowbottom

Gifford

Shaw

(1999) Defensible space theory and the police: Assessing the vulnerability of residences to burglary. Journal of Environmental Psychology 19: 117–129.

12.

Heck

Thomas

Tabata

(2012). Multilevel Modeling of Categorical Outcomes using IBM SPSS: Quantitative Methodology Series. New York, NY: Routledge.

13.

Hindelang

Gottfredson

Garofalo

(1978) Victims of Personal Crime: An Empirical Foundation for a Theory of Personal Victimization. Cambridge, Mass.: Ballinger.

14.

IBM Corporation (2011) IBM SPSS Missing Value 20. Available at: ftp://public.dhe.ibm.com/software/analytics/spss/documentation/statistics/20.0/en/client/Manuals/IBM_SPSS_Missing_Values.pdf (accessed 10 October 2015).

15.

ICVS International Working Group (2005) International Crime Victimization Survey (ICVS) Series. Available at: http://www.icpsr.umich.edu/icpsrweb/ICPSR/series/175 (accessed 12 August 2015).

16.

Messner

Zhang

. (2007) Risks of criminal victimization in contemporary urban China: An application of lifestyle/ routine activities theory. Justice Quarterly 24: 496–522.

17.

Miethe

Meier

(1994) Crime and its Social Context: Toward an Integrated Theory of Offenders, Victims, and Situations. Albany, NY: SUNY Press.

18.

Newman

(1972) Defensible Space: Crime Prevention Through Urban Design. New York: Macmillan.

19.

Phillips

Walker

(1997) State differences in burglary victimization in Australia: A research note. Journal of Sociology 33: 91–100.

20.

Pollock

Joo

Lawton

(2010) Juvenile arrest rates for burglary: A routine activities approach. Journal of Criminal Justice 38: 572–579.

21.

Raudenbush

Bryk

(2002) Hierarchical Linear Models: Applications and Data Analysis and Methods. 2nd ed. Thousand Oaks, CA: Sage Publications.

22.

Rosenfeld

(2011) The big picture: 2010 presidential address to the American Society of Criminology. Criminology 49: 1–26.

23.

Roundtree

Land

(1996) Burglary victimization, perceptions of crime risk, and routine activities: A multilevel analysis across Seattle neighborhoods and census tracts. Journal of Research in Crime and Delinquency 33: 147–180.

24.

Spano

Freilich

(2009) An assessment of the empirical validity and conceptualization of individual level multivariate studies of lifestyle/ routine activities theory published from 1995 to 2005. Journal of Criminal Justice 37: 305–314.

25.

Tabrizi

Madanipour

(2006) Crime and the city: Domestic burglary and the built environment in Tehran. Habitat International 30: 932–944.

26.

The United Nations Development Program (2002) Human Development Report. The United Nations Development Program. Available at: http://www.undo.org (accessed 22 May 2015).

27.

Tilley

Tseloni

Farrell

(2011) Income disparities of burglary risk. British Journal of Criminology 51: 296–313.

28.

Tseloni

Wittebrood

Farrell

. (2004) Burglary victimization in England and Wales, the United States and the Netherlands. British Journal of Criminology 44: 66–91.

29.

Uludag

Colvin

Hussey

. (2009) Democracy, inequality, modernization, routine activities, and international variations in personal crime victimization. International Criminal Justice Review 19:265–286.

30.

Van Kesteren

Van Dijk

Mayhew

(2014) The international crime victims surveys: A retrospective. International Review of Victimology 20: 49–69.

31.

Wayman

(2003). Multiple imputation for missing data: What it and how can I use it? Paper presented at the 2003 annual meeting of the American Educational Research, Chicago, IL.

32.

Wellsmith

Burrell

(2005) The influence of purchase price and ownership levels on theft targets. British Journal of Criminology 45: 741–764.

33.

Wilcox

Madensen

Tillyer

(2007) Guardianship in context: Implications for burglary victimization risk and prevention. Criminology 45: 771–803.

34.

Yuan

(2000) Multiple imputation for missing data: Concepts and new developments. Proceedings of the Twenty-Fifth Annual SAS Users Group International Conference (Paper No. 267). Cary, NC: SAS Institute.

35.

Zhang

Messner

Liu

(2007) A multilevel analysis of the risk of household burglary in the city of Tianjin, China. British Journal of Criminology 47: 918-937.