PRQXXX10.1177/1065912912 449698Political Research QuarterlyRosen

Political Research Quarterly 66(2) 306­–321 The Effects of Political Institutions on © 2011 University of Utah Reprints and permissions: sagepub.com/journalsPermissions.nav Women’s Political Representation: DOI: 10.1177/1065912912449698 A Comparative Analysis of 168 prq.sagepub.com Countries from 1992 to 2010

Jennifer Rosen1

Abstract Women’s political representation exhibits substantial cross-national variation. While mechanisms shaping these variations are well understood for Western democracies, there is little consensus on how these same factors operate in less developed countries. Effects of two political institutions—electoral systems and gender quotas—are tested across 168 countries from 1992 to 2010. Findings indicate that key causal factors interact with a country’s socioeconomic development, shifting their importance and possibly even direction at various development thresholds. Generalizing broadly across countries, therefore, does not adequately represent the effects of these political institutions. Rather, different institutional changes are advised to increase women’s presence in national governments.

Keywords women and politics, development, quotas, electoral systems, political representation

Women have secured the legal rights needed to partici- A rich tradition of scholars have analyzed political, socio- pate in politics in all but a handful of countries; however, economic, and cultural factors to explain the substantial their representation in national politics lags far behind cross-national variations in women’s representation, gener- their proportion in society. In mid-2011 women held 19.6 ally finding that level of development does not play percent of single/lower house parliamentary seats inter- a significant role in these explanations. This makes sense if nationally, 23 of the 189 highest ranking United Nations you consider that, as of early 2012, Rwanda, Cuba, diplomatic posts, and just 14 women held elected posts as Seychelles, South Africa, and Nicaragua ranked among the heads of state. The circumstance in many countries has top ten countries by percentage of women in lower/single mirrored the slow progression of women’s parliamentary houses of parliament, while women in countries such as representation globally (12 percent in 1985, 16.4 percent Britain, France, the United States, and Japan occupied rela- in 2005). From 1992 to 2011, women’s parliamentary tively low proportions of parliamentary seats (ranking 54th, representation in Brazil increased from 7.4 to 8.6 percent, 70th, 78th, and 106th respectively). I argue, however, that in India from 6.9 to 10.8 percent, in Russia from 8.7 to 14 level of development- when measured using economic and percent, and in the United States from 11 percent to a social proxies- interacts with key causal variables, shifting scarcely more impressive 16.8 percent. On the other hand, their importance and possibly even direction once a given countries such as Argentina, Costa Rica, Tanzania, and level is reached. In 1998 Richard Matland introduced the South Africa have seen women’s parliamentary represen- concept of a development threshold, a minimum level of tation nearly triple over the last two decades. In fact, development that is required in order for the mechanisms Rwanda’s 2008 parliamentary elections resulted in the found to be important by previous researchers to positively world’s first national legislative body with a female major- affect women’s political representation. To test Matland’s ity (56.3 percent). While women’s political advancement outside the industrialized world has accelerated in recent 1Northwestern University, Evanston, IL, USA years, research either has been limited to Western democ- Corresponding Author: racies or has presented contradictory conclusions regard- Jennifer Rosen, Northwestern University, Department of Sociology, ing how key causal mechanisms operate in the context of 1810 Chicago Avenue, Evanston, IL 60208. less developed countries. Email: [email protected] Rosen 307 hypothesis I highlight two political variables—electoral sys- 1990; Moore and Shackman 1996). Given these consid- tems and national gender quotas. In-depth examination of erable implications, it is important to ask which, if any, these variables offers particular significance because, as institutional mechanisms support increases in women’s political institutions, they are more easily manipulated than national political representation. ideological or socioeconomic factors to achieve the goal of gender parity in national politics. Results support Matland’s basic insight of a development threshold, indicating that Political Institutions across broad cross-national generalizations fail to adequately repre- Development Thresholds sent the effects of these political institutions. Research examining cross-national variations in women’s I assemble an original cross-national time-series data political representation has proliferated over the past few set for 168 countries from 1992 to 2010. Although I build decades (Hughes 2009; Kenworthy and Malami 1999; models that comprehensively analyze cross-national Krook 2010; Matland 1998; Moore and Shackman 1996; variations in women’s political representation, this is not Paxton 1997; Paxton and Hughes 2007; Paxton and my primary goal. Instead, I provide rigorous empirical Kunovich 2003, 2005; Reynolds 1999; Rule 1987; evidence to support Matland’s hypothesis of a develop- Viterna, Fallon, and Beckfield 2008). However, with a ment threshold, revealing how the causal processes of few important exceptions (Hughes 2009; Kenworthy and particular political institutions vary across countries Malami 1999; Krook 2010; Matland 1998; Viterna, depending on their level of socioeconomic development. Fallon, and Beckfield 2008), quantitative results rarely I demonstrate this in two ways. First, I replicate aspects differentiate between mechanisms at play across levels of of previous studies, but focus considerable attention on development. In fact, despite their universalizing asser- the ways in which electoral systems and national quotas tions, statistical results can suffer from regional biases differ by including interaction terms with level of devel- (often privileging Western democracies) rather than opment. Subsequently, I treat development as a scope describing global trends (Krook 2010). Consequently, condition, where countries are analyzed separately while the mechanisms shaping women’s representation in according to level of socioeconomic development. I con- Western, industrialized countries are generally agreed clude by arguing that, while most women in politics upon, there is little consensus regarding how these same researchers have generalized their findings across entire factors operate in the context of less developed countries populations of countries, the same mechanisms that increase (Hughes 2009; Krook 2010). Below, I consider what we the percentage of women in parliaments in developed do and do not know about the impact of key causal countries do not have an equivalent effect in less developed mechanisms across development thresholds. countries and vice versa. This empirical analysis provides evidence to suggest that the majority of existing models need to be reframed to account for the heterogeneity intro- Women’s Political Representation: Does duced by level of development. By presenting novel Development Matter? analysis from an original data set, I am also able to engage Cross-national analytic models have produced disparate the contradictory findings of the few studies that specifi- and oftentimes contradicting conclusions when applied to cally examine the causal mechanisms of women’s politi- less developed countries (Hughes 2009). Socioeconomic cal representation in less developed countries. development affects resource availability and personal The political advancement of women has broader opportunities and directly affects a population’s well- implications beyond the formal significance of justice being. Many believe that as the standard of living improves, and equality. Achieving gender balance in national gov- women’s roles revolve less around tasks necessary for ernments can measurably improve the quality of policy daily survival (i.e., collecting water), which enables making, both in relation to women’s specific needs and greater independence and involvement in the public sphere to the policy-making process more generally (Carroll (Krook 2010; Paxton and Kunovich 2005). As such, a 2001; Celis et al. 2008; Dahlerup 2010; Lovenduski and developed economy can create more opportunities for Norris 2004; Phillips 1995; Tolleson-Rinehart 2001). women to acquire the qualifications needed to serve in top Women’s political visibility is also theorized to have a government posts. Conversely, others argue that women symbolic effect, potentially increasing women’s status tend to enter politics during times of crisis, so that as eco- and reducing societal gender inequalities more broadly nomic distress becomes more severe women get involved (Hughes 2009; Paxton 1997; Paxton and Hughes 2007). in politics to advocate for changes that will ameliorate In fact, some scholars argue that gender parity in poli- their immediate situations (Del Campo 2005). Clearly, that tics is among the most important changes required to Rwanda and rank 1st and 2nd in women’s parlia- produce a system of in society (Chafetz mentary representation, respectively, suggests that one 308 Political Research Quarterly 66(2) cannot assume that women do better in Western, industri- factors, extensive theoretical revisions within the litera- alized nations, nor is it reasonable to suppose that the ture would be suggested. factors aiding women in Sweden are the same as those Four more recent studies have followed Matland’s advancing women in Rwanda. example of using development as a scope condition. Previous researchers generally use per capita GDP Kenworthy and Malami (1999) apply established models or energy use as proxies for a country’s level of develop- across a sample of 20 developed democracies and a diverse ment and find no direct, linear relationship to the group of 146 countries. While their results show few dif- percentage of women in parliament (Hughes 2009; ferences between the two samples, they do not analyze less Kenworthy and Malami 1999; Moore and Shackman developed countries in isolation. Viterna, Fallon, and 1996; Oakes and Almquist 1992; Paxton 1997; Viterna, Beckfield (2008) replicate Kenworthy and Malami’s anal- Fallon, and Beckfield 2008). This makes sense if one yses, but extend the models to a subset of 85 developing considers that some industrialized countries, such as the countries (defined by GDP per capita). The authors con- United States and France, have relatively low female clude that while level of development does not directly representation, while women in some poorer countries, affect women’s parliamentary representation, key causal such as Angola and Tanzania, occupy relatively large variables do vary significantly across samples. Hughes proportions of parliamentary seats. Matland (1998) and (2009) runs separate analyses across 36 high-income, 86 Krook (2010) use more nuanced measures of develop- middle-income, and 63 low-income countries, concluding ment that account for both social and economic factors.1 that previous models are less adequate when applied to Relying solely on economic performance overlooks the low-income countries. Finally, Krook (2010) uses QCA to fact that resources are oftentimes distributed unequally. analyze 22 Western democracies and 26 sub-Saharan Instead, a mixture of both economic and social proxies African countries, also finding that the causal importance provides a more accurate representation of a popula- of certain mechanisms varies between the two groups. tion’s well-being in a way that aggregated economics While these studies make progress by isolating the causal alone cannot (Inglehart and Norris 2003), which becomes mechanisms operating in less developed countries, they all more significant when looking at outcomes concerning arrive at different conclusions. social (in)equalities. Development has occasionally been treated as a “scope condition,” where separate populations are defined Electoral Systems: The Gender Advantage according to level of development. This method allows of Proportional Representation scholars to isolate those factors operating in a specific set A proportional-representation (PR) electoral system of countries, and determine whether or not the same has been highlighted by most scholars as a fundamental mechanisms work in similar fashions across groups of factor predicting cross-national variations in women’s countries (Mahoney and Goertz 2004). Matland (1998) parliamentary representation (Kenworthy and Malami was the first to apply existing models to separate samples 1999; Matland 1998; Paxton 1997; Reynolds 1999; Rule of twenty-four OECD and sixteen less developed coun- 1987; Viterna, Fallon, and Beckfield 2008). In these tries. His results show that the causal mechanisms empha- systems, votes are cast for political parties or a party’s sized by previous researchers, most notably electoral list of candidates, and the seats won is proportionate to system, only become significant as a country’s level of the percentage of votes received (e.g., if a party receives development increases. Unfortunately, because of limited 30 percent of the votes, they fill 30 percent of parliamen- sample size and lack of regional variation, his findings do tary seats). In contrast to majoritarian systems where not hold up to tests of robustness. Additionally, on the parties put forward a single candidate and therefore have basis of the time period of analysis, Matland cautions that an incentive to appeal to the lowest-common denomina- female representation may be too low in less developed tor of voter preferences, parties in PR systems field countries to discern broad patterns influencing representa- multiple candidates and benefit by appealing to a diverse tion. Today, levels of representation have soared in Africa cross-section of voters. According to Paxton (1997), a and Latin America, making it an apt moment to revisit his PR system may buffer women by making them less risky analysis. Matland’s study is also limited in the range of candidates in comparison to a majoritarian system where independent factors he considers. The proliferation of female candidates stand in direct competition with men. gender quotas since 1998 enables an investigation of this However, not all PR systems select candidates in the important equalizing mechanism, and comparative data same way. Legislators are chosen in the order they appear also facilitates the inclusion of more nuanced measures of on candidate lists, so the position of female candidates the electoral system. However, if Matland’s basic insight matters (Paxton and Hughes 2007). In closed-list PR sys- of a development threshold can withstand more rigorous tems, political parties decide the candidates’ position testing using a broader sample and larger set of causal on the list, while in open-list systems voters influence Rosen 309 candidate location. Evidence has been mixed as to which Reynolds (1999) and Paxton and Kunovich (2005) system is better for women since it depends on whether find that quotas do not have a statistically significant political parties or voters are more likely to advance effect on women’s representation. However, neither study women (Paxton and Hughes 2007). captures the time period when the most quotas were While the positive impact of PR systems on female adopted (1995-2004). Tripp and Kang (2008) and Hughes representation has become a sort of dogma in the litera- (2009) use more recent data in their analyses and arrive at ture (Tremblay 2008), results vary when it comes to the different conclusions regarding the measure’s statistical generalizability of these findings for less developed importance. While Hughes (2009) does not find statistical countries. Kenworthy and Malami (1999) and Viterna, significance, Tripp and Kang (2008) do; however their Fallon, and Beckfield (2008) find that PR systems have a measure combines both nationally mandated and volun- positive effect on women’s representation in all countries, tary party quotas while Hughes controls only for national while Matland (1998), Hughes (2009), and Krook (2010) quotas. Krook (2010) measures quotas based on either find that its significance only holds up in Western, indus- national codification or voluntary adoption by a major trialized democracies. This discrepancy suggests that the political party. She finds that it is neither a necessary nor effects of what is arguably the most important causal a sufficient factor leading to increases in women’s rep- mechanism are not fully understood and may operate dif- resentation in Western countries, but appears in three of ferently across groups of countries. the four QCA configurations in sub-Saharan African countries. Despite this discord among quantitative studies, sev- Quotas: National Mandates of Gender Parity? eral qualitative analyses show that quotas may be espe- Twenty years ago only a handful of countries had legis- cially important in emerging democracies and post-conflict lation requiring a minimum number of female parlia- societies (Bauer and Britton 2006; Dahlerup 2006; Krook mentarians, whereas in recent years over 50 countries 2006). In less developed countries equality between the have adopted such legislation (Krook 2006; Tripp and sexes is often seen as an outcome that must be instituted, Kang 2008). Gender quotas act as affirmative action rather than a product of greater development (Dahlerup policies intended to help women overcome obstacles to 2006; Dahlerup and Friedenvall 2005). Today advocates their election such as less political experience, cultural are less willing to wait the half century it took women to stereotypes, and/or incumbency. There are a variety of make gains subscribing to the “incremental approach” quotas that either reserve a set percentage of parliamen- taken in Scandinavia and many Western European coun- tary seats or mandate a minimum number of candidates tries (Dahlerup and Friedenvall 2005). Quotas tend to be that parties must fill with women. Many countries, particularly popular in emergent democracies and post- especially Western, industrialized democracies, have conflict societies because they confer an element of inter- opted for voluntary party quotas rather than those estab- national legitimacy, but also because there is a political lished by national legislation. These are not legally opening for quotas since these countries often begin from binding, but instead are regulated internally by indi- scratch designing political institutions and drawing up vidual parties. Nationally mandated quotas, on the other new constitutions (e.g., Afghanistan) (Bauer and Britton hand, are codified in either electoral or constitutional 2006; Dahlerup 2006; Krook 2006; Matland 2006; Tripp, law and are often accompanied by sanctions for Konate, and Lowe-Morna 2006). Conversely, across non-compliance. National quotas can address various developed democracies the call for quotas has most often stages of the nomination process, either regulating the originated from women’s groups inside political parties, number of women elected (reserved seat quotas) or culminating in internally regulated party quotas rather controlling women’s inclusion and/or placement on than national legislation (Dahlerup 2006; Celis et al. 2008; political parties’ candidate lists (legal candidate quota). Stevens 2007). Moreover, these countries often introduce Although national quotas encompass a diverse range of quotas after women have already attained 20 to 30 percent policies, leading experts in the field argue that all elec- of parliamentary seats, thereby minimizing their effects toral and constitutional quotas represent a nation’s (Dahlerup 2006). The popularity of gender quotas, there- attempt to “fast track gender balance in politics” fore, may be related to political transitions to democratic (Dahlerup 2006, 3), and therefore fall under the umbrella regimes, which encourages international acceptance and of national gender quotas. Quota legislation is generally provides an opportunity for women’s movements to laden with issues of compliance; however, women seek- become involved in national politics (Dahlerup 2006; ing parliamentary representation in countries with Hughes 2009). Thus, there are both external (desire for national mandates tend to have better outcomes com- international legitimacy) and internal (e.g., timing, con- pared to those with voluntary party quotas (Dahlerup flict, women’s movements) reasons why quotas are more 2006; Tripp and Kang 2008). likely to be adopted in less developed countries. It is also 310 Political Research Quarterly 66(2)

worth noting that larger statistical increases are possible in Women’s Parliamentary Representation less developed countries given that they begin, on aver- age, with low absolute levels of women’s representation. The dependent variable is the percentage of seats occupied by women in single or lower houses of a country’s main legislative body. Percentages range between 0 and 56.3 Data and Methods percent, with a mean of 10.8 percent throughout the time Using ordinary least squares (OLS) regression I exam- span. Data were collected from the Inter-Parliamentary ine political, socioeconomic, and cultural determinants Union and the Women in Parliament Dataset (Paxton, of cross-national variations in women’s parliamentary Green, and Hughes 2008). The outcome, therefore, con- representation across 168 countries from 1992 to 2010. cerns the number of women in parliament rather than the I also apply models across samples of developed, devel- effects of having more women in parliament (descriptive vs. oping, and least-developed countries, which allows me substantive representation). This article, however, is predi- to assess the effects of key causal mechanisms across cated on the belief that greater legislative participation of countries with varying levels of development. Data women at least increases the possibility that women’s inter- were compiled from a variety of sources including the ests will be better represented (Carroll 2001; Celis et al. United Nations, the World Bank, the Inter-Parliamentary 2008; Dahlerup 2010; Phillips 1995; Wängnerud 2009). Union, the CIA World Fact Book, the World Health Organization, and the Women’s Indicators and Statistics Database. Socioeconomic Development The robustness of my findings is verified through the Of primary importance to this analysis is the hypothesis use of partial plots and a variety of diagnostic tests that development shapes how electoral systems and quo- (leverage [hat], hettest, Ramsey “reset,” Cook’s D, etc.). tas affect women’s representation in national legislative Because of the small sample size of the discrete models, bodies. In most studies, level of development is mea- error terms were plotted and show a normal distribution. sured using GDP or energy consumption per capita as As a result of multicollinearity, level of development proxies, which fails to account for social development, a and geographic regions are not included in the same crucial factor in understanding gender equality. For model. The measure for developed countries is corre- example, a country such as Equatorial Guinea, which the lated at .83 with the dichotomous measure of Western World Bank ranked in the top 20 by GDP per capita in countries. The mean Vif for the full model is 2.14, and 2008, would be characterized as a developed country no variables have a Vif higher than 4 or a tolerance less using an exclusively economic definition. This ignores than .2. The results indicate that Seychelles (in 1992) the actual health and welfare of the population, as well as and Rwanda (from 2003 to 2010) are outliers. These the enormously skewed distribution of wealth, which cases are removed from the analyses both individually when taken into consideration makes it a developing and together, with no significant differences in the country. To address this I define development using the results. Because of the special importance of Rwanda, Human Development Index (HDI), a comparative mea- I choose to preserve these cases. Given the difficulty sure incorporating life expectancy at birth; adult literacy of instituting national gender quotas in majoritarian rates; combined primary, secondary, and tertiary gross electoral systems, I also assess the correlation between enrollment ratios; and the natural log of GDP per capita quotas and PR electoral systems (.11). at purchasing power parity into a single 100-point scale. Of the 192 independent states recognized by the According to the United Nations (2008), HDI values can United Nations in 2010, I have data for 168 countries. I be used to categorize countries as developed (≥.9), devel- do not analyze countries that lacked an elected national oping (.5 to .899), and least-developed (<.5). legislative body over the period, nor do I include coun- Correspondingly, countries are coded into three dummy tries considered dependencies or under trusteeship at any variables. As with any composite index, there are draw- point during the time frame (see Supplemental Appendix backs in terms of data quality and reliability. However, at http://prq.sagepub.com/supplemental/ for list of coun- the UN Development Program has gone to great lengths tries).2 Missing data are accounted for using multiple to systematize this measure, and it is widely used in the imputations in Stata 11.0, with all variables from the full development literature as a reliable indicator of societal model used to improve results. Imputed variables are well-being (Inglehart and Norris 2003). highly correlated with the variables needing imputations. Cases missing are proportionally representative of the three development groups.3 Running the analyses using a Women and Politics over Time dummy variable for missing data produced no significant Many factors that affect women’s political representation changes in the results. are processes that unfold over time. Gender quotas, for Rosen 311 example, commonly take a minimum of one election not (0). I include only nationally legislated quotas rather cycle for results to become visible. In addition, economic than voluntary party quotas because, as nationally man- crises or upturns may have lagged effects on political dated legislation, these quotas are generally accompanied outcomes. In spite of this, previous studies that examine by sanctions for non-compliance providing an avenue for variations in women’s parliamentary representation have enforcement. There are drawbacks to this measurement generally been at single points in time. Paxton (1997), choice however, especially for countries with dominant however, conducts a panel study across two points in party systems. For example, South Africa’s African time (1975 and 1988) to test for lagged effects, and her National Congress adopted a 30 percent quota for women results indicate that there are good reasons to believe on party lists, and as the leading political party this resulted time matters. Recently, leading scholars in the field have in a jump in women’s parliamentary representation from 3 called for more time-series analyses on the subject, argu- to 27 percent in a single election (1994). While more ing that it would enable researchers to find support for nuanced quantitative work on gender quotas is imperative, theories that have thus far generated weak evidence or to for the purposes of this analysis I include only those quotas problematize theories that have generated strong empiri- that are codified in national law, and therefore easier to cal support (Hughes and Paxton 2007; McConnaughy enforce and manipulate through legal channels. To exam- 2007; Paxton, Hughes, and Painter II 2009; Ritter 2007). ine whether the threshold mandated by the quota matters, I I include fixed effects for each year from 1992 to 2010 include a dummy variable for those quotas that require (reference) to address the oftentimes contradictory results women to fill 30 percent or more of parliamentary seats of how key causal mechanisms work in less developed and/or party list spots.4 countries, while also addressing unit heterogeneity, a To analyze the effects of PR electoral systems and common issue in cross-sectional time-series analyses. national gender quotas on women’s legislative representa- tion across development thresholds I include interaction terms of these two variables with the dichotomous measure Electoral Systems and Quotas across of development. I also run models across three separate Development Thresholds samples of countries grouped by level of development. To test Matland’s (1998) hypothesis of a development threshold and adjudicate between the disparate results reached by previous researchers looking at less devel- Political, Socioeconomic, and Ideological Factors oped countries, I focus on the effects of two important While most studies stress the principal importance of a political institutions: electoral systems and national country’s electoral system in facilitating or hindering gender quotas. Not only are these two variables among women’s access to political office, other country charac- the most important factors identified by previous teristics affect women’s entry into the political arena. researchers, but they also provide greater opportunities Many women in politics researchers employ an analytic for intervention than structural and ideological factors. framework that distinguishes between “supply” and Electoral systems are coded into four binary variables: “demand” side factors. While the supply side deals with majority/plurality (reference), mixed, open-list proportional- those aspects influencing whether or not women have the representation, and closed-list proportional-representation skills, experience, and desire to be considered in a pool of systems. If a given system is present it is coded as 1, other- potential candidates, the demand side addresses those wise it is coded as 0. While this coding system does not characteristics and institutions that facilitate (or impede) capture the full complexity and empirical messiness of women moving from the pool of candidates into elected these electoral systems, the simplification does allow for office (Paxton and Hughes 2007). I control for most fac- more straightforward analysis and also improves on the tors found to be important by previous scholars. The classifications used in previous studies, which generally do variables fall into one of three categories: political, socio- not distinguish between open- and closed-list PR systems. economic, and ideological/cultural factors. National quotas are operationalized using a comprehen- sive definition that includes both reserved seat and elec- toral quotas. While the former regulates the actual election The Political Structure of women and the latter addresses the candidates running Women’s ability to actually get elected is structured by for elected office, both aptly represent the many ways that the nature of the political system. Institutional differ- countries use national affirmative action strategies to ences across political systems shape the level of demand increase women’s political representation (Dahlerup 2010, for female candidates (Hughes 2009; Paxton and 19). Quotas are measured dichotomously, indicating Kunovich 2003; Stevens 2007). In addition to electoral whether the country has a nationally mandated quota (1) or system and quota variables, I include the number of years 312 Political Research Quarterly 66(2)

since women won the right to vote, a measure of govern- Ideological/Cultural Factors ment effectiveness, the Freedom House (2012) measure of political rights, and a dummy variable for countries Gender is a universal form of social categorization and a with a Marxist-Leninist national ideology. The longer basis for cultural norms, and therefore gender role ide- women have been legally defined as citizens (first gained ologies will also affect women’s access and electability suffrage), the more integrated women should be into the to political positions (Paxton and Kunovich 2003, 2005; country’s political fabric (Reynolds 1999). The variable Ritter 2007). I follow previous studies by including geo- is continuous, measuring the number of years since graphic regions and religion as ideological measures.7 women won voting rights. Women’s ability to win office Geographic fixed effects are common proxies for unmea- may also be affected by parliament’s capacity to actually surable ideologies and cultures unique to specific regions implement policies. If parliaments are effectively power- of the world. The seven regional categories are Western less, female legislators may be more palatable to stand as (reference), Scandinavia, Eastern and Central Europe, tokens in a symbolic government. Conversely, more Asia and the Pacific, Sub-Saharan Africa, Middle East effective parliaments may be correlated with stronger and Northern Africa, and Latin America and the democracy, which is theorized to benefit women’s access Caribbean. These measures also control for the possibil- to political positions. To test this link, I include both the ity of spatially correlated errors (Viterna, Fallon, and World Bank’s government effectiveness index and Beckfield 2008). In order to account for intra-country Freedom House’s political rights score. The government variations, I measure the percentage of adherents to vari- effectiveness index measures the ability of governments ous religious groups rather than controlling for religion to effectively deliver public services, the quality of the as a series of dummy variables (Paxton and Kunovich bureaucracy, the competency of public servants, and the 2003). Keeping in line with previous studies I include the independence of the civil service from political pressures following religious categories: Protestant (reference), (World Bank 2010).5 Values are indexed to have a mean Catholic, Orthodox, Muslim, and Other. of zero, with positive scores indicating better governance. The Freedom House political rights score combines esti- mates of government effectiveness with measures of Results electoral process autonomy and the extent of political Table 1 reports the OLS regression results for 7 models rights. Scores range from 1 (most free) to 7 (least free). run with the full sample of 168 countries. Table 2 shows Finally, a dummy variable is included for Marxist- the results when level of development is treated as a Leninist countries, which generally codify principles of scope condition, with separate analyses of developed, equality into law and therefore may have higher percent- developing, and least-developed countries. ages of women in parliament.6 Model 1 in Table 1 includes only the main theoretical variables and explains about 26 percent of the cross- national variation of women’s representation. In the most The Social Structure complete models (models 6 and 7), approximately 53 and Women’s position in society’s social structure affects the 60 percent of the variation is explained respectively.8 “supply-side” of politics (Hughes 2009; Paxton and Kunovich 2003; Stevens 2007). Political elites often have higher levels of education and professional experience. Parity in Proportional Representation Although empirical evidence is mixed, it’s hypothesized Consistent with the findings of previous researchers, my that access to economic and social capital influences the results show that PR systems are substantively and statis- factors that enable women to seek political office by pro- tically significant throughout all models. All things being viding them with the skills and experiences required of equal, countries with a PR system have on average more political officials (Hughes 2009; Kenworthy and Malami women in parliament than those countries with majoritar- 1999; Matland 1998; Paxton 1997; Paxton and Kunovich ian electoral systems. Closed-list PR systems are more 2003; Rule 1987; Tremblay 2008; Viterna, Fallon, and beneficial to female parliamentarians than open-list sys- Beckfield 2008). I include female labor force participa- tems. When compared to majoritarian systems and con- tion (as percentage of total labor force, logged), the per- trolling for all other factors, closed-list systems have an centage of women that are literate (logged), and fertility average of 4.6 percentage points more women in parlia- rates (births per ). These variables gauge the ment compared to 2.5 percentage points in open-list degree to which women are autonomous members of systems. In closed-list systems, voters are not able to society and provide proxy measurements of the pool of reorder candidate lists as they do in open-list systems, women eligible for elected public office. and this appears to benefit women in the full sample. Ψ

Model 7 .600 .595 ( n = 3,181) 0.646 (0.679) –0.490 (0.514) 1.80*** (0.385) 6.04*** (0.685) 2.60*** (0.237) 3.77*** (0.663) 1.57*** (0.348) 2.63*** (0.550) 4.46*** (0.312) 1.30*** (0.429) –1.36** (0.529) 117.7*** 16.62*** (0.855) –4.27*** (0.585) –4.25*** (0.785) –5.44*** (0.608) 16.56*** (0.816) 0.386*** (0.088) 0.049*** (0.009) –0.047*** (0.008) –14.31*** (3.41) –0.033*** (0.008) –0.036*** (0.008) –0.052*** (0.010) Ψ Model 6 .526 .521 ( n = 3,181) 3.25*** (0.490) 1.66*** (0.411) 8.11*** (0.629) 3.56*** (0.256) 4.94*** (0.706) 4.92*** (0.736) 2.09*** (0.647) 1.26*** (0.375) 2.45*** (0.594) 4.64*** (0.347) 2.52*** (0.482) –0.016* (0.008) 94.31*** 14.18*** (0.848) 0.390*** (0.093) 0.063*** (0.008) –9.24*** (1.15) –0.021** (0.008) –29.51*** (3.28) –0.032*** (0.008) –0.043*** (0.010) Ψ Model 5 .521 .516 ( n = 3,181) 4.29*** (0.437) 2.46*** (0.377) 8.24*** (0.561) 3.74*** (0.244) 4.61*** (0.706) 4.96*** (0.738) 2.19*** (0.649) 2.45*** (0.596) 1.42*** (0.362) 4.67*** (0.328) 2.61*** (0.460) 0.207** (0.083) 103.8*** 15.20*** (0.781) 0.059*** (0.008) –9.12*** (1.15) –35.81*** (2.86) Ψ Model 4 .470 .465 ( n = 3,181) 2.00** (0.620) –0.003 (0.085) 2.31*** (0.227) 5.45*** (0.733) 5.50*** (0.772) 3.56*** (0.674) 1.54*** (0.377) 4.29*** (0.335) 1.98*** (0.474) 92.93*** 10.60*** (0.766) 15.36*** (0.789) 0.053*** (0.007) –10.11*** (1.21) Ψ Model 3 .367 .362 ( n = 3,181) 0.452 (0.673) 5.67*** (0.798) 6.10*** (0.840) 8.31*** (0.597) 1.37*** (0.396) 3.29*** (0.357) 2.78*** (0.486) 70.29*** 14.01*** (0.623) –10.40*** (1.31) Model 2 .289 .287 ( n = 3,181) 7.1*** (0.886) 1.68** (0.707) 1.13** (0.418) 9.15*** (0.233) 6.07*** (0.843) 8.60*** (0.630) 3.04*** (0.377) 2.14*** (0.512) 161.1*** –10.76*** (1.39) Model 1 .257 .256 ( n = 3,181) 8.72*** (0.231) 4.91*** (0.435) 1.47*** (0.422) 4.17*** (0.366) 4.50*** (0.475) 219.6*** 10.71*** (0.430) 2 Eastern Europe/ Central Eastern Europe/ Sub-Saharan Africa Africa East/ NorthMiddle Latin America Labor Force Government effectiveness Government Political rights Political Quota × Developed PR × Developed Quota (>30%) Mixed system Mixed Closed PR system Open PR system 2 Geography Scandinavia Asia Asia/Pacific Constant R Muslim Other Ideological factors Orthodox Catholic Fertility Literacy Socioeconomic factors Marxist/Leninist Adjusted R F Note: Standard errors are in parentheses. OLS = ordinary least squares; PR = proportional representation. PR = proportional OLS = ordinary least squares; in parentheses. are errors Standard Note: the On average, –8.46 in 1992 to –2.31 2005. Coefficients range from and significant until 2006. negative are effects The fixed 1992–2010 (reference). years variables for Ψ Model includes dummy .44 each year. by coefficient decreases test. two-tailed p < .001, *** ** p < .01, * p < .05,

OLS Regression Results for the Full Sample, Percentage Women in Parliament 1992-2010. Women Percentage the Full Sample, Results for 1. OLS Regression Table Explanatory Variables

Years 1992-2010 Years Political factors Political Suffrage

Quota Development

Base model

313 314 Political Research Quarterly 66(2)

Table 2. OLS Regression Results for Developed, Developing and Least-Developed Countries, Percentage Women in Parliament 1992-2010. Developed Countries Developing Countries Least Developed Countries Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Explanatory Variables (n = 465) (n = 465) (n = 2,017) (n = 2,017) (n = 699) (n = 699) Base model Open-list PR system 12.09*** (1.25) 9.40*** (1.01) 1.69** (0.547) –0.549 (0.555) φ φ Closed-list PR system 8.79*** (1.32) 10.73*** (0.994) 2.90*** (0.430) 2.12*** (0.418) 5.83*** (0.686) 4.78*** (0.687) Mixed system 6.12*** (1.45) 5.90*** (1.33) 0.834 (0.495) –0.906 (0.472) 1.57 (0.823) 0.737 (0.714) All national quotas –1.08 (1.58) 1.26 (1.19) 4.83*** (0.534) 3.44*** (0.490) 9.34*** (0.772) 8.52*** (0.669) Years 1992-2010 Ψ Ψ Ψ Political factors Suffrage 0.165*** (0.021) –0.029** (0.011) –0.063* (0.027) Political rights –1.2 (0.757) 0.719*** (0.110) 1.09*** (0.169) Government 9.67*** (1.10) 0.504 (0.319) 5.97*** (0.635) effectiveness Socioeconomic factors Labor force 37.35*** (6.38) 7.33*** (0.788) 10.0*** (1.19) Literacy 10.04* (4.87) 2.40*** (0.820) 2.71*** (0.666) Fertility –11.26*** (2.21) –4.32*** (0.613) 3.85** (1.37) Cultural/ideological factors Orthodox –0.015 (0.037) –0.041*** (0.012) 0.100* (0.041) Catholic –0.019 (0.029) 0.043*** (0.009) 0.091* (0.036) Muslim 0.266*** (0.071) –0.024** (0.009) 0.048 (0.025) Other 0.011 (0.025) 0.020* (0.009) 0.054* (0.026) Constant 15.59*** (0.947) –188.1*** (42.09) 9.63*** (0.266) –20.60*** (4.70) 7.26*** (0.418) –43.56*** (6.15) R2 .179 .680 .071 .343 .252 .556 Adjusted R2 .172 .656 .069 .332 .248 .536 F 25.09*** 28.69*** 38.48*** 32.37*** 77.90*** 26.97***

Note: Standard errors are in parentheses. PR = proportional representation. φ All PR systems are closed list, therefore this variable is omitted. Ψ Model was run with fixed effects for years 1992-2010 (reference). Coefficients are smallest in developed countries and generally similar in developing and least developed countries. Coefficients in all models are negative and significant from 1992 to 2005. Coefficient sizes in developed countries range from –5.50 in 1992 to –3.89 in 2005, in developing countries from –6.62 in 1992 to –2.68 in 2005, and in least developed coun- tries from –6.94 in 1992 to –1.91 in 2005. *p<.05, **p<.01, ***p<.001, two-tailed test.

When PR system is interacted with level of development, women a 10.7 percentage point advantage in developed it becomes an even stronger predictor of variations in the countries, the benefit is only 2.1 and 4.8 percentage percentage of women in parliaments. Developed coun- points in developing and least-developed countries, tries with a PR system have 6.1 percentage points more respectively. Open-list PR systems are also more benefi- women in parliaments than developed countries with cial to women than majoritarian systems in developed other electoral systems (4.9 in the full model). These countries (conferring a 9.4 percentage point advantage); results show that women seeking political office in devel- though the effect is insignificant in developing countries.9 oped countries with PR systems have a very significant While mixed systems are less favorable than PR systems, advantage. To further illustrate how electoral systems they do confer an advantage to women seeking representa- work across thresholds of development, we can look at tion in the full sample when compared to majoritarian the scope models in Table 2. Closed-list PR systems are systems. However, looking at the scope models, it becomes the most beneficial to women across all countries; how- clear that these results are biased toward developed coun- ever, the effect size is two to four times larger in devel- tries. All else being equal, the coefficient fails to reach oped countries. While a closed-list PR system gives statistical significance in developing and least-developed Rosen 315

with nationally mandated quotas have between 2 and 4.9 percentage points more women in parliament than countries without such quotas. Quotas that mandate a 30 percent or greater threshold have an even stronger affect, with a 3.8 to 7.1 percentage point advantage over those countries that either lack quota legislation or have a quota mandating a threshold below 30 percent. Once quotas are interacted with level of development, however, the sign of the coefficients change from posi- tive to negative, but remain significant. This implies that a developed country with a quota actually has an inverse relationship to women’s parliamentary representation. In other words, when controlling for other factors, devel- oped countries with quotas have about 9.2 percentage points fewer women in parliaments than developed countries without quotas. These results, however, should be interpreted with caution. To begin with, it is important to note that during the time period of analysis only five developed countries instituted national quotas.10 Moreover, it is unlikely that quota legislation actively Figure 1. Average percentage of women in lower/single decreases the percentage of women in parliament. houses of parliament by level of development and electoral Instead, it is more probable that this is a spurious correla- system. tion, capturing the fact that the relationship is statisti- cally insignificant in developed countries. countries, indicating its effect is no different from majori- The discrete models in Table 2 give us a better sense of tarian systems. how quotas work across countries. Results show that Figure 1 clearly shows that while women do better in there is no statistical difference between developed coun- all countries that use a PR system, the effect size is much tries with and without quotas. By contrast, in developing larger in developed countries. This indicates that other and least-developed countries the relationship is signifi- factors are obscuring the beneficial effects of this elec- cant and positive. The effect size is strongest in least- toral system in less developed countries. Nevertheless, in developed countries where women experience about an contrast to Matland (1998) and Hughes’s (2009) findings, 8.5 percentage point benefit in comparison to similarly my results show that both open- and closed-list PR sys- situated countries without quotas. Least-developed coun- tems do benefit women in less developed countries. tries such as Rwanda, Uganda, and Burundi, which have However, it is also clear that these benefits do not work in all instituted national gender quotas, have considerably the same fashion across countries, which supports higher percentages of women in parliament than compa- Matland’s more basic insight about the presence of a rable countries in the region without quotas. In contrast to development threshold that must be met in order for the most cross-national analyses, these results indicate that full beneficial effects of a PR system to be seen. It is dif- there is a statistically significant relationship between ficult to draw a line delineating an exact threshold of nationally mandated gender quotas and increases in development and it is likely not linear as a PR system is women’s parliamentary representation in less developed more beneficial to women in least-developed versus countries. The findings also demonstrate that results developing countries. However, once countries achieve derived from large cross-national samples should be an HDI score above .9, the advantageous effects of PR interpreted with caution, as they have the potential to be systems double, triple, or even quadruple in size. biased toward one subset of countries. Figure 2 presents a graphical representation of these findings, showing that developed countries with quotas Mandating Equality through Quotas tend to have lower averages of women in parliament than In contrast to previous research, my results show that developed countries without quotas. Conversely, national the presence of nationally mandated gender quotas is quotas in developing and least-developed countries lead to significantly associated with cross-national variations significant increases in the average percentage of women in women’s parliamentary representation. Countries in parliament. These results show that a one-size-fits-all 316 Political Research Quarterly 66(2)

less developed countries rank among the highest when measuring women’s parliamentary representation. Instead, the effects of development interact with other variables in complex ways to drive levels of women’s representation. While this analysis does not allow us to pinpoint the exact threshold of development required for other variables to have an effect, we can identify how these variables work across three broad categories of socioeconomic development. Institutional features of political systems help to shape the demand for female politicians. Alone, political factors explain about 27 percent of cross-national variation. The most striking finding among the political factors is the effect of a Marxist/Leninist ideology, which confers a 14 to 16 percentage point advantage to women in parliament over countries that do not subscribe to such tenets. Although these countries are generally authoritarian regimes that are unlikely supporting women’s equality based on feminist principles of fairness, the national doc- trine of Marxism/Leninism emphasizes equality in all Figure 2. Average percentage of women in lower/single aspects of life. These systems assume both women and houses of parliament by level of development and national men will contribute to society outside of the home and gender quotas. family. Another interesting finding concerns the length of time since women won the right to vote. In both the full policy agenda advocating gender quotas to increase wom- sample and in developed countries, the more time that has en’s representation in national legislatures does not make elapsed since women’s suffrage, the larger the percentage sense. Rather, policies require greater nuances and should of women in parliament (although the effect is less than 1 be tailored to a country’s specific situation. percentage point). However, in developing and least- developed countries, every additional year since winning suffrage is associated with a slight decrease in the per- The Supply and Demand of Women’s centage of women in parliament. Many countries that Political Representation gained independence post-1945 also instituted female While highlighting the different effects of particular suffrage at the same time, meaning it occurred at a time poli-tical institutions on women’s parliamentary repre- when women’s rights were part of an international dia- sentation across development thresholds is the primary logue of priorities, as well as a method of attaining some intention of this article, other variables are also impor- international legitimacy. For example, both Mozambique tant to the analysis. First, we may note that annual fixed and Angola gained independence and passed women’s effects do not have much explanatory power (explain- suffrage in 1975, and women currently hold 39.2 and ing about 11 percent of the variation), but do exist from 38.6 percent parliamentary seats respectively. This is sig- 1992 to 2005 in all countries. In comparison to 2010, nificantly higher than other countries in the region that the preceding years have, on average, fewer women in granted suffrage rights earlier. Finally, many agree that parliament. Annual coefficient sizes are largest in least- stronger democratic political institutions will enable developed countries and decrease as level of develop- higher female representation in national politics. The ment increases. Second, socioeconomic development Freedom House measure of political rights, however, is alone explains about 18 percent of the variation in insignificant in developed countries and has a positive women’s parliamentary representation. Developed effect in both developing and least-developed countries. countries have, on average, an 11.5 percentage point Since a lower score on the political rights index corre- advantage over less developed countries, without sponds to greater freedoms, one might assume a negative controlling for other variables. The effect of this mea- association. Quotas, however, tend to be most popular in sure remains strong and significant across models. An post-conflict and even authoritarian regimes, potentially exclusive focus on development, however, can be mis- leading to higher female representation in countries con- leading and is not sufficient to explain why so many sidered to be less open and free. Rosen 317

While political features shape the demand for female interacts with key causal mechanisms to impact their politicians, social structure factors impact the supply of effects on women’s representation across countries. I qualified female candidates. Socioeconomic variables demonstrate this by focusing considerable attention on alone explain about 18 percent of the variation in wom- two political factors that have been identified by previ- en’s parliamentary representation cross-nationally. Both ous researchers as key causal factors. The significance education and labor force participation positively affect of using these variables to test Matland’s (1998) women’s parliamentary representation by putting more hypothesis of a development threshold is that as politi- women into the pool of potential political candidates. The cal institutions, they are more easily manipulated to effect sizes of both variables, however, are more than increase women’s political representation than factors double in developed countries. This suggests that there is such as cultural ideologies or social features. a stronger association between having a larger pool of In contrast to some previous research, I find that the “qualified” candidates and greater representation in devel- effects of PR systems are significantly related to increases oped versus less developed countries. In developed and in women’s parliamentary representation across all coun- developing countries, higher fertility rates are associated tries. However, the size of this effect is dramatically with significant decreases in the percentage of women in larger in developed countries, indicating that there is a parliaments (every additional child per woman associated threshold of development that must be reached before the with an average decrease of 11.3 and 4.3 percentage full benefits of this type of electoral system is seen. In points, respectively). However, this relationship differs in less developed countries, the benefits of PR systems may least-developed countries where higher fertility rates are be obstructed by barriers that prevent party lists from actually associated with greater percentages of women in going forward with female candidates. Matland (1998) parliaments. While fertility rates are considered a method argues that the failure of the PR system to work in wom- of showing how tied women are to the home (and there- en’s favor in less developed countries suggests that either fore excluded from the public sphere), in countries with demands for representation are not being forwarded by more traditional gender role ideologies those women that women—perhaps because they are politically inactive— represent an “ideal womanhood” may be more likely to or, within the parties, the perceived costs of nominating be seen as acceptable political symbols. Therefore, where women are so great that parties prefer to run virtually gender roles tend to idealize women as wives and moth- all-male slates rather than risk the wrath of the voters by ers, those women with more children may be considered nominating women. I believe the latter is true and that more electable to political office and therefore better without pressure for women’s inclusion, women may be candidates. unable to exploit the political opportunities provided by Ideological factors also impact the supply of female a proportional-representation system (Kittilson 2006). candidates by shaping the norms surrounding women’s The beneficial effects of PR systems may also be roles outside of the home. On their own, these factors obscured in less developed countries because of other explain about 11 percent of cross-national variations. hindrances, such as corruption, internal conflicts, and/or Religious affiliations are generally weak predictors of more traditional cultural ideologies relegating women to women’s representation both in the full and discrete mod- the home and family. Additionally, specific electoral sys- els, while geographic regions have stronger effects. tem features may differ in less developed countries, fur- Interestingly, the fixed effect for Sub-Saharan African ther interfering with female politicians’ bid for national countries fails to reach significance, indicating no statisti- office. This analysis shows that closed party lists can cal difference from Western countries. Women’s political mitigate any sexism among voters by preventing women representation has soared in Africa recently, mostly from being moved further down candidate lists (to essen- fueled by quota legislation and the democratic transitions tially unelectable positions). Furthermore, features such of several postconflict regimes. as district magnitude and electoral thresholds also affect women’s chances of winning national political positions (Bauer and Britton 2006; Matland 1998; Paxton and Discussion and Conclusions Hughes 2007). Since political parties are the gatekeepers This analysis shows that models designed to explain of candidate selection, variations in their structures and variations in women’s parliamentary representation are ideologies impacts electoral systems’ ability to aid wom- a good fit for the developed world, but caution should en’s representation (Kittilson 2006; Paxton and Hughes be taken when generalizing across countries with lower 2007). Finally, causal combination could also play a role levels of socioeconomic development. Using Matland’s in less developed countries, with the beneficial effects of (1998) research as a starting point, I show that development a PR system relying on the presence or absence of other 318 Political Research Quarterly 66(2) factors (such as strong Left parties and active women’s years after passing quota legislation in France the per- movements) (Krook 2010). These nuances need to be centage of women in parliament rose from 10.8 to 12.2 included in further quantitative analyses, while in-depth percent. The five developed countries with the highest case studies specifically comparing electoral systems in percentages of female parliamentarians all have PR sys- developed and less developed countries would offer tems, but none have national quotas. Rwanda is a post- insights as to why women do not experience the full conflict country with very low levels of social and rewards of a PR system in countries with lower levels of economic development; however, it does have a national socioeconomic development. quota and currently has the highest percentage of female In contrast to previous research, this analysis shows that parliamentarians in the world. Angola, Tanzania, nationally mandated quotas do have a statistical impact on Uganda, and Burundi all have relatively low levels of women’s parliamentary representation. However, while socioeconomic development, have adopted national legislation mandating a minimum percentage of parlia- quota legislation, and rank among the top twenty coun- mentary seats to be reserved for female candidates tries by percentage women in parliament. In fact, of the increases women’s representation in less developed coun- twelve less developed countries that rank in the top tries, it is not statistically effective in developed countries. twenty, nine of them have national gender quotas. This Industrialized democracies are less likely to adopt national pattern, however, does not hold for all countries. Women quota legislation, and if they do it often occurs once women in Brazil, Niger, Panama, and Armenia, all less devel- have already attained 20 to 30 percent of parliamentary oped countries with national gender quotas, make up seats, thereby limiting the beneficial effects (and curbing less than 10 percent of national parliaments. Likewise, statistical significance) (Dahlerup 2006). In addition, in in Ireland, Israel, and Greece, all developed countries newer development discourses on equality between the with PR electoral systems, women hold relatively low sexes are seen as something that must be instituted rather proportions of parliamentary seats. Testing the specific than a consequence of higher development (Dahlerup features of quotas and PR systems, as well as the possi- 2006). Quotas, therefore, tend to be more popular in semi- bility of causal combination, would greatly improve our democratic and postconflict countries, in part because they understanding of these differences. are perceived as conferring international legitimacy but More generally, these results suggest that rather than also because these countries often begin from scratch offering universal truths, comparative quantitative analy- building political institutions, drawing up new constitu- ses can suffer from regional biases (often privileging tions, and rebuilding parliaments, providing a political Western democracies) (Krook 2010). Generalizing opening for quotas (Matland 2006; Tripp, Konate, and broadly across countries, therefore, does not adequately Lowe-Morna 2006). Furthermore, these results show that represent the effects of key causal mechanisms. The the threshold of representation mandated by the quota leg- importance of determining the means to women’s politi- islation matters. Those quotas reserving 30 percent or more cal advancement goes beyond the formal significance of parliamentary seats and/or candidate spots for women have fairness and justice. Achieving greater gender equality in significantly more women in parliament, often far surpass- national governments improves the quality of policy ing the legislated threshold. making, both in relation to women’s specific needs and to These findings have the potential to alter the dis- the policy-making process more generally. Research course around how to increase women’s political repre- shows that when women are underrepresented, issues sentation. They indicate that a one-size-fits-all remedy such as sexual harassment, women’s health, and family to increase the proportion of women occupying seats in policies may receive less attention (Hughes 2009; national governments does not make sense. Rather, the Wängnerud 2009). In addition, a diverse group of legisla- transition to a PR electoral system considerably helps tors enhances the democratic process more broadly by women in the average developed country, but does less ensuring the representation of minorities, increasing the to aid women in less developed countries in the absence diversity of viewpoints, and heightening political compe- of other demands for representation. Conversely, women tition (Bauer and Britton 2006; Kenworthy and Malami in developing and least-developed countries generally 1999; Reynolds 1999). Given these substantial implica- experience substantial gains after the adoption of tions, it is important to understand which, if any, institu- national gender quota legislation, while this use of time tional mechanisms support increases in women’s political and/or resources may be ill advised in developed coun- representation. This research shows that different institu- tries. For example, the introduction of gender quotas in tional changes are advised to increase women’s presence Costa Rica increased women’s parliamentary represen- in national governments depending on contextual factors tation from 19 to 35 percent in one election, while six within specific country frameworks. Rosen 319

Appendix Percentage Women in Single/Lower House of National Parliament by Level of Development, 2010

Developed Countries Developing Countries (Cont’d) Developing Countries (Cont’d) Developing/LDC Countries (Cont’d) Australia (24.7%) Bolivia (25.4%) Mongolia (3.9%) Cameroon (13.9%) Austria (27.9%) Bosnia/Herzegovina (16.7%) Morocco (10.5%) Comoros (3%) Belgium (39.3%) Botswana (7.9%) Namibia (24.4%) Congo, Republic (7.3%) Canada (22.1%) Brazil (8.6%) Nicaragua (20.7%) Djibouti (13.8%) Denmark (38%) Bulgaria (20.8%) North Korea (15.6%) Equatorial Guinea (10%) (40%) Cape Verde (18.1%) Panama (8.5%) Ghana (8.3%) France (18.9%) Chile (14.2%) Papua New Guinea (0.9%) India (10.8%) Germany (32.8%) China (21.3%) Paraguay (12.5%) Kenya (9.8%) Iceland (42.9%) Colombia (12.7%) Peru (27.5%) Lao, PDR (25.2%) Ireland (13.9%) Costa Rica (38.6%) Philippines (22.1%) Lesotho (24.2%) Italy (21.3%) Croatia (23.5%) Poland (20%) Madagascar (12.5%) Japan (11.3%) Cuba (43.2%) Romania (11.4%) Mauritania (22.1%) Netherlands (40.7%) Czech Republic (22%) Russia (14%) Nepal (33.2%) New Zealand (33.6%) Dominica (12.5%) Samoa (8.2%) Nigeria (7%) Norway (39.6%) Dominican Republic (20.8%) Seychelles (23.5%) Pakistan (22.2%) Portugal (27.4%) Ecuador (32.3%) Slovakia (15.3%) Sao Tome Principe (18.2%) Spain (36.6%) Egypt (12.7%) Solomon Islands (0%) Tanzania (36%) Sweden (45%) El Salvador (19%) South Africa (44.5%) Togo (11.1%) Switzerland (29%) Estonia (22.8%) Sri Lanka (5.3%) Uganda (31.3%) United Kingdom (22%) Fiji (8.5%) St. Kitts/Nevis (6.7%) Yemen (0.3%) United States (16.8%) Gabon (14.7%) St. Lucia (11.1%) Zambia (14%) Georgia (6.5%) St. Vincent (14.3%) Developed/Developing Countries Grenada (13.3%) Suriname (9.8%) LDC Countries Cyprus (12.5%) Guatemala (12%) Swaziland (13.6%) Benin (10.8%) Greece (17.3%) Guyana (30%) Syria (12.4%) Burkina Faso (15.3%) Israel (19.2%) Honduras (18%) Tajikistan (19%) Burundi (32.1%) Luxembourg (20%) Hungary (9.1%) Thailand (13.3%) Central African Republic (9.6%) Singapore (23.4%) Indonesia (18%) Tonga (0.0%) Chad (5.2%) Slovenia (14.4%) Iran (2.8%) Trinidad and Tobago (28.6%) Congo, DR (8.4%) South Korea (14.7%) Jamaica (13.3%) Tunisia (27.6%) Cote d’Ivoire (8.9%) Jordan (10.8%) Turkey (9.1%) Eritrea (22%) Developing Countries Kazakhstan (17.8%) Turkmenistan (16.8%) Ethiopia (27.8%) Albania (16.4%) Kuwait (7.7%) Ukraine (8%) Gambia (7.5%) Algeria (7.7%) Kyrgyzstan (23.3%) Uruguay (15.2%) Guinea (0%) Andorra (35.7%) Latvia (20%) Uzbekistan (22%) Guinea Bissau (10%) Antigua and Barbuda (10.5%) Lebanon (3.1%) Vanuatu (3.8%) Haiti (11.1%) Argentina (38.5%) Lithuania (19.1%) Venezuela (17%) Kiribati (4.3%) Armenia (9.2%) Macedonia, F.Y.R. (32.5%) Viet Nam (25.8%) Liberia (12.5%) Azerbaijan (16%) Malaysia (9.9%) Zimbabwe (15%) Malawi (20.8%) Bahamas (12.2%) Maldives (6.5%) Mali (10.2%) Bahrain (2.5%) Malta (8.7%) Developing/LDC Countries Mozambique (39.2%) Barbados (10%) Mauritius (18.8%) Angola (38.6%) Niger (9.7%) Belarus (31.8%) Mexico (26.2%) Bangladesh (18.6%) Rwanda (56.3%) Belize (0.0%) Moldova (18.8%) Cambodia (21.1%) Senegal (22.7%) 320 Political Research Quarterly 66(2)

Acknowledgment 7. To retain the majority of my 168-country sample, I am not The author wishes to thank James Mahoney, Monica Prasad, able to utilize the more nuanced gender role ideology mea- Alberto Palloni, Jeremy Freese, and Charles Camic for their sure employed by Paxton and Kunovich (2003). helpful comments on earlier drafts of this paper. 8. Models are also run with dummy variables for each country, using Sweden as the reference. Electoral system variables are Declaration of Conflicting Interests insignificant in these models, while the measure for quotas is The authors declared no potential conflicts of interest with significant at the p <.001 level, with a coefficient of 3.45. The respect to the research, authorship, and/or publication of this model explains 85 percent of cross-national variation. article. 9. There are no least developed countries with open-list PR systems in the sample. Funding 10. Belgium, France, Portugal, Spain, and South Korea. The authors received no financial support for the research, authorship, and/or publication of this article. References Bauer, Gretchen, and Hannah Britton, eds. 2006. Women in Notes African parliaments. Boulder, CO: Lynne Rienner. 1. Although Hughes (2009) highlights results using energy Carroll, Susan. 2001. Representing women: State legislators as consumption per capita to distinguish a country’s develop- agents of policy-related change. In The impact of women in ment, she also tests her models using a scale from Inglehart public office, edited by Susan Carrol, 3-21. Bloomington: and Norris (2003) that relies on the Human Development University of Indiana Press. Index to group countries into the following categories: post- Celis, Karen, Sarah Childs, Johanna Kantola, and Mona Lena industrial (≥.9), industrial (.740-.899), and agrarian (≤.739). Krook. 2008. Rethinking women’s substantive representa- Her results remain stable regardless of the method used to tion. Representation 44 (2): 99-110. distinguish level of development. Central Intelligence Agency. 1992-2010. The World Fact book. 2. Models were run excluding countries that lack effective Washington, DC: Central Intelligence Agency. political systems (“politically closed authoritarian Chafetz, Janet Saltzman. 1990. Gender equity: An integrated regimes”; Diamond 2002). Coefficient sizes and explana- theory of stability and change. Newbury Park, CA: Sage. tory powers are significantly higher when excluding these Dahlerup, Drude, ed. 2006. Women, quotas and politics. New countries, although differences decrease as variables are York: Routledge Research in Comparative Politics. added. There are no differences in the significance or Dahlerup, Drude. 2010. From a small to a large minority: signs of any variable. While excluding these countries Women in Scandinavian politics. In Women, gender, and confirms and even strengthens the results of my models, politics: A reader, edited by Mona Krook and Sarah Childs, because of the exclusion of important cases (such as 225-30. New York: Oxford University Press. Rwanda, Vietnam, Burundi, etc.) I chose to maintain the Dahlerup, Drude, and Lenita Friedenvall. 2005. Quotas as a larger sample. In addition, these countries are included in “fast track” to equal representation for women: Why Scan- previous studies examining less developed countries dinavia is no longer the model. International Feminist Jour- (Hughes 2009). I do acknowledge, however, that lacking nal of Politics 7 (1): 26-48. an effective political system affects mechanisms by Del Campo, Esther. 2005. Women and politics in Latin America: which women enter politics. Perspectives and limits of the institutional aspects of women’s 3. Literacy rates are spotty from 1992 to 1994. Government political representation. Social Forces 83 (4): 1697-1725. Effectiveness is missing from 1992 to 1995, 1997, 1999, Diamond, Larry Jay. 2002. Thinking about hybrid regimes. and 2001. Imputations were checked to ensure normal Journal of Democracy 13 (2): 21-35. distributions. The imputed variables are correlated at .80 Freedom House. 2012. Freedom in the world country ratings: or higher. 1972-2011. Washington, DC: Freedom House. 4. Although a somewhat arbitrary distinction, 30 percent is Hughes, Melanie. 2009. Armed conflict, international linkages, theorized to be a threshold that enables a “critical mass” of and women’s parliamentary representation in developing women that can facilitate substantial changes (Bauer and nations. Social Problems 56 (1): 174-204. Britton 2006; Dahlerup 2006; Krook 2010). Hughes, Melanie, and Pamela Paxton. 2007. Familiar theories 5. Ninety-nine percent of the values fall between 2.5 and from a new perspective: The implications of a longitudinal –2.5. As with any composite index, issues of data quality approach to women in politics research. Politics and Gender and reliability must be considered. However, each country 3 (3): 370-78. estimate has a corresponding margin of error which Inglehart, Ronald, and Pippa Norris. 2003. Rising tide: Gender ensures that the index permits meaningful cross-country equality and cultural change around the world. New York: and over-time comparisons (World Bank 2010). Cambridge University Press. 6. China, Cuba, Lao PDR, North Korea, and Vietnam are International IDEA. 2006. Electoral system design. Stockholm, coded 1; all other countries are coded 0. Sweden: International IDEA. Rosen 321

Kenworthy, Lane, and Melissa Malami. 1999. Gender inequal- Paxton, Pamela, and Sheri Kunovich. 2003. Women’s political ity in political representation: A worldwide comparative representation: The importance of ideology. Social Forces analysis. Social Forces 78 (1): 235-68. 82 (1): 87-113. Kittilson, Miki Caul. 2006. Challenging parties, changing Paxton, Pamela, and Sheri Kunovich. 2005. Pathways to power: parliaments: Women and elected office in contemporary The role of political parties in women’s national political rep- Western Europe. Columbus: Ohio State University Press. resentation. American Journal of Sociology 111 (2): 505-52. Krook, Mona Lena. 2006. Reforming representation: The dif- Phillips, Anne. 1995. The politics of presence. Oxford: Claren- fusion of candidate gender quotas worldwide. Politics and don Press. Gender 2 (3): 303-27. Reynolds, Andrew. 1999. Women in the legislatures and execu- Krook, Mona Lena. 2010. Women’s representation in parlia- tives of the world: Knocking at the highest glass ceiling. ment: A qualitative comparative analysis. Political Studies World Politics 51 (4): 547-72. 58 (5): 886-908. Ritter, Gretchen. 2007. Gender and politics over time. Politics Lovenduski, Joni, and Pippa Norris. 2004. Westminster women: and Gender 3 (3): 386-97. The politics of presence. Political Studies 51 (1): 84-102. Rule, Wilma. 1987. Electoral systems, contextual factors, and Mahoney, James, and Gary Goertz. 2004. The possibility prin- women’s opportunities for election to parliament in twenty- ciple: Choosing negative cases in comparative research. three democracies. Western Political Quarterly 40:477-98. American Political Science Review 98 (4): 653-69. Stevens, Anne. 2007. Women, power and politics. New York: Matland, Richard. 1998. Women’s representation in national Palgrave Macmillan. legislatures: Developed and developing countries. Legisla- Tolleson-Rinehart. 2001. Do women leaders make a difference? tive Studies Quarterly 23 (1): 109-25. In The impact of women in public office, edited by Susan Matland, Richard. 2006. Electoral quotas: Frequency and Carrol, 149-65. Bloomington: University of Indiana Press. effectiveness. In Women, quotas and politics, edited by Tremblay, Manon. 2008. Women and legislative representation: Drude Dahlerup, 275-92. New York: Routledge Research Electoral systems, political parties, and sex quotas. New in Comparative Politics. York: Palgrave Macmillan. McConnaughy, Corrine M. 2007. Seeing gender over the short Tripp, Aili Mari, and Alice Kang. 2008. The global impact of and long haul. Gender and Politics 3 (3): 378-86. quotas: On the fast track to increased female legislative rep- Moore, Gwen, and Gene Shackman. 1996. Gender and author- resentation. Comparative Political Studies 41 (3): 338-61. ity: A cross-national study. Social Science Quarterly 77 (2): Tripp, Aili, Dior Konate, and Colleen Lowe-Morna. 2006. 273-88. Sub-Saharan Africa: On the fast track to women’s political Oakes, Ann, and Elizabeth Almquist. 1993. Women in national representation. In Women, quotas and politics, edited by legislature: A cross-national test of macrostructural gender Drude Dahlerup, 112-37. New York: Routledge Research theories. Population Research Policy and Review 12:71-81. in Comparative Politics. Paxton, Pamela. 1997. Women in national legislatures: A cross- United Nations. 2008. Human Development Indices. New York: national analysis. Social Science Research 26:442-64. United Nations Statistics Division. Paxton, Pamela, Jennifer Green, and Melanie Hughes. 2008. United Nations. 2010. The world’s women 2010. New York: Women in parliament, 1945-2003: Cross-national dataset United Nations Statistics Division. [Computer file]. ICPSR24340-v1. Ann Arbor: Inter-Univer- Viterna, Jocelyn, Kathleen Fallon, and Jason Beckfield. 2008. sity Consortium for Political and Social Research. How development matters: A research note on the relationship Paxton, Pamela, and Melanie Hughes. 2007. Women, politics, between development, democracy and women’s political repre- and power: A global perspective. Thousand Oaks, CA: Pine sentation. International Journal of Comparative Sociology 49 Forge Press. (6): 455-77. Paxton, Pamela, Melanie M. Hughes, and Matthew A. Painter Wängnerud, Lena. 2009. Women in parliaments: Descriptive II. 2009. Growth in women’s political representation: A lon- and substantive representation. Annual Review of Political gitudinal exploration of democracy, electoral system, and Science 12:51-69. gender quotas. Working Paper: 09-01. Center for the Study World Bank. 2010. Worldwide governance indicators. Wash- of Democracy. ington, DC: World Bank.