Power of Suggestion? Leadership Signals, Politics, Religion, and Women’s Support for the Disadvantaged

We assess the role of social signals about the appropriateness of women in leadership roles in either the political or religious domain. The relevant literature leads to expectations of a relatively clear effect on women’s efficacy levels when encountering social suggestions that women’s skills are better used in other ways. However, less certain is whether encouraging women away from leadership in religious organizations impacts their sense of effect on political outcomes, including policies in support of disadvantaged outgroups. Utilizing a framing experiment embedded in a statewide public opinion poll, we find that social suggestion that women should stay away from religious leadership has statistically significant and negative effects on efficacy levels among our sample’s subset of evangelical women. At the same time, these anti-religious leadership signals move evangelical women away from supporting policy statements benefitting a disadvantaged outgroup. Given that it is the power of suggestion regarding religious, not political, leadership that leads to the effects among evangelical women, we offer additional paths for future research to explore on this wider topic.

The lingering resonance in Bella Abzug's observation that women are well served by speaking "softly while carrying lipstick" belies the notion of gender equality in American politics, even in the decades since the much heralded "year of the woman" in-roads (see Paolino 1995). Indeed, and despite the role of feminist movements in raising public consciousness of gender inequality, there is little to suggest that women have overcome susceptibility to the diminishing effects of gender role positioning vis-à-vis men (Eagly 1987;Eagly et al. 2004). With acceptance of substantial political roles for women arguably never higher, whether and why women remain susceptible to suggestions that their political involvement is a misplaced priority are intriguing questions. The question we consider in this manuscript is the extent to which women are affected in their self-reported sense of political efficacy as a result of the power of suggestion about "appropriate" women's roles in political and religious settings-the latter having clear overlap with the civic skills and participation interest of the former (see Gilbert 2002, 2009). Our second, policy-oriented item of interest derives from this power of suggestion effect and regards the degree to which women, as they potentially ascend toward increased political status, begin to show less support for other scrutinized or disadvantaged political groups, perhaps as a result of their own maltreatment in politics, religion, and society. In this way, political and religious representations behind the relative political deprivation women encounter may take on a tangible, policy-oriented angle as women become less interested in offering support for causes and groups in need of help.
Drawing on social psychology theories about self-presentation and self-stereotyping in response to ingroup and outgroup signals, we assess whether randomly exposing adults to primes offering approval/disapproval of women's involvement as in political or religious contexts significantly affects (1) women's reported political efficacy (relative to men) and (2) policy preferences related to other political minorities. Note that our use of the term "disadvantaged" throughout this manuscript is intentionally broad, while our empirical assessment focuses on a narrow example of people facing the disadvantage of heightened prejudice and scrutiny because of cognitive or physical challenges. The usefulness of focusing on cognitive or physical development issues as the example of the "disadvantaged" here is that it should be far less correlated with partisanship and political ideology vis-à-vis other (more commonly referenced) points of disadvantage (e.g., race/ethnicity, religion, and gender). This enables us to disentangle some of the efficacy-related effects from our analysis on how women might respond to some of those in need. That said, there are clearly other ways to characterize the "disadvantaged" that are specific to different regional and international contexts.
Our assessment of religion's impact on women's sense of political efficacy is drawn from the political science literature's focus on religion as generating civic skills for adherents (i.e., Burns et al. 2001) but also adopts the perspective advanced by Djupe et al. (2007) that the religious resources available to women and men differ. The result, we suggest, is that women, who have been considered more interested in maintaining relationships and supporting social services from government, may be convinced to turn away from that position when the dual pressures of social suggestion and religious experience in being denied the same resource opportunities as men mount (see Shorrocks and Grasso 2020 for a comparative review of gender gaps in support of social welfare spending).
Consideration of women's political efficacy more broadly speaking is found throughout the political science literature. Overall, the story is of progress in moving toward greater participation, and assumption of leadership positions, over time. Andersen (1975) began looking at the changes in women's workforce participation and resulting changes in resources that she argued facilitated the changes in women's political participation for the better. However, scholarly focus on the question of resources required for political participation-including the experiences and impetus to support runs for leadership positions-showed that women seemed to lag behind men . By the late 1990s and early 2000s, resource models fell out of favor as women's education levels surpassed those of men and as women began to compete with their male counterparts for top jobs in all professions, but the question remained: Why are women not represented equally across levels of government? While some scholars continued to tackle this question (Baitinger 2015;Burns et al. 2001;Lawless and Fox 2005, 2010, 2015, others wondered what disadvantages women, men, and our political system as a whole continue to face by not being represented equally (Campbell and Wolbrecht 2006;Costain and Fraizer 2002;Gidengil et al. 2010).
As this literature shows, the question of political representation is generally treated descriptively from the standpoint of how many women are elected to office at all levels of government. However, we consider representation just as much a cognitive-based sense of how one perceives themselves in relation to the political system-including the sense of having power to engage in the political system-irrespective of the number of women holding positions of power, and this includes the way in which religious and political conceptions of self-agency in the wider world impact people. This means we transition from a focus on women as officeholders and leaders to one where the personal sense of self and meaning comes from social and religious reference points.
Of course, scholars have long considered the effects that religious beliefs and identity can have on views of gender roles, with women generally considered appropriate for subservient institutional roles in church and family (Richardson 1988;Ammerman 1990). That mainline and some charismatic evangelical denominations have made strides in fostered inclusion of women in ordained and other formal leadership posts in recent decades (see Chaves 1999) does not mean that the broader, and far more historic status quo of male-dominated faith communities has been substantially overcome in terms of the extent to which women may be made to feel inferior to men in terms of the contributions they make. Nevertheless, despite the history of male dominance in religious life, one of the most common differences sociologists have noted between women and men has to do with a heightened sense of religious commitment and interest among women (see Ferraro and Kelley-Moore 2000;Miller and Stark 2002). This is intriguing in that it makes the effect of religious identity salient for women, even as they are not able to partake in the full measure of institutional power and influence that men are (although some of this dynamic is institutionally specific-see Kraybill 2019). As with the variable nature of women's lived experience in politics, the question of whether the power of religious role suggestion might influence women's sense of efficacy when separate from, or combined with, suggestion about women's appropriate political roles is intriguing. It is so because of the test established between women's struggles for improving human rights through the work of religious institutions (which have largely limited leadership roles in said institutions). It might just be that women, eventually, lose momentum in their efforts to improve conditions for the disadvantaged, and suggestions about appropriate female leadership roles play a critical part in affecting that momentum.
To situate our understanding of the mechanisms at work in determining both lowered political efficacy and its follow-on push toward lower support for helping the disadvantaged, we offer a brief overview of the social identity mechanism we consider to be behind this effect.

The Precarious Position of Women in American Politics
Social identity boundaries may explain much of the longevity of traditional role stereotypes women encounter in both religious and political contexts (Tajfel et al. 1979;Tajfel 1982). While there is a developing literature on gender deviance-with traditional gender roles eschewed by men and women (see Doorn et al. 1994; Rudman and Glick 2012)-these roles maintain significant cache with much of the public (Witt and Wood 2010). For example, scholars continue to find men exhibiting higher tendencies toward social dominance when embracing a masculine gender identity through self-stereotyping (see Pratto et al. 1997;Foels and Pappas 2004). Meanwhile, women are usually stereotyped into the role of nurturer or caretaker (Bem 1983;Bryant and Check 2000;Egan and Perry 2001;Eagly et al. 2004;Bianchi et al. 2006).
To the extent that women come to associate themselves with the nurturer stereotype, their self-conceptions and behavior can be dominated by the male stereotypes of breadwinner, powerbroker, leader, etc. (Cejka and Eagly 1999;Ryan et al. 1983). Importantly, these feminine roles have been linked to lowered attention to public affairs versus men (Lane 1959;Welch 1977;Jennings 1979). The relatively stable differences in gender-based political attitudes are also well documented (Whitaker 2008), particularly in women's tendency to support more liberal political policies and candidates than men (Baxter and Lansing 1981;Rapoport et al. 1990;Kaufmann and Petrocik 1999;Kaufmann 2002;Norrander 2008). In terms of religious beliefs, women are a decidedly mixed group representing both liberal and conservative perspectives across traditions (see Neitz 1987;Manning 1997;Dochuk 2010). As such, there may be more variability in how women respond to religious expectations concerning gender roles, especially in terms of holding leadership positions over men.
These trends notwithstanding, there is reason to explicitly test the mechanisms by which gender stereotypes affect political outcomes in men and women. Part of this is methodological. While some scholars have made gender primes a key factor of their analysis (e.g., Schaffner 2005;Burdein et al. 2006), few studies use randomly assigned gender primes on population samples to assess direct effects. Additionally, despite the fact that Abzug's "lipstick" characterization remains relevant, women have actually improved in aspects of their political involvement over time (Rosenstone and Hansen 1993;Verba et al. 1995;Jamieson et al. 2002). Something, therefore, must be driving this change. Still another reason for this assessment concerns the source of the gender stereotypes themselves. If gendered expectations about leadership are rooted in social identity differences, it is important to compare effects when the same stereotype is offered by gender ingroup vs. outgroup members. However, "ingroup" and "outgroup" in this case is not limited to gender distinctions only. Instead, religious identity may also play a key role. If both gender and faith identities are active in setting expectations about "appropriate" women's roles, women may be motivated to conform their policy views in specific ways.
Central in the process of stereotype influence is the role played by an individual's desire to self-monitor and adapt behavior based on external stimuli and social comparisons (see Snyder 1979). At issue is self-presentation, as both women and men are susceptible to concerns about how others perceive them (including and especially members of the opposite sex) (Graziano and Bryant 1998). Extending these ramifications to the political and religious realms, gender outgroup expectation signals may be critical in determining one's sense of efficacy. However, by the same token, shared reality theory expects self-stereotyping to be premised on the most cognitively accessible social identities that one has (see Markus and Wurf 1987;Sinclair et al. 2006). Assuming that biological sex and associated gender identities are salient for most adults, it makes sense that signals containing both ingroup-and outgroup-derived gender expectations affect feelings of political efficacy and perceptions of other political groups.
In making one feel accepted and even emotive, ingroup identity processes become increasing relevant to a person's self-concept and political evaluations (Forgas 1995;Waschull 1995;Forgas and Bower 2001). The identity congruence one senses between self and a specific gender ingroup (and its associated norms) positively correlates with self-esteem (Bem 1983;Rowley et al. 1998). To the extent that feelings of domain-specific efficacy are determined in grade school, low self-esteem about political matters as driven by signals of ingroup norms in women may explain their generally lower levels of political interest from an early age (Easton and Dennis 1967;Sniderman 1975;Sigel and Hoskin 1981). Supporting this is Conway's (Conway et al. 1996) suggestion that belief among women that men are better able to understand politics (and, by extension, wielding power in groups) partly explains women's lower political interest. In this case, women have essentially told themselves that men are better able to handle politics. Hence, this gender stereotype, even if it did not originate in the gender ingroup, is reinforced by it.
The comparative effects from gender stereotypes perpetuated by ingroup and outgroup sources suggest that women may be susceptible to the political or religious powers of suggestion. This possibility is especially intriguing for how women react to dominant ingroup signals about their political participation/leadership in relation to what Conway et al. (1996) found. Hypothetically, if women are encouraged to participate politically by other women, does it galvanize women to take up support of other disadvantaged political groups as result? Or, do we see evidence that women, as they begin to feel more politically empowered-and coupled with societal shifts welcoming increased female activity-show less concern for other disadvantaged political groups (including those that are non-gender-differentiated)? Theoretically, lowered efficacy levels might mean that women abandon policy concerns for vulnerable populations because they feel not only as though they have the ability to help themselves but are unable to improve conditions for others. That said, and though the literature suggests that women are more attached to religion than men, there is reason to suspect differences in political efficacy levels when people are exposed to suggestions about religious leadership positions. Specifically, the literature does not make it clear that the denial of leadership opportunities for women in the religious domain spills into the political realm (and vice versa). This means that our design is able to test the comparative differences in effect between how the power of suggestion about leadership roles impacts political efficacy when the suggestion pertains to religion, not politics. If religious suggestion does not affect political efficacy, then it stands to reason that neither will it impact women's support for vulnerable populations (even if political suggestion has such an effect). The following section focuses our assessment on these and related questions.

Research Design
Our examination of gender stereotype signals on women utilizes data from a population survey experiment conducted as part of a statewide telephone poll of Missouri adults conducted over a 6-day period in late January 2013. Using Random Digit Dial on landline telephone numbers with Missouri telephone area codes, we garnered participation from 911 adults (18+) living in the state (AAPOR Response Rate 4, 14.5%; MOE, 3.2%). However, since we are interested in how the power of suggestion of a religious role for women affects efficacy and support for disadvantaged populations, we included oversamples of self-identified religious individuals using the General Social Survey screeners. Our sample included "born again" or evangelical Protestants (30% of respondents), Roman Catholics (26% of respondents), and mainline Protestants (20%). The remaining six percent were distributed between those claiming no affiliation (11%), Jewish (3%), agnostic (5%), Orthodox Christian (2%), Muslim (2%) Buddhist (1%), and Hindu (1%). In terms of gender, 51% of the sample are female, 48% are male, and 1% is non-binary. Though this is not a representative sample of the adult population, it is a useful sample from the standpoint that most subjects have at least a nominal religious identity that should be open to any effects from the power of religious suggestion concerning leadership roles.
Our experiment was basic in design. Based on the notion that priming certain items or concepts can have significant effects on survey response (see Fiske and Taylor 1991), we randomly primed one of two gender-oriented signal treatments during the course of the telephone survey. The CATI (Computer Assisted Telephone Interviewing) system used was programmed to randomly assign subjects to one of three conditions (the third was the control group receiving no prime). Both treatments contained a signal statement consisting of a single sentence report from "a recent study" (control group subjects were not asked to listen to anything-the survey simply continued to our various outcome measures). In keeping with our expected assessment of ingroup and outgroup gender signals, the cue statements varied according to whether they offered the suggestion that women's leadership skills were better used somewhere other than (1) politics (group one), (2) religion (group two), or a control with no suggestion offered.
The suggestive primes read as follows (treatment differences in parentheses): "According to a recent study, most people suggested that, while women can make good contributions as (political/religious) leaders, women are better off using their leadership skills in other fields." The number of assigned subjects to our treatment and control groups are: politics (299), religion (301), and control (311). Subjects were then asked a series of items tapping dimensions of their external and internal political efficacy. The internal efficacy questions follow Morrell (2003) use of the "well-qualified", "pretty good understanding", "good job in public office", and "well-informed" indicators 1 . Each was arranged on a 1-5 response scale, with 5 indicating "strongly agree". We used responses to these questions to generate an Internal Efficacy Index based on polychoric factor analysis (which considers ordinal scale items) (rotated eigenvalue = 2.55; Alpha = 0.89). The External Efficacy Index (also using polychoric factor analysis and the same 1-5 response scale as the Internal Index) (rotated eigenvalue = 1.26; Alpha = 0.64) utilizes subject response to the "public officials don't care" question featured in Atkeson and Carrillo (2007), as well as the standard "make a difference in politics" question. We also included the "most people can be trusted" measure to round out the External Index.

Efficacy and Suggestion Signals
Our first assessments focus on a series of difference of means tests using ANOVA. These represent the mean outcomes for the two groups vs. the control on the efficacy indices. We first separated women and men on the Internal Efficacy Index according to their assigned groups. Though it is clear our male subjects were entirely unaffected by the randomly assigned suggestion signals in our experiment, our women subjects were highly impacted, and this impact was due to the religious, not political suggestion. Specifically, women subjects showed significantly lower reported levels of internal efficacy when receiving the religious treatment (female mean of 2.4 versus male of 2.9, p < 0.01). Meanwhile, 1 The full question wording for these survey items are: "I consider myself well-qualified to participate in politics." "I feel that I have a pretty good understanding of the important political issues facing our country." "I feel I could do as good a job in public office as most other people." "I think I am well-informed about politics and government as most people." it appears that evangelical women exposed to the religious suggestion drive the downward turn in efficacy (mean of 1.7) versus Roman Catholics (2.5) and mainline Protestants (3.1).
In terms of mean scores on the External Efficacy Index, we again noticed that male subjects were not impacted by the suggestion signals, but women were highly responsive to them. In a repeat of the previous pattern, women subjects showed significantly lower external efficacy when being told that "most people" do not approve of women using their leadership in the religious realm (female mean of 2.5 versus male mean of 3.1, p < 0.01), and again, it was evangelical women driving the downward trend (mean of 1.5) versus Roman Catholics (2.6) and mainline Protestants (3.0).
Taken together, these efficacy findings suggest that women are quite sensitive to political leader signals, but specifically when those signals are applied to the religious, rather than the political, realm. In effect, then, both women and men appear to have tuned out the negative power of suggestion when it comes to political leadership. However, keeping women doubtful about their overall ability to effect political change is how evangelical women respond to the public's view of their role in religious leadership.

Gender Signals and Other Political Groups
Feelings of efficacy are but a part of the assessment we conduct. We wanted to see whether the leadership signals affected our subjects' support for a political group whose plight should be ready political territory for women (assuming the issue stereotypes hold). In this case, the timing of the statewide survey experiment provided a chance to assess how subjects viewed those suffering from developmental disabilities, as Missouri lobbyists for the developmentally disabled had made a major publicity push to educate the state through media ads about the need for opportunities and education for the developmentally disabled. The survey included a battery of opinion questions on this issue following our randomly assigned treatments. These questions were scaled on the same 1-5 Likert as for the two efficacy indices. We created an index (polychoric rotated eigenvalue = 2.82) (alpha = 0.74) from the following survey items: "Everyone should be provided the same opportunities to succeed in life." "Everyone can be productive members of society." "Even those with developmental disabilities can perform well on the job." "Employment rights should be provided to everyone, regardless of disability status." "Educational opportunities should not be limited on the basis of developmental disabilities." Based on findings in much of the literature, the general expectation was that female subjects should be more supportive of policy efforts to help those with developmental disabilities, and will, therefore, score higher on the issue index. In building on our prior efficacy findings, we used a path analysis model to assess direct and indirect treatment effects on the issue index. The direct effects run from the two treatment conditions to the issue index, while the indirect effects run from the treatment conditions through the two efficacy index scores (i.e., internal and external) for each subject to the issue index. Table 1 lists the coefficients' jackknife standard errors for the entire model, which is broken out by female and male subjects and includes controls for subject evangelical, Roman Catholic, or mainline Protestant identity, age, race (i.e., African American versus non-African American), and political party identity (i.e., Republican or Democrat). Given the findings from the difference of means tests above, we also include a multiplicative interaction term between assignment to the religious leadership treatment and subject evangelical identity (which we break down in the models according to subject gender).
Overall, the path results confirm the treatment effects on the two efficacy indices discussed previously. Additionally, in line with the efficacy index findings, male opinion about assistance for the developmentally disabled is not impacted by either the political or religious leadership suggestion (either directly or through the efficacy index scores). Instead, it is religious leadership disapproval that disengages female subjects-and evangelical women in particular (according to the interaction terms-from supporting the developmental disabilities items. This interaction effect drives the female score down significantly both directly from the religion leadership treatment to the disabilities issue score index and indirectly through both efficacy indexes to the disabilities issue index, and, again, it is evangelical women who drive the downward turn in support-with evangelical women showing a half point lower issue index score versus their Roman Catholic and mainline counterparts in the direct treatment, and a 0.4 drop when running through both the internal and external efficacy indices (as seen in the interaction term). This effect is noteworthy not only because of its magnitude, but because, again, our treatments focused only on societal suggestion about women in religious leaders. This is far afield of developmental disabilities issues but, in the larger picture, strikes at the heart of the kind of concern for the vulnerable that is normally associated with women (relative to men). OLS path coefficients in two-tailed tests. Jackknife standard errors using 911 replications with replacement in ( ). * = p < 0.05, ** = p < 0.01.

Discussion
What is most intriguing about these effects is that they show evangelical women to be subject to the suggestion of social pressure about their leadership in religious endeavors. That in and of itself is interesting, but, unlike what we might expect from the power of this suggestion, it is political outcomes that are impacted by the treatment-based social perspectives on women in religious, not political, leadership. Additionally, beyond the reduced feelings of effectiveness comes a reduction in support for statements about a vulnerable group and said group's appropriate role in society. It almost seems as though, when evangelical women are encouraged to look away from religious leadership roles, these women then see less of a need for fairness and equal treatment of others who need support in society. This may be thought of as a negative contagion effect.
Though exploratory, these results provide some indication that women-when sensing more control over their political domain-begin to sympathize less with other disadvantaged political groups than when political efficacy is at a lower threshold.
At the least, how the interplay between suggestions about women's leadership in religious settings, and how the feelings and perceptions coming from that suggestion away from leadership maps on to political outcomes-as seen in these results-needs additional unpacking. For example, does the effect hold when applied to policies aimed at aiding other disadvantaged groups? What about when gender-based cues are added, as in the case of women and men either approving or disapproving of women in leadership roles of various kinds? Finally, is our finding that women are not affected in their internal or external political efficacy when encouraged away from political leadership a reflection on the strides women have made in the political realm (and the corresponding lack of similar widespread leadership gains in some areas of Christianity, including evangelicalism)? It seems that the power of suggestion about women's leadership also contains a variety of as yet unexplored avenues for empirical scrutiny.