Female Representation and Legitimacy: Evidence from a Harmonized Experiment in Jordan, Morocco, and Tunisia

How does the gender composition of deliberative committees affect citizens’ evaluations of their decision-making processes? Do citizens perceive decisions made by gender-balanced, legislative bodies as more legitimate than those made by all-male bodies? Extant work on the link between women’s descriptive representation and perceptions of democratic legitimacy in advanced democracies finds the equal presence of women legitimizes decision-making. However, this relationship has not been tested in more patriarchal, less democratic settings. We employ survey experiments in Jordan, Morocco, and Tunisia to investigate how citizens respond to gender representation in committees. We find that women’s presence promotes citizens’ perceptions of the legitimacy of committee processes and outcomes and, moreover, that pro-women decisions are associated with higher levels of perceived legitimacy. Thus, this study demonstrates the robustness of findings from the West regarding gender representation and contributes to the burgeoning literature on women and politics.


INTRODUCTION
T he past two decades have witnessed a dramatic increase in the presence of women in decisionmaking bodies. Existing evidence from established democracies demonstrates that such representation increases citizens' perceptions of legitimacy in political institutions (Mansbridge 1999;Scherer and Curry 2010) and outcomes (Banducci, Donovan, and Karp 2004), thereby promoting institutional trust (Gay 2002;Ulbig 2007). Descriptive representation may even legitimize decisions which adversely affect women (Clayton, O'Brien, and Piscopo 2019). However, descriptive representation may not have uniform effects across different settings (Lee, Solberg, and Waltenburg 2021). Increased women's representation in decision-making bodies may engender backlash, especially in settings with conservative gender norms (Biroli and Caminotti 2020;Yildirim, Kocapınar, and Ecevit 2021).
We test whether recent findings linking women's descriptive representation and democratic legitimacy extend from established democracies with higher levels of gender progressive norms to less democratic, gender conservative contexts. To do so, we employ a survey experiment that varies two main treatment dimensions: a legislative committee's gender composition and its decision (expanding or limiting women's rights). 1 We implement the experiment as a harmonized study (Slough and Tyson 2023) in three Middle East and North African (MENA) countries-Jordan, Morocco, and Tunisia-which have more gender conservative societies and less democratic regimes than the sites of most previous studies on this topic. We focus on laws against domestic violence, which is estimated to afflict one in three women globally, or nearly 736 million women (World Health Organization 2021).
Contrary to expectations, we find that Jordanians, Moroccans, and Tunisians view equal inclusion of women in the decision-making process much the same as citizens in the West. Women's inclusion in decisionmaking and pro-women decisions increase respondents' perceptions of the legitimacy of processes and outcomes. Moreover, pro-women decisions increase respondents' expectations that the public will accept the committee's decision. Thus, a second importantand somewhat surprising-conclusion from the study is that respondents generally support increased penalties against domestic violence.

GROUP GENDER COMPOSITION AND LEGITIMACY
Descriptive representation, where representatives' demographic characteristics mirror the population from which they are drawn, is often conceptualized as "the politics of presence" (Mansbridge 1999). The argument for descriptive representation is based on the premise that elected officials are more likely to "act for" those with whom they share personal characteristics (Lovenduski and Norris 2003;Pitkin 1967). Descriptive representation should lead to fairer outcomes (Easton 1965;Gay 2002) and serve to cushion unfavorable decisions (Arnesen and Peters 2018). Thus, it can improve the quality of policies, particularly regarding women and other marginalized groups (Banducci, Donovan, and Karp 2004), and bolster the legitimacy of legislative bodies. 2 Studies exploring the link between women's descriptive representation and democratic legitimacy have proliferated over the past decade (Clayton, O'Brien, and Piscopo 2019;Lee, Solberg, and Waltenburg 2021). Research on the symbolic representation of women in politics (i.e., the attitudinal and behavioral effects of women's representation) (Lawless 2004) has found that women's numerical presence in decision-making bodies improves evaluations of decisions (i.e., substantive legitimacy) and increases trust in the institutions (i.e., procedural legitimacy) (Clayton, O'Brien, and Piscopo 2019). Yet other studies have shown that increased women's representation may lead to backlash against women (Krook 2015). Women's presence in previously male-dominated spaces may trigger "renewed determination by patriarchal forces to maintain and increase the subordination of women" (Walby 1993, 79). Backlash may manifest as violence against female politicians, be directed against women outside politics whose demands challenge the existing gender hierarchy (Berry, Bouka, and Kamuru 2020), or, we argue, be expressed in attempts to delegitimize genderbalanced committees.
Most studies to date have been conducted in contexts with more liberal gender norms, leaving open questions about how well findings travel. We expect backlash effects to be more pronounced in gender conservative societies, such as in the MENA, where patriarchal norms and discriminatory laws marginalize women within decision-making processes. Notably, most gender reforms across the MENA were introduced from the top to improve regimes' domestic and international reputations (Tripp 2019), with little effort to transform deep-rooted gender inequalities within the society and/or the economy. We anticipate backlash effects will be more pronounced in these settings given the lack of transformative measures to improve women's status overall. Because backlash against increased female representation is more likely to occur when women are increasingly visible as political actors (Berry, Bouka, and Kamuru 2020;Krook 2015), we posit that the equal presence of women in legislative bodies should have a negative impact on citizens' perceptions of their substantive and procedural legitimacy.
We anticipate this will be true even in less democratic regimes. Legislative assemblies in authoritarian regimes are often sites of co-optation, information signaling, and contestation over policy outcomes (Gandhi and Lust-Okar 2009). Autocratic legislatures may have more limited powers than those in more democratic settings, but parliamentarians still study, discuss, and approve or reject legislation (Shalaby and Elimam 2020). 3 Thus, we propose the following pre-registered hypotheses: 4 H1: Citizens will be less likely to agree that the committee made the right decision when the committee is gender-balanced (i.e., substantive legitimacy).
H2: Citizens will be more likely to report negative attitudes regarding the committee's decision-making process when the committee is gender-balanced (i.e., procedural legitimacy).
H3: Citizens will be less likely to believe that the general public will accept a decision made by a genderbalanced committee.
H4: Committee decisions supporting women will further increase the negative effect of gender-balanced committees for all outcomes in H1-H3.

GENDER NORMS, REGIMES, AND DOMESTIC VIOLENCE
We test our hypotheses in Jordan, Morocco, and Tunisia, focusing on deliberation over domestic violence penalties. This is an important, gendered issue in the MENA. 5 A majority of Arab Barometer participants reported in 2019 that domestic violence had increased since the start of the Coronavirus outbreak (Arab Barometer 2019). Even before that, and despite the highly sensitive nature of the issue, about one third of female respondents in Jordan (Clark et al. 2009) and around half of Moroccan (Kasraoui 2019) and Tunisian (Veen, Jrad, and Galand 2017) women reported experiencing domestic violence in their lifetime. Thus, examining this topic at this point in time lends additional value to this study.
Jordan, Morocco, and Tunisia have genderconservative societies. As shown in Figure 1a, respondents in each country are much more likely than Americans to agree with the statement that men make better political leaders than women. Conservative gender norms are also evident in the controversies over legislation regarding domestic violence, the focus of our study. All three countries have passed legislation outlawing domestic violence, but many still find the legislation insufficient. 6 Moreover, prominent political elites in Jordan (Watkins 2020), Morocco (Etezadi 2016), and Tunisia (Abdo-Katsipis 2017) have opposed strengthening domestic violence laws. Our survey also offers evidence that this issue is contested: about one in five Jordanians, one in four Moroccans, and one in three Tunisians believed that domestic violence penalties should not be raised.
The countries we study are governed by nondemocratic institutions (as shown in Figure 1b). Jordan and Morocco are monarchies with elected legislatures, which Freedom House deemed not free and partly free, respectively, in 2021. Tunisia's revolution in 2011 transformed it from an autocracy to a fledgling democracy, which Freedom House rated as free in 2021. Yet Tunisia experienced democratic backsliding, including the disbandment of parliament, just prior to the fielding of our experiment.
All three countries hold political deliberations over domestic violence penalties, in which both men and women participate. Jordan, Morocco, and Tunisia have gender quotas, and when we fielded the experiment, women parliamentarians made up about 12% of the elected legislature in Jordan, 24% in Morocco, and 26% in Tunisia. Majorities of respondents in all countries viewed our experimental scenario as realistic (see Figure A4 in the Supplementary Material).
Jordan, Morocco, and Tunisia thus allow us to examine findings from established liberal democracies in understudied settings. In particular, we study the effect of group gender composition on substantive and procedural legitimacy in cases with greater gender conservatism and less democratic institutions. We examine potential differences across the three cases, but our goal in doing so is to test if our findings are generalizable across these disparate settings. We do not expect or seek to explain cross-country variation.

RESEARCH DESIGN
We implemented phone-based survey experiments between November 2021 and March 2022 in Jordan, Morocco, and Tunisia. We asked 4,754 respondents a series of pre-treatment questions and then presented them with our experimental vignette: an excerpt from a mock radio show describing a legislative committee that decided whether to raise penalties on domestic violence. 7 In our experiment, we randomized the committee's gender composition (all male/gender-balanced) and its decision (expanding/limiting women's rights), resulting in a fully crossed 2 Â 2 experimental design. 8 Following the vignette, 9 respondents answered manipulation checks and questions related to our key outcomes. We identify the effects of gender balance on (H1) the evaluation of the committee's decision (a three-item index measuring belief that the committee made the right decision for all citizens, men, and women; α ¼ 0:804); (H2) attitudes toward the committee procedure (a two-item index measuring trust in committee and belief in committee fairness; α ¼ 0:668 ); (H3) perceptions that the general public will accept the committee's decision (a single-item measure); and (H4) that the committee decision moderates effects on the evaluation of the decision, procedure, and public acceptance (interactions of gender-balance and committee-decision treatments).
We estimate the following pre-registered ordinary least squares (OLS) regression, estimating the average treatment effect (ATE) of committee gender balance across all countries: Our main parameter of interest is β balance , representing the gender-balance ATE on a given outcome of interest (y ic ). To increase the precision of our estimate, we control for our second treatment (δ decision ), as well as respondents' country, gender, age, and education (represented by ψ i ). We supplement our main analysis with similar, country-specific OLS regressions. 10

RESULTS
In this section, we present the ATEs of our genderbalanced and committee-decision treatments. Figure 2 depicts the effects of our treatments on respondents' evaluation of the committee's decision (H1). The left side of Figure 2 shows that, in aggregate, gender balance modestly improved respondents' evaluation of the committee's decision by 7% of a SD. The pro-women decision treatment also improved respondents' evaluations of the committee's decision. Indeed, the decision appears to shape respondents' evaluations of substantive legitimacy more than the committee gender balance does. In the aggregate, the effect of the prowomen decision treatment is almost 8.5 times larger than the effects of the gender-balance treatment, and it is significant at the p ¼ 0:01 level in all countries. Thus, the results in Figure 2 stand in stark contrast to our preregistered expectations. Instead of backlash, we find that gender balance and pro-women decisions raise evaluations of the committee's decisions.
In Figure 3, we report the effects of our treatments on attitudes toward the committee procedure (H2). We find 7 A translated version of the vignette and an overview of survey methodology and measurement are provided in Appendix A of the Supplementary Material. In our vignette, the committee is unnamed, but recent examinations of abstraction in survey experiments suggest that an unnamed committee should not substantially impact inferences drawn (Brutger et al. 2022 Kristen Kao et al. that gender balance increases respondents' positive attitudes toward the committee by over 13% of a SD, and this effect is consistent across all countries. Pro-women decisions increased positive attitudes toward the committee, and they are about four times larger than the effect of the gender-balance treatment. Again, these results stand in contrast to our pre-registered hypothesis and they suggest that gender balance and pro-women decisions increase procedural legitimacy in the MENA. In Figure 4, we consider the extent to which our treatments shape respondents' expectations regarding public acceptance of the committee's decision. Our results suggest that gender balance does not have a precisely estimated effect on this outcome, but prowomen decisions increased respondents' expectations that the public will accept the committee's decision. In line with the findings reported in Figures 2 and 3, these stand in contrast to our pre-registered expectations.  Finally, in Figure 5, we consider whether the committee's decision in favor of women moderates the effects of our gender-balance treatment on our key outcomes from H1-H3 (H4). We regress the three outcomes over our two treatments and the interaction, BalanceÂPro, representing the moderating effect of pro-women decision treatment on the gender-balance treatment. We find that the committee's decision does not moderate the ATE of gender balance on our key outcomes and thus we find limited support for our fourth hypothesis.

DISCUSSION
The positive effects of gender balance and pro-women decision run counter to our theoretical expectations  and prompt us to ask why we do not find the backlash we anticipated. We consider five plausible explanations.
The first explanation is that respondents conceal their responses to public policy decisions, even if they express discriminatory personal beliefs. In authoritarian regimes that support women's rights, citizens may feel that they need to respond positively to pro-women decisions and gender-balanced committees. To examine this possibility, we analyze whether respondents who state that they support the regime are more likely to respond favorably to pro-women decisions and gender-balanced committees than those who do not. We find little evidence that attitudes toward the regime moderate the gender-balanced treatment effects (see Appendix C.9 and Figure A20 in the Supplementary Material).
A second explanation is that the results are only relevant to our issue area, domestic violence penalties. We leverage a second issue area treatment fielded in Jordan, on penalties for littering, and find that issue area does not moderate the effects of the genderbalance treatment. The decision treatment has a similar, albeit larger effect when the issue area is littering, likely reflecting the overall greater support for increasing penalties in this area (see Appendix C.8 of the Supplementary Material).
A third possibility is that there are significant heterogeneous effects, running in roughly equal and opposite directions, which lead to the appearance of small and statistically insignificant aggregate effects. We test whether sexist attitudes (see Appendix C.5 of the Supplementary Material), perceptions of gender norms (see Appendix C.6 of the Supplementary Material), or respondent's gender (see Appendix C.7 of the Supplementary Material) moderate the effects of the treatments on perceptions of substantive or procedural legitimacy. Overall, as further discussed in Appendix C.10 of the Supplementary Material, we find limited evidence in support of effect heterogeneity. This result may be driven in part by measurement challenges in our sexism indices, which we address in Appendix C.5 of the Supplementary Material by employing different measurement strategies. However, we encourage future research to build on our work and further explore heterogeneity using other innovative measures.
A fourth potential explanation is that the results reflect treated respondents' beliefs that it is socially unacceptable to express opposition to positions that improve women's rights. This may be particularly problematic in this study, as some have found that telephone surveys elicit greater social desirability bias than faceto-face surveys (Holbrook, Green, and Krosnick 2003). Yet, in line with recent studies (Blair, Coppock, and Moor 2020), we also have reason to doubt that such bias explains our results. Respondents express sexist attitudes in direct questions and that sexism is particularly prevalent in Jordan, the country in which gender balance had a positive and significant effect on substantive legitimacy.
In the Supplementary Material, we also implement a number of robustness and sensitivity checks. We test We are left to conclude that there may simply be less variation in attitudes toward women's representation than anticipated. Issues of domestic violence and gender representation have become globalized, with domestic leaders and international stakeholders pressing for changes in policies and practices around these issues. There may still be differences in opinions over what constitutes "domestic violence" or the roles of women in politics. But, as we show, there may be popular convergence in attitudes when it comes to some of the globally promoted policies that seek to improve women's welfare.

CONCLUSION
This study advances the literature on representation and bridges an important gap in our understanding on the intersection of gender and politics. Our results are both surprising and important. Despite the diverging contexts in which we implement our studies, women's representation and pro-women decisions appear to have positive effects on the legitimacy of decisionmaking bodies and their outcomes. Indeed, the effects of women's descriptive representation are similar to those found in studies from more gender liberal societies in Western democracies and it appears that citizens especially value pro-women decisions.
Our findings prompt scholars to delve deeper into understanding how context moderates the impact of descriptive representation on substantive and procedural legitimacy. Our study is unable to disentangle effects of less democratic regimes and patriarchal social norms, although it suggests that-at least taken together-they have less impact on the link between descriptive and substantive representation than one might expect. This should be more robustly tested, employing cross-national comparisons across a larger sample of cases that vary social norms and regime type. Furthermore, scholars should investigate the extent to which other contextual factors (e.g., intersectional identities 11 or cleavage structures) may moderate the relationship between descriptive representation and legitimacy.
This study also has important implications for policymakers. It suggests that policies aimed at promoting gender representation may enhance the legitimacy of institutions and policy outcomes, even in less democratic, gender conservative contexts. Yet it also raises questions about the relationship between gender representation, the legitimacy of these institutions, and human rights. Does increasing gender representation in such institutions help to stabilize authoritarian regimes, thwarting efforts at democratization or improved human rights? Or does increased legitimacy of such institutions strengthen potential loci of democratization? More research should be done to examine the implications of gender-balanced decision-making bodies in real-world contexts. Only by further exploring these outcomes can we fully understand the effects of gender quotas, campaign support, and other programs aimed at increased women representation.

SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S0003055423000357.

DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi. org/10.7910/DVN/5GP5SI.

ACKNOWLEDGMENTS
For helpful feedback on earlier drafts of this paper, we thank Amanda Clayton, Diana O'Brien, Jennifer Piscopo, and the participants of the Comparative Politics Colloquium at the University of Wisconsin-Madison, the 2022 American Political Science Association meeting, and the University of Gothenburg's Politics and Gender Workshop. We are also grateful to the four anonymous reviewers at the APSR, whose comments and suggestions have substantially improved the article. For assistance in implementing the research, we thank enumerators at Elka Consulting in Tunisia, Oneto-One for Research and Polling in Morocco, and NAMA in Jordan, and Kirk Ammerman and Erica Ann Metheney at the Governance and Local Development Institute for assisting in data collection; administrative assistance from Rose Shaber-Twedt, and research assistance from Jennifer Bergman.

FUNDING STATEMENT
We thank the Carnegie Corporation of New York for funding this project.