One of the most intriguing stylized facts in democratic politics is the presence of an ideological sex gap whereby women support left-wing parties more than men do (Chaney, Alvarez, and Nagler Reference Chaney, Alvarez and Nagler1998; Edlund and Pande Reference Edlund and Pande2002; Manza and Brooks Reference Manza and Brooks1998; Shapiro and Mahajan Reference Shapiro and Mahajan1986). This is taken to reflect the existence of sex differences in policy preferences: men and women have systematically different ideas with respect to what public policies ought to be implemented. Sex differences in policy preferences have been identified among the electorate (Alvarez and McCaffery Reference Alvarez and McCaffery2003; Edlund and Pande Reference Edlund and Pande2002; Funk and Gathmann Reference Funk and Gathmann2015; Thomas and Welch Reference Thomas and Welch1991), among public officials (Dolan Reference Dolan2000, Reference Dolan2002), and among politicians (Chattopadhyay and Duflo Reference Chattopadhyay and Duflo2004; Svaleryd Reference Svaleryd2009; Swers Reference Swers1998).
Just as left-wing preferences among voting women translate into a sex gap in voting, the prevalence of left-wing sympathies among female politicians explains why left-wing parties find it easier to recruit female candidates and to better guarantee descriptive representation. Recent evidence has shown that this sorting effect, whereby more women join (or vote for) left-oriented parties, is incomplete at best. Indeed, Webb and Childs (Reference Webb and Childs2012), Childs and Webb (Reference Childs and Webb2012), and Campbell and Childs (Reference Campbell, Childs, Celis and Childs2014, Reference Campbell and Childs2015) identify a significant within-party sex gap among U.K. Conservatives: female politicians and voters tend to take more left-wing positions on economic issues than their male colleagues. Celis, Roggemans, and Spruyt (Reference Celis, Roggemans, Spruyt, Spruyt and Siongers2014) find within-party sex gaps when studying voters’ feminist attitudes in Flanders. The existence of a within-party sex gap has by no means reached the status of a “stylized fact.” Still, its relevance is beyond doubt. It implies that parties’ positions and policies can be expected to depend on their internal sex balance. Moreover, from an electoral perspective, sex cues provide voters with low-cost information about representatives’ policy positions (McDermott Reference McDermott1998). Fulton and Ondercin (Reference Fulton and Ondercin2013) show how the sex of a candidate conveys information about his or her ideological positioning and how voters use this political sex stereotype to make swifter judgments.
The aim of the current article is to test for the prevalence of a within-party sex gap in a wider context than the existing literature. We do so in two ways. First, we review earlier literature on politicians’ policy preferences. While the focus of those studies was not on the identification of within-party sex gaps (such gaps were typically not even mentioned), several authors have implicitly tested for these. We briefly summarize their results. Second, we test for the presence of a sex gap in (budgetary) preferences using data from a large-scale survey among Flemish council members. The survey was sent in 2013 by email to 6,448 members of municipal councils. We collected expenditure preferences of 1,055 local politicians from 278 (of 308) municipalities. Our approach offers some important advantages. First, our focus on public expenditures allows us to cover a broad range of political activities and domains. Second, the choice for local politicians provides us with a large sample size. Third, the multiparty context allows to fully exploit the variation in party ideology among respondents. Finally, the Flemish quota rules (imposing equal presence of both sexes in candidate lists in addition to a placement mandate for top positions) promote the career opportunities of female candidates and their presence among representatives.
We find, both in the earlier literature and in our own empirical case, convincing evidence of within-party sex gaps.Footnote 1 Female politicians tend to take more left-wing positions than their male colleagues within the same party. Our own empirical work—focusing on a wide array of competences—reveals left-oriented sex gaps within parties for spending on crime, culture, and welfare. We find that such a sex gap is by no means universal: for half (four out of eight) of the policy domains no sex gap is identified. For spending on environmental programs, we actually find an inverted gap, with women taking more right-wing positions. A conclusion and discussion are provided in the final section.
SEX DIFFERENCES IN POLICY PREFERENCES
It is widely documented that women are more likely to vote for left-wing parties. However, there is no generally accepted theory that explains this sex difference. It is a complex phenomenon that results from biological, social, psychological, and economic factors (Howell and Day Reference Howell and Day2000). The (rational choice) idea that differences in socioeconomic status cause diverging attitudes out of self-interest is reasonable. From such a perspective, increasing divorce rates and female employment are believed to have provoked a change in the necessities of women since the 1970s. As a consequence, their voting behavior moved toward the left (Campbell and Childs Reference Campbell, Childs, Celis and Childs2014; Manza and Brooks Reference Manza and Brooks1998).
Several authors find sex differences in policy preferences among the electorate, public officials, and politicians for both nonbudgetary and budgetary instruments. Shapiro and Mahajan (Reference Shapiro and Mahajan1986), for example, study the evolution of policy preferences of U.S. citizens from 1960 until 1980. They find widening sex gaps in preferences for protective policies, compassion issues, abortion, women's rights, and force and violence issues. Edlund and Pande (Reference Edlund and Pande2002) show that women prefer higher levels of government spending. Others study specific expenditure areas and find that women, compared with men, want more spending on education, medical issues, welfare, and the environment but less on crime and the economy (Funk and Gathmann Reference Funk and Gathmann2015; Gidengil et al. Reference Gidengil, Blais, Nadeau, Nevitte, Tremblay and Trimble2003). Alvarez and McCaffery (Reference Alvarez and McCaffery2003) show that female voters would allocate a budget surplus toward child care, while men prefer tax reductions.
Similar results to those found in the electorate have been obtained in studies among civil servants and politicians. Female executives in the United States express greater support for workplace reforms that benefit women and support higher expenditures on welfare and the environment (Dolan Reference Dolan2000, Reference Dolan2002). Likewise, female politicians attach more importance to policy issues concerning welfare, gender equality, and the environment, while male politicians prioritize policies on fighting crime, the economy, and taxes (Fredriksson and Wang Reference Fredriksson and Wang2011; Leal Reference Leal2005; McEvoy Reference McEvoy2016; Poggione Reference Poggione2004). Analyzing spending preferences, Svaleryd (Reference Svaleryd2009) finds that female politicians prefer more spending on child care and education (relative to spending on elderly care) than men.
WITHIN-PARTY SEX GAPS
The literature cited here describes across-party sex gaps among voters, civil servants, and politicians. Recently, a few authors have explicitly focused on sex gaps within political parties. They analyze the different attitudes of men and women toward gender (feminist) issues or preferences on government taxation and spending within a party. Relevant contributions that explicitly study within-party sex gaps focus on the voters and politicians of the Conservative Party in the United Kingdom. Both female voters and female politicians of this party are found to have more leftist tax and spending attitudes than males (Campbell and Childs Reference Campbell, Childs, Celis and Childs2014, Reference Campbell and Childs2015; Childs and Webb Reference Childs and Webb2012). Their attitudes toward feminist policy correspond more with left-wing policies (Childs and Webb Reference Childs and Webb2012). Politicians from the Conservative Party who position themselves (relatively) more to the left are more likely to take liberal stances on gender issues and encourage the substantive representation of women. Interestingly, female voters also express more left-wing attitudes, but they do not position themselves more toward the left on an ideological left-right scale. They might not consider themselves more leftist or might base their party choice on other policy domains. Indeed, few Conservative voters regard gender issues as a priority (Webb and Childs Reference Webb and Childs2012).
Celis, Roggemans, and Spruyt (Reference Celis, Roggemans, Spruyt, Spruyt and Siongers2014) analyze within-party sex gaps using a larger set of parties. They use a sample of young (18- to 30-year-old) Flemish voters and focus on the attitudes toward feminist principles and feminist policies. They find that women have higher levels of feminist attitudes than male members adhering to the same political party. The most sizable sex gaps are observed among supporters of right-wing parties. The observation that within a party female politicians are more toward the left than male politicians is an important one. Voters lack the time and resources to inform themselves on candidates and rely on cues and political stereotypes with respect to candidates’ demographics to make (fast) voting decisions. The sex of a candidate is used as a cue for their ideological position: women (both Republican and Democrat) are perceived as being more leftist by the electorate (Dancey and Sheagley Reference Dancey and Sheagley2013; Fulton and Ondercin Reference Fulton and Ondercin2013; McDermott Reference McDermott1998).
Although only a few studies explicitly analyze within-party sex gaps, many researchers have implicitly tested for these. Indeed, empirical specifications that reveal a systematic effect from a respondent's sex on his or her preferences while controlling for party affiliation reveal a within-party sex gap. The econometric issue under consideration is well illustrated in Stadelmann, Portmann, and Eichenberger (Reference Stadelmann, Portmann and Eichenberger2014). Analyzing Swiss parliamentarians’ voting behavior, they find that female representatives’ voting behavior is less in line with the median voter than that of their male colleagues. However, when including a variable for respondents’ party affiliation, this sex effect dampens. They conclude that “female and male representatives adhere equally close to the majority's preferences if party affiliations are taken into account. This suggests that observed sex differences with respect to the national majority of voters may be reduced to an ideological left-right dimension” (Stadelmann, Portmann, and Eichenberger Reference Stadelmann, Portmann and Eichenberger2014, 355).
Multivariate analyses of politicians’ preferences typically include a party (or ideological) dummy as well as a variable capturing respondents’ sex (Boles and Scheuer Reference Boles and Scheurer2007; Poggione Reference Poggione2004; Stucky, Miller, and Murphy Reference Stucky, Miller and Murphy2008; Svaleryd Reference Svaleryd2009; Swers Reference Swers1998; Vega and Firestone Reference Vega and Firestone1995). All these authors find significant party and sex effects. It is noteworthy that these effects are systematically of opposite sign. For example, politicians from right-wing parties consistently report preferences for lower levels of spending (Svaleryd Reference Svaleryd2009). Controlling for this party effect, women are found to prefer higher spending. This reflects a within-party sex gap, with women having more leftist preferences (see also Stadelmann, Portmann, and Eichenberger, Reference Stadelmann, Portmann and Eichenberger2014). Similarly, a within-party gap is observed for actions on women-related issues (Swers Reference Swers1998; Vega and Firestone Reference Vega and Firestone1995), social welfare (Poggione Reference Poggione2004), arts legislation (Boles and Scheuer Reference Boles and Scheurer2007), and gun control (Stucky, Miller, and Murphy Reference Stucky, Miller and Murphy2008). In all cases, controlling for their party affiliation, women are found to take positions that correspond with the (more) leftist parties.
The question of why women do not sort themselves into more liberal (left-wing) parties, or—as this has identical empirical implications—why men do not sort themselves into more conservative (right-wing) parties, is an intriguing on. Technically, we see two possible mechanisms.
The first is that issue salience differs systematically. If female candidates are more liberal for “most” but not all issues, they may join a more conservative party if that party's position is more in line with their preferences on issues they consider crucial. These women may then be observed to have more liberal preferences on issues they consider less important. This could explain the empirical observation of a within-party sex gap.
The second mechanism may find its origin in the specific quota rules that are in place. For instance, in Flanders (our case), party lists in municipal elections need to be balanced: the number of male and female candidates should be equal (with the provision that a difference of 1 is allowed if the list length is uneven). Consider—for the sake of exposition—a situation with two parties: a liberal (left-wing) and a conservative (right-wing) party. In a simple version of this model (Version 1), we assume all female politicians to be more liberal than male politicians. In other words, there would be no overlap in preferences along an ideological dimension. In a more realistic version of the model (Version 2) such an overlap exists but we assume a relative overrepresentation of women (men) among politicians with liberal (conservative) preferences. If—consider the first version of the model—politicians sort into one of two parties according to their “true” preferences, all female (male) politicians would join the more leftist (rightist) party. As quota rules impose that all parties should have an equal number of male and female candidates, this “pure sorting” cannot be an equilibrium. To reach the 50/50 distribution of candidates, right-wing (left-wing) parties will have to recruit from the pool of liberal women (conservative men). Such a recruitment driven by the specific quota rules will result in a within-party gender gap: by recruiting liberal women (conservative men), both parties create a sex gap in preferences, with women having more leftist preferences. These conclusions also apply to Version 2 of the model unless the initial distribution of preferences is characterized by sufficiently strong overlap. In that case, parties can recruit the (quota-) required number of politicians among candidates with preferences that are in line with those of the party.
While the previous quota effect refers to a demand factor (parties “having to look” for candidates of a given sex), quotas may also initiate supply effects. Indeed, individuals’ party choice may not be based on ideology only. It is conceivable that they also take into consideration career opportunities. Assuming this to be the case, it may be that female (male) politicians see more career opportunities in the party whose platform does not (“really”) correspond with their preferences. Politicians trying to increase their career odds may decide to join the party where they are “scarce”: men join leftist parties (where women are overrepresented), and women join rightist parties (where men are overrepresented). Again, the result is that within-party sex gaps in preferences originate with female (male) candidates being more liberal (conservative).
EMPIRICAL STRATEGY AND DATA
We study expenditure preferences of local politicians in Flanders. The number of municipalities (308) and the multiparty context allow us to test for within-party sex gaps on a large population of politicians from several parties. Moreover, the broad range of competences of Flemish local governments allows us to identify possible sex gaps in preferences in many policy areas within the same institutional context.
Our empirical strategy consists of identifying possible within-party sex gaps in respondents’ stated preferences for (increases in) spending in specific policy areas. Comparing these biases (if present) allows us to establish whether female politicians express more leftist (rightist) spending preferences than their male colleagues from the same party. Parties at the municipal level in Flanders are either local factions of parties active at the federal (and regional) level in Belgium, or they are independent local parties. The latter are excluded from further analysis. To test for within-party sex gaps, we ask respondents to assess the level of spending (in each of 10 categories) within their municipality. We test whether—controlling for their party affiliation—female and male politicians respond differently. Crucially, we check whether such biases have an ideological dimension, or, to be more precise, whether female preferences can be labeled more “leftist” (or, for that matter, more “rightist”) than preferences of male members in the same party.
METHOD
A large-scale survey of local politicians’ budgetary preferences in Flanders provides us with the necessary data. The survey was sent in 2013 by email to all members of municipal councils whose details could be found online (n = 6,848). Apart from questions on budgetary preferences, respondents were presented with questions on municipal politics and their feelings of involvement with their party, municipality, and so on. After excluding incomplete responses and those from local parties, we obtained expenditure preferences of 1,055 representatives from 278 (out of 308) municipalities (almost four per municipality). These representatives belong to one of the six local factions of federal parties, which can be ordered along a left-right axis as follows: Groen (green party), sp.a (socialist party), CD&V (Christian democrats), Open Vld (liberal party), N-VA (Flemish nationalists), and Vlaams Belang (extreme right) (Deschouwer, Verthé, and Rihoux Reference Deschouwer, Verthé and Rihoux2013).Footnote 2
Respondents were asked to evaluate the spending in their municipality. We asked, “Do you feel your municipality spends too much, the right amount or too little in the following domains?”
Fourteen spending areas were presented. For the empirical analysis, these were grouped into 10 dependent variables. Answers relating to the first six areas—spending on administration, environment, mobility, road safety, regional planning, and crime—provide us with a first set of six dependent variables. Each takes the value 1, 2, or 3, depending on whether the respondent thinks “too much,” “enough,” or “too little” is spent. Higher values for the dependent variable indicate that respondents support an increase in spending. Four other dependent variables were constructed by combining the answers on the remaining spending areas. Nursery school and primary school are merged into a variable education, elderly care and social housing into welfare, culture and library into culture, and tourism and employment and trade into economy. Each time, the dependent variable takes a value of 1 (3) if a respondent answers that “too much” (“too little”) is spent on both subquestions; it takes a value of 2 in all other cases.
Our central question—whether within-party sex gaps occur—is tested by comparing the responses of male and female representatives. Given the trichotomous nature of our dependent variable, we estimate by ordered logistic regression. The idea is that “actual” preferences are a linear function of a vector (X) including a variable capturing the respondent's sex and a set of control variables including his or her party affiliation. For each of the expenditure categories (j = 1, … 10) and leaving out the respondent-specific indices for convenience, we have the following:
$$h^* = \beta ^{\prime}X + \varepsilon $$Where β′ is a vector of coefficients and ε is the error term. The actual preferences—h*—of respondents are in effect unobserved. What is observed are respondents’ stated preferences. Three possible answer categories were presented, and so the actual response h is derived as follows:
$$\matrix{ {h = 1} {{\rm if}} \hskip-2pc{h^* \le \mu _1} \cr {h = {\rm} 2} {{\rm if}} {\mu _1 \lt \,h^* \le \mu _2} \cr {h = {\rm} 3} {{\rm if}} {\hskip-2.2pc \mu _2 \lt {\rm} h^*} \cr} $$The μ’s are unknown parameters estimated by the parameters of the basic model and define the outer limits of the responses. This leads to our estimation equations:
$$\!\!\eqalign{h = {\rm} \beta _{0,m} + \beta _1\; Woman + \beta _2 \; Party + \beta _3\; Opposition + \beta _4\; Age + \beta _5 \; Marital \cr \quad + {\rm} \beta _6\; Education{\rm} \; level + \beta _7\; Children \; f\hskip-1.3ptor{\rm} \; m = 1,{\rm}. {\rm}. {\rm}. {\rm}, {\rm} 278.} $$Politicians’ stated preferences depend on individual, party, and municipality characteristics. Our central focus is the presence of sex differences in preferences and, more specifically, how these relate to the preferences within a respondent's party. The sex dimension is captured by the variable woman, which takes a value of 1 for female respondents and 0 otherwise. The party affiliation is picked up by the inclusion of party, a vector of dummies for the local factions of the major nationally (and regionally) organized parties in Flanders (the center party CD&V is used as reference category).Footnote 3
The coefficients on the party dummies indicate whether party members in general are more (or less) supportive of increases in expenditures in the given category than members of the center party. A positive (negative) value for β1 indicates that female respondents tend to find—more than men within the same party—that spending in the policy domain should be increased (decreased). A within-party sex gap is revealed if β1 is significantly different from zero. To assess the ideological dimension of this gap, the sign of the coefficient is compared to the empirically identified ideological dimension of preferences (see following section). For areas in which more spending is classified as being a left-wing policy, a positive (negative) value for β1 reveals that female party members take a more left-wing position than their male colleagues. For areas in which more spending is classified as right wing, a negative (positive) value for β1 reveals a more right-wing position of female respondents.
Politicians’ assessments of the level of spending may depend on party characteristics. Ashworth and Heyndels (Reference Ashworth and Heyndels1997, Reference Ashworth and Heyndels2000) show that politicians assess policies differently depending on their party's status. Politicians belonging to opposition parties tend to be more critical of current policy. We capture the respondent's party's current political status through the variable opposition, which equals 1 when the party is part of the opposition (0 otherwise).
The other variables control for individual-, party-, and municipality-specific determinants of a respondent's preferences. At the individual level, we control for the respondent's age, marital status, education level, and whether he or she has children. Age is introduced as preferences may depend on (both life and political) experience, as well as to allow for the possibility that preferences change with getting older (Campbell and Childs Reference Campbell, Childs, Celis and Childs2014).Footnote 4 The respondent's marital status distinguishes between respondents living—either married or not—together with a partner (marital = 1) or not (marital = 0). There is evidence that being single affects spending preferences—if only because being single implies a higher risk of being in a more vulnerable economic situation (Campbell and Childs Reference Campbell, Childs, Celis and Childs2014; Chaney, Alvarez, and Nagler Reference Chaney, Alvarez and Nagler1988; Finseraas, Jakobsson, and Kotsadam Reference Finseraas, Jakobsson and Kotsadam2012). A respondent's level of schooling is measured by education level. This is a binary variable indicating whether the respondent has at least a bachelor's degree (education level = 1) or not (education level = 0). Education has been shown to influence policy attitudes toward the welfare state (Brooks and Svallfors Reference Brooks and Svallfors2010; Svallfors Reference Svallfors2011). Furthermore, we include the variable children, which indicates whether the respondent has children (children = 1) or not (children = 0). Having children may influence preferences (Campbell and Childs Reference Campbell, Childs, Celis and Childs2014) and generate a larger level of altruism (Schlozman et al. Reference Schlozman, Burns, Verba and Donahue1995).
Finally, budgetary preferences can be expected to depend on the budgetary, economic, sociological, and political context in the municipality. This is controlled for by introducing municipality-specific intercepts (β0,m).Footnote 5
RESULTS
Before presenting the regression results, we identify ideological dimensions for each of the spending categories. Rather than identifying these on an à priori basis, the survey data allow us to follow an empirical approach. By comparing the average responses on each of our 10 (sets of) questions for left- and right-wing politicians, we identify empirically whether a preference for “more spending” in a specific policy domain can be labeled “left wing” or “right wing.” In practical terms, we compare the average answers from politicians from the left-wing parties, Groen and sp.a, with the answers from politicians from the more right-wing parties, Open Vld, N-VA, and Vlaams Belang. If left-wing politicians report significantly (Wilcoxon Mann Whitney test; p < 5%) higher (lower) values for the dependent variable than right-wing politicians, we take this as empirical support for labeling “more spending” on the given policy domain as “leftist” (“rightist”). Table B in the appendix in the online version of this article summarizes the results.Footnote 6 More spending on administration, environment, culture, welfare, mobility, and education is identified as left wing. More spending on the economy and crime is labeled right wing. Preferences for road safety and regional planning have no statistically significant ideological dimension and are therefore excluded from further analyses.Footnote 7
In a second step, we compare the average responses of men and women. If the average response of women is larger (smaller) than that of men in the case of a policy labeled left wing (right wing), then we consider this as supporting the standard within-party sex bias hypothesis, with women being more left wing. The results are presented in Table C in the appendix. For example, as mentioned in the previous paragraph, “more spending on welfare” is identified as being a left-wing policy (see Table B in the appendix). We observe (in Table C in the appendix) that female—relative to male—respondents report a stronger preference for more spending in welfare. This reveals their more left-wing position.
The “ideological” radar (Figure 1) summarizes all this information. The center of the radar represents right-wing preferences. The labels on the arrows identify left-wing preferences. As mentioned, these correspond to a preference for more spending in all categories except for the economy and crime. In the latter cases, left-wing politicians prefer lower levels of spending than right-wing politicians. The dark-gray and light-gray shapes connect the stated preferences of men and women, respectively. The data in the radar indicate how much the average responses by men and women diverge from the overall average. If men and women report (on average) an identical preference for spending in a specific area, then the divergence from the average is 0 (in which case the light-gray and the dark-gray shapes coincide). For example, for spending on welfare, women tend to be more in support of more spending, as their average response (2,403) is 2.7% higher than the general average (2,341), while that for men (2,325) is 0.7% lower.
(Fe)male politicians’ preferences by policy area.
The figure reveals that in six out of eight cases, women have a stronger preference for left policies: the light-gray shape lies more toward the outer edge than the dark-gray shape. More precisely, the radar shows that women are to the left of men for spending on administration, environment, culture, education, welfare, and crime. Still, the leftist position of women is not observed in all cases. When it comes to spending on mobility and the economy, women are more right wing than men.
Figure 1 also gives a first indication of how much preferences differ by sex. For example, while sex differences are limited for spending areas as the economy and administration, they are much more pronounced for welfare, crime, and especially culture, where the light-gray and dark-gray lines are far apart.
A more systematic analysis of within-party sex gaps is done through the ordered logit regressions, results of which are summarized in Table 1 and Table 2. We present two sets of results. The first set (Table 1) gives an overview of the regressions including all individual, political, and contextual (municipality-specific dummy) control variables mentioned earlier. In Table 2, only the variables for which coefficients were significant at the 10% level are included using a stepwise procedure. In general, the models explain politicians’ preferences quite well. We find significant effects from the individual-, political-, and contextual-level variables.
Support for an increase in expenditures: Ordered logit (CD&V as reference category)
Standard errors in parentheses.
*** p < .01; ** p < .05; * p < .1.
Support for an increase in expenditures: Stepwise ordered logit (CD&V as reference category)
Standard errors in parentheses.
*** p < .01; ** p < .05; * p < .1.
Controlling for respondents’ party affiliation (party), a significant sex coefficient shows that men and women within the same party have systematically different preferences. Such sex gaps are identified in four expenditure areas: environment, crime, culture, and welfare. No systematic sex gap is apparent for spending on administration, mobility, education, and the economy. Importantly, these gaps are not homogeneous in terms of their ideological dimension. Indeed, while we find women to have more left-wing preferences than men when it comes to spending on crime, culture, and welfare, they report more right-wing preferences with respect to spending on the environment. Relative to men, women prefer lower expenditures on crime but more on culture and welfare. Table B in the appendix identifies such preferences as leftist. The former findings support the existing literature (reported in the section “Within-Party Sex Gaps”), which typically finds that women take more left-wing positions and that women have a more positive attitude (than men) toward welfare policy and culture but a more negative attitude toward military interventions and spending on defense (Funk and Gathmann Reference Funk and Gathmann2015 Leal Reference Leal2005; Swers Reference Swers1998). The marginal effects (presented in Table D and Table E in the appendix) show that women are 5% and 6% more likely than men within the same party to find more should be spent on culture and welfare, respectively, while they are 6% less likely to find that more should be spent on crime.
For spending on the environment, women take more right-wing positions (while more spending in this area is identified as a left-wing preference in Table B in the appendix, we find female representatives to be less supportive of an increase). This finding is unexpected. We observe a strong (and highly significant) ideological dimension whereby left-wing parties are (much) more in favor of spending in this category. We do, however, also observe a strong sex effect whereby women tend to take a more right-wing position. The marginal effects (in Table E in the appendix) demonstrate that women (relative to men within the same party) are 6% less likely to find more should be spent on the environment. Such a finding is in contrast to à priori expectations and to earlier evidence that reports women to have more positive attitudes toward environmental policies (Fredriksson and Wang Reference Fredriksson and Wang2011; Funk and Gathmann Reference Funk and Gathmann2015). Looking at the raw data, we find that these results are driven by the responses of CD&V politicians. In this party, men are significantly more likely to find that too little is spent on the environment. For the leftist parties the picture that emerges is more in line with the expectations: women, while insignificantly, find more than men that more should be spent on the environment. In the other parties (Open Vld, N-VA, and Vlaams Belang) spending preferences in this area are not significantly influenced by sex. Speculatively, the reason may be that environmental issues are highly salient for women (so that their party choice is driven by it) but less so for men. Further research is needed to clarify this.
The general conclusion is thus that within-party sex gaps in preferences exist in several—but not all—spending areas. Whereas the data lend support to the view in the literature that women tend to take more left-wing positions, this is not a “universal law.” First, we identify situations in which spending preferences do not have an ideological dimension. Second, even though the left-leaning sex gap is more prevalent (as is documented in the literature review), we find that women take more right-wing positions for spending on environmental issues.
Turning to the control variables in our estimation equation, we find, unsurprisingly, that the power to determine current policy also influences its assessment. Members of the opposition clearly report different spending preferences than majority members: compared with the latter, members of opposition parties tend to find that more spending is needed in all policy categories except for administration and mobility. Further, older respondents prefer higher spending on crime, welfare, and culture but want less spending on education. Living together with a partner makes respondents more willing to spend on education. Respondents with a higher education level prefer higher spending on the environment and lower spending on education and crime than less educated respondents. Finally, parents want more spending on administration.
The models as presented in the Table 1 and Table 2 test for the presence of a “general” within-party gap: controlling for their party, women are more liberal. Of course, it may be the case that such a gap is more prominent in some parties while being absent (or of the opposite direction) in other parties. To allow for this possibility, we reestimated the model, now including interaction effects between the sex variable woman and the party dummies (party). The results are reported in Table H in the appendix. While the pattern that is revealed is rather complex, some interesting conclusions at the party and spending area level can be drawn. First, we observe at least one statistically significant sex gap for each of the parties. Second, significant sex differences are more common among parties in the center of the ideological spectrum (notably CD&V and Open Vld).Footnote 8 Finally, significant sex gaps are identified in all spending areas except mobility and education. Especially the latter may be surprising, as previous literature has identified this area as “feminine.” Most importantly, Table H in the appendix documents that the significant within-party sex gap found in our main tables aggregates the effects in the individual parties.
CONCLUSION
The observation that more women vote for (affiliate with) left-wing parties was recently supplemented by the finding that female voters and politicians within parties tend to take more left-wing positions. While the contributing authors typically concentrate on a single policy area or one political party, we provide more extensive evidence of within-party sex gaps.
First, we show that within-party sex gaps have been implicitly studied in the literature. While the focus was not on the identification of such gaps, several authors have implicitly tested for them. The general conclusions are that within-party sex gaps are common and that women adhere to more left-wing positions. Second, we study a wide range of expenditure categories in a multiparty context to gain a broader view on within-party sex gaps. Using a large data set on stated preferences by 1,055 members of municipal councils in Flanders, we identify that—across parties—female representatives tend to have more leftist preferences than their male colleagues in six out of eight policy domains. Our results show within-party sex gaps in four policy domains: women (relative to men in the same party) want more spending on culture and welfare but less spending on crime and the environment. Importantly, we find that the ideological dimension of within-party sex gaps is not by definition left wing. We find that women within a party have more leftist preferences for spending on culture, crime, and welfare but more right-wing preferences for spending on the environment. Therefore, our analysis provides strong evidence of the existence of within-party sex gaps in a number of policy domains. At the same time, it shows that such a gap is absent in other domains. Crucially, our analysis suggests a more complex picture with women sometimes taking more left-wing and sometimes more right-wing positions than their male colleagues in the same party.
Our results highlight the existence of within-party sex gaps in Flemish municipalities. These results generalize the findings of Webb and Childs (Reference Webb and Childs2012), Childs and Webb (Reference Childs and Webb2012), and Campbell and Childs (Reference Campbell and Childs2015) studying sex differences in preferences among politicians, members, and voters of the Conservative Party in the United Kingdom and of Celis, Roggemans, and Spruyt (Reference Celis, Roggemans, Spruyt, Spruyt and Siongers2015) studying within-party sex gaps on feminist attitudes among Flemish voters.
To what extent our results can be generalized to other regions and countries depends on the institutional and cultural context. Our respondents operate in a multiparty setting of representative democracy at the local level. Generalization is thus most realistic to comparable settings. Multiparty representative democracy typically allows for politicians’ discretion in policy making—in stark contrast to situations of direct democracy in which the (median) voter has a dominant position. In the latter setting, we would expect politicians’ preferences to be more in line with that (unique) median voter and male and female politicians’ preferences not to differ (Stadelmann, Portman, and Eichenberger Reference Stadelmann, Portmann and Eichenberger2014). Our respondents are active at the municipal level. Generalizing the results to other levels of government should be done with caution, if only because of differences in scale: councils are relatively small (7 to 55 councilors). While party discipline tends to be high in Flemish councils (as is often the case in European politics), their small size may leave more room for “personal” preferences (to play a role in actual policy making). Small party fractions—often consisting of two or three representatives (the average is almost five)—likely imply that representatives (our respondents) take a central position when deciding on the “party line.” Finally, the specific quota rules that are in place, notably their neutrality (party lists need to have an equal number of male and female candidates) may be responsible for sorting effects that create within-party sex gaps (as discussed in the section “Within-Party Sex Gaps”).
Our findings may help solve a puzzle that is implicit in earlier literature. Indeed, an observation of a within-party sex gap that is systematically to the left would raise the question why women (or men) do not change parties: if one finds himself to be more leftist (rightist) than other party members, a rational response could be to move to a party on the left (right). As a consequence, within-party gaps would disappear. If, of course, as we observe, women sometimes position themselves to the left and sometimes to the right, the existing situation may be an equilibrium outcome and within-party sex gaps may be sustainable. More generally, a within-party sex gap may be the outcome of our model in the section “Within-Party Sex Gaps,” where initial preferences are such that women are more leftist and where quota rules constrain parties in their choice of candidates and create incentives for career-seeking politicians to join parties with nonmatching ideology.
The strong power positions of individual representatives and female representation in the local (Flemish) government raise the question whether, and under what conditions, the descriptive representation of women translates into substantive representation. Exploring this link between (sex-related) preferences and actual policy (see Svaleryd Reference Svaleryd2009) in Flemish municipalities is an interesting route for further research. From a policy perspective, the observation of systematically different preferences between male and female party members is, of course, of the utmost importance. It shows that the politician's sex affects his or her preferences in a systematic way. This provides a rationale for voters using candidates’ sex as a voting cue in low-information elections as well as for the idea that the workings of political institutions is dependent on their sex composition. Such an observation can be seen as lending support to institutional measures like quota rules that aim to guarantee a sex balance.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S1743923X18000685.
