This item is only available as the following downloads:
1 by Christian Chessman A thesis submitted in partial fulfillment of the requirements for graduation with honors from the Department of Political Science
2 Theories, Theorizing, and War: An Introduction In 1991, International R elations (IR) theorist Steven Walt accurately characterized war and its causes as the central focus of international relations theory broadly and secu rity studies specifically (Walt 1991) While the scope of both security studies and international relations theory has expanded immensely since Walt wrote about the state of the field, war still retains an undeniable centrality in the study of both. Though the scope, means, and magnitud e of war have changed significantly over the last century war remain s a constant feature both of global politics and the scholarship that attempts to define and understand it While the task of analyzing the effects of war on the individual, state, and international system is important, equally important is the task of understanding underlying causes of war, as well as factors that might contribute to the severity and escalation of war. Mainstream international relations theorizing is replete with diverse approaches to explaining and understanding war (Keohane and Martin 2003, Chapter 3 ) Inherent to any theoretical approach to explaining war are a series of assumptions about wh at information qualifies as relevant data, how that relevant data should be operationalized and represented, and how those operationalizations should be interpreted. The se assumptions shape the ability of those theories to describe, prescribe, and predict behavior for states because they constrain the possible visions of the world in which those states operate (Mitchell, Diehl, McLaughlin 2012, 22 ) 2009, 21). The degree of distortion stemming from the assumptions underlying a theory is magnified in a variety of ways. If such assumptions are the lens through which a theory describes the world, then the inherent
3 limits in data interpretation and acquisition are dirt on that lens. Incomplete datasets, inaccurate data collection, and the difficulty in fully operationalizing complex social phenomena further limit the descriptive potential of a given theoretical approach. Theoretical frameworks thus establish th e parameters of the world which a researcher can see, painting a necessarily incomplete and partial picture Vasquez argues this inherent partiality is not cause to abandon the scientific study of warfare or overarching theoretical frameworks for understan ding international relations and war (Vasquez 2009 20) To the contrary the risk of partiality requires implies the need to conscientiously identify biases and theoretical predispositions, and eliminate or minimize their effect to the fullest extent poss ible As such, inquiry into the assumptions underlying mainstream international relations theory generally and theorization of the causes of war specifically is a necessary precondition to determining both the interactions between and nature of the factors that contribute to war (Dillon and Reid 2000, 133). While mainstream international relations theory has not fully embraced feminist approaches to theorizing conflict, liberal feminist schools of thought are gaining increasing traction in the literature. for conflict ( see generally Caprioli 2001; Hudson et al 2012). This thesis builds on that literature and finds them ultimately unsupported. After making substantive contributions to the empirical methods deployed in analyzing the feminist peace this thesis tests and finds empirical support for the hypothesis that the welfare of women in a given state does not predict an increased or decreased likelihood of conflict. Finally, this thesis will attempt to explain the disparity between the claims o f the feminist peace and the results of the data analysis using insights from other
4 branches of feminist regarding the conflation of sex and gender. It will not attempt to argue that feminist thought has nothing to contribute to mainstream international re lations theory; instead, it will argue that the nature of that contribution is something other than the feminist peace thesis advanced by the liberal feminist school of thought. In doing so, this thesis will attempt to advance the dialogue regarding the re levance and relative importance of sex and gender to conflict theorizing in a manner that is accessible to the mainstream. Sex and Gender in International Relations In 1988, J. Ann Tickner introduced international relations theory to the idea that sex wa s a factor in explaining international relations generally and interstate conflict specifically (Tickner 1988). In the decades that followed the literature examining the interconnection and relationship between international relations of sex, gender dynamics, and war has developed substantially and developed in a manner that can be broadly described as feminist. Despite a growing literature that utilizes and recognizes sex, gender, or both as relevant to the constitution of war, mains tream international relations theory tends to ignore gend er in its analysis (Tickner 1997 ). Laura Sjoberg argues that the lenses suggest a group of causal variables in war decision making that enrich current M ainstream international relations theory tends to ignore the role sex and gender play in for a variety of re asons, not the least of which is unfamiliarity with the conceptual tools utilized by feminist theorists. J. Ann Tickner notes that engagements between mainstream international s and other kinds
5 (Tickner 1997, 629) The gap between ki 2003, 292). Serious feminist engagement with mainstream frameworks for explaining and understanding war must include explanation of the analytical tools it wields and the concepts that are underlie such tools. Sex and gender are two commonly confused concepts that serve as the foundation of most if not all feminist analysis. Judith Stiehm points out that in security studies specifically and internatio nal relations theory generally, theorists talk use the term gender when they mean to talk about sex ( 2 010, 22). Despite consistent interchangeable use conflation of sex and gender is incorrect both in conversation generally and in context of feminist international relations specifically. The ts that result from as well as the chromosomal arrangement itself These theorists see conflict relevant traits as biologically determined or predisposed. For example, such theorists may see women as inherently pea ceful, inherently passive, or inherently empathetic and therefore less likely to resolve disputes with aggression and violence. Feminist theorists who focus on sex theorize about the material wellbeing of biologically female individuals, and the relationship between the welfare of those individuals and international processes. Gender is distinctive from sex in a variety of ways. C rawford and Unger characterize 421 ). To the extent that gender can be described a social construction, it is distinct from sex though
6 categorization is a part of gender (Sjoberg 2013, 5) Sjoberg explains that gender can The element of mutual agreement upon which gender norms rely makes the social, political, and nder are not universal and in fact may even contradict ac ross cultures (Najmabadi 2005, C h apter 2). These rules prescribe not difference, but derived inequalit e[s] a self reinforcing inequality of power both Feminist theorists who foreground gender resultantly avoid centering their analysis on biological women, and instead focus on the way in which gendered norms, expectations, and symbols interact with and constitute politics between states. These theorists attempt to understand the nder as a particular kind of power In that sense, gender theorists recognize that it is possible for and that it is the nature of the conduct not the biological nature of the actor that affects the propensity of people and states to enter and escalate conflicts. The choice to foreground sex or gender in feminist analysis has the capacity to affect the choice of empirical research methods. Theorists whose scholarship focuses on the interrelation between sex and international conflict study phenomena which are we ll suited to operationalization Sex and the attendant measu res of sex equality in society tend to be concrete,
7 discrete, and measurable on large scales For example, common measures of sex equality may include maternal mortality, adolescent fertility, secondary education, shares of parliamentary seats, and labor force partic ipation rates (Caprioli 2000) As a result, scholars who foreground sex are methodologically opened to positivist analysis. In contrast, theorists who focus on gender regularly question the applicability of positivist methods to their analysis and to feminist theoretical frameworks overall In foregrounding gender, these theorists see implicitly agreed upon norms as constitutive of international relations generally and international conflict specifically. Such norms do not lend themselves to easy m easuremen t or operationalization, and are arguably too complex to ever be meaningfully represented by any reduction to a variable (Tickner 2005, 20). The questions and answers which gender theorists see as relevant to international conflict have an inheren t intangibility that resists commensuration and operationalization. Such resistance may also derive from constructivist insights that the concepts in question are relationally constituted (Steans 2003, 434) thus poorly suited to casual, linear, and time b ound analysis. For example some gender theorists see day to day violence as both implicated by and constitutive of gender norms and expectations (Cuomo 1996, 33). This thesis does not need to fully resolve the scope of explanatory ability wielded by pos itivism to perform its analysis The existence of methodological and epistemological pluralism among feminist schools of thought indeed may be understood as a strength, rather than a weakness. Diverse perspectives with differing base assumptions provide mu tually beneficial methodological insights that illuminate otherwise unexposed questions Rather than approaching disagreement as fundamentally deconstructive, this thesis characterizes method differences as a way to expose and avoid potential pitfalls in the conduct of research. It concurs with and
8 draws from John Hoffman various approaches to river with numerous contradictory currents rather than a series of rivers flowing in different and Situated in Sex Studies : A Review of Feminist Conflict Literature If feminist explanatory i nsight into war and conflict is indeed a river, then it is a river with many powerful and overlapping currents. These currents substantively differ on the nature and relevance of materiality, language, race, colonization, and biology. In an attempt to situ ate the contribution of this thesis into ongoing discussion regarding sex, gender, and warfare, it will examine the various prominent schools of international relations thought on the subject. One of the earliest schools of feminist thought on international relations is feminist constructivism. l: international relations are constructed when people talk, follow rules and norms, are guided by world views or institutions, perform rituals, and engage in various social as the result of actively shapes the nature and composition of in stitutions, as well as the distribution and identity roles to whi ch states, subnational units, and individuals conform. When Al Gore accused the 2000 election he implicitly challenged the gendered identity role that George Bush represented as feminine. In a similar vein, t h e second
9 have and thus the world we message of super expression in cases where political leader (Reardon 1993, 30 31). Constructivist feminists interested in warfare argue these gendered ideas manifest themselves i n ways that increase the likelihood of conflict escalation and violence. For example, the gendered assumptions underlying international affairs may provoke states into voke to conflict escalation (Peterson and Runyan 1999, 4). Po ststructuralist feminist international relations bears several elements in common with constructivist feminist international relations, but add several caveats that create stark differences between the two schools of thought. Both agree that states cannot and should not be treated as homogenous, unitary actors; both agree that states are constitutively gendered; both agree that symbolic constructions of masculinity and the embodiment and pursuit of such constructions can lead to conflict escalation in the m anner described above. Poststructuralist feminist international relations theory differs, though, in two critical ways; first, it identifies different mechanisms by which identity and power re lations are created. Second, it argues current configuration s of gendered power structures are impermanent, contextual, and transient.
10 With respect to the latter, poststructuralists question the assumption that existing gender roles are natural, permanent, or universal. Laura Shepherd argues that unlike constructivist to be a transhistorical or universal system of identity production, nor is it assumed that individuals experience gender in the same way, even within a particular social/politi I Poststructuralist feminists also contest the mechanism by which gendered power relations are created. Constructivists either fail to provide a mechanism by which power operates (Locher (Shepherd 2007, 246). Poststructuralists argue instead that constitution of masculinity and femininity especially as they relate to conflict and warfare 07, 247). These configurations are the result of the iterated conduct of individuals and states, which both reinforces and is reinforced by mutually shared gendered expectations and roles. States and individuals shape a particular gender role as they act it out, while simultaneously give legitimacy to it by conforming to it. The relationship between gendered behavior and gendered expectations is thus bidirectional. In that sense, poststructural (Connell, cited in Sjoberg 2013, 5). Poststructural insights about power, history, and identity are extended to particular cultural locations by postcolonial feminists. Postcolonial feminists draw from the notion that gender relations, roles, and power manifest differently in different contexts to account for the
11 effect that colonization has on gender They add depth to poststructural theories of war by examining the gendered role that imperialism, imperialist power hierarchi es, and imperialist political economic organizations play in contributing to international conflict (Mohanty 1988, 64 65 ). Each aforementioned school of thought attempt to explain international conflict and war making by foregrou nding gender analysis Lib eral feminist conflict theorists stand in stark contrast to these schools of thought by foregrounding biological sex as the explanation for conflict. These theorists examine indicators of security and welfare for biological women and argue that these indicators influence the likelihood that a given state will be prone to aggression and violent conflict ( e.g. Caprioli and Boyer 2001; Caprioli 2005; Hudson et al 2012 C hapter 5 ). Liberal feminists attempt to explain a connection between sex and conflict through a variety of theoretical approaches ( Agacinski 2001, 14 ) The modeling school of thought argues that children are exposed to several fundamental types of difference from a young age, the most significant of which is sex. To explain why sex is such an important axis of difference, l iberal feminists often turn to evolutionary psychology They argue that sex difference is the foundation of other form impressionable young children see sex inequality as the mistreatment of others and export that learned behavior to other situations (Hudson et al 2012 251 ). Liberal feminists argue that when states mistreat women specifically the citizenry of the state learn s that it is appropriate to mistreat others generally and becomes more prone to aggressive responses overall.
12 Other liberal feminists take evolutionary psychology a step further, and argue that women have an innate biological opposition to conflict Hunt argues that indicators of female political participation are predictors for state levels of aggression because female leaders are innately less prone to conflict (2007). These feminists argue women have a natural femininity that makes them prefer peaceful solutions to posturing and violence o women thus make for better leadership (Hunt 2007, 109). McDermott and Cowden (2001) develop this concept and argue that women have inherent maternal instincts that tend towards protective and nurturing behaviors in place of aggressive and violent behavi ors. McDermott and Cowden (2001) conduct ed a psychological experiment on one hundred male and female participants. In each case, they created a male female dyad and simulated international crisis situations. The results indicated that men were substantial ly more likely to be aggressive and conflict prone, prioritizing expansion and accumulation of weapons. In contrast, women tended to be more interested in diplomacy and peaceful resolution of tensions. The authors use these findings to conclude that inter national conflict would decrease in the real world much as it did in the simulation, were women leaders put in positions of authority (McDermott and Cowden 2001). Liberal feminists ultimately conclude that the mere presence of women in political leadership positions specifically and the increased societal propensity fo r conflict (Hudson et al 2012, 102 103). Several strands of liberal feminist thought take the claim of innate biological female regime type. Hudson et al make the unsubstantiated claim that
13 the greater attention given to social welfare, legal protection, and transparency in government 2012, 100) They take these indicators as evidence of rule of law and flourishing civil society, and argue that reatment of women, then, may affect societal propensity to adopt a Hudson et al 2012, 100) To the extent regime type can function as a predictor of propensity for monadic conflict liberal feminists argue that sex equality is a predictor for conflict. Other branches of liberal feminist thought maintain that state level indicators of welfare for women are relevant to conflict because they serve as a proxy for economic growth and de velopment. Esther Boserup (199 7) grounded this position on a survey of thirty three countries, where she examined the percentage of women among agricultural hired labor and self owned farms. She found that when women were segregated from mainstream farming enterprises and Because women were limited to the margins of agricultural innovation, they were less likely to be exposed to new methods of cultivation and innovation. As a res ult, economic output from agriculture primarily operated by women lags behind overall innovative development, producing suboptimal economic results. Boserup argues this effect is most profoundly pronounced in developing nations, where economics are primari ly driven by agricultural productivity. Amartya Sen (1989) expands this argument beyond developing states using the concept of economic specialization. She argues that the sexual division of labor in sexually unequal societies is not an efficient allocati on of labor, leaving much labor either unutilized or underutilized. This occurs because societies are often faced with a choice of placing women in equal economic roles, or maintaining a sex hierarchy. In sexually unequal society, women are placed in roles that do not utilize their full potential to contribute economically. Using an
14 interregional study of Asia and Africa, Sen divides the territory into sub regions and examines (1989) She found that countries whose ratio of males to females in gainful employment was closest to their male to female population ratio were more economically successful as a measure of GDP and per capita income. 1 According to Sen, t here is a clear empirical relationship demonstrating that countries are more economically successful when they allow women to be full participants in the labor force (Sen 1989). Andrew Mason and Elizabeth King (2001) confirm this economic prediction in perhaps the most thorough examination of the linkage between sex equality and economic development. Using 25 country years for 63 countries and encompassing OECD and developing stat es, they malnutrition, more illness, and more deprivations of other k Mason and King identify several connections between sex equality and economic productivity, including maternal health and welfare during pregnancy. A dearth of natal care l eads to higher infant mortality and increases the l ikelihood that children who do survive will do so with severe physical or mental disabilities, reducing the ability of those children to contr ibute to economic increases the likelihood of trade interdependence and itself leads to higher levels of prosperity (2012, 98). Mary Caprioli (2000) subjects the liberal feminist school of thought to empirical analysis and finds preliminary support for it She specifically argue s that sex equality is a positive 1 Sen excepts China because its female infanticide policy skews the data
15 indicator of the likelihood that a state will be conflict prone. Caprioli expanded the scope of this project in a work she co wrote with Valerie Hudson Bonnie Baliff S panvill, and Chad Emmett (2012). Faced with a dearth of sex equality data, Hudson, Baliff Spanvill, Emmett and Caprioli created and utilized the WomenStats Database (www.womanstats.org), with multidimensional indicators for sex equality and welfare. This dataset is one of the most comprehensive repositories of sex equality data and includes data on over 175 states (Hudson et al 2012). To the Global Peace Index or GPI. Using the GPI as their conflict data and the WomenStats data as their sex equality data Hudson et al. argue that the physical security and material equality of women will positively predict lower levels of state aggression while lower levels of physical security for women will predict higher levels of state a ggression (Hudson et al 2012; 2 09). Hudson et al (2012) analyze this data using bivariate regressions and controlling for democracy. They conclude that there is a rob ust predictive relationship between higher levels of sexual equality and security, and lower levels of aggression. There are a variety of statistical and theoretical issues underlying the liberal feminist explanation for the relationship between sex equality and conflict. Several liberal schools of thought including the modeling school of thought and the economic sch ool of thought do not account for the necessary time lag between a change in sex equality and a change in propensity for conflict. For example, if the level of sex equality changes in the year 2001 2002, the modeling school does not predict the state wil l be less violent in 2002; it predicts the state will be less violent when the cohort of children who grew up in 2002 become adults and take political office. There is a temporal discrepancy between when the modeling school predicts sex equality would chan ge people, and when people would begin to change the state. Similar temporal
16 discrepancies exist for the economic school; as measures of social and political equality increase, the sexual division of labor does not automatically decrease. Because many of the indicators of sex equality are only partially, tangentially, or indirectly related to the economic participation of women, it is unlikely that change in the former would immediately or quickly result in change in the latter. There are similar methodol ogical questions for the liberal feminists that see an inherent nature to women. This position called biologically essentialism reduces the conduct of women to their genetic conduct. This reduction is both intellectually specious and inconsistent with the empirical record surrounding the conduct of women (Sjoberg and Gentry 2007). There simultaneously being inaccessible to men, by nature of biology. There are plenty of men who are maternal and plenty of women who are not maternal. The presumption that the mere presence of a woman in governance will decrease the likelihood that a state is conflict prone relies on inaccurate generalizations (Sjoberg and Gentry 2007). A f urther criticism comes from p ost structural feminists who question all three schools of thought as simplis tic to the point of inaccuracy. Each school of thought removes sex inequality from the contexts in which it occurs and attempts to make universal gen eralizations about its nature Such universal generalizations necessarily lack explanatory power because they are incomplete. For example, t he way in which children interpret sex inequality as well as what counts as sex inequality occurs vary across cu lture, time, and space. These variations alter the ways in which sex inequality connects to and intersects the broader social experience of individuals in any given culture. The same holds true for the ways in which women are socialized to behave and cond uct themselves; different cultures prioritize different types of
17 femininity at different times, which implies that a universal femininity shared by all women cannot be accurate. Even the economic school of thought relies on context insofar as the interplay between the indicators of economic growth and sex equality varies in different social and political contexts. Liberal feminist schools of thought rely on the assumption that the operationalized measures of sex inequality and sex equality have the same cu ltural, social, and political effects in all societies for which they attempt to explain conflict. To the extent that this assumption is faulty and does not accurately reflect social and political realities across global settings for conflict, the data sho uld be inconsistent with it. The Challenge of Context: A Research Question This thesis challenges the empirical, theoretical, and normative support for the liberal feminist explanation of conflict. It does so in two ways; first, it does so by examining and questioning the underlying theoretical frameworks advanced by theorists who see a connection between sex and conflict. Second, this thesis asserts that the best empirical data do not support a predictive connection between sex and conflict, e ither in initiation or severity. It ultimately of military aggression. To do so, this thesis builds on the literature in a number of ways. First, it brid ges a theoretical gap between poststructuralist feminist arguments and liberal feminist theoretical claims This thesis examines the universal sociological and psychological claims made by liberal feminists through the lens of poststructuralist insights re garding the significance of culture and historical context. It also exposes several other theoretical blind spots in the liberal feminist
18 choice to foreground sex as the determinative factor of conflict. Such explanations approach gender gender roles, and gendered expectations b e read unproblematically 246) rather than a s social construct s that affect the expectations of both men and women. Maintaining Methods and Determining Data : A New Approach The most recent and most compelling case for the liberal feminist approach to conflict is made in Sex and World Peace by Hudson et al (2012). This thesis relies in part on the methods utilized by Hudson et al, but develops them in a variety of ways. It expands both the quality and quantity of data used in answering the question posed by the feminist peace. It will perform the first longitudinal study of empirical relationship between sex and world peace in the field, and examine the ways in which temp oral variance might affect the validity of previous studies. It will also utilize a different conflict dataset than other liberal feminist theorists (e.g. Caprioli and Boyer 2001; Hudson 2010; Hudson et al 2012). Rather than following these feminists in us ing the General Peace Index, this thesis will utilize the mainstream Militarized Interstate Dispute (MIDs) database. Additionally, it ide ntifies a severe endogeneity problem that Sex and World Peace does not correct. Finally though this thesis also uses b ivariate ordinary least squares regressions for its data analysis it does so using longitudinal data for sex equality and the leading indicators of both sex equality and state aggression in their respective fields. T he Sex and World Peace authors utilize data from a single year 2006 for a total of 105 data points (Hudson et al 2012). They do not test their hypothesis over time, which presents a number of methodological questions. First, the lack of temporal variance creates th e possibility that 2006 w as an exceptional outlier year for conflict. Without the ability to examine a trend over time, Hudson et al cannot rule out the possibility that their conclusions are nothing more than
19 coincidental happenstance. Second, the use of a single year leaves op en the possibility of a year specific spurious cause. The lack of temporal variance means that any spurious cause that existed for the duration of the year would be necessarily undetected by the data. Indeed, the of longitudinal analysis by changing the unit of analysis from countries to country years, and analyzing data from 1970 to 2002. The data used by Hudson et al also exclude approximately 40% of the world from their analysis without explanation (Hudson et al 2012, 215). The N for their bivariate regressions varies between 105 and 140, and the United Nations cu rrently recognizes 193 member states (United Nations 2013). This thesis uses all countries for whom data is available in all years, further developing the analysis performed by Hudson et al. There are further issues with the data analysis conducted by Sex and World Peace Hudson et al utilize the Global Peace Index (GPI) for their conflict data. The majority of theorists who discuss do so using the Correlates of War database or the Militarized Interstate Disputes (MID) database (Braithewaite 2010). The dis parity between datasets makes it difficult for the results found by Hudson et al to interact with the mainstream, because they are founded measures that are incons istent with the measures chosen by the MIDs database, including but not limited to; number of domestic refugees, level of violent crime, number of jailed persons, imports and exports of conventional weaponry, and ease of access to small arms (Global Peacef ul Index 2013, 87). None of these measures are included in the MIDs database as indicators or measures of conflict. As a result, the results found by Hudson et al are not generalizeable to the field even if they are statistically sound. This thesis prefer s the MIDs
20 database not only because the majority of the field uses it, but because it contains more effective operationalizations of the concepts in question. The Global Peacefulness Index is a measure of for conflict (Global Peaceful Index 2013, 4). The difference between positive peace and warfare is well documented (eg, Cuomo 1996) and makes the GPI inapt for measuring the way in which sex equality affec ts propensity for warfare. Finally, the GPI has not been subjected to the same empirical tests of reliability that scholars have performed on the MID database (Braithewaite 2010). As a result, this thesis utilizes the MID database for its conflict data. For sex equality, this thesis relies on five separate indicators of sex equality. First, this thesis used two aggregate indicators of gender equality collected by the United Nations, which are widely used in field (Oxaal and Baden 1997). Specifically, it utilizes the data from t he Gender Development Index (GDI) as well as the Gender Empowerment Measure (GEM) GDI is a composite indicator of the differences for men and women on life expectancy at birth, adult literacy rate, combined gross school enrolment, and male to female ratio of estimated earned income. GEM is a composite indicator of sex specific inequality issues, like years since suffrage and eligibility for public office shares of parliamentary seats, and labor force participation rates and income ratios I also created a composite variable of the GDI and GEM based on standard deviations from the mean of each, respectively, preferring the GDI where both exist (operationalized as GENEQUALALL). Finally, I utilized the Cingranelli Richards (CIRI) meas ure s of political equality for women (operationalized as WOPOL ) and social equality for women (operationalized as WOSOC) Though focused on measures of human rights generally, the CIRI dataset also includes data l egal protection and equal treatment This measure has
21 also gained widespread acceptance in the field as an accurate measure of human rights generally Richards 2010, 402). I utilize the combination of datasets for several reasons. First, as detailed above, both all three enjoy wide acceptance in the field as accurate measures of sex equality and inequality. Second, the combination of datasets is necessa ry because of the sheer dearth of data on sex equality. Varying levels of political interest in sex equality have led to inconsistent funding and inconsistent data collection that means any single dataset falls short of thoroughly representing sex equa lity. Third, the indicators of sex equality utilized by each dataset have sufficient overlap to justify their mutual use. All all identify variables representi all identify variables representing Given the overall limit on available sex data, the overlap between indicator purpose and measure is sufficient for data analysis, albeit not ideal. F rom the available gender equality data, I created a dataset with 1,497 data points that utilized the country year as a unit of analysis including a total of 141 countries This allows a direct comparison between the country year level of gender equality a conflict during that year. For conflict data, I use the Militarized Interstate Dispute (MID) database indicator on states which originate conflict. 2 A MID is a militarized interstate dispute, which could be as serious as an in terstate war (a 21 on the scale) and as mild as a threat to use force (a 1 on the scale). I utilized three variables in measuring conflict; first, I coded a variable that summed the number of MIDs a country had initiated during the given country year (MID INIT). This variable ranged from a base of zero MIDs initiated to a theoretically unlimited ceiling of total possible MIDs initiated. Second, I utilized the MID variable that described the 2 Faten Ghosn and Glenn Palmer, Codebook for the Militarized Interstate Dispute Data, Version 3.0 (http://cow2.la.psu.edu).
22 severity of the conflict (MIDHOSTILE). This variable ranged from 0 (no MID initiated) to 5 (full scale war). Finally, I created a composite indicator of both the severity and frequency of MIDs (AGGRESS1). I added a series of controls to contribute to the strength of the research. I controlled for several dominant explanations for conflict in the literature. Though the literature on democracy and economic growth only confirm that they hold in dyadic analyses (e.g. Russet 1993), I have controlled for both to prevent the possibility of an unconfirmed trend from interfering with the data. To control for democracy, I used the Polity IV dataset, which is most commonly used by researchers on democracy and peace to measure regime type (Russett 1993) It is a 20 point scale which describes state regime type each year since either the state inception or 1800. Since the scale typically varies from 10 to 10, I rescaled the values to a range of 0 to 20. I used the Polity IV score for each relevant country year in the dataset I compiled (variable 2.4 in the dataset). It is a composite measure which includes competitiveness of executive recruitment, openness of executive recruitment, constraint on the chief executive, and c ompetitiveness of political participation. I also accounted for economic growth by controlling for country GDP as reported by the United Nations development reports. For dependent variables whose units are integers (MIDINIT and AGGRESS1), I utilized bivariate ordinary least squares regressions that controlled for heteroscedasticity to evaluate the existence of a significant correlation between sex equality and levels of state aggression. I also ran multivariate ordinary least squares regressions, cont rolling for regime type gross domestic product, the end of the Cold War and heteroscedastic errors The bivariate correlation demonstrated whether or not a relationship between the two variables exists and the multivariate regression established that any demonstrated correlation was not spurious The control for
23 heteroscedasticity corrects for error variance that is not distributed normally and might otherwise jeopardize the regressions. F or the ordinal dependent variable (MIDHOSTILE), I ran multivariate ordered logistic regressions which controlled for regime type, gross domestic product, the end of the Cold War, and heteroscedastic errors. Because MIDHOSTILE is an ordinal variable, the most appropriate regression to test the relationship between sex equ ality and this measure of state aggression is the multivariate ordered logistic regression These methods attempt to evaluate whether there is a clear relationship between sex equality and state aggression, using a dataset with almost 1,400 country years of gender, dispute initiation, and democracy data points over a 32 year period. Analyzing the Data The majority of models generated by the regressions were not statistically significant. There was no statistically significant relationship between level of hostility (MIDHOSTILE) and sex equality. There was also no statistically signif icant relationship between both level of hostility and number of initiated MIDs (MIDINIT) and sex equality. [table 1] There were three statistically significant models where where sex equality and one measure of state aggression were negatively correlat ed initially appearing to suggest that states with higher levels of sex equality make less confl ict. However, a closer examination of those three models suggests that sex equality has miniscule eff ects at most [table 2] Several elements are worth emphasizing. First, t he extremely low R 2 indicates that all three models explain less than a single percent of the variance in the MID initiation data. This
24 implies that any relationship between sex equality and MID initiation likely explains very little v ariance in the propensity of a state to initiate a MID. Second, the relationship itself is not very strong because the coefficients are lightly positive at best In an ordinary least squares regression, the coefficient indicates the unit increase in the de pendent variable ( in this case, conflict) for each integer unit increase in the independent variable ( in this case, the particular metric of sex equality). WOPOL has a four unit range (varying from zero to four), which means that the maximum possible chang e it can make to the dependant variable is four multiplied by .068, or .272 units. This model predicts that a society in which every indicator measured by WOPOL was fully present would only initiate .272 fewer MIDs. At best, WOPOL is not an effective predi ctor for conflict initiation. Similar constraints exist on the contributions by GDI and GENEQUALALL where the maximum change in MID initiation predicted is .359 and .39 fewer MID initiations, respectively. In short, even if society had full sex equality in every indicator of every variable for equality, the world would experience an identical number of MIDs as it does now. These bivariate regressions show that the relationship that Hudson et al (2012) find between sex equality and peace in Sex and World Peace does not hold over time, more data points, or more reliable measures of either sex equality or state aggression. The relationship between sex equality and MID initiation had little explanatory power, an d there was no relationship between sex equality and level of hostility. At worst, the empirical support for the relationship between sex and conflict is nowhere near as robust as Hudson et al assert. At best, sex equality is not a meaningful predictor of state aggression.
25 Conclusions about Conflict: Gender, not Sex? This thesis has examined the empirical claims made by liberal feminists regarding the interconnection of sex equality and state aggression. It builds on the liberal feminist conflict literature by increasing the quality of data utilized, increasing the amount of time analyzed, and dev eloping the underlying theoretical framework for predicting conflict. After doing so, it finds that the data does not support the assertion of a robust interconnection between sex equality and state aggression. The interconnection between sex equality and MID initiation had a low explanatory power less than a single percen tage and such low coefficients that even if all measures of sex equality were fully present, the models predict the same number of initiated MIDs. There was no data supporting a correlation between sex equality and level of MID hostility. In making these conclusions this thesis hopes to build on the literature in two ways. First, it tests the feminist peace theory using the best indicators of sex equality and state aggression Hudson et al (2012; 2009) in Sex and World Peace and related work use unconven tional indicators of peace and unconventional datasets, which prevent generalization to other literature surrounding conflict. contextualizes the argument to the conventional indic ators of military conflict and aggression. Second, this paper builds on the literature by adding temporal variance. Hudson et al only examine country years of data spanning from 1970 to 2002 The longitu dinal analysis conducted by this thesis gives it a more comprehensive view and helps to control for effects from this current decade that may otherwise confound or provide a spurious relationship. The result of this comprehensive analysis was to
26 question a nd ultimately discredit the liberal feminist claim sex equality decreased the likelihood of conflict While this work has intervened in the literature in a meaningful way, it is not without threats to its own internal and external validity. Given more ti me and resources, I would like to be able to control for more variables that the literature holds to be important, including but not limited to type of financial system (Gartze 2007), trade interdependence (Oneal and Russett 2001), and a claimed general tr end of decrease of war (Pinker 2011; Goldstein 2011). While the data used in this paper significantly expands on both the quantity and quality of data used in the Sex and World Peace book, it could include more country years, utilizing and converting more gender equality data, including other years of the GDI and GEM as well as other measures, like the World Economic Forum Gender Gap Index. An expansion of this paper may also recast the unit of analysis as dyads rather than individual country years to bes t comport with the majority of conflict theorizing in the field. There is a larger statistical concern that sex equality and conflict are endogenous. As wars progress, they tend to decrease sex equality in a variety of ways. Sjoberg (2013) describes the ways in which wars disproportionately leave women economically, socially, and politically disadvantaged in society (Chapter 9). In addition to being inordinately targeted for gendered war tactics like rape, women are also at a series of economic and social disadvantages. Sjoberg documents how women are empirically unemployed first and at a faster rate than men during wartime; that healthcare for women suffers first, with a particular drop in prenatal care; that destruction of agricultural land has a dispara te impact on women because they are more likely to rely on subsistence agriculture to eat; and that women are more commonly forced into illicit economic tactics such as prostitution for the purpose of supporting their families (Sjoberg 2013,
27 Chapter 9). Th e analysis performed by Sjoberg indicates that the relationship between sex equality and conflict is bidirectional, and suggests that the variables are co constitutive. One way to further develop this research project in the future would be to time lag th e data to test for the endogeneity problem. This allows for a scholar to test whether or not the variables manifest as endogenous and, if they do, how they correlate to each other once that endogeneity is corrected. While the evidence presented in this thesis does suggest that the feminist peace is not supported empirically, it does not suggest that feminist though has no contributions to make to the field of international relations. It also does not suggest condemning efforts to engage, include, and assist women in the political, social, and economic spheres. Rather, it suggests that the choice to foreground sex as the explanation for international conflict is methodologically inappropriate. Gender roles, expectations, and norm s may still play a role in the construction and conduct of international relations. Because gender deals with the conduct of a particular agent rather than the biology of that agent, its relevance is not negated by this scholarship. Instead, a focus on gen der as explanatory of international conflict can lead to insights about the ways in which norms and masculine expectations can cause and contribute to war. This focus on gender may perhaps provide more empirical insight into the nature, duration, and freq uency of conflict between states.
28 [Table 1 ] Regression Model Significance Dependent Variable Independent Variable Result A G GRESS1 GDI Model not significant. AG G RESS1 GEM Model not significant. A G GRESS1 GENEQUALALL Model not significant. A G GRESS1 WOPOL Model not significant. A GG RESS1 WOSOC Model not significant. MIDINIT GDI Low explanatory value. MIDINIT GEM Model not significant. MIDINIT GENEQUALALL Low explanatory value. MIDINIT WOPOL Low explanatory value. MIDINIT WOSOC Model not significant. MIDHOSTILE GDI Model not significant. MIDHOSTILE GEM Model not significant. MIDHOSTILE GENEQUALALL Model not significant. MIDHOSTILE WOPOL Model not significant. MIDHOSTILE WOSOC Model not significant. [Table 2] Significant Bivariate Regressions between MID Initiation and Sex Equality Measure of State Aggression Measure of Sex Equality Number of Cases Model Significance Model Coefficient Model R 2 MIDINIT GDI 1372 .0004 .359 .0081 MIDINIT GENEQUALALL 1497 .0016 .065 .0062 MIDINIT WOPOL 1300 .0353 .068 .0022