The effects of detrimental drinking patterns and drug use on female homicide victimization rates across Europe

Abstract

Spanish

French

This study examines the extent to which drinking patterns and drug use can explain cross-national variations in female homicide victimization across Europe. Given the limitations in measuring femicide consistently across a large number of countries, this study uses mortality data on recorded female deaths as a proxy to explain differences in levels of this kind of violence against women across Europe. In particular, it focuses on national-level patterns of alcohol and drug use as predictors, controlling for other known structural correlates. Contrary to findings of previous studies, cultural drinking patterns were not significantly related to female homicide victimization in this sample of countries, but detrimental drug use was.

Keywords

Alcohol drugs Europe femicide gender homicide violence

Cross-national comparisons of social problems, including femicide, are important for situating the phenomena within a broader social context to raise awareness of the social, political, and cultural factors that shape social interactions at the individual level. Such macro-level analyses allow researchers and practitioners to develop more comprehensive policy solutions to these problems (Lynch and Pridemore, 2011). Unfortunately, comparing femicide across a large number of countries, even within one region, is difficult because the level of detail needed to distinguish femicides from other homicides with female victims is not currently part of the protocols of international data collection agencies. However, we can use recorded female homicide victimization rates as a proxy for measuring the volume of femicides across a large number of countries. By examining patterns and correlates of recorded female homicide victimization rates across Europe, we can establish a baseline portrait of lethal violence against women that can later be refined when better cross-national femicide data become available.

Homicide is typically considered a male phenomenon because in all countries that report data, males comprise the majority of both victims and offenders. According to the United Nations Office on Drugs and Crime (UNODC), in 2011 ‘the global male homicide rate [was] almost four times that of females (9.7 versus 2.7 per 100,000)’ (UNODC, 2013a: 13). Nonetheless, females still comprise one-fifth of all known homicide victims worldwide and the percentage of female victims in Europe is the second highest (28%) in the world, just behind Asia (29%) (UNODC, 2013a: 13).

Across Europe, there is considerable variation in the number of recorded female homicide victimizations. Police data collected by the United Nations show that in 2010 women constituted approximately 50% of known homicide victims in the Czech Republic, Denmark, Norway, Croatia, Slovenia, Germany, Luxembourg, and Switzerland. In contrast, women were homicide victims in fewer than 20% of the recorded cases in Albania, Bulgaria, Ireland, and Moldova (UNODC, 2013b).

The few studies that have examined variations in recorded female homicide victimization rates have emphasized structural predictors, such as material resources, the demographic composition of the population, social integration and social control, and gender roles and status. Very little attention has been paid to the macro-level effects of drinking cultures in cross-national analyses, despite the fact that intra-national studies have shown an important relationship between detrimental drinking patterns and violence (Parker and Auerhahn, 1998; Pridemore, 2002, 2004; Room et al., 2005) and none, to the author’s knowledge, has considered serious drug use.

For example, individual-level studies conducted in several nations have shown that alcohol and drugs often play a role in violence against women (e.g., Chervyakov et al., 2002; Dawson and Gartner, 1998; Mouzos and Makkai, 2004; Sela-Shayovitz, 2010; Sharps et al., 2001), yet the extent to which national-level differences in substance use, cultural behaviors associated with alcohol and drug use, and state policies regulating substance sales and use can explain variations in violence against women across countries has not been well established. By linking ‘women’s victimization to an ecological framework’ we can examine ‘factors that are not only individual, but societal and community based’ (Barberet, 2014: 95).

Both detrimental drinking patterns and drug use should be considered as correlates in cross-national homicide studies for three reasons: (1) the physical and psychological effects of serious drinking and drug use could increase the likelihood of aggressive behavior (Goldstein, 1985), (2) serious substance use, particularly drug use, could increase antisocial and aggressive behavior related to the economic pressures of continual use (Goldstein, 1985), and (3) serious substance use can loosen self- and social controls over violent behavior in certain social settings (Parker, 1998; Pridemore, 2002, 2004). Additionally, given the fact that female homicide victimization is more likely to result from family-related and intimate partner violence than male victimization, detrimental substance use is likely to negatively affect family material resources and social integration, thereby increasing women’s risk for violent victimization. This study empirically tests how well detrimental drinking patterns and drug use can explain variations in recorded female homicide victimization across a sample of 34 European countries, controlling for other known correlates.

Theoretical framework and prior literature

Most quantitative, cross-national homicide studies examine structural correlates derived primarily from a combination of strain, social disorganization, and routine activities theories of crime (Gartner, 1990; Howard et al., 2000; Kivivuori et al., 2012; Neapolitan, 1997; Stamatel, 2009a, 2009b). The most popular theoretical framework using this integrated approach was developed by Gartner, who argued that ‘the characteristics of a population commonly associated with the homicide rate can be grouped into four categories, which constitute the exogenous structural and cultural contexts for homicide’ (1990: 94). The first category, the material context, draws primarily from Durkheim’s anomie theory (1951 [1897]) and considers the absolute and relative deprivation of economic resources as stressors motivating antisocial and aggressive behaviors. The integrative context is derived from social control and social disorganization theories and includes institutions of social control and mechanisms of social integration that serve to bond people to the normative social order. When these ties and institutions are weakened, then violence is likely to increase. The demographic context captures the composition and activities of the population. Informed by routine activities theory, certain types of individuals (e.g., young males) are at higher risk for violent offending and victimization in some places and certain kinds of behaviors (e.g., working outside of the home) increase opportunities for criminal behavior. Finally, the cultural context, in Gartner’s (1990) formulation, considers exposure to legitimate forms of violence, such as wars or the death penalty, to increase the risk for illegitimate violence due to modeling and normative expectations of violence as a means of conflict resolution.

Structural predictors of the first three of Gartner’s contexts of violence have been tested extensively in the quantitative, cross-national homicide literature, but much less attention has been given to the cultural context due to the challenges of measuring culture quantitatively. Recent reviews of the literature have shown quite mixed results regarding the extent to which these structural correlates meet theoretical expectations regarding their relationships to cross-national homicide variations (Nivette, 2011; Pridemore and Trent, 2010; Trent and Pridemore, 2012). The most empirical support has been found for the material and integrative contexts, which will thus serve as the starting point for the theoretical model adopted for this study.

In order to account for differences in the extent of male and female homicide victimization across countries, Gartner et al. (1990) reformulated the demographic context, described above, as the situational context. Drawing from routine activities theory, this context captures the extent to which people are exposed to potential offenders, thereby increasing their risk for criminal victimization. In order to explain differences in rates of female homicide victimization across countries, Gartner et al. (1990) argued that in countries where more women had non-traditional gender roles, their exposure to potential offenders, and hence their risk of violent victimization, was greater than in countries with a stricter adherence to traditional gender roles. However, these authors also argued that this relationship would be mitigated by increasing women’s status over time through long-term social and economic gains.

The empirical support for gender roles as a correlate of recorded female homicide victimization cross-nationally has been fairly consistent (Cutright and Briggs, 1995; Gartner et al., 1990; Stamatel, 2014; Stamatel, under review); however, the empirical evidence regarding the relationship between female status and recorded female homicide rates has varied by measures of status and sample composition (Chon, 2013; Stamatel, 2014; Yodanis, 2004). Given that this study focuses on recorded female homicide victimization, the theoretical framework will include this situational context, in addition to the material and integrative contexts, as described above.

Finally, the cultural context as presented in Gartner’s (1990) original study was operationalized in terms of the number of battle deaths and the existence of the death penalty, as indicators of the extent to which violence is sanctioned by the state. These indicators are not particularly relevant for a sample of European countries in the 21st century. Instead, the cultural context for this study will be formulated in terms of detrimental drinking patterns and drug use.

The few studies that have examined the macro-level relationship between alcohol consumption and homicide rates have emphasized the cultural underpinnings of drinking behavior. For example, Parker argued that ‘alcohol selectively disinhibits violence depending on contextual factors such as characteristics of the situation, the actors involved, relationships among the actors, and the impact of bystanders’ (1998: 12). Additionally, Pridemore noted that ‘social tolerance for heavy drinking can result in lower self- and informal social control over drinking behavior’ (2004: 1035).

The limited empirical research that exists has established a relationship between different kinds of drinking cultures and rates of violence. Parker (1998) distinguished between a dry drinking style, characterized by spirits consumption and binge drinking, which is more likely to lead to violence, and a wet drinking style, typified by wine consumption in frequent but moderate amounts, which is less likely to facilitate violent behavior. His research showed that countries with a mixed drinking style, exhibiting characteristics of both dry and wet cultures, had higher rates of violence than other countries ‘because of the ambivalent nature of alcohol’s place in the social and cultural framework’ (Parker, 1998: 16). Additionally, he argued that this mixed pattern of drinking increased the risk of violence for women because the dry drinking style, while dangerous, is often adopted by men in segregated social situations, whereas the mixed style could allow for more interaction between sexes during risky drinking activities.

In contrast, Rossow (2001) found that among European nations, the relationship between alcohol consumption and homicide was strongest in the Nordic countries because of a more ‘explosive’ drinking pattern characterized by greater intoxication compared to moderate drinking (2001: 79). Rossow also found that the effect of alcohol consumption on homicide was greater for males than females because men more frequently drink heavily compared to women.

Finally, research examining differences in recorded homicide rates between Western and Eastern Europe have attributed some of these differences to different drinking cultures. Pridemore (2002, 2004) has shown that heavy episodic drinking of distilled spirits in private or semi-private settings contributes to the high homicide rate in Russia. In a cross-national study, Bye found that post-communist countries had a stronger relationship between alcohol consumption and homicide compared to Western Europe because of ‘more detrimental drinking patterns’ (2012: 240). Stamatel (under review) also found that higher levels of spirits consumption in Eastern European countries was positively associated with homicide compared to Western Europe and Central Asia. Both Bye (2012) and Stamatel (under review) found the same relationship for both male and female homicide victimization rates.

Interestingly, cross-national research on the relationship between detrimental drug use and homicide rates is virtually nonexistent. Parker (1998) noted that this relationship is difficult to study cross-nationally because of large differences in attitudes toward drug users, the legality of certain drugs, and regulatory attention paid to illegal drugs across countries. While there is undoubtedly significant variation in attitudes toward regulation of some kinds of drugs, there is more normative consistency regarding detrimental drug use leading to overdose deaths as a global health problem (UNODC, 2014). Some patterns of detrimental drug use mirror those of detrimental alcohol use with the majority of heavy users being male and greater mortality due to drugs found in Eastern Europe compared to Western Europe, so it is reasonable to expect severe drug abuse to have the same effect on macro-level homicide variation as detrimental alcohol use.

Parker and Auerhahn (1998) cited Goldstein’s (1985) typology as a starting point for understanding the relationship between detrimental drug use and homicide. He proposed three ways that drug use and drug trafficking could influence violent behavior: (1) psychopharmacological effects of certain drugs could increase propensities toward violence, (2) economic-compulsive violence that stems from the need or desire to obtain drugs, and (3) systemic violence that often accompanies the business of drug sales and distribution.

Much like the argument for detrimental alcohol use, detrimental drug use can increase the material motivations to engage in violence, weaken social integration, particularly within families where women are more likely to be victims of violence, and shape opportunities, especially for men, to engage in violence. Research on the relationship between drug use and homicide at the macro-level has been hampered by the lack of reliable and comparable information about drug use across countries, but estimates of drug-related deaths are a good proxy for detrimental drug use, at the extreme end of the substance use continuum.

To summarize, this article starts with the theoretical framework proposed by Gartner (1990) to explain cross-national homicide victims and expands and adapts it in three ways. First, it focuses on the parts of the original model that have received the most empirical support (material and integrative contexts). Second, it includes factors that were developed in other works (Gartner et al., 1990) to specifically explain recorded female homicide victimization (situational context). Finally, it reformulates the cultural context to focus on factors relevant to contemporary Europe and often neglected in the literature (Parker, 1998). Figure 1 illustrates the theoretical framework for this study and shows how the concepts will be operationalized for the empirical analysis. A fuller description of the variables is provided in a later section.

Figure 1.

Theoretical model.

Hypotheses

This study examines the extent to which the material, integrative, situational, and cultural contexts can explain recorded cross-national female homicide victimization rates across Europe. The first hypothesis tests the first three contexts, which have been shown to be relevant correlates of total homicide victimization and/or female homicide victimization in the existing literature. The material context is defined as economic deprivation that produces social strain that could lead to violence, which for this study is measured by level of economic development and poverty level. The integrative context is defined as the strength of social institutions that tie people to the normative order, such as families. The situational context is defined by gender roles and female status that are expected to shape routine activities that affect the risk of violent victimization.

Hypothesis 1: Countries with lower levels of economic development, higher levels of poverty, less family integration, less traditional gender roles, and lower female status will have higher recorded female homicide rates than countries with higher levels of economic development, lower levels of poverty, more family integration, more traditional gender roles, and higher female status.

The more important contribution of this study to the existing literature is the reformulation of the cultural context in terms of permissiveness for or acceptance of detrimental drinking patterns and drug use. Previous studies have used sales data for different alcoholic beverages to approximate drinking styles. This study instead uses a classification of the riskiness of a country’s drinking pattern to better capture the social and cultural dimensions of alcohol consumption:

Hypothesis 2: Countries with riskier drinking patterns will have higher recorded female homicide rates than those with less risky patterns, controlling for other structural correlates of homicide.

Similarly, it is expected that detrimental drug use will have a similar effect on female homicide rates as detrimental alcohol use:

Hypothesis 3: Countries with more detrimental drug use will have higher recorded female homicide victimization rates than those with less detrimental drug use.

Data and measures

As mentioned in the introduction, femicide is very difficult to measure reliably across a large number of countries because the information needed to determine whether a homicide is femicide is not routinely recorded by international, and even sometimes national, data collection agencies. The term femicide implies intent when defined as ‘the killing of women because they are women’ (Academic Council of the United Nations System, 2014), and also knowledge of the gender of both victims and offenders if defined as ‘the killing of females by males because they are female’ (Radford and Russell, 1992, original emphasis). Unfortunately, cross-national homicide data rarely include enough information to distinguish between femicide and female homicide victimization for a large number of countries, although there have been some improvements in data collection across Europe (see Liem et al., 2013; Martinez and Schrottle, 2006; Palma-Solis et al., 2008).

Nonetheless, we can reasonably use female homicide victimization as a proxy for femicide as national and international data have shown that a substantial proportion of female homicides is committed within the context of family-related and/or intimate partner violence, thereby establishing the gendered nature of violence that distinguishes femicide from homicide (Nikolic-Ristanovic, 2004; Stöckl et al., 2013; UNODC, 2013a). Although this measure is only a rough approximation of femicide until better data become available, by examining patterns and correlates of female homicide victimization we can establish a baseline for understanding violence against women in a cross-national context.

Consistent with other recent literature studying cross-national homicide variation, the dependent variable is the female homicide victimization rate per 100,000 people recorded by the World Health Organization. This is generally considered by criminologists to be a more reliable and valid measure than police-recorded crime statistics because the determination of the cause of death is made by medical examiners rather than police and, therefore, is not subject to known police biases in recording crime statistics (Howard et al., 2000; LaFree, 1999; LaFree and Drass, 2002; Stamatel, 2009a). There are still organizational and definitional issues that can create inconsistences in the recordings of deaths cross-nationally, but these are generally seen to be less egregious than police biases within the cross-national research community (LaFree and Drass, 2002; Neuilly, 2011; Stamatel, 2009a).

In order to compare the results of this study to two other recent cross-national analyses of European homicide patterns, the dependent variable was constructed as a five-year average from 2003 to 2007. Multiple-year averages are often used for fairly rare events in order to smooth annual fluctuations in the data. The two other recent studies had used pooled data from four time periods circa 1990, 1995, 2000, and 2005 (Stamatel, 2014; Stamatel, under review). This study is a cross-sectional analysis only of the last time period because the key independent variables measuring alcohol and drug use are not available for such a long time series. Table 1 shows the recorded female homicide victimization rate for the 34 countries in the sample circa 2005. They range from a low of 0.22 in the United Kingdom to 11.01 in Russia.

Table 1.

Point estimates for key variables by country.^a

Country	Average female homicide victimization rate	Drinking pattern (coded 1–4)	Drug mortality rate
Albania	1.28	Medium (3)	4.0
Austria	0.69	Least (1)	34.6
Belarus	5.00	Very (4)	17.9
Belgium	1.42	Least (1)	10.9
Bulgaria	1.08	Somewhat (2)	7.3
Croatia	1.18	Medium (3)	30.1
Cyprus	0.64	Least (1)	15.4
Czech Republic	0.86	Medium (3)	16.5
Denmark	0.63	Somewhat (2)	76.1
Estonia	3.74	Medium (3)	177.2
Finland	1.20	Medium (3)	91.7
France	0.42	Least (1)	7.1
Germany	0.54	Least (1)	22.7
Greece	0.53	Somewhat (2)	14.2
Hungary	1.56	Medium (3)	4.2
Iceland	2.06	Somewhat (2)	220.7
Italy	0.37	Least (1)	9.4
Latvia	0.53	Medium (3)	5.2
Lithuania	4.47	Medium (3)	22.2
Macedonia, FYR	1.59	Medium (3)	15.9
Moldova	5.12	Very (4)	17.4
Netherlands	0.56	Least (1)	12.5
Norway	0.77	Medium (3)	56.5
Poland	0.76	Medium (3)	7.9
Portugal	0.72	Least (1)	3.8
Romania	1.67	Medium (3)	2.1
Russia	11.01	Very (4)	89.8
Slovakia	1.15	Medium (3)	28.2
Slovenia	0.98	Somewhat (2)	19.8
Spain	0.62	Least (1)	45.3
Sweden	0.31	Somewhat (2)	66.2
Switzerland	0.75	Least (1)	25
Ukraine	5.87	Very (4)	43.3
United Kingdom	0.22	Medium (3)	52.8

None of the apparent outliers in this table were determined to be influential cases for the regression analyses, as determined using Cook’s D.

In order to test Hypothesis 1, the material context is operationalized by two variables. Gross domestic product (GDP) per capita is a standard measure of macro-level economic well-being in the cross-national homicide literature. Economic deprivation is measured by the infant mortality rate, which is a common measure of poverty. There is some debate in the cross-national homicide literature about whether absolute deprivation (poverty) or relative deprivation (income inequality) is more salient for driving motivations toward violence (Messner et al., 2010; Pridemore, 2008, 2011). Because both measures are currently used in the literature, they were both tested in preliminary analyses for this study. The Gini index, which is the most common measure of income inequality, was not statistically significant – compared to other cross-national studies with global samples, the Gini index for European countries may not vary enough to be significant for this sample – so it was dropped from the final model to preserve degrees of freedom for a small-sample analysis. Both GDP/capita and the infant mortality rate, which were retained for the final analyses, were obtained from the World Bank’s World Development Indicators (WDI; World Bank, 2014).

Consistent with previous studies, the integrative context is measured with the crude divorce rate (number of divorces per 1000 population) from the United Nations Demographic Yearbook (2015). For the situational context, traditional gender roles were operationalized by the adolescent fertility rate from the WDI, which is the number of births per 100,000 women aged 15–19, as women in traditional societies are more likely to marry younger and bear more children than women in less traditional societies who may delay family roles due to educational or career goals. Female status, capturing long-term social and economic gains for women, was measured as the percentage of a country’s tertiary school enrollments who are women, also from the WDI. All of these variables are common measures for structural covariates of cross-national homicide differences (see LaFree, 1999; Nivette, 2011; Pridemore and Trent, 2010; Trent and Pridemore, 2012, for more detailed reviews of this literature and the related measures).

In order to test Hypotheses 2 and 3, new measures were added to capture detrimental drinking and drug use. In previous cross-national studies, alcohol styles were typically measured by the amount of consumption of certain types of alcoholic beverages, with high levels of spirits consumption being the most likely to be associated with violence (Bye, 2012; Rossow, 2001; Stamatel, under review). This measure, however, is problematic in that it only captures recorded alcohol consumption, which is kept in sales and taxation records, and it ignores unrecorded consumption, which can include homemade products, illegally distributed products, or those purchased in other countries than where they were consumed (WHO, 2014). Unrecorded consumption is believed to be linked to heavy drinking occasions, which has implications for the study of violence (Rehm et al., 2014). The WHO estimates unrecorded consumption based on population-weighted data from countries that provided this information, but they are still not the best indicators of drinking culture, just consumption. Even the WHO acknowledges that ‘[m]easuring drinking patterns to account accurately for the impact of alcohol consumption on people’s health and wellbeing is more complex than simply ascertaining the amount of alcohol consumed’ (2014: 35). Instead, Rehm et al. (2013) developed a patterns of drinking index to measure how people drink, not just how much they drink. The scale ranges from 1 (least risky pattern of drinking) to 4 (most risky pattern of drinking) and is comprised of the following measures: ‘the usual quantity of alcohol consumed per occasion, festive drinking, proportion of drinking events when drinkers get drunk, proportion of drinkers who drink daily or nearly daily, drinking with meals, and drinking in public places,’ where the items of drinking daily and drinking with meals correspond to less risky behavior while the other items indicate more risky behaviors (WHO, 2014: 35; see also Rehm et al., 2013). The drinking patterns for the 34 countries in this study are listed in Table 1.

Data for detrimental drug use were obtained from the United Nations Office on Drugs and Crime’s (UNODC) World Drug Report (2014). It is difficult to obtain comparable drug use and abuse data cross-nationally, as discussed above, so the drug mortality rate per 100,000 people aged 15–64 from the UNODC was used as a measure of detrimental drug use. There could be severe drug abuse that does not necessarily lead to death, either due to type of drug use or medical interventions, but that could precipitate violence, which would not be captured by this measure. However, because of significant variations in public attitudes toward drugs and legal enforcement of drug violations, there is currently not a better, standardized measure across a large number of countries (Parker, 1998). How the sample countries fare on this measure is also available in Table 1.

Table 2 provides descriptive statistics for all of the independent variables and Table 3 shows the bivariate correlations. Multicollinearity is often a problem with aggregated data. Table 3 shows that GDP/capita, the infant mortality rate, and the adolescent fertility rate are highly correlated, which is not surprising given that they are all indicators of level of development. The implications of these high correlations on the regression analyses is discussed in the section below.

Table 2.

Descriptive statistics for all variables.

Variable	Mean	SD	Min.	Max.
Average female homicide victimization rate	1.91	2.28	0.22	11.01
GDP/capita	222.33	179.31	8.31	657.67
Infant mortality rate	7.19	4.98	2.4	19.80
Divorce rate	2.27	0.93	0.08	4.20
Adolescent fertility rate	16.65	9.47	4.78	41.98
Females as % of tertiary school enrollment	69.16	21.73	29.63	101.69
Drinking pattern (1 = least risky, 4 = very risky)	2.35	1.04	1.00	4.00
Drug mortality rate	37.47	48.01	2.10	220.70

Table 3.

Zero-order correlations for all variables.

	1.	2.	3.	4.	5.	6.	7.	8.
1. Female homicide victimization rate	1.00
2. GDP/capita	−0.49	1.00
3. Infant mortality rate	0.51	−0.70	1.00
4. Divorce rate	0.64	−0.08	0.15	1.00
5. Adolescent fertility	0.52	−0.67	0.80	0.26	1.00
6. Females as % of tertiary school enrollment	0.21	0.29	−0.40	0.26	−0.27	1.00
7. Drinking pattern	0.64	−0.56	0.58	0.41	0.62	0.14	1.00
8. Drug mortality rate	0.21	0.35	−0.25	0.19	−0.02	0.43	0.13	1.00

Results

The regressions were estimated using ordinary least squares (OLS). The dependent variable was highly skewed to the right, so it was logged to correct for heteroskedasticity. Table 4 shows the regression coefficients for four models. Model 1 tested Hypothesis 1 regarding the material, integrative, and situational contexts that have been established as affecting cross-national female homicide victimization in previous studies. The first two significant coefficients support existing studies. Countries with fewer material resources (lower GDP/capita) and less social integration (higher divorce rates) have higher recorded female homicide victimization rates. Female status, as measured by higher education enrollment, is significant but positive, which is contrary to expectations. Gartner et al. (1990) argued that higher status would decrease women’s risk for violent victimization, although subsequent studies have produced mixed results regarding the relationship between female status and homicide victimization cross-nationally (Chon, 2013; Cutright and Briggs, 1995; Stamatel, 2014; Stamatel, under review; Yodanis, 2004). These inconsistencies can be attributed, in part, to difficulties finding appropriate measures of female status. In this case, although female participation in higher education has been used in previous studies as a measure of status, it is not clear that this variable measures status separately from gender roles. If interpreted as a measure of non-traditional gender roles, then the positive finding is consistent with the theoretical expectation that when women do not conform to traditional gender roles, they are at greater risk for violent victimization.

Table 4.

OLS regression coefficients and standard errors.

Context	Variable	Model 1	Model 2	Model 3	Model 4
Material	GDP/capita	−0.002* (0.001)	−0.002 (0.001)	−0.003** (0.001)	−0.003** (0.001)
	Infant mortality rate	0.067 (0.039)	0.057 (0.040)	0.078* (0.037)	0.082* (0.035)
Integrative	Divorce rate	0.379* (0.123)	0.349* (0.128)	0.359** (0.115)	0.370** (0.109)
Situational	Adolescent fertility	0.001 (0.019)	−0.003 (0.020)	−0.020 (0.019)	−0.019 (0.018)
	Females as % of tertiary school enrollment	0.012* (0.006)	0.010 (0.006)	0.006 (0.006)	0.007 (0.005)
Cultural	Drinking pattern		0.147 (0.157)	0.054 (0.146)
	Drug mortality rate			0.007* (0.002)	0.007** (0.002)
	Constant	−1.594 (0.580)	−1.613 (0.582)	−1.075 (0.560)	−1.051 (0.548)
	R²	0.66	0.67	0.74	0.74

p < 0.05, **p < 0.01, ***p < 0.00.

Two variables, infant mortality rate and adolescent fertility, were not statistically significant, as they have been in other studies (Gartner, 1990; Stamatel, 2014; Stamatel, under review). This is likely due to multicollinearity, which makes it more difficult to reach statistical significance (Gujarati and Porter, 1998). As mentioned above, Table 3 shows that GDP/capita, infant morality, and adolescent fertility are highly correlated. Infant mortality and adolescent fertility also had the highest variance inflation factors (above 3) in all of the models, verifying the presence of multicollinearity. Previous studies where these variables have been significant for European nations have used pooled time series data, which not only increases the sample size, but also the amount of variation for each variable, somewhat mitigating multicollinearity.

Model 2 includes the same variables as Model 1 and adds the measure of cultural drinking patterns. Contrary to theoretical expectations, the drinking variable is not statistically significant. An F-test also indicates that Model 2 is not a better fit for the data than Model 1. The drinking pattern variable is a better measure of the cultural aspects of drinking that are related to violence than consumption measures, but the four-point scale of the drinking pattern variable is weakly continuous, which could account for the lack of statistical significance. Two other measures of detrimental drinking patterns were tested to see if they produced better results. A dummy variable of less risky versus more risky drinking patterns was created from the original four-point scale. Additionally, a continuous measure of spirits consumption from the WHO, which has been used in previous studies (Bye, 2008, 2012; Rossow, 2001; Stamatel, under review), was tested. Neither of these alternatives was statistically significant (results not shown). The other studies that have found a relationship between spirits consumption and total and/or female homicide victimization had different sample compositions and used different estimation techniques (time series or pooled time series). In this cross-sectional sample of 34 European nations, detrimental drinking culture is not significantly related to female homicide victimization.

Finally, Model 3 adds the drug mortality rate as a measure of detrimental drug use. This variable is positive and significant, as expected, indicating that higher levels of detrimental drug use correspond to higher recorded female homicide victimization rates. Once again, the variable measuring detrimental drinking patterns is not significant. Additionally, in this model, GDP/capita and the divorce rate are still statistically significant, and now so is the infant mortality rate. Interestingly, neither of the variables capturing the situational context for female homicide victimization is significant. As discussed above, the lack of significance for the adolescent fertility rate could be due to multicollinearity. Nonetheless, the surprisingly weak performance of the situational context that was theoretically constructed to explicitly explain gender differences in violent victimization certainly warrants further consideration of both theoretical mechanisms and macro-level measures to explain cross-national differences in female homicide victimization.

Given that detrimental drinking patterns were not statistically significant in either Model 2 or 3, an F-test was performed comparing Model 3 with the drinking pattern variable to Model 4 without it. The test showed that Model 4 was the better fitting model for these data. This is also evident from the fact that the R² does not improve with the addition of the drinking variable when comparing Models 3 and 4. Additionally, the p-value for the drug mortality rate in Model 4 decreases when the drinking variable is removed.

Discussion and conclusions

This study started with a theoretical framework originally developed by Gartner (1990) and elaborated upon by Gartner et al. (1990) and Parker (1998) that explained differences in female homicide victimization rates in terms of four social and cultural contexts: material, integrative, situational, and cultural. More specifically, I argued that detrimental drinking patterns and drug use are important aspects of the cultural context of nations that are often overlooked in quantitative, cross-national analyses of female victimization.

Both detrimental drinking and drug use can contribute to violence through physical, psychological, economic, and situational effects (Goldstein, 1985; Parker 1998; Parker and Auerhahn, 1998; Pridemore, 2002, 2004; Rossow, 2001). Additionally, detrimental substance use is likely to diminish family material resources and social integration and increase marital strain, thereby increasing women’s risk for violent victimization in particular. This study expanded upon previous ones by using a more nuanced measure of detrimental drinking to capture the social and cultural contexts of drinking behaviors that are potentially dangerous. It was also the first study to consider detrimental drug use as a predictor of recorded, cross-national female homicide victimization.

The results from this study produced three new insights about correlates of female homicide victimization in Europe. First, the lack of significance of the cultural drinking pattern variable, and the alternate measures, was quite surprising and contrary to previous findings. Some of the inconsistent findings in the cross-national crime literature are likely due to variations in sample composition, time frames, and type of analysis (Stamatel, 2006, 2009a). In fact, this study is different from previous ones focusing on Europe in all of these respects.

Additionally, while the drinking pattern scale attempts to capture the social and cultural aspects of drinking across countries, the reduction of very interesting components into a four-point scale may be obscuring precisely the nuances we are aiming to measure. For example, in this sample, 28 of the 34 countries fall in the two middle categories of the scale. Case-level data, such as that collected by the European Homicide Monitor, would be helpful to better understand the relationship between alcohol and homicide and to better identify true femicide cases, but these data are not currently available for a large number of countries (Liem et al., 2013).

Second, the significant effect of detrimental drug use on recorded female homicide victimization is a new finding, as this variable has not been examined in the cross-national homicide literature. The theoretical expectations for this variable were similar to those for detrimental drinking so it is interesting that one is statistically significant while the other is not. The Global Study on Homicide 2013 produced by the UNODC acknowledges the connection between detrimental substance use and homicide, but the causal mechanisms are vague. As this report states, ‘[i]t is particularly challenging to disentangle the various components of illicit drug-related homicide, as this type of violence is the product of the interaction of a variety of different complex causes and underlying factors’ (UNODC, 2013a: 73). We need better data, and better theories, to begin to disentangle these processes. Additionally, we need to consider how gender intersects with other factors. For example, the European Homicide Monitor data show that in Finland, 15.5% of male homicide victims and 12.6% of female homicide victims were under the influence of illicit drugs, which ‘is indicative of the multiple effects that illicit drugs can have on violence’ (UNODC, 2013a: 74). We need to more clearly conceptualize, and eventually test, both the gendered nature of social and cultural patterns related to female homicide victimization, as well as the gendered consequences of such patterns that contribute more specifically to femicide.

Finally, the fact that there was very little support for the situational context in this study warrants further consideration, as it was the context that was most clearly related to gender. Problems with multicollinearity and measurement were already discussed above, but further consideration needs to be paid to the gendered mechanisms driving female homicide variations at the macro-level. The significance of the divorce rate variable indicates that family situations matter, but what are other ways that family dynamics might shape macro-level variations? For example, other cross-national homicide studies have examined the role of social support (McCall and Brauer, 2014). Perhaps the presence or absence of government-level support for women and/or families affects the extent to which women are at risk for violent victimization. More generally, just like the broader field of cross-national criminology, research on macro-level differences in female victimization would benefit from greater attention to social process theories (Schaible, 2012).

There are several methodological limitations that must be considered when interpreting these findings. First, the sample size is small, which limits both the power of the analyses and the number of independent variables that can be tested in any given model. As discussed in the methods section, some alternative measures were tested in preliminary analyses but dropped from the final models so as to not overfit the models. Even with the small number of independent variables, the amount of variation in female homicide victimization rates across Europe explained in the best model (Model 3 in Table 4) is quite high at 74%.

Second, as discussed in the methods section, some measures of theoretical concepts are imperfect due to data availability. This is a common problem for quantitative cross-national analyses where researchers must rely on international data collection agencies for reliable and comparable data across a large number of countries (Stamatel, 2009a). This limitation is particularly salient for the insignificant findings. For example, female education as a measure of female status was not significant, contrary to theoretical expectations and some prior research. It would be premature to conclude from this analysis that female status is not related to female homicide victimization, but rather than we need to continue testing other measures of status, especially those that are most relevant in the European context.

Finally, the lack of comparable measures of femicide across a large number of countries means that we must be careful about applying these results to femicide. As with any international data source for crime, underreporting is always a concern. Additionally, while most female homicides are likely femicides, there could be other factors related to gender roles or substance abuse that would further distinguish these two types of events that cannot yet be tested with this kind of analysis. Nonetheless, this study gives us a starting point for further research on the cross-national correlates of femicide.

To conclude, the material and integrative contexts continue to be important factors to understand female homicide victimization differences across Europe. Additionally, detrimental drug use is a new correlate that deserves additional consideration in future theoretical and empirical studies. However, measures of gender roles, gender status, and detrimental alcohol use were not significant in this study, contrary to theoretical expectations. While measurement issues are always a challenge for cross-national studies, we also need more refined theories of why recorded female homicide rates vary across space and time. We also need better cross-national data to distinguish femicides from female homicide victimization in order to examine similarities and differences between these two manifestations of violence against women.

Footnotes

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Author biography

Janet P Stamatel is an Assistant Professor in the Department of Sociology at the University of Kentucky. She received her PhD in Sociology from the University of Chicago. Her research interests include cross-national crime comparisons, the consequences of macro-level social changes on social order, and quantitative methods.

References

Academic Council of the United Nations System (2013) Femicide: A Global Issue that Demands Action. Vienna: United Nations.

Barberet

(2014) Women, Crime and Criminal Justice: A Global Enquiry. New York and Abingdon: Routledge.

Bye

(2008) Alcohol and homicide in Eastern Europe: A time series analysis of six countries. Homicide Studies 12(1): 7–27.

Bye

(2012) Alcohol and homicide in Europe. In: Liem

MCA

Pridemore

(eds) Handbook of European Homicide Research. New York: Springer, pp. 231–245.

Chervyakov

Shkolnikov

Pridemore

McKee

(2002) The changing nature of murder in Russia. Social Science and Medicine 55(10): 1713–1724.

Chon

(2013) A spurious relationship of gender equality with female homicide victimization: A cross-national analysis. Crime and Delinquency. Epub ahead of print 12 July 2013. DOI: 10.1177/0011128713492497.

Cutright

Briggs

(1995) Structural and cultural determinants of adult homicide in developed countries: Age and gender-specific rates, 1955–1989. Sociological Focus 28(3): 221–243.

Dawson

Gartner

(1998) Differences in the characteristics of intimate femicides: The role of relationship state and relationship status. Homicide Studies 2(4): 378–399.

Durkheim

(1951[1897]) Suicide. New York: Free Press.

10.

Gartner

(1990) The victims of homicide: A temporal and cross-national comparison. American Sociological Review 55(1): 92–106.

11.

Gartner

Baker

Pampel

(1990) Gender stratification and the gender gap in homicide victimization. Social Problems 37(4): 593–612.

12.

Goldstein

(1985) The drugs/violence nexus: A tripartite conceptual framework. Journal of Drug Issues 15(4): 493–506.

13.

Gujarati

Porter

(2008) Basic Econometrics, 5th edn. Boston: McGraw-Hill/Irwin.

14.

Howard

Newman

Pridemore

(2000) Theory, method, and data in comparative criminology. In: Criminal Justice 2000, Volume 4: Measurement and Analysis of Crime and Justice. Washington, DC: US Department of Justice, National Institute of Justice, pp. 139–211.

15.

Kivivuori

Savolainen

Danielsson

(2012) Theory and explanation in contemporary European homicide research. In: Liem

MCA

Pridemore

(eds) Handbook of European Homicide Research. New York: Springer, pp. 139–211.

16.

LaFree

(1999) A summary and review of cross-national comparative studies of homicide. In: Smith

Zahn

(eds) Homicide: A Sourcebook of Social Research. Thousand Oaks, CA: Sage, pp. 139–211.

17.

LaFree

Drass

(2002) Counting crime booms among nations: Evidence for homicide victimization rates, 1956 to 1998. Criminology 40(4): 769–800.

18.

Liem

Ganpat

Granath

. (2013) Homicide in Finland, the Netherlands, and Sweden: First findings from the European Homicide Monitor. Homicide Studies 17(1): 75–95.

19.

Lynch

Pridemore

(2011) Crime in international perspective. In: Wilson

Petersilia

(eds) Crime and Public Policy. New York: Oxford University Press, pp. 5–52.

20.

McCall

Brauer

(2014) Social welfare support and homicide: Longitudinal analyses of European countries from 1994 to 2010. Social Science Research 48 (November): 90–107.

21.

Martinez

Schrottle

(2006) State of European research on the prevalence of interpersonal violence and its impact on health and human rights. Coordination Action on Human Rights Violations (CAHRV).

22.

Messner

Raffalovich

Sutton

(2010) Poverty, infant mortality, and homicide rates in cross-national perspective: Assessments of criterion and construct validity. Criminology 48(2): 509–537.

23.

Mouzos

Makkai

(2004) Women’s experience of male violence: Findings from the Australian component of the International Violence Against Women Survey (IVAWS). Australian Institute of Criminology.

24.

Neapolitan

(1997) Cross-national Crime: A Research Review and Sourcebook. Westport, CT: Greenwood Press.

25.

Neuilly

M-A

(2011) Impact of medico-legal practices on mortality statistics and their use in comparative research. Victims and Offenders 6(3): 306–320.

26.

Nikolic-Ristanovic

(2004) Post-communism: Women’s lives in transition. Feminist Review 76(1): 2–4.

27.

Nivette

(2011) Cross-national predictors of crime: A meta-analysis. Homicide Studies 15(2): 103–131.

28.

Palma-Solis

Vives-Cases

Álvarez-Dardet

(2008) Gender progress and government expenditure as determinants of femicide. Annals of Epidemiology 18(4): 322–329.

29.

Parker

(1998) Alcohol, homicide, and cultural context. Homicide Studies 2(1): 6–30.

30.

Parker

Auerhahn

(1998) Alcohol, drugs, and violence. Annual Review of Sociology 24: 291–311.

31.

Pridemore

(2002) Vodka and violence: Alcohol consumption and homicide rates in Russia. American Journal of Public Health 92(12): 1921–1930.

32.

Pridemore

(2004) Weekend effects on binge drinking and homicide mortality: Preliminary evidence for the social connection between alcohol and violence in Russia. Addiction 99(8): 1034–1041.

33.

Pridemore

(2008) A methodological addition to the cross-national empirical literature on social structure and homicide: A first test of the poverty-homicide thesis. Criminology 46(1): 133–154.

34.

Pridemore

(2011) Poverty matters: A reassessment of the inequality–homicide relationship in cross-national studies. British Journal of Criminology 51(5): 739–772.

35.

Pridemore

Trent

CLS

(2010) Do the invariant findings of Land, McCall, and Cohen generalize to cross-national studies of social structure and homicide? Homicide Studies 14(3): 296–335.

36.

Radford

Russell

DEH

(1992) Femicide: The Politics of Woman Killing. Woodbridge, CT: Twayne.

37.

Rehm

. (2013) The comparative risk assessment for alcohol as part of the Global Burden of Disease 2010 Study: What changed from the last study? The International Journal of Alcohol and Drug Research 2(1): 1–5.

38.

Rehm

. (2014) A systematic review of the epidemiology of unrecorded alcohol consumption and the chemical composition of unrecorded alcohol. Addiction 109(6): 880–893.

39.

Room

Babor

Rehm

(2005) Alcohol and public health. The Lancet 365(9458): 519–530.

40.

Rossow

(2001) Alcohol and homicide: A cross-cultural comparison of the relationship in 14 European countries. Addiction 96(1): 77–92.

41.

Schaible

(2012) Overcoming the neglect of social process in cross-national and comparative criminology. Sociology Compass 6(10): 793–807.

42.

Sela-Shayovitz

(2010) The role of ethnicity and context: Intimate femicide rates among social groups in Israeli society. Violence Against Women 16(12): 1424–1436.

43.

Sharps

Campbell

Gary

Webster

(2001) The role of alcohol use in intimate partner femicide. American Journal on Addictions 10(2): 122–135.

44.

Stamatel

(2006) Incorporating socio-historical context into quantitative cross-national criminology. International Journal of Comparative and Applied Criminal Justice 30(2): 177–207.

45.

Stamatel

(2009a) Contributions of cross-national research to criminology at the beginning of the 21st century. In: Krohn

Lizotte

Hall

(eds) Handbook on Crime and Deviance. New York: Springer, pp. 3–22.

46.

Stamatel

(2009b) Correlates of national-level homicide variation in post-communist East-Central Europe. Social Forces 87(3): 1423–1448.

47.

Stamatel

(2014) Explaining variations in female homicide victimization rates across Europe. European Journal of Criminology 11(5): 578–600.

48.

Stamatel

(under review) Unpacking the socio-historical context of gender-specific homicide variation across Europe and Central Asia.

49.

Stöckl

Devries

Rotstein

. (2013) The global prevalence of intimate partner homicide: A systematic review. The Lancet 382(9895): 859–865.

50.

Trent

CLS

Pridemore

(2012) A review of the cross-national empirical literature on social structure and homicide. In: Liem

MCA

Pridemore

(eds) Handbook of European Homicide Research. New York: Springer, pp. 111–135.

51.

United Nations Demographic Yearbook (2015) Demographic and social statistics [data]. Available at: unstats.un.org/unsd/demographic/products/dyb/dyb2.htm (accessed 5 April 2015).

52.

UNODC (United Nations Office on Drugs and Crime) (2013a) Global Study on Homicide. Available at: www.unodc.org/gsh/ (accessed 5 April 2015).

53.

UNODC (United Nations Office on Drugs and Crime) (2013b) United Nations surveys on crime trends and the operations of criminal justice systems [data]. Available at: www.unodc.org/unodc/en/data-and-analysis/United-Nations-Surveys-on-Crime-Trends-and-the-Operations-of-Criminal-Justice-Systems.html (accessed 5 April 2015).

54.

UNODC (United Nations Office on Drugs and Crime) (2014) World Drug Report 2014 – Use of Drugs. Available at: public.tableausoftware.com/views/WDR2014-Useofdrugs (accessed 5 April 2015).

55.

WHO (World Health Organization) (2014) Global status report on violence prevention 2014. Available at: www.who.int/violence_injury_prevention/violence/status_report/2014/en/ (accessed 5 April 2015).

56.

World Bank Group (2014) World Development Indicators [data]. Available at: data.worldbank.org/data-catalog/world-development-indicators (accessed 5 April 2015).

57.

Yodanis

(2004) Gender inequality, violence against women, and fear: A cross-national test of the feminist theory of violence against women. Journal of Interpersonal Violence 19(6): 655–675.