Running Backwards and in High Heels

Abstract

The question of whether voter bias exists toward female politicians remains unsettled. Although anecdotal accounts of gender inequality abound, systematic research demonstrates that women “do as well as men” when they run. Previous work suggests that these conflicting observations result from an omitted variables problem. Specifically, if women are higher quality than men, and if quality is omitted from models of vote-share, then voter bias may be concealed. Using a unique measure of incumbents’ political quality, the author’s research documents a sex-based quality gap and importantly, is the first to link the quality gap to the gender parity in electoral success.

Keywords

women gender sex vote-share election Congress bias discrimination sexism quality

“Sure . . . [Fred Astaire] was great, but don’t forget that Ginger Rogers did everything he did . . . backwards and in high heels.”

Frank and Ernest, cartoon, 1982

The most recent presidential election sparked renewed hope for women’s advancement in electoral politics. But at the same time, it also exposed the persistence of barriers for aspiring female politicians in America. In contrast to her male competitors, Hillary Clinton often received disparaging comments about her appearance and manner of dress. Even her agency over her career accomplishments was cast into doubt—often being attributed to her husband—as Maureen Dowd (2008) of the New York Times opined, “It’s odd that the first woman with a shot at becoming president is so openly dependent on her husband to drag her over the finish line.” In reflecting on the primary process, House Speaker Nancy Pelosi remarked, “Of course there is sexism, we all know that . . . it is a given” (Sweet 2008). And CBS News anchor Katie Couric observed, “One of the great lessons of the campaign is the continued—and accepted—role of sexism in American life” (Seelye and Bosman 2008).

Over the past several decades, the scholarship on women in politics has made many enduring contributions to our stock of knowledge regarding the unique experiences female politicians confront on their path to political office. But, for the myriad ways in which women’s experiences diverge from men’s, the literature also portrays some striking similarities—most notably, in the success rates of male and female politicians. Contrary to the conventional wisdom, the quantitative research consistently downplays discrepancies in men’s and women’s ability to attract votes and win elections, even in the face of controls for party, seniority, and officeholder status (Black and Erickson 2003; Burrell 1994; Darcy and Schramm 1977; Darcy, Welch, and Clark 1994; Dolan 2004; Duerst-Lahti 1998; Ekstrand and Eckert 1981; Hedlund et al. 1979; Seltzer, Newman, and Leighton 1997). Because men and women perform equally well in models of vote-share, much of the literature concludes that voters are unprejudiced against women (but see as exceptions, Ambrosius and Welch 1984; Fox and Smith 1998).

Nevertheless, individual-level research on sex stereotypes paints a different portrait. This scholarship convincingly demonstrates that voters hold stereotypical views that often disadvantage female candidates for national office (Kahn 1994, 1996; Sapiro 1981), as well as shows that stereotypes shape citizens’ propensity to vote for them (Sanbonmatsu 2002a). Moreover, recent research suggests that voters prefer government to be male dominated (Dolan and Sanbonmatsu 2009). How can individual-level voters hold stereotypes and preferences that work against women candidates yet these biases fail to emerge in aggregate-level election returns?

Inferences that the electoral environment is gender-neutral because women “do as well as men” when they run rest on the assumption that men and women are qualitatively similar in terms of all of the characteristics that influence electoral outcomes except for one—gender. But, if men and women are distinct regarding characteristics that influence their electability—for instance, if women hold higher political quality than men but only perform at parity with men in the electoral arena—then this would be evidence of gender discrimination to the extent that women have to work harder than men to achieve similar electoral results. At the individual level, voters may stereotype female candidates and prefer males; however, these effects will be masked in the aggregate if women candidates hold superior quality and if quality is omitted in the vote-share model.

Other scholars have recognized that a major deficiency of earlier research is the absence of a control for political quality (Black and Erickson 2003; Ekstrand and Eckert 1981; Milyo and Schosberg 2000). However, building satisfying measures for the underlying construct for quality are difficult. As a result, the empirical implications of a sex gap in quality have not been brought to bear on election outcomes. My analysis is unique from previous work because I engage in a direct test of the hypothesis that a sex gap in quality masks bias against female politicians. Using an innovative informant-based measure of incumbent political quality, coupled with election results from the real world, my article bridges the divide between inference and evidence.

My results provide reason for concern about women’s political equality. I show that when the incumbent’s political quality is omitted, men and women enjoy equivalent support at the ballot box. However, once the heterogeneity in political quality is taken into account, the sex of the incumbent emerges as a significant negative predictor of vote-share—even after controlling for a number of alternative explanations. Representing a straightforward case of omitted variables bias, the effect of sex on election outcomes is attenuated in the absence of a political quality measure. Once the sex-based quality discrepancy is controlled, a net vote disadvantage of approximately 3 percent emerges for women incumbents.

This gendered quality gap challenges more sanguine interpretations about the health of our democratic system because it provides an alternative basis for the gender parity in electoral success so often cited in the women in politics literature. If women are higher quality than men, then inferences that the electoral environment is gender-neutral because women “do as well as men” when they run are fundamentally problematic, arousing significant questions about democratic fair play. Moreover, if bias exists, then expanding the “supply” of female candidates may not entirely address women’s underrepresentation problem. Rather, this implies that much of the impetus for change will depend on the “demand” for women as politicians.

Sex and Electoral Success: Empirical and Theoretical Discrepancies

Anecdotal evidence suggesting gender inequality pervades American politics. In the 112th Congress, women occupy just 17 percent of U.S. Senate and U.S. House seats. At the state and local levels, only six states are headed by a female governor, and female mayors lead only seven out of the hundred largest U.S. cities (Center for American Women in Politics 2010c). Given such glaring inequalities, it may be tempting to conclude that sex discrimination remains plentiful in contemporary American politics.

However, much of the scholarship on women in politics dispels these suspicions. In one of the first studies to cast doubt on the claim of sexism in elections, Darcy and Schramm (1977, 10) report that “the electorate is indifferent to the sex of congressional candidates . . . voters are not keeping women from serving in the House.” And although they do not rule out the possibility that sex discrimination existed in a bygone era, Darcy, Welch, and Clark (1994, 73) similarly conclude that “the political system and cultural milieu no longer present . . . barriers to women. . . . If more women run, more women would be elected.” Reinforcing their argument, the authors show that their results persist despite controls for incumbency, seniority, and partisanship.

More recently, the results of the 2010 election seem to validate the interpretation that women are treated equitably in American elections. Women candidates declared victory in 51.4 percent of the House races and 50.0 percent of the Senate races in which they competed (Center for American Women in Politics 2010a).

Nevertheless, there is a growing consensus that voters hold preferences for male officeholders and rely on gender stereotypes to infer candidate traits, issue competencies, and ideologies. As Sanbonmatsu and Dolan (2009, 486) put it, “women’s vote-getting ability does not necessarily mean that voters react to men and women candidates in the same way.” In their analysis of the 2006 American National Election Studies (ANES), Dolan and Sanbonmatsu (2009) report that the average voter prefers that men occupy 60 percent of political offices. whereas only 10 percent of respondents preferred that government be composed of a majority of women. Although election results demonstrate no discrepancy in men’s and women’s electability, these aggregate-level statistics mask individual-level preferences for male politicians and male-dominated government.

This preference for male leaders may be traced back to gender stereotypes that individuals hold. The women in politics scholarship marshals strong evidence that voters hold and evaluate candidates on the basis of sex stereotypes to infer candidate traits, issue competencies, and ideologies. In terms of traits, individuals expect women to be more compassionate, moral, and honest (Alexander and Andersen 1993; Kahn 1992, 1994, 1996; McDermott 1998). Voters see female candidates as being more competent on “compassion” issues, like education, helping the poor, health care, the environment, and women’s issues (Alexander and Andersen 1993; Herrnson, Lay, and Stokes 2003; Huddy and Terkildsen 1993b; Kahn 1992, 1994, 1996; Sanbonmatsu 2002a; Sapiro 1981). And in terms of ideology, women are expected to be more liberal than men (Huddy and Terkildsen 1993b; Koch 2000; McDermott 1998).

Most importantly, these sex stereotypes about issue competence, personality traits, and ideology have been linked to voting preferences (Dolan 2010; Herrnson, Lay, and Stokes 2003; Koch 2000; McDermott 1998; Sanbonmatsu 2002a). All of this portrays a significant theoretical and empirical paradox: at the individual level, voters stereotype based on sex and hold preferences for male candidates and officeholders; at the same time, aggregate-level election results depict gender parity and an absence of bias. How can women “do as well as men” when they run—even in terms of national-level office, like Congress—if voters hold stereotypes that work to women’s disadvantage and act upon them with respect to voting preference?

Resolving the Paradox: The Gender Gap in Quality

Underlying the inference that women receive equitable treatment in elections because they “do as well as men” when they run is the premise that men and women are qualitatively similar in terms of all of the characteristics relevant to electoral performance except for one—gender. But, what if this premise were wrong? What if women were more likely than men to possess qualities that voters valued?

Recognizing the flaws inherent in this premise, one strand of the literature investigates selection mechanisms that may manufacture gender gaps in electorally relevant characteristics—in particular, quality. For instance, gender differences in self-selection (Fox and Lawless 2005; Fulton et al. 2006), recruitment from political parities (Fox and Lawless 2005; Niven 1998; Sanbonmatsu 2006), and the dual structure of American primary and general elections (Lawless and Pearson 2008; Palmer and Simon 2006) are selection mechanisms that may contribute to qualitative differences between male and female candidates who successfully emerge.

In their study of men and women in pipeline professions, Fox and Lawless (2005) report that because women view themselves as less qualified than men, and because they are less likely than men to be recruited by party leaders, many objectively qualified women are either voluntarily or involuntarily censored from the pool of potential candidates. From this, they reason that those women who survive the process of elimination ought to be more qualified than their male counterparts.

Although this research illuminates how the earlier stages of the political process might give rise to a quality differential, some scholarship goes further to draw an explicit link between sex-based discrepancies in quality and gender equity in electoral success—concluding that women are subjects of bias to the extent that they have to work harder than men in developing political quality just to achieve similar electoral results.

For instance, Lawless and Pearson (2008) examine the amount of primary challenger opposition solicited by male and female politicians. They reveal that women are more likely than men to be challenged in both their same- and opposite-party primaries and intuit that the women who survive their primaries emerge as higher quality candidates as a result of the more rigorous process. In conclusion, they attach the inferred quality discrepancy to the apparent absence of voter bias: “To make it through the primary process, women must be stronger candidates, or at least candidates who are willing to endure greater challenges, and more challengers, than their male counterparts face. Women, in other words, have to be ‘better’ than men in order to fare equally well” (Lawless and Pearson 2008, 78).

These are arresting claims, but ones that would be more convincing if they withstood a direct test of the hypothesis that voter discrimination is masked by the gender gap in quality. Nevertheless, this hypothesis has eluded such a direct test, largely due to problems operationalizing political quality. In the absence of research that directly engages these empirical problems, the normative concerns raised by the prospect of gender inequality remain unsettled. To wit, this article presents the first attempt to assess whether men and women are treated equitably in the political arena after controlling for the confounding effects of the sex-based heterogeneity in quality. And as a consequence, it elevates our understanding of a problem of substantial normative concern.

Conceptualizing Political Quality

Scholars comparing the electoral performance of men and women acknowledge that a major deficiency of this line of inquiry is the absence of a control for political quality (Black and Erickson 2003; Ekstrand and Eckert 1981; Milyo and Schosberg 2000). However, the near intractability of measuring quality stymies the development of an entirely satisfying measure that moves beyond the standard variable for previous office-holding experience.

The challenges inherent in quantifying quality are not particular to the gender politics literature—scholars of congressional elections more generally have wrestled extensively with this problem (Bond, Covington, and Fleisher 1985; Jacobson 1978; Krasno and Green 1988). The simplest and most common solution is to draw on previous office-holding experience as a proxy for quality. And while this approach holds significant intuitive appeal—in order to win, officeholders have, presumably, had to run a campaign, appeal to voters, and face opponents—it does suffer from some shortcomings.

Most importantly, office holding may overlap with the concept of quality to some degree, but it is neither necessary nor sufficient for quality to be manifest. For instance, winning elective office may be partly attributable to the candidate’s quality, but it is also due to “other aspects of the environment that have nothing to do with quality, such as favorable shifts in district partisan composition, serious missteps by opposing candidates, or national political tides” (Stone, Maisel, and Maestas 2004, 480). Treating quality as previous office holding neglects a variety of factors that both contribute to winning and are uncorrelated with attributes of the individual candidate.

Likewise, this conception overlooks the existence of high-quality candidates who have never held elective office before. Individuals may lose their bid for prior elective office for a variety of reasons that have little to do with their quality—for instance, partisan tides may turn on a high-quality candidate and cause her to lose the election. As a result, using previous office holding to operationalize quality may overestimate the quality of individuals who have held office and underestimate the quality of those who have not (Squire 1992).

Beyond these concerns, experimental research suggests that quality and experience constitute distinct dimensions in the minds of voters and that variations in quality alter the individual-level vote intention and the aggregate vote-share of incumbents (Kulisheck and Mondak 1996; Mondak 1995; Stone et al. 2010). Indeed, equating quality with office holding means that incumbents cannot be distinguished in terms of their quality. But this interpretation is not realistic. Officeholders of all stripes vary with respect to their dedication, legislative ability, integrity, and so forth. Office holding as a proxy of quality neglects these important sources of variation.

While no indicator can possibly quantify all of the relevant characteristics comprising quality, one approach to the problem is to draw on the opinion of informed district observers who are steeped in local-level politics and who are acquainted with the personal and political characteristics of local politicians. Because this measure for quality is derived from the opinion of district observers who are familiar with the personal and political strengths of the incumbent, it reflects information that goes beyond conventional measures relying on office-holding experience.

Moreover, the measure has been shown to relate strongly to legislative behavior and is highly predictive of vote-share (Stone et al. 2010). And because real-world election results can be merged into the data set along with the measure for quality, the data present the first opportunity to assess the electoral effects of gender after controlling for this confounding variable.

Data and Method

Toward this end, I enlist the ratings of 2,672 political activists and potential candidates dispersed across a national random sample of 200 congressional districts in 1998 (Stone and Maisel 2003; Stone, Maisel, and Maestas 2004).¹ The activist sample was generated by randomly selecting an equal number of Democratic and Republican national convention delegates and county chairs in each congressional district (n = 1,160). The potential challenger survey consisted of state legislators from the congressional district, as well as potential challengers who were named by the activists as individuals who would make strong House candidates (n = 1,512). The activists and potential challengers are both treated as “informants” who are knowledgeable about local politics—particularly, characteristics of the congressional district as well as the incumbent.²

As an election year, 1998 was a relatively placid one for gender politics. Although previous research shows that women benefit when issues play into their stereotypic strengths and suffer when the agenda is dominated by men’s issues (Dolan 2004; Herrnson, Lay, and Stokes 2003; Kahn 1994, 1996; Lawless 2004), the election of 1998 gave neither sex a distinct advantage. Being six years removed from the much heralded “Year of the Woman” and three years prior to the onset of the “War on Terrorism,” the political agenda was neither favorable nor hostile toward women. Much of the focus in Washington was on the Clinton impeachment debacle; but despite this controversy, the president’s party picked up five seats in the House and broke even in the Senate (Barone and Ujifusa 2000). With respect to gender politics, the status quo prevailed: women broke even in the Senate (at 9 percent) and barely increased their margins in the House from 11 percent to 12 percent (Center for American Women in Politics 2010d).

One year in advance of the election, informants were asked to evaluate the political quality of the incumbent in their congressional district in terms of ten items: personal integrity, dedication to public service, grasp of the issues, ability to find solutions to problems, ability to work with political leaders, public speaking ability, ability to stay in touch with the district, ability to provide constituency service, ability to bring federal funds to the district, and legislative accomplishments. In their assessment, informants used a 7-point scale ranging from extremely strong (+3) to extremely weak (−3). By tapping the incumbent’s personal commitment, character, and style, along with his or her professional accomplishments, this measure conceives quality as an essential attribute of individuals, that is, distinct from their previous office-holding status.

In addition, this conception of quality highlights attributes that voters might value in their own right, as opposed to more strategic campaign skills (Krasno and Green 1988; Stone and Simas 2010). In contrast to campaign skills, like who has the most name recognition or money, this view holds that voters respond to virtues like competency, integrity, and dedication and value politicians who champion district causes. To be sure, campaign skills promote a candidacy, but the ability to generate name recognition or money is likely a consequence of more intrinsic components of personal quality, like who is a capable and faithful public servant. Operationalizing quality in this way emphasizes those attributes that conceptually ought to alter voters’ preferences.

My analysis recognizes that the informants’ estimates of the incumbent’s quality may be subject to partisan and gender bias.³ Respondents who share the same partisanship and/or sex of the incumbent may systematically rate those incumbents more favorably due to their shared characteristics. I constructed variables for “shared partisanship” (different = −1, same = 1) and “shared gender” (different = −1, same = 1) and ran regressions using both of these variables as predictors of each of the ten quality items. The regression coefficients for shared partisanship and shared gender yield the extent of the bias in the measure—either positive or negative. I subtracted this “bias factor” from each of the ten individual quality items to purge the measure of the effects of partisanship and gender. I construct the overall political quality index by taking the average of these purged measures and aggregating them up to the district level.^4,5

I control for other factors regarded in the congressional elections literature as influential in determining incumbent vote-share, such as: whether the race is contested,⁶ as well as contested by an experienced challenger,⁷ the partisan composition of the congressional district,⁸ the seniority of the incumbent,⁹ and the partisanship of the incumbent.¹⁰ To gauge the value added of the informant-generated quality measure over and above the conventional variable for prior office-holding experience, I use the Almanac of American Politics to code whether or not the incumbent held previous state legislative office before entering Congress (Barone and Ujifusa 1998).¹¹

Notably, this design is oriented toward incumbents only, which may arouse concerns about the generalizability of the results. However, by focusing on incumbents only, the analysis is more likely to underestimate the impact of sex discrimination than it is to overstate it. Because information about incumbents is abundant, voters are less likely to rely on categorical cues like gender to shape their decisions. As a result, voter bias ought to be more difficult to find in the incumbent’s election performance.

Contrast this situation with challengers, who face a deficit of information. Because voters have less information about challengers, they are more likely to rely on “simple” voting cues—like gender—to inform their assessments (McDermott 1998). For instance, Kahn (1992, 507) finds that voters more frequently rely on sex stereotypes—and these stereotypes are stronger—when evaluating challengers rather than incumbents: “[because] people receive less information about challengers . . . people may therefore turn to sex stereotypes to fill in the information gaps.” Similarly, Sapiro (1981, 73) notes that “It is clear that gender provides a cue for candidate evaluation, especially under conditions of low information.” While Huddy and Terkildsen (1993b, 134) find that “in the absence of specific information about a candidate’s political beliefs, gender appears to be the primary cue used by participants to infer a candidate’s political outlook.”

Simply put, if voters are less likely to support the incumbent on the basis of her sex, then this effect ought to be even greater for challengers who are more likely to be the subjects of sex-based heuristics. The challenger’s vote ought to be even more susceptible to gender bias than the incumbent’s.

Results and Analysis

Table 1 explores the relationship between the variables of interest and sex. Consistent with previous research, men and women do not significantly differ in the vote-share they receive. In addition, women appear no more likely to encounter an experienced challenger, they are indistinguishable in terms of seniority, they have similar previous office-holding experience, they are equally likely to find themselves in a key race, and they raise equivalent amounts of money as men. However, Table 1 demonstrates that women are significantly more likely to face a general election challenger, they represent “safer” partisan districts, they are more Democratic, and they are judged to be higher quality than men.

Table 1.

Comparison of Means: Independent Variables by Sex

	Men		Women
	M	SD	M	SD
Incumbent party vote-share	71.495	14.635	67.659	12.751
Contested race	0.750	0.434	0.857*	0.359
Experienced challenger	0.208	0.407	0.238	0.436
District partisanship favors incumbent	49.162	10.051	53.738**	11.712
Seniority of incumbent	5.006	3.957	3.762	3.032
Incumbent is democrat	0.429	0.496	0.762***	0.436
Incumbent prior office-holding experience	0.488	0.501	0.429	0.507
Incumbent quality	0.998	0.483	1.256***	0.497
Congressional Quarterly key race	0.071	0.258	0.048	0.218
Incumbent spending (log)	13.125	0.658	13.158	0.654
Incumbent previous vote-share	63.914	13.102	60.777	18.488

Note: Analysis excludes open seats. Due to directional hypotheses, all significance tests are one-tailed.

p ≤ .10. **p ≤ .05. ***p ≤ .01.

To evaluate the relationship between these variables and the incumbent’s vote, I estimate five ordinary least squares (OLS) regression models in Table 2. The baseline model 1 incorporates traditional predictors of vote-share and shows that favorable district partisanship significantly enhances the incumbent’s electoral support, while being challenged, especially by someone who is experienced, significantly erodes the incumbent’s vote-share. The coefficient for female incumbent is negative but is not statistically different from zero, which is consistent with the previous research documenting gender parity in vote-share. In the absence of a control for political quality, women do as well as men.

Table 2.

Ordinary Least Squares Regression Model: Incumbent Party Vote-Share

	Model 1		Model 2		Model 3		Model 4		Model 5
	Regression coefficient	SE	Regression coefficient	SE	Regression coefficient	SE	Regression coefficient	SE	Regression coefficient	SE
Female incumbent	−2.288	1.946	−1.992	1.933	−3.775**	1.827	−3.199**	1.652	−2.779**	1.651
Contested race	−24.346***	1.361	−24.227***	1.350	−23.219***	1.285	−21.610***	1.220	−21.250***	1.223
Experienced challenger	−6.209***	1.322	−6.561***	1.320	−5.873***	1.228	−3.578***	1.164	−3.388***	1.158
District partisanship favors incumbent	0.370***	0.072	0.360***	0.071	0.459***	0.068	0.387***	0.065	0.349***	0.067
Seniority of incumbent	0.096	0.138	0.069	0.137	−0.103	0.131	−0.119	0.118	−0.182*	0.121
Incumbent is Democrat	−0.768	1.181	−0.914	1.172	−1.116	1.089	−0.936	0.983	−0.823	0.976
Incumbent prior office-holding experience			2.196**	1.046	2.161**	0.979	1.326*	0.893	1.364*	0.886
Incumbent quality					5.183***	1.100	4.745***	0.995	4.435***	0.999
Congressional Quarterly key race							−7.873***	1.817	−7.764***	1.802
Incumbent spending (log)							−3.051***	0.897	−2.576***	0.921
Incumbent previous vote-share									0.083**	0.041
Constant	72.746***	3.640	72.380***	3.611	62.355***	3.811	105.620***	12.852	96.174***	13.583
Number of observations	187		187		186		186		186
F test	92.39***		81.32***		78.53***		81.38***		75.63***
Adjusted R²	0.747		0.751		0.770		0.813		0.816

Note: Analysis excludes open seats. Due to directional hypotheses, all significance tests are one-tailed.

p ≤ .10. **p ≤ .05. ***p ≤ .01.

Although Table 1 showed no average difference in the prior office-holding experience of men and women, the mean only describes the average, and not the spread of the variable—it is still possible that this variable’s distribution differs for men and women and that the heterogeneity in prior office-holding accounts for the gender parity in vote-share depicted in model 1. To test this proposition, model 2 includes all of the variables contained in model 1 but also controls for the incumbent’s previous office-holding experience as an indicator of quality. As expected, office-holding experience exerts a positive and significant impact on vote-share. Incumbents who have held prior office garner about 2 percent more electoral support than those who lack such experience. The inclusion of the prior office-holding term does not alter the size or significance of any of the variables in the model. And most importantly, controlling for this variable has no perceptible impact on the female coefficient. All and all, model 2 suggests that if there is an electoral bias against women that is masked by differences in men’s and women’s quality, prior office-holding experience does not tap the underlying source of the heterogeneity.

To test whether the informant-generated quality indicator alters the substantive meaning of the results, model 3 incorporates this variable. The magnitude and significance of the variables for contested race, favorable district partisanship, seniority, and partisanship continue to be stable. Prior office-holding experience continues to exert a positive and significant effect on vote-share, even after the informant-based measure for quality is included, suggesting that these two variables reflect independent sources of electoral advantage. For instance, prior office-holding experience may capture more strategic campaign skills and political experience, while the informant-based quality measure taps more intrinsic components of the incumbent’s personality, character, and performance.

Importantly, the variable for incumbent quality is statistically and substantively significant—the difference between the highest and the lowest quality incumbent is approximately 12 percentage points. And once this more nuanced measure of quality is taken into account, the gender of the incumbent emerges as a significant negative predictor of the vote.

This finding suggests that the gender parity in election outcomes cited in previous research is rooted in an omitted variables bias. Male and female incumbents garner equal electoral support in models 1 and 2 when the sex-based heterogeneity in quality is neglected, but once the difference in men’s and women’s performance, service, integrity, and dedication are controlled for, the disadvantage for women becomes apparent. To perform on par with men, women incumbents would need to be approximately one standard deviation greater on the quality scale than their male counterparts. This suggests that if women withdrew from making additional investments in political quality, over and above that of men’s, then they would encounter significant electoral sanctions—in fact, they appear to escape the electoral effects of gender discrimination precisely because they exceed their male counterparts in this respect.

However, alternative explanations should also be considered. Might these results be the artifact of gender differences in unobserved factors? For instance, not all female incumbents enjoy the same amount of monetary resources, press coverage, name recognition, or, more generally, campaign effectiveness. If any of these factors influences the vote, are correlated with gender, and are omitted, then the effect of sex in the vote-share model will be biased. There are a variety of characteristics that shape the incumbent’s likelihood of victory, some of which may be traced back to heterogeneity in the incumbent’s campaign effectiveness, media treatment, and political resources; but others may be linked to external conditions in the district—like political scandal or an unpopular local policy or project.

To account for the incumbent’s likelihood of victory, I incorporate variables for whether the race was ranked as a “Key Race” by Congressional Quarterly (CQ), as well as include a measure for the log of the incumbent’s spending.¹² The CQ measure taps how close or competitive the race is expected to be given factors such as local and national political tides, incumbent scandal, and district preferences. The incumbent spending measure also reflects race competitiveness. In the research on congressional elections, there is ample evidence that “the more the incumbent spends, the worse s/he does” (Jacobson 1978, 1980, 2004). Of course, incumbent spending does not cause them to become vulnerable; rather, incumbents spend when they face a credible challenger and feel uncertain about their prospects of victory. In this sense, the log of the incumbent’s spending also ought to relate to the incumbent’s chances of winning.

Model 4 shows that the incumbent spending and CQ key race variables are both negative and significant. Holding all of the other variables at their mean or mode, incumbents who are locked in a competitive contest are predicted to win 58.2 percent of the vote, while their counterparts in more lopsided districts are expected to capture 66.1 percent of the vote. Incumbents appear to spend reactively—the most lavish spenders only win an estimated 61.7 percent of the vote, while more disciplined spenders (presumably, safe incumbents) are estimated to garner 73.6 percent of the vote-share.

But even after the variables capturing the incumbent’s vulnerability are included in model 4, the sex of the incumbent remains appropriately signed, similarly sized, and substantively and statistically significant. In addition, most of the other variables in the model maintain similar-sized coefficients and significances, suggesting that the CQ and incumbent spending measures reflect information pertinent to the race at hand and uncorrelated with most of the other variables in the model. The two coefficients that are moderated by the inclusion of the CQ and spending terms—district partisan favorability and experienced challenger—are causally prior to the CQ ranking and incumbent spending. Even so, district partisanship and the presence of an experienced challenger still exert significant independent effects.

Model 5 considers the potential for vote-share to be autoregressive: past performance may predict current performance. If there are intrinsic characteristics of the candidate or campaign that make them more successful than others, it may be important to take the candidate’s baseline performance into consideration. Moreover, there may be temporal effects of the prior campaign that persist to influence the current campaign—for instance, the campaign’s prior conduct may leave voters more or less favorably disposed toward the candidate. To the extent that the CQ measure and incumbent spending variables are imperfect indicators of the incumbent’s prospects of victory, then there may be other sources of unobserved heterogeneity that explain the negative and significant coefficient for a female incumbent.

To account for these alternative hypotheses, model 5 includes a lagged dependent variable—the incumbent’s previous vote-share.¹³ Including the lagged dependent variable constitutes a strict test of the hypothesis that the gender of the candidate independently influences elections, as these specifications are known to bias coefficient estimates downward (Achen 2000), making the estimates much more conservative.

As expected, the lagged term is positive and statistically significant. But even with the inclusion of the lagged term, the coefficients for the presence and experience level of the challenger, district partisanship, prior office-holding experience of the incumbent, incumbent quality, key race, and spending all remain similarly sized and highly statistically significant.

And most importantly, the coefficient for female continues to be stable and significant—after controlling for the gender gap in quality, women receive an approximate 3 percent vote disadvantage. Remarkably, this result persists even after taking a variety of alternative explanations into account.^14,15 Although this effect may seem rather small, it can certainly make the difference between the success or failure of a candidate in a hotly contested race that is evenly matched. Indeed, in 1998 alone, there were ten races that were decided by a 3 percent margin or less (Congressional Quarterly 2010).

Interestingly, women candidates are more likely than men to lose by a 3 percent margin or less. Although women were 27.6 percent of congressional candidates overall (Center for American Women in Politics 2010b), they were 50 percent of the candidates who lost by less than 3 percent. The five women who lost by this margin were all challengers, but if women challengers confront just as much, if not more, bias than incumbents—that is, they lack incumbency advantages, they are lesser known, and so voters may be more likely to rely on gender cues and stereotypes—then these five women might have won if they had been men.

Conclusion and Implications

Women may do as well as men when they compete for electoral support; however, the mechanism that produces this result is not because men and women are treated equitably by the electorate. After controlling for a variety of alternative explanations, my results consistently show that men and women systematically vary with respect to their political quality, and the gender parity in electoral success is directly attributable to this gap in quality. Although the scholarship on stereotypes uncovers strong evidence of the use of sex-based cues at the individual level, empirical evidence of bias at the aggregate level has been suppressed by the omission of an intervening variable in the causal model linking gender to election outcomes. Identifying this omitted variable, producing theoretical explanations for its significance, and directly testing this hypothesis on election data has been the principle contribution of this article. Relative to men, women have to work harder at developing greater political quality to be equally competitive.

While these results may merit revisions to our understanding of gender discrimination against women in politics, it would be premature to suggest that this is the final word. For one, the theoretical and empirical underpinnings of the gendered quality gap remain unclear. Although previous research suggests that the quality disparity stems from women enduring a more competitive electoral process, the research falls short of drawing a direct link between the causal process and the gendered discrepancy in quality (Fox and Lawless 2005; Lawless and Pearson 2008).

In an attempt to formulate a stronger link between the causal mechanism and the quality gap, in other work I explore how perceptions of quality and bias interact with gender to alter run decisions (Fulton 2011). Using a national random sample of potential candidates, I find that women who deem the political environment as unequal are more responsive than their male counterparts (and women who perceive no sex bias) to their self-assessed quality. To counteract the effects of discrimination, women who anticipate sexism withhold their candidacies until they attain a quality threshold. Thus, the expectation of bias “weeds out” low-quality female potential candidates—women who anticipate sexism develop more skills and resources before they decide to run. To the extent that the perception of bias is pervasive, women’s heightened attention to quality will alter the characteristics supplied by male and female candidates in the aggregate.

Moreover, future research should be aimed at determining whether gender bias exists in alternative contexts. For instance, a variety of scholars have demonstrated that when voters have scarce information, they are more reliant on categorical cues than in high information contexts (Huddy and Terkildsen 1993a; Kahn 1992, 1994, 1996; McDermott 1998; Sapiro 1981). To the extent that voters have less information about challengers than incumbents, they may be even more dependent on gender stereotypes to evaluate challengers. There is some preliminary evidence to support the expectation that female challengers are more highly qualified than male challengers (Fulton et al. 2006), but whether this quality gap alters challengers’ electoral support is an empirical question that remains yet answered.

Beyond the distinction between challengers and incumbents, there are persistent questions about how the level of office shapes the treatment of female candidates. The scholarship on sex stereotypes suggests that voters prefer candidates to exhibit stereotypically “masculine” traits in national-level offices and stereotypically “feminine” traits in less prestigious and powerful offices (Huddy and Terkildsen 1993a; Kahn 1994, 1996). If this is the case, what does this mean for women seeking high-level political positions—like senator or president? Are voters less supportive of women in higher office?

Naturally, questions remain about whether the quality gap exerts similar effects on vote-share across partisanship. Previous research documents a variety of nuanced ways in which partisanship and gender interact (King and Matland 2003; Koch 2000, 2002; Ondercin and Welch 2009; Sanbonmatsu 2002a, 2002b, 2004). Although my data show a consistent quality gap between male and female Democrats and Republicans alike—Democratic women = 1.200, Democratic men = 0.973, p < .05; Republican women = 1.426, Republican men = 1.017, p < .05—due to the small number of female Republicans in Congress, it is not possible to parse out whether the gap in quality exerts similar effects on electoral performance across partisanship. As their presence in Congress increases, future research should take advantage of the larger number of female Republicans.

From the perspective of normative democratic principles, these findings may arouse some concern. However, there is at least one more optimistic implication of this analysis: if women are more likely than men to possess integrity, dedication, problem-solving skills, collaborative leadership, faithful representation, and legislative ability—all of the factors that comprise the measure of political quality—then citizens represented by female members of Congress ought to receive more competent representation. Over the long term, if citizens come to recognize and reward trustworthy, effective, and faithful legislators, then women’s descriptive representation ought to improve. Examining these dynamics may prompt a reappraisal of the value of female legislators and alter the demand for female politicians in the future.

Footnotes

Acknowledgements

The author is grateful to Walter Stone, L. Sandy Maisel, and Cherie Maestas for generously sharing their data. This article has benefited from the comments of Judith Baer, Jon Bond, Sally Friedman, Kim Hill, Paul Kellstedt, Michael Koch, Walter Stone, Michelle Taylor-Robinson, and the American Politics Program participants at Texas A&M University. The author thanks the editors and the anonymous reviewers for their careful reading of the manuscript and for offering thoughtful comments about it.

An earlier version of this article was presented to the American Political Science Association’s Annual Meeting in Chicago, Illinois. Supplemental materials are available on the web at .

The authors declared that they had no conflicts of interests with respect to their authorship or the publication of this article.

The author(s) disclosed receipt of the following financial support for the research and/or authorship of this article:The data collection was supported by grants from the Carnegie Corporation of New York and from the National Science Foundation (SBR-9515450).

Notes

References

Achen

Christopher H.

2000. “Why Lagged Dependent Variables Can Suppress the Explanatory Power of Other Independent Variables.” Prepared for the Annual Meeting of the Political Methodology Section of the American Political Science Association. http://polmeth.wustl.edu/media/Paper/achen00.pdf .

Alexander

Deborah

Andersen

Kristi

1993. “Gender as a Factor in the Attribution of Leadership Traits.” Political Research Quarterly 46:527-45.

Ambrosius

Margery M.

Susan

Welch

. 1984. “Women and Politics at the Grassroots: Women Candidates for State Office in Three States, 1950-1978.” Social Science Journal 21 (1):29-42.

Barone

Michael

Ujifusa

Grant

, eds. 1998. The Almanac of American Politics. Washington, DC: National Journal Press.

Barone

Michael

Ujifusa

Grant

, eds. 2000. The Almanac of American Politics. Washington, DC: National Journal Press.

Black

Jerome H.

Erickson

Lynda

. 2003. “Women Candidates and Voter Bias: Do Women Politicians Need to Be Better?” Electoral Studies 22 (1):81-100.

Bond

Jon R.

Covington

Cary

Fleisher

Richard

. 1985. “Explaining Challenger Quality in Congressional Elections.” Journal of Politics 47:510-29.

Burrell

Barbara

. 1994. A Woman’s Place is in the House: Campaigning for Congress in the Feminist Era. Ann Arbor: University of Michigan Press.

Center for American Women in Politics. 2010a. “Election 2010: Scorecard for Women Candidates.” http://www.cawp.rutgers.edu/fast_facts/elections/candidates_2010.php .

10.

Center for American Women in Politics. 2010b. “List of Women Candidates—Election 1998: Congressional and Statewide Elective Races—Winners and Losers.” http://www.cawp.rutgers.edu/fast_facts/elections/documents/Can1998Results-Hi.pdf .

11.

Center for American Women in Politics. 2010c. “Women in Elective Office 2010.” http://www.cawp.rutgers.edu/fast_facts/levels_of_office/documents/elective.pdf .

12.

Center for American Women in Politics. 2010d. “Women in the U.S. Congress 2010.” http://www.cawp.rutgers.edu/fast_facts/levels_of_office/documents/cong.pdf .

13.

Congressional Quarterly. 2010. “Voting and Elections Collection.” http://library.cqpress.com/elections/ .

14.

Darcy

Schramm

Sarah Slavin

1977. “When Women Run Against Men.” Public Opinion Quarterly 41 (1): (1977):1-12.

15.

Darcy

Welch

Susan

Clark

Janet

. 1994. Women, Elections and Representation. Lincoln: University of Nebraska Press.

16.

Dolan

Kathleen A.

2004. Voting for Women: How the Public Evaluates Women Candidates. Boulder, CO: Westview Press.

17.

Dolan

Kathleen A.

2010. “The Impact of Gender Stereotyped Evaluations on Support for Women Candidates.” Political Behavior 32 (1):69-88.

18.

Dolan

Kathleen A.

Sanbonmatsu

Kira

. 1994. “Gender Stereotypes and Attitudes Toward Gender Balance in Government.” American Politics Research 37 1994: 409-28.

19.

Dowd

Maureen

. 2008. “Two Against One.” New York Times, January 23. http://www.nytimes.com/2008/01/23/opinion/23dowd.html .

20.

Duerst-Lahti

Georgia

. 1998. “The Bottleneck: Women Becoming Candidates.” In Women and Elective Office, 15-25. New York: Oxford University Press.

21.

Ekstrand

Laurie E.

Eckert

William A.

1981. “The Impact of Candidate’s Sex on Voter Choice.” Western Political Quarterly 34 (1):78-87.

22.

Fox

Richard L.

Lawless

Jennifer L.

2005. It Takes a Candidate: Why Women Don’t Run for Office. Cambridge, MA: Cambridge University Press.

23.

Fox

Richard L.

Smith

Eric R. A. N.

1998. “The Role of Candidate Sex in Voter Decision-Making.” Political Psychology 19:405-19.

24.

Fulton

Sarah A.

2010. “What Underlies the Gendered Quality Gap? The Role of Perceptions in Shaping the Supply of Female Candidates.” Paper presented at the American Political Science Association’s annual meeting, Washington, DC.

25.

Fulton

Sarah A.

Maestas

Cherie D.

Maisel

L. Sandy

Stone

Walter J.

2006. “The Sense of a Woman: Gender, Ambition and the Decision to Run for Congress.” Political Research Quarterly 59:235-48.

26.

Hedlund

Ronald D.

Freedman

Patricia K.

Hamm

Keith E.

Stein

Robert M.

1979. “The Electability of Women Candidates: The Effects of Sex Role Stereotypes.” Journal of Politics 41:513-24.

27.

Herrnson

Paul S.

Lay

J. Celeste

Stokes

Atiya Kai

. 2003. “Women Running ‘as Women’: Candidate Gender, Campaign Issues, and Voter-Targeting Strategies.” Journal of Politics 65:244-55.

28.

Huddy

Leonie

Terkildsen

Nayda

. 1993a. “The Consequences of Gender Stereotypes of Women Candidates at Different Levels and Types of Office.” Political Research Quarterly 46:503-26.

29.

Huddy

Leonie

Terkildsen

Nayda

. 1993b. “Gender Stereotypes and the Perception of Male and Female Candidates.” American Journal of Political Science 37:119-47.

30.

Jacobson

Gary C.

1978. “The Effects of Campaign Spending in Congressional Elections.” American Political Science Review 72:469-91.

31.

Jacobson

Gary C.

1980. Money in Congressional Elections. New Haven, CT: Yale University Press.

32.

Jacobson

Gary C.

2004. The Politics of Congressional Elections. 6th ed. New York: Pearson Longman.

33.

Kahn

Kim Fridkin

. 1992. “Does Being Male Help? An Investigation of the Effects of Candidate Gender and Campaign Coverage on Evaluations of U.S. Senate Candidates.” Journal of Politics 54:497-517.

34.

Kahn

Kim Fridkin

. 1994. “Does Gender Make a Difference? An Experimental Examination of Sex Stereotypes and Press Patterns in Statewide Campaigns.” American Journal of Political Science 38:162-95.

35.

Kahn

Kim Fridkin

. 1996. The Political Consequence of Being a Woman: How Stereotypes Influence the Conduct and Consequences of Political Campaigns. New York: Columbia University Press.

36.

King

David C.

Matland

Richard E.

2003. “Sex and the Grand Old Party: An Experimental Investigation of the Effect of Candidate Sex on Support for a Republican Candidate.” American Politics Research 31:596-612.

37.

Koch

Jeffrey W.

2000. “Do Citizens Apply Gender Stereotypes to Infer Candidates’ Ideological Orientations?” Journal of Politics 62:414-29.

38.

Koch

Jeffrey W.

2002. “Gender Stereotypes and Citizens’ Impressions of House Candidates’ Ideological Orientations.” American Journal of Political Science 46:453-62.

39.

Krasno

Jonathan S.

Green

Donald P.

1988. “Preempting Quality Challengers in House Elections.” Journal of Politics 50:920-36.

40.

Kulisheck

Michael R.

Mondak

Jeffery J.

1996. “Candidate Quality and the Congressional Vote: A Causal Connection?” Electoral Studies 15:237-53.

41.

Lawless

Jennifer L.

2004. “Women, War and Winning Elections: Gender Stereotyping in the Post-September 11th Era.” Political Research Quarterly 57:479-90.

42.

Lawless

Jennifer L.

Pearson

Kathryn

. 2008. “The Primary Reason for Women’s Underrepresentation? Reevaluating the Conventional Wisdom.” Journal of Politics 70:67-82.

43.

McDermott

Monika L.

1998. “Race and Gender Cues in Low-Information Elections.” Political Research Quarterly 51:895-918.

44.

Milyo

Jeffrey

Schosberg

Samantha

. 2000. “Gender Bias and Selection Bias in House Elections.” Public Choice 105 (1-2):41-59.

45.

Mondak

Jeffrey J.

1995. “Competence, Integrity, and the Electoral Success of Congressional Incumbents.” Journal of Politics 57:1043-069.

46.

Niven

David

. 1998. The Missing Majority: The Recruitment of Women as State Legislative Candidates. Westport, CT: Praeger.

47.

Ondercin

Heather L.

Welch

Susan

. 2009. “Comparing Predictors of Women’s Congressional Election Success: Candidates, Primaries and the General Election.” American Politics Research 37:593-613.

48.

Palmer

Barbara

Simon

Dennis

. 2006. Breaking the Political Glass Ceiling. New York: Routledge.

49.

Sanbonmatsu

Kira

. 2002a. “Gender Stereotypes and Vote Choice.” American Journal of Political Science 46:20-34.

50.

Sanbonmatsu

Kira

. 2002b. “Political Parties and the Recruitment of Women to State Legislatures.” Journal of Politics 64:791-809.

51.

Sanbonmatsu

Kira

. 2004. Democrats, Republicans and the Politics of Women’s Place. Ann Arbor: University of Michigan Press.

52.

Sanbonmatsu

Kira

. 2006. Where Women Run: Gender and Party in the American States. Ann Arbor: University of Michigan Press.

53.

Sanbonmatsu

Kira

Dolan

Kathleen

. 2009. “Do Gender Stereotypes Transcend Party?” Political Research Quarterly 62: 485-94.

54.

Sapiro

Virginia

. 1981. “If U.S. Senator Baker Were A Woman: An Experimental Study of Candidate Images.” Political Psychology 3 (1/2):61-83.

55.

Seelye

Katharine Q.

Bosman

Julie

. 2008. “Media Charged with Sexism in Clinton Coverage.” New York Times, June 13. http://www.nytimes.com/2008/06/13/us/politics/13women.html .

56.

Seltzer

Richard

Newman

Jody

Leighton

Melissa Voorhees

. 1997. Sex as a Political Variable: Women as Candidates and Voters in U.S. Elections. Boulder, CO: Lynn Reinner.

57.

Squire

Peverill

. 1992. “Challenger Quality and Voting Behavior in Senate Elections.” Legislative Studies Quarterly 17: 247-63.

58.

Stone

Walter J.

Fulton

Sarah A.

Maestas

Cherie D.

Maisel

L. Sandy

. 2010. “Incumbency Reconsidered: Prospects, Strategic Entry and Incumbent Quality in U.S. House Elections.” Journal of Politics 72:178-90.

59.

Stone

Walter J.

Maisel

L. Sandy

. 2003. “The Not-So-Simple Calculus of Winning: Potential U.S. House Candidates Nominations and General Election Prospects.” Journal of Politics 65:955-71.

60.

Stone

Walter J.

Maisel

L. Sandy

Maestas

Cherie D.

2004. “Quality Counts: Extending the Strategic Politician Model of Incumbent Deterrence.” American Journal of Political Science 48:479-95.

61.

Stone

Walter J.

Simas

Elizabeth N.

2010. “Candidate Valence and Ideological Positions in U.S. House Elections.” American Journal of Political Science 54:371-88.

62.

Sweet

Lynn

. 2008. “House Speaker Nancy Pelosi says—‘I’m a Victim of Sexism Myself All the Time.’” Chicago Sun-Times, June 24. http://blogs.suntimes.com/sweet/2008/06/house_speaker_nancy_pelosi_say.html .