The Relationship Between Gender Equality and Rates of Inter- and Intra-Sexual Lethal Violence: An Exploration of Functional Form

Abstract

Considerable attention has been given to the relationship between gender equality (GE) and levels of violence perpetrated by men against women in recent decades. Yet, the current state of the literature is equivocal. High levels of GE have been hypothesized to both decrease and increase men’s violence. This study proposes a theoretical account that integrates the ameliorative and backlash theses and offers a possible explanation for some of the inconsistent findings. Specifically, we argue that both ameliorative and backlash processes operate but that their relative strength and salience vary at different levels of GE. As a result of the interplay between these counter-balancing forces, the relationship between GE and levels of men’s violence men against women (and other men) is hypothesized to be curvilinear rather than linear. The results of multivariate analyses of male and female inter- and intra-sexual homicide offending offer robust support for this hypothesis.

Keywords:

• gender equality
• homicide
• male violence

Acknowledgment

We are grateful to Lawrence Raffalovich for helpful comments on a previous draft of this manuscript.

Notes

1. Our review of the literature on the macro level causes of men’s violence against women is descriptive and as such does not take into account variation in methodological quality of studies, which could be accomplished in a formal meta-analysis. See Pratt and Cullen (2005) for an application of meta-analysis to the literature on the macro-level causes of crime more generally.

2. McCloskey (1996) found that reports of wife abuse were higher for couples in which the income gap between husbands and wives was small (as wives’ incomes approached husbands’ incomes, abuse was more likely).

3. The homicide data came from the Uniform Crime Reports: Supplemental Homicide Report, 1976-2003 (ICPSR 4351). We utilized a weight (WTIMPUS) that adjusts counts based on the UCR estimate of homicide and imputed offender data. Missing offender data (including sex) are estimated using an algorithm based on victim data (Fox, 2004).

4. Cities were retained in the sample if they had SHR data for at least one year in each time period.

5. The Gini coefficient ranges from 0 (i.e. perfect equality) to 1 (i.e. one income recipient receives all the income).

6. We included percent unemployed as a covariate following McCall et al. (2010). We found, however, that it was too collinear with other covariates to be included. When unemployment is included the coefficient does not obtain significance; further when unemployment is substituted for economic disadvantage it does obtain significance. These alternatives do not impact other results. We decided to retain economic disadvantage and drop unemployment.

aAbbreviations for inter- and intra-sexual homicides represent the offender’s sex followed by the victim’s sex.

7. The detailed results of these robustness tests are not reported but are available upon request. As noted above, we examined the results of the OLS regression of each of the four logged homicide rates on the independent and control variables as another robustness test. Though considered less appropriate given the distribution and low base rates of these dependent variables (e.g. Osgood, 2000), the OLS analyses replicated the negative binomial results for the male-female, female-male, and female-female homicides in both decades. In the case of the male-male models, a curvilinear relationship was detected in 2000 only and only without robbery held constant.

8. We explored the manifestation of the cross-sectional curvilinear relationship in the context of change analyses. In a supplemental analysis, we used OLS to regress the change in male-female homicide between 1990 and 2000 (male-female homicide rate for 2000 minus male-female homicide rate for 1990) on change in gender equality, initial level (1990) of gender equality, a product term for initial level and change, and changes in the controls. We found that both the change in gender equality and the initial level of gender equality were associated with a decrease in male-female homicide. Furthermore the coefficient for the significant product term (p < .05, one tailed) was negative suggesting a declining effect (in strength) of increases in equality between 1990 and 2000 when 1990 equality was high. This result supports the logic underlying our explanation of the curvilinear relationship. Importantly, while change in equality and the initial level of equality were both negatively associated with male-male and female-male homicide, the product term was not significant. That said, the inclusion of change in the robbery rate, rendered all three gender equality variables non-significant in the male-female equation (though change in equality and 1990 equality remained significant in the male-male and female-male models).