The Glass Ceiling in Politics

Abstract

There is a scarcity of women at the apex of political power, as well as a lack of methods to disentangle the potential sources of this under-representation. This article suggests a four-step method to test for whether women’s under-representation can be explained by the existence of a “glass ceiling.” We emphasize that this concept implies discrimination in promotions within the political organization, that the discriminatory promotions increase in severity at the top levels of power, and that they increase in severity during an individual’s career trajectory. The proposed method is applied to subnational politics in Sweden, a long-standing world leader in women’s descriptive representation. The results support the conjecture that a glass ceiling is hindering elected women’s rise to political power in this context.

Keywords

glass ceiling political careers subnational politics women and politics supply of politicians gender inequality racial inequality

A glass ceiling is a political term used to describe the unseen, yet unbreachable barrier that keeps minorities and women from rising to the upper rungs of the corporate ladder, regardless of their qualifications or achievements.

—U.S. Department of Labor (1995)

There is an unequal distribution of political power in most contemporary democracies. Women make up 21.8% of the world’s parliamentarians, but only 7.8% of its heads of government and 5.9% of its heads of state (UN Women, 2014). Their absence from positions of power suggests that society may be drawing its political leaders from an overly narrow pool of human talent (Murray, 2014), and that women’s preferences may be severely under-represented in political decisions (Brollo & Troiano, 2013; Chattopadhyay & Duflo, 2004; Phillips, 1995; Svaleryd, 2009; Thomas, 2002). There is a growing interest in women’s ascendance up the political hierarchy, including their access to ministerial portfolios, party leadership positions, and head of government (Jalalzai, 2013; Krook & O’Brien, 2012; O’Brien, 2015). Yet this work often lacks a mechanism for explaining women’s promotion from one level of power to the next.

This article suggests an empirical methodology with testable criteria to establish whether a political minority is held back from the upper levels of political organizations by a glass ceiling. We also show how this methodology can be applied by testing whether a glass ceiling exists for women in the case of Sweden. This case study draws on comprehensive panel data for Swedish politicians and constitutes a research contribution in its own right. The Swedish case provides an interesting testing ground, partly because the core political institutions of multiple political parties, proportional representation, and list-based selection are the most common institutional setups in the world. Sweden also stands out as a long-standing world leader in women’s numerical political representation. Arguably, if a glass ceiling exists in this context, it is also likely to exist in other places.

Our article seeks to increase the methodological rigor of understanding how under-represented groups fare in political institutions (Bloemraad, 2013; Childs & Krook, 2009; Driscoll & Krook, 2009; McKay & Krook, 2011). We extend the growing literature on the under-representation of women in political leadership by shifting the focus from macro-systematic explanations to party-level and micro-level explanations. Rather than comparing countries, we use detailed panel data for politicians, and for the promotion decisions among these politicians by hundreds of local party groups.¹

We argue that the concept of a glass ceiling represents a specific pattern of career disadvantages that can explain the lack of women in top positions (Baxter & Wright, 2000a, 2000b; Cotter, Hermsen, Ovadia, & Vanneman, 2001; Feree & Purkayastha, 2000; U.S. Department of Labor, 1991, 1995). There are two main features of this pattern. First, there must be discriminatory barriers to women’s career advancement (i.e., their slower advancement must stem from sex-based discrimination rather than differences in work experience, formal merits, or personal preferences). Second, these discriminatory barriers must increase for positions higher up in the organizational hierarchy. In the article, we outline four criteria with testable empirical implications, which can be taken to the data to examine the existence a glass ceiling.

Our effort to define the metaphor of the glass ceiling can help clarify the growing academic and public debate on women in politics. This metaphor has been used as a blanket statement to describe the declining proportion of women at the top of political hierarchies. Scholarly work has thus overlooked the fact that the concept entails discriminatory promotions between candidates with equal qualifications and more intense discrimination in higher organizational ranks (for examples of these omissions, see Jalalzai, 2013; Jalalzai & Krook, 2010; Kenski & Falk, 2004; Kropf & Boiney, 2001; Palmer & Simon, 2001, 2010; Thomas & Adams, 2010; Trimble & Arscott, 2003; UN Women, 2014; Verge, 2010). Our article calls for a more stringent use of this metaphor by recognizing that it refers not only to a declining proportion of minorities at the top of organizations but also to a specific reason why that pattern is observed.

Differentiating the glass ceiling effect from other drivers of vertical inequality is crucial from a policy standpoint. For example, it is commonly argued that an insufficient number of qualified candidates is causing women’s under-representation in top political positions—a situation referred to as the pipeline problem (Carroll, 1994; Carroll & Strimling, 1983; Darcy, Clark, & Welch, 1994; Herrick, 1995; Kobayashi, 2004; Norrander & Wilcox, 1998; Rodriguez, 2003). According to this assessment of the problem, appropriate policy measures to foster a more equal division of power could include gender quotas or efforts to bolster recruitment and training. As these policies can deliver a richer pipeline of qualified candidates to entry-level positions, top positions would then become more equally divided over time. By contrast, addressing inequality that is caused by a glass ceiling requires strategies to directly target the norms and practices that underpin discriminatory promotions. Improving the pool of qualified candidates is of little use if qualified women are systematically overlooked when candidates advance to the higher organizational levels.

Our case study of local Swedish parties shows that conditional on observable qualifications, the proportion of women shrinks as we move up the political hierarchy (meeting Criterion 1 of “Conditional Vertical Inequality”). This inequality is also more severe for the top positions than for the lower ones (meeting Criterion 2 of “Bottom-to-Top Inequality Acceleration”). In a more data demanding test that relies on individual panel data, we find that conditional on observable qualifications, women’s promotions are slower than men, (meeting Criterion 3 of Career Advancement Inequality). Finally, we find some evidence that women’s and men’s career trajectories diverge over time, although we lack sufficient data on complete career trajectories to conduct the most rigorous test possible (thus providing suggestive evidence in support for Criterion 4 of Diverging Career Trajectories).

Our sensitivity analysis shows that these findings exist also in a sample of politicians who did not have preschool children during their time in elected office. Extending the analysis to subsamples for Sweden’s three largest political parties gives suggestive evidence that the glass ceiling exists in parties with both left and right ideologies.

That our results are observed in the Swedish case is interesting for several reasons. In the time period that we study, most of the local parties already had proportions of elected women that exceeded the cutoff for a political “minority,” ~30%. This means that a glass ceiling is observed despite the common understanding that gender should largely have lost its importance as a salient divisor of the internal work environment of these organizations (Dahlerup, 1988; Kanter, 1977). As countries around the world take steps to increase women’s numerical representation, our study cautions that higher numbers of women by no means guarantee equal influence.

Four Criteria for the Glass Ceiling in the Political Sector

In this section, we combine insights from previous research with our understanding of the political sector to formalize the concept of the glass ceiling and derive testable criteria. A useful point of departure for this discussion is that discrimination is the foundation of the glass ceiling. This implies that a person’s gender reduces her probability of being appointed to influential positions in an organizational hierarchy (U.S. Department of Labor, 1991, 1995). In addition, the glass ceiling is different from other forms of inequality because it requires that the obstacles to upward advancement increase in severity at the higher levels of the organization.

The most comprehensive operational definition of the glass ceiling is provided by Cotter et al. (2001), who break the complicated metaphor down into four criteria. We modify the wording of these criteria, interpret the content to fit the political sector, and discuss their empirical implications.

Criterion 1: Conditional Vertical Inequality

Cotter et al.’s first criterion is that there is a gender inequality in holding higher offices that is not explained by other job-relevant characteristics of the politician. In technical terms, there should exist a residual difference in the proportion of women or minorities in higher political offices after controlling for differences in experience, education, abilities, motivation, and other characteristics. This criterion reflects the idea of differential recruitment intensity for equal qualifications, which is present in the political science literature with respect to recruitment into political organizations. For example, Norris and Lovenduski (1995) argue that party recruiters may value a specific set of qualifications differently depending on the candidate’s sex.

In empirical efforts to control for all job-relevant characteristics, potential control variables should be screened for endogeneity concerns, because we risk “over-controlling” for differences in qualifications if those differences were formed within the organization and as a consequence of a discriminatory promotion process. For example, political parties’ internal resources may be allocated to junior male politicians by senior male politicians. If resources are allocated via homosocial ties in this way, junior men end up with more resources than junior women, and thus become more likely to advance. Controlling for the amount of resources per person in a regression that has upward career moves as the outcome variable would thus explain at least some of the gender difference in promotions. This approach would, however, “over-control” for job-relevant qualifications, because the resource allocation would be part of the discriminatory phenomenon that we wanted to keep in the residual.

Given the challenge with endogenous controls, which control variables should be included? The answer is that we should include qualifications that can be seen as mostly external to the political promotion process itself, for example, education level and previous work experience. Another important control is work experience in the organization. In any organization, length of experience is a main ingredient in career promotions. There are also reasons to believe that this control captures some gender differences that are not due to discrimination. For example, women will have shorter tenures if they enter politics when they are older or if some women are more likely to leave politics because of family concerns (i.e., Childs, 2004; Sapiro, 1982; Thomas, 2002).

Criterion 2: Bottom-to-Top Inequality Acceleration

The second glass ceiling criterion is that gender inequality in holding influential positions is more pronounced for higher political offices than for lower offices. This represents the metaphor of a ceiling that hinders upward movement and restricts access to offices above a certain level in the organization (Baxter & Wright, 2000a; see also Albrecht, Bjorklund, & Vroman, 2003; Morgan, 1998; Prokos & Padavic, 2005).

The empirical implications of the second criterion have been subjected to scholarly debate. Earlier work argued that the criterion was met if women’s disadvantage in advancement increased at each hierarchical level (Baxter & Wright, 2000a). A critique of this definition, which was subsequently acknowledged by Baxter and Wright (2000b), held that a constant disadvantage across levels should be sufficient. Feree and Purkayastha (2000) convincingly stated that from the candidate’s perspective, a constant disadvantage at each career step would produce a cumulative (and thereby greater) disadvantage at the upper steps. A woman who starts at the entry level faces some disadvantage in proceeding to the next level, more disadvantage at the next career step, and so on. A constant disadvantage for women at each level thus produces a increasing cumulative disadvantage for the individual and a falling proportion of women as we move up the organizational ranks. Cotter et al. (2001) argue that a constant disadvantage should be considered evidence of a glass ceiling, but that a growing disadvantage should make us “even more confident that the glass ceiling exists” (p. 659).

A second reason to require a constant estimate of women’s disadvantage across hierarchical levels, rather than a growing disadvantage, is the possibility of gender differences in unobserved qualifications (Feree & Purkayastha, 2000). If women are subjected to a more stringent selection process (due to discrimination) as they move up the organizational hierarchy, women in the available pool of potential candidates for subsequent promotion may be increasingly more qualified than men. This difference could exist along a dimension that is unobserved by the researcher and hence not captured by the control variables. This would mean that an estimated constant (or even decreasing) chance of advancement would still represent more discrimination at the top of the hierarchy, as the pool of available women has become superior to that of men.

Criterion 3: Career Advancement Inequality

The third and fourth glass ceiling criteria both aim to disentangle inequality in representation at different levels in the political hierarchy from inequality in advancement into those higher levels (Cotter et al., 2001). The third criterion states that there is a gender inequality in the probability of advancing to a higher level, not merely in the probability that the average woman will occupy those higher positions. Testing this criterion requires longitudinal data for the career trajectories of individuals over time and not just cross-sectional information on the proportion of women at each organizational level.

Criterion 3 addresses some possible weaknesses of cross-sectional comparisons. First, this method fails to account for the fact that not all individuals enter the organization at the lowest level, and that men may be more likely than women to enter directly at higher levels. This situation will show an over-representation of men at higher levels, especially when the regression includes a control variable for experience so that only persons with the same length of political tenure are compared. In a cross-sectional setting, an over-representation of men with shorter tenures in high positions will thus be interpreted as a glass ceiling, but this is not correct, because new “hires,” rather than discriminatory promotions within the organization, caused the vertical inequality. Directly comparing actual promotions sidesteps this concern.

Another weakness of the cross-sectional approach is that a lower proportion of women in top positions may be the consequence of a “class cliff” rather than a “glass ceiling” (O’Brien, 2015; Ryan & Haslam, 2005, 2007). If women have shorter tenures in top posts (i.e., they fall off the glass cliff), this will produce small proportions of top-ranked women, but not because of discriminatory promotions. Again, a longitudinal comparison of inequalities in actual promotions rules out this alternative explanation of the small proportion of women at the top.

Criterion 4: Diverging Career Trajectories

The fourth criterion is that gender inequalities in holding influential positions increase during the course of a political career, and the disadvantage grows more for more influential positions. This means that the career paths of men and women who enter the organization at the same hierarchical position should have more similar career developments at first, but then increasingly diverge over time as women are prevented from advancing.

Criterion 4 specifically states that the glass ceiling should become more salient during an individual’s career trajectory. If we observe a difference in the probability of promotion for a given length of experience and other job-relevant qualifications, as required by Criterion 3, this difference could be driven by unequal chances of promotion among newcomers in the organization. We thus need additional testing to evaluate whether inequalities are greater for more senior members.

To outline how to test for growing gender differences in the returns to experience for more influential positions, we can imagine a man and a woman who start out together at the entry level of the political organization. Criterion 4 dictates that the probability that they will reach the middle level after, say, two election periods is more similar than the probability that they will have reached the top level after four periods. If a glass ceiling exists, we should observe a more dramatic career divergence for the same length of experience for the top positions in the hierarchy compared with the less influential positions.

Summary and Discussion

A test procedure for the four glass ceiling criteria is outlined in Table 1 below. One way to proceed is to start with the first criterion, check each observable prediction, and move onto the next criterion if the previous one is met. For example, if we do not observe a gender difference in the probability of holding higher position(s) before including control variables, we can conclude that there is no glass ceiling in the organization.

Table 1.

Summary of the Four Glass Ceiling Criteria, Observable Implications, and Data Requirements.

1: Conditional vertical inequality	There is a gender inequality in holding higher offices that is not explained by other job-relevant characteristics of the politician.
Observable implications	Women have a lower probability of holding the higher position(s) in the organization.
	This inequality persists after including control variables for job-relevant qualifications.
Data requirements	Cross-sectional data that include at least two hierarchical levels and control variables for key qualifications.
2: Bottom-to-top inequality acceleration	Gender inequality in holding influential positions is more pronounced for higher political offices than for lower offices.
Observable implications	Women’s disadvantage in holding higher positions is at least as large for top position(s) as for mid-level position(s), and at least as large for mid-level as for entry-level position(s).
	This pattern is not affected by the inclusion of controls for job-relevant qualifications.
Data requirements	Cross-sectional data that include at least three hierarchical levels and control variables for key qualifications.
3: Career advancement inequality	There is gender inequality in the probability of advancing to a higher level.
Observable implications	Over time, the average woman has an average lower probability of moving from a lower position to a higher one than the average man.
	This difference exists across the political hierarchy, and remains after controlling for job-related qualifications.
Data requirements	Individual-level panel data that include key qualification variables. At least some individuals in the data set must move up to higher positions over time.
4: Diverging career trajectories	Gender inequalities in holding influential positions grow over the course of a political career, and the disadvantage grows more for more influential positions.
Observable implications	As the length of politicians’ tenure increases, women’s disadvantage in reaching top positions also increases.
	This differential remains after controlling for job-relevant qualifications.
Data requirements	Individual-level panel data that include key qualification variables. The longer the time span of the data, the better.

Moving down the list of criteria, the tests become increasingly data demanding. Proceeding through the tests may be viewed as a way for the researcher to become increasingly certain (or not) of the existence of a glass ceiling. Cross-sectional data that support the first two criteria would represent a strong indication, but compelling evidence requires all four criteria to be met. It is also important to recognize that the glass ceiling is only one of several dimensions along which discrimination could occur. For example, the glass ceiling targets discrimination in the career progression of men and women who remain in the organization, but it does not capture discrimination that causes women to quit politics at a faster pace than their male colleagues.

Case Study and Data: Swedish Municipal Politics

We now describe our test case for the four glass ceiling criteria. We use individual-level panel data for Swedish municipal politicians, a data set that contains all elected politicians in seven elections, 1988-2010, in each of Sweden’s 290 municipal assemblies. Each assembly has between 31 and 101 members who are elected via a flexible list proportional representation (PR) system with one preferential vote allowed per voter.

The political power of the municipality is concentrated in the municipal council board. Its chairperson is appointed by the largest party in the governing coalition after the general election, and the remaining board seats are distributed among all parties according to the number of seats they hold in the full assembly. The council board chair is the top political executive of the municipality, the equivalent to the mayor in other countries. He or she is the main architect of the development strategy of the municipality and leads the political and practical executive work of the council board (Jonsson, 2003).

Specific policy areas are dealt with in subcommittees, and the governing majority appoints the chair of each committee. Most municipal councils have seven subcommittees, including in the areas of education, social work, building permits, and land use.

All positions within the council board and the subcommittees are appointed by the local parties after the election, and are usually handled by the same nomination committee that put together the electoral ballot. In general, appointments are made in accordance with the rank order of the electoral ballot. For example, the council board chair holds the top position on his or her electoral ballot in nine out of 10 municipalities (authors’ own data). Similar to the nomination processes for the electoral ballot, there is no formal application process for the top appointments. Instead, the nomination committee usually approaches individual candidates with requests. In a recent survey among candidates in the 2010 local election, 80% of the respondents said that they appeared on the ballot as a consequence of “being asked and saying yes,” a proportion that was even larger among politicians with longer experience (Järnbert & Wilén, 2013).

According to Swedish law, all political parties must submit their electoral ballots to the electoral authority and include the personal identification number of every politician on the list. Using this information, we requested Statistics Sweden to link every politician on every electoral ballot to yearly panel data on a host of socioeconomic variables from the administrative records for the full period (1988-2010). These data are not self-reported, have few missing values, and are highly accurate. The variables include sex, age, educational attainment, occupation sector, and the birth year of every child of every politician.

Testing the glass ceiling criteria requires that we break the political organization down into hierarchical levels. In most cases, this should be straightforward, as political systems and organizations around the world tend to have predictable and comparable promotion ladders that politicians climb over the course of a career (Fenno, 1973; Hagevi, 2010; Norris, 1997). We divide the Swedish municipal political hierarchy into three levels, L1-L3, according to the following list and illustrated in Figure 1 below:

a. Level 1—elected politicians: those who are elected but do not belong to the group of top and mid-level politicians.

b. Level 2—mid-level chairs: those who are appointed as the chair of a subcommittee or municipal council, but not to the municipal council board.

c. Level 3—top chairs (mayors): those who are appointed to chair the municipal council board.

Figure 1.

Political organization of a typical Swedish municipality, including the operationalization of hierarchical levels L1-L3.

Using this operationalization requires that we limit our data sample to the largest political party in the governing majority in each municipality and election period. These parties appoint both the council board chair (L3) and the majority of the subcommittee chairs. Within these parties, there is hence a clear organizational ladder ranging from lower to higher appointments.

Our operationalization of the political hierarchy is supported in qualitative research that describes political appointments and careers in Swedish municipalities (Jonsson, 2003; Nilsson, 2001). A 2010 survey that asked municipal politicians to rank political positions in terms of their influence clearly placed the council board chair at the top, followed by the chair of the municipal council, and then by the subcommittee chairs (Gilljam, Karlsson, & Sundell, 2010). Counting the chair of the municipal council (corresponding to the speaker of the house of parliament) as a mid-level chair rather than a top chair is further supported by previous qualitative work (see Montin, 2007). Our suggested hierarchy of positions is also reflected in the monetary compensation of the positions. The positions of entry-level politicians are honorary, with low piece-rate payments for individual meetings. Among mid-level politicians, large municipalities usually grant part-time or even full-time political employment to one or several individuals. The top position is a full-time political job in all but a handful of the smallest municipalities.

There are several reasons why it would have been less appropriate to define the political hierarchy by ranking subcommittees in order of political importance (see approaches by Baekgaard & Kjaer, 2012; Carroll, 2008; Yule, 2000). Wide (2011) argues that the standard method of ranking committees from high-powered (usually financial and economic areas) to low-powered areas (usually care and family areas) is not appropriate in the Swedish municipal context. Municipalities’ main areas of political authority largely coincide with traditional “female” and “soft” issue areas, which makes these committee chairs (at least) as politically important as those of the budgetary or technical committees.

Information on the exact appointments of each politician was extracted from a mandatory survey collected by Statistics Sweden following the 2006 and 2010 elections. The sample used in the analysis comprises politicians from the largest majority parties and covers 34% of the total number of elected politicians in the Swedish municipalities. In this sample, women make up 48% of the entry-level politicians, 38% of the mid-level chairs, and 28% of the top chairs. As expected, the sample has an over-representation of certain political parties. The Social Democrats account for 46% of the sample and the Conservatives for 32%, compared with 39% and 24%, respectively, in the full data set.²

The proportion of women is highly similar in the full data set (0.44) and in the sample (0.45). Descriptive statistics for other variables show the same similarity (see Online Appendix Table A1 for details). In both the sample and the full data set, the most salient differences between the male and female politicians lie in their age and education levels. Female politicians are younger and more likely to have tertiary education than their male colleagues.

Measuring Qualifications for Political Appointments

The glass ceiling criteria require that we hold constant all job-relevant qualifications to analyze whether equally qualified women face a disadvantage in the political organization relative to men. This makes the empirical measures of qualifications crucial. The more accurate the qualification controls, the more accurate the claim that residual gender differences in career progress can be attributed to discrimination.

There are two important challenges to choosing the appropriate qualification measures. As mentioned above, we run the risk of controlling for characteristics that are qualifications for higher positions, but that are also outcomes of previous discrimination. In the Swedish context, preference voting is a likely example of such a variable. If parties discriminate against women in their external communications and internal support of candidates, women may obtain fewer preference votes on average. Controlling for preference votes as a qualification will hence capture some of the discrimination that we want to leave in the residual.

The second challenge is that the very notion of a “qualified candidate” is subjective. Women may have different life experiences, skills, and (socialized) personality traits that party elites may consider as differentially valuable merits for higher office. Our approach is to depart from the status quo of characteristics that are generally used to capture politician quality and that have traditionally helped male candidates get into power (see, for example, O’Brien & Rickne, 2014; Schwindt-Bayer, 2011). If anything, the potential omission of women-specific qualifications that this implies makes us less likely to find evidence of a glass ceiling.³

The length of job experience is a key qualification for top positions in any sector. Climbing the career ladder inside the political organization via consecutive promotions over time is the standard way to reach the top posts in most political contexts around the world (Fenno, 1973; Hagevi, 2010; Norris, 1997; for a different conclusion on Latin America, see Escobar-Lemmon & Taylor-Robinson, 2005; Heath, Schwindt-Bayer, & Taylor-Robinson, 2005). Previous experience in political office is a common measure of a politician’s “quality,” particularly in studies of U.S. politics (Cox & Katz, 1996; Hirano & Snyder, 2013). The idea is that skills improve over time as politicians develop task-specific human capital.

We measure experience as the number of previous terms that a politician has served in the municipal assembly. Because the earliest available year in our data is 1988, the variable is truncated at five election periods. We count previous periods of office back to the 1988 election for politicians elected in 2006, and back to 1991 for those elected in 2010. All periods of office holding are included, irrespective of whether they are consecutive.

The second measure of qualifications is the level of education, a variable broadly considered to capture enhanced practical skills, signaling ability, and civic engagement (Franceschet & Piscopo, 2012; Schwindt-Bayer, 2011). We include dummy variables for seven categories of educational attainment, ranging from having 9 years of education to having a doctorate degree.

Our third qualification variable is age (following, for example, Davis, 1997; Franceschet & Piscopo, 2012), which we capture using dummy variables for five age intervals: 18-29, 30-49, 50-60, 61-64, and 65 or above.

The fourth measure captures the politician’s occupation sector in the non-political labor market, departing from the view that the skill sets developed in different labor market sectors differ in their relevance for the political arena (Messner & Polborn, 2004). We chose to be agnostic on which occupations are more beneficial for the political sector by including binary indicators for 10 aggregate employment sectors. The fifth and final measure is a binary indicator of whether the politician has previous experience as a parliamentarian.

As discussed above, we do not include a politician’s preference vote support as a qualification measure. In Sweden, voters have the option of casting one voluntary preference vote, and usually do so for a politician in the top portion of the ballot. As a consequence, the votes have little impact on who is elected, but previous research has shown that they are important for the distribution of powerful appointments within the party (Folke, Persson, & Rickne, in press). The receipt of preference votes is, however, likely to be endogenous to the glass ceiling effect. Discriminatory treatment of male and female candidates by the party elite in the public promotion of candidates (and in support of personal electoral campaigns) will be reflected in their vote shares. More importantly, discriminatory promotions to top posts are highly likely to boost vote counts through a pure order effect on the electoral ballot (see Blom-Hansen, Elklit, Serritzlew, & Villadsen, 2015; Esteve-Volart & Bagues, 2012).

Although our data are of uniquely good quality for measuring qualifications, important factors may remain unobserved. A common argument in the debate about women’s political careers is that women are less ambitious and expend less effort on their political positions because they prioritize family-related tasks instead. In our baseline analyses, we follow Cotter et al. (2001) and exclude family-related variables from the set of baseline qualification controls. In the final section of the article, we support this approach by showing that our results are not sensitive to excluding all politicians who have children from the analysis.

Although we do not have a measure of ambition in our data set, we can use other sources of survey data to compare male and female politicians on this dimension.⁴ A 2012 survey of all municipal politicians, which had a response rate of 80%, asked politicians to state their desired future tenure by responding to the question of “how long they wish to remain an elected representative.” Figure 2 shows the distribution of responses divided by gender and for the same sample of parties that is used for the main analysis of this article, politicians at the lower two hierarchical levels, and in the largest majority parties.

Figure 2.

Distributions of self-declared desired future tenure among male and female politicians, 2012.

Figure 2 does not reveal a gender difference in desired future tenure, a result that corresponds to similar surveys in both Sweden and abroad. For example, Järnbert and Wilén (2013) found no gender difference among municipal politicians in the proportion that agreed with the statement that their candidacy for office was “motivated by the ambition to start a political career” (Järnbert & Wilén, 2013). Öhberg (2010) found no significant gender difference in career ambitions among parliamentarians, measured as self-stated desired positions 10 years into the future.⁵ A comparable study among U.S. members of Congress by the Center for American Women and Politics (2001) found that female state legislators were “as likely as men to aspire to another political office as well as to leadership positions within their legislatures.” Finally, gender parity in progressive career ambitions has been documented in diverse contexts such as Latin America (Schwindt-Bayer, 2011), India (Geissel & Hust, 2005), and Germany (Davidson-Schmich, forthcoming).

Empirical Method and Results

We are now ready to outline the empirical methodology and the empirical results for the four glass ceiling criteria:

Criterion 1: There is a gender inequality in holding higher offices that is not explained by other job-relevant characteristics of the politician

To test Criterion 1, we estimate the regression equation

Y_{i t} = a_{i t} + β_{1} w_{i} + λ^{'} E_{i t} + η^{'} Q_{i t} + γ_{p t} + ε_{i t},

where the outcome variable, $Y_{i t}$ , is a binary indicator for holding a position higher up in the political hierarchy. We run two versions of the regression, one where the outcome takes the value 1 for the two upper levels, and otherwise 0, and one where the outcome takes the value 1 for the top level only, and otherwise 0. The first independent variable, $w_{i}$ , is the binary indicator for being a woman. We estimate Equation (1) using a conditional logit estimator. This means that the estimate of interest, $β_{1}$ , captures the average woman’s odds of the outcome variable divided by the average man’s odds, the so-called odds ratio. An estimate below 1 implies that the average woman’s odds are lower than the average man’s. To hold qualifications for higher office constant, the equation includes a variable vector $E_{i t}$ that contains the five binary indicators for the number of previous periods in office, as well as the vector of our other qualification controls, $Q_{i t}$ .

The conditional logit estimation is a logit specification that allows for fixed effects in panel data. These fixed effects are necessary for our analysis to capture gender differences at the level of the (discriminatory) promotion choice. We want to compare women’s career performance relative to their male colleagues within the same organization and at the same time, that is, within the same local governing party rather than across parties, regions, or time periods. To achieve this, all our regression specifications include fixed effects for each local party interacted with fixed effects for each election period. These variables are jointly denoted and consist, hence, of one dummy variable for every combination of local party and election period, for example, the Social Democrats in the municipality Upplands Väsby Municipality in the 2006 election. Because fixed effects for this many groups (two election periods multiplied by 290 local parties) cannot be estimated with a regular logit model, we resort to the conditional logit estimation method.

The results are presented in Table 2. Columns 1 and 2 examine inequalities in holding any type of chair position. In Columns 3 and 4, we estimate the inequalities in holding the top position only. For each outcome, we first present the results both with and without the control variables for qualifications and experience.⁶ To save space, the estimates for the control variables are not reported here (see Online Appendix Table A2 for the full set).

Table 2.

Conditional Logit Estimates of the Odds Ratio Between Women and Men in Holding Any Chair Position (L2 or L3 Relative to L1), and in Holding the Council Board Chair Position (L3 Relative to L1 and L2), Compared With Men.

	Any chair position (L2 + L3) = 1, otherwise 0		Council board chair (L3) = 1, otherwise 0
	1	2	3	4
Woman	0.59*** (0.03)	0.59*** (0.03)	0.50*** (0.05)	0.51*** (0.06)
Observations	8,899	8,839	8,860	8,605
Qualification controls		yes		yes

Note. Fixed effects for municipalities, interacted with fixed effects for election periods, are included in all specifications. Robust standard errors in parentheses. The qualification control variables are five binary indicators for one to five periods of previous experience, binary indicators for six education levels, binary indicators for five age categories, binary indicators for 10 occupation sectors, and one binary indicator for previous experience as a parliamentarian.

Significant at 10%. **Significant at 5%. ***Significant at 1%.

The results in Table 2 show that Criterion 1 is met. The average woman has 40% lower odds of holding any chair position (L2 + L3) than the average man, and the average woman has 50% lower odds of holding the top post (L3) compared with the average man. In Columns 2 and 4, we can also see that neither of these estimates is affected by adding the control variables for job-relevant qualifications.

Criterion 2: Gender inequality in holding influential positions is more pronounced for higher political offices than for lower offices

To test Criterion 2, we follow Baxter and Wright (2000a) and compare inequalities in the odds of holding the higher of two “adjacent” positions. In our context of three hierarchical levels, comparisons are made in two subsamples of data. The first sample contains the lowest and the middle levels (dropping the top level), and the other sample contains middle and the top levels (dropping the lowest level).

The results are presented in Table 3. Support for Criterion 2 requires statistically significant inequalities that are either of constant size across the two levels or larger for the top level than for the middle level. If the estimated inequality is smaller for the higher level, this means that we do not find support for the criterion. Because the two inequalities are estimated in two separate samples, we cannot directly test whether they are of the same size or not. Instead, we compare the estimates in a common-sense manner based on both size and precision. For example, if the standard errors on the estimates are very large, we simply will not be able to say anything about Criterion 2.

Table 3.

Conditional Logit Estimates of the Odds Ratio Between Women and Men in Holding a Mid-Level Chair Position Versus Being Elected, and of Holding a Council Board Chair Position Versus Holding a Mid-Level Chair Position.

	Mid-level chair position = 1, elected politician = 0		Top-level chair position = 1, Mid-level chair position = 0
	1	2	3	4
Woman	0.64*** (0.03)	0.63*** (0.04)	0.70*** (0.07)	0.63*** (0.08)
Observations	8,137	8,085	2,614	2,536
Qualification controls		yes		yes

Note. Fixed effects for party group, interacted with fixed effects for election periods, are included in all specifications. Robust standard errors in parentheses. The qualification control variables are five binary indicators for one to five periods of previous experience, binary indicators for six education levels, binary indicators for five age categories, binary indicators for 10 occupation sectors, and one binary indicator for previous experience as a parliamentarian.

Significant at 10%. **Significant at 5%. ***Significant at 1%.

The results in Table 3 show a large and statistically significant disadvantage for women’s probability of holding the higher of the two positions in both of our pair-wise comparisons. Both of the estimated inequalities are of similar size. In fact, they are identical when the control variables are included. In both cases, the average woman’s odds of holding a mid-level position compared with an entry-level position is 37% lower than the average man’s odds. Both estimates are also estimated with high statistical precision, which means that we have high confidence that Criterion 2 finds support in the data.

Criterion 3: There is a gender inequality in the probability of advancing to a higher level.

The third criterion adds a dynamic element to the analysis. We examine the pattern of promotions between the 2006 and 2010 elections. Equation (1) is estimated using a binary outcome variable that takes a value of 1 for politicians who moved to a higher level in 2010 compared with 2006, and 0 for those who remained at the same level. As we do not want to confound our estimates of promotions with exits, we omit politicians who left politics between the two elections.

We estimate Equation (1) separately for promotions between L1 and L2, and promotions from either L1 or L2 to L3. The estimate of $β_{1}$ will now capture the average woman’s promotion odds relative to the average man’s. In the estimation of promotions to the top position, we also condition on the current position (L1 or L2). By including this control, we make sure that our estimate of interest does not capture an over-representation of men at the mid-level position, where the promotion probability for the top position is greater than that from the entry level.

The analysis is restricted to political parties that were the largest majority party in both 2006 and 2010. In this setting, our fixed effects mean that we drop all groups that had a constant outcome variable across both years, that is, where the distribution of positions remained the same across individuals in both years. The results are presented in Table 4, which follows the same structure as Tables 2 and 3.

Table 4.

Conditional Logit Estimates of the Odds Ratio Between Women and Men in Transitioning Into Positions Higher Up in the Political Hierarchy (2006-2010).

	Promotion from L1 to L2 = 1, No promotion = 0		Promotion from L1 or L2 to L3 = 1, No promotion = 0
	1	2	3	4
Woman	0.74* (0.11)	0.64** (0.11)	0.80 (0.24)	0.53* (0.18)
L2			3.50*** (0.90)	4.07*** (1.33)
Observations	1,205	1,202	787	757
Qualification controls		yes		yes

Significant at 10%. **Significant at 5%. ***Significant at 1%.

The results in Table 4 show support for Criterion 3. Conditional on the controls for qualifications, the average woman has 36% lower odds of being promoted from an entry-level position to a mid-level position and 47% lower odds of being promoted to the top position from either of the two lower positions. As the sample size for this analysis is much smaller than when we test for the two previous criteria, the estimation is less precise. The estimate of women’s disadvantage in promotions to the top level is quantitatively large but only statistically significant at the 10% level.

Criterion 4: Gender inequalities in holding influential positions grow over the course of a political career, and the disadvantage grows more for more influential positions

We have panel data for political appointments in two election periods, which makes it empirically challenging to test Criterion 4. Following the cross-sectional approach of Cotter et al. (2001), we compare the probability that women hold positions of influence across increasing levels of previous experience in pooled data for the two cross sections. To meet Criterion 4, we expect that women’s disadvantage will grow with the number of previous periods in office. We also expect that this growth is more salient for the top position than for mid-level positions.

We start with a descriptive analysis. Figure 3 shows the proportion of men and women who have reached the upper hierarchical levels by their total number of previous periods in office. It is apparent that experience in office is positively correlated with holding a position of influence for both men and women. When we examine the top position (the right-hand side of the figure), this inequality widens with the level of experience. This suggests that the pay-off of experience is larger for men. When we examine the middle level (left-hand side), the difference between men and women is constant. The descriptive analysis hence provides support for Criterion 4 on all dimensions but one, as there is not a widening gap for the mid-level position. However, the key part of Criterion 4 is a widening gap for the top position, which we do find.

Figure 3.

The proportion of men and women who hold influential positions (L2 or L3 in the left figure, L3 in the right figure) by total tenure in elected office (pooled 2006 and 2010 data).

To substantiate the findings from the descriptive analysis, we use formal regression analysis to estimate the following:

Y_{i t} = a_{i t} + β_{1} w_{i} + β_{2} w_{i} E_{i t} + β_{3} E_{i t} + η^{'} Q_{i t} + γ_{p t} + ε_{i t},

where the parameter of interest for testing Criterion 4 is $β_{1}$ , the estimate for the interaction term between being a woman, $β_{2}$ , and the number of periods of previous political experience, $E_{i t}$ . If women’s average disadvantage grows with political experience, as suggested by Criterion 4, this interaction term should be negative. Also, we would expect the negative relationship to be more pronounced for the top position. The results from this analysis are presented in Table 5, which follows the same structure as the previous tables.

Table 5.

OLS Estimates of the Odds Ratio Between Women and Men in Holding Any Chair Position, or in Holding the Council Board Chair Position, as a Function of Previous Experience.

	Any chair position (L2 + L3) = 1, otherwise 0		Council board chair (L3) = 1, otherwise 0
	1	2	3	4
Woman	−7.53*** (1.09)	−8.75*** (1.19)	−0.77 (0.50)	−1.80*** (0.54)
Woman × Periods of experience	−0.29 (0.62)	−0.46 (0.61)	−1.39*** (0.41)	−1.28*** (0.40)
Periods of experience	8.45*** (0.40)	8.73*** (0.42)	3.93*** (0.27)	4.54*** (0.27)
Observations	9,099	9,034	9,099	9,034
Qualification controls		yes		yes

Note. Fixed effects for party group, interacted with fixed effects for election periods, are included in all specifications. Robust standard errors in parentheses. The qualification control variables are binary indicators for six education levels, binary indicators for five age categories, binary indicators for 10 occupation sectors, and one binary indicator for previous experience as a parliamentarian. OLS = ordinary least squares.

Significant at 10%. **Significant at 5%. ***Significant at 1%.

Because the interaction term is the variable of interest, we can no longer use a conditional logit. As shown by Ai and Norton (2003), the logit estimate for an interaction terms cannot be interpreted as a marginal effect. In fact, the estimate may even be of the opposite sign of the actual marginal effect. Because of this, we use an ordinary least squares (OLS) to estimate Equation (2). In the online appendix (Table A3 and Figures A1 and A2), we use the software package of Norton, Wang, and Ai (2004) to further discuss and illustrate the problems associated with the logit estimates.

The results in Table 5 corroborate the descriptive analysis in Figure 3. The estimate for the interaction term in Columns 1 and 2 is not statistically significant and is fairly close to zero, showing that women’s average disadvantage in holding a position at either the middle level or the top level is constant across levels of experience. In contrast, the estimates in Columns 3 and 4 show that the disadvantage in holding the top position increases with experience. The more senior that men and women become in the organization, the larger the gender gap in the probability of reaching the very top. These results meet Criterion 4 to the extent that the career disadvantage of reaching the top grows with experience, which is the key element of Criterion 4. However, we do not find support for a growing inequality during the career trajectory in access to the middle level position.

The results in this section should also be interpreted with some caution because they are estimated using cross-sectional rather than panel data. This approach runs the risk of confounding career and cohort effects. Because politicians with more experience come from older cohorts, the estimates could be picking up more salient gender disparities among older politicians compared with younger ones.

Family Concerns and Political Parties

This section extends our baseline analysis in two ways. Our first extension addresses the concern that gender differences in the prioritization of family-related caregiving responsibilities could be driving our baseline findings. We examine this explanation by restricting our sample to politicians who never had a child between the ages of 0 and 6 during any of their years as elected representatives. We find that the baseline results are not sensitive to performing the estimation for politicians without preschool children only, with the caveat that the estimates are less precise due to the smaller sample size (see online appendix, Tables A4-A7 and Figure A3).

Our second extension considers differences in women’s career patterns across political parties. Previous research has shown that parties with leftist ideologies are more likely to have female friendly institutional structures, such as gender quotas, and are more likely to appoint female leaders (Caul, 2001; Kittilson, 2006; Krook, 2009; O’Neill & Stewart, 2009). Examining political parties in isolation reduces our sample size, but we can at least test Criteria 1 and 2 for the three largest political parties in our data set, the Social Democrats, the Conservative Party, and the Center Party. The Social Democrats are a leftist party that adopted a zipper gender quota for alternating male and female names on all party ballots in 1993. The Conservative Party and the Center Party do not have gender quotas, but the Center Party stands out as the first party to appoint a female party leader, and both parties have female party leaders today.

Table 6 shows the proportions of women across the three hierarchical levels in each of the three parties. The Social Democrats have the highest proportion of women on all three levels, followed by the Center Party. In all three parties, the proportion of women drops dramatically as we move up the hierarchy.

Table 6.

Proportion of Women on the Three Hierarchical Levels in the Three Largest Parties (Pooled Data for 2006 and 2010).

	Social democrats (%)	Center party (%)	Conservative party (%)
L1: Elected (without chair position)	53.2	44.8	41.3
L2: Mid-level chair position	39.9	35.9	34.5
L3: Top-level chair position	32.1	31.7	22.2

Table 7 presents the results of testing Criteria 1 and 2 on each party in isolation. As in the previous table, we show the estimate for women’s disadvantage in holding the mid-level position in Columns 1 and 2, while the results for women’s disadvantage in holding the top position are shown in Columns 3 and 4. The upper half of the table contains the results for Criterion 1 and the lower half for Criterion 2.

Table 7.

Test of Criteria 1 and 2 in the Three Largest Parties.

	Criterion 1
	Any chair position (L2 + L3) = 1, otherwise 0		Council board chair (L3) = 1, otherwise 0
	1	2	3	4
Social Democrats
Woman	0.56*** (0.03)	0.52*** (0.04)	0.49*** (0.06)	0.39*** (0.06)
Observations	4,939	4,916	4,939	4,900
Center Party
Woman	0.60*** (0.08)	0.58*** (0.11)	0.64** (0.14)	0.69 (0.20)
Observations	1,010	998	1,010	936
Conservative Party
Woman	0.67*** (0.06)	0.75*** (0.08)	0.47*** (0.09)	0.58*** (0.12)
Observations	2,885	2,856	2,885	2,736
	Criterion 2
	Mid-level chair position = 1, elected politician = 0		Top-level chair position = 1, Mid-level chair position = 0
Social Democrats
Woman	0.59*** (0.04)	0.57*** (0.05)	0.75* (0.11)	0.54*** (0.10)
Observations	4,654	4,632	1,442	1,430
Center Party
Woman	0.61*** (0.10)	0.54*** (0.12)	0.97 (0.26)	0.72 (0.26)
Observations	834	828	275	251
Conservative Party
Woman	0.75*** (0.07)	0.80** (0.09)	0.56*** (0.12)	0.58** (0.14)
Observations	2,629	2,601	841	799
Qualification controls		yes		yes

Significant at 10%. **Significant at 5%. ***Significant at 1%.

The results for Criterion 1 show a disadvantage for women in all parties and for both positions. Criterion 1 is met for the Social Democrats and the Conservative Party, while the evidence is inconclusive for the Center Party. The Center Party has the smallest estimated disadvantage for the top position, and the estimate is not statistically significant after adding the control variables, but given the smaller sample size, we cannot be fully confident in rejecting Criterion 1.

An interesting pattern emerges in the results for Criterion 2. The results for the Conservative Party show a larger disadvantage for the top position than for the middle position, indicating a glass ceiling even according to the most demanding definition. For the Social Democrats, the estimated disadvantage is large and of similar size for both of the pairs of adjacent categories, meaning that the second criterion for the glass ceiling is met by the less demanding version of the criteria. In the Center Party, the estimated disadvantage for the second step is clearly smaller, implying that the second criterion is not met.⁷

The party heterogeneity shown in this section is interesting from several perspectives. It shows that a high level of descriptive representation, like in the Social Democrats, does not necessarily equal leadership in vertical equality. The zipper gender quota of the Social Democrats had guaranteed near-numerical parity in the party’s elected seats for over a decade prior to our period of analysis. Corresponding to this effort, the descriptive statistics showed that this party had larger proportions of women in all three hierarchical positions that we study. Nevertheless, women continue to face a glass ceiling that blocks equal access to higher positions. The Conservative Party has lower levels of female numerical representation on all three levels, and we find strong evidence of a glass ceiling. The results for the Center Party also show lower levels of female representation on all hierarchical levels compared with the Social Democrats, but in contrast to the Conservative Party, we do not find that a glass ceiling can explain this pattern.

Discussion and Conclusion

This article contributes to understanding the absence of women from the top positions of political power. We formalize criteria and suggest an empirical strategy to test one potential explanation of this phenomenon: that a glass ceiling hinders women from climbing to the top of organizational hierarchies. This explanation firmly ties career disadvantages to the promotion process within the political organization. It thus differs from macro-level explanations (features of the electoral and party system, macro-level cultural norms, etc.) as well as from other micro-level explanations (a weak supply of experienced and otherwise qualified candidates, or that women prioritize caregiving work in the family over their political career).

Our empirical analysis on data for subnational political organizations in Sweden shows that a glass ceiling exists for female politicians. All of the four criteria for a glass ceiling find support in our analysis, although evidence for the forth criterion should be interpreted with caution due to limited access to panel data for politicians’ full career trajectories. Finding these results in the Swedish context can be seen as remarkable. It shows that even in a country with a long history of efforts to improve women’s political representation, a glass ceiling can continue to hinder women’s access to top posts.

These results indicate that some efforts to promote vertical equality will likely be more productive than others. First and foremost, as vertical inequality cannot be explained by women having less political experience or weaker qualifications, a “wait and see” approach is not sufficient. Even direct efforts to improve the supply of qualified women will likely fall short as long as equally qualified women are denied equal promotion chances. More successful strategies should target the norms, behaviors, and practices involved in parties’ promotion processes. In the Swedish context, such strategies should focus on recruiting women for top appointments.

Splitting the data set into separate analyses for each of the three largest parties leads to different recommendations for improving women’s representation. For the Social Democrats, a left-leaning party, numerical representation at the lower levels is near parity, but a glass ceiling remains for reaching the top. In the Conservative Party, women’s numerical representation is considerably lower and women also face a glass ceiling, indicating that strategies to improve women’s representation could address both the weak inflow of women at the entry level and the glass ceiling. In the Center Party, we do not find evidence of a glass ceiling; therefore, a targeted strategy could focus on recruiting more women rather than focusing on existing women’s promotion chances.

Fully identifying the channels of discriminatory promotions is outside the scope of this work. Previous work, however, suggests that we should not think of these results as the consequence of conscious acts of discrimination on the part of key actors in the promotion process. Notably, discrimination may be largely or entirely unintentional on the part of recruiters and/or party elites. As these groups are mostly male, similarity of socioeconomic characteristics, attitudes, values, and personalities between this in-group (perceiver) and candidates for promotion may be interpreted as qualifications (Bjarnegård, 2009; Lipman-Blumen, 2000; Piliavin, 1987). In turn, social similarities may lead to an assumption of competence in a way that favors the dominant group (Holgersson, 2003; Klahr, 1969; Schlozman, Burns, Verba, & Donahue, 1995).

Factors outside the political party may also affect recruiters’ impressions of the available candidates. Media bias against female politicians is widely documented, including an imbalance of focus of political commentary on their physical appearance and private lives rather than their political positions (Kahn & Goldenberg, 1991; Ondercin & Welch, 2005). Perceived voter bias against women as political leaders could also factor into promotion decisions (Norris & Lovenduski, 1995).

There are many avenues for future research. Perhaps most importantly, the four testable criteria for the glass ceiling could constitute a basis for future tests in other contexts. Given a sufficient number of observations, Criteria 1 and 2 could be applied to cross-sectional data to rule out the existence of a glass ceiling, or provide suggestive evidence of its existence. It these two conditions are met, further tests on individual panel data (to compare promotions over time) could also be added to bolster the evidence of a glass ceiling. If possible, whole career trajectories should be compared to test the most data demanding criterion of diverging career trajectories. However, even in the absence of such tests, analysis on cross-sectional and/or panel data of more limited time ranges can still give convincing evidence for or against the existence of a glass ceiling.

In other contexts, slightly different data could be used. In the U.S. case, comparisons could include movements across hierarchies of appointments both within and across chambers of legislatures, as well as to top executive positions. Other contexts could also enable analysis that covers more hierarchical steps in the political organizations, for example, the stages prior to elected office. For parties without zipper gender quotas, the earlier career steps of becoming a party member, being nominated to a lower rank on a party list, and climbing to an electable position on these lists could be relevant bottlenecks worthy of further analysis.

Another avenue for future work could combine promotion data with data sets on work effort and/or politicians’ preferences to more fully control for characteristics that remained unobserved in this study. In our case, the conclusion about the glass ceiling for women was sustained in a subsample of politicians without children. In a separate survey data set, we also showed highly similar levels of desired political tenures (a proxy for ambition) for men and women in lower levels of the political organization. With more detailed data on attitudes and work efforts of politicians, it would be possible to further ensure that residual disparities in promotion probabilities are not due to gender differences in these factors, while also carefully considering the potentially endogenous nature of such measurements.

Footnotes

Acknowledgements

We thank Diana O’Brien and seminar participants at Rutgers University for helpful comments. We thank Jonas Ahlerup, Johan Arntyr, Sirus Dehdari, and Elin Molin for excellent research assistance, and Kelley Friel for editorial assistance.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Financial support from the Swedish Research Council and the Torsten and Ragnar Söderberg Foundation is gratefully acknowledged.

Notes

Author Biographies

Olle Folke is a researcher at the Department of Government at Uppsala University and affiliated to the Research Institute for Industrial Economics. His primary research fields are applied political economics and gender and politics.

Johanna Rickne is a researcher at the Research Institute for Industrial Economics. Her research interest includes gender and politics and labor economics.

References

Norton

(2003). Interaction terms in logit and probit models. Economics Letters, 80, 123-129.

Albrecht

Bjorklund

Vroman

(2003). Is there a glass ceiling in Sweden? Journal of Labor Economics, 21, 145-177.

Baekgaard

Kjaer

(2012). The gendered division of labor in assignments to political committees: Discrimination or self-selection in Danish local politics? Politics & Gender, 8, 465-482.

Baxter

Wright

E. O.

(2000a). The glass ceiling hypothesis: A comparative study of the United States, Sweden, and Australia. Gender & Society, 14, 275-294.

Baxter

Wright

E. O.

(2000b). The glass ceiling hypothesis: A reply to critics. Gender & Society, 14, 814-821.

Bjarnegård

(2009). Men in politics: Revisiting patterns of gendered parliamentary representation in Thailand and beyond (Doctoral dissertation). Uppsala University, Uppsala, Sweden

Bloemraad

(2013). Accessing the corridors of power: Puzzles and pathways to understanding minority representation. West European Politics, 36, 652-670.

Blom-Hansen

Elklit

Serritzlew

Villadsen

L. R.

(2015, October). Ballot position and election results: Evidence from a natural experiment. Annual Meeting of the Danish Political Science Association, Vejle.

Brollo

Troiano

(2013). What happens when a woman wins an election? Evidence from close races in Brazil (CAGE Online Working Paper Series). Coventry, UK: Department of Economics, University of Warwick.

10.

Carroll

(1994). Women as candidates in American politics (2nd ed.). Bloomington: Indiana University Press.

11.

Carroll

(2008). Committee assignments: Discrimination or choice? In Reingold

(Ed.), Legislative women: Getting elected, getting ahead (pp. 135-156). Boulder, CO: Lynne Rienner.

12.

Carroll

Strimling

W. S.

(1983). Women’s routes to elective office: A comparison with men. New Brunswick, NJ: Center for the American Women and Politics.

13.

Caul

(2001). Political parties and the adoption of candidate gender quotas: A cross-national analysis. The Journal of Politics, 63, 1214-1229.

14.

Center for American Women and Politics. (2001). Women state legislators: Past present and future. Highlights and major findings. Retrieved from http://www.capwip.org/readingroom/cawp-womenstateleg.pdf

15.

Chattopadhyay

Duflo

(2004). Women as policy makers: Evidence from a randomized policy experiment in India. Econometrica, 72, 1409-1443.

16.

Childs

(2004). A feminized style of politics? Women MPs in the house of commons. The British Journal of Politics & International Relations, 6, 1-22.

17.

Childs

Krook

M. L.

(2009). Analyzing women’s substantive representation: From critical mass to critical action. Government and Opposition, 44, 125-145.

18.

Cotter

D. A.

Hermsen

J. M.

Ovadia

Vanneman

(2001). The glass ceiling effect. Social Forces, 80, 655-681.

19.

Cox

G. W.

Katz

J. N.

(1996). Why did the incumbency advantage in U.S. house elections grow? American Journal of Political Science, 40, 478-497.

20.

Dahlerup

(1988). From a small to a large minority: Women in Scandinavian politics. Scandinavian Political Studies, 1, 275-298.

21.

Darcy

Clark

Welch

(1994). Women, elections, and representation. Lincoln: University of Nebraska Press.

22.

Davidson-Schmich

(forthcoming). Gender quotas and democratic participation: Recruiting candidates for elective office in Germany. University of Michigan Press. 978-0-472-12161-8.

23.

Davis

R. H.

(1997). Women and power in parliamentary democracies: Cabinet appointments in Western Europe, 1968-1992. Lincoln: University of Nebraska Press.

24.

Driscoll

Krook

M. L.

(2009). Can there be a feminist rational choice institutionalism? Politics & Gender, 5, 238-245.

25.

Escobar-Lemmon

Taylor-Robinson

M. M.

(2005). Women ministers in Latin American government: When, where, and why? American Journal of Political Science, 49, 829-844.

26.

Esteve-Volart

Bagues

(2012). Are women pawns in the political game? Evidence from elections to the Spanish Senate. Journal of Public Economics 96, 387-399.

27.

Fenno

R. F.

(1973). Congressmen in committees. Boston, MA: Little, Brown.

28.

Feree

M. M.

Purkayastha

(2000). Equality and cumulative disadvantage: Response to Baxter and Wright. Gender & Society, 14, 809-813.

29.

Folke

Persson

Rickne

(in press). The primary effect: Preference votes and political promotions. American Political Science Review.

30.

Franceschet

Piscopo

J. M.

(2012). Gender and political backgrounds in Argentina. In Dahlerup

(Ed.), The impact of gender quotas (pp. 43-56). Oxford, UK: Oxford University Press.

31.

Geissel

Hust

(2005). Democratic mobilisation through quotas. Commonwealth & Comparative Politics, 43, 222-244.

32.

Gilljam

Karlsson

Sundell

(2010). Politik på hemmaplan. Tiotusen fullmäktige-ledamöter tycker om politik och demokrati [Local politics. Ten thousand municipal politicians opine about politics and democracy]. Stockholm, Sweden: SKL Kommentus AB.

33.

Hagevi

(2010). Anciennitetsprincipens uppgång och fall [The rise and fall of the seniority principle]. In Brotén

Holmberg

(Eds.), Folkets representanter: En bok om riksdagsledamöter och politisk representation i Sverige (pp. 241-265). Gothenburg, Sweden: University of Gothenburg.

34.

Heath

Schwindt-Bayer

Taylor-Robinson

(2005). Women on the sidelines: Women’s representation on committees in Latin American legislatures. American Journal of Political Science, 49, 420-436.

35.

Herrick

(1995). A reappraisal of the quality of women candidates. Women & Politics, 15(4), 25-38.

36.

Hirano

Snyder

J. M.

(2013). Primary elections and the quality of elected officials (Mimeo). New York, NY: Columbia University.

37.

Holgersson

(2003). Rekrytering av företagsledare: En studie i homosocialitet [Recruitment of business leaders: A study in homosociality]. Stockholm, Sweden: Ekonomiska forskningsinstitutet vid Handelshögskolan.

38.

Jalalzai

(2013). Shattered, cracked, or firmly intact? Women and the executive glass ceiling worldwide. New York, NY: Oxford University Press.

39.

Jalalzai

Krook

M. L.

(2010). Beyond Hillary and Benazir: Women’s political leadership worldwide. International Political Science Review, 31(1), 5-23.

40.

Järnbert

Wilén

(2013). Folkvaldas villkor i kommunfullmäktige [Municipal politician’s work conditions in the municipal councils] (Report from Statistics Sweden’s Division for Democracy Statistics). Stockholm: Statistics Sweden.

41.

Jonsson

(2003). Kommunstyrelseordförande: Kommunledare med politisk förankring [Municipal council board chair: Municipal leader with a political mandate]. Västra Frölunda, Sweden: Nya Doxa.

42.

Kahn

K. F.

Goldenberg

E. N.

(1991). Women candidates in the news: An examination of gender differences in U.S. Senate campaigns. Public Opinion Quarterly, 55, 180-199.

43.

Kanter

R. M.

(1977). Men and women of the corporation. New York, NY: Basic Books.

44.

Karlsson

Gilljam

(Eds.). (2014). Svenska politiker: Om de folkvalda i riksdag, landsting och kommuner [Swedish politicians: Elected representatives in parliament, counties, and municipalities]. Stockholm, Sweden: Santérus Förlag.

45.

Kenski

Falk

(2004). Of what is that glass ceiling made? A study of attitudes about women and the oval office. Women & Politics, 26(2), 57-80.

46.

Kittilson

M. C.

(2006). Challenging parties, changing parliaments—Women and elected office in contemporary Western Europe. Columbus: Ohio State University Press.

47.

Klahr

(1969). Decision making in a complex environment: The use of similarity judgments to predict preferences. Management Science, 15, 593-618.

48.

Kobayashi

(2004). Has the closed door opened for women? The appointment of women ministers in Japan. Political Science & Politics, 37, 63-64.

49.

Krook

M. L.

(2009). Quotas for women in politics: Gender and candidate selection reform worldwide. New York, NY: Oxford University Press.

50.

Krook

M. L.

O’Brien

(2012). All the president’s men? The appointment of female cabinet ministers worldwide. The Journal of Politics, 74, 840-855.

51.

Kropf

M. E.

Boiney

J. A.

(2001). The electoral glass ceiling: Gender, viability, and the news in US Senate campaigns. Women & Politics, 23, 79-103.

52.

Lipman-Blumen

(2000). Connective leadership. New York, NY: Oxford University Press.

53.

McKay

Krook

M. L.

(Eds.). (2011). Gender, politics and institutions: Towards a feminist institutionalism. Hampshire, UK: Palgrave MacMillan.

54.

Messner

Polborn

M. K.

(2004). Paying politicians. Journal of Public Economics, 88, 2423-2445.

55.

Montin

(2007). Moderna Kommuner [Modern municipalities]. Malmö, Sweden: Liber.

56.

Morgan

L. A.

(1998). Glass-ceiling effect or cohort effect? A longitudinal study of the gender earnings Gap for Engineers, 1982 to 1989. American Sociological Review, 63, 479-493.

57.

Murray

(2014). Quotas for men: Reframing gender quotas as a means of improving representation for all. American Political Science Review, 108, 520-532.

58.

Nielsen

(2001). På och av – om uppdragsvillighet, rekrytering och avhopp i den kommunala demokratin [On and Off—Motivation for public office, recruitment and resignations in the municipal democracy]. In Swedish Government (Ed.), Att vara med på riktigt. Bilagor till betänkande av Kommundemokratikommittén (pp. 17-108). Stockholm, Sweden: Fritzes.

59.

Nilsson

(2001). Den lokalpolitiska karriären: En socialpsykologisk studie av 20 kommunalråd [The local political career: A social psychology study of 20 full-time local politicians] (Doctoral dissertation). Lund University, Lund, Sweden.

60.

Norrander

Wilcox

(1998). The geography of gender power: Women in state legislatures. In Thomas

Wilcox

(Eds.), Women and elective office (pp. 103-117). New York, NY: Oxford University Press.

61.

Norris

(Ed.). (1997). Passages to power: Legislative recruitment in advanced democracies. Cambridge, UK: Cambridge University Press.

62.

Norris

Lovenduski

(1995). Political recruitment: Gender, race and class in the British parliament. Cambridge, UK: Cambridge University Press.

63.

Norton

E. C.

Wang

(2004). Computing interaction effects and standard errors in logit and probit models. Stata Journal, 4, 154-167.

64.

O’Brien

(2015). Rising to the top: Gender, political performance, and party leadership in advanced industrial democracies. American Journal of Political Science, 59, 1022-1039.

65.

O’Brien

Rickne

, (2014) Gender quotas and women’s political leadership. American Political Science Review.

66.

Öhberg

(2010). Politiker med karriärambitioner – en omöjlig självklarhet [Politicians with career ambitions—A contradiction.] (Doctoral dissertation). Gothenburg University, Gothenburg, Sweden.

67.

Ondercin

H. L.

Welch

(2005). Women candidates for congress. In Thomas

Wilcox

(Eds.), Women and elective office: Past, present, and future (2nd ed., pp. 60-80). New York, NY: Oxford University Press.

68.

O’Neill

Stewart

(2009). Gender and political party leadership in Canada. Party Politics, 15, 737-757.

69.

Palmer

Simon

(2001). The political glass ceiling: Gender, strategy, and incumbency in US House elections, 1978-1998. Women & Politics, 23, 59-78.

70.

Palmer

Simon

(2010). Breaking the political glass ceiling: Women and congressional elections. New York, NY: Routledge.

71.

Phillips

(1995). The politics of presence. Oxford, UK: Oxford University Press.

72.

Piliavin

(1987). Age, race, and sex similarity to candidate and voting preference. Journal of Applied Social Psychology, 17, 351-368.

73.

Prokos

Padavic

(2005). An examination of competing explanations for the pay gap among scientists and engineers. Gender & Society, 19, 523-543.

74.

Rodriguez

V. E.

(2003). Women in contemporary Mexican politics. Austin: University of Texas Press.

75.

Ryan

M. K.

Haslam

A. S.

(2005). The glass cliff: Evidence that women are over-represented in precarious leadership positions. British Journal of Management, 16, 81-90.

76.

Ryan

M. K.

Haslam

A. S.

(2007). The glass cliff: Exploring the dynamics surrounding the appointment of women to precarious leadership positions. Academy of Management Review, 32, 549-572.

77.

Sapiro

(1982). Private costs of public commitments or public costs of private commitments? Family roles versus political ambition. American Journal of Political Science, 26, 265-279.

78.

Schlozman

K. L.

Burns

Verba

Donahue

(1995). Gender and citizen participation: Is there a different voice? American Journal of Political Science, 39, 267-293.

79.

Schwindt-Bayer

L. A.

(2011). Women who win: Social backgrounds, paths to power, and political ambition in Latin American legislatures. Politics & Gender, 7, 1-33.

80.

Svaleryd

(2009). Women’s representation and public spending. European Journal of Political Economy, 25, 186-198.

81.

Thomas

Adams

(2010). Breaking the final glass ceiling: The influence of gender in the elections of Ellen Johnson-Sirleaf and Michelle Bachelet. Journal of Women, Politics & Policy, 31, 105-131.

82.

Thomas

(2002). The personal is the political: Antecedents of gendered choices of elected representatives. Sex Roles 47, 343-353.

83.

Trimble

Arscott

(2003). Still counting: Women in politics across Canada. Toronto, Ontario, Canada: University of Toronto Press.

84.

UN Women. (2014, March 11). Progress for women in politics, but glass ceiling remains firm [Press release]. Retrieved from http://www.unwomen.org/en/news/stories/2014/3/progress-for-women-in-politics-but-glass-ceiling-remains-firm#sthash.7FqvEsLe.dpuf

85.

U.S. Department of Labor. (1991). A report on the glass ceiling commission. Washington, DC: Author.

86.

U.S. Department of Labor. (1995). Good for Business: Making Full Use of the Nation’s Human Capital. Washington, DC: Author.

87.

Verge

(2010). Gendering representation in Spain: Opportunities and limits of gender quotas. Journal of Women Politics & Policy, 31, 166-190.

88.

Wide

(2011). Kvinnors vertikala marginalisering i kommunalpolitiken i Sverige [Women’s vertical marginalization in Swedish Municipalities] (Mimeo). Umeå, Sweden: Umeå University.

89.

Yule

(2000). Women councilors and committee recruitment. Local Government Studies, 26(1), 31-54.