Abstract
Recent evidence suggests that women overcome the potential negative impact of gender stereotypes by emerging when they are stronger candidates than men. I leverage an original survey of state legislative candidates to determine whether women devote more time to their campaigns. I find that women on the whole, and those who had previously been elected to a political office in particular, invested more of their personal time into the campaign than men. This difference is driven by the fact that women are more likely to forgo employment during the election. These findings suggest that women are more likely than men to arrange their personal obligations in such a fashion that they can run stronger campaigns.
A growing number of studies have recently sought to determine how a candidate’s gender affects his or her decision to enter politics, and how well-prepared he or she is for the campaign. Women face stereotypes and double standards in media coverage of their candidacies, but there is little evidence that, in the aggregate, women perform worse than men in terms of either fundraising or vote totals. One potential reason for the apparent gender parity in these areas is that women wait for races in which they stand a good chance of winning, and demonstrate higher average “quality” as candidates, which allows them to outpace any gender-related disadvantages they may face (Fulton, 2012, 2014). In this article, I engage the related question of whether women devote more of their personal time to their campaigns than men and whether they set aside other obligations to do so. I find that women do indeed devote more of their personal time to the campaign effort, largely because they work less at non-campaign jobs during the election.
Gender and campaign effort
Both prospective candidates (Lawless and Fox, 2010) and the public at large (Dowling and Miller, 2015) view women candidates as disadvantaged relative to men, and women are less likely to see themselves as well-qualified to run for office (Lawless and Fox, 2010). The widely held perception that women fare worse as candidates fuels a wide gulf between the two genders in terms of political ambition, which in turn results in fewer women entering politics (Fulton et al., 2006; Lawless and Fox, 2010). Yet despite conventional wisdom about the electoral prospects of women candidates, once incumbency is accounted for the women who do enter races raise as much or more money than men (Burrell, 2008), and non-incumbent women not only perform on par with men in terms of vote percentage (Dolan, 2004; Smith and Fox, 2001), but also report winning races for state and local offices at rates similar to men (Lawless and Fox, 2010). It is certainly possible that men and women perform equally well because the latter do not face gender-related disadvantages as candidates. However, there is a robust literature suggesting that women face double standards in media coverage (Carroll and Dittmar, 2010; Falk, 2008; Fowler and Lawless, 2009; Fox, 1997; Kahn, 1996). Voters apply stereotypes as well, viewing women as more liberal (Koch, 2000, 2002), more compassionate, and better-suited to the policy arena of “women’s issues” such as abortion or child care (see Burrell, 2008; Herrnson et al., 2003; Kahn, 1996; Koch, 1999; Plutzer and Zipp, 1996). Stereotypes about women’s policy competence do appear to affect how voters evaluate them (Dolan, 2010), and Koch (2002) found that voters’ judgment of women candidates as more liberal can make them appear to be farther from the voter’s ideal point than they actually are.
There is an emerging explanation for why, if women candidates do face gender-related disadvantages, they still perform comparably to men in the aggregate. For instance, despite the gender gap in ambition, women looking to move from the state legislature to Congress display more strategic behavior than men, entering races when political conditions are especially conducive to their candidacies (Fulton et al., 2006). Women might be so calculating out of a belief that they must be exceptionally well-prepared if they are to overcome the perceived gender-based disadvantage. The result is that among the candidates who actually begin a campaign, women enjoy an advantage in average candidate quality, which leads to aggregate gender parity with men on observable outcomes such as vote totals or money raised Fulton (2012, 2014). Put another way, women attempt to overcome the negative effects of stereotypes by being better candidates.
One heretofore unexamined aspect of “being better” is whether women devote more of their personal time to legislative campaigns. I engage a simple question: Do women candidates spend more time on their campaign than men? Since candidates’ personal effort pays dividends for challengers in terms of votes (Miller, 2016), it stands to reason that the strongest candidates will be those willing to commit more of their time to the campaign. This is particularly true in state legislative elections, which compared with the more professionalized, media-driven environment of a congressional campaign are more likely to require personal effort on the part of the candidate, such as field canvassing or phoning (Hogan, 1997). Candidates looking to wage a high-quality state legislative campaign should therefore be more likely to wait for a time when they can fully commit to the race. Thus, if women believe they must out-work their male counterparts, they should be more likely to relinquish non-campaign obligations such as outside employment, and should invest more of their personal time in the campaign relative to men. Given the observed strategic behavior of “quality” women candidates, this effect should be especially acute among women who have previously been elected to a public office, such as school board or city council.
Data and method
I employ original survey data from the major-party, lower-house candidate populations in 18 states during the 2008 general election. Initial contact was made in early October, with collection ceasing in early December. The included states are: Alaska, Arizona, Colorado, Connecticut, Delaware, Hawaii, Iowa, Maine, Michigan, Minnesota, Missouri, Montana, New Mexico, Ohio, Rhode Island, Vermont, West Virginia, and Wisconsin. These states are geographically diverse and range in ranking of Squire’s (2007) index of legislative professionalization from third (Wisconsin) to forty-third (Maine), accounting for a broad range of conditions in which state legislative campaigns are conducted. The survey was deployed in several waves via three media: mailed instruments, an electronic interface, and telephone. Of 2,971 candidates in the population, the survey yielded 1,022 responses overall, for a response rate of 34.4%, which is consistent with previous surveys of candidate populations (Francia and Herrnson, 2003; Howell, 1982). I retain survey data from candidates who faced major-party opposition and who responded to the relevant behavioral questions. There were 820 such candidates in the sample, of which 273 were incumbent members of their respective state houses. I supplement the survey data with information about each candidate’s vote totals and fundraising performance, obtained from the appropriate regulatory agencies of each state. The survey instrument and sample information can be found in Miller (2014).
Respondents were asked a series of questions about their demographics, attitudes, and political experience, but two measures are particularly important to the present analysis. First, respondents entered the number of hours that they personally devoted to a number of activities “during the first week of October”. Specifically, candidates quantified the number of hours they spent on fundraising, public speeches, field activity, electronic campaigning, media relations, research, strategy, phoning voters, sending mailings, and the courting of interest groups. 1 The bulk of my analysis compiles these activities into additive indices of either “total time”, reflecting the total number of weekly hours that candidates reported personally devoting to their campaign, or separates time into fundraising, field activities, and all other activities (see Miller, 2014, 2016).
I also asked candidates to categorize their occupational status (outside of legislative responsibilities) during the campaign as “not working”, “employed part-time”, or “employed full-time”. 2 Candidates are grouped into categories based on their answer to this question, via indicator variables for whether the candidate was in a given work category. Similarly, candidates are placed into three “experience” categories: Incumbents, candidates who have previously been elected to a political office, and those with no prior office-holding experience. In tandem, the measures of the candidates’ weekly hours and his/her responsibilities outside of the campaign yield insight into the extent to which the candidate was committed to the race on a daily basis, and as political experience has been often-utilized as a shortcut measure of candidate quality (Jacobson, 1989; Van Dunk, 1997), it seems prudent to account for experience in the analysis of campaign commitment.
To gauge whether women spent more time campaigning, I employ negative binomial models in which the (rounded) number of hours that a candidate reported devoting to the campaign serve as the dependent variable. 3 Likewise, I utilize logistic regression models of the binary outcome of whether a candidate was in a particular work category. The independent variable of interest in all models is an indicator coded 1 if the candidate was a woman. I show at least two specifications of each model: a binary regression of the dependent variable on the gender indicator, as well as a multivariate specification with theoretically relevant controls. All multivariate models also include state fixed effects. In the logistic regression models of whether a candidate was in a particular work category, I also include multiplicative interaction terms to test for the possibility that either experience or incumbency moderates the relationship between gender and propensity to work.
Throughout, I employ the political experience categories as a control variable. Where relevant for a given model, I include as controls indicators for the candidate’s political party, self-identified race, and whether the election was for an open seat or occurred in a multi-member district. I also include observed characteristics of the legislative district, such as the percentage of the two-party general election vote that the candidate’s party had received in the previous (2006) general election, as well as others obtained from Lilley et al. (2007), such as urbanity, (logged) population, and (logged) household income.
Results
I first consider the amount of time that candidates reported devoting to their campaign, by their level of experience. Figure 1 depicts the mean number of hours that candidates spent on fundraising, “field” activities such as canvassing or sign-posting, and all other campaign activities. 4 Within every experience category, women reported spending more weekly hours on their campaign than men. The difference is substantively small among inexperienced candidates, where women outworked men 48.6 hours to 47.0 hours. Incumbent women, however, devoted 47.2 hours to their campaign: exactly 5 hours more than incumbent men. An even larger gap is evident among experienced challengers. Non-incumbent women with political experience spent 47.9 hours on their campaign on average. This figure is higher than any other candidate category, and is over 7 hours more than experienced men devoted to their campaigns. Figure 1 is therefore consistent with the notion that women who entered a state legislative campaign did so strategically, investing more time in the effort than men with similar backgrounds.
Weekly hours devoted to campaign activities, by experience and gender.
Table 1 contains negative binomial regression coefficients for models which explore the relationship between gender and campaign time in greater detail. 5 The dependent variable in these models is the (rounded) number of weekly hours that the candidate reported spending on the campaign. Given the results depicted in Figure 1 above, Model 5 unsurprisingly returns a positive, statistically significant coefficient for the variable indicating the candidate is a woman, suggesting that among experienced candidates, women devote more of their personal time to the campaign. The same is true in Model 1, which employs data from all candidates.
Negative binomial regression coefficients and robust standard errors: number of weekly hours devoted to campaign.
Robust standard errors in parentheses. Multivariate models include state fixed effects.
p < 0.05. One-tailed tests on gender coefficients.
Dependent variable is number of hours candidates reported spending on the campaign.
However, both models capture only the bivariate relationship between gender and effort. When the appropriate controls are included in the even-numbered models, the coefficient for the female indicator becomes substantively smaller, and in no case achieves statistical significance. That said, the predicted number of hours spent on the campaign from the even-numbered models in Table 1 closely mirrors the pattern observed in Figure 1. When the other variables are held at their mean levels for each gender, inexperienced women are predicted to spend about 48 hours on the campaign: about 2 hours more than inexperienced men. The difference is much higher for experienced women, who outwork their male counterparts by a margin of 53 to 45.5, while the predicted 47 hours spent by incumbent women is about 5.5 more than male incumbents.
Taken together, the models in Table 1 suggest that women devote more time to their campaigns, possibly because they are less likely than their male peers to work during the campaign. Such a relationship would be consistent with the idea that women feel a greater need to eschew other obligations so they can focus on running a strong race. Figure 2 depicts the percentage of candidates, by gender and experience, who worked either full-time or part-time, or who did not work during the campaign, and demonstrates that regardless of their experience level, men were more likely to report working full-time than women. Indeed, a majority of non-incumbent men worked full-time during the election, while a majority of women in all three candidate groupings worked less than full-time. As with the analysis of campaign time, the largest gap between men and women appears among candidates with political experience: While more than 60% of experienced men worked full-time, only 27% of women with political experience did, and experienced women were almost twice as likely as experienced men to not work at all during the race.
Percentage of candidates not working, working part-time, and working full-time, by gender and experience.
It is worth noting that Figure 2 is also consistent with the notion that experienced women opt out of work by strategic calculus. If women were more likely than men to not work by virtue of unobserved factors such as family structure or wealth, the rates of non-work should be comparable for women across experience categories. However, women with political experience were almost twice as likely as inexperienced women to not work at all during the campaign. This stark difference implies that the propensity to focus on the election is not driven solely by gender; rather, it is consistent with the premise that “high-quality” women candidates choose to put other obligations aside during the campaign.
Table 2 contains logistic regression coefficients for models of two dependent variables: whether the candidate did not work at all, and whether the candidate worked less than full-time. All four additive models are suggestive that women are significantly more likely to either not work or to work less than full-time. Models 11 and 14 include interaction effects intended to test whether political experience moderates the apparent relationship between gender and work status. One of the two interaction terms is significant in each model, suggesting that political experience is indeed a moderating force.
Logistic regression coefficients and robust standard errors: determinants of whether candidates worked, or whether they worked less than full-time.
Robust standard errors in parentheses. Multivariate models include state fixed effects.
p < 0.05. One-tailed tests on gender coefficients and interaction terms and components.
Dependent variable is indicator for whether candidate did not work (1–3) or worked less than full-time (4–6).
To aid in interpretation, I now turn to Figure 3, which depicts the predicted probabilities of not working obtained from Model 11 (left pane), and the predicted probabilities of working less than full-time obtained from Model 14 (right pane) for candidates of both genders and all three experience levels. In addition to the point estimate of the predicted probability, Figure 3 also depicts 90% confidence intervals. The predicted probability of not working (left pane) is higher for women than men at all three experience levels. However, the difference is substantively meaningless for inexperienced candidates, and the confidence intervals for incumbent women and men display substantial overlap. The space between women and men is larger, about 18 points, for experienced candidates, indicating once again that the probability that an experienced woman chooses not to work is significantly higher than it is for experienced men.
Predicted probability of not working, or working less than full-time, by experience and gender.
A similar pattern is evident in the right pane, which depicts the predicted probability of working less than full-time. Again, women appear more probable than men to work less than full-time at each experience level. Moreover, while the gap is substantively small for incumbents, it is large for both experienced and inexperienced challengers. Indeed, the predicted probability of working less than full-time is roughly 21 and 35 points higher for inexperienced and experienced women, respectively, relative to men. Taken together, my analysis is therefore strongly suggestive that women with political experience devote more effort to their campaigns than men on a weekly basis because they are also more likely to forgo employment during the election.
Discussion/conclusion
Previous research has argued that women overcome gender-related disadvantages by being better candidates. I engage a related but open question: Do women candidates expend more effort on their campaign? I find that women candidates devote more time to their campaign because they are more likely to forgo working at a job during the election. In tandem with existing work (namely, Fulton 2012), these findings provide further evidence that women believe that in order to win, they must be better candidates than men. My results suggest that this belief is especially strong for women who have previously held elected office. While the survey data do not support definitive conclusions about why experienced women are particularly likely to invest more of their time in the campaign, my findings are consistent with the notion of a more strategic “quality” woman candidate who enters a race under favorable conditions and works hard to affect the desired result (see Fulton et al., 2006). Eschewing full-time work may just be another symptom of women being particularly good at “picking their spot”. That said, further research should more fully examine the strategic considerations of women entering political races, and experienced women in particular.
My analysis is not without limitations. One potential issue is the reliability and/or internal validity of self-reported campaign time, which serves as a dependent variable for a portion of my analysis. Although retrospective self-reporting of work time has been demonstrated to be fairly accurate (Jacobs, 1998), it is possible that candidates might have inflated the number of hours they reported devoting to their campaign, since doing so would make them appear to be more “hard-working”. Fortunately, such an inflation, assuming that it occurred in a relatively uniform fashion, would affect only the intercept in a regression equation, and would bias neither the coefficients nor the standard errors in the models reported above. Moreover, the mean level of self-reported weekly campaign time was about 47 hours, which seems plausible in a contested state legislative election. As such, there is no obvious reason to doubt the results of my analysis on the basis of self-reported time alone.
It is also possible that rather than eschewing work strategically in order to campaign harder, women are simply less likely to have work obligations to begin with. The fact that experienced women displayed such aberrant behavior relative to women who lacked political experience, however, is not consistent with this narrative. If time flexibility drives the entry decision, then it seems reasonable to expect both inexperienced and experienced women to exhibit higher (relative to men) rates of sub-full-time employment. However, only women with political experience, who should be most likely to make strategic decisions about the campaign, display higher rates of non-employment than men. So while it is possible that causation flows in the opposite direction, I believe it is unlikely given the data presented above. Future work could directly engage the strategic considerations that women make before they enter the race, perhaps by soliciting responses about income, wealth, or family structure.
Footnotes
Acknowledgements
I thank an invaluable team of research assistants for their data collection efforts: Rebecca Dittrich, Benjamin Gitlin, Thomas Hudson, Christopher Martin, Rob Morrissey, and Zachary Newkirk. Any mistakes, oversights, or omissions belong solely to me.
Funding
Data collection for this research was generously supported by the National Science Foundation (grant number SES-0819060). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.