Violent crime and female victimization: evidence from metropolitan regions in South Korea

ABSTRACT

In the economics of crime literature, victimization by crime has received less attention than the supply side of crime. This article investigates the relationship between violent crime and female victimization. We show that violent crime increases with both the overall female exposure and female proportion in the victim-target group. Potential interactions between these female characteristics and income inequality are also shown to influence the incidence of violent crime. Empirically, we introduce proxies for female characteristics that better reflect our hypotheses – for example, young unmarried female-headed households (for exposure to crime) and new job gains among females (for economic status). Using a panel of South Korean metropolitan regions, 2000–2011, we find that a certain limited change in these female characteristics could account for as much as 16% of violent crime.

KEYWORDS:

• Women risk
• labour markets of women
• economic change of women
• social disorganization
• income inequality

JEL CLASSIFICATION:

• K42
• K14
• D63

Acknowledgements

Constructive suggestions and encouragement by the editor and the referee are kindly acknowledged. We would like to thank David Allen, Bruce Benson, Seokju Cho, Jaesung Choi, Hyunchul Kim, Hyunseok Kim, Minseong Kim, Hojun Lee, Junsang Lee, and Sungkyu Park for their helpful comments on the earlier versions of this paper. We also thank insightful conversations on female victimization and crime with Judges Byung Koo Cho and Kisang Jung, as well as attentive comments from the participants in the Korean Law and Economics Association Meetings, the Western Economic Association International Conference, and the seminars at Sungkyunkwan University and the LEC (Law and Economics with Courts) Forum. Finally, Kyongpyo Ko and Joowon Lee have provided excellent research assistance. This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea grant: [Grant Number NRF-2015S1A5A2A03049110]. The usual caveat applies.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ The response of males was about 1.7 times higher (11.4%). This difference is not large in absolute terms, but further evidence provided below and in Section II will be sufficient to prove the seriousness of female victimization.

² Female victimization has also become a major social problem in other countries such as Germany, the UK, Canada, Central America (e.g. El Salvador, Guatemala, Mexico), and Australia.

³ The ‘regulatory crimes’ category involves violations of numerous administrative regulations that have been declared to be crimes through political processes. Thus, they are differentiated from ‘customary crimes’, which include homicide, robbery, arson, rape and sexual assaults, theft, assault, fraud, vandalism, and so on, based on traditional definitions under criminal law.

⁴ McCollister, French, and Fang (2010, 25) estimated the social costs of various crimes using US data and found that violent crimes incur the highest costs. Although we initially intended to include serious assault, its consistent time-series figures are simply unavailable in Korea, due to frequent revisions of the various laws governing the scopes of different assaults (e.g. Act of Crimes, Act of Punishing Violence, and other Acts) in the 2000s and, for other reasons.

⁵ The Global Study on Homicide is based on the United Nations Office on Drugs and Crime Homicide Statistics dataset and covers 219 countries and territories since 2012. Also, a recent report by the Violence Policy Center (2015) reveals that, in the US in 2013, 1615 females were murdered by males in single victim/single offender incidents, as submitted to the Federal Bureau of Investigation’s (FBI) for its Supplementary Homicide Report. In particular, the US Department of Justice confirmed that women are significantly more likely to be victims of violent crimes committed by intimate partners.

⁶ In criminology and sociology, modern victimization studies were fully launched with the advent of lifestyle-exposure theory (Hindelang, Gottfredson, and Garofalo 1978) and routine activity theory (Cohen and Felson 1979) in the late 1970s. Cohen, Kluegel, and Land (1981) subsequently merged the two in the ‘opportunity theory’. After about a decade, opportunity theory became the modus operandi for the current major theories of victimization (Meier and Miethe 1993). Cohen, Kluegel, and Land (1981) argued that exposure to crime (defined as the physical visibility and accessibility of persons or objects to potential offenders at any given time or place) is indicative of one’s vulnerability to crime. They claim that the factors for target attractiveness include income, protective environment, and the capacity of people to resist attack

⁷ Nonetheless, the premise of female vulnerability against being victimized may not hold in certain situations, as a referee legitimately pointed out. For example, there may be various enforcement measures that mitigate any perception of different propensities to resist (e.g. court’s harsher punishments against female victimization). Also, citizens judging themselves to be at greater risk (i.e. females) may take greater precautions in various ways. We have scrutinized these possibilities in terms of substantive laws and related surveys, but concluded that our assumption of female’s relative vulnerability would still hold. We are grateful for the attentive comment.

⁸ Intuitively, potential criminals are more likely to encounter female workers because females in the target group are more exposed.

⁹ Although violent crime is less strongly affected by economic incentives than property crime, economic incentives still appear to play an important role in the incidence of violent crime, based on the existing literature and the survey results. Or, given other non-economic aspects of violent crime, economic incentives are worthy of investigating, at least as the trigger of violent crime at the margin.

¹⁰ Before 1997, Korea’s income distribution was considered to be one of the most equitable among the OECD member countries. By 2011, however, the income gap between the richest and the poorest 10% of the population (i.e. the S90/S10 ratio) was 10.7 in Korea, which is the 9th highest among 33 member countries (OECD 2013, 4). The World Economic Forum (World Economic Forum 2014) defines this widening income inequality into ten global risks of highest concern in 2014, declaring that income disparity is among the most worrying issues.

¹¹ Today, the notion of moral insensitivity seems to be a dominant factor in understanding the nexus between $\mathrm{I}\mathrm{N}\mathrm{Q}$ and ${\mathrm{C}}^{\mathrm{v}}$. Although one cannot explicitly explain the envy effect, the existing literature has either assumed the existence of the envy effect or simply suggested the envy effect as an ex-post explanation for the empirical results that were obtained. Refer to Witt, Clarke, and Fielding (1998, 266), Fajnzylber, Lederman, and Loayza (2002b, 1328), or Sachsida et al. (2010, 95).

¹² In fact, this recommendation came from various studies on labour markets or income distribution such as Blau and Kahn (1992) and Altonji and Blank (1999). They emphasized the importance of gender inequality in various aspects. Increasing female participation in the labour force may tend to drive down male wages and/or employment probabilities, as a referee correctly suggested. However, we leave a separate investigation of this longer-term effect for future research.

¹³ A simple illustration that represents these conjectures of conflicting forces can be made in terms of our model. $\mathrm{F}\mathrm{E}\mathrm{M}$ includes the characteristics of female exposure and economic status. Suppose that $\mathrm{I}\mathrm{N}\mathrm{Q}=\mathrm{g}\left(\mathrm{F}\mathrm{E}\mathrm{M}\right)$ and $\mathrm{F}\mathrm{E}\mathrm{M}=\mathrm{h}\left(\mathrm{I}\mathrm{N}\mathrm{Q}\right)$ where $\mathrm{\partial }\mathrm{I}\mathrm{N}\mathrm{Q}/\mathrm{\partial }\mathrm{F}\mathrm{E}\mathrm{M}=\mathrm{\partial }\mathrm{g}/\mathrm{\partial }\mathrm{F}\mathrm{E}\mathrm{M}<0$ and $\mathrm{\partial }\mathrm{F}\mathrm{E}\mathrm{M}/\mathrm{\partial }\mathrm{I}\mathrm{N}\mathrm{Q}=\mathrm{\partial }\mathrm{h}/\mathrm{\partial }\mathrm{I}\mathrm{N}\mathrm{Q}>0$. Also, assume that ${\mathrm{C}}^{\mathrm{v}}={\mathrm{C}}^{\mathrm{v}}\left(\mathrm{g}\left(\mathrm{F}\mathrm{E}\mathrm{M}\right),\mathrm{h}\left(\mathrm{I}\mathrm{N}\mathrm{Q}\right)\right)$ for demonstrative convenience. Then, with the simplified feature of $\mathrm{d}\mathrm{F}\mathrm{E}\mathrm{M}=\left(\mathrm{\partial }\mathrm{F}\mathrm{E}\mathrm{M}/\mathrm{\partial }\mathrm{I}\mathrm{N}\mathrm{Q}\right)\times \mathrm{d}\mathrm{I}\mathrm{N}\mathrm{Q}$, the total derivative of ${\mathrm{C}}^{\mathrm{v}}$ with respect to $\mathrm{I}\mathrm{N}\mathrm{Q}$ is as follows: $\left(\mathrm{\partial }{\mathrm{C}}^{\mathrm{v}}/\mathrm{\partial }\mathrm{I}\mathrm{N}\mathrm{Q}\right)+\left[2\times \left(\mathrm{\partial }{\mathrm{C}}^{\mathrm{v}}/\mathrm{\partial }\mathrm{F}\mathrm{E}\mathrm{M}\right)+\left(\mathrm{\partial }{\mathrm{C}}^{\mathrm{v}}/\mathrm{\partial }\mathrm{I}\mathrm{N}\mathrm{Q}\times \mathrm{\partial }\mathrm{I}\mathrm{N}\mathrm{Q}/\mathrm{\partial }\mathrm{F}\mathrm{E}\mathrm{M}\right)\right]\times \left(\mathrm{\partial }\mathrm{F}\mathrm{E}\mathrm{M}/\mathrm{\partial }\mathrm{I}\mathrm{N}\mathrm{Q}\right)$. The first term indicates the marginal effect of income inequality. The second term represents the conflicting effects of $\mathrm{I}\mathrm{N}\mathrm{Q}$ on ${\mathrm{C}}^{\mathrm{v}}$ via $\mathrm{F}\mathrm{E}\mathrm{M}$. The first component in the bracket increases ${\mathrm{C}}^{\mathrm{V}}$, while the second component decreases ${\mathrm{C}}^{\mathrm{v}}$ by lowering $\mathrm{I}\mathrm{N}\mathrm{Q}$. Perhaps, these conjectures may be pertinent more for a longer-term effect of $\mathrm{I}\mathrm{N}\mathrm{Q}$. Nonetheless, as they are potentially important, we will attempt to empirically determine the sign of the conflicting forces by including an interaction term. We are grateful to a commentator for helping us address this issue more clearly.

¹⁴ The grouping reflects the availability of other regional data and the homogeneity in the office sizes.

¹⁵ For example, Raphael and Winter-Ebmer (2001), Gould, Weinberg, and Mustard (2002), and Lin (2008) argued that the overall unemployment rate may not help identify the marginal criminal. Thus, researchers attempted to identify crime-prone groups such as younger, male, or less-educated classes (Tauchen 2010).

¹⁶ Gottfredson and Hirschi (1990) argued that weak family structures with children exhibiting low self-control could be a source of criminal behaviour. $\mathrm{D}\mathrm{I}\mathrm{V}\mathrm{O}\mathrm{R}\mathrm{C}\mathrm{E}$ and $\mathrm{F}\mathrm{E}\mathrm{M}\mathrm{H}\mathrm{E}\mathrm{A}\mathrm{D}$ have similar effects on $\mathrm{C}{\mathrm{R}}^{\mathrm{v}}$, but $\mathrm{F}\mathrm{E}\mathrm{M}\mathrm{H}\mathrm{E}\mathrm{A}\mathrm{D}$ may capture other characteristics since it can be an outcome of being a widow or single (except for being a divorcee).

¹⁷ An anonymous referee provided a couple of warnings against the legitimacy of using this variable as a proxy for alcohol consumption. For example, a possibility exists that lower income individuals may tend to spend less per unit of alcohol consumed. However, although based on a limited survey report, we reached a conclusion that the alcohol expenditures would have at least a monotonic relationship with the amount consumed even if such a possibility holds. A more detailed explanation is available from the authors.

¹⁸ We are deeply indebted to an anonymous referee for emphasizing this aspect of the model, which ultimately led us to explore the following empirical investigation.

¹⁹ As another example, inequality was argued to increase the unsuccessful individuals’ frustration and thus enhances the tendency to commit crime against either the rich or the poor (Kawachi et al. 1999; Pare and Felson 2014).

²⁰ Nonetheless, existing empirical investigations have produced mixed results. For example, Fajnzylber, Lederman, and Loayza (2002a, 2002b), Lederman, Loayza, and Menéndez (2002), and Saridakis (2004) confirmed a positive relationship, while estimates were statistically insignificant in Doyle, Ahmed, and Horn (1999), Neumayer (2005), and Haddad and Moghadam (2011). Chintrakahn and Herzer (2012) even found a negative relationship.

²¹ It is also called the employment-to-population ratio, which is, according to Paul Ashworth, one of the best measures of labour market conditions.(http://money.cnn.com/2012/10/18/news/economy/other-unemployment-rate/index.html)

²² Two variables increased with the same average annual growth rate of 0.2% for our sample period. In estimation, the coefficient of $\mathrm{E}\mathrm{X}\mathrm{P}\mathrm{O}\mathrm{S}\mathrm{U}\mathrm{R}\mathrm{E}2$ had a correct sign but was insignificant. We then speculated that the effect of $\mathrm{E}\mathrm{X}\mathrm{P}\mathrm{O}\mathrm{S}\mathrm{U}\mathrm{R}\mathrm{E}2$ might vary depending on the age group. First, the distribution of female victims of ${\mathrm{C}}^{\mathrm{v}}$ was heterogeneous across age groups. Moreover, based on Women’s Lives through Statistics (2015) by Statistics Korea and the Ministry of Gender Equity and Family, the time spent outside, for example, for income-generating work and for commuting was greatest for those in their 30 s, followed by those in their 40 s and 50 s. Thus, we decided to focus on these age groups for $\mathrm{E}\mathrm{X}\mathrm{P}\mathrm{O}\mathrm{S}\mathrm{U}\mathrm{R}\mathrm{E}2$. This is equivalent to effectively identifying a victim-prone group, similar to identifying crime-prone groups for potential criminals in the economics of crime literature.

²³ That figure is estimated to be close to one in every four households in 2015. This does not appear to be a unique phenomenon of Korea. According to the US Census Bureau, the number of one-person households is increasing. In 1970, only 17% of US households were considered one-person and this figure grew steadily over the past four decades, reaching 25.5% by 2000 and 27.4% in 2012. Also, Euromonitor International reports that this phenomenon is most distinct in Northern European countries such as Sweden (47%) and Norway (40%).

²⁴ According to a survey of 570 female one-person households by the City of Seoul in 2012, the fear of crime (e.g. rape and sexual assaults) was the most difficult problem for 77% of the respondents.

²⁵ Due to unavailability of $\mathrm{S}\mathrm{T}\mathrm{A}\mathrm{T}\mathrm{U}\mathrm{S}2$ after 2010, we applied an extrapolation for years 2010 and 2011.

²⁶ Since the data of job gains among females were unavailable across regions, we approximated it by calculating annual differences in total female workers at all-size establishments. The differences would reflect the new employment across regions and, thus, the differences in the female’s economic status.

²⁷ To the best of our knowledge, this is the first attempt to include ${\mathrm{P}}_{\mathrm{p}}$ in studies of the nexus between violent crime and female victimization.

²⁸ A possible reason for the greater deterrence of ${\mathrm{P}}_{\mathrm{p}}$ is that the sample average of ${\mathrm{P}}_{\mathrm{p}}$ is only about the half of ${\mathrm{P}}_{\mathrm{a}}$. We suspect that the law of diminishing marginal returns would hold in the production of deterrence. We have not undertaken the same task, but our general sense is that public prosecutors and their inputs, among different law enforcers in the entire criminal justice system, may play the most dominant role in explaining the variation of crime rates in Korea, as was recently suggested by Entorf and Spengler (2015) for the case of Germany. On average, the total number of prosecutors, who have the exclusive authority over public prosecution in Korea, is only about 1260 nationwide for a country that has a population of nearly 50 million.

²⁹ Because violent crimes increased during the sample period, part of the estimates may reflect an upward trend in violent crimes. To deal with potential bias, we included a dummy variable that is 1 for the years 2005, 2007–2010 – a period of surge in violent crimes.

³⁰ Disappointed by this result, we attempted to use alternative wages or income (e.g. another industry’s wages, family wages, regional income, etc.), but we did not have significant estimates.

³¹ For instance, Saridakis (2004) argued that alcohol consumption should not be omitted from any violent crime specification when modelling violent offences.

³² As is customary practice in the literature, we conducted the endogeneity test for deterrence variables (${\mathrm{P}}_{\mathrm{a}}$ and ${\mathrm{P}}_{\mathrm{p}}$) in the two final specifications (i.e. $\mathbf{F}\mathbf{E}\mathbf{M}\mathrm{\_}\mathbf{E}2\mathrm{\_}\mathbf{S}2$ and $\mathbf{F}\mathbf{E}\mathbf{M}\mathrm{\_}\mathbf{E}3\mathrm{\_}\mathbf{S}3$). Following Kim and Kim (2015), in terms of the instrumental variables for ${\mathrm{P}}_{\mathrm{a}}$, we used ‘police manpower per 100,000’ and ‘one-year lagged ${\mathrm{P}}_{\mathrm{a}}$’. For ${\mathrm{P}}_{\mathrm{p}}$, ‘the number of prosecutors per 100,000’ and ‘one-year lagged ${\mathrm{P}}_{\mathrm{p}}$’ were used. The Durbin–Wu–Hausman (DWH) tests confirmed exogeneity for ${\mathrm{P}}_{\mathrm{a}}$ and ${\mathrm{P}}_{\mathrm{p}}$. Furthermore, we also tested potential endogeneity for the female characteristics. $\mathrm{C}{\mathrm{R}}^{\mathrm{v}}$ rose over the period 2000–2011, and the values of female characteristics (i.e. female exposure and economic status) overall increased, too. This co-movement indicated a potential problem of omitted variables. In terms of instrumental variables for the six female-related variables, we used, for each variable, the ‘one-year lagged value’ and the ‘first-differenced value’ (Doyle, Ahmed, and Horn 1999; Fajnzylber, Lederman, and Loayza 2002a). The DWH statistics (${\chi }^2$) ranged from 0.23 to 1.51, confirming exogeneity of the six variables. Detailed test results are available upon request.

³³ Furthermore, we found the reporting rates for the UK in Brand and Price (2000). We recalculated the variables in question by the multipliers therein. We were then surprised to discover that re-estimation results were largely similar to what we obtained based on the reporting rates of Korea. We are grateful to an anonymous referee for motivating us to undertake this task of facilitating the reporting rates.

³⁴ There is severe multicollinearity between $\mathrm{G}\mathrm{I}\mathrm{N}\mathrm{I}$ and the interaction term (correlation coefficient = 0.873). To overcome this purely ‘statistical problem’, we monotonically transformed $\mathrm{E}\mathrm{X}\mathrm{P}\mathrm{O}\mathrm{S}\mathrm{U}\mathrm{R}\mathrm{E}2$ to a discrete variable that takes a value of 1 if its value is within the top 20% in the sample. The correlation coefficient was then 0.143 so that the multicollinearity problem disappeared. This method was used for $\mathbf{I}\mathbf{N}\mathbf{T}\mathbf{E}\mathbf{R}\mathbf{A}\mathbf{C}\mathbf{T}2$ where the original correlation coefficient was as high as 0.927, and was also applied for $\mathrm{S}\mathrm{T}\mathrm{A}\mathrm{T}\mathrm{U}\mathrm{S}2$, $\mathrm{E}\mathrm{X}\mathrm{P}\mathrm{O}\mathrm{S}\mathrm{U}\mathrm{R}\mathrm{E}3$, and $\mathrm{S}\mathrm{T}\mathrm{A}\mathrm{T}\mathrm{U}\mathrm{S}3$.

³⁵ CPTED is a multi-disciplinary approach to deterring criminal behaviour through environmental design for cities and architecture. The basic idea of CPTED is to use interdisciplinary knowledge and creativity to build surroundings in ways that reduce the fear from, and prevent the incidence of, crimes (Cozens 2008; OECD 2011).

³⁶ A series of criminological literature suggests that voting is an important indicator for a reduction in female victimization since it leads to political avenues to improve female safety (Lee 2008; Xie, Heimer, and Lauritsen 2012). Although the degree of effectiveness regarding this argument may be subject to empirical verification in Korea, more serious research on this area also seems to be of benefit to derive efficient policies in the longer term.

Additional information

Funding

This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea grant: [Grant Number NRF-2015S1A5A2A03049110].