INTRODUCTION
As scholars have sought to understand why women are vastly under-represented in political office, they have found that the nature of political party recruitment plays an important role. The dominance of men in the existing party networks and the gendered nature of party leaders’ social and professional contacts mean that women are much less likely to be identified as prospective candidates and get recruited. But what happens when recruitment efforts include women?
The research on gendered responses to recruitment has produced little consensus. Observational quantitative data suggest that men and women respond to recruitment similarly, but evidence gathered from party leaders and other political elites indicates that recruitment of women is significantly more difficult. We conducted a field experiment to test how men and women respond to various political party recruitment messages. In partnership with the Utah County Republican Party, we invited 11,610 of the Party's most active caucus-goers to a “Prospective Candidate Information Seminar” (PCIS). We randomly varied the content of the invitation to test whether different messages were more or less successful at recruiting women and men.
We find some differential results between messages; however, due to lower than expected response rates (about 1% overall), it is difficult to draw firm conclusions about the effects of the treatment messages. However, we find that, on average, women were about half as likely as men to respond to party recruitment. Thus, while we are not able to definitively distinguish between the various recruitment messages, our results clearly demonstrate that responses to recruitment are strongly gendered.
THEORETICAL BACKGROUND
Prior research of individuals in candidate pools and those in political office has found little support for the idea that politically engaged women seek office as a natural outgrowth of their political interest and desire for power. Extensive interviews with professionals in fields from which politicians are most often drawn show that women are much less likely to have considered running for office than their male counterparts, even when controlling for political interest and qualifications (Lawless and Fox Reference Lawless and Fox2010). More than half of female state legislators report that they had never seriously thought about running for office until someone else suggested it to them (Carroll and Sanbonmatsu Reference Carroll and Sanbonmatsu2013). These data cast doubt on the universality of the standard political ambition model for candidates in American politics.
Instead, Carroll and Sanbonmatsu (Reference Carroll and Sanbonmatsu2013) argue that women tend to decide to run for office via a “relationally embedded” process. In this model, party recruitment is an important catalyst for sparking women's political ambition. Empirical research supports this perspective. Professional women who were recruited for office by party leaders, office holders, or political activists are more than twice as likely to consider running for office and significantly more likely to have taken some concrete steps toward running than those who were not (Lawless and Fox Reference Lawless and Fox2010, 109). And when asked about the most important factor that motivated them to run for office, 23.8% of female state representatives (vs. 14.8% of male state representatives) and 14.9% of female state senators (vs. 7.8% of male state senators) identified party leader recruitment (Carroll and Sanbonmatsu Reference Carroll and Sanbonmatsu2013). Furthermore, interviews with political elites support the claim that “women's initial decision to run for office is more likely than men's to involve recruitment” (Sanbonmatsu Reference Sanbonmatsu2006, 27). In fact, there is even some reason to believe that recruitment may close the gender gap in political ambition: Professional women who recall being recruited by party officials are equally likely to consider running for office as men who recall being recruited (Fox and Lawless Reference Fox and Lawless2010). In other words, observational data suggest that difference between men and women is not how they respond to recruitment but whether they are recruited to begin with.
On the other hand, the conventional wisdom expressed in interviews with party elites suggests that even when party leaders try to recruit women, they are often much more hesitant than men to respond positively to recruitment. As one party leader observed about recruitment: “Men are much more willing to jump into [running for office] than women. You need to push women a lot harder to do it” (Sanbonmatsu Reference Sanbonmatsu2006, 126).
Because of the observational nature of existing data, it is difficult to untangle the complicated causal story surrounding political ambition and recruitment. Lawless and Fox's (Reference Lawless and Fox2010) data, for example, rely on respondents’ recollections of recruitment, which is likely to be measured with noise. Further, recruitment is likely to be endogenous to political ambition, and it is rarely done in the kind of systematic way that makes it easy to compare outcomes.
Hence, we design the first field experiment that measures gendered responses to political party recruitment efforts. While our experiment is not the first attempt to experimentally study the effect of recruitment messages on political ambition (see Broockman, forthcoming), it is the first to consider gender, to use party-sponsored messages, and to involve a behavioral outcome with a costly time commitment. Moreover, our experimental design uses a unique subject pool of highly politically active individuals. Experimental research on the behaviors and motivations of elites is quite rare, but it is extremely important because it provides an additional level of external validity (Grose Reference Grose2014). Finally, we study Republican Party members, which is important because the gender gap among candidates and office holders is much larger in the Republican Party than in the Democratic Party (Carrol and Sanbonmatsu Reference Carroll and Sanbonmatsu2013). Hence, we believe this research is an important contribution to the growing body of experimental literature on gender and political ambition.Footnote 1
EXPERIMENTAL DESIGN
In cooperation with the Utah County Republican Party (UCRP), we designed a field experiment in which we organized a free, party-sponsored PCIS. The seminar took place in June 2013 and consisted of presentations by local party leaders on topics such as fundraising, networking, and managing a campaign team. Using the party database of caucus-attendees in Utah County, we mailed invitations to participate in the seminar four weeks prior to the event. We invited the 11,610 most active male and female UCRP caucus-goers to attend this seminar.Footnote 2 Male participants (N = 5,561) were randomly assigned to a control invitation or one of the nine theoretically motivated treatment invitations. Female participants (N = 6,049) were randomly assigned to a control or one of the 10 treatments (the same treatments as the men, plus a female-specific treatment). We describe the treatments below and include the full text of each invitation in Appendix 3.
All invitations were addressed to members of the UCRP and were selected to ensure that only one individual received an invitation per household. These invitations outlined the basic description of the seminar and provided a link to the website where individuals could obtain additional information and register to participate in the PCIS. To enter the website, participants had to use an individualized username identification (ID) located on their invitation. This allowed us to identify those who responded to the invitation and collect data on three behavioral outcomes: (1) who logged on to the website, (2) who registeredFootnote 3 for the seminar, and (3) who attended the seminar.
Ten days prior to the seminar, we made 7,709 reminder robocallsFootnote 4 to the households for which we had phone numbers. We also sent registration reminder emails to the 8,395 subjects with known email addresses. The complete text of the robocall and email messages is included in Appendices 6 and 7 respectively. Tables 1 and 2 in Appendix 5 report the breakdown of the reminder messages by gender and treatment. There were no significant differences between the treatment groups with regard to how many subjects received these follow-up messages. We also find that the contact rate was typically higher for women, therefore we do not expect that these gender asymmetries could be responsible for the results that we observe below. Moreover, we find no significant change in the number of logins or registrations in the three days following the robocalls, suggesting that they were ineffective in increasing mobilization. In addition, Tables 3 and 4 report response rates and comparison of means tests for the subsample of subjects who received follow-up messages. These results are qualitatively identical to the results observed in the full sample. Finally, all subjects received a mailed postcard reminder in the week of the seminar. The postcard text is included in Appendix 8. The seminar lasted for 3.5 hours. On the day of the seminar, we recorded subjects’ attendance and distributed an exit survey at the conclusion of the seminar for feedback and further measurement of the subjects’ interest in candidacy. The exit survey is included in Appendix 2. For a detailed analysis of the effect of the seminar on participants, see Preece and Stoddard (Reference Preece and Stoddard2014).
Response Rates of Subjects by Gender
Note: Standard errors are in parentheses.
Mean Website Logins, Registrations, and Attendance Rates by Treatment and Gender
Notes: Standard deviations are in square brackets. Standard errors are in parentheses.
*Marginal statistical significance relative to control at the 10% level, two-sided t-test.
TREATMENTS AND HYPOTHESES
Based on existing literature and discussions with party leaders, we designed 10 versions of the mailed invitation. The header of each invitation differed in the following way for each treatment (see Appendix 3 for the full invitations):
Control (C): “Have you considered running for office?”
Less Competitive (T1): “Your chance of being elected may be better than you think—nearly half of seats are uncontested in Utah elections.”
H1: T1 should boost response rates vis-à-vis the control.
Competitive (T2): “Do you like debating about politics and thrive in competitive environments?”
H2: T2 should decrease response rates vis-à-vis the control for women.
Qualified (T3): “You’re in the top 10% of our most active party supporters, making you highly qualified to run for office.”
H3: T3 should increase response rates vis-à-vis the control.
Challenging qualifications (T4): “Do you think you have the qualifications it takes to run for office?”
H4: T4 should decrease response rates vis-à-vis the control.
Service (T5): “Do you want to make a difference by improving the lives of your family and community members?”
H5: T5 should increase response rates vis-à-vis the control.
Resources and support (T6): “Learn about resources and support available to run for office.”
H6: T6 should increase response rates vis-à-vis the control.
Time commitment (T7): “Did you know that almost all elected offices in Utah are part-time positions?”
H7: T7 should increase response rates vis-à-vis the control.
Election rates (T8): “Did you know that when women run, they fundraise and win elections at the same rate as men?”
H8: T8 should increase women's response rates vis-à-vis the control.
Civic duty (T9): “‘One of the deepest manifestations of civic responsibility is running for office.’—Chair of the Utah County Republican Party.”
H9: T9 should increase response rates vis-à-vis the control.
Women's appeal (T10): “‘Nowhere outside the maternal circle does woman shine in her full glory as on the [political] platform.’—Martha Hughes Cannon, Utah Pioneer and America's first female state senator.” (Sent to women only.)
H10: T10 should increase response rates vis-à-vis the control.
RESULTS
Table 1 reports the distribution of our dependent variables by gender using a comparison of means. Clearly, across all treatments and dependent variables, men and women in our experiment exhibit large differences in their response to party recruitment. On average, women in our sample were half as likely as men to log onto the website, to register, and to attend the seminar, resulting in a significant gender gap in political ambition. Since gender is not randomly assigned, it is difficult to draw conclusions regarding causal effects. However, the results strongly suggest that women, on average, may be significantly less responsive than men to political party recruitment efforts.
Table 2 reports subjects’ response rate by gender and treatment with two-tailed t-tests. The results indicate that the three treatments that significantly increase women's political ambition in our experiment are the qualified treatment, the resources and support treatment, and the civic duty treatment. The qualified treatment more than doubles the share of women who log on to the website and attend the seminar (p = 0.128 and 0.083, two-tailed respectively). The resources and support treatment also result in a 25% increase in women's attendance relative to the control (0.005 vs. 0.000, p = 0.083, two-tailed) while having no statistically significant effect on the share of logins and registrations. And lastly, the civic duty treatment has a positive effect on women's registration rates, relative to the control (0.000 vs. 0.005, p = 0.083, two-tailed). For men, emphasizing the part-time nature of political office increased the number of website visits—but not registration or attendance. We saw no statistically significant effect of the other treatments on participant's political ambition.
It is likely that many of our null results are because of a lack of statistical power, and it is possible that some of our statistically significant results are a consequence of the multiple comparisons problem. Our power calculations assumed a 3% response rate based on our discussions with the party leadership. The actual response rates for logging on to the website averaged only 1% across treatments.Footnote 5 We therefore lack the ability to thoroughly evaluate our hypotheses about individual treatments and state more definitively which messages are the most effective in prompting increases in women's political ambition. Nevertheless, we believe that our findings will provide future researchers a helpful baseline when designing similar experiments as well as pointing gender scholars in fruitful directions for further research.
CONCLUSIONS
It is well documented that women are less likely than men to be encouraged to run for office by political elites (Crowder-Meyer Reference Crowder-Meyer2013; Lawless and Fox Reference Lawless and Fox2010; Niven Reference Niven1998). While there is some observational research to suggest that when women are recruited, they respond similarly to men (Fox and Lawless Reference Fox and Lawless2010), party leaders often report that women are difficult to recruit. Our experiment aimed at reconciling this seemingly contradictory evidence by considering whether certain messages are more or less effective at recruiting women and men to consider candidacy. Our results clearly show that when invited by their party to take steps toward running for office, politically active women are much less likely to do so than politically active men. But recruitment messages that reinforce a woman's qualifications emphasize available resources and support, and focus on office-holding as a civic duty may be more effective than other types of messages.
It is possible that our results are specific to this particular type of recruitment effort—although candidate training seminars are becoming increasingly common, especially for recruiting women (Hennings Reference Hennings2011; Rozell 2008; Sanbonmatsu et al. Reference Sanbonmatsu, Carroll and Walsh2009; Sanbonmatsu, forthcoming). The results may also be sensitive to the particular population we studied: Utah County, Utah is a conservative, Republican-dominated area. However, recruitment in the Republican Party is particularly interesting to study because, in contrast to the Democratic Party, growth in the proportion of female officeholders in the Republican Party has stagnated or reversed over the past two decades (Carroll and Sanbonmatsu Reference Carroll and Sanbonmatsu2013). This reality makes studying recruitment efforts in the Republican Party particularly interesting.
While our results do not provide a definitive answer into which recruitment messages increase women's political ambition the most, this study provides an important first step to understanding gender differences in responses to party recruitment. Our results also show the important role that experimental methods can play in studies of recruitment and how working with parties and political elites can be an effective way to study these questions (Grose Reference Grose2014). As researchers continue to work with parties to design similar experiments, we can learn more about the role of various recruitment methods and the differences between male and female candidacy decisions.
SUPPLEMENTARY MATERIAL
For supplementary material for this article, please visit Cambridge Journals Online. http://dx.doi.org/10.1017/xps.2014.17.
