Title: Has the Gender Gap in Voter Turnout Really Disappeared?
André Blais, and Elisabeth Gidengil
This paper can be referred to as follows:
Kostelka, Filip, André Blais, and Elisabeth Gidengil, forthcoming,“Has the Gender Gap in
Voter Turnout Really Disappeared?”, West European Politics, doi:
We thank Damien Bol, the members of the Research Chair in Electoral Studies at the University
of Montreal, the participants in the 2016 ECPR general conference in Prague, and this journal’s
reviewers and editors for helpful comments on the previous drafts of this manuscript.
According to conventional wisdom, the traditional gender gap in voting has disappeared
or even reversed in most established democracies. Drawing on the existing literature on sex
differences in political engagement and on pioneering voter turnout theories, this article
questions the conventional assumption and hypothesises that women still participate at lower
rates in less important elections. It systematically tests this hypothesis by exploring the impact
of sex on voter turnout in different electoral arenas. The empirical analyses of two cross-
national datasets (Making Electoral Democracy Work and the European Election Study)
demonstrate that although there is generally no gender gap in first-order elections, women tend
to vote less than men in second-order contests. This reflects women’s weaker interest in politics
and their lower levels of knowledge about politics in second-order electoral arenas.
The question of sex differences in electoral participation has a long pedigree in political
science (e.g. Gosnell, 1930; Tingsten 1963; Verba et al. 1978). Writing in 1955, Duverger (p.
14) highlighted “a problem of women non-voters.” Lower voting rates on the part of women
were typically attributed to social norms, structural and situational factors, or women’s late
enfranchisement (e.g. Norris 2002; Franklin 2004; Mayer 2010: 116). The latter explanation is
corroborated by the fact that in established democracies the gender gap in turnout gradually
shrank as new cohorts of women entered the electorate. Thus, by the turn of the 21st century,
Verba et al. (1997: 3) were referring to “the well-known equivalence in voter turnout” in U.S.
elections. More broadly, researchers have observed that, “traditional gender differences in
voting participation diminished in the 1980s and 1990s, or even reversed, in many advanced
industrialized countries” (Norris 2002: 96).
However, these optimistic assessments are mostly based on national elections. In
today’s era of multilevel governance, it is essential to examine the relationship between sex
and electoral participation at different levels of government. Drawing on pioneering works in
political science, recent research on sex differences in political engagement, and large-scale
survey data from national, subnational and supranational elections, we challenge the
widespread belief that established democracies have achieved full sex equality in voter turnout.
We start by formulating a theoretically grounded hypothesis regarding sex differences
in voter turnout in second-order elections. In line with earlier research, we argue that voting in
these elections is more strongly driven by political interest and political knowledge. When
asked in surveys, women in most established democracies express lower levels of political
interest and give fewer correct answers to political knowledge questions than men (Inglehart
1981; Verba et al. 1997; Burns et al. 2001; Gidengil et al. 2004; Stolle and Gidengil 2010;
Kittilson and Schwindt-Bayer 2010, Fraile 2014, Dassonneville and McAllister 2018).
Accordingly, we should expect a gender gap in turnout in second-order elections. We also
discuss other potential explanations of sex inequality in voter turnout and explain why they are
unlikely to account for the difference between first-order and second-order elections. We
subsequently test these explanations along with our main hypothesis by including control
variables corresponding to the various explanatory factors.
We validate our theory with a systematic study of voting in different types of elections
using two different cross-national datasets that span 26 countries and include 42,415
respondents. Our analysis proceeds in three steps. First, we examine whether there is a gender
gap in turnout in different types of elections. Our analyses confirm that there is no gender gap
in first-order elections, which are typically elections to the national parliament. In contrast,
turnout tends to be higher among men in second-order elections, which comprise most
subnational and supranational elections. Second, we use a non-linear adaptation of the Blinder-
Oaxaca decomposition and logistic regressions to ascertain the sources of these gender gaps.
The results corroborate our expectations: the turnout gaps in second-order elections reflect sex
differences in political interest and political knowledge. Third, we conduct additional analyses
to show how political interest and political knowledge explain the varying effects of sex on
voter turnout across election types.
Theory and Hypotheses
The political science literature shows that sex inequalities persist in some non-electoral
forms of political participations such as protest participation, political persuasion and
campaigning (e.g. Inglehart and Norris 2003; Norris 2007; Coffé & Bolzendahl 2010; Ondercin
and Jones-White 2011; Beauregard 2014; Córdova and Rangel 2017). However, it is generally
considered that women vote at the same rate as men in most advanced democracies (Verba,
Burns and Schlozman 1997; Norris 2002; Mayer 2010). As a result, relatively little recent
research tackles sex differences in voter turnout (but see Harell 2009). Interestingly, two studies
have reported as a secondary finding that women participate less than men in European
Parliament elections (Soderlund et al. 2011; Blais and Kostelka 2015). Women are also less
likely than men to vote in Swiss federal elections (Engeli et al. 2003; Stadelmann-Steffen and
We argue that these cases are not exceptions but reflect a broader pattern of sex
differences in participation between less important (second-order) and more important (first-
It has long been recognized that voter turnout depends heavily on psychological
engagement with politics, particularly in less important elections (Campbell 1960; Campbell
et al. 1960). According to this line of thought, the electorate can be divided into core voters,
who are politicised and interested in politics, and peripheral voters, who are less
psychologically involved in politics. While core voters go to the polls in most circumstances,
peripheral voters “have less intrinsic interest in political matters” and vote “only when there is
strong stimulation to do so” (Campbell 1960: 409).
Perhaps the most important criterion that determines the importance of an election in
an established democracy is election type. While voter turnout typically varies by only three
percentage points from one election to another election of the same type within countries (Blais
2007), the differences across election types often exceed ten percentage points.
Except for a
few specific cases discussed below, participation is considerably lower in subnational and
supranational elections. For a number of reasons, these electoral contests tend to be “low-
Córdova and Rangel’s (2017) results suggest that the traditional gender gap may still exist in national elections.
However, their Table A1 in the Electronic Appendix reveals that this is the case only in young democracies and
it is neither substantively nor statistically significant in any Western democracy but Switzerland.
For instance, in those countries and elections that we study below, the average within-country variance in official
voter turnout rates across election types reaches 12.5 percentage points (elections covered by the Making Electoral
Democracy Work project) and 26.1 percentage points (elections covered by the European Election Study).
stimulus elections” (or second-order) when compared to “high-stimulus” (first-order) national
elections (Reif and Schmitt 1980). The stakes are lower in the former. The policy competence
of municipal, regional and European authorities is usually limited and, in most cases, the central
authorities have a preponderant say on the most important policy issues and manage the
greatest share of the public budget (Moravscik 2002). Moreover, in European Parliament (EP)
elections the link between election outcomes and policy seems particularly tenuous (Hix 2006).
Partly as a result of lower stakes, subnational and supranational elections tend to generate less
party mobilisation and parties’ campaign expenditures are usually much lower than in national
elections (van Klingeren et al. 2015). At the same time, media coverage is considerably less
intense than in national contests (de Vreese, Lauf and Peter 2007). In short, subnational and
supranational elections provide citizens with an environment that is less conducive to electoral
participation: the stakes are less clear, information is scarcer and social pressure to participate
The less voter friendly environment of second-order elections should particularly
impact citizens who are less interested in politics. Their feeble interest translates into a weak
intrinsic motivation to vote and a lack of information. If mobilisation by politicians, the media
and/or the community is weak, such peripheral voters can be expected to turn out in smaller
numbers. The act of voting in second-order elections is more demanding: it is costlier and less
rewarding. Citizens with little interest in politics are likely to vote in the most important
elections while eschewing in large numbers other electoral contests where the incentives to
participate are weaker. This implies that the differences in voter turnout between first-order
and second-order elections should be predominantly driven by variable participation on the
part of individuals with low interest in politics. From this perspective, voter “turnout in any
specific [type of] election is largely a question of how many of the less interested […] make
the effort to vote.” (Campbell 1960: 400).
The degree of psychological involvement in politics can be measured in two ways. The
more traditional method relies on self-reported political interest. One of the strongest
individual-level predictors of voter turnout (Brady et al. 1995; Blais 2000), interest in politics
is typically acquired in the process of socialisation and remains largely stable throughout one’s
lifetime (Cambpell et al. 1960; Prior 2010). As a result, variation in turnout across election
types is unlikely to reflect variation in the degree of subjective political interest but rather
variation in the degree to which subjective interest matters for participation.
Another measure of psychological involvement in politics is political knowledge,
which has been found to foster voter turnout even after controlling for subjective political
interest (Coulson 1999; Milner 2002: 43-44; Larcinese 2007; Ondercin and Jones-White 2011).
This measure is different in nature since it is typically more context specific. Less politicised
citizens may be relatively knowledgeable about national-level political actors, thanks to
exposure to nationwide media, and yet show a sketchier grasp of subnational or supranational
politics. Such variation in political knowledge across levels of government reflects at least two
attributes that matter for voter turnout: the level of attention paid to the given political arena at
a particular moment of time and the availability of information. In contrast to subjective
political interest, variation in voter turnout is likely to be driven by variation in the amount of
political knowledge and less so by variation in the effect of political knowledge.
The scientific literature consistently shows that, despite the disappearance of gender
gaps in turnout in national elections, when asked in surveys, women express lower levels of
political interest and give fewer correct answers to political knowledge questions than men
(Inglehart 1981; Verba et al. 1997, Burns et al. 2001; Gidengil et al. 2004; Kittilson and
Schwindt-Bayer 2010; Fraile 2014, Dassonneville and McAllister 2018). Although this may be
partially due to a gender bias in measurement
and does not necessarily apply to all
subdimensions of political interest (Campbell and Winters 2008; Coffee 2013), it is widely
held that these differences are generally nonetheless real and originate in the process of political
socialisation (Bennett and Bennett 1989; Verba, Burns and Schlozman 1997: 1070; Norris
2007: 733-735). This implies that women are overrepresented among peripheral voters who
have a lower propensity to vote in the absence of a strong stimulus (see also Harell 2009).
Consequently, although women may vote at the same rates as men in national electoral
contests, where the environment is the most conducive to participation, a gender gap can be
expected in second-order elections, where psychological engagement in politics matters more.
To some extent, voting in second-order elections is likely to resemble engagement in other,
more demanding, forms of political participation such as attending public meetings or joining
in a protest, where women still participate less than men largely due to their lower levels of
political interest and political knowledge (Verba, Burns and Schlozman 1997; Ondercin and
However, not all non-national elections are second-order and not all national elections
are first-order (see Schakel and Jeffery 2013). For instance, in subnational entities with a strong
regional identity, such as Catalonia or Quebec, the question of regional autonomy and the
presence of regional parties may increase the perceived stakes and hence interest in the regional
Mondak and Anderson (2004) estimate that as much as half of the gender gap in political knowledge is
attributable to sex differences in the propensity to guess. Other scholars have suggested that conventional
measures systematically underestimate women’s political knowledge because they implicitly limit the scope of
politics to the traditional arenas of electoral and legislative politics (Smiley 1999). However, even if political
knowledge questions are subject to these gender biases, this does not undercut their usefulness for predicting voter
turnout since, to some extent, voting is similar to answering a knowledge-based question: if women are unsure
about their preference, they may decide not to vote.
election. This will be reflected in a voter turnout that is as high or even higher than in national
elections. In such cases, regional elections qualify as first-order elections and we can expect
women to be as likely to vote as men. Conversely, some national contests may be much less
mobilising because their outcomes are not consequential. This is exemplified by Switzerland,
where the most important decisions are taken through referenda and government composition
is determined by a long-term power-sharing deal (Franklin 2004, Blais 2014). In such second-
order national elections, peripheral voters participate less than core voters and so we can expect
women to vote at lower levels than men, since they are more likely to be peripheral voters.
Hypothesis 1: Women vote at the same rates as men in first-order elections but they
vote at lower rates in second-order elections.
Hypothesis 2: Women vote less than men in second-order elections because they are
less psychologically engaged in politics.
Psychological engagement in politics is not the only potential source of sex differences
in voter turnout. The scientific literature advances at least three other types of explanation for
why women may be less politically active than men (Norris 2007). They pertain to gendered
political practices and institutions, a lack of resources, and fewer opportunities to be mobilised.
However, in most cases, these explanations are unlikely to account for the sex difference in
voter turnout between first-order and second-order elections.
From the perspective of the current research question, the first factor is of little help. In
the case of EP elections, the gender gap is unlikely to be due to male-dominated institutions or
practices. In all EU member states, EP elections are run using proportional representation,
which tends to be favourable to female candidates (Kittilson and Schwindt-Bayer 2010).
Accordingly, EP elections have systematically yielded a significantly higher share of female
deputies than national elections and “the European Parliament is heralded as one of the most
gender-equal elected bodies in the world” (Fortin-Rittberger and Rittberger 2014: 1).
The potential for sex differences in resources to explain the variation in the gender gap
in voter turnout across election types is also limited. Burns et al. (2001) find that the availability
of free time has no effect on the political participation of women or men, and that women who
work full-time and have children are actually more likely to participate than their stay-at-home
counterparts. Similarly, even though education, along with age, is the most important
individual predictor of voter turnout (Blais 2000: 52), in most developed countries women
today have higher average levels of educational attainment than men (cf. Pekkarinen 2012).
Moreover, a seminal study of the resource model of political participation concludes that
“political interest is much more important than resources if our main project is to explain voting
turnout” (Brady at al. 1995: 283).
The third factor is party mobilisation. When media coverage is sparser, as it typically
is in second-order elections, the burden of direct mobilisation falls upon political parties. Given
that women are less likely than men to belong to a political party or to be involved in political
networks (Norris 2007), they may be more difficult to reach. As a result, men may be more
likely than women to be contacted by canvassers in second-order elections. In the same vein,
women are less likely to experience indirect mobilisation, given the gendered nature of
associational involvement and social networks. Women are more likely to belong to
associations that relate to the traditional domains of the home and community and that reinforce
their private sphere roles (Norris and Inglehart 2005). Importantly, there is less discussion of
politics in these female-dominated associations (Gidengil et al. 2005) and so women are less
likely to be exposed to mobilising influences. Similarly, women’s informal social networks
tend to be less diverse than men’s (Gidengil et al. 2006). Whether these differences are
sufficiently strong to account for a gender gap in second-order elections is an open question.
Data and Methods
The following empirical analyses employ two datasets. The first comes from the
Making Electoral Democracy Work project (MEDW, Blais 2010; Stephenson et al. 2017). It is
particularly suitable for tackling our main research question since, in contrast to most available
alternatives, it provides the same kind of information about voter behaviour in elections at
different levels of government (national, subnational and supranational) and thus allows for
comparisons across election types. It contains survey data from five countries (Canada, France,
Germany, Spain and Switzerland), which were gathered between 2010 and 2015. In each
country, two sub-national administrative units (referred to here as ‘regions’) were studied.
Altogether, we use twenty-three samples: nine from national elections, nine from subnational
elections and five from supranational elections.
Another major advantage of these data is that
the surveys are short two-wave panels, conducted about one week before and one week after
the given election. All the independent variables tested below were measured before the
election whereas the dependent variable was measured afterwards. This allows for much
greater confidence in the direction of causality. The other dataset is the 2014 European Election
Study (EES) (Schmitt et al. 2016), which enables us to test our findings on a larger pool of 24
The samples are from Ontario and Quebec in Canada, Île de France and Provence in France, Lower Saxony in
Germany, Madrid and Catalonia in Spain; and Lucerne and Zürich in Switzerland.
The complete list can be found in the Electronic Appendix.
EU member states.
Another key advantage of the ESS data is that they allow for comparing
the same individuals in both national and European elections (but, importantly, not subnational
We distinguish five types of elections: national elections, subnational elections in
Catalonia and Quebec, subnational elections in other regions, supranational elections, and
As discussed above, we expect voters to consider the first two types of
electoral contests (national elections, and subnational elections in Catalonia and Québec) to be
more important than the other types. This is confirmed empirically in the MEDW surveys,
which asked citizens to rate the importance of the current election on a 0 to 10 scale. Table 1
shows that subnational elections in Catalonia and Quebec are seen to be as important as national
elections (i.e. the reference category). In contrast, the perceived importance of the other three
types is both substantively and significantly lower. Consequently, we refer to the first two
election types as “first-order” and the other three as “second-order”. We expect women to vote
at lower rates than men in second-order elections since participation in less important elections
is costlier and less gratifying for those who are less psychologically engaged in politics. It
should be emphasized that our theory does not postulate that perceived importance differs by
sex, and we find no sex difference.
The 24 countries are Austria, Bulgaria, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Germany,
Hungary, Ireland, Italy, Latvia, Lithuania, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia,
Spain, Sweden, and the United Kingdom. Countries that use compulsory voting are excluded.
Drawing on earlier research (Blais 2014), we expect national elections in Switzerland to be less relevant than in
other countries and we therefore pool them together with Swiss subnational elections. This analytical choice is
further validated by Table EA 6 in the Electronic Appendix, which shows that the estimated gender gap is of
comparable magnitude in both election types.
We reran the analysis presented in Table 1 including an interaction between sex and election types (see Table
EA 5). The results reveal that women rate all elections as less important than men do. Women may be more
We first measure the impact of sex on voter turnout in each type of election (Hypothesis
Then, we test Hypothesis 2 via a non-linear decomposition of the effect of sex. Finally,
we explore in greater detail the variation in the explanatory factors by sex and election type.
Table 1 Perceived Importance of Election by Election Type (MEDW Data)
Catalonia & Québec:
Note: MEDW data. OLS Regression. The dependent variable is importance of the current election (scale ranged
from 0 to 10). The reference category is National Elections. The category Subnational does not include
subnational elections in Catalonia and Québec, which are a separate category. Robust standard errors in
parentheses. + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.
In the MEDW dataset, Interest is a self-reported rating of general political interest on a
scale from 0 (no interest in politics) to 10 (strong interest). To facilitate comparison with other
interested in local politics than men as some studies have found (Coffé 2013), but they do not rate local elections
as more important than men do or more important than national elections (in both cases, it is rather the opposite).
The data are weighted to correct for over-reporting and correspond to the real voter turnout rate. This weighting
has no impact on the substantive results of our study. More generally, it is true that a sex difference in over-
reporting could, in theory, affect our analysis. We cannot address this directly since we do not have a validated
measure of voter turnout at our disposal. However, previous studies of validated voter turnout (Karp and
Brockington 2005) and survey experiments (Morin-Chassé et al. 2017) did not find consistent or statistically
significant sex differences in over-reporting.
variables, we divide it by ten. Political knowledge is evaluated by asking respondents to match
pictures of party leaders competing in the election with their respective party names. In each
survey, there were typically 5 competing parties and 10 pictures (5 pictures of party leaders
and 5 pictures of other political personalities). The resulting variable Knowledge gives the
proportion of correct matches (between 0 and 1). In the EES data, Interest is a 4-point scale
recoded to vary between 0 and 1. The measure of political knowledge gives the share of correct
answers to two factual questions about the European Union (whether it is true that “Switzerland
is a member of the EU” and “Each Member State elects the same number of representatives to
the European Parliament”) and two questions about the national political system (whether
“there are [150 % of the correct number] members in the lower house of national parliament”
and whether “[name of the head of government] belongs to [the name of the correct party]”).
When compared to the MEDW measure, the EES indicators tap knowledge that is less election-
specific and more likely to be acquired in school, thus providing a weaker measure of
psychological involvement in politics.
Controls are included for resources and exposure to mobilisation in order to test for
alternative explanations, as discussed in the previous section. In the MEDW data, resources
include Age (a continuous variable), Tertiary education (a dummy variable) and the fact of
having Children (a dummy variable).
Mobilisation corresponds to dummies that ascertain
whether the respondent was contacted by a political party (Party) or was encouraged to vote
for a political party by a friend (Friend), family member (Family) or spouse (Spouse). In the
EES data, in addition to the variables Age, Tertiary education and Children, which are
measured exactly the same as in the MEDW data, resources include dummy variables
pertaining to employment: Employed and Self-employed (the reference category being those
Unlike the EES data, the MEDW unfortunately does not offer variables pertaining to employment.
who declare no formal employment). The data also offer a ten-point social status scale (recoded
to vary between 0 and 1), where the extreme points correspond to the “lowest” and “highest”
levels in society, respectively. Mobilisation is operationalized through two dummies: Trade
union member and Party closeness. The latter is coded as 1 for those who report feeling close
to a political party.
Controls are also included for attachment to the given level of government, using the
extent to which citizens consider voting in the current election to be a civic duty on a scale
from 0 to 1 in the case of the MEDW data
and two dummy variables indicating attachment
to the European and national levels of government, respectively, in the case of the EES data.
More details and descriptive statistics are presented in the Appendix.
Election Type and the Gender Gap in Voter Turnout
To test our first hypothesis regarding the presence of a gender gap in voting in second-
order elections, we run a logistic regression for each election type. The dependent variable is
reported turnout and the main independent variable is the dummy Female. All models also
include election-region (or, in the EES, election-country) dummy variables. Our interpretation
is based on the average marginal effects (AME). These are average changes in the probability
of voting resulting from a one-unit change in the independent variable and are estimated using
the observed values method (Mood 2010).
The results presented in Table 2 clearly validate our expectations. The analyses of the
MEDW data show that the effect of sex in first-order elections is substantively small and
The use of a more direct question on attachment (Q43A-D) yields substantively similar results but, since the
variable is available only for a subset of surveys, we opt for the civic duty question.
statistically insignificant. In contrast, women participate substantially less in all types of
second-order elections: subnational, supranational, and Swiss. In those electoral contests, the
gender gap in the predicted probability of voting reaches 8, 9, and 16 percentages points,
respectively. We show in the Electronic Appendix that these differences also largely hold when
the data are broken down by country or region.
The EES data, which allow us to compare the participation of the same individuals in
different contests, give a similar picture. In national elections, there is no sex difference in
participation, but the difference is both substantively and statistically significant in
supranational elections. The substantive magnitude of this difference is smaller than in the
MEDW data, which is not surprising given that EES data contain a broader sample of countries
across which sex inequality varies. We explore this variation in Figure 1, which plots the
estimates of the gender gap in voter turnout in national and supranational election by country.
In addition, the figure plots the sex difference in the drop-off between national and
supranational elections, i.e. the difference between the share of women and the share of men
who vote in national elections but do not vote in EP elections. Following our theory, we
distinguish two groups of countries: those with a large sex difference in political interest and
those with a small difference.
In the latter group, which includes mostly North European
countries, women seem to vote at slightly higher rates than men and there is typically little
See Table EA 6. The only exception is the Madrid region where no significant gender gap was recorded in
We use political interest since the subsequent analyses show that it accounts for a greater share of the gender
gap in voter turnout than political knowledge. The small sex difference group corresponds to the top 20 % of
countries with a substantively weak (<0.04) and, typically, statistically insignificant sex difference on the 0-1
scale of political interest.
difference between the national and supranational levels.
Likewise, the differences in the
national-supranational drop-off tend to be substantively small and statistically insignificant. By
contrast, in the former, more sex-differentiated group, women tend to vote at lower rates than
men in supranational elections but there is no consistent sex difference in national elections.
By the same token, the gender differences in drop-off rates tend to be larger and, in many cases,
statistically significant. This suggests that, in those countries in which women have less interest
in politics than their male counterparts, they vote at lower rates than men in supranational
Figure 1 Gender Gap in Voter Turnout by Country (EES data)
An important exception is Sweden where women vote at higher rates than men in supranational elections.
Table 2 Gender Gap in Voter Turnout by Election Type (Hypothesis 1)
Predicted voting rate of
Note: Logit Regression. Average marginal effects (AME) with robust standard errors in parentheses. + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.
Of course, we cannot draw far-reaching conclusions about individual cases since the
estimations plotted in Figure 1 suffer from noise due to sampling variance and relatively small
sample sizes. However, taken together, these patterns confirm that the aggregated results
presented in Table 2 are not driven by a few extreme country-case studies. They also provide
some preliminary evidence for Hypothesis 2, which we test with more advanced tools in the
Decomposing the Gender Gap in Voter Turnout in Second-Order Elections
To test Hypothesis 2, we employ an adaptation of the Blinder-Oaxaca decomposition
for non-linear regression models that was proposed by Fairlie (2005).
Decomposition is a
classic method for the analysis of political and economic inequalities in terms of sex or race
(Blinder 1973, Oaxaca 1973, Dow 2009, Kim 2010). Starting from a regression analysis, it
decomposes differences in mean outcomes across two groups into two parts: explained and
unexplained (see Equation 1). The explained part corresponds to group differences in the levels
of the independent variables and the unexplained part to group differences in the magnitude of
regression coefficients (as well as the constant and the contributions of unobserved variables,
see Fairlie 2005: 3). In the context of the current research question, the difference in mean
outcomes that we decompose is the difference between men’s and women’s average voting
rate as shown in Table 2. The explained part indicates how much of the gender gap in voter
turnout is due to unequal endowments in the factors we discussed above.
If Hypothesis 2 is
We executed the decomposition via the Stata package Fairlie (Jann 2008).
Following other studies using the decomposition methods (see Jones 1983, Fairlie 2005), we do not interpret
the unexplained part because of its ambiguity.
correct, sex differences in political interest and political knowledge account for most of the
Equation 1 The non-linear decomposition proposed by Fairlie 2005
Note: The M and W superscripts stand for Men and Women. The function F is logit.
The decomposition necessitates the choice of reference regression coefficients that are
applied to the explained part of the equation. Like Fairlie 2005, we opt for the regression
coefficients from a pooled regression model (including both male and female respondents)
but the alternative use of male () or female coefficients (yields very similar results.
The results of the decomposition are displayed in Tables 3 (MEDW data) and 4 (EES
They show that differences in endowments explain most of the gender gap for all three
types of second-order election. The share of the explained part varies from 68 % in Swiss
elections to 102 % in subnational elections. The latter share suggests that, if women and men
had exactly the same levels for all the independent variables included in the regression model,
women’s voting rates would be slightly higher than men’s in non-Swiss subnational elections.
Following Fairlie 2005 (p. 309, footnote 8; see also Jann 2008: 458), the pooled model used for extracting the
coefficients contains the group dummy (Female).
Note that, as the decomposition includes only individuals with non-missing values on all the independent
variables, the measured gender gap is slightly larger than in Table 2.
Table 3 The Non-Linear Decomposition of the Effect of Sex (MEDW)
Gender gap to explain
Individual contributions to the explained part
% of the
% of the
% of the
Interest in politics
Duty to vote
Region dummy 1
Region dummy 2
Region dummy 3
Region dummy 4
Note: Regional dummies are Lower Saxony (1), Ontario (2), Marseille (3), Madrid (4), and Paris (reference) in
the case of subnational elections; Provence (1), IDF (2), Lower Saxony (3), Madrid (4), and Catalonia (reference)
in the case of supranational elections; and Lucerne national (1), Lucerne subnational (2), Zürich national (3) and
Zürich subnational (reference) in the case of case Swiss elections. + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.
Table 4 The Non-Linear Decomposition of the Effect of Sex (EES Data)
Note: The reference category for country dummies is France. + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.
Importantly, the results strongly support Hypothesis 2. In all cases, interest in politics
and political knowledge account for at least two-thirds of the explained part. Interestingly, in
all types of elections in the MEDW data, the contributions of the two factors are comparable.
The only exception is the EES data, where knowledge of European politics explains much less
than interest in politics (the ratio is 18 % to 93 % of the explained part). This relatively weaker
Individual contributions to the explained part
% of the
Trade union Member
23 Country Dummies
contribution of political knowledge is consistent with the less election-specific nature of the
EES measure when compared to the MEDW measure (see Data and Methods).
With respect to the control variables, the only consistent and moderately strong
contribution comes from women’s tendency to consider voting in second-order elections as
less of a duty than men. It accounts for 9 % (Swiss elections) to 21 % (supranational elections)
of the explained part of the gender gap. The EES data nonetheless show that when perceptions
of different levels of government are measured on a more general level they have virtually no
explanatory power in terms of the gender gap in voter turnout.
To complement these analyses, we replicated logistic regressions of voter turnout from
Table 2 incorporating, in turn, the control variables, and political interest and political
knowledge as predictors. The results are presented in the Electronic Appendix (Tables EA 7
and 8) and, like the decomposition method, provide strong support for Hypothesis 2. In all
types of second-order elections, the gender gap in voter turnout remains substantively and
statistically significant in the presence of control variables, enabling us to rule out the
alternative explanations discussed above. In contrast, the inclusion of political interest and
political knowledge renders the gap negligible and statistically insignificant.
All in all, the results clearly validate the prediction that unequal psychological
engagement in politics is the major source of the gender gap in voter turnout in second-order
The only exception are Swiss elections where the gap is still significant but decreases from 11 to 5 percentages
points. This residual gap is probably due to the singularity of the history of female suffrage in Switzerland. As
Swiss women were enfranchised at the federal level only in 1971, it is possible that the effect of the lower degree
of psychological engagement in politics is reinforced by additional generational and cultural barriers to women’s
elections. In the next subsection, we explore in greater detail the distribution of psychological
engagement by sex and by election type.
Sources of the Gender Gap in Interest and Knowledge
On average, men are more interested and know more about politics than women do.
This is evident from the results presented in Table 5. Models 1 and 3 simply regress Interest
and Knowledge on sex and indicate that women’s scores are significantly lower in every case
in both the MEDW and EES data. While interest, as expected, does not vary by election type
(see Model 2), Model 4 (MEDW data) and the difference between Models 6a and 6b (EES
data) corroborate our expectations and show that the gender gap in knowledge is larger in
The question that remains to be answered is why the discrepancy in psychological
engagement in politics plays a smaller role in first-order than in second-order elections. The
level of interest in politics is not election-specific but our theory predicts that this factor matters
more in less mobilising contexts. When there are fewer incentives to vote, the impact of
political interest (or lack of it) can be expected to increase. To demonstrate this empirically,
we ran an additional analysis with the MEDW data, pooling all elections. We added a dummy
variable for second-order elections and an interaction between the dummy and Interest (see
Model 1 in Table EA 4 in the Electronic Appendix).
Table 5 Levels of Interest in Politics and Political Knowledge by Sex
Female * Subnational in
Catalonia and Québec
Female * Subnational
Female * Supranational
Female * Swiss
Subnational in Catalonia
Note: OLS Regression. Robust standard errors in parentheses. + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.
Figure 2 Effect of Political Interest and Knowledge by Election Type
The left part of Figure 2 shows how the predicted probability of voting in first-order
and second-order elections varies as a function of interest in politics.
Fully in line with the
theory first advanced by Campbell in 1960 (see above), the probability of voting is the same
regardless of election type (slightly below 0.7) for high levels of interest in politics (score 1 on
the 0-1 scale). Moreover, as predicted, the weaker the interest, the larger is the difference in
the probability of voting in first-order and second-order elections. For the lowest values of self-
declared interest (0), the probability of voting is 48 percent in first-order elections but only 33
percent in second-order elections. The effect of political interest (the difference between 0 and
1) is thus much stronger in second-order elections (37 percentage points) than in first-order
elections (22 percentage points). This explains why the gender gap in interest matters much
more for voter turnout in second-order than in first-order elections. As an aside, this finding
The figure draws on average marginal effects, i.e. the predicted probabilities are calculated for all individuals
keeping these individuals’ characteristics at their original values except for Interest (in Model 1), Knowledge (in
Model 2), and Second-order elections.
has implications for other disadvantaged groups that are less interested in politics such as young
or less educated adults.
In contrast to interest in politics and, as our theory predicts, political knowledge is an
election-specific factor. We do not expect its effect to vary across election types and indeed it
does not, as the right part of Figure 2 reveals (see also Model 2 in Table EA4). However, as
we have noted above, the gender gap in knowledge is larger in second-order elections. Men are
better able than women to identify the political leaders who run in the most important elections,
but the discrepancy is even larger in less important contests.
The findings confirm the conventional wisdom in political science that there is no
gender gap in voter turnout in national elections in Western democracies. There is also no
evidence of sex differences in subnational elections that are seen to be important by voters. The
gap, however, still exists in second-order elections: subnational and supranational elections in
most countries and regions, and both national and cantonal elections in Switzerland. The
optimistic assessment that sex has no effect on turnout, dominant in the scientific literature,
needs to be revisited. We have identified two reasons why women participate at lower rates
than men in these contests.
First, as has been well documented by earlier work, women are less interested in
politics. Drawing on classic political science theories, we show that this sex difference matters
In the Electronic Appendix, we present additional analyses to show that young adults, parents and less educated
respondents are more likely to drop out of voting in supranational elections and that this effect is diminished once
political interest is controlled for (see Table EA 13).
more in less mobilising contexts. In elections of lower importance, the fact of being
(un)interested in politics matters more. With respect to subnational elections, our results show
that even if women are more interested in local issues than men as earlier studies suggest, this
does not compensate for the lack of interest in politics in general and does not prevent a gender
bias in turnout.
Second, as in the case of political interest, there seems to be a persistent sex
difference in political knowledge that is detrimental to women’s electoral participation. This
difference is amplified in less important electoral arenas. Controlling for these two factors
effectively eliminates (or, in the Swiss case, substantially reduces) the gender gap in voting in
The best way to eliminate the gender bias in second-order elections is to reduce the sex
differences in political interest and political knowledge. This may prove particularly difficult
given their tenacity despite the reduction of gender inequalities in other areas. As noted above,
their roots are typically seen in the process of political socialisation. This means that change
has to come essentially through generational replacement as a result of a transformation in the
way that young people construct their political identity and learn about politics. Such a
transformation may necessitate a targeted action plan by public authorities, whose benefits may
extend well beyond full gender equality at the ballot box.
Beauregard, Katrine (2014). ‘Gender, political participation and electoral systems: A cross-
national analysis’, European Journal of Political Research 53:3, 617–634.
As with supranational elections, the presence and degree of the gender gap in subnational elections is likely to
vary cross-nationally as a function of gender differences in political interest and political knowledge (and also
these elections’ importance). For instance, in contrast to the countries studied here, Childs (2004) reports no
gender gap in voter turnout in Great Britain’s local elections.
Bennett, Linda L. M. and Stephen Earl Bennett (1989). ‘Enduring Gender Differences in
Political Interest The Impact of Socialization and Political Dispositions’, American
Politics Quarterly 17:1, 105–122.
Blais, André (2000). To Vote or Not to Vote?: The Merits and Limits of Rational Choice
Theory, Pittsburgh: University of Pittsburgh Press.
Blais, André (2007). ‘Turnout in Elections’, in Hans-Dieter Klingemann and Russell J. Dalton
(eds). The Oxford Handbook of Political Behavior. Oxford: Oxford University Press,
Blais, André (2010). ‘Making Electoral Democracy Work’, Electoral Studies 29:1, 169–170.
Blais, André (2014). ‘Why is Turnout So Low in Switzerland? Comparing the Attitudes of
Swiss and German Citizens Towards Electoral Democracy’, Swiss Political Science
Review 20:4, 520–528.
Blais, André and Filip Kostelka (2015). ‘The Decision to Vote or Abstain in the 2014 European
Elections’, European Journal of Social Sciences :53-1, 79–94.
Blinder, Alan (1973). ‘Wage Discrimination: Reduced Form and Structural Estimates’, Journal
of Human Resources 8:4, 436–455.
Brady, Henry E., Sidney Verba, and Kay Lehman Schlozman (1995). ‘Beyond SES: A
Resource Model of Political Participation’, American Political Science Review, 89:2,
Burns, Nancy, Kay Lehman Schlozman and Sidney Verba (2009). The Private Roots of Public
Action, Harvard University Press.
Campbell, Angus (1960). ‘Surge and Decline: A Study of Electoral Change’, Public Opinion
Quarterly 24:3, 397–418.
Campbell, Angus, Philip E. Converse, Warren E. Miller and Donald E. Stokes (1960). The
American voter, Chicago: The University of Chicago Press.
Campbell, Rosie and Kristi Winters (2008). ‘Understanding Men’s and Women’s Political
Interests: Evidence from a Study of Gendered Political Attitudes’, Journal of Elections,
Public Opinion and Parties 18:1, 53–74.
Childs, Sarah (2004). ‘A British Gender Gap? Gender and Political Participation’, The Political
Quarterly 75:4, 422–424.
Coffé, Hilde (2013). ‘Women Stay Local, Men Go National and Global? Gender Differences
in Political Interest’, Sex Roles 69:5-6, 323–338.
Coffé, Hilde and Catherine Bolzendahl (2010). ‘Same Game, Different Rules? Gender
Differences in Political Participation’, Sex Roles 62:5-6, 318–333.
Córdova, Abby and Gabriela Rangel (2017). ‘Addressing the Gender Gap The Effect of
Compulsory Voting on Women’s Electoral Engagement’, Comparative Political Studies
Coulson, Tony (1999). ‘Voter Turnout in Canada: Findings from the 1997 Canadian Election
Study’, Electoral Insight - Participation In The Electoral Process, 1:2, 18–22.
Dassonneville, Ruth and Ian McAllister (2018). ‘Gender, Political Knowledge, and Descriptive
Representation: The Impact of Long-Term Socialization’, American Journal of Political
Science 62:2, 249–265.
de Vreese, Claes H., Edmund Lauf and Peter Jochen (2007). ‘The media and European
Parliament elections: Second-rate coverage of a second-order event?’, in Wouter van der
Brug and Cees van der Eijk (eds). European Elections & Domestic Politics: Lessons from
the Past and Scenarios for the Future. Notre Dame: University of Notre Dame Press,
Dow, Jay K. (2009). ‘Gender Differences in Political Knowledge: Distinguishing
Characteristics-Based and Returns-Based Differences’, Political Behavior 31:1, 117–
Engeli, Isabelle, Thanh-Huyen Ballmer-Cao, and Marco Giugni (2006). ‘Gender Gap and
Turnout in the 2003 Federal Elections’, Swiss Political Science Review, 12:4, 217–242.
Fairlie, Robert W. (2005). ‘An extension of the Blinder-Oaxaca decomposition technique to
logit and probit models’, Journal of Economic and Social Measurement 30:4, 305–316.
Fortin-Rittberger, Jessica and Berthold Rittberger (2014). ‘Do electoral rules matter?
Explaining national differences in women’s representation in the European Parliament’,
European Union Politics 15:4, 496–520.
Fraile, Marta (2014). ‘Do Women Know Less About Politics Than Men? The Gender Gap in
Political Knowledge in Europe’, Social Politics: International Studies in Gender, State
& Society 21:2, 261–289.
Franklin, Mark N. (2004). Voter Turnout and the Dynamics of Electoral Competition in
Established Democracies since 1945, Cambridge University Press.
Gidengil, Elisabeth, André Blais, Neil Nevitte and Richard Nadeau (2004). Citizens,
Vancouver: UBC Press.
Gidengil, Elisabeth, Elisabeth Goodyear-Grant, Neil Nevitte and André Blais (2005). ‘Gender,
Knowledge, and Social Capital’, in Brenda O’Neill and Elisabeth Gidengil (eds). Gender
and Social Capital. Routledge, 241–272.
Gosnell, Harold F. (1930). Why Europe votes., Chicago, Ill.: The University of Chicago Press.
Harell, Allison (2009). ‘Equal Participation but Separate Paths?: Women’s Social Capital and
Turnout’, Journal of Women, Politics & Policy 30:1, 1–22.
Hix, Simon (2006). ‘Why the EU needs (Left-Right) Politics? Policy Reform and
Accountability are Impossible without It’, Notre Europe Policy Paper, 19, 1–27.
Inglehart, Margaret L. (1981). ‘Political Interest in West European Women: An Historical and
Empirical Comparative Analysis’, Comparative Political Studies 14:3, 299–326.
Inglehart, Ronald and Pippa Norris (2003). Rising Tide: Gender Equality and Cultural Change
Around the World, Cambridge University Press.
Jann, B. (2008). ‘The Blinder-Oaxaca decomposition for linear regression models’, Stata
Journal 8:4, 453–479.
Jones, F. L. (1983). ‘On Decomposing the Wage Gap: A Critical Comment on Blinder’s
Method’, Journal of Human Resources 18:1, 126–130.
Karp, Jeffrey A. and David Brockington (2005). ‘Social Desirability and Response Validity: A
Comparative Analysis of Overreporting Voter Turnout in Five Countries’, Journal of
Politics 67:3, 825–840.
Kim, Chang Hwan (2010). ‘Decomposing the Change in the Wage Gap Between White and
Black Men Over Time, 1980-2005: An Extension of the Blinder-Oaxaca Decomposition
Method’, Sociological Methods & Research 38:4, 619–651.
Kittilson, Miki Caul and Leslie Schwindt-Bayer (2010). ‘Engaging Citizens: The Role of
Power-Sharing Institutions’, The Journal of Politics 72:4, 990–1002.
Larcinese, Valentino (2007). ‘Does political knowledge increase turnout? Evidence from the
1997 British general election’, Public Choice 131:3-4, 387–411.
Mayer, Nonna (2010). Sociologie des comportements politiques, Paris: Armand Colin.
Milner, Henry (2002). Civic Literacy: How Informed Citizens Make Democracy Work,
Mondak, Jeffery J. and Mary R. Anderson (2004). ‘The Knowledge Gap: A Reexamination of
Gender-Based Differences in Political Knowledge’, The Journal of Politics 66:2, 492–
Mood, Carina (2010). ‘Logistic Regression: Why We Cannot Do What We Think We Can Do,
and What We Can Do About It’, European Sociological Review 26:1, 67–82.
Morin-Chassé, Alexandre, Damien Bol, Laura B. Stephenson, and Simon Labbé St-Vincent
(2017). ‘How to Survey About Electoral Turnout? The Efficacy of the Face-Saving
Response Items in 19 Different Contexts’, Political Science Research and Methods, 5:3,
Norris, Pippa (2002). ‘Women’s Power at the Ballot Box’, in Rafael López Pintor and Maria
Gratschew (eds). Voter Turnout from 1945 to 2000: A Global Report on Political
Participation. IDEA, 95–104.
Norris, Pippa (2007). ‘New Feminist Challenges to the Study of Political Engagement’, in
Russell J. Dalton and Hans‐Dieter Klingemann (eds). The Oxford Handbook of Political
Behavior. Oxford University Press, 724–743.
Norris, Pippa and Ronald Inglehart (2005). ‘Gendering Social Capital: Bowling in Women’s
Leagues?’, in Brenda O’Neill and Elisabeth Gidengil (eds). Gender and Social Capital.
Oaxaca, Ronald (1973). ‘Male-Female Wage Differentials in Urban Labor Markets’,
International Economic Review 14:3, 693–709.
Ondercin, Heather L., and Daniel Jones-White (2011). ‘Gender Jeopardy: What is the Impact
of Gender Differences in Political Knowledge on Political Participation?’, Social Science
Quarterly, 92:3, 675–694.
Pekkarinen, Tuomas (2012). ‘Gender Differences in Education’, Nordic Economic Policy
Review 1, 165–196.
Prior, Markus (2010). ‘You’ve Either Got It or You Don’t? The Stability of Political Interest
over the Life Cycle’, The Journal of Politics, 72:3, 747–766.
Reif, Karlheinz and Hermann Schmitt (1980). ‘Nine Second-Order National Elections – a
Conceptual Framework for the Analysis of European Election Results’, European
Journal of Political Research 8:1, 3–44.
Schakel, Arjan H., and Charlie Jeffery (2013). ‘Are Regional Elections really “Second-Order”
Elections?’, Regional Studies, 47:3, 323–341.
Schmitt, Hermann; Hobolt, Sara B.; Popa, Sebastian A.; Teperoglou, Eftichia; European
Parliament, Directorate-General for Communication, Public Monitoring Unit (2016):
European Parliament Election Study 2014, Voter Study, First Post-Election Survey.
GESIS Data Archive, Cologne. ZA5160 Data file Version 4.0.0, doi:10.4232/1.12628
Smiley, Marion (1999). ‘Democratic Citizenship: A Question of Competence?’, in Stephen L.
Elkin and Karol Edward Soltan (eds). Citizen Competence and Democratic Institutions.
University Park: Penn State Press, 371–383.
Söderlund, Peter, Hanna Wass and André Blais (2011). ‘The impact of motivational and
contextual factors on turnout in first- and second-order elections’, Electoral Studies 30:4,
Stadelmann-Steffen, Isabelle, and Daniela Koller (2014). ‘What Type of Resources?
Household Effects and Female Electoral Participation’, Swiss Political Science Review,
Stephenson, Laura B., André Blais, Damien Bol and Filip Kostelka (2017). Making Electoral
Democracy Work. V1, 2017, Harvard Dataverse, available at
Stolle, Dietlind and Elisabeth Gidengil (2010). “What Do Women Really Know? A Gendered
Analysis of Varieties of Political Knowledge,” Perspectives on Politics 8:1, 93–109.
Tingsten, Herbert (1963). Political behavior: studies in election statistics, Bedminster Press.
van Klingeren, M., M. Orozco, J. Spanje and Claes de Vreese (2015). Party Financing and
Referendum Campaigns in EU Member States, Brussels: European Parliament.
Verba, Sidney, Nancy Burns and Kay Lehman Schlozman (1997). ‘Knowing and Caring about
Politics: Gender and Political Engagement’, The Journal of Politics 59:4, 1051–1072.
Verba, Sidney, Norman H. Nie and Jae-on Kim (1978). Participation and Political Equality:
A Seven-Nation Comparison, University of Chicago Press.