Abstract
How candidates are selected, such as through nomination by party elites or election by primary voters, potentially influences the underrepresentation of women in political office. Partisan differences suggest that primary voters in right-leaning parties might select fewer female nominees than would their left-leaning counterparts even if both parties’ elites are equally likely to select female nominees. This hypothesis is confirmed by an analysis of lieutenant governors in the United States, a position that varies in whether candidates are appointed by party elites or elected by primary voters. In cases where lieutenant-governor candidates are appointed, Democratic and Republican gubernatorial nominees are equally likely to choose female running mates; where primary voters select the lieutenant governor, Republicans are less likely to nominate women.
Keywords
A number of studies find that voters do not discriminate much against female candidates (Burrell 1992; Dolan and Lynch 2014; Fulton 2014). Why, then, are women so underrepresented in positions of political power? Scholars who find voters to be open to female candidates often argue that party gatekeepers’ bias keep women out of office by systemically denying them equal ballot access and support (e.g., Lühiste 2015; McElroy and Marsh 2010). This implies that positions more controlled by party elites should see fewer female officeholders. Yet evidence suggests the opposite; women obtain a greater share of appointed rather than elected mayoralties and judgeships in the United States (Alexander 2011; Bratton and Spill 2002; Holmes and Emrey 2006). 1 Parties even use elections strategically to avoid complying with gender-balance rules, exploiting primaries’ tendency to underproduce female winners (Baldez 2007).
The pattern of relatively woman-friendly electorates would seem to contradict the finding that electorates in practice advance fewer women into positions of power. How can we resolve this paradox? One much-studied possibility is that women are culturally or psychologically more averse to political campaigns, 2 thus competing less often for elected office even if they are equally qualified or likely to win (Flammang 1997; Fowler and McClure 1989; Fox and Lawless 2004). The gendered effect of primaries is then not because of which candidate the voters select, but because the set of candidates available in primaries tends to be more dominated by men.
Another, less studied, possibility is that parties are not all alike; party elites and primary electorates may pointedly differ from general-election voters, varying in their openness to female nominees. Even with the same institutions, then, parties may produce differing shares of female candidates, and the disconnect between gender-neutral electorates and gender-imbalanced offices could arise because aggregate surveys conceal divergent patterns among different parties (Carroll and Sanbonmatsu 2013). In particular, left- and right-leaning parties—for example, Democrats and Republicans—might differ in their treatments of women, for several reasons. They have different cultures, which affect their respective abilities to recruit women as electoral candidates, and electorates respond differently to female candidates from each party (Carroll and Sanbonmatsu 2013; Elder 2012; Hayes 2011). This may make it harder for Republicans to get women through primary elections, even if the electorate as a whole would vote for female candidates and even if the party’s elites avidly seek gender balance. 3 Those elites have much more direct control over appointed positions, by contrast, and so are more likely to be able to achieve parity there.
If this is the case, any difference between the gender composition of elected and appointed positions would be stronger among Republicans. We thus contribute to the literature exploring distinctions between appointment and election by focusing on partisan implications. Moreover, our empirical test analyzes the lieutenant governorship, an office further up the ballot and thus typically more visible than the mayoral or judicial positions used in previous research on gender differences between appointment and election. Crucially, lieutenant governors’ manner of selection varies: in some states, parties choose their candidates for lieutenant governor via primary election; in others, gubernatorial candidates appoint their own running mates to the ballot. Anecdotally, the two parties appear to take different approaches to nominating women to this office, with Republicans more reluctant to admit that sex played any role in candidate selection. For example, among currently sitting governors who chose their running mates, (Democrat) Andrew Cuomo made it known that he explicitly sought a female lieutenant-governor nominee (Orden 2014), whereas (Republican) Mike Pence and (Republican) Terry Branstad adamantly insisted that they chose the best possible candidate, who simply happened to be female (Henderson 2010; LoBianco 2012). The relationship among partisanship, gender, and lieutenant governors especially matters because of lieutenant governorships’ asymmetrical role in getting women into high-powered offices. Almost half (7 out of 15) of female Republican governors to date ascended, unelected, from a lieutenant governorship; the proportion for Democrats is merely 2 out of 22. 4
The evidence presented below supports the hypothesis of partisan differences in the effect of candidate-selection mechanisms. Both major parties put forward female lieutenant-governor candidates more frequently when choosing nominees by appointment, but the appointment-election difference is especially large among Republicans. In other words, although there is little difference between Republican and Democratic gubernatorial nominees in the proportion of women they choose as running mates, a significantly smaller proportion of Republican than Democratic primary elections for lieutenant governor produce a female nominee.
Theory and Previous Literature
Research on the differential effects of appointment and election are part of a larger literature on how candidate-selection mechanisms interact with gender (Bullock and Maggiotto 2003; Fiber-Ostrow 2012). Previous studies have seen a variety of relationships between primary elections and women’s success in obtaining nominations. Although, as noted above, appointment often appears to be the fastest route for women to get access to political power, some authors have found other relationships. Martinez (1989) argued that party institutions, in organizing primaries, produced more gender-balanced slates than did gubernatorial nominees selecting their own running mates. Goelzhauser (2011) found that elected offices diversify faster than appointed in states with liberal populations, but not in states with conservative populations.
These findings, especially the effect of electorate conservatism, hint at why different political parties may obtain divergent outcomes when using primaries to select their candidates. Regarding gender, Republican women have a harder time winning primary elections (King and Matland 2003), fitting with a broader tendency for reactions to candidate gender to be filtered through partisan attitudes (Krupnikov and Bauer 2014). But the results, by themselves, do not explain why primary elections and appointments should produce different outcomes: ultimately, appointers should be responsive to the electorate. If those selectors are strategic and seeking instrumentally to win elections, both primary-election voters and party elites will naturally attempt to choose candidates who will appeal in the general election. One might therefore expect that the means of selection would not greatly affect political outcomes: observed differences in outcome between primaries and appointments present a social-science puzzle.
Fortunately, it is a puzzle for which the literature provides several answers, many of which hint at partisan differences that could influence how successful women are at capturing the nomination. One mechanism would question the assumption that nominators are strategic: past studies argue strongly that not everyone is. In particular, evidence for strategic behavior by primary voters is mixed at best (Abramowitz, McGlennon, and Rapoport 1981; Kube and Puppe 2009). Even concerted campaigns to engineer strategic voting in primary elections have seen little, if any, effect (e.g., Stephenson 2011). Instead, many primary voters seem to make their choices based on expressive rather than instrumental motivations (Hamlin and Jennings 2011), and as would be expected if primary voters were not wholly worried about backward induction, success in primaries appears to reflect different factors from success in general elections—especially for female candidates (Ondercin and Welch 2009). Elites, in contrast, may be likelier to strategically consider the general electorate’s attitude toward female candidates and to anticipate less obvious effects of selecting a female candidate, such as its effects on available general-election campaign options (Windett 2014). These ideas fit with the emphasis, ubiquitous running-mate-selection studies, on the calculated nature of the choice (Sigelman and Wahlbeck 1997; Uscinski 2012). 5 Candidates who select their running mates appear to pay intense attention to how the general electorate will react to nominee characteristics, suggesting they will be attentive to the implications of gender.
Suppose elite nominators converge on some strategic outcome, but primary electorates instead vote more sincerely—or, perhaps, are less informed and resort to heuristics when evaluating female candidates (Ditonto, Hamilton, and Redlawsk 2014). What consequences would we expect for nominations? As those identifying as Republicans tend to be systematically less supportive of female candidates and politicians (Dolan 2010), the different parties’ primary electorates would be unlikely to converge in their nominations. Instead, regardless of any tendency for both parties to be equally willing to nominate women when elites select the candidate, Republican primary voters will tend to favor female candidates less often than will their Democratic counterparts.
Even if elites and primary voters are both strategic, they may pick winners differently. In particular, primary voters may misapprehend the general electorate’s preferences in ways that better informed party elites do not. This need not be because primary voters are more extreme (Hirano et al. 2010), though extreme primary voters could certainly produce or reinforce the outcome hypothesized here. Instead, variation in electorate demographics may be key: older voters, especially men, are less open to voting for women (Flannelly 2002; Mo 2015), and are also the sort of habitual voters who turn out disproportionately for primaries. As the mass public tends to assume its own positions are generally moderate and popular—and so likely to be winners in the election—primary voters may diverge more from what the electorate wants than party elites do (Delavande and Manski 2012). Under this mechanism, the difference between primary voters and elites may not be the same across parties. If one party were to especially draw its support and primary-election votes from older voters, that older party would be less apt to nominate female candidates if and only if the primary electorate chose the nominee. With the Republicans’ enduringly older base of support (Crittenden 1962; Fisher 2008), Republican primaries could act as a particular hurdle to female candidates. 6 The demographic divergence between parties’ electorates would lead Republican primary voters to nominate relatively few female candidates even if the two parties’ elites are similar in their nomination patterns.
Alongside these possibilities where (some) nominators are thinking strategically, partisan differences could arise without recourse to anyone making conscious calculations but instead simply expressing their true, and differentially gendered, preferences. Although most accounts of political elites do suggest at least some strategic behavior, they—including gubernatorial nominees in particular—also make some choices that appear nonstrategic (Brown 2013; Rasch 2014). Nonstrategic behavior may be particularly likely in questions of gender, as the voting public’s explicit demand for gender balance in nominations is often weak (Bruhn 2003). Any configuration of outcomes could then reflect the direct preferences of the nominators involved. This points to the possibility of partisan differences in attitudes toward female candidates across primary-election voters. Partisan differences in elite attitudes are somewhat more difficult to assess because of more limited survey data and higher likelihood of social desirability bias in answers, but if, for example, elites are socialized into norms of leadership and equality regardless of party—or come from distinct strata of society with particular attitudes about gender and candidates—the electorate-level differences might not appear among elites. Then, too, the method of candidate selection would be pivotal for partisan differences in gender of nominees.
Whether because of different levels of strategic thinking among different potential nominators, because attempts to strategize lead primary electorates and party elites to divergent conclusions, or because of simple variation in preference among the possible nominators, then, Republicans and Democrats could be expected to see dissimilar effects of having primaries select their general-election candidates. All the mechanisms point to one central hypothesis: Republican primary voters will select fewer female nominees than would their Democratic counterparts even if both parties’ elites are equally likely to select female nominees. Although the above theory does not specifically predict the relative propensities of Democratic primaries, Democratic elites, and Republican elites to put forward female candidates, the arguments do suggest that Republican primaries should produce particularly few female nominees.
The above logic is independent of many traditional factors spurring partisan differences in propensity to put forward female candidates. General electorates’ different reactions to Democratic and Republican female candidates (Hayes 2005; Koch 2000; Sanbonmatsu and Dolan 2009) imply that the payoff to nominating women differs by party, as descriptive-representation senses of engagement mostly occur with copartisans (Reingold and Harrell 2010), and the partisan gender gap (Burden 2008) means there are more Democratic women voters to entice than Republicans. Furthermore, when gender does motivate voters to deviate from their party identifications, Democratic female candidates generally see larger effects than do Republicans (Brians 2005; Plutzer and Zipp 1996). Insofar as these factors produce partisan differences in the strategic value of nominating women—mostly, by making it more electorally profitable for Democrats to nominate women—they apply equally to primary- and elite-selected candidates. Similarly, it is possible that being the very first female nominee for an office and hence a political novelty may hurt a candidate’s chances of election (O’Regan and Stambough 2011). Even if Republican female candidates are rarer and hence more novel than are Democratic female candidates, this novelty penalty on its face is constant across the various institutional mechanisms for nominating candidates.
Candidates for Lieutenant Governor
Addressing the hypothesis empirically requires evidence from an office that varies in whether it is elected or appointed. We focus on the office of lieutenant governor, candidates for which are selected by primary election only in some states. Lieutenant governorships have the additional benefit of having in recent decades been relatively accessible to female candidates—in Fox and Oxley’s (2003) influential classification on the stereotypical gender of state executive offices, the lieutenant governor is classified as “neutral” rather than, as most offices are, “masculine.” In addition, previous studies suggest that running-mate sex influences vote choices in gubernatorial elections and that the preference for female lieutenant governors varies across electoral contexts (Fox and Oxley 2005).
Despite lieutenant governors’ generally low profile in political science, they have long been acknowledged to have real institutional power (Isom 1938; Kettleborough 1917). Although these formal powers—beyond the not insignificant role of being a heartbeat away from the governorship—vary from state to state, they have generally increased over time (Declercq and Kaminski 1979; Winder and Hill 2006). 7 This is in common with some other statewide executive offices, which the lieutenant governorship often bears strong relationships with and indeed is sometimes indistinguishable from; in some states, the official having replacement-governor responsibilities is even given a title other than “Lieutenant Governor,” such as “Secretary of State.” 8 This exemplifies the continuum of powers that lieutenant governors hold: at the extreme end, they assume roles and titles that in other states would be housed in other offices.
As important as official powers are the lieutenant governorship’s informal role and relatively public position. Seconds-in-command are frequently in a privileged position to observe and learn from the political leadership of the government (Prémont 2008). With a statewide elective office near the top of the ballot, moreover, lieutenant governorships—and campaigns for lieutenant governorships—provide visibility that makes them effective staging posts in the careers of those with progressive political ambition (Schlesinger 1966). The campaigns even can have extensive coverage of issues and policies, especially where the lieutenant governor runs in a separate election from the governor (Cook, Hartwig, and Wilcox 1992).
To test our hypothesis, we assembled a dataset including all Democratic and Republican general-election candidates for lieutenant governor from 1986 to 2014. 9 For the states (Arizona, Oregon, and Wyoming) where Secretaries of State assume the governorship on its vacancy, those secretaries—effectively lieutenant governors with a different title—are included as well. We exclude states where governors’ replacements are chosen not through a general election but by the state legislature (Maine, New Hampshire, Tennessee, West Virginia), as the theories above concern how the public reacts to or is anticipated to react to the selected candidate. For similar reasons, New Jersey does not enter the dataset until its 2009 election.
Institutions vary in their treatment of lieutenant-governor candidates in ways beyond who selects the nominees. In particular, the lieutenant governorship is in some states a separate line on the ballot, elected in its own right. In other states, governor and lieutenant governor run as a unified ticket, with votes for one candidate going also to the linked candidate in the other office. This variation relates to candidate-selection mechanisms: every state where candidates select running mates also joins the two on the general-election ballot. Where primary electorates select lieutenant-governor candidates, by contrast, the general-election race is only sometimes joint. This could potentially affect outcomes for our central hypothesis if primary voters, as they select their preferred candidates, take into account whether the lieutenant-governor candidates they nominate will eventually share a ballot line with the gubernatorial nominee. Relatively little evidence from previous literature suggests that this or other similar processes contribute substantially to lieutenant-governor nominations, so we pool all the primary-selected candidates in our main reported results below, though we return to the issue in a later section. 10
The key dependent variable is the sex of the lieutenant-governor candidates; for the regression models, we code female candidates as 1 and male candidates as 0. Different states have historically had varying propensities to nominate females to this office; Figure 1 shows this, with darker shading indicating a larger proportion of nominees being female.
Proportion of lieutenant-governor nominees female, 1986–2014, by state.
The key independent variables are party affiliation—a binary variable indicating whether the ticket is Republican—and the lieutenant-governor candidate-nomination procedure. In most states, primary-election voters select the lieutenant-governor nominee, but in 16 states, the gubernatorial nominees select their running mates. This provides two categories of state: those with elected and those with appointed nominees, shown in Figure 2. 11 This schema abstracts away from some complexities. In some states, for example, parties can overrule primary votes with nominating conventions where party activists select nominees, as when no candidate receives a sufficiently large share of the primary-election vote. We include such cases as elected, as these mostly are one-off episodes in states (such as Virginia) that otherwise have primary voters choose the candidate.
Method of lieutenant-governor candidate selection as of 2012.
As both a final summary of the data and an initial plausibility test of the hypothesis, Figure 3 shows what proportion of lieutenant-governor nominees were female for each combination of party affiliation and selection method. The figure also notes, with the vertical bar, the 95% confidence interval around the proportion, as well as the number of observations with the relevant combination of party and selection method.
Sex of lieutenant-governor candidates (1986–2014), by party and selection process.
As expected, Figure 3 shows that gubernatorial appointment produces women nominees more often than does primary election. Also as expected, females are somewhat more common among Democratic nominees than among Republican nominees. This difference, though, varies by the selection mechanism. Among appointed nominees, the two parties have nearly the same proportion of female candidates, with only a slight increase among Democrats. When primary elections select candidates, by contrast, Democrats are a third more likely to choose a woman (a difference statistically significant with 95% confidence). Both parties thus seem relatively open to female nominees when party elites, in the form of the gubernatorial nominee, select candidates: at least in this context, only moderate bias against women appears. It is only when primary elections come in that Republicans lag in their nomination of female candidates.
While Figure 3 is indicative, it does not control for other factors that may drive parties’ nomination patterns. For example, culture or other state institutions could mean that some states systematically are more receptive to female candidates (Windett 2011); if these features happened to correlate with the mechanism by which lieutenant-governor candidates were selected, the figure’s relation could be spurious. The next section attempts to rule this out by considering more sophisticated models including control variables.
Results
Our formal tests for differences are logistic regressions predicting whether a party’s nominee for lieutenant governor for a given state-year is female. The models control for several variables. One of these is whether the party’s nominee for governor is female, as parties may support female candidates less where women are already installed in positions of political authority (Bratton and Spill 2002). Another important variable is the state’s partisan landscape, as a long though debated literature suggests that women may be more likely to get on the ballot when there is little hope of victory (Carroll 1985; Gertzog and Simard 1981). Importantly for our hypothesis, some have argued that this pattern may be especially common among Republicans (Cooperman and Oppenheimer 2001; Sanbonmatsu 2002; Stambough and O’Regan 2007). To account for this, we use the proportion of the vote the party slate received in the general gubernatorial election. 12 Having incumbents running as nominees for governor or lieutenant governor may also affect women’s access to the ballot, as historically most incumbents have been both male and able to secure renomination easily; when incumbents are running, then, women may be less likely to be even seriously considered for nomination. Conversely, when running against an incumbent, parties may try to combat their opponent’s superior name recognition by capturing novelty effects, as by putting forward a female candidate.
Other features also influence states’ likelihoods of having female candidates. Voters often take candidate gender to signal ideology or other traits to which conservatives and liberals may react differently (e.g., Dolan 2010; Fulton and Ondercin 2013). Populations that are (say) more liberal might in this case be more acceptant of female candidates; if parties or candidates anticipate this, it may influence the pool of nominees. These ideologies are measured using the Enns and Koch (2013) index of state mass-public attitudes, giving the proportion of population that is liberal and the proportion that is conservative (with the moderate proportion being the omitted category). 13 Paralleling this demand-side force for additional female candidates is a differing availability of supply. Because many candidates for higher office emerge from the state legislature (Maestas et al. 2006), a more heavily female state legislature may increase women’s chance of nomination for statewide office. To account for this, we control for the proportion of the state legislature that is female in the relevant state-year. This variable also reflects local populations’ willingness to accept female elected officials, which could additionally influence how likely primary voters or strategic elites are to put forward female nominees.
Finally, to account for differences across states in political culture or other characteristics, the models include state fixed effects. These should account for features that lead states to diverge in their candidate-selection mechanisms, whether for historical (e.g., the era in which their constitutions and hence institutions were devised) or strategic reasons. Similarly, fixed effects for time control for the generally increasing propensity for women to obtain political nominations over the decades. In reported results, time is measured as two-year electoral cycles, though yearly fixed effects produce similar results for the coefficients of interest.
Some of these control variables restrict the sample in various ways. Variables relating to the gubernatorial candidacy rule out observations from Oregon, where elections for governor and for Secretary of State (Oregon’s second-in-line to the governorship) occur in alternate elections. Special elections to fill a vacant lieutenant governorship (e.g., as happened in Arkansas in 1996 or Louisiana in 2010) similarly fall out of presented models. Meanwhile, in its very short history of having a lieutenant governor, New Jersey has only had female candidates nominated; as a result, state fixed effects eliminate those observations from the model.
Column A in Table 1 reports the baseline model results. As would be expected from the previous literature, the estimated effect of having an appointed lieutenant governorship is positive and statistically significant—all else equal, nominees selected by the gubernatorial candidate are more likely to be female—whereas Republican tickets, also statistically significant, are associated with a negative coefficient. As to the interactive variable of interest, its predicted effect is statistically significant and almost exactly equal to (indeed, very slightly larger than), and opposite in sign to, the effect of being a Republican. This echoes Figure 3’s suggestion of a substantial gap between Republicans’ and Democrats’ propensity to nominate female candidates for lieutenant governor—but only when it is primary voters who do the nominating.
Logistic Models Predicting Femaleness of Major-Party Lieutenant-Governor Candidates, 1986–2012.
Note. Numbers in parentheses give robust standard errors.
Indicates coefficients that are statistically distinguishable from zero (95% confidence, one-tailed).
As to the statistically significant control variables, female lieutenant-governor nominees are more likely with incumbent governors, but less likely with female gubernatorial candidates and high-vote-share candidates. More conservative and (in some specifications) more liberal states are less apt to nominate women. The effect of liberalism stems at least in part from multicollinearity among the regressions’ independent variables: liberal states tend to have more women in the state legislature and as gubernatorial nominees. When these variables are removed from the models, the effect of liberal populations attenuates. As is typical, the interaction variables also show signs of multicollinearity and increased standard errors; this most notably contributes to the uncertainty in the estimate for appointed running mates. That estimate is, as Figure 3 implied, substantively quite large—approximately twice the size of the effect of being a Republican ticket or of the interaction term itself —yet fails to attain standard levels of significance. The multicollinearity is compounded for that variable because the state fixed effects generally coincide with candidate-selection mechanisms.
Because of this, and because fixed effects obscure the mechanisms of state differences, columns B and C of Table 1 replace the state fixed effects with measures of several additional state-invariant features. 14 Two variables capture Elazar’s (1984) political-culture index: one measuring moralism, the other individualism. In each case, the state is scored at three points if the relevant culture is the state’s sole culture. If a state has two local cultures, the more important gets a score of 2 and the less important culture gets a score of 1. Cultures not associated with a state at all take a value of zero. Traditionalism is the implied baseline category.
Two features more specific to the lieutenant governorship deserve special mention. One is the title. As mentioned above, some states have no one with the title of “Lieutenant Governor,” but instead have the Secretary of State assume the governorship on the death, resignation, or removal from office of the governor. 15 Nominators could treat the office differently because of its different label, perhaps because of more comfort bestowing the title “Secretary” to women (Formanowicz et al. 2013; Melcher et al. 1992) given the longstanding association of secretarial work with women’s employment (Beaton, Tougas, and Laplante 2007). Secretary-of-State nominees are always chosen through primary elections instead of gubernatorial nominee appointment (and always run separately, not on a unified ticket with the governor), so any such association between gender and title could confound the results. A binary variable indicating whether the office’s title is “Secretary of State” rather than “Lieutenant Governor” aims to control for this.
The other notable feature of the lieutenant governorship is its varying responsibilities in different states. Women generally obtain more nominations to low-power positions (Diamond 1977; Mandel 1981; Rosenthal 2005), which might imply less access to the ballot in states with relatively powerful lieutenant governors. Winder and Hill (2006) measure two separate facets of lieutenant governors’ institutional power: one (often constitutionally assigned to the office) relating to the legislature, the other (over which the governor has more discretion) relating to the governor and executive. The Winder–Hill measures only exist for offices titled “Lieutenant Governor”; models including those variables thus exclude the states with lieutenant-governor-equivalent Secretaries of State and, consequently, the variable controlling for that title. 16
The relationship among partisanship, candidate-selection mechanism, and gender looks somewhat different with these measures included in the models, as columns B and C of Table 1 show. The central hypothesis maintains its effect, with Republicans substantially more likely to have a female lieutenant-gubernatorial candidate when that candidate is appointed rather than elected, all else equal. However, gubernatorial appointment on its own—that is, as the model is specified here, for Democrats—does not always associate with more female nominees. Columns B and C show much smaller differences in predictions between elected and appointed Democratic nominees, and fall short of standard statistical significance benchmarks.
Most control variables in these models have, like the interaction variable, effects similar to those in column A. Several of the new control variables have notable effects: moralism strongly associates with greater probability of female lieutenant-governor candidates, with a 2-point shift from traditionalism to moralism having an effect comparable with moving from a Republican to a Democratic slate when primary voters select the candidate. With the measures of lieutenant governors’ powers, legislative powers associate with lower likelihood of women receiving major-party nominations, whereas executive powers have a rather smaller and less statistically significant effect in the opposite direction. This fits with the idea that formally powerful offices are less often offered to women, even if elites such as the governor are willing to direct more (executive) powers to female officeholders once they are in office.
To further illustrate the results, Figure 4 shows estimates for a model like that of column A in Table 1, with covariates set to means. For comparison, each selection mechanism-party combination is shown over a range of the Enns and Koch (2013) state conservatism percentages. In three of the figure’s panels, the predictions are very similar: an estimated 30% likelihood of a female lieutenant-governor nominee when a state is 30% conservative, falling to approximately 25% likelihood when the state is 45% conservative. The rightmost panel, where Republicans select their nominees via primary elections, looks quite different. There, at the same levels of state conservatism, the probability of a female nominee is only 18% and 14%, respectively. Compared to other combinations of partisanship and selection mechanism, that is, Republican primary electorates appear about 40% less likely to nominate females to be lieutenant governor. This is an even starker partisan difference than seen in Figure 3; in that figure’s uncontrolled comparison, both parties’ primary electorates appeared less likely to select female candidates, though the effect was stronger among Republicans. The Table 1 and Figure 4 results, in adding control variables, suggest that the primary-election effect is actually exclusive to Republicans.
Estimated probabilities of female lieutenant-governor nominees, by party, selection mechanism, and state conservatism.
The analysis thus produces two main findings. One concerns contexts where lieutenant-governor candidates are appointed. There, Democratic and Republican gubernatorial nominees appear comparably likely to choose female running mates. This result sits in marked contrast to that of Bratton and Spill (2002), who find Democrats markedly more likely to use appointment power to bring women into judicial office. What explains the difference with the result here? It could be that the parties have different motivations in the two situations. Judicial appointments typically have more power than visibility, whereas a lieutenant-governorship nomination supplies more visibility than power—especially for the candidates whose ticket loses in the general election. Just as Republicans reach out to other traditional Democratic constituencies such as racial minorities by providing members of those constituencies with disproportionately visible roles (Hutchings et al. 2004; Philpot 2004), they may use lieutenant-governor nominations as a prominent but low-commitment means of demonstrating their openness to women. Judicial positions, with their relative obscurity, send such signals far less effectively.
The other main finding concerns cases where primary electorates select the candidates. There, Republicans appear less likely to nominate women. The data here cannot determine the precise mechanism producing this outcome: Republican women could be particularly averse, because of unwillingness to compete against copartisans or lower self-confidence, to facing the tumble of two campaigns and so especially unlikely to run when a primary election is necessary, or, perhaps, some key bloc of Republican primary voters could be less acceptant of female candidates. Whatever mechanism is in effect, the results suggest that these partisan differences may mediate the effect of different institutional arrangements for selecting candidates.
About Those Separate Ballot Lines
The above discussion focused on the distinction between the solidly shaded and merely (cross-)hatched states in Figure 2—that is, on whether or not gubernatorial candidates choose their own running mates. Because every state with appointed candidates also places the governor and lieutenant governor on a single unified ticket, though, one may wonder whether appointment really causes the effects seen in Table 1, or whether separate elections, which happen to correlate strongly with gubernatorial appointment of candidates, are responsible. After all, the institutions in question do have some of the same substantive implications; being linked on the ballot likely increases gubernatorial candidates’ interest in lieutenant-gubernatorial candidate characteristics, which may influence candidate-selection processes and outcomes. Indeed, recent work on lieutenant-governor candidates has focused on ticket splitting when the two offices run separately—especially in the context of descriptive representation of discriminated-against groups (Hood, Kidd and Morris 2015). Furthermore, the pool of candidates willing to put themselves forward for an independent race may differ from that of those comfortable campaigning on a joint ticket.
To test this, Table 2 repeats the analysis in Table 1, but replaces the variables measuring whether the governor appoints their running mates with those measuring whether the two offices are elected separately in the general election (control variables, omitted for brevity, produce very similar results to those seen in Table 1). In these models, the interaction of the candidate-selection institution with Republican affiliation never has a particularly large or statistically significant effect—even though they presumably capture via multicollinearity some of the effects seen in Table 1. This is not to say that separate elections are irrelevant: outside of the interaction, they do at times reduce the probability of female nominees in a statistically significant way. However, Table 2 overall indicates that the separate-election process is a relatively minor contributor to the central finding of partisan interactions with candidate-selection mechanisms. 17
Logistic Models Predicting Femaleness of Major-Party Lieutenant-Governor Candidates, 1986–2012.
Note. Numbers in parentheses give robust standard errors; control variables, omitted for brevity, are as in Table 1.
Indicates coefficients that are statistically distinguishable from zero (95% confidence, one-tailed).
Conclusion
This article opened with a puzzle: if the electorate is, as previous research suggests, more willing to vote for qualified female candidates than gatekeepers are to nominate women, why do appointed positions often see greater proportions of female nominees than do elected positions? Many answers to this question are possible; this article looked at possible differences across parties. Although partisan cultures may impede women’s pursuit of candidacies in right-wing parties, the evidence here suggests that, in the United States, Republican elites—at least those who face an election themselves—promote women to the ballot just as much as Democrats do, whether through personal conviction or the cold logic of electoral imperatives. Thus, though elite gatekeepers have a major role in determining how many women represent the party on the ballot (Nechemias 1987; Rule 1981; Sanbonmatsu 2002; 2006), they may not be responsible for the varying proportions of women representing different parties. Even if Republican elites and Democratic elites are equally inclusive in their candidate lists, primary electorates may produce substantial partisan gaps, as observed in the results here.
Naturally, candidates who will face an election alongside the lieutenant-gubernatorial candidate being appointed are a special class of elites. Future research may explore whether the partisan behaviors observed here carry over to other appointment settings, where the appointee need not stand for election. Also worth examining, perhaps through interviewing of party elites, is the reason for the seemingly larger disjunction between nominations and primary elections among Republicans. The mechanisms laid out in this article, following the previous literature, focused on divergent behavior between the two parties’ primary electorates (e.g., the Republican primary electorate being ideologically further from the general electorate, so that it selects candidates less in line with what strategic candidates would). Other explanations are also consistent with the observed outcome. The patterns observed here could also be explained through candidate behavior, with Republican women relatively less willing to put themselves in public, primary-election competition with other Republicans because of particular personality or cultural characteristics most common among Republican women, or because the party’s greater ideological homogeneity (Grofman et al. 1999) produces less ideological differentiation to motivate female candidate entry. Further observation of the dynamics of partisan selection process in elected versus appointed contexts could clarify whether or how these processes influence different parties’ tendencies to select female candidates. For example, focusing on women’s willingness to contest primary elections in different parties could suggest to what extent candidate rather than nominator behavior drives the partisan differences in candidate gender. The ideas could also be considered outside the United States (Hinojosa 2012).
Alongside these substantive questions, this discussion demonstrates possibilities in the seldom-studied lieutenant governorship. With its variation in selection and powers across states and over time, the office offers fertile ground for testing a wide range of theories about political institutions. Its relatively high visibility, especially during campaigns, also presents an important potential stepping stone in political careers, one useful for studying politicians’ professional trajectories. This is particularly important for underrepresented groups—not just women, but also ethnic, religious, or cultural minorities—as they advance into high office. The findings here may indeed suggest further research possibilities based on whether having primary elections to the lieutenant governorship limits women’s, especially Republican women’s, political careers more generally. By making one step of the political ladder harder to reach, this candidate-selection mechanism may have implications for how frequently women attain other offices as well.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.