The gender dynamics of elections in the United States have changed
dramatically since the 1980s. As recently as 20 years ago, women held only
two state governorships, two Senate seats, and barely 5% of seats in the
House of Representatives. For almost the entirety of U.S. history, women
candidates were novelties. Today, though the overall number of women
serving in these offices is still low, women are regularly running for
high-level elective positions, and one woman, Hillary Clinton, is even the
early front-runner to receive her party's nomination for president in
2008. Consistent with the rise in numbers of women candidates, recent
examinations of voting behavior have found that women tend to perform just
as well as men at the ballot box (Dolan 2004;
Seltzer, Newman, and Leighton 1997). This is not
to say that the current context of elections is gender neutral.
Investigators have continued to find myriad ways in which the electoral
process is gendered, including candidate recruitment (Sanbonmatsu 2006), voter perceptions that men are
better at handling certain policy issues than women and vice versa (Lawless 2004), and campaign strategy (Fox 1997). More generally though, as intolerance of
women in the political sphere has declined and women have begun succeeding
as candidates for public office, political parties and other electoral
gatekeepers have begun to strategize about when it is best to try to
capitalize on the new prominence of women candidates.
One of the first, and perhaps most dramatic, examples of the strategic
selection of a woman candidate occurred when Walter Mondale chose
Geraldine Ferarro to be his running mate on the Democratic presidential
ticket in 1984. The hope of the Mondale campaign was that by selecting
Ferraro, many independent and moderate female voters would flock to the
ticket (Bonk 1988). While no other presidential
candidate has opted for the strategy of selecting a woman, this decision
has been taking place at the state level with great regularity. In 24
states, the governor and lieutenant governor run on the same ticket. And
male gubernatorial candidates have been running with women in high
numbers. Of all the elected positions in American politics at the state
level, a higher percentage of women run as candidates and serve as the
lieutenant governor when running on the same ticket with the governor than
for any other elective office (CAWP 2005). In
2002, for example, in races where the governor and lieutenant governor ran
on the same ticket, a majority of lieutenant gubernatorial
candidates were female.
Clearly, the strategy of running with a woman has become widespread
practice for male gubernatorial candidates. While it can be difficult to
know precisely why specific women have been chosen as running mates, a few
examples from recent elections can be instructive.1
The selection of lieutenant governor candidates is typically
accomplished via a primary election, at a party convention, or directly by
the gubernatorial candidate.
In 1994, campaign analysts suggested
that Republican candidates George Pataki in New York and George Voinovich
in Ohio both were compelled to select a female running mate to demonstrate
that they were sensitive to the concerns of women.
2 Gail Collins, “Capt. Al Flies Starship GOP,”
Newsday, May 25, 1994, 6; Mike Curtin, “Voinovich Aims for
Appalachia,” Columbus Dispatch, January 20, 1994,
C2.
In Minnesota in 2002, gubernatorial Republican candidate Tim
Pawlenty went to the unusual step of selecting state representative Carol
Molnau four months before the nominating convention, where running mates
are usually announced. Pawlenty asserted that he wanted to “shake up
the race” and show he was running an “unconventional
campaign.”
3 Dane Smith, “Pawlenty
Announces Running Mate; Rep. Carol Molnau from Cologne is his Choice for
Lieutenant Governor,” Star Tribune, January 15, 2002,
B3.
Highly strategic selections of women running mates also
appeared to occur in Iowa, Wisconsin, and Pennsylvania in 2002.
What we do not know is whether this strategy is successful. In this
article, we directly examine the electoral effects of having a woman as a
running mate in gubernatorial elections. This research represents the
first examination of the electoral importance of running mates in
gubernatorial contests and also of the role gender plays in that process.
To assess whether running with a woman candidate has a significant impact
on voters, we analyze election exit poll data for gubernatorial elections
in selected years between 1990 and 2002. We conclude that running-mate sex
does influence vote choice in gubernatorial elections, even when we take
into account a wide range of individual-level and electoral
characteristics that are related to voter preference. Our research makes
two contributions to the literature on gender and elections. First, from a
practical perspective, we determine whether the electoral strategy of
running with a woman is effective—does it lead to more votes? This
will allow us to extend the research on gender and voting to a previously
unexamined office. Secondly, and more broadly, our research allows us to
gauge whether and to what degree strategic calculations about gender are
influencing the outcomes of state elections.
THE ELECTORAL IMPACT OF RUNNING WITH A
WOMAN: HYPOTHESES AND EXPECTATIONS
What is the impact on voter choice if a male gubernatorial candidate
has a woman lieutenant governor candidate as his running mate? In
developing our specific hypotheses to answer this question, we must
acknowledge that there is no research examining the impact of running-mate
selection on gubernatorial vote choice. The literature on vote choice in
gubernatorial elections focuses primarily on traditional predictors (e.g.,
party, ideology, and so forth); economic conditions, both for the
individual voter and the entire state; and the national political context
(Atkeson and Partin 1995; Carsey and Wright 1998; Svoboda
1995). In light of the lack of research in this area, we must draw
upon the women and electoral politics literature and apply it to predict
how voters might respond to women lieutenant governor candidates.
Ultimately, we believe that the sex of the running mate will influence
voters. The literatures on candidate stereotyping and campaign strategy
provide clear indicators that having a woman running mate will be a
significant factor. The position of lieutenant governor is an executive
office, a type of position for which women have had a harder time gaining
acceptance (Adams 1975; Dolan
1997). Yet it is often a position with little political power, one
that is clearly subordinate to the position of governor, and therefore may
be viewed by reluctant or stereotyping voters as more appropriate for
women (Huddy and Terkildsen 1993b). Overall,
while we expect that running-mate sex will matter, because of the
constantly evolving role of gender in electoral politics, the effects
should not be static. The effects of a woman running mate will matter with
some voters in some electoral environments. Specifically, we develop five
hypotheses that draw upon past findings regarding voter reactions to women
candidates. Our goal here is to be thorough in combing the literature for
reasonable expectations regarding how and when the sex of a running mate
will influence vote choice for the candidate at the top of the ticket.
Therefore, we do not a priori anticipate that we will find more support
for one of the following hypotheses, and the hypotheses themselves are not
mutually exclusive. Instead, we propose what we feel is a comprehensive
list of possible hypotheses generated from existing literature.
The Positive Association Hypothesis suggests that voter
perceptions of a male gubernatorial candidate may improve by the selection
of a woman running mate. By running with a woman, these candidates
demonstrate the quality of inclusiveness, because women are still a small
minority in most elective positions. Further, the male gubernatorial
candidates may also have transferred to them many of the positive traits
and issue competencies that voters associate with women electoral
candidates. Gender stereotyping scholars have found that women politicians
are identified as more compassionate, willing to compromise, and oriented
toward people. Male politicians are stereotyped as more assertive, active,
and self-confident. In terms of issue expertise, female candidates are
perceived as better suited than men to address issues such as education,
heath care, and helping the poor, while male candidates are seen as more
competent to address economic and business issues as well as crime (Burrell 1994; Huddy and Terkildsen
1993a; Leeper 1991; Rosenwasser and Dean 1989; Rosenwasser and Seale 1988; Sapiro
1981–82). By running with a woman on the ticket, male
gubernatorial candidates might become associated with many of the positive
traits and issues associated with women, which are particularly well
received by voters in state-level politics (Adams
1975; Huddy and Terkildsen 1993b). If
this assessment holds true, we expect that all male candidates running
with a woman lieutenant governor candidate will receive more votes than
candidates running with a man.
The Ideological Balancing Hypothesis also relies on the
stereotyping literature. In terms of belief stereotypes, female candidates
are viewed as more liberal than male candidates (Alexander and Andersen 1993; Koch
2000; McDermott 1997, 1998). In fact, Koch (2002)
has shown that women candidates are perceived to be more liberal than they
actually are. In contrast, to determine the ideological leanings of male
candidates, voters rely on the candidate's party affiliation rather
than on gender stereotypes of men (Koch 2002).
These findings suggest that running with a woman will cause a male
gubernatorial candidate to appear more liberal than if he ran with a man.
If the Ideological Balancing Hypothesis holds true, we would expect
Republican male candidates to benefit from running with a woman. A
conservative man (as voters perceive Republican men to be) paired with a
liberal woman (as voters perceive all women to be) is a more ideologically
balanced Republican ticket than is one with two men (conservative plus
conservative, in the voters' minds). Alternatively, we would expect
Democratic male candidates to perform worse when they run with women. In
this case, voters would perceive both Democratic candidates to be liberal,
but since women are perceived to be more liberal than men, a male-female
ticket appears more ideologically extreme to voters than would a male-male
ticket. In deriving these expectations, we are relying on the
well-established supposition that American voters prefer moderate to
ideologically extreme candidates. In addition to these overall effects,
this hypothesis also suggests that certain voters, particularly
ideological moderates, will be influenced more than others. If voters are
reacting to a ticket's ideological balance, we would expect moderate
voters to be especially drawn to Republican male gubernatorial candidates
with women running mates and deterred from voting for Democratic
candidates who have a woman running mate.
The Appeal to Women Voters Hypothesis is based on the premise
that men select women running mates to help appeal to women voters. The
conventional reasons suggested for putting a woman on a ticket are to
demonstrate that a male candidate is comfortable with and also in touch
with the concerns of female voters. In support of this reasoning, several
studies of voters suggest that women voters may prefer women candidates
(Dolan 1997, 2004;
Plutzer and Zipp 1996; Rosenthal 1995; Sanbonmatsu
2002). If this Appeal to Women Voters Hypothesis holds true, then,
we expect that male gubernatorial candidates running with a woman will
perform better among women voters than male gubernatorial candidates
running with a man.
The previous three hypotheses all assume that the sex of the running
mate will influence voters regardless of the electoral context. Yet prior
research demonstrates that the role of gender in elections can be
conditioned by environmental factors (e.g., Dolan
2004; Herrnson, Lay, and Stokes 2003).
Thus, we identify two additional hypotheses that bring into focus two
features of the electoral context that we expect may influence the degree
to which candidate sex will matter. The first of these, the Social
Acceptance Hypothesis, anticipates change over the course of the four
elections that we study. Since 1990, as women have become more frequent
candidates, public acceptance of women candidates has grown tremendously
(Thomas and Wilcox 1998). We would expect that
this changed environment should lead to increased voter support for
gubernatorial tickets containing women in more recent years compared to
1990.
The second contextual hypothesis is the Issue Environment
Hypothesis. This hypothesis recognizes that not all electoral
environments are the same. Specifically, the policy issues that dominate
in an electoral contest are not always identical. Some research has shown
that women candidates perform better in electoral environments where the
most important policies being debated are “women's
issues” (Burrell 1994; Fox 1997; Lawless 2004). As
the literature on gender stereotyping discussed earlier finds, voters
ascribe different policy expertise to women and men candidates. Therefore,
in electoral contexts that are dominated by “women's
issues,” such as education or health care, the advantage of running
with a woman candidate should be greater than in environments dominated by
“men's issues,” such as crime or the economy.
While each of these hypotheses identifies the manner in which having a
woman running mate will impact vote choice, it is important to acknowledge
the literature suggesting that having a woman running mate could have
little or no impact on voter behavior in gubernatorial elections. For one,
lieutenant governor candidates have quite low visibility. While there is
no survey concerning voter knowledge of lieutenant governor candidates, we
know from studies of Congress that voters recall the name of the incumbent
less than half the time and challengers less than 20% of the time (Jacobson 2004, 123). It seems reasonable to assume
that similar or even lower percentages of voters could recall the name of
lieutenant governor candidates. Secondly, assuming that voters do know the
name and identity of the lieutenant governor candidate, there is no reason
to expect that candidate sex will affect vote choice. Research across the
1990s has found that there is little or no evidence of voter bias for or
against women candidates (Burrell 1994; Dolan 2004; Seltzer, Newman, and
Leighton 1997). Given the low-information aspect of lieutenant
governor candidacies and the lack of gender bias in voting in general
elections, we might find that having a woman running mate will neither
hurt nor help male gubernatorial candidates with the voters. Our general
expectation, though, as we have detailed, is that running-mate sex will
influence voters' preferences in gubernatorial contests.
DATA AND STATE CONTEXT
To conduct our analyses, we rely primarily on general election exit
polls conducted by the Voter News Service (VNS) and made available by the
Inter-university Consortium for Political and Social Research at the
University of Michigan. More specifically, we use the national exit poll
files for 1990, 1994, 1998 and 2002.4
Those
who watched the electoral returns in 2002 might recall that reports of
exit poll results were discontinued early that evening. This was due to a
problem with VNS's computer system, not to any problems with the
questionnaire or sampling procedures. After the election, the data were
analyzed by a panel of academics who “judged the data to be
comparable in quality to previous exit polls” (Introduction, Voter
News Service 2002 exit poll documentation).
Not all states hold
their gubernatorial elections during the same year, of course. Of the 24
states where the governor and lieutenant governor run on the same ticket,
19 hold gubernatorial elections in midterm election years. In contrast,
only 4 states (Indiana, Montana, North Dakota, and Utah) hold these
elections during presidential election years, while Kentucky's
gubernatorial elections occur in odd years. By focusing on the midterm
election years from 1990 through 2002, our analyses attempt to capture
voters from 19 of the 24 states that pair their gubernatorial and
lieutenant gubernatorial candidates on the same ballot.
5 Unfortunately, though, the national exit poll files do not
contain respondents from all of these 19 states in each election.
Specifically, the 1990 file did not include South Dakota voters, while
voters from both Alaska and South Dakota were excluded in 1994. In 1998,
New Mexico is not represented, and Alaska and Hawaii were not included on
either the 1998 or 2002 files.
Table 1 presents a list of these 19 states.
While it would not be accurate to label these as fully representative of
all American states, they do vary along important electoral dimensions. In
contests for state offices (both legislative and executive), a majority
have competitive two-party systems, as assessed by Austin Ranney's
index of state party competition. Four states have either a Democratic
dominant system or a competitive-leaning Democratic system, while four
other states are competitive-leaning Republican or Republican dominant
states. Ten of the 19 states have individualistic political cultures,
seven are moralistic, and two are characterized as having a traditional
culture. A state's political culture is related to the presence of
women in electoral politics; those with traditional cultures witness fewer
women candidates, and moralistic cultures, with their emphasis on fairness
and equality, foster more female candidacies (Norrander
and Wilcox 1998; Rule 1990). A more
direct measure of a state's willingness to elect women is presented
in the final column of Table 1: the percentage
of a state's legislators that are women (averaged across the four
election years of our study). There is considerable range across these
states for this measure, with a low of 11.1% for Pennsylvania and a high
of 33.8% for Colorado. Further, all geographic regions of the United
States are represented in these 19 states, albeit some more heavily than
others. In particular, midwestern states are much more common than are
southern or southwestern states.
Characteristics of states holding gubernatorial elections on midterm
electionyears
In the analyses that follow, we examined only those contests where the
gubernatorial candidate was male. Our hypotheses assume that the candidate
on the top of the ticket is male, and so races with female gubernatorial
candidates were excluded from all analyses.6
Among states in our analyses, there were four female
gubernatorial candidates in 1990 (the Democratic candidate in Kansas and
the Republicans in Alaska, Nebraska, and Pennsylvania), four in 1994 (the
Democrats in Illinois and Iowa and the Republicans in Hawaii and
Maryland), three in 1998 (the Democrats in Colorado and Connecticut and
the Republican in Maryland), and four in 2002 (the Democrats in Kansas,
Maryland, Massachusetts, and Michigan).
For the states for which
we have exit poll data available, and when the gubernatorial candidate was
male, 42% of the lieutenant gubernatorial candidates in 1990 were female.
Female running mates were even more common in more recent years. In fact,
a majority of gubernatorial running mates were women in 1994, 1998, and
2002 (57%, 59%, and 53%, respectively).
DOES RUNNING WITH A WOMAN HELP? EXAMINING
VOTER PREFERENCES
To begin our empirical examination, we analyze vote choice in
gubernatorial contests. Specifically, we estimated a series of logistic
regression models that predict two-party support for the Democratic
governor/lieutenant governor ticket.7
Vote choice for the gubernatorial ticket was assessed by the
following exit poll question: “In today's election, did you
just vote for: [name of Democratic candidate], [name of
Republican candidate], Other: Who?” We recoded responses so
that a vote for the Democratic ticket was coded as 1 and a vote for the
Republican ticket was coded as 0.
We have included a number of
variables designed to test our hypotheses. Two variables (female
Democratic lieutenant governor and female Republican lieutenant governor)
assess the overall impact of running-mate sex on gubernatorial vote choice
(refer to the Appendix for the coding of variables included in the
models). The inclusion of these two variables allows us to test the
Positive Association and Ideological Balancing Hypotheses. Support for the
Positive Association Hypothesis would be provided if the coefficient for
the Democratic lieutenant governor variable is positive and significant or
if a negative and significant coefficient for the sex of the Republican
lieutenant governor candidate emerges (since the dependent variable is
support for the Democratic ticket). Finally, if both of the lieutenant
governor variables are significant and negative, the Ideological Balancing
Hypothesis would be supported. If voters are basing their choices on
ideological balancing, we should see stronger support for Democratic
tickets with male versus female running mates since voters tend to
perceive women as more liberal than men. In contrast, voters should show
more support for Republican tickets with female versus male running mates
for this balancing hypothesis to be supported.
To test further the Ideological Balancing Hypothesis, we included
interaction terms. Since this hypothesis assumes that moderates will be
especially likely to engage in ideological balancing, a dummy variable for
moderates was interacted with each of the two female lieutenant governor
variables. This hypothesis predicts a decrease in moderate support for
Democratic tickets containing women running mates and an increase in
moderate support for Republican tickets with women. Keeping in mind that
the dependent variable is support for the Democratic ticket, negative
coefficients for both of these interaction terms would demonstrate support
for the Ideological Balancing Hypothesis. Similarly, interaction terms
assess the Appeal to Women Voters Hypothesis. In this case, a female voter
dummy variable was interacted with the two lieutenant governor candidate
dummies. If tickets with female running mates are more appealing to women
voters, the interaction term for the Democratic lieutenant governor
candidate will be positive, while the interaction term for the Republican
candidate will be negative.
Our models also include control variables, some of which capture
elements of the state's electoral context. The first (safe Democratic
contest) is a race outcome prediction. Data for this variable were
collected from Congressional Quarterly, whose staff labels each
gubernatorial race as safe or likely for Democrats, safe or likely for
Republicans or a toss-up. These predictions are published a few weeks
before the election. This summary measure incorporates a wide range of
relevant electoral circumstances, such as whether either of the candidates
is an incumbent, the strength of any challengers, the party congruency of
candidates and the state, and financial strength of the candidacies. Since
vote choice is based not only on voter attitudes but also on the strength
of candidates who appear on the ballot, it is important that our models
control for the latter.8
Including this
variable has an additional benefit. It is possible that women have been
selected as running mates for only certain types of contests, such as when
the gubernatorial candidate was likely to win. If this were the case, then
our analyses of voter preferences could demonstrate that voters prefer
female to male running mates, when an alternative explanation would be
that voters prefer strong over weak gubernatorial candidates. Thus,
controlling for the strength of the gubernatorial candidates is crucial
for us to test accurately our hypotheses.
We also include
variables that capture a state's openness to voting for women, under
the assumption that voters are currently more likely to vote for women in
states that have a tradition of electing women (
Hill
1981; but see
Hansen 1993). To assess a
state's likelihood to elect women, we include the percentage of women
serving in the state's legislature. Since we expect the presence of
women in the state legislature to be related to voting for female running
mates (rather than related to our dependent variable, vote preference for
the Democratic ticket), we interacted this variable with the sex of the
Democratic and sex of the Republican gubernatorial candidates.
9 These interaction terms are highly correlated
with one of their component variables. Specifically, the correlation
between the women in state legislature by sex of Democratic running-mate
interaction term and sex of Democratic running-mate variable is higher
than .9 for each year. Similarly, the Republican running-mate interaction
term and sex of Republican running-mate variable are also correlated at .9
or greater. Multicollinearity is thus present in the models, possibly
inflating the standard errors (Schroeder, Sjoquist, and
Stephan 1986). Yet this problem is less severe when
multicollinearity is present between interaction terms and their component
variables largely because the standard errors for interaction terms are
quite likely to be smaller than are standard errors present in simple
additive models. This difference is due to the fact that standard errors
for interaction terms are conditional: The effect of an independent
variable on the dependent variable depends upon the value of a second
independent variable. Therefore, as Robert Friedrich (1982, 803) concludes, “Though a multiplicative
term and its constituent variables are often highly correlated, this
multicollinearity does not pose problems for the interpretation of the
regression results” (also see Jaccard, Turrisi,
and Wan 1990). Further, solving for multicollinearity is typically
achieved by removing one of the variables that is correlated with another.
For us, this would entail removing the women in state legislature by sex
of running-mate interaction terms, but in so doing, we risk specification
bias and its accompanying biased coefficient estimates (Schroeder, Sjoquist, and Stephan 1986). Given the
unfortunate choice of selecting between two potential
problems—specification bias or multicollinearity—and given
that multicollinearity is less concerning when due to the inclusion of
interaction terms, we opted to keep the state legislature interaction
terms in our model.
Finally, we include a number of individual-level predictors that other
researchers have found to be related to vote choice. In particular, voting
theories and past research find that gubernatorial vote preference is
influenced by political attitudes, such as party identification, ideology,
and presidential approval (Atkeson and Partin
1995; Carsey and Wright 1998; Svoboda 1995). Perceptions of the national economy and
of one's personal financial situation influence the vote for governor
(Cook, Jelen, and Wilcox 1994; Stein 1990), as do demographics such as sex (Cook, Jelen, and Wilcox 1994). Data for these
variables were contained on the exit polls.10
For the years of our analyses, VNS used two versions of their
exit poll questionnaire, with approximately one-half of respondents
receiving each version. The questions of interest to us were not contained
on both questionnaire versions for any of the four years, and so each of
our models includes roughly one-half of the possible respondents in each
state.
We estimated models separately for each year, results of which appear
in Table 2 (the bolded variables are the ones
that directly test the hypotheses; nonbolded variables are controls).
Considering first the control variables, the electoral context faced by
the voters was related to their vote choices. For all years, in states
where the gubernatorial contest was predicted to be safer for the
Democratic candidate, more voters supported the Democratic
gubernatorial–lieutenant gubernatorial ticket. The presence of women
in the state legislature was also related to vote choice, but not always
in the anticipated direction. Voters were more likely to support
Democratic tickets with female running mates in states with more women in
the legislature for a single year (1998). However, voters in states with
more women in the legislature were actually less likely to support
Democratic tickets with female running mates in 1994 and 2002, and less
likely to support such Republican tickets in 1998. While we are not the
first to find that the presence of women in a state's legislature can
be negatively related to the presence of women in its executive offices
(Hansen 1993), we are not certain what accounts
for this result. Since the lieutenant governor position is a subordinate
one, perhaps voters in states with fewer women in the state legislature
(and likely more traditional attitudes toward women's roles) will be
more supportive of tickets containing women running mates. Alternatively,
it is possible that voter preference for women lieutenant governors (a
statewide office) would be related to the presence of women in leadership
roles, rather than simply their numbers in the legislature. Just as the
likelihood of a state adopting women's health policies is more likely
when women serve in leadership positions in the state legislature (Tolbert and Steuernagel 2001), voters could be more
attracted to gubernatorial tickets containing women when women serve in
higher-profile legislative roles in a state.
Logistic regression models of vote choice for Democratic
gubernatorial/lt. gubernatorialcandidates
Turning to the individual-level control variables, there is much
consistency in the results across the four elections. The key political
attitudes of party identification, presidential approval, and liberal
ideology were related to vote choice in expected directions for all years.
For only 1998 and 2002 were economic attitudes significantly related to
gubernatorial voting. Preference for Democratic over Republican tickets
was more likely for those who felt that the national economy was strong in
1998 and poor in 2002, likely because evaluations of the national economy
are tied to presidential evaluations. Finally, voter sex, when controlling
for a range of other characteristics, is not related to gubernatorial vote
choice for any of the four elections.
More importantly for our research, we find that the sex of the
lieutenant gubernatorial candidate does influence voter choice. However,
the effect of running-mate sex was not constant across years or parties.
In 1990, for neither party was overall voter support for gubernatorial
tickets influenced by the sex of the running mate. In this year, though,
we do find one pocket of voter hostility toward female running mates:
Women voters were less likely to vote for Republican tickets containing
female versus male running mates, in contradiction to the Appeal to Women
Voters Hypothesis. Some hostility toward Republican female running mates
also existed in 1994, this time among moderate voters. For Democratic
tickets, however, men running with women were advantaged this year, as the
Positive Association Hypothesis predicts.
Four years later, however, Democratic gubernatorial candidates were
disadvantaged when running with women. When the Democratic running mate
was female, voters were less supportive of the Democratic ticket than when
the running mate was male. Republicans, on the other hand, received more
voter support when running with women. This overall pattern does support
the Ideological Balancing Hypothesis, although we do not find that the
vote choices of moderate voters were influenced by running-mate sex in
1998, as this hypothesis also predicts. Moderates thus do not appear to
base their vote choices on whether the gubernatorial ticket is
ideologically balanced (in this or any of the other years). Finally, in
2002, we find no evidence that voter preference for Republican tickets was
influenced by running-mate sex (among all voters, moderates, or women).
Democratic gubernatorial candidates, in contrast, did benefit when running
with female running mates among all voters (supporting the Positive
Association Hypothesis) and among women voters (as the Appeal to Women
Voters Hypothesis predicts).
SOCIAL ACCEPTANCE OR FRIENDLY ISSUE
ENVIRONMENTS?
At first glance, our results demonstrate some support for the Social
Acceptance Hypothesis that voters would be less hostile to tickets
containing women in more recent elections. This result is clearest among
Republicans, where in 1990 and 1994, subgroups of voters (women and
moderates, respectively) clearly preferred Republican tickets with male
versus female running mates. This hostility disappears in 1998 (when
voters prefer female running mates) and in 2002 (when running-mate sex was
not related to vote choice). Among Democrats, the over-time trend
demonstrates the irrelevance of running-mate sex in 1990, voter preference
for female running mates in 1994, for male running mates in 1998, then
again for female running mates in 2002. With the interruption of 1998,
then, the Social Acceptance Hypothesis also appears to have some support
for Democratic tickets. Yet these analyses do not allow us to sort out
whether decreasing voter hostility to female running mates was due to
greater social acceptance of women candidates or due to differences in the
issue environments surrounding these elections. The trends we uncovered
could have been due to variation, across election contests, in whether
women-friendly issues dominated political discourse.
To systematically examine the relationship between voter preference
for female running mates and the issue environment, we rely on two
different types of data. First, we turn to a measure from the exit polls
that assesses which issue voters deemed to be most important to them on
each election day. Unfortunately for our purposes, this particular poll
question asked voters specifically which issue mattered most in their
congressional vote. While the issues at play in gubernatorial elections
are often the same as those that dominate congressional elections, the
issue environments are not always identical.11
David E. Rosenbaum, “The 1998 Elections: The
Nation—Governors,” New York Times, November 4, 1998,
B5.
There is, however, no other source of voter-issue salience
for gubernatorial elections across all of our states, and so if we wish to
know which issues were on voters' minds in the voting booth, we must
rely on this exit poll measure. We collapsed voters' responses to
this query into two categories: 1) women's issues and 2) men's
or neutral issues. We rely on the gender-stereotyping literature to
classify the issues; women's issues include health care, education,
social security, the environment, and abortion.
The Issue Environment Hypothesis predicts that female running mates
will fare better when women's issues dominate the issue space. To
test this hypothesis, we compared the percentage of voters who felt a
women's issue mattered most for their vote to the sex of the
lieutenant gubernatorial candidate on their ballot, separately by party.
We then linked this comparison to the results from Table 2, which demonstrate when (and which) voters
preferred female over male running mates. If our multivariate results can
be explained by the issue environment, in situations where voters showed a
preference for female running mates, voter salience of women's issues
should be higher when the lieutenant gubernatorial candidate was female.
So, for example, our Table 2 results
demonstrated that voters preferred tickets with female Democratic running
mates in 1994 (to ease interpretation, Table 2
results are summarized in Table 3). If women
running mates were advantaged by the issue environment in this year, we
would see a higher salience of women's issues among voters who were
faced with a ticket containing a female running mate. A glance at the
middle two columns of Table 3 does not
demonstrate such a difference; the salience of women's issues did not
differ by Democratic running-mate sex for 1994. Of the other three
situations when voters preferred female running mates (Republican tickets
in 1998, Democratic tickets in 2002, and women voters' support for
Democratic tickets in 2002), in only one do we find support for the Issue
Environment Hypothesis. In 2002, 43.2% of voters in states where the
Democratic running mate was female ranked women's issues as mattering
most to their vote, whereas only 39.7% of voters faced with a male
Democratic running mate felt women's issues were most important to
their vote.
Analysis of issueenvironment
A corollary to this hypothesis is that female running mates should be
disadvantaged when voters are especially unconcerned with women's
issues. Yet in the three situations where we uncovered voter hostility
toward female candidates (women voters choosing among Republican tickets
in 1990, moderate voters choosing among Republican tickets in 1994, and
all voters choosing among Democrats in 1998), the salience of women's
issues in voters' minds did not differ by running-mate sex. Thus,
when using voter attitudes to test the Issue Environment Hypothesis, we
find little support that it explains why voters prefer women running mates
in some years rather than others.
Our second measure of the degree to which women's issues existed
in the political discourse surrounding gubernatorial elections comes from
news accounts of these elections. Specifically, we examined newspaper
stories about gubernatorial election outcomes that were printed the day
after election day in each state contained in our multivariate analyses.
In the rare case when there were no stories on the day after election day,
we read stories that were published in the days leading up to the
election. News articles were obtained via a search of the Lexis-Nexis
archive. This search resulted in sufficient data for all years except
1990, where we could not locate any stories about the gubernatorial
elections in eight of the states in our analysis (largely because many
newspapers are not archived in Lexis-Nexis until after 1990).
To determine the salience of women's issues in each state's
gubernatorial election, we read the news stories for each year and
calculated the percentage of all issues mentioned in the stories that were
women's issues. We then compared the salience of women's issues
in the state to the sex of the Democratic and Republican lieutenant
gubernatorial candidates. Results of this analysis appear in the final
columns of Table 3. The cases for these data are
states, not voters, and so we are able to tabulate results only for the
overall effect of running-mate sex on voter choice and cannot present
results for specific types of voters. Using the same analytical method we
employed with the voter data, we find little support for the Issue
Environment Hypothesis. For the three situations when our regression
analyses demonstrate voter preference for women running mates, in only one
was the mention of women's issues in the gubernatorial election
higher for female versus male running mates (among Republican tickets in
1998).12
The difference in mean percentages
for some of these comparisons is large, but fails to reach statistical
significance for two reasons. First, the number of cases is low (under 10)
for each comparison group. Second, the variation within many of these
groups is quite large.
Further, for the one situation where
voters preferred male running mates, the prevalence of women's issues
in the election discourse did not differ by running-mate sex.
DISCUSSION AND CONCLUSION
Ultimately, our research provides the first empirical assessment of
whether seeking office with a woman running mate influences a
candidate's electoral prospects. Our results clearly indicate that
the selection of a woman running mate does indeed impact vote choice in
gubernatorial elections. Even when controlling for many key individual
voter characteristics and electoral circumstances, gubernatorial vote
choice was influenced by the sex of the lieutenant governor candidates in
each of the four elections we examined. Above and beyond the normal
predictors of the vote in these contests, the choice of running mate, at
least in terms of his or her sex, matters. For example, voters clearly
preferred viable candidates, such as those with the advantages of
incumbency. Yet in some situations, vote selections were also influenced
by whether a woman was on the ticket.
Among our three specific hypotheses predicting that running with a
woman would generally help male gubernatorial candidates, we find some
support for each. The strongest support appears for the Positive
Association Hypothesis, in that this prediction is supported in two
different election cycles (1994 and 2002) when voters overall preferred
Democratic tickets with female versus male running mates. In contrast,
support for the Appeal to Women Voters Hypothesis was mixed but generally
not very strong. Female voters were actually less likely to vote for
tickets containing women running mates in 1990. In 2002, however, women
voters showed a preference for female running mates. This could signal an
over-time change, with women being more likely to vote for tickets
containing women now compared to in the past. Analyses of future election
cycles will be necessary to determine whether this is indeed a trend.
We uncovered support for the Ideological Balancing Hypothesis in 1998,
when voters overall preferred Republican tickets with female running mates
and Democratic tickets with male running mates. Why would ideological
balancing have occurred only for this year? In retrospect, we believe that
broader factors explain this result. Gubernatorial election results are
often influenced by national trends, as was the case for the years of our
analysis. The “story” of 1994 was voter preference for
Republicans, both in Congress and the states, while bleak economic
conditions influenced gubernatorial voting in 2002 and, to a lesser
extent, 1990.13
Maureen Groppe and Jennifer
Babson, “Republican Surge Gives Party Bigger Share of
Statehouses,” CQ Weekly, November 12, 1994, 3247; Jonathan
Allen and Rebecca Adams, “States' Priorities Vs. Funds,”
CQ Weekly, November 9, 2002, 2945; Holly Idelson,
“Governors Find Re-Election a Trickier Proposition,” CQ
Weekly, November 10, 1990, 3838.
In 1998 though, voters
across the United States demonstrated a preference for moderate
candidates.
14 Alan Greenblatt with Karen
Foerstel and Geoff Earle, “For GOP, a Few Points of Light,”
CQ Weekly, November 7, 1998, 2999; Geoff Earle, “Governors:
States of Suspense,” CQ Weekly, October 24, 1998,
2876.
Journalists in particular singled out specific Republican
gubernatorial candidates who benefited from the voters' centrist
mood: Tom Ridge (PA), George Pataki (NY), John Rowland (CT), and Paul
Cellucci (MA)—the final three of whom ran with female running mates.
Given Bill Clinton's success two years earlier in winning reelection
by running as a centrist, and given the fact that the 1998 elections
occurred under the cloud of much national partisan fighting over the
Clinton-Lewinsky scandal, it is not surprising that voters turned to
moderate gubernatorial tickets in this year.
Another conclusion from our research is that the effect of
running-mate sex on gubernatorial vote choice is not static, but rather
depends upon the electoral context. Among the many contextual factors that
could influence voter preferences, our results most consistently
demonstrate that the increasing visibility of female politicians has
translated into voter support for tickets containing women running mates.
Further, a women-friendly issue environment did contribute occasionally,
but not always, to voters selecting tickets containing women. However,
women running mates were not disadvantaged when the issue environment was
instead dominated by men's issues. Finally, national-level trends,
such as voter moods, can influence whether running-mate sex will matter
for voters in the polling booth. Thus, our primary conclusion here, and it
is a significant one, is that the electoral context interacts with voter
attitudes in important ways in determining when voters will turn to
tickets containing female running mates.
Finally, our results show the importance gender continues to play in
electoral politics at the state level. As with many electoral decisions,
the selection of a running mate is almost entirely a strategic decision
whereby political actors are looking for some sort of campaign advantage.
We conclude that the seemingly innocuous and low-profile choice of a
female rather than a male gubernatorial running mate does have
consequences in the voting booth. Further, we find that the preference for
women running mates is more common among all voters, rather than only
subsets of voters (such as women or moderates), suggesting that the
importance of selecting a female running mate is widespread across the
electorate. Campaign strategists probably assume this to be true, as they
are increasingly utilizing gender considerations to appeal to voters. In
response, voters appear to be reacting to these cues.
Appendix: Description and Coding of
Variables
Electoral Context Variables
Female Democratic/Republican Lt. Governor: 0 if candidate is male,
1 if female.
Safe Democratic Contest: Congressional Quarterly's
prediction of gubernatorial race outcome; 1 = safe Republican, 2 = likely
Republican, 3 = toss-up, 4 = likely Democratic, 5 = safe Democratic
Women in State Legislature: percentage of a state's legislators
that are female.
Exit Poll Questionnaire Items (including full question
wording)
Democratic Party Identification: “No matter how you voted today,
do you usually think of yourself as a Democrat, Republican, Independent,
Something else?” Recoded as 1 = Republican, 2 = Independent, 3 =
Democrat.
Presidential Approval: “Do you approve or disapprove of the way
George Bush [1990]/Bill Clinton [1994,
1998]/George W. Bush [2002] is handling his job as
president?” Recoded as 0 = disapprove, 1 = approve.
Liberal and Moderate: “On most political matters, do you
consider yourself liberal, moderate, conservative?” Liberal: 1 =
liberal, 0 = other. Moderate: 1 = moderate, 0 = other.
National Economic Conditions: “Do you think the condition of the
nation's economy is excellent, good, not so good, poor?”
Recoded as 1 = poor, 2 = not so good, 3 = good, 4 = excellent.
Personal Financial Situation: “Compared to two years ago, is
your family's financial situation better today, worse today, about
the same?” Recoded as 1 = worse today, 2 = about the same, 3 =
better today.
Female: “Are you male, female?” Recoded as 0 = male, 1 =
female.
