Societal gender inequality and the gender gap in safety perceptions: Comparative evidence from the International Crime Victims Survey

Abstract

This study considers whether societal gender inequality moderates the relationship between gender and perceptions of personal safety. Pooled 1992–2005 rounds of the International Crime Victims Survey, comprising more than 285,000 respondents from 75 countries, are used to estimate multilevel models of safety perceptions, with a cross-level interaction specified between gender and gender inequality. We find that the gender gap in safety perceptions, although statistically significant in all countries, is largest in countries exhibiting high gender equality and smallest in countries with high gender inequality. This is explained entirely by variation in men’s safety perceptions; male respondents perceive themselves as safer in a milieu of gender equality, but less safe in a milieu of gender inequality. In contrast, the safety perceptions of female respondents are uncorrelated with societal gender inequality.

Keywords

Gender gender inequality safety perceptions fear of crime

Introduction

Research has increasingly focused on the role of gender inequality and its relation to violence against women. Gender inequality is both ideological and structural. It is defined ideologically by the beliefs, norms, and values that a society holds about the status and place of women. Structurally, it can be gauged by women’s access to resources and the positions they hold in society (Dobash and Dobash, 1979; Yodanis, 2004). Studies suggest that, when gender equality is high, the power differential between men and women is diminished. When women have increased access to jobs and political participation, and their role in society is not restricted based on their gender, this leads to a more egalitarian status quo (Whaley et al., 2013). However, theorists also note that high levels of gender equality can be regarded as a threat to male power, generating a backlash that leads to increased violence toward women in an effort to restore the gender status quo favoring men (Lauritsen and Heimer, 2008). Despite concerns about such a backlash, it is believed that in societies with high levels of gender inequality, which are characterized by the restricted role of women in society, violence against women can be expected to be widespread (Whaley et al., 2013). Elizabeth Stanko (1985) suggests that, in a male-dominated society, women’s heightened fear of victimization should also be expected. If the extent of gender inequality can predict levels of violence against women, can these measures provide us with a window into reported levels of fear?

Research on fear of crime has found that gender is among its most reliable predictors, with women consistently more fearful of victimization than men (LaGrange and Ferraro, 1989; Lane, 2012). This is despite women’s lower rates of victimization compared to men.¹ Given evidence that women’s victimization is higher in contexts with higher levels of gender inequality (Miller, 2008; Sanday, 1981), is there also a relationship between women’s status in society and their levels of fear? Similarly, does men’s status in society help explain their feelings of fear? Most research theorizes why women’s fear is higher than men’s, rather than considering why men’s fear is lower, or whether it varies across place. Moving beyond individual reports of fear of crime, can societal levels of gender inequality tell us something about the gender gap in fear? Moreover, is this gender gap as invariant as current research would suggest?

This study uses the International Crime Victims Survey to conceptualize gendered safety perceptions as a multilevel phenomenon. Pooling data from dozens of countries and multiple survey rounds, gender is theorized and measured as both an individual- and a country-level construct. The latter appears in the form of the gender inequality index, created by the United Nations Development Programme (2010) from indicators of reproductive health, political empowerment, and labor market participation. The objective is to consider whether and how safety perceptions are gendered to a different degree by the extent of gender inequality. In the language of the multilevel models employed in the analysis, a cross-level interaction between gender and gender inequality allows the size of the gender gap in safety perceptions to vary by societal gender inequality.

Gender gap in fear of crime and perceived safety

Research has consistently demonstrated the existence of a gender gap in perceptions of safety, fear of victimization, and perceived risk of crime. Since the earliest public opinion and victim surveys began probing respondents’ perceptions about safety and victimization, scholars have noted a pronounced gender gap. Erskine (1974) reports that, when respondents to a 1965 Gallup survey were asked ‘Is there any area around here – that is, within a mile – where you would be afraid to walk alone at night?’, 17 percent of men responded affirmatively compared with 49 percent of women. Both percentages climbed through the rest of the 1960s and into the early 1970s, but did so more rapidly for women.² Ennis (1967) reports from a 1967 National Opinion Research Center survey item asking ‘How safe do you feel walking alone in your neighborhood after dark?’ that 83 percent of men reportedly felt ‘very safe’ or ‘somewhat safe,’ compared with 56 percent of women.³

The gender gap is largely invariant to how anxiety about crime and safety is measured in surveys. LaGrange and Ferraro (1989) report from a large telephone survey that women are significantly more fearful of many forms of crime, and report a significantly higher likelihood of becoming a victim of personal crime (although not of property crime). They report that women are roughly three times as fearful as men of walking alone in their neighborhood at night.

The gender gap is also observed cross-nationally. Chon and Wilson (2016) use data from the fourth and fifth rounds of the International Crime Victims Survey, comprising 50 countries, to find that gender (sex category) is correlated in the expected manner with the probability of feeling ‘very unsafe’ or ‘a bit unsafe’ walking alone after dark, as well as with the perceived risk of burglary. Although they document a significantly larger gender gap in highly developed countries than in less developed countries, the gap nevertheless persists in all contexts.

The persistent correlation between gender and perceived safety has been noted as a paradox in light of the disproportionately higher likelihood of criminal victimization among men.⁴ There are a number of explanations used to explain this gender gap, many of which explain the rationale for women’s higher reported levels of fear using perceptions and experiences that are sex or gender specific. According to the physical vulnerability hypothesis, women are more fearful because they feel less capable of physically defending themselves or feel they are more likely to be injured in an assault (Jackson, 2009; Killias, 1990; Killias and Clerici, 2000; Skogan and Maxfield, 1981). A related explanation is the shadow of rape hypothesis, where the possibility that any given personal contact crime may escalate into sexual assault or rape heightens women’s fearfulness (Ferraro, 1995, 1996; Fisher and Sloan, 2003; Gordon and Riger, 1989; Warr, 1984, 1985). The ecological vulnerability hypothesis attributes this paradox to fearful reactions to verbal and sexual harassment more commonly experienced by women in public spaces (Mellgren and Ivert, 2019; Smith and Torstensson, 1997; Smith et al., 2001; Tuerkheimer, 1997; Walklate, 1997). The common thread of this class of explanations – implicit if not explicit – is that women’s anxiety about crime and safety stems from a generalized fear of men, especially unknown men (see Hollander, 2001; Stanko, 1995).

Another class of explanations turn the analytical lens toward men and prioritize cultural norms of masculinity – better known as hegemonic masculinity (see Connell, 1987, 2000). For example, the social desirability hypothesis implies men are susceptible to messages about masculinity that forbid admission of fear and vulnerability, least of all to survey takers (Sutton and Farrall, 2005; Sutton et al., 2011). Alternatively, the fearless men hypothesis asserts that masculine bravado serves as an expression of invulnerability in the face of objectively higher victimization risk (Goodey, 1997; Hollander, 2001). We also propose the offender as victim hypothesis, which implies that, because men are more frequent offenders, victimhood is akin to an ‘occupational hazard’ and thus not to be feared. What these three explanations have in common is denial of vulnerability, rooted in hegemonic masculinity, which manifests variously as concealing fearfulness, projecting fearlessness, or accepting victimization as an unavoidable consequence of criminal involvement (see Stanko and Hobdell, 1993).

A third class of explanations, closely related to the second, treats gender as a dynamic and situated accomplishment – referred to as doing gender (see West and Fenstermaker, 1995; West and Zimmerman, 1987). According to this perspective, masculinity and femininity are not properties of individuals but are socially constructed in interactions. While gender is an emergent feature of situations, it is subject to normative conceptions of approved behaviors and attitudes for particular sex categories, although it is far from completely determined by them. The gendered identity hypothesis posits that endorsement of more feminine or more masculine attitudes and behaviors corresponds with crime perceptions that align in a gendered way, among both sex categories (Cops and Pleysier, 2011). As this review illustrates, there is an abundance of theoretical explanations for the gender gap in fear of crime, perceived crime risk, and safety perceptions. Current research provides ambiguous support in whether or not safety perceptions are gendered based on the extent of gender inequality. The explanations tend to be rooted in individualistic or situational influences, although they need not be. The hegemonic masculinity and doing gender perspectives are friendly to the notion that gender is accomplished differently in different societal contexts; we explore this as a theoretical possibility in the next section.

Gender inequality and the societal context of perceived safety

To explore gender beyond the individual level we consider it at higher levels of aggregation, in the form of societal gender inequality. Empirical research incorporates various measures of gender inequality as a societal characteristic and have found it meaningful in accounting for gendered crime patterns. A small number of cross-national studies have examined the nature of the relationship between women’s status and lethal violence.⁵ Gartner et al. (1990), analyzing homicide data from 18 democracies from 1950 to 1980, report that women’s non-traditional roles reduce the gap between women’s and men’s risk of homicide victimization, although higher female status increases this gap.⁶ Yodanis (2004) aggregates data from respondents in 27 countries surveyed in the 1989 round of the International Crime Victims Survey (ICVS), and finds that gender inequality is associated with a culture of violence against women.⁷ The education and occupational status of women are inversely correlated with the overall prevalence of sexual violence, wherein the higher status of women corresponds with lower rates of sexual violence. Whaley and Messner (2002) examine the effects of gender equality on the ‘gendering’ of lethal violence and find that, for a sample of US cities, gender equality is positively related to rates of both male-on-female homicide and male-on-male homicide in Southern cities. They also find that gender equality is negatively associated with male-on-male homicide in cities located in other regions.

Although not a study of the gendering of lethal violence or perceived safety, Savolainen et al. (2017) examine correlations between societal-level gender inequality and self-report delinquency in surveys from 30 countries in the International Self-Report Delinquency Survey. They construct a normative measure of patriarchy by aggregating responses from the World Values Survey, and employ a structural measure of gender inequality from the United Nations.⁸ Their findings indicate the gender gap in self-report delinquency is considerably larger in more patriarchal and gender-unequal countries. Irrespective of whether gender inequality is measured normatively or structurally, the wider gender gap in these countries stems largely from the lower volume of delinquency among young women, and only partly from the higher volume of delinquency among young men. These studies help identify patterns that may speak to how gender equality can have an impact on fear of victimization, to the extent that it threatens male dominance and challenges stereotypical notions of masculinity/femininity.

Current theories provide ambiguous support in whether or not safety perceptions are gendered based on the extent of gender inequality. As these theories are anchored in different frames, either outcome can be explained. Hypotheses about the nature of the relationship between societal gender inequality and the gender gap in perceived safety are comparably varied in light of this complexity. On one hand, we can hypothesize that countries with more gender inequality will exhibit a larger gender gap in perceived safety. Societal gender inequality – including women’s access to jobs, political participation, healthcare, and other social institutions and resources relative to men – has a large role in shaping norms and values and the individual lived experience of being a woman or man. Gender inequality corresponds with women’s restriction to the domestic sphere and higher levels of economic dependency. Although this might restrict their routine activities and limit their opportunities for victimization in public settings, it is also a cultural context of control over women that could promote anxiety about female safety (see Bograd, 1988). In contexts of high gender inequality, violence against women is believed to be more normalized and widespread, even as it is disproportionately likely to take place in the home and at the hands of intimates. Alternatively, in terms of men’s perceived safety, high gender inequality also may correspond with restrictive definitions of masculinity, wherein expressions of fear are not manly and are a sign of weakness, driving down men’s concerns about their safety (Stanko, 1990). This would lead us to predict that gender inequality widens the gender gap in perceived safety, either by driving up women’s anxiety and/or by decreasing men’s (self-reported) anxiety.

We can also hypothesize that countries with more gender inequality will exhibit a narrower gender gap in perceived safety. We can make a case for this through routine activities, where the theoretical emphasis is on exposure, particularly in public settings. As individuals spend less time away from home, the likelihood of becoming a victim of predatory crime declines. Since societal gender inequality restricts women’s access to activities outside of the home, their perceived safety might be higher despite the normalization of domestic violence. Alternatively, gender inequality may correspond with several factors that increase men’s safety concerns. First, prior research suggests that in contexts of gender inequality, general aggression and violence are also higher (Gartner et al., 1990; Sanday, 1981; Whaley and Messner, 2002), thus increasing men’s risks for victimization. Reid and Konrad (2004) explain that men’s fear of crime actually surpasses women’s fear when the effect of perceived risk of victimization of certain gender-neutral crimes (such as robbery) is greater for men than for women. Second, to the extent that men in societies with high gender inequality are expected to act in hypermasculine ways to align with cultural standards of manhood, men may place themselves in situations wherein the chances of being victimized are higher, thus driving up their reported levels of fear. These arguments would lead us to predict that gender inequality narrows the gender gap in perceived safety, either by reducing women’s reported levels of fear and/or by increasing men’s anxiety about their safety. Different theories and insights account for these hypotheses; the objective of the analysis is to tease out plausible mechanisms that consider whether and how safety perceptions are gendered to a different degree by the extent of gender inequality.

Data and methods

The data from this study are drawn from the ICVS (Van Dijk et al., 2007), and are supplemented with information drawn from a variety of country-level databases. The ICVS is a comparative survey of criminal victimization, initially conducted in 1989. Subsequent rounds (‘sweeps’) of data collection occurred in 1992, 1996, 2000, and 2004/5.⁹ This study uses all surveys conducted since 1992, the year that a survey item inquiring about safety perceptions was first fielded. After elimination of respondents missing information on safety perceptions and gender, the working sample is 287,266 respondents from 75 countries (these are listed in Appendix A in the online Supplemental Material).

Respondent-level measures

All variables and their descriptives are shown in Table 1. The dependent variable, safety perceptions, is an item from the survey inquiring about the degree to which the respondent feels safe versus unsafe after dark. This variable was selected because it was asked of a broad cross-section of respondents during the time period under consideration, and bears the closest resemblance to a measure of fear of victimization. The four response categories are feeling ‘very unsafe’ (1), ‘bit unsafe’ (2), ‘fairly safe’ (3), and ‘very safe’ (4). A higher value on this measure thus indicates that a respondent feels safer after dark.

Table 1.

Descriptive statistics.

Variable	Valid N	Mean (SD)	Min, Max
Dependent variable
After-dark safety perceptions	287,266	2.78 (0.96)	1, 4
1 Very unsafe	287,266	12.0%	0, 1
2 Bit unsafe	287,266	23.7%	0, 1
3 Fairly safe	287,266	39.3%	0, 1
4 Very safe	287,266	25.1%	0, 1
Respondent-level (level 1) regressors
Male	287,266	48.5%	0, 1
Age	269,671	6.00 (3.34)	1, 12
Household size	261,558	3.72 (1.82)	1, 10
Occupational status
Working	276,176	52.9%	0, 1
Looking for work	276,176	7.8%	0, 1
Keeping home	276,176	11.6%	0, 1
Retired or disabled	276,176	17.4%	0, 1
In school	276,176	10.3%	0, 1
Years of education	211,640	11.93 (3.83)	0, 25
Relative income
Upper 25 percent	247,064	25.6%	0, 1
50–75 percent	247,064	26.3%	0, 1
25–50 percent	247,064	28.5%	0, 1
Lower 25 percent	247,064	19.7%	0, 1
Marital status
Single	282,073	27.2%	0, 1
Married	282,073	62.0%	0, 1
Divorced	282,073	4.9%	0, 1
Widowed	282,073	6.0%	0, 1
Victim of robbery	285,509	7.1%	0, 1
Victim of assault	274,332	10.5%	0, 1
Victim of burglary	285,206	10.9%	0, 1
Victim of attempted burglary	284,103	9.5%	0, 1
Victim of car theft	285,756	4.7%	0, 1
Victim of theft from car	285,333	15.4%	0, 1
Victim of personal theft	285,105	18.5%	0, 1
Country-level (level 2) regressors
Polity score	144 (72)	6.72 (5.22)	–9.00, 10.00
Homicide rate per 100,000	133 (70)	7.66 (12.47)	0.50, 67.18
Gender inequality index	142 (70)	0.32 (0.18)	0.05, 0.69
Gender inequality components
Reproductive health	142 (70)	0.59 (0.08)	0.50, 0.80
Political empowerment	142 (70)	0.52 (0.12)	0.16, 0.70
Labor market participation	142 (70)	0.61 (0.07)	0.47, 0.85

Notes: N = 287,266 respondents from 75 countries. Means of binary variables are shown as percentages. Estimates are weighted for the respondent-level regressors but are unweighted for the county-level regressors. For the country-level regressors, the first sample size refers to the number of countries by the number of ICVS survey rounds (‘sweeps’) with valid data, and this is the basis of the descriptive statistics. The second sample size (in parentheses) refers to the number of countries with valid data in at least one survey round.

Source: International Crime Victims Survey integrated database, 1992–2005 survey years.

The distribution of safety perceptions, by gender, is provided in Figure 1. Female respondents report feeling less safe after dark than their male counterparts: 45 percent of female respondents report feeling a bit unsafe or very unsafe, whereas 26 percent of male respondents report the same sentiments. All conventional statistical tests (for example, t-test, chi-square test) are highly significant, and confirm the expectation that women generally report feeling less safe than men.

Figure 1.

Distribution of safety perceptions, by gender.

At the respondent level, age, household size, occupational status, education, relative income, marital status, and prior victimization are included as covariates. These are commonly used variables in victimization and fear of victimization studies. Age is measured in five-year intervals but is included as a continuous regressor since close inspection reveals a roughly linear relationship with safety perceptions. Household size is the actual number of residents, top-coded at 10. Occupational status is dummy coded to classify respondents as to whether they are working, looking for work, keeping home, retired or disabled, or still in school. Years of education is a continuous measure of the number of years of schooling. Relative income is dummy coded to classify respondents into income quartiles (for example, lower 25 percent). Marital status is dummy coded to classify respondents as single, married or living together, divorced, or widowed. Prior victimization is measured by a set of seven dummy indicators for whether respondents have, in the five years prior to the survey, been a victim of robbery, assault, burglary, burglary attempt, car theft, theft from car, or personal theft.

Country-level measures

The key country-level measure is the gender inequality index, measured by the United Nations Development Programme’s gender inequality index. Introduced in the 2010 Human Development Report and backfilled in five-year intervals to 1990, this index comprises indicators of reproductive health, political empowerment, and labor market participation. Closely related to the human development index, the gender inequality index seeks to quantify ‘the distribution of achievements between women and men’ (United Nations Development Programme, 2010: 7), in particular, the loss in achievements due to gender disparities in the three criteria (see also Gaye et al., 2010). The gender inequality index is available in five-year intervals between 1990 and 2010, the years that fully encompass the ICVS. The closest available year is matched to ICVS respondents from a given country in a given year. Note that the match is made based on the calendar year the ICVS survey was administered to a given respondent, rather than the year of the round of data collection (for example, 1992, 1996, 2000, 2004/5).

The first dimension of the gender inequality index measures inequality in reproductive health using two indicators: maternal mortality ratio and adolescent fertility rate. The second dimension measures inequality in political empowerment with two indicators: secondary education (separately measured for men and women) and female parliamentary representation. The third dimension measures inequality in labor market participation using labor force participation rates measured separately for men and women. Details on the construction of the gender inequality index and its components are provided by the United Nations Development Programme (2010).

As country-level control variables, we include measures of the political environment and lethal violence for the calendar year the ICVS survey was administered. The polity score is a measure of democratic governance developed by the Polity IV Project and disseminated by the Center for Systemic Peace (Marshall et al., 2018). It is updated annually and assigns numerical values derived from evaluations of political participation, election openness, and checks on executive authority, with the highest value identifying strongly democratic regimes and the lowest value identifying strongly autocratic regimes. The homicide rate per 100,000 controls for the volume of lethal violence in a country proximate to when respondents reported their safety perceptions. This information is obtained from the United Nations Office on Drugs and Crime, whose intentional homicide database includes information from 1995 to the present.

Analysis

Because the 287,266 respondents surveyed in the ICVS are nested within 75 countries, a two-level hierarchical model is estimated. The model is parameterized with respondents at level 1 and countries at level 2. The random effect is assumed to be normally distributed across countries and captures the balance of unobserved variables that characterize mean differences in country-level safety perceptions. The dependent variable is ordinal, but a normal-error model is estimated for ease of interpretation. However, inferences from the normal-error models are confirmed from ordered logistic regression models. Because of the sampling design of the ICVS, all models are estimated using individual survey weights. Note also that all models utilize cluster robust standard errors, where clusters are defined by countries.

Recall that gender (level 1) and gender inequality (level 2) are the focal variables in this analysis. In the first multilevel formulation, the level-1 coefficient for gender quantifies the expected difference in safety perceptions of a male respondent compared with a female counterpart, whereas the level-2 coefficient for gender inequality captures the difference in average safety perceptions for respondents who reside in two countries that differ by one unit in gender inequality. In a second multilevel formulation, the interaction between gender and gender inequality is included, which is a ‘cross-level interaction’ that allows the size of the gender gap in safety perceptions to differ by societal gender inequality.

Because as much as 25 percent of the sample is missing values on any given regressor (for example, years of education), to avoid deterioration of the sample size respondents with missing values are flagged with a dummy variable for each such regressor. These are then included as additional control variables in order to adjust coefficients for peculiar missingness patterns. The same procedure is used for countries that lack information on homicide or gender inequality in a given survey year. We finally perform a variety of sensitivity analyses, and wait to comment on these at relevant points below.

Results

A graph of the relationships between gender, gender inequality, and safety perceptions in the ICVS is provided in Figure 2. The figure sorts countries by their mean gender inequality between 1995 and 2005, and plots mean safety perceptions for men and women. There are three important features of the graph. First, viewing the vertical axis, in all countries surveyed, males possess higher mean safety perceptions than females. Second, viewing the horizontal axis, in countries with a higher rank on the gender inequality index, mean safety perceptions tend to be lower. This is indicated by the fact that the linear fits to the male and female country-specific means both slope downward. Third, there is suggestive evidence from the graph that the size of the gender gap in safety perceptions gets smaller as a function of gender inequality. This is apparent from an examination of the linear fits for men and women, which exhibit mild convergence among countries with higher gender inequality.

Figure 2.

Mean safety perceptions, by gender and gender inequality.

Prior to estimation of the full specified regression models, an intercept-only multilevel model is estimated. This ‘unconditional’ model indicates there is significant heterogeneity in mean safety perceptions across the 75 countries. Specifically, the intraclass correlation is 0.170, meaning 17 percent of the total variance in safety perceptions exists across countries, with the balance accounted for by within-country variance across respondent characteristics. This model thus confirms the relevance of the analytical approach adopted below.

The coefficients and standard errors from the multilevel regression models are shown in Table 2. Recall that higher values of the dependent variable indicate respondents feel safer after dark. Model 1 is a standard multilevel model with no interactions. To summarize, we focus our attention first on the respondent-level (level-1) regressors. Male respondents report feeling significantly safer after dark relative to females. Younger respondents, respondents from larger households, and those who are working (relative to those who are retired or looking for work) or who are single also report feeling significantly safer after dark. Respondents who are better educated feel significantly safer after dark, as do respondents with higher relative income. Finally, respondents who have experienced any form of previous victimization – either personal or property – report feeling significantly less safe after dark. Among the country-level (level-2) regressors, the homicide rate but neither the gender inequality index nor the polity score is correlated with mean safety perceptions.

Table 2.

Multilevel linear regression model of after-dark safety perceptions.

Variable	Model 1			Model 2
	Coeff.	SE	Sig.	Coeff.	SE	Sig.
Respondent level (level 1) regressors
Male	0.425	0.021	***	0.540	0.058	***
Age / 10	−0.091	0.026	***	−0.088	0.026	***
Household size / 10	0.181	0.049	***	0.180	0.049	***
Occupational status (ref. = Working)
Looking for work	−0.036	0.016	*	−0.035	0.016	*
Keeping home	−0.031	0.022		−0.047	0.020	*
Retired or disabled	−0.136	0.017	***	−0.139	0.017	***
In school	−0.001	0.017		−0.001	0.018
Years of education / 10	0.070	0.024	**	0.071	0.024	**
Relative income (ref. = Upper 25 percent)
50–75 percent	−0.048	0.015	**	−0.046	0.015	**
25–50 percent	−0.088	0.019	***	−0.085	0.019	***
Lower 25 percent	−0.098	0.022	***	−0.093	0.021	***
Marital status (ref. = Single)
Married	−0.040	0.019	*	−0.041	0.019	*
Divorced	−0.051	0.025	*	−0.051	0.025	*
Widowed	−0.040	0.020	*	−0.037	0.020
Victim of robbery	−0.190	0.025	***	−0.188	0.025	***
Victim of assault	−0.178	0.019	***	−0.175	0.020	***
Victim of burglary	−0.118	0.013	***	−0.119	0.013	***
Victim of attempted burglary	−0.172	0.013	***	−0.171	0.013	***
Victim of car theft	−0.088	0.010	***	−0.088	0.010	***
Victim of theft from car	−0.076	0.011	***	−0.075	0.011	***
Victim of personal theft	−0.082	0.012	***	−0.083	0.012	***
Country-level (level 2) regressors
Polity score	0.003	0.007		0.002	0.008
Homicide rate per 100,000 (logged)	−0.107	0.053	*	−0.111	0.054	*
Gender inequality index	−0.440	0.595		−0.198	0.599
Cross-level interactions
Male × Polity score	–	–		0.002	0.004
Male × Homicide rate per 100,000	–	–		0.008	0.016
Male × Gender inequality index	–	–		−0.521	0.128	***

Notes: N = 287,266 respondents from 75 countries. Estimates are weighted at the respondent level. Cluster robust standard errors are shown. Included but not shown are an intercept and dummy variables for survey round (‘sweep’) and for regressors with missing data. Positive coefficients signify that respondents feel safer after dark when there is an incremental increase in the value of the regressor. The intraclass correlation for both models is 0.118.

p < .05, **p < .01, ***p < .001 (two-tailed tests).

Source: International Crime Victims Survey integrated database, 1992–2005 survey years.

The relative magnitude of the regressors with respect to the strength of their relationship with safety perceptions can be judged by z-scores, of which the largest are gender (+20.32), average of prior victimization (for example, −8.80, but ranging from −6.67 for personal theft to −13.28 for attempted burglary), average of relative income (−4.11), household size (+3.72), age (−3.55), average of occupational status (−2.99), education (+2.95), average of marital status (−2.05), and homicide rate (−2.02). We can also derive the effect size to estimate the practical significance of the gender differences observed in Model 1.¹⁰ The effect size for gender in this model is d = 0.46. Considering that 0.2 is widely regarded as ‘small,’ whereas 0.5 is ‘medium,’ we conclude that the gender differences in safety perceptions are substantial, and are both statistically and substantively significant.

Model 2 includes the cross-level interaction between gender and gender inequality. Inclusion of the interaction modifies the interpretation of the ‘main effects’ for gender and gender inequality. First, the main effect for gender indicates the gap in safety perceptions between males and females who live in countries with perfect gender equality (that is, gender inequality index = 0). Countries with near-zero gender inequality in our database (0.056 to 0.083) include, for example, Sweden, the Netherlands, and Denmark. In these countries, the expected gender gap holds up strongly, whereby males feel significantly safer after dark than females. The practical significance of the gender gap in safety perceptions in gender-equal societies is best expressed by the effect size, d = 0.58.

Second, the main effect for gender inequality indicates mean safety perceptions among the female residents (that is, male = 0) of countries with incrementally higher gender inequality. The coefficient is statistically flat, suggesting gender inequality is uncorrelated with females’ after-dark safety perceptions. In other words, female respondents of countries with the lowest and highest levels of gender inequality do not differ in their mean safety perceptions.

Third, the coefficient for the interaction between gender and gender inequality represents a contrast in the slope of the relationship between gender inequality and mean safety perceptions among male residents compared to female residents. The fact that it is negative and highly statistically significant indicates that gender inequality is inversely correlated with male safety perceptions, compared to the null correlation with female safety perceptions. Specifically, while the slope of the relationship between gender inequality and safety perceptions is −0.198 among women, it is −0.719 among men (−0.198 − 0.521). The implication for the gender gap in safety perceptions is that, in a societal milieu of gender equality, the gap in safety perceptions between men and women is considerably larger than it is in a societal milieu of gender inequality, and this gap seems to be driven by the perceptions of men rather than of women. Stated differently, male respondents feel significantly safer in countries with high gender equality than they do in countries with high gender inequality. But female respondents feel equally (un)safe in countries with high gender equality versus high gender inequality.¹¹

Figure 3 provides a graph of the information conveyed by the regression models in Table 2. The left-hand-side graph corresponds with the results in Model 1, the multilevel model with no interaction, and is provided strictly for comparison. The right-hand-side graph corresponds with the results in Model 2, the multilevel model with the cross-level interaction, and produces the key result. In countries with low gender inequality, the gender gap in safety perceptions is large and in the expected direction, as indicated by the lack of overlap in confidence intervals for female and male respondents. By comparison, in countries with the highest levels of gender inequality – in our sample, these countries include India, Egypt, and Uganda, with values on the gender inequality index larger than 0.658 – the gender gap, while still statistically significant, is substantially smaller. Indeed, the gender gap in safety perceptions in countries with the most gender inequality (0.182) is just one-third the size of the gender gap in safety perceptions in countries with the most gender equality (0.540).¹²

Figure 3.

Predictive margins of the cross-level interaction between gender and gender inequality.

Sensitivity analyses

The results reported above are from normal-error multilevel models, the coefficients of which are easily summarized and interpreted. As a sensitivity analysis, ordered logit models were also estimated, since the dependent variable possesses ordered response metric. All of the basic relationships described above are replicated in these models. Namely, the coefficients for gender and the cross-level interaction are statistically significant. In addition, we dichotomized safety perceptions (0 = very unsafe or bit unsafe; 1 = fairly safe or very safe) and estimated logistic regression models, and again all of the significance patterns were replicated.

Because the ICVS has undergone multiple rounds of data collection, surveys were fielded in more than one ‘sweep’ in a sizable number of countries. Repeated cross-sections are thus technically nested within countries, allowing for estimation of a three-level model of safety perceptions. When we estimate this model, the key results are virtually identical to what is reported above. These results are tabulated and graphed in Appendixes B and C in the online Supplemental Material.

Finally, it is possible to treat country-level unobservables as fixed rather than random, and thereby relax the stringent assumption that the unobservables are uncorrelated with the included regressors. We do so by dummy coding countries and including them as regressors. The coefficients for the polity score, homicide rate, and gender inequality index remain identified because they are matched to the year of survey administration for each respondent, which means they are time varying (otherwise, they would be completely absorbed by the dummy variables). When we estimate this fixed-effects model, the key results are little changed from what is reported in Table 2. These results are tabulated and graphed in Appendixes D and E in the online Supplemental Material.

Discussion

This study is motivated by the question of whether and how macro-level indicators of gender inequality modify the micro-level relationship between gender and perceived safety. We use the ICVS to explore gendered safety perceptions as a multilevel phenomenon, by considering how the size of the gender gap (at the individual level) is altered by the extent of gender inequality (at the country level). Congruent with previous criminological research, we find women report significantly more concerns about their safety compared with men. Although our results suggest this is true no matter the level of gender inequality, it is also the case that the gender gap is significantly (and substantially) larger in societies characterized by gender equality, but smaller in societies characterized by gender inequality. Our results also indicate this pattern – a larger gender gap in gender-equal societies than in gender-unequal societies – is driven by the safety perceptions of men rather than of women in these different societal milieus. In fact, and unexpectedly, the safety perceptions of women are inelastic with respect to societal gender inequality. The theoretical task, therefore, is to explain why female safety perceptions are unrelated to gender inequality, whereas male safety perceptions are correlated with gender inequality, in such a way that the size of the gender gap diminishes in societal gender inequality.

A first pass at an explanation is informed by studies of the macro-level relationship between gender inequality and violence. Violence tends to be higher in gender-unequal societies and, because men are frequently the victims of violence (notwithstanding domestic violence), this has implications for their overall risk of victimization (Gartner et al., 1990; Sanday, 1981; Whaley and Messner, 2002). In a series of follow-up models with the ICVS in which we use the indicators of personal victimization as outcomes (not shown), the gender inequality index is positively and significantly correlated with mean risk of robbery, burglary, attempted burglary, and personal theft. Therefore, to the extent that gender inequality is correlated with objective victimization risk, and especially objective violence risk, this could explain men’s greater concern about their safety (see Reid and Konrad, 2004). Our results do not refute this as an explanatory mechanism, but, to the degree the homicide rate and prior personal victimization successfully proxy for differential objective risk, our findings suggest this is an incomplete explanation. The gender inequality index remains a strong modifier of the gender gap with these regressors included.

A related explanatory mechanism is rooted in differential risk exposure via gendered routine activities. If gender inequality implies that men spend more of their time outside the home and in the public arena, their elevated safety concerns might be simply explained by their increased exposure to predatory victimization. In a follow-up model with the ICVS in which we use as an outcome a measure of the frequency of going out (not shown), there is a significant cross-level interaction between gender and gender inequality that is only partly in the expected direction. Among female respondents, high gender inequality is correlated with lower mean frequency of going out, implying the adoption of more home-based routine activities, consistent with expectations. Among male respondents, however, there is no correlation between gender inequality and the frequency of going out, which is contrary to expectations. When we control for this variable directly in the model of safety perceptions, the gender inequality index remains a strong modifier of the gender gap in safety perceptions. Although an imperfect test, explanatory mechanisms that appeal to gendered routine activities do not suffice to explain the pattern of our findings.

Considering the inability of these traditional criminological mechanisms to fully explain our findings, we cannot rule out that this study speaks to how public space is utilized in ways that are difficult to measure and quite independently of the nature of the activities that take place there. The fact that men’s safety perceptions, in particular, vary directly with gender inequality speaks to the gendering of public space in different societal contexts. Men’s greater concern for their safety in contexts of high gender inequality might be better understood by exploring how these contexts motivate men and women to act in gender stereotypical ways (Bengtsson, 2016; West and Zimmerman, 1987). Gender and power are relational constructs, and serve simultaneously as outcomes of, and justifications for, social arrangements that legitimize the most fundamental division of society. Societal gender inequality perpetuates masculine domination of the established order, involving submission to schemes of perception, thought, and action demanding that ‘real men’ be ever ready to pursue glory and distinction in the public sphere (Bourdieu, 2001). An essential theoretical point, which can be easily overlooked, is that masculine domination implies not only the domination of men over women in these spaces, but the domination of men over other men, as well (Messerschmidt, 1993).

This can also be explained through Furstenberg’s (1971) study, which seeks to explore whether the concern about crime is resentment of social change and resistance to further alterations in the status quo or whether public reactions to crime are justified due to increasing crime rates. Furstenberg’s study (using interviews with 1545 Baltimore residents) finds that one’s discontent with changing social conditions is associated with high apprehension about the crime situations and that concern about crime is at least in part an expression of resentment of changing social conditions.¹³ Although in his study changing social conditions are defined in the realm of racial justice, we could extrapolate these findings to hypothesize about public sentiments on crime within social conditions where levels of gender equality are shifting. Based on this we can consider that higher levels of fear among men are a result of growing tension and efforts to strive for gender equality (specifically in countries with high levels of gender inequality, with a poor status quo on gender equality).

Limitations

It is important to point out several limitations to this study and its conclusions. First, in its current form, the gender inequality index includes three dimensions: reproductive health, political empowerment, and labor market participation. Measuring these dimensions across a diverse set of nations, using the same indicators, gives rise to the legitimate question of whether they are equally relevant and meaningful in all contexts. This is a longstanding concern in the comparative research tradition. Second, we would have liked to have access to a more direct measure of fear of victimization. Instead, we are able to use only an imperfect proxy measured by how safe respondents feel after dark. Although we are confident that this item possesses convergent validity with fear of victimization, it might be contaminated with other sources of variation not directly related to crime and victimization.

Third, our measure of gender inequality is also an imperfect proxy. The United Nations Development Programme has invested a great deal in the construction and validation of the gender inequality index – it was created as a solution to certain limitations of the better-known human development index. Nevertheless, the UN’s reliance on ‘objective’ indicators fails to tap into societal definitions of gender roles and expectations. Fourth, while we are able to incorporate weights at the individual level, countries are otherwise equally weighted in the analysis.

Despite these and other shortcomings, this study provides evidence that variation in the size of the gender gap in safety perceptions is partly a function of societal gender inequality. It will be important in future research to confirm whether or not this finding holds up in other data sources that are capable of improving upon the limitations of the present study.

Conclusion

Congruent with previous criminological research, we found that women report more safety concerns compared with men, and, although we discovered this was true no matter the level of societal gender inequality, the female–male gap was largest in gender-equal societies and smallest in gender-unequal societies. Although men perceive themselves to be significantly safer than women no matter the societal context, the gender gap in countries with high gender equality is substantially larger than in countries with high gender inequality. Our data suggest the size of the gender gap is driven strictly by the perceptions of men, whose safety concerns are more highly elastic with respect to societal gender inequality. On the other hand, and unexpectedly, the safety perceptions of women are not correlated with gender inequality. Where perceptions of personal safety are concerned, then, the findings from our study suggest the chief beneficiaries of societal gender equality are men.

Supplemental Material

Appendix_7 – Supplemental material for Societal gender inequality and the gender gap in safety perceptions: Comparative evidence from the International Crime Victims Survey

Supplemental material, Appendix_7 for Societal gender inequality and the gender gap in safety perceptions: Comparative evidence from the International Crime Victims Survey by Ntasha Bhardwaj and Robert Apel in European Journal of Criminology

Footnotes

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Ntasha Bhardwaj

Supplemental material

Supplemental material for this article is available online.

Notes

References

Bailey

Peterson

(1995) Gender inequality and violence against women: The case of murder. In: Hagan

Peterson

(eds) Crime and Inequality. Stanford, CA: Stanford University Press, 174–205.

Bengtsson

(2016) Performing hypermasculinity: Experiences with confined young offenders. Men and Masculinities 19(4): 410–428.

Bograd

(1988) Feminist perspectives on wife abuse: An introduction. In: Yllö

Bograd

(eds) Feminist Perspectives on Wife Abuse. Beverly Hills, CA: Sage, 11–26.

Bourdieu

(2001) Masculine Domination. Translated by R

Nice

. Stanford, CA: Stanford University Press.

Chon

Wilson

(2016) Perceived risk of burglary and fear of crime: Individual- and country-level mixed modeling. International Journal of Offender Therapy and Comparative Criminology 60: 308–325.

Cohen

(1988) Statistical Power Analysis for the Behavioral Sciences (2nd edn). Hillsdale, NJ: Lawrence Erlbaum Associates.

Connell

(1987) Gender and Power. Stanford, CA: Stanford University Press.

Connell

(2000) The Men and the Boys. Berkeley, CA: University of California Press.

Cops

Pleysier

(2011) ‘Doing gender’ in fear of crime: The impact of gender identity on reported levels of fear of crime in adolescents and young adults. British Journal of Criminology 51: 58–74.

10.

Dobash

(1979) Violence against Wives: A Case against the Patriarchy. London: Open Books.

11.

Ennis

(1967) Criminal Victimization in the United States: A Report of a National Survey. National Opinion Research Center, University of Chicago.

12.

Erskine

(1974) The polls: Fear of violence and crime. Public Opinion Quarterly 38: 131–145.

13.

Ferraro

(1995) Fear of Crime: Interpreting Victimization Risk. Albany, NY: State University of New York Press.

14.

Ferraro

(1996) Women’s fear of victimization: Shadow of sexual assault? Social Forces 75: 667–690.

15.

Fisher

Sloan

III (2003) Unraveling the fear of victimization among college women: Is the ‘shadow of sexual assault hypothesis’ supported? Justice Quarterly 20: 633–659.

16.

Furstenberg

(1971) Public reaction to crime in the streets. The American Scholar 40(4): 601–610.

17.

Gartner

Baker

Pampel

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

18.

Gauthier

Bankston

(1997) Gender equality and the sex ratio of intimate killing. Criminology 35: 577–600.

19.

Gaye

Klugman

Kovacevic

Twigg

Zambrano

(2010) Measuring key disparities in human development: The gender inequality index. Human Development Research Paper 2010/46. New York: United Nations Development Programme.

20.

Goodey

(1997) Boys don’t cry: Masculinities, fear of crime and fearlessness. British Journal of Criminology 37: 401–418.

21.

Gordon

Riger

(1989) The Female Fear. New York: The Free Press.

22.

Haynie

(1998) The gender gap in fear of crime, 1973–1994: A methodological approach. Criminal Justice Review 23: 29–50.

23.

Hollander

(2001) Vulnerability and dangerousness: The construction of gender through conversation about violence. Gender & Society 15: 83–109.

24.

Jackson

(2009) A psychological perspective on vulnerability and fear of crime. Psychology, Crime and Law 15(4): 2–24.

25.

Killias

(1990) Vulnerability: Towards a better understanding of a key variable in the genesis of fear of crime. Violence and Victims 5: 275–295.

26.

Killias

Clerici

(2000) Different measures of vulnerability in their relation to different dimensions of fear of crime. British Journal of Criminology 40: 437–450.

27.

LaGrange

Ferraro

(1989) Assessing age and gender differences in perceived risk and fear of crime. Criminology 27: 697–719.

28.

Lane

(2012) Theoretical explanation for gender differences in fear of crime. In: Routledge International Handbook of Crime and Gender Studies. London: Routledge, 55–70.

29.

Lauritsen

Heimer

(2008) The gender gap in violent victimization, 1973–2004. Journal of Quantitative Criminology 24: 125–147.

30.

Marshall

Gurr

Jaggars

(2018) Polity IV Project: Political Regime Characteristics and Transitions, 1800–2017. Vienna: Center for Systemic Peace.

31.

Mellgren

Ivert

(2019) Is women’s fear of crime fear of sexual assault? A test of the shadow of sexual assault hypothesis in a sample of Swedish university students. Violence Against Women 25: 511–527.

32.

Messerschmidt

(1993) Masculinities and Crime: Critique and Reconceptualization of Theory. Lanham, MD: Rowman & Littlefield.

33.

Miller

(2008) Getting Played: African American Girls, Urban Inequality, and Gendered Violence. New York: New York University Press.

34.

Rader

(2004) The threat of victimization: A theoretical reconceptualization of fear of crime. Sociological Spectrum 24: 689–704.

35.

Rader

May

Goodrum

(2007) An empirical assessment of the ‘threat of victimization’: Considering fear of crime, perceived risk, avoidance, and defensive behaviors. Sociological Spectrum 27: 475–505.

36.

Reid

Konrad

(2004) The gender gap in fear: Assessing the interactive effects of gender and perceived risk on fear of crime. Sociological Spectrum 24(4): 399–425.

37.

Sanday

(1981) The socio-cultural context of rape: A cross-cultural study. Journal of Social Issues 37: 5–27.

38.

Savolainen

Applin

Messner

Hughes

Lytle

Kivivuori

(2017) Does the gender gap in delinquency vary by level of patriarchy? A cross-national comparative analysis. Criminology 55(4): 726–753.

39.

Skogan

Maxfield

(1981) Coping with Crime: Individual and Neighborhood Reactions. Beverly Hills, CA: Sage.

40.

Smith

(1989) Social relations, neighbourhood structure and the fear of crime in Britain. In: Evans

Herbert

(eds) The Geography of Crime. London: Routledge.

41.

Smith

Torstensson

(1997) Gender differences in risk perceptions and neutralizing fear of crime: Toward resolving the paradoxes. British Journal of Criminology 37: 608–663.

42.

Smith

Torstensson

Johansson

(2001) Perceived risk and fear of crime: Gender differences contextual sensitivity. International Review of Victimology 8: 159–181.

43.

Stanko

(1985) Intimate Intrusions: Women’s Experience of Male Violence. London: Routledge & Kegan Paul.

44.

Stanko

(1990) Everyday Violence: How Women and Men Experience Sexual and Physical Danger. London: Pandora Press.

45.

Stanko

(1995) Women, crime, and fear. Annals of the American Academy of Political and Social Science 539: 46–58.

46.

Stanko

Hobdell

(1993) Assault on men: Masculinity and male violence. British Journal of Criminology 33: 400–451.

47.

Sutton

Farrall

(2005) Gender, socially desirable responding and the fear of crime: Are women really more anxious about crime? British Journal of Criminology 45: 212–224.

48.

Sutton

Robinson

Farrall

(2011) Gender, fear of crime, and self-presentation: An experimental investigation. Psychology, Crime & Law 17: 421–433.

49.

Tuerkheimer

(1997) Street harassment as sexual subordination: The phenomenology of gender-specific harm. Wisconsin Women’s Law Journal 12: 167–206.

50.

United Nations Development Programme (2010) Human Development Report 2010 – The Real Wealth of Nations: Pathways to Human Development. 20th Anniversary Edition. New York: Palgrave Macmillan.

51.

Valentine

(1992) Images of danger: Women’s sources of information about the spatial distribution of male violence. Area 24(1): 22–29.

52.

Van Dijk

Van Kesteren

Smit

(2007) Criminal Victimization in International Perspective: Key Findings from the 2004–2005 ICVS and EU ICS. Onderzoek en Beleid, WODC, No. 257. The Hague: Boom Legal Publishers.

53.

Walklate

(1997) Risk and criminal victimization: A modernist dilemma. British Journal of Criminology 37: 35–45.

54.

Warr

(1984) Fear of victimization: Why are women and the elderly more fearful? Social Science Quarterly 65: 681–702.

55.

Warr

(1985) Fear of rape among urban women. Social Problems 32: 238–250.

56.

West

Fenstermaker

(1995) Doing difference. Gender & Society 9: 8–37.

57.

West

Zimmerman

(1987) Doing gender. Gender & Society 1: 125–151.

58.

Whaley

Messner

(2002) Gender equality and gendered homicides. Homicide Studies 6(3): 188–210.

59.

Whaley

Messner

Veysey

(2013) The relationship between gender equality and rates of inter- and intra-sexual lethal violence: An exploration of functional form. Justice Quarterly 30(4): 732–754.

60.

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.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.17 MB