Democracy, Gender Equality, and Collective Violence

Transcription

1 Democracy, Gender Equality, and Collective Violence - Situating East Asia- Elin Bjarnegård Erik Melander To be presented on Tuesday, April 03, at 4:00 PM, in the Greenroom. Panel TD65: Regional Peace Studies: Exploring Regional Trends in Peace and Armed Conflicts.

2 Abstract We contribute to answering the puzzle of why democracy seemingly doesn t matter for collective violence internally by applying insights and arguments from the research on gender and war, thereby fusing two theoretical discourses. Research has found strong relationships between gender equality and lower levels of collective violence, such as war and violations of personal integrity rights. Much research has tried to find any influence of democracy on collective violence within societies, but the results are inconclusive. We argue that democracy has the potential to lower levels of violence, but only when dominant norms in society have become less warlike. We further argue that warlike norms are intimately linked to forms of masculinity that glorify violence and prescribe the subordination of women. Such norms are likely to result in societal gender inequality. We test our argument about the way democracy and gender equality interact with regard to two different dependent variables: use of torture and onset of intrastate armed conflict, and find support for our hypotheses. In the light of these insights we address the puzzle of the role, or perhaps rather the absence of the role, of gender equality in explaining how states in East Asia keep internal peace. On some indicators of gender equality, many states in East Asia show very high levels. In many of these states with seemingly more equal relations between men and women intrastate armed conflicts have been rare. Yet, these states tend to be very oppressive and undemocratic. At the same time, these East Asian states tend to disregard women s social rights even if more formal indicators of gender equality, such as the percentage women in parliament, would suggest that women and men are not that unequal in the region. The arguments and findings of this paper sheds new light on how undemocratic political institutions combined with low levels of women s social rights most likely have contributed to the oppressive nature of the internal peace in many of the East Asian states.

3 Introduction What does it take for democracy to matter for collective violence? This paper revisits this fundamental question in light of the new knowledge about the importance of societal gender equality for the reduction of violent conflict. Numerous studies have shown that countries that exhibit high levels of gender equality in different spheres of society also tend to be more peaceful fewer internal conflicts (e.g. Caprioli 2005; Melander 2005; Gleditsch, Wucherpfennig et al. 2011), and less human rights abuse on the domestic arena (e.g. Melander 2005), fewer external conflicts (e.g. Caprioli 2000; Regan and Paskeviciute 2003; Caprioli 2005), and, when international disputes do occur, they are treated with less aggression (e.g. Caprioli 2003) and there are better conditions for subsequent peacekeeping (e.g. Gizelis 2009). Thus, diverse forms of gender equality seem to go hand in hand with decreased conflict levels in various arenas. This strong and robust relationship has been interpreted in different ways, few of which have emphasized the role of democracy (exceptions include Melander 2005; Bjarnegård and Melander 2011). Although democracy has been included as a control variable in most of the above studies, this paper argues that its real impact has been overlooked. Meanwhile, recent research on the causes of civil war has concluded that the level of democracy seems to b e unrelated to the risk of civil war (Strand 2006; Vreeland 2008), and has instead shifted the focus to other characterizations of regime types. For example, Fjelde (2010) finds that single-party authoritarian regimes are particularly resilient to armed challenges to their authority. This disappointing absence of a clear, peace-promoting effect of democracy is puzzling, and directly challenges the many recommendations to promote democracy as a way of building peace in conflict-ridden societies (see e.g., Melander et al, 2004). We demonstrate that democracy does matter for the reduction of collective violence of different kinds, and we specify how. We show that the function of democratic political institutions with regard to lowering the risk of violence is to enable the pacifying effect of the

4 norms associated with gender equality to influence policy. The flip side of this insight is that when in a democracy the oppression of women is particularly harsh the accompanying norms of belligerence and intolerance may poison policy through the open political institutions, and thereby increase the risk of violence. In line with this argument, we show that in a society characterized by dismal respect for women s rights the risk of intrastate armed conflict is actually higher when institutions are democratic then if they are authoritarian. That the pacifying influence of gender equality is modified by the level of democracy was shown by Melander (2005). In this article we draw the implications of this interplay for understanding the role of democracy in mitigating or aggravating conflict. We also develop the explanation for why and how gender equality matters for conflict. By bringing back democracy into the picture, we can offer a new and nuanced interpretation of the role of gender equality. While others have argued that the mechanism binding gender equality and peace together is the increased influence of peaceful-minded women (Caprioli and Boyer 2001; Caprioli 2003; Regan and Paskeviciute 2003), we argue that the combination of constraints that democracy puts on its leaders, on the one hand, and broad societal normchanges altering the attitudes of women and men alike, on the other, is what shapes society into a more tolerant and peaceful one. We move on by reviewing literature about democracy and democratic quality. We show that arguments about democracy are largely lacking in the literature about gender and peace, and we discuss how and why such arguments are needed. We also demonstrate how we depart from earlier interpretations of the relationship between gender equality and peace. We then elaborate on our operationalizations and on our statistical model, before putting our theoretical argument to empirical scrutiny in a regression model. In the light of these insights we then address the puzzle of the role, or perhaps rather the absence of the role, of gender equality in explaining how states in East Asia keep internal peace. On some indicators of gender equality, many states in East Asia show very high levels.

5 In many of these states with seemingly more equal relations between men and women intrastate armed conflicts have been rare. Yet, these states tend to be very oppressive and undemocratic. At the same time, these East Asian states tend to disregard women s social rights even if more formal indicators of gender equality, such as the percentage women in parliament, would suggest that women and men are not that unequal in the region. The arguments and findings of this paper sheds new light on how undemocratic political institutions combined with low levels of women s social rights most likely have contributed to the oppressive nature of the internal peace in many of the East Asian states. Does Democracy Matter? In its most fundamental meaning, democracy is a manifestation of the incontestable equal worth of all human beings. One person-one vote means that, at least in a free and fair democratic election, everyone counts as equal. As a result, there have been formidable expectations on democracy to deliver all kinds of desirable societal results, such as gender equality and peace (Bjarnegård and Melander 2011). However, the formal democratic equality ideal is of course not directly translated to democratic societies. Democracies, just like other types of regimes, exist in societies that exhibit different degrees of equality and inequality between different groups and individuals. Formal democratic institutions are not enough to ensure a population that values equality among its citizens. Democratic institutions can be installed in and even imposed on a country. There is an expectation that formal democracy will foster democratic citizens, but empirical research on the interplay between democracy and gender equality has, by and large, been inconclusive. Countries with higher levels of democracy do not necessarily have more inclusive political bodies or a population with more gender equal values (e.g. Paxton 1997; Jaquette and Wolchik 1999; Inglehart, Norris et al. 2002; Bjarnegård and Melander 2011). Inglehart, Norris and Welzel see both gender equality and democratization as parts of a broader postmodernization trend,

6 but they also acknowledge that the two do not automatically follow upon each other (Inglehart, Norris et al. 2002). Certainly, there are countries with democratic systems where women still are regarded as second class citizens, not enjoying the same social rights of work and income as their male fellow-citizens. India is one example of a country with a long standing democracy where women s social rights are consistently low and where extreme measures of gender equality, such as the sex ratio at birth, also provide evidence of son preference in certain parts of the country. Other countries have climbed quite high on certain indicators of societal gender equality such as work-force participation and political representation, while maintaining a non-democratic regime. Many communist states, such as Cuba and China, are examples of this combination. It is thus important to separate formal, or basic, aspects of a democratic procedure from more qualitative societal components such as gender equality, tolerance, corruption or conflict - so that we can investigate to which extent one follows upon the other, and how they interact and impact on one another (O'Donnell 1994; Inglehart, Norris et al. 2002; Hadenius and Teorell 2005; Bjarnegård and Melander 2011). Another basic feature of democracy is that it is a way of settling disputes and solving conflicts peacefully, by debating in parliament and voting in polling booths instead of taking to arms. But just like in the case of equality, it can be questioned to what extent this is an ideal typical illustration of the normative value of democracy, and to what extent democracy actually empirically matters for avoiding violent outbreaks and resolving armed conflict. Democracy has been prominent in the literature on peace and conflict, and the most important association that has been established is the so-called democratic peace theory, which holds that two democratic regimes are very unlikely to fight each other. There is, however, little consensus about how democracy relates to collective violence within a society. Strand (2006) and Vreeland (2008) showed that when using appropriate measures for the level of democracy, no significant pacifying effect of democracy could be found. The study of the role of regime type for collective violence has instead come to focus mostly on other aspects of

7 political institutions than democracy. For example, Fjelde (2010) demonstrated that singleparty authoritarian regimes are particularly unlikely to become embroiled in intrastate armed conflict, and Gleditsch and Ruggeri (2010) found that controlling for the level of democracy, the ascension to power of a leader through irregular means increases the risk of civil conflict. 1 Despite the paucity of clear and robust results, democracy is often prescribed as a good inhibitor to violent conflict in the policy literature (see, e.g., Melander et al. 2004). Against this background, the question of why democracy seems incapable of significantly reducing the risk of internal collective violence is arguably of paramount importance. Literature on gender and conflict has, for the most part, not seriously considered what the interplay between democracy, gender equality and conflict might look like. The aim of this paper is therefore to revisit democracy within the context of gender and conflict, and to bring together lessons learned from different fields. If we limit our definition of democracy to the most basic criteria, it means that the more democratic a country is, the less arbitrarily selected are the rulers and the more constrained are they by rules and regulations about what they can, can t and should do. The more democratic a country is, the more established are various representative channels - elections being the most obvious example - through which political attitudes and ideas of the citizens are communicated to its representatives. The more democratic a country is, the more important is it for the leadership of a country to acknowledge the views of its constituents and possibly also to act accordingly. Leaders in a democracy are held accountable for their actions by the citizens, and if they depart too far from the values of the voters, they will lose legitimacy and, ultimately, the election and political power. In this way, then, democracy matters for the likelihood that values and attitudes in the population will translate into actual governmental actions and policies. Trumbore and Boyer (2000) downplay the differences between the actions of democratic and autocratic states in this regard, but mainly in times of international crises. They claim that democratic leaders are generally involved in a two way process in which leaders have to look to constituents as well as to international counterparts,

8 and try to balance the wishes and demands of the micro and macro level. During crises, however, democracies tend to converge towards autocracies, becoming more closed and less attuned to the political concerns of citizens. This indicates that there is, certainly, a difference in how different political systems respond to and interact with the attitudes of its citizens. Democracies are hardly surprising and by their very institutional design - more responsive. It is, however, important to note that the claims, wishes and attitudes to which they respond can and do vary greatly between different democratic countries. The importance of the democratic system for questions of gender equality and violent conflict is thus in its interaction with the norms and values of its citizens. This proposition - that democracy is, in a sense, an efficient mirror and at least a higher likelihood of having the attitudes of the citizens being reflected in leadership and policy is also bad news for the expectations we can have on democracy. If attitude changes do not go hand in hand with democratization, then any attitudes equal or not equal, tolerant or intolerant, peaceful or belligerent may be mirrored in domestic and foreign policies. Democracy matters but how it matters depend on its citizens. Changing Norms and Attitudes The literature about gender equality and peace generally takes its starting-point in a gender gap in public opinion when it comes to the state s use of violence. In most, but not all (e.g. Tessler and Warriner 1997; Tessler, Nachtwey et al. 1999), studies of public opinion, men are more likely to advocate or support violent solutions in general, such as using military means or initiating a war in a foreign country than women are (Conover and Sapiro 1993), and men express more belligerent attitudes than women when it comes to concrete questions of foreign and domestic policy (Page and Shapiro 1992; Peterson 1992; Tickner 1992; Wilcox, Hewitt et al. 1996). The argument then goes that if women are less likely to support the use of force, the inclusion of women as equal members of society, with equal political, social and

9 economic access, would impact foreign policy (Caprioli 2005, p. 56). The idea is that states in general externalize their domestic political attitudes and behavior and that the norms of the citizens become the norms of the state (e.g.doyle 1986; Russett 1990; Peterson 1992; Tickner 1992; Maoz and Russett 1993; Togeby 1994). As argued above, we find it likely that the degree to which the norms of the citizens become the norms of the state is likely to vary depending on the level of democracy. A merely procedural democracy is not necessarily populated by tolerant citizens. This is a notion that this field has hitherto failed to incorporate into its hypotheses, with the notable exception of Melander (2005b). Most studies acknowledge that the gender gap identified might very well be due to socialization and constructed gender roles rather than to essentialist sex differences but they nevertheless use this gap between men and women as the starting-point and as an argument for why societal gender equality matters for issues of war and peace. Many studies result in direct policy recommendations about lowering fertility rates, increasing the work force participation of women or ensuring the political representation of women in order to reduce violent conflict (e.g. Caprioli 2000; Regan and Paskeviciute 2003). We depart from a primarily constructivist view of gender and argue that gender roles and attitudes about the suitable roles of women and men in society can and do change. Roles and expected behavior of men and women vary greatly across time and space. We should thus not focus so much on the societal influence of individuals (women) who at present hold more dovish attitudes, which would imply using the existing and potentially temporary gender gap instrumentally. Instead, we should focus on the prospects for diminishing the gender gap altogether and understanding where, and under which circumstances both men and women, but especially men, may give up on belligerent attitudes and instead embrace more tolerant and equal values. The differences in values between societies and entire populations, rather than within societies and male and female members of their populations, should be in focus. We thus concur with Melander who, drawing on Tickner (1992, 2001), and Goldstein (2001) argued that in societies characterized by more equal gender roles, the norms of respect and

10 inviolability that characterize an individual's closest relations are also expected to carry over to more distant relations, thereby strengthening societal norms that reject abuse and violence (2005b, p 698; see also Gleditsch et al 2011, p 268). The remaining question is, however, whether it takes a democratic regime for these norms of respect and individual inviolability to translate into government policy. Inglehart and Norris (2003) in their large study on gender equality and cultural change around the world, confirm the claim that gender differences are, indeed, smaller in societies with more tolerant, gender equal and postmaterial values. Broadly, economic development gives rise to a modernization process that changes attitudes and values towards more gender equal values. Such values and norms, emphasizing the equal rights and values of the two groups of the population that are relevant in every society in the world, also reflects deeper values of tolerance and respect for human rights. In many cases, they also lead to support for democracy. There is a need to be cautious when it comes to certain indicators of gender equality that may reflect artificial changes, rather than real norm changes. This includes representation of women in parliament in countries where gender quotas are used to political bodies without any real influence. Again, communist countries are examples of this (see Bjarnegård and Melander 2011). For China, fertility rate can also be seen as an artificial measurement, not really indicating that women are empowered enough to choose to have fewer children, but rather that the state has imposed a one-child-policy. The sex-ratio at birth is an extreme example of the fact that girls and boys are not seen as having equal value in China, with 40 million females missing due to selective abortion (Hudson et al. 2008/2009). We thus need to be careful to select measurements of societal gender equality that are likely to mirror real attitudinal norm changes, not just cosmetic government policies. Hypotheses

11 Our argument laid out in the preceding sections can be summarized as follows. Democratic political institutions are themselves to some extent manifestations of the ideals of the equal worth of all human beings and peaceful settlement of disputes. In practice, democratic institutions tend to translate the prevailing norms of society into policy, although this reflection of societal norms in policy outcome cannot be expected to be perfect. If the actors competing for power are guided by intolerance, unwillingness to compromise, and belligerence, the democratic political institutions are unlikely to produce policy outcomes that are in line with the democratic ideals of equal worth and nonviolence. Instead, oppressive, exclusionary and violent policy outcomes are more likely to result. The peacepromoting potential of democratic political institutions consists in enabling more pacific norms to influence policy. A prime mover in transforming a society from intolerance, unwillingness to compromise, and belligerence is the redefinition of gender roles from militarized masculinity to gender roles that emphasize equality between women and men, and deemphasize the male warrior role. This is why we expect democratic political institutions in combination with societal gender equality to result in less collective violence, and why we expect that democratic political institutions in a society with low levels of societal gender equality does not result in lower risks of violence, and may instead even increase the risks. We formulate the following general hypothesis to test our argument: Given democracy, higher levels of societal gender equality reduce levels of collective violence in a society. In order to make our test stronger we will examine two very different forms of collective violence within a society, namely torture and intrastate armed conflict. We thus present the following two hypotheses that differ only with regard to the dependent variable: H1: Given democracy, higher levels of societal gender equality reduce torture. H2: Given democracy, higher levels of societal gender equality reduce intrastate armed conflict.

12 Hypotheses 1 and 2 do not predict anything about the effect of democracy when levels of societal gender equality are low. We argued, however, that in the absence of societal gender equality democracy is insufficient to reduce violence, and we test this with our third and fourth hypotheses: H3: Given low societal gender equality, higher levels of democracy are not associated with less torture. H4: Given low societal gender equality, higher levels of democracy are not associated with less intrastate armed conflict. H3 and H4 are deliberately formulated so that we allow for the possibility that democracy turns out to be systematically related to worse violence in the absence of societal gender equality, although our arguments do not require this pattern. Data, Research Design, and Methods To test our arguments about the interaction of democracy and gender equality in determining the level of violence in society we study two dependent variables, namely and indicator of the frequent use of torture and an indicator of the onset of intrastate armed conflict. Data on torture The data we use for the incidence of widespread torture comes from the CIRI Human Rights Data Project, hereafter CIRI, (Cingranelli and Richards, 2010), and were downloaded via the Quality of Government Dataset (QoG) (Teorell et al., 2011). CIRI defines torture as follows: "The purposeful inflicting of extreme pain, whether mental or physical, by government

13 officials or by private individuals at the instigation of government officials. Torture includes the use of physical and other force by police and prison guards that is cruel, inhuman, or degrading. This also includes deaths in custody due to negligence by government officials (Short Variables Description, p 4). Our first dependent variable is a binary indicator taking the value 1 if torture was practiced frequently in a given year according to CIRI, and 0 otherwise. Data on intrastate armed conflict Our second independent variable is the onset of intrastate armed conflict. Data on involvement in conflict was collected by the Uppsala Conflict Data Program (UCDP) in collaboration with PRIO (Gleditsch, Wallensteen et al. 2002), and has since been updated and expanded by UCDP. For this article we use the UCDP data on onsets of intrastate armed conflict (Harbom and Wallensteen, 2010). Intrastate armed conflict is in line with the Uppsala definition conceived of as a contested incompatibility that concerns government and/or territory where the use of armed force between two parties, one of which is the government of a state and the other one or more opposition groups, results in at least 25 battle-related deaths in a year. Our dependent variable is coded 1 if a country experiences the onset of intrastate armed conflict in a given year. In order to qualify as an onset, a recurrence of fighting in a previously active conflict must have been preceded by a lull of inactivity of five years in that conflict. In robustness tests we change this period of required conflict inactivity. Since several conflicts often are active in one and the same conflict we do not drop years of ongoing conflict inactivity after the year of onset. This way we allow all conflict onsets that fulfill the inactivity requirement to count.

14 Data on gender equality Our measure of societal gender equality comes from the CIRI, and is a four-step indicator of the extent of social rights enjoyed by women in the society in question. The variable Women s Social Rights ranges from 0 to 3. CIRI defines the four steps of the variable as follows: A score of 0 indicates that there were no social rights for women in law and that systematic discrimination based on sex may have been built into law. A score of 1 indicates that women had some social rights under law, but these rights were not effectively enforced. A score of 2 indicates that women had some social rights under law, and the government effectively enforced these rights in practice while still allowing a low level of discrimination against women in social matters. Finally, a score of 3 indicates that all or nearly all of women s social rights were guaranteed by law and the government fully and vigorously enforced these laws in practice. (Short Variable Descriptions, p 8) The rights that CIRI take into account when coding this variable include equal rights to inheritance; travel abroad; obtain a passport; confer citizenship to children or a spouse; initiate a divorce; own, acquire, manage, and retain property brought into marriage; to participate in social, cultural, and community activities; education; choose a residence/domicile; freedom from female genital mutilation of children and of adults without their consent; and freedom from forced sterilization. Data on democracy We use two alternative ways of measuring democracy for our tests. When Frequent Torture is the dependent variable we use the Polity2 index of democracy. It ranges from -10 to 10, with -10 representing the least democratic states and 10 representing states with the most democratic political institutions. In Polity2 years coded as periods of transition during which new institutions are established are prorated across the span of the transition so as to avoid systematically losing data because of missing values. 2 Years coded as complete collapse of

15 central political authority are for the same reason recoded to a neutral score of 0. We add two dummy variables indicating regime collapse and transition, respectively, so as to capture the potentially peculiar effects of these country years. When the onset of intrastate armed conflict is the dependent variable we instead use Executive Constraints, one of several institutional component indicators that together are used by the Polity project to derive the composite Polity2 index of democracy. The reason why we use this indicator Executive Constraints instead of the full Polity2 index when analyzing armed conflict is that research has shown that the Polity2 index is constructed in such a way that violent conflict influences the value of the Polity2 index, which for the purposes of analyzing the causes of violent conflict is a problem in that the dependent and independent variables partially overlap by construction (Strand 2006; Vreeland 2008). Executive Constraints indicates on seven ordered steps (from 0 to 6) the degree to which the executive of a government is constrained by any accountability group, such as a legislature. The lowest value, 0, indicate regimes with no such constraints, and the highest value, 7, denote regimes with parity between executive and legislature, or the subordination of the executive to the legislature. The value of Executive Constraints is not influenced by the presence of intrastate armed conflict by construction, and will thus be a suitable indicator for this study when armed conflict is the dependent variable. Executive Constraints is highly correlated with Polity2 (0.948), and in practice the two variables thus measure almost the same thing. Control variables Our tests will take into account a large number of control variables. In the results section we will report on those models in which the control variables turned out to be statistically

16 significant, so that they in that sense can be said to belong in the model in question. In the section on robustness checks we will then list a large number of additional control variables that were tried but that were not statistically significant. When studying the effects of women s social rights one may wonder whether the social status of women is strongly influenced by some other factor that also could be related to the dependent variable, so that any association found between gender equality and violence is spurious. We want to control for such potentially confounding variables. Inglehart and Norris (2003) found that attitudes to gender equality are systematically related to Islam, in such a way that Muslims tend to be less favorable inclined to gender equality. Unless we control for any potential effect of Islam we could thus mistakenly conclude that gender inequality is related to the dependent variable when the true relationship could be between Islam and the dependent variable. For this reason, we will add the percentage of the population who are Muslims as a control variable, and we will do the same also for Catholics, Protestants and Other Denominations. These percentages were coded for 1980 by La Porta et al. (1999), and we got the data via the QoG dataset. Two general socio-economical variables that often turn out to be related to political phenomena having to do with violence in different forms are the total population and the gross domestic product per capita of a country. These two variables are typically log transformed to capture a decreasing marginal effect and are included in many studies. We will also control for log of total population and log of gross domestic product. The data is based on the Penn World Tables 5.6, but uses additional sources and imputation techniques to reduce problems with missing values (Gleditsch 2004, 2002). We will also control for ongoing intrastate armed conflict using data from UCDP, and for regime durability. The latter variable is coded by the Polity project and indicates the number of years since the most recent regime change (defined by a three point change in the polity

17 over a period of three years or less) or the end of a transition period defined by the lack of stable political institutions. We got the data on Regime Durability from the QoG dataset. A number of additional control variables that were not statistically significant are mentioned in the section on robustness and presented in the online appendix accompanying this paper. Research Design Our unit of analysis is the country year. The time period, , is limited by the availability of data on Women s Social Rights and population. In order to test for the hypothesized interplay between democracy and gender equality we follow standard procedures and use multiplicative interaction terms in our multiple regression models. We call the first interaction term Women's Social Rights X Polity, and it is constructed by multiplying Democracy (Polity2) with Women s Social Rights. Our second interaction term is Women's Social Rights X Constraints, and we produce by multiplying Democracy (Constraints) with Women s Social Rights. We use cross-tabulations with percentages as an intuitive way of examining bivariate and trivariate relationships, and logistic regression for multivariate analysis. Many control variables were added to the baseline models shown one at a time, but all of them failed to exceed standard thresholds for statistical significance. These controls are reported in the section on Robustness and presented in the Online Appendix accompanying this paper. To control for time dependence we followed the recommendations in Beck et al. (1998). Accordingly, we used a control variable counting the number of years a country has refrained from frequent torture and an alternative count variable counting the number of years a country had been at peace. These count variables were used together with cubic splines.

18 Results We open the analysis with bivariate cross-tabulations and percentages.

19 Table 1. The overall relationship between Women s Social Rights and Frequent Torture. Frequent Women's Social Rights Torture Total , , Total 632 1, , Pearson chi2(3) = Pr = Table 1 shows a clear bivariate relationship between Women s Social Rights and the resort to Frequent Torture by agents of the state, so that Frequent Torture is much less frequent in countries with higher levels of Women s Social Rights. Next we will show the bivariate relationship between gender equality and torture when the sample of country-years is split into two sub-samples, one with non-democracies (Polity2<=0), and one with more democratic regimes (Polity2<0).

20 Table 2. The Relationship between Women s Social Rights and Frequent Torture among Less Democratic States Frequent Women's Social Rights Torture Total Total , Pearson chi2(3) = Pr = Table 2 above shows that there is no significant relationship between Women s Social Rights and Frequent Torture when the institutional setting is less democratic. This is in line with our expectations.

21 Table 3. The Relationship between Women s Social Rights and Frequent Torture among More Democratic States Frequent Women's Social Rights Torture Total , Total , Pearson chi2(3) = Pr = In Table 3, in contrast, a strong relationship between Women s Social Rights and Frequent Torture is evident when the sub-sample is reduced to the more democratic states. Thus, among the more democratic regimes, states are much less prone to torture when the level of Women s Social Rights in society is higher.

22 Table 4. The Relationship between Women s Social Rights and Conflict Onset among Less Democratic States Onset of Intrastate Armed Women's Social Rights Conflict Total , Total , Pearson chi2(3) = Pr = Table 4 above shows no statistically significant association between gender equality and the risk of conflict onset in the subsample of less democratic states.

23 Table 5. The Relationship between Women s Social Rights and Conflict Onset among More Democratic States Onset of Intrastate Armed Women's Social Rights Conflict Total , Total , Pearson chi2(3) = Pr = Table 5 shows a strong and statistically highly significant bivariate relationship between gender equality and the risk of conflict onset among more democratic states. In this subsample of more democratic regimes, the onset of intrastate armed conflict is much higher when Women s Social Rights are low or very low. Next we turn to multivariate logistic regression so as to be better able to control for many potentially confounding variables.

24 Table 6. Predicting Torture and Conflict with Interaction Terms Frequent Torture Conflict Onset Frequent Torture Lag 2.047** (0.129) Intrastate Armed Conflict 1.024** (0.147) Democracy (Polity2) (0.018) Women's Social Rights * (0.097) (0.390) Women's Social Rights X Polity ** (0.012) Log of population 0.125** (0.026) (0.109) Religion: Catholic 0.004* (0.002) Regime Durability ** (0.003) Women's Social Rights X Constraints * (0.088) Democracy (Constraints) 0.282** (0.101) Log of GDP per capita ** (0.157) Constant ** * (0.460) (2.102) Observations Robust standard errors in parentheses. * significant at 5%; ** significant at 1% [Comment on coefficients and significance in Table 6] As pointed out by Ai and Norton, the interaction effect can be non-zero even if the coefficient of the interaction term (Women s Social Rights X Polity in the first model, and Women s Social Rights X Constraints in the second model) is 0; the statistical significance of the interaction cannot be tested with a simple t-test on the interaction term; and the sign of the interaction term is not necessarily the same as the sign of the interaction effect (2003, 124). To examine the effects we used the simulation approach presented in Brambor et al

25 (Brambor, Clark et al. 2006). In the following graphs the dashed lines represent the 95% confidence bands around the estimated effects. The control variables were set at the following values: Figure 1. Effect of Women's Social Rights As Democracy Changes Improvement in Women's Social Rights from 0 to 1 (full range: 0 to 3) Change in Torture Risk Democracy (Polity2) The graph above shows the decrease in the risk of frequent torture associated with an improvement of women s social rights by one step from zero to one. The confidence bands do not include a zero effect when the level of democracy is higher than zero on the 21-point Polity2 scale that ranges from -10 to 10 (the x-scale). In other words, in the upper half of the democracy range are improved social rights for women associated with a statistically significant reduction in the risk that torture will be widely used. The magnitude of the reduction in risk is very substantial, from a predicted risk among full democracies (Polity=10)

26 of frequent torture of 25% when women s social rights are minimal to a risk of 16% when the rights are slightly better.

27 Figure 2. Effect of Democracy as Women's Social Rights Changes Improvement in Democracy (Polity2) from 0 to 1 (full range: -10 to 10) Change in Torture Risk Women's Social Rights The next graph (above) illustrates the varying effect of democracy given different contexts in terms of women s social rights. The confidence bands around the predicted effect do not include a zero effect when women s social rights are at level one, two, or three. When women s social rights are at the minimum level (i.e., 0), democracies are not more averse to frequent torture. At the other end of the scale, when women s social rights are the highest, even very slight improvement in democracy (one step higher on the 21-point Polity2 scale) is associated with significantly lower torture risk. Again, the magnitude of the risk reduction is dramatic.

28 Figure 3. Effect of Women's Social Rights Given Democracy Improvement in Women's Social Rights from 0 to 1 (full range: 0 to 3) Change in Onset Risk Democracy (Executive Constraints) The effect shown on the y-axis above is very strong in substantial terms. The predicted onset risk given executive parity or subordination (xconst=7) is 4.1 per cent when women s social rights are at the minimum, but 1.7 per cent when the rights are one unit higher. Thus, the onset risk in democracies is halved when women s social rights improve slightly.

29 Figure 4. Effect of Democracy Given Women's Social Rights Improvement From No Constraints to Parity or Subordination Change in Onset Risk Women's Social Rights The graph above (Figure 4) shows that for most societies the level of democracy is inconsequential for the risk of intrastate armed conflict onset. Even more noteworthy is the leftmost part of the graph, which shows that fully constrained regimes are significantly more likely to experience conflict onset than the most unconstrained regimes when the context is one of a society with minimum social rights for women. Robustness

30 This section is not written yet. Situating East Asia In order to assess regional patterns of violent conflict, geographical regions need to be situated in the global context. The global analysis that we have conducted here can thus be utilized in order to shed light on different regions. One such interesting regional pattern is the relative peace of East Asia after How can this peace be explained? And how can we gauge its depth? Situating East Asia within the results of this global study gives us some interesting clues that are worth pursuing in future research. The East Asian region has several examples of states that score high on certain indicators of gender equality, such as representation of women in parliament and workforce participation of women. Also, in many of these states with seemingly more equal relations between men and women, intrastate armed conflicts have been rare. The findings of this paper suggests that caution is necessary when it comes to interpreting this as a sign that gender equality has contributed to the East Asian Peace in this direct way. First of all, in a region where many countries are still not fully democratic, gender equal norms do not stand a good chance of translating into policy, leading to less oppression. When coupling this insight with the knowledge that the gender equality measures of many countries might be of a largely cosmetic nature, such as fertility rates in China or representation numbers in Communist countries, we realize that gender equality values that represent real values of tolerance and the equal value of all men and women in connection with democratic regimes are probably rather rare in East Asia. This indicates that the depth of the East Asian Peace should be questioned. Even if certain types of conflicts are absent, repression and

31 violation of personal integrity is still rather common, and the equality values and democratic institutions are lacking in many countries. We still, however, need more systematic research on actual values of the East Asian population, and we believe that survey research focusing on the issues of violence and gender equality are called for in several countries in East Asia. The data used for this study point to some countries in East Asia in which democratic political institutions together with low respect for women s social rights occur concurrently with torture and onsets of armed conflict. For example, Philippines was for many years a democracy with low social status for women (Women s Social Rights=1 or 0), and frequent torture. Several rebel groups have been actively fighting the government under the same period. Another example from the East Asian region is Indonesia. When Indonesia democratized in the late 1990s, the social status of women remained low (1 or 0), torture was frequent, the Aceh conflict became active again, and East Timor seceded from Indonesia under much bloodshed. In contrast, the two countries in East Asia with the deepest peace would be Mongolia and Taiwan, judging solely from the variables focused in this study. Discussion Our tests have failed to falsify the hypotheses that we derived from our arguments about the role of the interplay of democratic institutions and gender equality in determining the level of violence in a society. In sum, our results suggest that democracy has the potential to reduce torture only when the level of discrimination of women in terms of their social rights is low or very low. In societies with high respect for women s social rights, democracy is on the other hand strongly associated with less torture. Conversely, stronger social rights for women are not associated with aversion to torture when the institutional setting is undemocratic. But among more democratic regimes, there is a strong relationship between more respect for

32 women s social rights and aversion to torture. Among societies with equally democratic institutions are those societies that discriminate less against women in terms of their social rights much less prone to use torture. The findings with regard to the risk for the onset of intrastate armed conflict are parallel, with one important difference. Given a context of somewhat democratic or very democratic political institutions, women s social rights are very strongly associated with peace. In less democratic states, gender equality does not influence the risk of conflict onset. The level of democracy is unrelated to the risk of intrastate armed conflict for most types of societies, namely those with at least a low level of respect for women s social rights. Somewhat surprisingly and very worryingly, more democratic political institutions are associated with a higher risk of armed conflict in the societies with the very lowest level of respect for women s social rights. Our interpretation is that these unfortunate societies are beset by the combustive combination of a political system based on negotiation and mutual tolerance (as indicated by the executive being constrained by and accountable to the legislature), and uncompromising and oppressive norms of militarized masculinity pervading society (as indicated by the total disregard for women s social rights). We believe that the results presented above provide strong support for our argument that democracy needs to be taken into consideration when the relationship between gender equality and violent conflict is to be assessed. Theories on the effects and functions of democratic political institutions must, we submit, integrate these insights on the interplay between open political institutions for channeling political conflict and the societal norms that fill these channels of political participation with content. In particular, this study points to the key importance of norms regarding the appropriateness of violent domination by men over women in influencing societal outcomes having to do with collective violence, such as the prevalence of torture and civil war.

33 Our thesis on the crucial importance of the interplay of democratic political institutions and norms of gender equality in explaining collective violence has the potential to contribute to unlocking several empirical puzzles. One such puzzle is the role, or perhaps rather the absence of the role, of gender equality in explaining how states in East Asia keep internal peace. On some indicators of gender equality, many states in East Asia show very high levels. In many of these states with seemingly more equal relations between men and women intrastate armed conflicts have been rare. Yet, these states tend to be very oppressive and undemocratic. At the same time, these East Asian states tend to disregard women s social rights even if more formal indicators of gender equality, such as the percentage women in parliament, would suggest that women and men are not that unequal in the region. The arguments and findings of this paper sheds new light on how undemocratic political institutions combined with low levels of women s social rights most likely have contributed to the oppressive nature of the internal peace in many of the East Asian states.

34 References

35

36 They also found a linear pacifying effect of democracy when they had controlled for irregular leader entry and other relevant variables. However, tests discussed later in this paper show that irregular leader entry has no statistically significant effect when societal gender equality is taken into account. What is more, the pacifying direct effect of democracy found by Gleditsch and Ruggeri can no longer be discerned. For example, country X has a POLITY score of -7 in 1957, followed by three years of -88 and, finally, a score of +5 in The change (+12) would be prorated over the intervening three years at a rate of per year, so that the converted scores would be as follows: ; ; ; and (Marshall & Jaggers, 2002: 16)See also and Jaggers & Gurr (1995).