The Inclusion of Women in Cabinets under Autocracy

We start with the ideal-type autocracy. Since elections only serve as pro forma acclamations, political competition is characterized by an evolving balance of power under conditions of anarchy (Svolik Reference Svolik2012). Therefore, the primary threats to the ruler come through extra-legal methods, either from the regime in the form of coups or from the broader civil society in the form of revolutions (De Mesquita et al. Reference De Mesquita, Smith, Siverson and Morrow2005; Geddes, Wright, and Frantz Reference Geddes, Wright and Frantz2014; Svolik Reference Svolik2012).

We expect the demand for women in cabinet to be low in autocracies. This is not to say that the underlying gender-related attitudes of the general public are antagonistic toward women. Instead, private attitudes are effectively crowded out by other issues that relate more directly to the leader’s survival. Therefore, autocrats must first and foremost appoint individuals who can help them survive threats.

Research in authoritarian politics has firmly established that coups by elites represent the most common threat that autocrats face (Svolik Reference Svolik2012). To keep the elite under control, the autocrat has an incentive to appoint loyal individuals (Egorov and Sonin Reference Egorov and Sonin2011) and co-opt potentially dangerous groups into organizations such as the cabinet, legislatures, and the party (Arriola, DeVaro, and Meng Reference Arriola, DeVaro and Meng2021; Arriola and Johnson Reference Arriola and Johnson2014; Blaydes Reference Blaydes2010; De Mesquita et al. Reference De Mesquita, Smith, Siverson and Morrow2005; Gandhi Reference Gandhi2008; Lust-Okar Reference Lust-Okar2006).

The autocrat can avoid a revolution by using repression (Gerschewski Reference Gerschewski2013). This causes them to rely on people who can keep “subversive” parts of the population subdued (Scharpf and Gläßel Reference Scharpf and Gläßel2020). This often involves using violence on a mass scale to subdue real or perceived enemies of the revolution (Kim Reference Kim2018). Due to historical processes and discrimination, men may be more readily available to occupy positions based on perceptions of credible threats of violence, making them more suitable (in the eyes of the leader) for governance in a repressive regime.

By contrast, organizations in which women have been more prominent, such as popular movements (Beckwith Reference Beckwith2001), typically exert power through nonviolent means and so cannot ransom the autocrat to the same degree. While we do not claim that women are intrinsically less violent than men, we contend that social pressures tend to restrict women from using violent tactics to achieve political ends to a greater degree than men (Best, Shair-Rosenfield, and Wood Reference Best, Shair-Rosenfield and Wood2019). Without being able to threaten violence credibly, we posit that women will be unable to extract concessions from the leader under an ideal-type autocracy. Additionally, public feminist movements are less prevalent to begin with in autocracies, given preference falsification of private attitudes in the face of social pressures (Kuran Reference Kuran1997).

Nonetheless, autocrats can have reasons to pick women to serve in the cabinet. They may want to positively signal to international donors, lenders, and investors by appointing women (Donno, Fox, and Kaasik Reference Donno, Fox and Kaasik2021; Fallon, Swiss, and Viterna Reference Fallon, Swiss and Viterna2012; Kroeger and Kang Reference Kroeger and Kang2022). Furthermore, as we argue later on, autocrats will have incentives to pick female leaders from popular movements to incorporate those groups into their support coalition if the regime evolves toward a more competitive form of authoritarianism. Despite these reasons, there is little pressure on the autocrat to select female ministers on the whole.

The main sources of potential cabinet ministers in autocracies are usually groups of inner-circle elites since these groups have resources and are willing to use violence to unseat the leader through a coup. Inner-circle elites typically come from the military, the ruling party, or the royal family (Barnes and O’Brien Reference Barnes and O’Brien2018; Gandhi Reference Gandhi2008; Geddes, Wright, and Frantz Reference Geddes, Wright and Frantz2018), all groups that are predominantly male. The military is the clearest example of an organization in which men dominate the upper echelons and upon which the autocrat relies to keep the population in line through repression (Barnes and O’Brien Reference Barnes and O’Brien2018; Svolik Reference Svolik2012).

Autocratic parties are designed to extend autocratic power and durability (Geddes Reference Geddes1999). They can do so by improving elite cohesion, monitoring citizens, providing patronage, co-opting opposition groups, and funneling state benefits to the elite (Magaloni and Kricheli Reference Magaloni and Kricheli2010; Miller Reference Miller2020; Slater Reference Slater2010). Therefore, they serve a very different purpose compared with parties in democracies where they are vehicles for electoral competition, meaning that parties need to be responsive and help mobilize voters (Aldrich Reference Aldrich1995; Downs Reference Downs1957). As a result, autocratic parties are less susceptible to change, and promotion within the party will be based on different criteria such as loyalty to the leader, willingness to commit violence, participation in corruption, and, to a lesser degree, competence (Egorov and Sonin Reference Egorov and Sonin2011; Scharpf and Gläßel Reference Scharpf and Gläßel2020). Autocratic parties have, therefore, in many cases devolved into de facto old boys’ clubs. While women can join authoritarian parties, they are often placed in legislative positions with little real influence or pigeonholed into gendered roles (Bjarnegård and Melander Reference Bjarnegård and Melander2013; Fallon, Swiss, and Viterna Reference Fallon, Swiss and Viterna2012; Harsch Reference Harsch and Smith2014; Stockemer Reference Stockemer2017; Tripp Reference Tripp2001).

Dynastic ties have been shown to help women access political office (Baturo and Gray Reference Baturo and Gray2018). This could be a factor in monarchies, where bloodline trump social forces, and the supply pool, therefore, are primarily the leader’s family. However, most of the world’s remaining monarchies are located in Middle Eastern countries (Gerring et al. Reference Gerring, Wig, Veenendaal, Weitzel, Teorell and Kikuta2021), where many regimes actively promote highly patriarchal norms, with some even continuing to bar women from public office at the turn of the century (Norris and Inglehart Reference Norris and Inglehart2001). As a result, it is reasonable to assume that the supply pools of ministerial candidates in autocracies are overwhelmingly male.

The Inclusion of Women in Cabinets in Democracies

In the ideal-type democracy, power is transferred according to the results of free and fair elections, and accountability is institutionalized (e.g., Dahl Reference Dahl1971).Footnote ³ Therefore, democratic leaders need to pick cabinet members who can contribute toward re-election, which may pressure the leader to select more women for four reasons.

First, the establishment of civil liberties that accompany democratization provides opportunities to politicize women’s interests. The full inclusion of women into formal democratic institutions requires a shift whereby women’s interests, previously constrained in the private sphere, are imagined as a matter of rights in the public political sphere (Rai Reference Rai1994). Civil rights counter the social pressures that lead to the falsification of true preferences (Kuran Reference Kuran1997), enabling feminist organizing (Wang et al. Reference Wang, Lindenfors, Sundstrom, Jasson, Paxton and Lindberg2017).

Second, the entrepreneurial force that materializes this fuel of (less organized) preferences is feminist activists (Annesley, Beckwith, and Franceschet Reference Annesley, Beckwith and Franceschet2019). Scholars studying cabinet selection processes in democracies demonstrate how gender in cabinets gains electoral salience as gender balance gets established as part of public perceptions of representational legitimacy (Annesley, Beckwith, and Franceschet Reference Annesley, Beckwith and Franceschet2019; Franceschet, Annesley, and Beckwith Reference Franceschet, Annesley and Beckwith2017, 235). While we do not suggest that democracies deterministically increase female cabinet ministers, we argue akin to Annesley, Beckwith, and Franceschet (Reference Annesley, Beckwith and Franceschet2019) that democratic elections establish “concrete floors” of cabinet participation through which incumbent and opposition leaders face punishment at the polls should they fall short. Moreover, as Annesley, Beckwith, and Franceschet (Reference Annesley, Beckwith and Franceschet2019) and other authors (e.g., Bauer and Darkwah Reference Bauer and Darkwah2022) also demonstrate, this concrete floor tends to rise as contenders for office propose more gender-inclusive cabinets. While democracies may be limited in enabling constituent principals to monitor their representative agents, they can facilitate the struggle for collective principles such as inclusion (Hayward Reference Hayward, Benhabib, Shapiro and Petranovich2007).

Third, once selectors use their power to include more women in cabinets, further electoral benefits arise in appointing women.Footnote ⁴ Women make up around 50% of the population in all countries and, as such, constitute the single biggest national cross-cutting cleavage (Teele Reference Teele2018). A group of studies examining gender effects in more generalized candidate choice scenarios has revealed that female candidates and politicians are just as favored or slightly more favored relative to their male counterparts regardless of the gender of the respondent (e.g., Bridgewater and Nagel Reference Bridgewater and Nagel2020; Dassonneville, Quinlan, and McAllister Reference Dassonneville, Quinlan and McAllister2021; Schwarz and Coppock Reference Schwarz and Coppock2022). Furthermore, women are perceived to be more competent in areas relating to social policy (Dolan Reference Dolan2010; Huddy and Terkildsen Reference Huddy and Terkildsen1993). A large body of research further suggests that significant differences in policy preferences exist between men and women (e.g., Aidt and Dallal Reference Aidt and Dallal2008; Kittilson Reference Kittilson2011) and that social policy becomes more balanced between these competing preferences when more women attain positions of political power (Atchison Reference Atchison2015; Clayton and Zetterberg Reference Clayton and Zetterberg2018; Kittilson Reference Kittilson2011). All this leads gender-parity cabinets to receive public praise and positive media coverage (Beckwith and Franceschet Reference Beckwith and Franceschet2022).

Fourth, the portfolio specificity of cabinet positions can insulate the executive arena from some of the demand-side social expectations that hold women back in arenas such as the legislature (e.g., Dolan Reference Dolan2010; Huddy and Terkildsen Reference Huddy and Terkildsen1993; Teele, Kalla, and Rosenbluth Reference Teele, Kalla and Rosenbluth2018). If gender-based social roles operate as a heuristic in the absence of candidate quality information (e.g., Alexander and Andersen Reference Alexander and Andersen1993), the specification of portfolio and expertise in cabinet positions may dampen the relevance of social roles, stereotypes, and statistical discrimination. In fact, examining party leaders, Dassonneville, Quinlan, and McAllister (Reference Dassonneville, Quinlan and McAllister2021) find that female party leaders are more favored than their male counterparts among experienced politicians.

Overall, there are substantial reasons to expect that democratic leaders have incentives to appoint women to the cabinet. This will particularly be the case where traditional gender norms and associated constraints have declined.

We also contend that democratic institutions foster conditions that result in an increased supply of female cabinet members. Cabinet ministers in democracies are usually selected from among members of parliament, party members, business leaders, or technocrats with expert knowledge (Escobar-Lemmon and Taylor-Robinson Reference Escobar-Lemmon and Taylor-Robinson2016). Although men have traditionally dominated these groups, they are more open to change than elite groups in autocracies, offering women more pathways to power. As a result, democracies are better than their autocratic counterparts at capturing developments at a mass level.

While democratic parties vary substantially to the degree to which they promote women (Caul Reference Caul1999), all significant parties are vote-seeking organizations (Aldrich Reference Aldrich1995; Downs Reference Downs1957).Footnote ⁵ Consequently, we expect them to be more open toward change, easier to join, and more responsive toward pressure for including underrepresented candidates compared with their autocratic counterparts. In short, we expect that they will include women to greater degrees.

Along with joining a political party, women can also be assigned cabinet appointments based on expertise in a given subject and can thus be drawn from academia, the civil service, or third-sector charities (e.g., Tripp Reference Tripp2001). Over the last decades, we have seen women increasingly access the labor force and, in many places, match or surpass the educational attainment of men. We also see developments at the elite level. Countries around the world are electing an increasing number of women to their parliaments (Wängnerud Reference Wängnerud2009) and women are increasingly obtaining leadership positions in the public sector, business, and academia (Huang et al. Reference Huang, Gates, Sinatra and Barabási2020). Therefore, there is a larger share of women in the supply pool of candidates in democracies relative to their autocratic counterparts, particularly in recent years.Footnote ⁶

Taken together, democracies hold more favorable conditions for women’s promotion to cabinet positions both through the voter-derived demand for balanced cabinets and the underlying composition of societal groups that typically make up the cabinet. Our main points are summarized in Table 1.

Table 1. Demand and Supply Forces in Cabinet Selection of Female Ministers

Transition, Consolidation, and Gender Equality

The previous sections argued that the nature of political competition in democracies should result in higher levels of female participation in government relative to autocracies. However, autocracies are increasingly adopting democratic institutions, blurring the line between the two regime types, meaning that the ideal types used to construct our theoretical argument rarely exist. Instead, we are seeing countries moving along a continuum (Levitsky and Way Reference Levitsky and Way2010). Furthermore, most countries have not been democratic throughout, but have instead experienced transitions from autocracy. While democratization opens a path for women to access political offices that were essentially built in their absence, new democracies and old democracies are not the same. Instead, new democracies have path dependencies from previous authoritarian regimes which hold back increased gender equality in government. We point to three factors that hold women back in new democracies.

First, a history of autocracy can depress the number of women in the pipeline to power (Lawless and Fox Reference Lawless and Fox2005). This is particularly challenging in new democracies where existing structures and patriarchal social norms may have stunted women’s potential political careers (Lee and McClean Reference Lee and McClean2022). In the early days of democracy, women will have rationally made life choices and arrangements that maximize their interests under the old, male-centric regime. Women might have falsified private preferences due to repression (Kuran Reference Kuran1997), calculated that pursuing a political career is too risky, and accrued sunk costs. In addition, women may not be connected to the political networks that serve as pathways to power, and political entrepreneurs may not approach women if they are not a part of the political system (Goyal Reference Goyal2019). However, certain components of democracies can decrease the costs facing women. For example, freedom of expression and freedom of organization can give women opportunities to speak about their preferences and drop pretenses of preference falsification. Moreover, continued experience of democracy generates more viability and trust in the institution, resulting in more women opting for careers in politics. In addition, democratization and political change enable women’s organizations to push for greater political representation at lower levels (Hughes and Tripp Reference Hughes and Tripp2015). Thus, although women often play a significant role in democratization and women’s organization is pivotal in putting pressure on the authoritarian regime, men tend to have a head start in the pipeline to power after democratization.

Second, authoritarian elites may cling to power despite democratization. Authoritarian elites often play a significant role in writing national constitutions before transitions occur in such a way that political competition is limited, albeit legally (Albertus and Menaldo Reference Albertus and Menaldo2018). Furthermore, authoritarian elites also secure positions of political power which limits the ability of a new political elite to gain power (Loxton and Power Reference Loxton and Power2021; Miller Reference Miller2021). A regime transition may not necessarily entail a transition of members of the elite, at least in the short run. Nevertheless, as authoritarian influence in executive politics wanes, the previous elite may give way to a new and potentially more gender-balanced elite.

Third, as the relative proportion of women increases in the pool of qualified ministerial candidates, a dynamic previously perceived as women’s underrepresentation may be reappraised as the overrepresentation of relatively less talented men (Murray Reference Murray2014). At the same time, voters may find it easier to punish the overrepresentation of less qualified men when the alternative (more qualified female candidates) becomes apparent. Furthermore, establishing and elevating concrete floors may reduce the political knowledge gap between men and women, particularly among younger cohorts less affected by authoritarian legacies (Dassonneville and McAllister Reference Dassonneville and McAllister2018), creating a virtuous cycle of younger voters challenging male overrepresentation in politics to a greater extent than voters of previous generations.

Consequently, democracy is, in the light of our research question, better captured as a historical phenomenon, taking place over the long run rather than as a level defined cross-sectionally (Gerring et al. Reference Gerring, Knutsen, Maguire, Skaaning, Teorell and Coppedge2020; Gerring, Thacker, and Alfaro Reference Gerring, Thacker and Alfaro2012). Thus, we expect the closest association between regime type and gender balance in cabinets to be in a country’s historical experience of democracy. As a result, we follow Gerring, Thacker, and Alfaro (Reference Gerring, Thacker and Alfaro2012) in operationalizing democracy as a “stock” variable, in essence adding up the years for which a country has had the experience of democracy. In summary, our main hypothesis is the following:

Measuring Women’s Representation in Cabinet

To test whether democracies select more women than autocracies, we use a new dataset on cabinet members, WhoGov (Nyrup and Bramwell Reference Nyrup and Bramwell2020). The dataset contains detailed information on cabinet members for every July in the period of 1966–2021 in all countries with a population of more than four hundred thousand citizens. In total, the dataset contains yearly information on 56,063 cabinet members in 177 countries, adding up to 8,814 country years, although not all are included throughout the analysis due to missingness on other variables. In addition to gender, the dataset includes information on the type of portfolio, enabling us to look at whether women gain access to high-prestige cabinet positions.

Throughout the analysis, we primarily rely on the share of female cabinet members as the dependent variable. To construct this measure, we select all full-ranking cabinet members and calculate the female share for every country in every year. Thus, we exclude junior ministers, the leader herself, and noncabinet officials when constructing the measure.

While the theory focuses on the share of female ministers, we triangulate our results using two alternative measures of female representation that are sensitive to the importance of the different ministries. Thereby, we are able to show that women are also qualitatively assigned more power in democracies.

First, we calculate the share of high-prestige ministerial posts occupied by women. We use the classification included in WhoGov as a baseline. Here, portfolios are divided into three levels. High-prestige portfolios are positions distinguished in terms of their visibility and significant control over policy. The minister of defense, finance, foreign affairs, and home/interior are included in this category. This also includes the deputy prime minister and, in presidential systems, the prime minister. Ministries that control significant resources, but have lesser status, are classified as medium prestige. These are, for example, agriculture, education, and transportation. Lastly, low-prestige positions are characterized by less resources and refer to ministries like youth, culture, and sports.Footnote ⁷ We make some exceptions to the general classification. For example, the minister of natural resources is considered highly prestigious in OPEC+ countries (including former members of OPEC). For more details on the coding, see Appendix A of the Supplementary Material. If the same person controls multiple portfolios or ministries, we only include the most prestigious position.

Lastly, we create a weighted share of female ministers. We give high-prestige portfolios a score of 3, medium-prestige portfolios a score of 2, and low-prestige a score of 1. Then we add all scores that are held by women in a given year and divide it by the total sum for a cabinet in a given year. An example is shown in Appendix B of the Supplementary Material.

Measuring Democracy

There are many indexes of democracy and the choice of an index can be important for results (Gerring et al. Reference Gerring, Knutsen, Maguire, Skaaning, Teorell and Coppedge2020; Munck and Verkuilen Reference Munck and Verkuilen2002). We rely on the V-Dem project’s Polyarchy measure, which is based on Dahl (Reference Dahl1971) and has five components that are combined into the Polyarchy scale, which ranges from 0 (most autocratic) to 1 (most democratic). We choose this measure because it has broad coverage (covering the same period as WhoGov, 1966–2021), distinguishes between democracies and autocracies based on both the contestation and participation criteria, is continuous (since our theory perceives regime types as levels rather than binary categories), is comparatively transparent, and is gathered in a rigorous manner (Teorell et al. Reference Teorell, Coppedge, Lindberg and Skaaning2019). Furthermore, we can deconstruct the measure, allowing us to look at different components of democracy.

To ensure that our results are not driven by the choice of democracy index, we triangulate the main findings using one additional continuous measure of democracy, namely Polity (Marshall, Gurr, and Jaggers Reference Marshall, Gurr and Jaggers2019), and two binary measures, namely Boix, Miller and Rosato (Reference Boix, Miller and Rosato2013) and the Democracy-Dictatorship Index (Cheibub, Gandhi, and Vreeland Reference Cheibub, Gandhi and Vreeland2010). The results are almost identical and can be found in Appendix C of the Supplementary Material.

To test our theory, we operationalize democracy both as a level and as a stock. Similarly to Gerring, Thacker, and Alfaro (Reference Gerring, Thacker and Alfaro2012), we create the stock measure of democracy by summing up each country’s score from 1900 (or, if the country gained independence after the year 1900, the first year) to the given year, applying an annual depreciation rate. We mainly rely on a depreciation rate of 95%, but vary it to show that the results are consistent.Footnote ⁸ Substantively, the stock variable implies that a country’s stock of democracy stretches back over time but that recent years receive more weight than distant years. We standardize the stock variable to a bound from 0 to 1 to ease interpretation and comparability. Thus, a score of 1 or close to 1 represents countries with the most democratic history, such as Sweden or Denmark, whereas the countries with a low stock of democracy, such as North Korea or Eritrea, get a score of 0 or close to 0. In Appendix D of the Supplementary Material, we provide simulations to give the reader an impression of how the stock of democracy varies dependent on the depreciation rate.

Estimation

The empirical analysis consists of both descriptive statistics and time-series cross-national estimations in which we regress the measures of female representation on measures of democracy, along with controls and fixed effects. We mainly focus on the determinants of the within-country variation over time in the share of female ministers, w, in country c at date t using a linear model. The estimated equation is

(1) $$ \begin{array}{rl}{w}_{ct}=\theta {d}_{ct-1}+\gamma {x}_{ct-1}+{\mu}_c+{\alpha}_t+{\eta}_{ct},& \end{array} $$

where $ {w}_{ct} $ is the share of female ministers in country c at year t. The main independent variable of interest is the lagged measure of democracy denoted by $ {d}_{ct-1} $ . $ {\mu}_c $ is a country-fixed effect, $ {\alpha}_t $ is a year dummy, $ \gamma {x}_{ct-1} $ are other time-variant controls lagged by 1 year, and $ {\eta}_{ct} $ is the error term. We cluster the standard errors by country to allow for arbitrary within-country correlations in the errors.

By including country and year dummy variables in all specifications, we control for fixed country characteristics, such as history and culture, which might affect the share of women in government, and global macro-trends, such as rising levels of women’s representation. In addition, we include a battery of controls. These are discussed and presented in the analysis. We refer to Appendix B of the Supplementary Material for a detailed description of each variable, Appendix E for links to all datasets used in the analysis, and Appendix F for descriptive statistics on all variables used in the analysis.

Democratization and the Proportion of Women in Government

Figure 2 plots the proportion of women in cabinet, the level of polyarchy, and the stock of polyarchy in 20 countries that have democratized since 1966. The countries are selected to secure diversity in geography, time of democratization, and prior type of authoritarian regime. In many countries, for example, Argentina, Bulgaria, and Indonesia, we see a rapid increase in the polyarchy score but a comparatively slow increase in both the stock of polyarchy and the proportion of female ministers. This suggests that the effect of democracy on the share of female cabinet members is not immediate but rather the result of a process where women gradually gain access to government.

Figure 2. Share of Women in Cabinet during Democratization across 20 Countries

Next, we test this systematically. The results are reported graphically in Figure 3. The figure shows the regression coefficients when we regress the three measures of female government representation on both the level and the stock of Polyarchy.

Figure 3. Regression Analysis on the Three Measures of Female Representation

Note: Independent variable is level or stock of Polyarchy. The models are based on an OLS with country-clustered standard errors. The thick bars show the 90% confidence intervals, whereas the thin bars show the 95% confidence intervals. The measure of Polyarchy is lagged by 1 year and no other controls are included. For full model results, see Supplementary Table G1 for “Share female” ( $ N=8,388 $ country years), Supplementary Table G2 for “Share female weighted” ( $ N=8,388 $ country years), and Supplementary Table G3 for “Share female high prestige” ( $ N=8,365 $ country years).

In the first model, we include only the level of democracy. Here, we see that democracy is significantly and positively correlated with women’s representation in government. This finding is independent of how we measure female representation. Substantively, we see that the most democratic country, on average, has 19% more female ministers than the least democratic.

In the second model, we include year-fixed effects. This takes into account longitudinal trends. Most importantly, we have seen an increase in both the number of democratic countries and the proportion of women in government across the world. Not surprisingly, including year-fixed effects decreases the size of the coefficient, albeit remaining statistically significant. Thus, women’s higher representation in government in democracies is not just an artifact of the rise of democracy (Huntington Reference Huntington1993; Teorell et al. Reference Teorell, Coppedge, Lindberg and Skaaning2019).

Third, we include country-fixed effects and, therefore, take into account time-invariant country-specific variables such as history or culture, meaning that we only look at variation in women’s representation and democracy within the same country. This diminishes the size of the coefficient, causing the effect of democracy to lose significance. Thus, we do not find that a country, on average, has significantly more women in government when it has a higher level of democracy.

However, when we operationalize democracy as a stock in the fourth model, we find that democracy is highly predictive of the share of women in government. In other words, when a country has more experience with democracy, the share of women in government increases. Furthermore, as shown in model 5 relative to model 4, the effects are larger as we place more emphasis on the past by reducing the depreciation rate. When using the 95% stock measure, a country with the democratic history of Denmark (in 2021) is predicted to have a 16% higher proportion of female cabinet members, relative to a country with the “democratic” history of North Korea, all else being equal.

We find that the results are substantively similar when using alternative measures of women’s representation that take into account the importance of the portfolios women are assigned to, meaning that women are assigned more portfolios, but also more important portfolios, when a country has more experience with democracy.

Testing for Potential Confounders

Many other factors could confound the relationship, even when including two-way fixed effects. We may particularly worry about time-variant variables, such as economic development, or other stock variables, such as GDP per capita. To counter some alternative explanations, we run a number of alternative specification tests in Table 2. Below, we only show the results for the share of female ministers, but in Appendixes H and I of the Supplementary Material, we find that the results are comparable when using the weighted share of female ministers and the share of female ministers in high prestige positions.

Table 2. Specification Tests

Note: All right-side variables lagged by 1 year. Country-clustered standard errors in parentheses. Estimator: OLS (ordinary least squares). FE = fixed effects (country and year), RE = random effects. * $ p<0.05 $ .

Model 1 shows the base model, where we only include the measure of the stock of democracy with country- and year-fixed effects. In model 2, we include the year as a trend instead of using fixed effects. This is another way of controlling for the effect of time by controlling for possibly spurious correlations between the measure of democracy and any similarly-trended independent variables. Using this measure of time increases the coefficient for the stock of democracy relative to the benchmark. Model 3 includes several time-varying measures of economic growth, such as GDP per capita and urbanization. Most importantly, we also include the log of GDP per capita, which also can be perceived as a stock measure for the level of development.Footnote ⁹ Likewise, we include measures of human development in model 4. Democracy may be related to both economic development (Acemoglu et al. Reference Acemoglu, Naidu, Restrepo and Robinson2019; Przeworski et al. Reference Przeworski, Alvarez, Cheibub and Limongi2000) and human development (Gerring, Thacker, and Alfaro Reference Gerring, Thacker and Alfaro2012). Furthermore, economic growth is known to be related to women’s representation (Matland Reference Matland1998). Coefficient estimates for the stock of democracy are stable across these tests and comparable in magnitude. Thus, the finding is not an artifact generated by increases in human development or economic capacity in democratizing countries and our findings are, therefore, not due to modernization (Lipset Reference Lipset1959).

In model 5, we control for other measures of women’s empowerment, such as women’s political empowerment index and the share of female legislators. These controls can, to some degree, be considered “bad controls,” since, as discussed in the theory, it is likely that the effect of democratization runs through these variables. However, we want to ensure that the findings are not caused by women’s general political empowerment in democratizing countries, but that there is something specific about democracy. We find that the stock of democracy remains significant and that the share of female legislators and women’s political rights are related to more female cabinet ministers. In contrast, the women’s political empowerment index is, surprisingly, related to fewer women in government. It should be noted that the controls for women’s empowerment are highly correlated and we should, therefore, be cautious when interpreting them in the same regression.

Model 6 includes several indices that focus on institutional capacity from the V-Dem dataset, such as property rights, the rule of law, judicial constraints, and party institutionalization, and shows a similar estimate for the measure of democracy. This suggests that the results are driven by the democratic feature and not other capacity-related features of institutions. Including these indices also mitigates a potential threat to inference stemming from country experts assigning a high score to a country along some institutional parameter during a period in which a country has more female ministers. If so, the relationship could be spurious and a product of coding circularity (Bizzarro et al. Reference Bizzarro, Gerring, Knutsen, Hicken, Bernhard, Skaaning and Coppedge2018).

In model 7, we include a lagged dependent variable. The coefficient for the lagged dependent variable is large (75.05), which is not surprising given that a country with more female ministers in one year will most likely have more ministers in the next year. The inclusion of the lagged dependent variable means that the coefficients for the stock of democracy capture only the short-term effects and we only estimate how the stock of democracy at $ t-1 $ affects the share of female ministers from $ t-1 $ to t. We find a strong, positive, and significant effect of the stock of democracy on the share of female ministers in the short run.

Lastly, in model 8, we include a number of fixed covariates to the benchmark model, such as ethnic fractionalization, percent Muslim, and the continent. These variables, which change little or not at all across the period of observation, lead us to replace country-fixed effects with a random-effects model. The coefficient for the stock of democracy remains significant in this specification.

While we include a long range of tests in Table 2, the list is not exhaustive. Hence, we include further robustness tests in the Supplementary Material. In Appendix J of the Supplementary Material, we show that the longer a view we have on history, that is, a lower deprecation rate, the stronger the association between the stock of democracy and the share of female ministers. Furthermore, we show that the results are comparable across different measures of democracy in Appendix C of the Supplementary Material, both when we use dichotomous measures, namely Boix, Miller and Rosato (Reference Boix, Miller and Rosato2013) and the DD index (Cheibub, Gandhi, and Vreeland Reference Cheibub, Gandhi and Vreeland2010), and another fine-grained measure, Polity IV (Marshall, Gurr, and Jaggers Reference Marshall, Gurr and Jaggers2019). In addition, we find that the results are similar when using imputed data for missing values, as seen in Appendix K of the Supplementary Material.

The analysis supports our argument that it is not the level of democracy that explains women’s access to public office but rather the historical experience with democracy. However, to strengthen the credibility of our theory, we proceed by testing further implications of the theoretical model.

Dividing Democracy into Its Composite Measures

In Figure 4, we split the measure of Polyarchy into its composite measures and run the models separately for the stock of each measure, including country- and year-fixed effects.

Figure 4. Which Component of Polyarchy Best Predicts the Share of Female Ministers?

Note: Share of female ministers. Based on separate models using an OLS regression with country- and year-fixed effects. Country-clustered standard errors in parentheses. The thin bars show the 95% confidence intervals, whereas the thick bars show the 90% confidence intervals. The measure of democracy is lagged by 1 year and no other controls are included. For full model results, see Supplementary Table L1: column 1 for Stock of Clean Elections ( $ N=8,144 $ ), column 2 for Stock of Freedom of Expression ( $ N=8,304 $ ), column 3 for Stock of Associational Autonomy ( $ N=8,365 $ ), column 4 for Stock of Elected Officials ( $ N=8,082 $ ), and column 5 for Stock of Suffrage ( $ N=8,255 $ ).

The measures of clean elections and freedom of expression are significantly and positively related to the share of female ministers, whereas associational autonomy is positive but outside conventional bounds for statistical significance. On the other hand, the measures of elected officials and suffrage are negative. At first, these results may seem puzzling, but elected officials and suffrage can be seen as minimum requirements for democracy and, therefore, “hollow” measures of democracy. A country can have sham elections with full suffrage, causing it to score high on these two measures, while having limited or no competition for power in reality. Meanwhile, the three measures focusing on whether the leader loses (at least some) control over the electoral process and, as a result, faces electoral competition, are all positive. This indicates that only real electoral competition changes the motives for selecting female ministers. This is in line with our theory and supports the idea that competitive elections and “real” democracy are important for the inclusion of women in government.

Transition to and from Democracy

Next, we investigate the implications of transitions to and from democracy on the share of female ministers when using a binary measure. This can be seen as a more general test of our main argument. Following our theoretical argument, we would not expect to see a drastic increase in the share of women being included in government immediately after democratization. Instead, the increase should be slow and gradual.

We test this in Figure 5, where we compare the share of female ministers before and after a country democratizes according to Boix, Miller, and Rosato (Reference Boix, Miller and Rosato2013). We use the share of ministers relative to the year of a democratic transition (year 0) to better take into account that democratization is clustered in time (Huntington Reference Huntington1993) and to better isolate the “effect” of democratization. As predicted, there is no sudden, large jump in the proportion of female ministers immediately following democratization, albeit we do see a small and significant increase in the share of female ministers in the year of democratization compared with the preceding autocratic period. Interestingly, we see little to no increase in the share of female ministers in the years immediately after democratization. However, four years after democratization, the share of female ministers starts increasing markedly, and eight years after democratization, we find that there is a significantly larger proportion of the cabinet that is female relative to the year of democratization. We may see these jumps because of four-year election cycles.

Figure 5. Democratic Transitions and the Share of Female Ministers

Note: The share of female ministers relative to the year of a democratic transition according to Boix, Miller, and Rosato (Reference Boix, Miller and Rosato2013) which have been merged with WhoGov, so the year of democratization corresponds to the first democratic cabinet. Ninety democratizations are included in the analysis. See Supplementary 96 M1 for full model results.

Split Samples

In the last section of the analysis, we investigate the relationship between regime type and female cabinet representation in subsets of the full sample to explore mechanisms further. The results are shown in Figure 6.

Figure 6. Split Sample Tests

Note: Dependent variable is the share of female ministers. Independent variable is the stock of Polyarchy. The models are based on an OLS with country-clustered standard errors. The thick bars show the 90% confidence intervals, whereas the thin bars show the 95% confidence intervals. The measure of stock of Polyarchy is lagged by 1 year and no other controls are included. For full model results, see Supplementary Table N1: column 1 for OECD ( $ N=1,734 $ ), column 2 for Non-OECD ( $ N=6,654 $ ), column 3 for GDPpc $ >7,000 $ ( $ N=4,000 $ ), column 4 for GDPpc $ <7,000 $ ( $ N=3,616 $ ), column 5 for Pre-1993 ( $ N=3,551 $ ), and column 6 for Post-1993 ( $ N=4,837 $ ).

We begin by dividing the sample into countries in the OECD (in 2018) and those that are not. Countries in the OECD are mainly established democracies integrated in the global economic system and have no recent history of being colonized. Furthermore, these countries have received the bulk of scholarly attention. We do not find conclusive evidence of a significant relationship between the stock of democracy and the share of female ministers in OECD countries. One possibility is that the result lacks statistical power because most OECD countries are older democracies (and, therefore, at the higher end of the global sample of stock of democracy). Thus, our independent variable may not capture enough variation within this narrower subset of well-established democracies. Here, other explanations, such as Franceschet, Annesley, and Beckwith’s (Reference Franceschet, Annesley and Beckwith2017) model of formal and informal selection rules, may be better at capturing the nuanced ways in which different types of established democracies continue on their paths to gender parity. In comparison, we see that the results are strongly significant in non-OECD countries, where most countries range from the lower to high-intermediate zone of global democratic stock. This suggests that our theoretical model is particularly well-suited to explain variation in this group of emerging democracies that are often overlooked in the literature.

Subsequently, we run the analysis within a sample of richer countries (GDP per capita $ >\$7,000 $ ) and poorer countries (GDP per capita $ <\$7,000 $ ) (in the given year) and find that democratic experience matters more in poorer countries. This reiterates that the finding has little to do with the level of development and highlights that democracy is of extra relevance in poorer countries for helping women access positions of power.

In the last two columns, we find that the point estimate is higher after the fall of the Soviet Union (1993–2021) than during the Cold War (1966–92). The fall of the Soviet Union was followed by democratic transition and consolidation in Africa, Eastern Europe, and elsewhere. Furthermore, there has been a higher demand for women in government in more recent years due to the decline in traditional gender norms, which, as discussed in the theory, should particularly have an impact on democracies. Lastly, some governments try to cater to international organizations and donors by appointing more women. For example, this was the case for a group of post-communist countries trying to obtain EU membership (Bego Reference Bego2014). Combined, these factors push the estimate upward in the period after the Cold War.