Democracy, Representation, and Women: A Comparative Analysis

Transcription

1 Democratization ISSN: (Print) X (Online) Journal homepage: Democracy, Representation, and Women: A Comparative Analysis Manon Tremblay To cite this article: Manon Tremblay (2007) Democracy, Representation, and Women: A Comparative Analysis, Democratization, 14:4, , DOI: / To link to this article: Published online: 24 Jul Submit your article to this journal Article views: 1361 View related articles Citing articles: 9 View citing articles Full Terms & Conditions of access and use can be found at Download by: [ ] Date: 07 October 2016, At: 15:11

2 Democracy, Representation, and Women: A Comparative Analysis MANON TREMBLAY The value of political equality is central to normative theories of democracy: it is argued that women are equal citizens and therefore should share equally with men in public decisionmaking. By contrast, empirical theories typically define democracy by the presence of institutions such as a multi-party political system and competitive elections. In fact, the concept of empirical democracy may encompass a complex range of practical realities, including cultural, socio-economic, and political factors. This multifaceted understanding of empirical democracy is used to help explain the descriptive representation of women in national parliaments. The goal of the article is to advance understanding about which factors help or hinder women in entering parliament in countries defined as democratic. Its principal argument is that these factors differ according to the length of the democratic experiment. The analysis shows first that the proportion of women in parliaments is influenced by a plurality of interacting factors. Second, it shows that the mix of factors influencing the percentage of female legislators differs according to the length of the country s democratic experience. In countries where democracy has prevailed for only a short time, the voting system is the most important factor explaining the proportion of women parliamentarians. By contrast, in well-established democracies the most powerful explanation is found in an egalitarian conception of gender roles. The conclusions reminds us that empirical democracy is a complex, heterogeneous, and multifaceted phenomenon; consequently, analysis intended to reach a better understanding of the presence of women in parliaments must refrain from uniformly applying indicators to all countries, the realities of which are, in fact, quite different. Key words: democracy; parliaments; representation; women Introduction Historically, the very design of democracy and representation excluded women. In ancient Greece, cradle of the democratic experiment, democracy was rigorously exclusive: the citizen was male and born of Athenian parents. Theorists of the social contract and popular sovereignty Hobbes ( ), Grotius ( ), Locke ( ), Madison ( ), Pufendorf ( ), and Rousseau ( ) at best ignored women, and at worst endorsed their confinement to the private and family spheres. The great revolutions that paved the way for representative democracy the Glorious Revolution of 1688 in England, the American Revolution ( ), and French Revolution (1789) hardly meant the access of women to the governance of the nation. In France, the idea of the Manon Tremblay is Professor at the School of Political Studies at the University of Ottawa, Ontario, Canada. Democratization, Vol.14, No.4, August 2007, pp ISSN print/ x online DOI: / # 2007 Taylor & Francis

3 534 DEMOCRATIZATION abstract individual the notion underpinning Republican universalism precluded discussion of the political citizenship of women until the beginning of the 21st century, when a highly stimulating debate began on the subject of parity. 1 Today, representative democracy presented as an ideal includes women. Unfortunately, practice does no honour to this ideal: as of October 2006, the average percentage of women in lower or single houses of some 189 parliaments was 17 per cent. 2 This proportion varies greatly between countries. For example, the most recent election for the Rwandan Chamber of Deputies, held in September 2003, resulted in a new record for women in national parliaments 48.8 per cent were female members of parliament (MPs). That is to say, Rwanda achieved quasi-parity. This contrasts with data from other countries: 20.8 per cent of the Canadian House of Commons (following the 2006 elections), 17.3 per cent of the Italian Camera dei deputati (2006), 12.2 per cent of the French Assemblée nationale (2002), 8.3 per cent of the Indian Lok Sabha (2004), and 2.0 per cent of the Egyptian Majlis Al-Chaab (2005). About ten countries have no woman member in their lower or single house (for example, Micronesia, Nauru, Tuvalu, and the United Arab Emirates). Moreover, there seems to be no connection between the proportion of women holding seats in a legislative assembly and what the author would call the quality of democratic life in the country in question: many countries little-known for their democratic virtues, such as Cuba, Mozambique, Rwanda, and Tanzania, exhibit percentages of women parliamentarians between 30 per cent and 48.8 per cent, while in Canada, France, Japan, Switzerland, the United Kingdom, and the United States the proportion of women legislators does not even reach 25 per cent. However, Nordic countries do not fit either category: these countries combine high-quality democratic governance and enviably large proportions of women in their parliaments. The value of political equality is central to normative theories of democracy: it is argued that women are equal citizens and therefore should share equally with men in public decision-making otherwise, there is a democratic deficit. By contrast, empirical theories define democracy by the presence of institutions such as: freedom to form and join organizations; freedom of expression; the right to vote in elections and to stand for public office; the right of political leaders to compete for support and votes; freedom of information and availability of alternative sources of information; free and fair elections; and institutions for making government policies that depend on votes and other expressions of preference. 3 In fact, empirical democracy is a much more complex concept that encompasses a range of practical realities, including cultural, socio-economic, and political factors. For instance, how can democracy work if the population lacks the resources to satisfy its basic needs in health and education? How can democracy work when the majority of the people live in poverty? How can democracy work without freedom of the press and the right of association? How can democracy work where the dominant religion prescribes that a wife must obey her husband? In sum, how can democracy deliver its virtues when the cultural, socio-economic, and political contexts are hostile to the very principle of gender equality? It is this multifaceted understanding of empirical democracy that will be used here in order to advance understanding of the descriptive representation of women in national parliaments.

4 DEMOCRACY, REPRESENTATION, AND WOMEN 535 Starting from empirical theories that a democracy must be understood not as a homogenous entity but rather as a complex, heterogeneous, and multifaceted phenomenon, the principal argument of this paper is that the factors that help or hinder women in entering parliament differ according to the length of the democratic experiment. In other words, the factors affecting the presence of women in parliaments are not consistent, but vary according to how long a country has been democratically governed. The account is structured as follows: first, the relevant literature on the determinants of the proportion of women in national parliament is reviewed; second, the research design is described; third, the main results of the statistical analysis are presented; finally, the major findings are summarized and conclusions drawn. Review of the Literature Studies have shown that a range of factors influence women s access to legislative arenas. These factors can be grouped into three broad categories: cultural, socioeconomic, and political. While such categorization simplifies analysis, these factors actually combine and overlap in influencing women s access to power. They therefore pose a problem in terms of causality: do cultural factors precede socio-economic and political factors, or is it the opposite way around? For example, must there first be a culture of gender equality before women in large numbers can access legal careers and achieve economic independence, which are both major advantages in stepping into the political arena? Because such questions remain unresolved, it is probably more instructive to assume that cultural, socio-economic, and political factors interact to create a dynamic that acts as a global incubator for the election of women. Cultural Factors Culture refers to the values, standards, beliefs, and attitudes that underpin a society and its institutions and which animate the population s ways of being, talking, and doing. Religion, education, and views of gender-based social roles are the primary cultural factors identified as determinants of the proportion of women in parliaments. Generally, Protestantism as dominant religion (as opposed to other religions), women s access to university education, and egalitarian gender roles are variables positively associated with women s access to parliaments. 4 Moreover, Inglehart and Norris, Mateo Diaz, and Paxton and Kunovich 5 have all shown that culture, especially a conception of equality between women and men, is a more influential variable than voting systems (especially list proportional representation) in determining the proportion of women in parliaments. The importance of culture can be explained, Paxton and Kunovich 6 propose, by the fact that this variable is superimposed, in some ways, on socio-economic and political factors. Socio-economic Factors Socio-economic factors shape the conditions that lead women to envision careers in politics. As Norris and Lovenduski note, we are talking here about the supply of candidates. 7 One theory is that if there are few women in politics, it is because

5 536 DEMOCRATIZATION women are under-represented in the milieus where parties identify and recruit their potential candidates. 8 An improvement in women s socio-economic conditions should therefore favour an increase in their presence in parliaments. Variables considered in this category include the type of society (agricultural, industrial or post-industrial), the Human Development Index (HDI), the birth rate, the proportion of women in the labour market, the female/male revenue ratio, the per capita Gross National Product (GNP), public expenditure on education and healthcare, and the urbanization rate. Studies show that the proportion of women in parliaments is positively influenced by factors such as participation in the labour market (particularly in specialized employment), a high score on the Human Development Index, a postindustrial society, and a developed welfare state. 9 Political Factors Political factors shape the demand for candidates. More specifically, these factors influence the selection (that is, the choice of which citizens from the eligible pool will contest the election) and the election of candidates. In terms of women s representation, political factors belong to two dimensions: the political rights of women and the political regime. The first dimension refers to the political citizenship of women. It is primarily measured by the year women achieved the right to vote in national legislative elections. A large number of studies have demonstrated a connection between this measure and the proportion of women in parliaments. 10 Few variables have been developed to understand the specific mechanisms by which the political rights of women affect their presence in parliaments. However, many aspects of women s political rights may influence the outcome. These include the year women won the right to run for office in a national election, the year the first woman was elected to parliament, and the year of the first time a woman accessed government, whether a woman has already held the position of president or premier, and so on. This study takes such measures into account (see the analysis below). Unlike the political rights of women, the political regime (the second dimension of political factors discussed here) has received much attention from researchers. Here, researchers have drawn on a wide range of variables to explore the impact of the political regime on the proportion of women members of legislatures. These variables include the state structure (unitary or federal), the structure of parliament (unior bi-cameral, the number of seats, the maximum length of a legislature), the nature of the legislative career (the turnover rate of parliamentarians), the party system (the number of parties contesting elections, the number of effective parliamentary parties, the ideologies of the parties that form the government, how candidates are selected, and so on) and obviously the electoral system itself (the type of voting system, the district magnitude, the nature of the lists closed, open or flexible, and so on). Matland as well as Oakes and Almquist 11 support the idea that the number of legislative seats plays a role in the proportion of women elected. A low turnover rate of parliamentarians has long been identified as an obstacle for women s access to legislative assemblies. 12 Norris and Inglehart 13 believe that parliaments where many parties are represented offer female candidates more possibilities

6 DEMOCRACY, REPRESENTATION, AND WOMEN 537 to be elected than two-party parliaments. Political parties tending toward the left and centre of the ideological spectrum offer better possibilities for women wishing to obtain a seat in parliaments than do right-tending ones, particularly where a party positioned on the left or centre forms the government. 14 If quotas appear to have a direct and immediate impact on the feminization of parliaments, it is important to note that many criteria must be present for quotas to reach their objectives. For example, they must pre-empt resistance by mandating a gender-sensitive order for classifying candidates on lists and by imposing penalties for offences against the law. 15 The electoral system is the element of the political regime that has generated the richest literature on the determinants of women s access to parliamentary arenas. Generally speaking, parliaments elected by proportional representation show higher percentages of women than parliaments constituted by plurality/majority systems. 16 Nonetheless, many studies also present a more complicated view of the association too frequently assumed between proportional representation and a substantial presence of women in parliaments. 17 Indeed, voting systems do not consistently act independently of their contexts; how else could we explain that parliaments formed through the same voting system show different proportions of women parliamentarians, or the fact that the percentage of women in any given parliament varies across time even if the voting system remains constant? The performance of voting systems depends on the characteristics that structure them and the actors that use them to their advantage. Thus, for example, high vote/seat proportionality fosters a high turnover rate of parliamentarians that, in turn, improves women s chances of being listed in eligible positions. 18 In the same vein, a threshold of representation (probably the most powerful tool through which voting procedures can influence proportionality 19 ) that is set at a relatively high level reduces the number of parties between which the seats in parliament are shared; each party thereby obtains more seats and can therefore assign some to women, whether voluntarily (that is, to balance its representation) or mechanically (because the number of seats to be filled reaches further down the list to where female candidates are often placed). 20 However, the opposite effect can also be argued: a high-level threshold of representation is an obstacle for the parliamentary representation of the small parties for which women are often candidates. Many studies have also shown that the size of electoral districts (the other mechanism responsible for the degree of proportionality of a voting system) has influenced women s ability to be elected: the more seats there are per electoral district and, especially, the more seats a party can hope to win the higher are the chances of women obtaining parliamentary seats. 21 Then there are the parties, the true masters of the parliamentary representation of women. According to Matland and Montgomery, The electoral system directly affects female legislative representation, because it shapes the recruitment strategies of party gatekeepers at the nomination phase. 22 Indeed, the voting system may best be characterized as an intermediary variable between the political parties and the legislative representation of women, as it is the voting system that shapes parties strategies for selecting candidates. This perspective is endorsed by Farrell, who argues

7 538 DEMOCRATIZATION that it is not the electoral system which is at fault [for the under-representation of women in parliaments] so much as the party selection committees. 23 Of course, in almost all electoral systems, it is the parties and not the electorate that controls the selection process of candidates and therefore the composition of parliaments. However, the nature of the control that the parties exercise over the selection of candidates varies greatly according to the voting procedure and its particular characteristics. 24 In terms of proportional representation, closed lists and lists without preferential voting give political parties nearly total control over the composition of parliament, as opposed to open lists with panachage or combination of methods whereby the voters play a role in designating the people, and not only the parties, who will represent them. There is, however, no consensus as to which closed or open lists best foster the election of women. 25 In single transferable vote systems, parties have indirect (in other words, preliminary) control over the composition of parliament: female and male candidates must obtain endorsement from the party for which they are campaigning, but votes for endorsed candidates are preferential votes entrusted by the electorate to the individual. These votes, and not the ranking established by the parties, will therefore determine who will hold a seat in parliament. 26 However, the few countries that use a single transferable vote system show low proportions of women in their parliaments. 27 As for plurality/majority systems, the American formula of primaries does not confer any power on the parties to select candidates for Congress, while the Australian, British, and Canadian parties firmly hold the reins of the candidate selection process. Women may, however, benefit from having a standardized and centralized selection process of candidates, as this would allow national elites to more easily adopt positive affirmative measures and quotas. 28 What is certain is that the candidate selection process is not neutral: it is a process of exclusion and inclusion, meaning that some would-be legislative candidates are systematically ruled out, while others are systematically ruled in. Norris and Lovenduski suggest that the party selectors (that is, the party leaders, officers, activists, members, and voters) choose their candidates according to an ideal type of member of parliament that the researchers call homo politicus. For them, [t]he system has been designed to select a standard model candidate who is articulate, well educated and typically employed in a professional career, in business as an executive or manager, in education as a school teacher or university lecturer or in the law as a practising barrister or lawyer [...] By defining the appropriate qualifications for a career in politics in such a way then certain types of candidates will tend to be successful. As a result women, workingclass candidates and those from the ethnic minorities will tend to be consistently disadvantaged. 29 Gender quotas are one strategy used to counteract the kind of informal (or systemic) discrimination that selectors may impose against would-be female candidates. 30 Since the beginning of the 1990s, approximately 90 countries have adopted some form of affirmative action (be it reserved seats, legal constitutional or legislative quotas, or party quotas) to advance the representation of women in lower or single houses of their parliaments. Gender quotas ensure that women make up a certain

8 DEMOCRACY, REPRESENTATION, AND WOMEN 539 minimum percentage of the candidates on each party s list (or minimum percentage of MPs in the case of reserved seats). Today, most quota systems aim to ensure that women constitute at least a critical minority of per cent. Feminist mobilizations on the electoral scene, both in parties and in civil society, are another strategy to counteract the discrimination that party selectors may perform against women would-be candidates. 31 Rule goes so far as to state that such mobilization can be a counterweight to the resistance that certain voting procedures present to the election of women: [n]egative electoral system features have been overcome by women s political mobilization. 32 It goes without saying that feminist mobilizations and quotas are not opposing strategies since, as Krook 33 argues, women s mobilization is one of the causal accounts explaining why gender quotas are adopted. Apart from the effect of voting systems as such, many other aspects of elections can influence women s access to legislative assemblies. For example, many studies have shown that the electorate does not exercise negative discrimination toward candidates but neither does it discriminate positively. 34 Other current research casts doubt on the hypothesis that women typically stand as candidates in electoral districts considered to be already lost, suggesting that they are, at least, no more likely than men to do so. 35 And while the cost of electoral campaigns has long been considered a major obstacle to the election of women, many recent works have questioned its relevance in explaining the low proportion of women in parliaments. 36 In sum, there is an array of factors (cultural, socio-economic, and political) influencing women s access to parliaments around the world. That said, it is now time to turn to the more specific question: what factors help or hinder women in entering parliament in countries defined as democratic? As mentioned above, democracy does not necessary constitute a springboard for the feminization of parliaments (as we see from the relatively low proportion of women in many well-established democracies, such as Japan). Intriguingly, though, the proportion of women in parliaments is more and more seen as an indication of the state s quality of democratic life. This shift in paradigm was made possible by a host of political events (including the third wave of democratization, the end of the Cold War, and the political restructuring of former Warsaw Pact member countries) and, especially, by the mobilizations of the women s movement in political and electoral arenas both national and international. Many international instruments, such as the Convention on the Elimination of All Forms of Discrimination against Women (1979) and the Universal Declaration on Democracy (1997), are fundamentally marked by this new paradigm. Following this reasoning that democracy can no longer be conceived of without women the analysis here is limited to the 89 countries classified as democratic in 2005 by the Gastil Index. 37 The next task is to target within states of this type (which, by the way, do not each offer an equal quality of democratic life) the factors that contribute to or limit the election of women to parliaments. Research Design The principal argument of this article is that the factors that help or hinder women in entering parliament differ according to the length of the democratic experiment.

9 540 DEMOCRATIZATION In order to gain a better understanding of women s legislative representation in democratic countries, the following research design was developed. The dependant variable chosen is the proportion of women in the lower or single houses of parliaments (as of February 2006). The countries included are those that are deemed free according to the Freedom House rating and democratic according to the Gastil index, which is based on Freedom House data. This criterion may, however, be misleading in that a country could be free in 2005 but not before, as in the case of Senegal. So the analysis separates countries free in 2005 into two categories: those that were free at least half of the time between 1972 and 2003, and those that were free for less than half of that time. The first category includes 48 out of the 89 countries that were free in The democratization that occurred in many countries following the changes of the 1990s has without doubt created pressure for the feminization of parliaments, since it was at this point that the presence of women in parliaments became a criterion of primary importance in evaluating the quality of democracy achieved by given countries. This then creates the category of democratic states since at least 1990, which is made up of 49 countries. For the most part, these two categories overlap. 38 The Appendix lists the countries in each category. The mean proportion of female legislators in the 41 countries that have been free for less than half of the period is 15.7 per cent, while for the 48 countries that have been free for at least half of that time the corresponding proportion is 19.6 per cent. This gap is not, however, significant at to the conventional.05 level (7.5 per cent). The independent variables are those identified by previous research (discussed above in the Review of the Literature) as having an impact on the election of women. They are classified according to two categories: (i) Cultural and socio-economic: This set of variables refers to the shift from traditional toward modern society. It is postulated that modernity (broadly defined) enlarges the pool of eligible women by offering them the necessary conditions to be involved in politics. 39 The cultural and socio-economic indicators used here include, among others: the female literacy rate, the proportion of women enrolled in tertiary education; gender role values (traditional or egalitarian 40 ); the HDI score; the score on the gender empowerment index; the fertility rate; Gross Domestic Product per capita; the level of poverty (as measured by the Gini Index 41 ); the ratio of female to male earned income; the female economic activity rate as a percentage of the male rate; the percentage of all professional and technician workers who are female; the urban population as a percentage of the whole; and the level of public expenditure on education and health. (ii) Political: Variables in this category concern two dimensions: the state framework of governance and the achievements of women in politics. The former encompasses indicators such as: the structure of the state (unitary or federal); the structure of the national legislature (for example, the number of houses one or two, the maximum length of the legislature, the number of seats in the lower or single house); the electoral system (plurality/majority, proportional or combined); the level of proportionality; the number of electoral districts;

10 DEMOCRACY, REPRESENTATION, AND WOMEN 541 the district magnitude; the party list (closed or open); the electoral threshold; the electoral formula (such as d Hondt, Hare, or St Laguë); the party system (for example, the effective number of electoral and parliamentary parties); the conditions of candidacies (whether a monetary deposit is required for candidacies; whether candidates campaign costs are subsidized); and electoral turnout. Indicators of the political achievements of women include: the year women got the right to vote and to stand for national elections; the year the first woman was elected to the lower or single house; the presence of affirmative action measures for women in politics (such as gender quotas); and whether there has ever been a woman head of state (as president) or chief of government (as prime minister). The principal sources of data 42 are: the CIA World Fact Book, 2005; the Elections Around the World database, 2005; Freedom House, 2005; the Global Database of Quotas for Women, 2005; the Human Development Report, 2003; International IDEA (IDEA s Electoral System Design project; IDEA Finance database); the Interparliamentary Union Parline Database and Women in National Parliaments listing, 2005; the Political Database of the Americas (PBDA); the Richard Rose International Encyclopaedia of Elections, 2000; and lastly, the World Values Survey. The statistical analyses presented here are a series of bivariate (Pearson) correlations, and multivariate (OLS) regressions. Results Bivariate Analysis Let us first assess the influence that independent cultural, socio-economic, and political variables have on the proportion of women members of parliaments. Table 1 presents the coefficients of simple (Pearson) correlations (without any prior controls). An initial observation concerns the difference between the visible, distinct impact of the indicators in countries which have been free at least half of the time between 1972 and 2003, and the lesser impact of the same indicators in countries that have not been free at least half of this time. That is, these indicators seem more relevant for the former group of countries than for the latter one. Matland 43 has noted the same phenomenon in his study on developed and developing countries: factors known to influence the proportion of women in parliaments (such as women s participation in the labour force) prove to be non-determinant in developing countries. A second observation is that, for countries free at least half of the time from 1972 to 2003, the results generally support the conclusions of other studies (for example, the impact of proportional voting systems and egalitarian attitudes to gender roles). Nevertheless and herein lies a third observation some coefficients are intriguing, such as the non-significance of the indicators female professional and technician workers as percentage of total, number of seats, electoral threshold, electoral formula, and monetary deposit, candidacies. Evidence from previous studies has shown the positive impact of the first two indicators on women s access to parliaments (see the Review of the Literature above). Regarding the other

11 542 DEMOCRATIZATION TABLE 1 COEFFICIENTS OF SIMPLE CORRELATIONS (WITHOUT ANY PRIOR CONTROLS) Free less than half of the period Free at least half of the time from 1972 to 2003 Cultural and socio-economic variables Female literacy rate.217 (40).427 (45) Per cent of women in enrolled Data not available.480 (34) in tertiary education Gender roles values.371 (18).669 (29) Human development index.226 (32).512 (44) Fertility rate.185 (41).435 (48) Gross domestic product.413 (41).510 (48) Level of poverty (Gini Index).119 (26) (28) Ratio of estimated female to.001 (18).431 (31) male earned income Female economic activity rate (30).322 (38) as per cent of male Female prof. and technical (20) (31) workers as per cent of total Urban population.228 (32).474 (44) Public expenditure health.220 (32).486 (44) Public expenditure education.270 (29).191 (42) Political variables Structure of the state (unitary or (41).213 (48) federal) Uni/bicameral (41).010 (48) Max. length of legislature.204 (41) (48) Number of seats.127 (41).149 (48) Voting system.526 (40).577 (48) Level of proportionality.530 (27).412 (34) Number of constituencies (39) (48) Constituency magnitude.396 (39).511 (48) Open/closed lists.049 (28).150 (27) Electoral threshold (14).493 (11) Electoral formula.084 (22).050 (24) Number of parties competing.130 (38).242 (47) ENPP.077 (36).379 (45) Monetary deposit, candidacies.119 (36).252 (45) Candidates campaign costs.210 (32).572 (44) subsidized Electoral turnout.117 (38).190 (48) Variables relating to women s political status Gender empowerment index.231 (29).473 (38) Year, vote national elections (41) (48) Year, stand national elections (41) (48) Year, 1st woman elected to the (38) (47) lower or single house Quotas.168 (41).313 (48) Woman as president or PM.148 (41).159 (48) ENPP is Effective Number of Parliamentary Parties. Significant at the 0.05 level. Significant at the 0.01 level

12 DEMOCRACY, REPRESENTATION, AND WOMEN 543 indicators, one may reasonably suggest that a high electoral threshold (which, in absence of apparentement, 44 bars legislative representation to the small parties for which women are often candidates), an electoral formula favourable to large parties (such as d Hondt and Imperiali highest average), and the obligation to pay a deposit to stand for elections exert a negative impact on women s parliamentary representation. Instead of this, the bivariate correlations reveal no significant relationship. The modest significance of the quotas indicator is also surprising. The result obtained for the maximum length of legislature indicator is also interesting because it suggests that legislative mandates that are shorter (and therefore that give rise to more frequent elections) favour women s access to parliament. However, the example of the House of Representatives in the United States shows the limitations of this reasoning. Finally, the non-significance of the open/closed lists indicator highlights the absence of consensus on the question of which lists open or closed favour the election of women. In sum, the results of the bivariate analysis tend to support the principal argument of this paper that factors influencing the percentage of women in parliament vary according to the length of the democratic experiment. However, it is likely that these relationships are much more complex than can be described by bivariate analysis, particularly because several factors interact to influence the feminization of legislative assemblies. Multivariate Analysis In the second step, we assess the impact of the independent cultural, socio-economic and political variables on the proportion of women members of parliaments, this time using multivariate analysis to also consider the influence these variables exert on one another. Although one of the strengths of multivariate analysis is that it takes into account the influences that independent variables have on each, this also produces weaknesses. In particular, the results of OLS regressions analysis may be distorted by collinearity relationships. For instance, the correlation between the HDI score and gender role values indicators is (P ¼ 0.002), and between the voting system and the level of proportionality it is a strong (P ¼ 0.000). The condition index allows researchers to summarize the state of the colinearity relationships in a regression. A common rule of thumb is that a condition index over 15 indicates a possible multicollinearity problem, and a condition index over 30 suggests a serious multicollinearity problem. Table 2 shows the standardized beta coefficients of OLS regressions performed on the percentage of women in the lower or single houses of parliaments in countries which have been free for at least half of the time between 1972 and None of the models displayed in Table 2 have a condition index higher than 15. The analysis is based on the 48 countries that have been free for at least half of the time from 1972 to 2003, a period for which the gender role values index is available. Model A includes two cultural and socio-economic indicators: gross domestic product and gender role values. Results confirm those obtained by other studies, 45 notably that socio-economic factors have a secondary influence and that the cultural dimension (particularly an egalitarian conception of gender roles) has a

13 544 DEMOCRATIZATION TABLE 2 STANDARDIZED BETA COEFFICIENTS (OLS REGRESSIONS ANALYSIS), AS FOR THE DEPENDENT VARIABLE THE PERCENTAGE OF WOMEN IN LOWER OR SINGLE HOUSES OF PARLIAMENT IN COUNTRIES FREE AT LEAST HALF THE TIME FROM 1972 TO 2003 Model A Model B Model C Model D Cultural and socio-economic dimension Gross domestic product Excl. Excl. Excl. Gender roles values Political dimension Voting system ENPP Women s political status Women s rights index Excl. Quotas index Nb Adjusted R square Condition index Note: Result excluded. P 0.05; P 0.01; P primary influence on the proportion of women MPs. Because the gross domestic product indicator adds nothing to the explanation, it was withdrawn from Models B, C and D. Model B adds the important factor of the electoral design to the OLS regressions analysis. This model is the most convincing in explaining the proportion of women in the lower or single houses of parliaments. Indicators were entered as they appear: gender role values, voting system, and Effective Number of Parliamentary Parties (ENPP). First, it seems that an egalitarian conception of gender roles (and notably the conviction that women and men should share political positions equally) may inspire electoral design (as shown by recent experiences in Afghanistan and Iraq) or lead to the adoption of measures to improve women s chances of being elected in politics (as in Belgium and the United Kingdom). Second, it is a well-established fact that the voting system has a major impact on the effective number of parliamentary parties. 46 All the variables satisfy the minimal 0.05 level of significance, and the low condition indexes suggest that multicolinearity problems are absent. The results in Model B confirm those obtained by other studies, notably the positive impacts of an egalitarian conception of gender roles and a proportional voting system on the election of women in parliaments. 47 These results also go some way toward confirming the influence, previously suspected, of the effective number of parliamentary parties: having more parties in legislature offers women more opportunities to be elected. 48 Models C and D which add to the analysis the dimension of women s political rights are interesting. Model C shows that the years in which women earned the rights to vote and to stand in national elections, as well as the year when the first woman entered the parliament, apparently have no influence on the proportion of female legislators. 49 This contradicts the conclusions of other studies which have shown a relationship between the year women won the right to vote in national elections and the percentage of women MPs. 50

14 DEMOCRACY, REPRESENTATION, AND WOMEN 545 As for Model D, results may be more discouraging than intriguing. These results demonstrate that quotas (whether reserved seats, legal constitutional or legislative quotas, or party quotas 51 ) are not a significant factor in explaining the percentage of women parliamentarians. This observation is disappointing, considering the expectation, held by many, that these mechanisms would increase the number of women in parliaments. One might argue that it is not quotas but penalties for non-compliance that have an impact on the proportion of seats occupied by women in parliaments. Here again this indicator is not significant (data not shown). The French experience with parité is a sad but obvious example of the limits of quotas: despite constitutional, legislative, and party quotas as well as financial penalties imposed on political parties for contravention of the parity electoral law, in the 2002 elections, the year the parity law was first applied at the legislative level, the proportion of women in the Assemblée nationale increased only marginally from 10.9 per cent to just 12.2 per cent. As Sineau and Tremblay 52 note, the French political parties have preferred to pay penalties than to select female candidates. As mentioned in the Review of the Literature, if quotas can be justified to overcome the aversion of the population toward the political participation of women, their influence is still not guaranteed when it comes to recalcitrant political parties hostile to women s presence in their ranks. The statistical non-significance of quotas as a determinant of the presence of women in parliaments is a very surprising result, which certainly calls for further analysis. We now turn to countries that have been free for less than 50 per cent of the time between 1972 and Table 3 shows the standardized beta coefficients of OLS regressions analysis performed on the proportion of female legislators. An initial general observation is that the proportion of total variation explained by the four models used to analyse these countries is much less than for countries that have TABLE 3 STANDARDIZED BETA COEFFICIENTS (OLS REGRESSIONS ANALYSIS), AS FOR THE DEPENDENT VARIABLE THE PERCENTAGE OF WOMEN IN LOWER OR SINGLE HOUSES OF PARLIAMENT IN COUNTRIES FREE FOR LESS THAN HALF THE TIME FROM 1972 TO 2003 Model A Model B Model C Model D Cultural and socio-economic dimension Gross domestic product Gender roles values.437 Excl. Excl. Excl. Political dimension Voting system Max. length of legislature.225@ @ Women s political status Women s rights index Quotas index.199 Has had woman as president or PM.233@ Nb Adjusted R square Condition index Excl.: Indicator excluded; :P 0.05, :P 0.01, :P P 0.10.

15 546 DEMOCRATIZATION been free for at least half of the relevant period. Indeed, for the latter countries, the adjusted R-square varied between 40.5 per cent and 63.4 per cent, while it was between 37.2 per cent and 40.8 per cent for the others (taking 39 countries into account). This difference is not surprising when we consider the more modest scope of Pearson correlations obtained for countries free for less than half of the time between 1972 and 2003 (see Table 1). Second, it must be noted that in these models the gender role values indicator is absent and has been replaced by the indicator for Gross Domestic Product. On one hand, data for the gender role values indicator is available for only 17 countries in this category, limiting considerably the possible scope of OLS regressions analysis. On the other hand, the average of the gender role values indicator in countries free for less than 50 per cent of the time between 1972 and 2003 is significantly less than in countries that have been free for at least half this period (on a scale of per cent, 51.1 per cent compared with 62.4 per cent [P 0.01]). This result helps us to understand that in juxtaposition with other indicators (such as voting system ), the gender roles values indicator loses all its statistical weight, unlike the gross domestic product indicator (keeping in mind that the relationship was the opposite in Table 2). A final comment concerns the condition indexes, which were higher in Table 3 than Table 2, suggesting possible multicolinearity problems within those models. These results confirm that the explanatory model for the proportion of women in parliaments differs according to the length of the country s democratic experience. This supports the principal argument developed in this paper. In countries that were free for less than half of the time between 1972 and 2003, the gender roles values indicator has no statistical effect, while the voting system indicator is an imposing presence in the foreground of all explanatory models. In countries that were free for at least half of this period, an egalitarian conception of gender roles clearly dominates explanatory models, to such a point, in fact, that it relegates the voting system indicator to an almost insignificant status. It might be worth exploring the hypothesis that a plurality/majority voting system does not constitute a barrier to the election of women in parliaments in societies where an egalitarian conception of gender roles dominates. A good illustration of this is the intriguing case of Québec 53 where, despite a first-past-the-post system and a strong and well-established twoparty arrangement, in per cent of members of the Québec s Assemblée nationale are women. The role played by the maximum length of legislature indicator is quite surprising: the fact that this coefficient is positive in Models B, C, and D indicates that longer legislatures favour the election of women to parliaments. This result is counterintuitive because it is reasonable to believe the opposite: shorter legislatures should mean more frequent elections, which should in turn increase the opportunities for women to run for office and, possibly, to be elected. Moreover, the positive effect of the maximum length of legislature indicator cannot be explained by collinearity with the voting system indicator (Pearson correlation: 2.008). In short, this indicator deserves more attention from researchers. Like Models C and D in Table 2, Model C in Table 3 confirms the statistical nonsignificance of the women s rights index and quotas index indicators in

16 DEMOCRACY, REPRESENTATION, AND WOMEN 547 determining the proportion of women MPs. However, Model D in Table 3 reveals an interesting finding: a woman president or prime minister influences positively though modestly the percentage of women in parliaments. Although this result is very preliminary and has yet to be supported by further research, it provides supporting evidence for a traditional explanation of the low presence of women in politics that is, the lack of role models. It seems that the presence of women in the top echelons of power contributes to larger proportions of women parliamentarians, at least in countries with less experience of democracy. In sum, the results of both the bivariate and multivariate analyses support the main argument of this paper: women s access to parliaments does not follow a single pattern but is instead structured by an array of factors, the importance of which is determined, at least in part, by the length of the democratic experiment. Conclusion What factors help or hinder women in entering parliament in countries defined as democratic? In answering this question, we must return to a point made at the outset: in practice, democracy is a complex, heterogeneous, and multifaceted phenomenon. The results drawn from the analyses carried out as part of this research illustrate this: there is no homogenous and consistent set of variables that can explain, as if by magic, the proportion of women in parliaments. Rather, these factors vary according to the context in which they are found and, notably, according to the length of the democratic experiment. Therefore, an initial conclusion is that we should refrain from uniformly applying indicators to all countries, whose realities are, in fact, quite different. The results presented here show that the explanatory power of indicators and regressions models varies according to the length of each country s democratic experiment. For example, though an egalitarian conception of gender roles is a quite convincing indicator in explaining the proportion of female MPs in countries that have been free for at least half of the time between 1972 and 2003, the relevance of this explanation fades when it comes to countries that have been free for less than half of the same period. It would be valuable to refine the indicators used to be more sensitive to the cultural, socio-economic, and political factors prevalent in these countries. A second conclusion is that there are variables we should integrate into our models of analysis but which are not easily adapted to a statistical approach. As mentioned in the Review of the Literature, several researchers have shown that feminist mobilizations on the electoral scene both in parties and civil society mobilizations that are without doubt a component of empirical democracy are influential strategies in increasing the number of female legislators. For Rule, 54 such mobilizations may counterbalance the negative effects that some voting systems have on the election of women. The difficulty then is to translate these mobilizations into a workable and significant numerical indicator for the statistical analysis. The final conclusion brings out the complexity of the relationships between empirical democracy, voting systems, and an egalitarian conception of gender

17 548 DEMOCRATIZATION roles. In countries where democracy has prevailed for only a short time, the voting system constitutes the most important factor in explaining the proportion of women parliamentarians, while in well-established democracies this explanation is found in an egalitarian conception of gender roles. The relegation of the voting system to the background in established democracies makes it necessary to qualify an argument that too many political scientists and feminist activists endorse and transmit without critical evaluation: that proportional representation favours the election of women and majority voting discourages it. This relationship, which is often presented as a causal effect, should be the subject of further research, after which we might find that it may well fall from the pedestal it currently occupies in the women and electoral politics field. ACKNOWLEDGEMENTS This article was prepared while the author was a visiting researcher at the Australian National University (ANU). She thanks the Department of Political Science in the Faculty of Arts for welcoming her. The paper on which the article is based won the Gender & Politics Wilma Rule Award 2006 of the International Political Science Association. The author would also like to thank Merrindahl Andrew (ANU) for her linguistic assistance. NOTES 1. See Joan Wallach Scott, Parité! Sexual Equality and the Crisis of French Universalism (Chicago, IL: University of Chicago Press, 2005) (accessed 14 November 2006). 3. Robert A. Dahl quoted in Arend Lijphart, Democracies. Patterns of Majoritarian and Consensus Government in Twenty-One Countries (Oxford: Oxford University Press, 1984), p See, among others: Pippa Norris, Conclusions: Comparing Passages to Power, in Pippa Norris (ed.), Passages to Power. Legislative Recruitment in Advanced Democracies (Cambridge: Cambridge University Press, 1997), pp ; Pippa Norris and Ronald Inglehart, Cultural Obstacles to Equal Representation, Journal of Democracy, Vol. 12, No. 3 (2001), pp ; Dawn Nowacki, Women in Russian Regional Assemblies: Losing Ground, in Richard E. Matland and Kathleen A. Montgomery (eds), Women s Access to Political Power in Post-Communist Europe (Oxford: Oxford University Press, 2003), pp ; Pamela Paxton, Women in National Legislatures: A Cross-National Analysis, Social Science Research, Vol. 26, No. 4 (1997), pp ; Jacqueline Peschard, The Quota System in Latin America: General Overview, in International IDEA, The Implementation of Quotas: Latin American Experiences (Stockholm: International IDEA, 2003), pp. 20 9; Andrew Reynolds, Women in the Legislatures and Executives of the World: Knocking at the Highest Glass Ceiling, World Politics, Vol. 51, No. 4 (1999), pp ; Wilma Rule, Electoral Systems, Contextual Factors and Women s Opportunity for Election to Parliament in Twenty-Three Democracies, Western Political Quarterly, Vol. 40, No. 3 (1987), pp Ronald Inglehart and Pippa Norris, Rising Tide. Gender Equality and Cultural Change around the World (Cambridge: Cambridge University Press, 2003), pp ; Mercedes Mateo Diaz, Representing Women? Female Legislators in West European Parliaments (Colchester: ECPR Press, 2005), p. 64; Pamela Paxton and Sheri Kunovich, Women s Political Representation: The Importance of Ideology, Social Forces, Vol. 82, No. 1 (2003), pp Paxton and Kunovich (note 5). 7. Pippa Norris and Joni Lovenduski, Political Recruitment. Gender, Race and Class in the British Parliament (Cambridge: Cambridge University Press, 1995), pp R. Darcy, Susan Welch and Janet Clark, Women, Elections, & Representation, 2nd edition (Lincoln: University of Nebraska Press, 1994), pp Richard E. Matland, Women s Representation in National Legislatures: Developed and Developing Countries, Legislatives Studies Quarterly, Vol. 23, No.1 (1998), pp ; Richard E. Matland,