Gendered framing in human rights campaigns

Abstract

Human Rights Organizations (HROs) often rely on gender stereotypes in their advocacy campaigns. Yet, we know little about the effects of such essentializing stereotypes on the success of HRO campaigns. Using a survey experiment, we examine how human rights campaigns emphasizing sex and gender roles of the victims affect individuals’ attitudes on the campaign issue and their willingness to take direct action. Although we find no direct effect of gendered personal frames in an HRO campaign on levels of consensus or action mobilization, there are indirect effects. When victims are portrayed as holding stereotypically feminine roles, respondents are more likely to identify them as innocent and vulnerable. These evaluations lead respondents to identify the violence described as a fundamental human rights violation, and to participate in direct action to end the abuse. However, as these effects are indirect and often substantively trivial, we recommend HROs avoid using such problematic messaging strategies.

Acknowledgments

The authors would like to thank the editorial team and two anonymous reviewers at the Journal of Human Rights and participants at the conference on The Social Practice of Human Rights at the University of Dayton, October 2–3, 2015, for their helpful comments. Thanks also to Amber Garcia-McBride and Amanda Murdie for valuable modeling and methodological advice. Thanks to Danica Genners, Sarah Huffman, and Michaela McNaughton for their research assistance. And thanks to Andrew Mingione for his graphic design work. This research was performed with College of Wooster Human Subjects Research Committee approval [#HSRC 2015/04/25].

Notes

1 Much of the human rights literature conceptualizes HRO success as the ability to apply pressure to states or other actors and to prevent or end rights abuses (Hertel 2006; Becker 2013; Brysk 2013). Here, we define HRO success or efficacy in terms of building consensus and action mobilization at the individual level, as this micromobilization is a necessary precondition for overall rights protection and harm reduction (McEntire et al. 2015).

2 See, for example, Keck and Sikkink’s (1998: 165–195) discussion of the emergence and evolution of the women’s rights transnational advocacy network, as well as Carpenter’s (2007) analysis of absence of HRO advocacy on the issue of children born of wartime rape.

3 Conrad et al.’s (2018) and Bracic and Murdie’s (2019) findings on the public’s differential attitudes toward and recognition of rights abuses, based on victims’ identities is instructive here.

4 In addition to the examples discussed above, consider also Carpenter (2005: 304–305), Moorti (2011), Amnesty International (2015), Human Rights Watch (2017), and Associated Press (2019).

5 The occupational stereotyping literature refers to fields that are viewed as being for men or for women as sex-typed, which contrasts with our choice of language of gendered social roles. Gendered social roles is a broader term to encompass inferences based on social constructs of gender regarding a wider variety of social roles (e.g., roles in the home and family, occupational).

6 Scholars have recently written critically of the ethics of crowdsourced research, and MTurk, in particular (Marinova, 2016; Williamson, 2016). The concerns raised include below-minimum-wage compensation for research participations, absence of employee benefits and other legal protections, alienation, and the lack of worker enrichment that results from piecemeal labor. Most problematic, according to Marinova (2016: 426), is the cover of legitimacy that academics provide to corporations like Amazon’s MTurk by using its services to conduct research: “Even though most would agree that participating in a research experiment does not constitute actual labor, by fielding experiments through [MTurk] we nevertheless support a legal and economic model based on employment insecurity, financial and legal asymmetry between employers and employees, and absence of labor representation.” We have reflected on these critiques carefully; agreeing with many of the author’s conclusions, we are actively considering alternative online sites for future data collection initiatives.

7 ANOVA analyses reveal that there are no statistically significant differences across groups with regard to mean age (F = .289, df = 8, p = .970), education level (F = 1.242, df = 8, p = .271), or partisan affiliation (F = .742, df = 8, p = .655). With regard to sex of the respondent, a chi-square test ($x^2=16.146, 8 df, p=.040)$ indicates that the composition of male and female respondents differs across treatment groups.

8 As a robustness check, in all models we also examined the impact of each treatment independently on the dependent variables in bivariate models. The results are not substantively different. Because participants were randomly assigned to the treatment groups within the experiment, we did not expect respondent demographics to impact our results. However, we ran all models with control variables measuring various sociodemographic characteristics of the respondents (age, sex, race, political party affiliation, education, etc.) that have been shown to affect public opinion on human rights violations (see Richards et al. 2012). These models are available in the online appendix, in Tables OA7–OA9 and OA11). In light of our argument about the gendered framing of human rights appeals, we are most interested in the effects of respondents’ sex on our outcome variables. Based on prior research, we do not expect male and female respondents to react differently to the stimuli. This is because men and women similarly internalize gender stereotypes (Hentschel, Heilman, and Peus 2019) and both men and women across countries score highly on benevolent sexism measures (Glick and Fisk 2001; de Lemus, Moya, and Glick 2010). Respondents’ sex had no significant effect, with the single exception that men were less likely than women to view the victim as vulnerable (see Table OA9).

9 We were surprised to find that seeing an HRO campaign featuring a female victim of violence decreased respondents’ characterization of the victim as innocent (see Model 1, Table 2). We suspected that the “resisting arrest” element of the masculine stereotypic role frames (for both men and women victims) may explain this counterintuitive finding. Limiting the analysis to only the treatments with feminine stereotypic role descriptions (thereby excluding the resisting arrest element), we can examine differences between treatments with a male vs. a female victim. As expected, after removing all treatments in which victim was described as resisting arrest, HRO campaigns featuring a female victim of violence (compared to a male victim) significantly increased respondents’ characterization of the victim as innocent.

10 The effects of the treatments on the other “feminine” and “masculine” traits are listed in Tables OA8–OA10 in the online appendix. Note that the described gender roles of the victims influence how respondents see them, increasing their association of the victims with stereotypically feminine characteristics, and decreasing their association of the victims with stereotypically masculine characteristics.

11 Running the same model with the full sample of respondents who saw any version of the sleep deprivation treatment yields substantively similar results; perceptions of innocence are statistically significant, but you need to move from the “not at all” to “very well” categories before you observe a substantive difference in the effect on consensus mobilization.

12 Upon reflection, we were concerned that our operationalization of masculine stereotypic gender roles—in particular, describing the victim as resisting arrest—may have led respondents to see such victims as less innocent not because they held masculine traits (powerful, brave, assertive), but because resisting arrest is often perceived as something only the guilty would do (see also note 5). However, this aspect of the narrative was not as salient for readers as were descriptions of the violation and its impact on the victim’s health. Three-quarters of respondents cited these latter elements as most important, whereas only 14 percent of respondents said the same about the description of the victim resisting arrest. As a robustness check, we reran all models, adding the variable “resistarrest” (which = 1 if the respondent indicated that this was one of the two most important elements of the narrative) as a control variable. As one might expect, people who thought resisting arrest was a salient part of the story were, in fact, less likely to view the victim as innocent or vulnerable. All results in Models 1–7 and 9–10 are substantively the same. With the addition of resistarrest in Model 8, perceptions of the victim’s vulnerability is no longer a statistically significant predictor of signing the petition against sleep deprivation.

13 We also examined the impact of each treatment variable independently on the dependent variables in bivariate (excluding all control variables) and multivariate (including all control variables) models. The results are substantively the same, unless otherwise noted.

Additional information

Funding

This research was supported by the Faculty Development Fund at the College of Wooster.

Notes on contributors

Michele Leiby

Michele Leiby is an associate professor of political science at the College of Wooster. Her research and teaching interests lie at the intersection of comparative politics and international relations, with a focus on human rights and political violence, transitional justice, and gender and conflict processes.

Angela L. Bos

Angela L. Bos is a professor of political science at the College of Wooster. Her research focuses on how gender and gender roles shape political socialization, engagement, and evaluations.

Matthew Krain

Matthew Krain is a professor of political science at the College of Wooster. His work examines the causes and consequences of repression and large-scale political violence, and the role of both the state and the international community in causing, mitigating, or preventing conflict and violence.