Civil War and Female Empowerment

Recent research has directed attention to the transformative potential of war for female empowerment. As a disruptive shock, armed conflict can create a window of opportunity for advancing the societal role of women. We complement this research agenda by looking at how conflict severity and termination condition the outcomes for women in the aftermath of civil conflict. We expect that both level of violence and mode of resolution affect subsequent female empowerment, where severe conflicts ending by a negotiated settlement have the greatest transformative potential. Consistent with expectations, we find that post-conflict improvements in female empowerment occur primarily after high-intensity civil conflicts. However, subsequent tests reveal that this effect is driven largely by conflicts terminated by peace agreements. The greatest improvement in female empowerment is seen when peace agreements have gender-specific provisions. These results support calls for a sustained effort toward mainstreaming gender issues in conflict resolution and peacebuilding processes.


Introduction
Women's status and political influence worldwide has improved substantially over the past decades, both in relative and absolute terms. For instance, progress is observed in girls' educational attainment, with gender parity in primary education being achieved in most countries (United Nations 2015). Further, women's political influence has been fast-growing; women's average share of parliamentary membership nearly doubled between 1995(Inter-Parliamentary Union 2015. Yet, gender gaps are still persistent in most countries (UN Women 2016; World Economic Forum 2017). 1 These gaps are particularly extensive in non-democratic countries with low socio-economic development. Unsurprisingly, some of the worstperforming countries in terms of gender gaps in education, work force, and political participation are countries with durable civil war and fragile governance, such as Yemen, Chad, Syria, and Mali (World Economic Forum 2017). As Melander (2016, 197) puts it, "The strongest pattern in civil war is probably its gendered nature." Low female empowerment cannot by explained simply as an adverse outcome of armed conflict, however; incumbent regimes in many states willfully ignore demands for democratic reform or social justice, including calls for closing the gender gap. At the same time, the ending of destructive civil wars may create a potential for rapid and transformational normative and political change. Indeed, some of the most encouraging improvements in women's access to parliamentary power have been seen in post-conflict countries. Post-war countries like Rwanda, Burundi and Uganda are all above the global average in closing the gender gap (World Economic Forum 2017).
This pattern can be detected not only in unique, well-selected cases; recent comparative research also shows how, somewhat paradoxically, the destructive effects of war may open a window of opportunity for women to strengthen their position in society (Anderson 2016;Anderson and Swiss 2014;Hughes 2009;Hughes and Tripp 2015;Mageza-Barthel 2015;. In one of the first cross-national studies putting this to test, Hughes and Tripp (2015) find that the ending of civil conflict enables higher growth in women's legislative representation in Africa. More recently, Webster, Chen, and Beardsley's (2019) comprehensive global analysis of conflict and female empowerment provide compelling empirical evidence that war can disrupt societies and foster transformational changes in women's role in society, at least in the short to medium term.
There are three ways in which civil conflict may stir up gender roles. First, war destructs the gendered divisions of labor and opens up space for women to enter traditionally male-dominated areas and professions. Second, war fosters women's mobilization and pro-social behavior, creating a push for continued mobilization. Third, women's entry into new jobs and the public sphere creates normative changes in attitudes toward women. In total, these processes are argued to contribute to increased political empowerment for women. 2 We seek to advance the literature on civil war and female empowerment further by presenting theoretical and empirical innovations along three dimensions. First, much of preceding work centers on the consequences of major civil (and interstate) wars under the, often tacit, assumption that only large wars hold the potential to systematically change gender roles and attitudes. In contrast, we evaluate the sensitivity of women's standing to different levels of civil conflict severity. Second, while existing theory predicts conflict-driven impacts on gender roles and female mobilization to influence subsequent female empowerment, we argue that the nature of conflict resolution also plays a central role. Negotiated peace agreements, especially those that explicitly address women's role in society, are expected to facilitate greater improvements in formal gender equity than conflicts that fade away or end by military victory to one side. Third, empowerment is a multifaceted phenomenon, and the causal processes outlined above may have greater influence on some dimensions than other. In this study, we focus on two complementary aspects of female empowerment: individual civil liberties and political participation.
To test the empirical merit of the assumptions about gendered implications of civil war, we conduct a statistical analysis of changes in female empowerment across 160 independent countries for the period 1975 to 2017. We find that countries escaping major civil war experience substantial improvements in women's civil liberties and political participation the initial post-war phase whereas no effect is detected for minor conflicts. However, the lethality of civil war also affects its mode of termination; many of the most severe conflicts end by a negotiated settlement. In subsequent tests, we find that much of the post-war improvement of female empowerment attributed to conflict severity in fact is driven by negotiated settlements. Peace agreements containing gender-specific provisions have a particularly large effect on improvements in women's political participation. The latter is powerful evidence in support of UNSCR 1325 and underscores the importance of involving women in negotiations and settlements in order to create a more egalitarian society.
The paper is structured as follows. First, we outline various ways in which civil conflict can increase female empowerment and develop a theoretical argument for how the severity and ending of the conflict shape this relationship. We then present the data material and research strategy, followed by a documentation and interpretation of the empirical findings. We end by considering some implications of these results for future research and policy.

Previous Research and Theory
In existing research on gender inequality and conflict, a near-consensus finding is that gender inequality is associated with increased risk of civil war (e.g. Dahlum and Wig 2020). Emerging evidence also suggests that armed conflicts and wars, despite their many devastating impacts on society, can facilitate female empowerment (Webster, Chen, and Beardsley 2019;Hughes and Tripp 2015). The main argument sustained through this research is that war constitutes a shock that shatters both institutions and society at large, creating a window of opportunity for women to advance their formal rights and privileges. In the following, we briefly discuss common theories about how armed conflicts can reshape societal gender roles, from which we develop a set of arguments and associated testable expectations relating to how the severity of civil conflict as well as the manner in which it ends inform the potential for subsequent improvements in women's role in society. We start by clarifying what we understand as women's empowerment and the scope of the paper.
Female empowerment is a wide notion encompassing several aspects of a woman's authority (see, for example, Narayan 2005), but for the purpose of this study we focus on women's political empowerment. 3 We understand political empowerment as a multi-dimensional concept that covers inclusion and agency both at the individual level as well as at the level of formal political institutions. At the individual level, female empowerment includes women's right to discuss freely, engage in civil society organizations (CSO), and have access to a fair judicial system. At the institutional level, women's empowerment concerns their formal participation and inclusion in decision-making institutions, such as national parliaments. Our understanding is in line with Sundström et al. (2017). 4 We focus on civil conflict. 5 Although some of the arguments outlined below also apply to interstate war, we maintain that the defining character of internal conflict, pitting incumbent regimes against non-state challengers over incompatible political objectives, reflects a unique potential to reshape fundamental aspects of the society's polity.

Civil War as a Critical Juncture
Armed conflict is sometimes described as development in reverse (Collier et al. 2003;Gates et al. 2012). Aside from human and material losses in fighting, severe conflict often deters financial investments and long-term planning, ruins the tourism industry and other economic activities, causes massive human displacement, and erodes norms and the social fabric conducive to a prospering society. Some studies suggest that war has especially damaging consequences for women. For example, it has been shown that more women than men die in the aftermath of conflict (Ormhaug, Meier, and Hernes 2009), that women's life expectancy is reduced more than men's (Ghobarah, Huth, and Russett 2003;Plümper and Neumayer 2006), that more women than men are forcibly displaced as refugees (Buvinic et al. 2013), and that maternal mortality rates are elevated in the aftermath of conflict (Urdal and Che 2013). In addition, sexual violence is directed (mostly) at women as a weapon of war (Cohen 2013;Cohen and Nordås 2014). Yet, the ways in which conflicts cause severe material and social damages may, paradoxically, serve as mechanisms that can facilitate increased space, opportunities, and privileges for women when the gun smoke clears.
Within research on female empowerment, women's advancement is often portrayed as an incremental process (Blumberg 1984). The rise of women in politics and other traditionally male-dominated spheres of society is dependent on a gradual development where women expand their human capital through increased education and participation in the labor market. On the other hand, it is argued that in some cases, women may take a sudden and big leap forward, helped by some dramatic event or crisis (Hughes and Tripp 2015). The notion of critical juncture, a concept first developed by Rokkan and Lipset (1967) and further developed to explain regime dynamics (Collier and Collier 1991), explains how a crisis or cleavage can create opportunities for political entrepreneurs to change the political status quo. Correspondingly, policy windows, a notion developed by Kingdon (1995), is an opening, commonly caused by crisis, where actors can push forward for a wanted development. Moreover, theory on social movements emphasize how mobilization is dependent on, inter alia, the opportunity structure, in which agents for change exploit political openings caused by landmark events (Tilly, McAdam, and Tarrow 2001). While neither of these concepts was developed with a gender focus in mind, they are useful also for theorizing how women may exploit political openings and the power vacuum that sometimes appear in the wake of civil war.
From the extant literature we can identify three central mechanisms through which armed conflict facilitates positive change in women's empowerment. First, civil war may create a demand for women to take on new positions in society. Such new-earned responsibilities can extend from becoming head of the household and managing the family's economy via entry into traditionally male-dominated jobs to joining men in combat on the battlefield (Kaufman and Williams 2010;Ní Aoláin, Haynes, and Cahn 2011;Thomas and Bond 2015). Moreover, civil war may facilitate increased female participation in political decision-making. For instance, psychological experiments find that people perceive women to be better suited to lead in times of crisis (Bruckmüller and Branscombe 2010). The demand for women to engage politically may be helped further by the framing of women as the more "peaceful" gender (Rudman, Goodwin, and Dovidio 2004;Wood and Ramirez 2018), who have a "natural" aversion against violence (Caprioli and Boyer 2001;Hudson et al. 2012), and who are less corrupt than men. 6 A second pathway through which civil war may influence women's empowerment relates to female mobilization and increased pro-social behavior. Several studies document how, during war, many women enter the public realm through setting up grassroot projects, organizing peace movements, and mobilizing in civil society organizations, all of which can be crucial for advancing women's rights and participation in a post-war context (Anderson 2016;Merrill 2017;Viterna and Fallon 2008;Waylen 2007). In parallel, conflict-related sexual violence can strengthen the social cohesion of a community, as well as trigger a collective solidarity response (Gilligan, Pasquale, and Samii 2014;Voors et al. 2012;Berry 2015;Kreft 2019). For example, evidence from Sierra Leone shows that families of victims to sexual violence are more likely to engage in community organizations and social events (Koos 2018). Unlike the first mechanism's emphasis on "pragmatic" effects related to male vacuum and mistrust, the mobilization mechanism speaks of female agency and deliberate actions when the opportunity arises.
A third conflict-related pathway refers to normative changes that may emerge as a knock-on effect of the two previous mechanisms. Historically, women have been viewed as belonging to the private realm, leaving men to take on public positions, leadership roles, and to serve in the military. Changes to the division of labor also change perceptions of what a woman can or cannot do; when women assume these roles, it shows the society at large that women can perform the same tasks as men.
In the words of an interviewee from Liberia, "[t]he war brought a lot of evils, and a lot of good things. It brought out our leadership abilities in women" (Fuest 2008, 202). Importantly, these processes also can increase women's self-awareness and reveal how important their contributions can be (Kaufman and Williams 2010). Consequently, mobilizing for peace can be a stepping stone toward continued female mobilization and a normative acceptance of women's public engagement after war, resulting in (inter alia) increased share of women in formal political institutions and new opportunities for women to take part in peace-building processes (e.g., .
There is emerging empirical support for each of these mechanisms, even if much of the evidence is indicative or case-specific due to data limitations that presently prevent investigating such processes in a more detailed, comparative causal inference framework. However, what is largely missing from this literature is a deeper understanding of the conditions that make these mechanisms most likely to manifest themselves in measurable improvements in women's empowerment after civil war. In the following, we outline a simple theoretical storyline that connects the empowerment mechanisms to three central features of the previous conflict: its level of severity, its mode of resolution, and its facilitation of a peace agreement with specific gender provisions.

Battlefield Severity and Post-conflict Empowerment
Civil conflicts take on many forms, and it is often much easier to tell two conflicts apart than identifying their commonalities. Despite encouraging findings from recent scholarship about a general association between armed conflict and female empowerment (e.g., Hughes and Tripp 2015;Webster, Chen, and Beardsley 2019), it is clear that civil conflicts vary widely in the extent to which they manage to inspire positive shifts in women's political status in society. An important reason for this variation, we contend, is contextual heterogeneity, or differences in fundamental features of the recently ended conflict. 7 The first and arguably most defining feature of armed conflict that can have a substantial bearing on post-conflict sociopolitical development is its level of severity. The more destructive the conflict in terms of material costs and loss of lives, the greater the shock to society. An armed conflict that directly affects a sizable share of the population will generate more opportunities for women to join new segments of the labor market, engage in social movements and community activities, and have greater influence on social and political norms, with cascading impacts on attitudes toward women through the pathways discussed above. This intuition is commonly, if sometimes tacitly, acknowledged in empirical research, which often is restricted to analysis of major civil (and interstate) wars.
Present evidence for a conditioning role of battlefield severity in shaping postconflict empowerment is limited and mixed. Studies of female parliamentary representation in Sub-Saharan Africa have found a larger increase in response to major wars than low-intensity conflicts (Hughes 2009;Hughes and Tripp 2015) whereas Webster, Chen, and Beardsley (2019) report a negative but dissipating short-term effect of number of battle-deaths on women's empowerment. Notable differences in samples, conflict types, and analytical approaches between these studies complicate direct comparison but we conclude that current understanding suggests a positive association between civil conflict severity and post-conflict change. Our first testable hypothesis is: H1: Women's empowerment increases more after major civil wars than after minor conflicts, ceteris paribus.

Peace Agreements and Post-conflict Empowerment
The manner in which conflicts end can have a profound effect on subsequent statebuilding. For example, civil wars that end in military victory to one side generally tend to produce a more lasting peace than those that end in negotiated settlements, although peace agreements (PAs) can have a positive impact on the durability of peace under the right circumstances, and their record has improved considerably in the post-Cold War era (Caplan and Hoeffler 2017;Mason and Greig 2017;Licklider 1995;Shair-Rosenfield and Wood 2017). The negotiated manner through which formal PAs are reached imply that competing stakeholders get a chance to voice their grievances and raise their demands. Since the number of issues on the table when post-conflict state-building is negotiated will be greater, required solutions will be more comprehensive than when the winners take all. Commonly, PAs involve power sharing arrangements with minority protection, which rarely can be designed and implemented without reforms to judicial and political institutions. It is unsurprising, therefore, that conflicts that are settled by a PA see larger improvements in civil liberties and are better at reducing socioeconomic inequalities (Stewart and Daga 2017). Likewise, PAs may facilitate liberalization of the electoral system, such as the introduction of gender quota (Anderson and Swiss 2014;Lukatela 2012). This broader effort to achieve postconflict stability and prosperity is likely to shape both men and women's civil liberties and political freedoms (e.g., through democratic reforms), but given that women on average have fewer rights from the outset, their potential for positive change is greater.
Post-conflict female empowerment also may emerge from more conscious processes. Even though women traditionally have been denied a seat at the negotiation table (Bell and O'Rourke 2010;Krause, Krause, and Bränfors 2018), the negotiation process may provide opportunities outside the formal channels, such as through reconciliation ceremonies, regional consultations, lobbying, and media campaigns (Bouta, Frerks, and Bannon 2005). For example, Bell (2018) shows how women can navigate the peace process at different stages to try to influence gender equality even in the absence of a stark focus on gender from the outset of the negotiations. The international community also has greater leverage to influence post-conflict development when the conflict is ended by a negotiated settlement. Among other things, countries with formal PAs receive more aid than other post-conflict countries (Stewart and Daga 2017), implying greater donor influence over recipient countries. Hence, female empowerment is expected to be a distinct outcome of a PA, also in the absence of an explicit focus on women's interests in the accords. Our second hypothesis is: H2: Women's empowerment increases more after formal peace agreements than after other civil conflict endings, ceteris paribus.

Gender Provisions and Post-conflict Empowerment
Just like conflicts come in many forms, so do peace agreements. Recent research has shown that there is considerable variation in the extent to which PAs are gender sensitive, i.e. whether they include specific provisions for empowering girls or women (Bell and O'Rourke 2010). One manner in which the mechanisms presented above might materialize is through increased war-time female agency forming the contents of negotiated settlement (e.g., True and Riveros-Morales 2019). The inclusion of gender provisions in a PA reflects an explicit awareness of the significance of gender equity in the negotiations and an agreed-upon intention to strengthen women's position in society. The explicit reference to gender or women can be leveraged by women's movements in the post-conflict environment to pressure the government to follow up on issues of gender equality. The extent to which women are recognized in the PA typically reflects their mobilizing capacity, their access to negotiation fora, and the presence of supportive international mediators (Bell 2018). All else equal, PAs focusing on women's rights thus should produce better outcomes for women in the post-conflict stage than PAs lacking such provisions.
An important policy relevant question remains, however: does the empowerment effect of gender provisions depend on successful implementation or is legal recognition of such policies in the PA framework sufficient to facilitate strengthening of women's political status? Providing a definitive answer to this question is challenging, although recent and emerging research suggests that gender provisions are most likely to engender real improvements in women's political status when women take active part in elite peace processes and the resulting PA provisions are accompanied by specific gender quotas (True and Riveros-Morales 2019). In other words, substantive involvement may matter more than token representation. However, separating between what is implementation of PA-specific gender provisions (explanans) and what are new liberating policies and regulations that emerge as a consequence of the PA (explanandum) is not trivial, especially when working with aggregate data in a comparative large-N framework. Given that our ambition is to evaluate and identify general conflict-related conditions under which female empowerment is most likely to materialize, we are unable to provide empirical insights into the question of de jure versus de facto gender provisions here. Instead, we more simply propose: H3: Peace agreements that explicitly address women's status have a larger positive effect on subsequent female empowerment than other peace agreements, ceteris paribus.
Before presenting the research design and the results from the empirical analysis, two issues related to the temporal dimension of women's empowerment deserve attention. Although the causal pathways proposed above (participation, mobilization, normative change) describe processes emerging during civil war, a positive impact on female empowerment may not be detectable until after fighting stops. During civil war, a state of emergency likely prevents normative and livelihood changes from manifesting themselves in observable dimensions of formal individual-or institutional-level female empowerment due to postponement of national elections and reforms that otherwise would reflect shifting public sentiments. At the same time, the notion of civil war ending as a critical juncture implies that effects are immediate but not necessarily long-lived. Accordingly, our primary focus is on female empowerment during the immediate post-conflict years, and our argument is not necessarily inconsistent with views that women often experience a backlash in newfound freedoms in the aftermath of war (Ní Aoláin, Haynes, and Cahn 2011).

Data and Methods
To test the hypotheses, we construct a country-year dataset covering all independent states between 1975 and 2017 (N ¼ 6,131). The temporal domain is determined by the availability of the peace agreement data, and the same sample is used in all main models to facilitate direct comparison of results. We use two complementary dependent variables to capture theorized improvements in women's participation in society at large as well as in formal political institutions. To this end, we employ data available from the Varieties of Democracy project (V-Dem). The first outcome variable is an additive index, women's civil liberties index (CL), composed of three variables: women's freedom of discussion, women's participation in CSOs, and women's access to justice (Coppedge et al. 2019). These components take into account both formal discrimination, e.g. due to governmental laws, as well as informal discrimination, e.g. from cultural norms, and reflect women's empowerment in an absolute sense. The second index, women's political representation (PR), is taken from V-Dem as is, and measures the extent to which women are descriptively represented in formal political positions, including in parliament. 8 Unlike CL, PR expresses the situation for women relative to men's. Both indices range from 0 to 100, with higher values signifying higher levels of empowerment. In accordance with the proposed hypotheses emphasis on growth, we operationalize our dependent variables as change in each of these indices, defined as y t À y tÀ1 .
Data on civil conflicts are taken from UCDP/PRIO Armed Conflict Dataset v.19.1 (Pettersson, Högbladh, and Ö berg 2019;. Since we expect conflict-driven changes in women's political standing to materialize primarily in the immediate post-conflict phase, the independent variables are specified as decay functions, defined as x ¼ 2 À ð t a Þ , where t is time in years since the last year of conflict activity and a is the designated half-life parameter. Based on preliminary inspection, we set the half-life parameter to two years, which implies that the anticipated conflict-related effects of severity, PAs, and gender provisions will halve every other year after the conflict has ended. 9 We assign the first calendar year at peace after conflict (t) a decay value of 1; three years after the conflict (tþ2) corresponds to a decay value of 0.5; five years at peace (tþ4) gives x ¼ 0.25 and so on. In other words, the decay function acts as a variable weight that reduces the influence of the conflict-specific variables as time passes. All countries are assigned the value 0 until they experience their first post-conflict phase, and the decay is reset to 1 in the initial year of the next post-conflict phase.
To test if the severity of the civil conflict matters for outcomes for women (H1), we specify separate decay functions for the aftermath of minor conflicts (less than 1,000 battle deaths) and major conflicts (at least 1,000 battle deaths). These variables are created by taking binary indicators of the accumulated intensity level (as coded in the UCDP dataset) multiplied with the decay, such that the minor conflict ending variable maintains a score of 0 (or continues its decay after an earlier minor conflict) after major conflict, whereas the same is true for the major conflict ending variable after minor conflict. In additional tests, documented in the online Appendix, we use the (log) accumulated number of battle-related deaths in the previous conflict in interaction with the decay as a more nuanced measure of intensity. Battle deaths data come from UCDP (Pettersson, Högbladh, and Ö berg 2019), supplemented with data from PRIO for the period before 1989 (Lacina and Gleditsch 2005). All models include a control for ongoing civil conflict, leaving peace as the reference category.
To test H2, we use data on peace agreements from UCDP, version 19.1 (Pettersson, Högbladh, and Ö berg 2019; Harbom, Högbladh, and Wallensteen 2006). First, we identify the subset of civil conflicts that ended in a peace agreement, signed during the last two years of the conflict. In a subsequent step, we generate a postconflict PA decay using the same procedure as described above, i.e. interacting a PA dummy with the two-year decay function, such that the influence of the settlement fades with time. Due to the scarcity of conflicts ending with PA, we do not distinguish between minor or major PA endings, but we control for major conflict ending in the PA models since conflict severity might affect the likelihood of a negotiated settlement, such that PA potentially acts as a mediator between severity and women's empowerment. In the PA models, we also control for conflicts ending in other ways (e.g., military victory or ceasefires without formal settlement), as well as ongoing civil conflict, implying that the reference category becomes country-years at peace.
The peace agreements dataset also includes information on whether the peace agreement included any gender provisions. 10 To evaluate H3, we thus code separate decay functions for PAs with gender provisions versus PAs without. We only look at the subset of peace agreements that ended the conflict (based on the dichotomous variable "Ended" in the UCDP peace agreement dataset). Again, we control for major conflict ending, conflicts ending without PA, and ongoing conflict, such that the estimated effect of gender provisions can be compared to other modes of resolution as well as to remaining at peace.
All models contain a limited number of control variables to minimize the risk of spurious results. First, since the potential for improvement in women's political power depends on the point of departure, we control for the level of female empowerment on the corresponding CL/PR dimension, measured in the previous year. 11 We also account for non-stationary variables by means of a linear time trend, defined as year minus 1974. We further control for income level using log-transformed real Gross Domestic Product (GDP) per capita in constant 2005 US dollar . Economic development is linked to increased gender equality through improvements in education, job opportunities, and welfare (Dollar 2004;Kabeer and Natali 2013), and wealthy countries also have less conflict (Gartzke 2007). We also control for women's education level, using the mean years of schooling for women aged twenty to twenty-four, expressed in years. The variable, taken from Wittgenstein Centre Data Explorer (Lutz, Butz, and KC 2014), comes in five-year intervals and was linearly interpolated between observation points. 12 To minimize influence of reverse causality, empowerment levels, education level, and GDP per capita are all specified with a one-year lag.
We do not include a control for democracy in the main models. Conceptually, democracy (however defined) refers to institutions, rules, and regulations that extend far beyond those enabling or limiting rights or privileges to specific gender categories. Accordingly, institutional reforms toward or away from democracy can happen without implications for women's political status, and changes in the latter need not be accompanied by, or have measurable implications for, other formal aspects of governance. However, because conflict resolution and peacebuilding sometimes imply democratic reforms that both affect the likelihood of a lasting peace and have direct implications for formal gender roles in society, controlling for democracy may lead to biased estimates. We test the inclusion of democracy in sensitivity tests and controlling for democracy does not substantively challenge the conclusions drawn here.
All statistical models presented below are estimated by means of Ordinary Least Squares (OLS) with panel-corrected standard errors (PCSE). The PCSE model accounts for problems related to heteroskedastic residuals and autocorrelation (Beck 2001;Beck and Katz 1995). In sensitivity tests, we estimate OLS models with country fixed effects to isolate the cross-temporal drivers of women's empowerment. These tests, as well as descriptive statistics, are documented in the online Appendix.

Results and Discussion
Before we put the formulated hypotheses to test, we briefly inspect some relevant patterns in the underlying data. Panel A in Figure 1 reveals growth in global average female empowerment, using the two complementary dimensions of civil liberties and political representation. While both dimensions follow the same overall upward trajectory, women have always fared better with regard to civil liberties than formal political representation. Panel B provides deeper insight into how the two dimensions compare for individual country-years. As expected, most observations fall close to the 45 diagonal, meaning that countries score similarly good (or bad) on both indices, but we also note a number of deviations from this overall pattern.
Next, we assess the extent of empirical support in favor of Hypothesis 1, that major internal conflict has a more pronounced effect on female empowerment than minor conflicts, other factors held constant. Table 1 reports the effects of civil conflict endings on women's civil liberties (Model 1) and political representation (Model 2). Judging by the sign of the coefficients for major conflict endings, the immediate aftermath of civil war is associated with a statistically significant improvement in both civil liberties and political representation. The coefficient for the civil liberties outcome implies an average increase of around 1% in the first postconflict year, compared to a year at peace. Although this effect may seem small, and it tapers off over time, the cumulative effect over the first five years is nearly three points growth in CL. The corresponding effect for political representation is smaller, at around 0.7% for the first year at peace. Consistent with results from some earlier research (Hughes and Tripp 2015), we find that minor conflict endings have a weak and non-significant effect on women's formal political status. Although the estimated coefficients for major conflict ending in Model 1 and 2 are not significantly larger than those for minor conflict ending (i.e., their 95% confidence intervals overlap), the substantive difference in effect sizes still lead us to tentatively conclude in favor of Hypothesis 1, that major civil conflicts have a more pronounced effect on women's subsequent political empowerment than minor conflicts. The controls in Table 1 behave mostly as expected; higher average levels of female education correspond to higher growth in women's political status, whereas level of economic development more broadly has a weak negative influence, all else equal. We also find that ongoing civil conflict has a larger and more consistent negative effect on civil liberties than on political representation, reflecting a tightening of state security measures that often infringe on individual freedom and liberties during the height of war. Existing levels of women's CL/PR participation, captured by the lagged level of the dependent variables, are negatively associated with the likelihood of improvement. In other words: the growth potential is higher for less egalitarian societies. Overall, the coefficients for the controls are quite small, suggesting that female empowerment is not a phenomenon that exhibits high volatility from year to year.
We now turn to the evaluation of Hypothesis 2, that civil conflicts ending by a formal peace agreement are associated with a stronger post-war empowerment effect. In line with expectations, Table 2 shows that the average growth in women's civil and political privileges is much higher after a negotiated agreement (conflict ending w/ PA) than after military victory or stalemate (conflict ending w/o PA). In Model 3, the estimated increase in civil liberties exceeds 3% during the first postconflict year, which accumulates to a nine-point growth in CL during the first five post-conflict years. This effect is more than three times as large as the effect we found for major conflict endings (Model 1) and also significantly larger than the effect for non-negotiated conflict endings. The effect of PA on institutional empowerment, Model 4, is twice as large as the effect of major conflict found in Model 2 and also much larger than the coefficient for conflicts terminating in other ways, even though there is a slight overlap in the 95% confidence intervals for the PA versus w/o PA variables in Model 4.
Interestingly, we note that major conflict ending, which was found to increase growth in women's civil and political rights, no longer is statistically significant once we account for how conflicts ended. In other words, there is little indication that conflict severity acts as a confounder that produces a spurious statistical relationship between conflict ending and female empowerment. Rather, we interpret this result as evidence that the major conflict ending effect reported in Table 1 was largely picking up the correlated but unobserved effect of PA. Taken together, these results undermine the empirical support for Hypothesis 1 but provide compelling evidence that negotiated settlements have especially beneficial societal consequences (H2). Figure 2, panel A and B, provides a visual representation of the results documented in Table 2. The positive effect of the early post-conflict years on women's civil liberties and political representation are clearly detectable, but we also note the higher estimated growth rate and the narrower confidence interval for the first dependent variable, consistent with the coefficients in Models 3 and 4.
Hypothesis 3 stated that peace agreements containing specific gender provisions are associated with an even more pronounced positive effect on post-conflict female empowerment than PAs without such provisions or conflicts ending in other ways. The results from this test are documented in Models 5 and 6 in Table 3, and visual representations of the main effects are shown in Figure 2, panel C and D. Here, results differ somewhat between the two empowerment dimensions. For civil liberties, we observe little difference between peace agreements with versus without a gender profile; both outcome types are associated with significantly higher growth in women's status than the average country-year at peace, and the magnitude of the effect is similar to the aggregate PA effect reported in Model 3. For political representation, however, we find that the positive effect of peace agreements reported in Model 4 is attributable primarily to those agreements that explicitly address women; the estimated effect for PA ending w/ gender is five times greater than that for PA ending w/o gender, even though the large standard error around the latter point estimate implies overlap in 95% confidence intervals. A potential explanation for   the relatively higher effect of gender provisions on political representation is that gender provisions address the political inclusion of women rather than their civil liberties. Post-conflict countries might be willing to improve the gender balance of the political institutions without simultaneously improving civil liberties. Again, major conflict endings are insignificantly related to female empowerment, implying that the positive effect in Table 1 is driven mostly by conflicts that ended by peace agreement (for political representation, mostly PAs with gender provisions). In sum, Hypothesis 3 receives mixed support; explicit gender provisions are important to strengthen the influence of PAs on women's representation in political office but they matter less for women's civil liberties, where the main distinction lies between PA of any kind and conflict ending in other ways. Note: OLS coefficients with t-scores based on panel-corrected standard errors in parentheses. CL is civil liberties; PR is political representation. The conflict ending variables are specified as decay functions of time since the last year of the conflict. *p < 0.05; **p < 0.01.

Robustness
In order to assess the robustness of the results reported here, the models have been subjected to a number of sensitivity tests, all of which are reported in the online Appendix. First, we implement a complementary test of how the severity of civil conflict impacts women's post-conflict empowerment by replacing the binary minor vs. major conflict coding with the log of battle-related deaths accumulated over time, as well as a post-conflict battle deaths decay-function, analogous to the simpler minor and major conflict decays. The results, shown in Table A4, reveal that a higher number of battlefield casualties is associated with a statistically significantly greater improvement in women's civil liberties and political representation, although when we also account for conflicts ending by peace agreements, the severity effect is no longer statistically significant (similar to the difference between Table 1 and Table 2 above). This supports our conclusion that conflict severity at most affects women's post-war empowerment indirectly, by increasing the likelihood of the hostilities terminating in a formal peace agreement, although the latter's effect is found to be more generic and robust (i.e., it applies also to minor conflicts). Further, we test alternative half-life parameters for the decay function to see if the most basic effect of the post-conflict period can be reproduced. In the main models, we employed a decay variable that assumed a halving of the post-conflict effect every second year. In sensitivity tests, we use alternative half-life parameters, set to four, six, eight, ten, and twelve years. As we would expect, the longer we assume the post-conflict effect to last, the smaller the coefficient, but even when extending the half-life of the decay function to twelve years, we detect a positive and statistically significant effect of conflict endings by PA on both dimensions of female empowerment. These results are displayed in Tables A5 to A9. Next, as an alternative to the decay functions, we estimate a set of models where we use simpler dummies representing the first few years after the conflict is terminated. Specifically, we estimate alternative post-conflict dummies for the initial two, four, and six years after the conflict. These models, shown in Table A10 to A12, generally show weaker effects for major conflict ending but the PA effect on women's empowerment is reproduced.
Fourth, we check that observed improvements in the two dependent variable indices are not driven by only one of the indicators that make up the indices. In Tables A13 to A15, we show that civil wars terminated by peace agreement are associated with a positive and statistically significant improvement on all individual indicators.
As a fifth sensitivity test, we re-estimate the main models while controlling for democracy, see Table A16. This test serves to address possible concerns that the reported PA effect on women's civil liberties and political representation in essence is capturing general liberalization and democratization processes, correlated with female empowerment but not strictly addressing the status of women. As we would expect, democracy is associated with higher growth in female empowerment, all else equal, but even so, the results for the post-conflict variables remain largely unaltered.
In further tests, we re-run the models while excluding countries that never experience conflict in the analysis period, to allow for a more homogenous and relevant sample. As shown in Table A17, this alteration did not significantly change our results. Next, to account for the changing international environment after the Women, Peace, and Security agenda was implemented with UN Security Council Resolution (UNSCR) 1325 in year 2000, which might have a systemic effect on the prospect for female empowerment, we interact the post-conflict variables with a dummy for years from 2000 onward. The results, documented in Table A18, remain consistent with those reported here, although we note that growth in female empowerment after conflict is not higher in the years after 2000, not even for conflicts that ended by peace agreement. This is contrary to the findings of Hughes and Tripp (2015). At the same time, UNSCR1325 has been criticized for slow implementation and lack of results (e.g., Pratt and Richter-Devroe 2011), which could explain why we do not see a more pronounced growth in post-conflict empowerment after year 2000.
As a more flexible approach to pick up particular system-level effects that could affect empowerment, we also run models with calendar-year fixed effects, presented in Table A19 Here, the parameter estimates for the post-conflict variables are less precisely estimated, due to multicollinearity correlations between a subset of the year constants and PA endings, although the size of the estimated effects are very similar to those included in Tables 1 to 3. Lastly, we consider the robustness of the main results to including country-fixed effects to address potential concerns with omitted variable bias and to isolate the temporal dimension of the post-civil conflict effect on changes in women's legal and political standing in society. Reassuringly, the results are substantively similar to those reported above.
In sum, we find limited evidence to substantiate Hypothesis 1, since the effect of conflict severity on post-conflict empowerment is mediated by peace agreements, which are found to have a stronger and more consistent influence on women's subsequent political empowerment. We confirm Hypothesis 2, that civil conflicts (at any level of intensity) ending by peace agreement are associated with a substantial growth in post-conflict empowerment. We also find some support for Hypothesis 3, that the growth potential in post-conflict empowerment is especially high after peace agreements with gender provisions, although this effect was found only for women's formal political representation.

Discussion
Arguably the most important and policy-relevant finding from this study is that civil conflicts terminating in peace agreements produce better outcomes for women than conflicts terminating in other ways. The positive effect of peace agreements is very consistent and robust to alternative model specifications, supporting claims that the negotiated termination of civil war opens a window of opportunity for improving women's position in society. This also demonstrates that societal change does not always follow an incremental process but can sometimes be subject to rapid improvements when the moment is ripe.
The fact that peace agreements have a positive average effect on the empowerment of women may seem like a trivial discovery. Yet, this result was not necessarily a given, considering that women rarely enjoy the privilege of sitting by the negotiation table and most peace agreements fail to address the situation of women as victims of violence and marginalized members of society. However, after year 2000 and the adoption of UNSCR 1325, there has been an increase in the number of peace agreements that include reference to women (Bell and O'Rourke 2010), and in some peace processes women also have played elite roles (True and Riveros-Morales 2019). Where women's access to the negotiations is more restricted, they can inform the process through more informal channels.
Post-war Liberia is one example where women gained some influence; not by invitation to the peace negotiations, but by extensive activism and political lobbying that contributed to placing women's rights on the agenda (Fuest 2008;. The Accra Agreement that that marked the end of the civil war included references to gender, including institutional reform to promote women's rights. 13 Another example is Nepal; The Comprehensive Peace Accord Signed between Nepal Government and the Communist Party of Nepal (Maoist) 14 in 2006 contained several references to women, including problems related to discrimination by gender, and violence against women, despite the fact that women were absent from the formal peace negotiations (Ariño 2008). Subsequent interim political fora and institutions included women 15 and there were some advancements in women's empowerment (e.g., Yadav 2016). Absent the potential for societal transformation embedded in the peace agreement process, it is unlikely that the women's movement would have succeeded in their push for gender reforms in the aftermath of the conflict. However, the situation for women in post-conflict societies should not be idealized, as there is no shortage of documentation of women's suffering during war and many have documented the potential backlash in women's status after progress have been achieved, showing that the effect does not automatically last forever.
The findings reported here support previous studies that find that female empowerment improves more rapidly in post-war countries (Hughes and Tripp 2015;Hughes 2009;Webster, Chen, and Beardsley 2019). What this study brings in addition is an explicit consideration of how key characteristics of recently ended civil wars shape subsequent opportunities for women to strengthen their formal role in society. In line with expectations, we find that more severe civil wars are associated with greater post-war improvements in women's empowerment. However, subsequent models revealed that the conflict severity effect works mostly through increasing the likelihood of negotiated settlements. Peace agreements, particularly those that contain gender-specific provisions, have a pronounced positive impact on post-conflict female empowerment whereas conflicts that end in other ways show little general tendency to catch up on (and exceed) what's lost in terms of gender equity during war.

Conclusion
This article investigates the conditions under which civil war can bring about increased political empowerment of women. We complement previous research on this topic by looking into how the severity and type of conflict ending play into post-war empowerment. Consistent with our theoretical argument, we find that civil wars that end by peace agreement are more likely to produce positive changes during the immediate post-conflict rebuilding phase that conflicts that end in other ways, and the effect is particularly notable for peace agreements with gender provisions. The results are robust to a wide selection of sensitivity tests.
These insights carry immediate policy relevance. First, the data provide compelling evidence that investing in negotiations that facilitate formal peace agreements are likely to produce important dividends for societal reconstruction and gender equity-as long as peace prevails. This should serve as a powerful voice in support of mediation efforts by the UN and the international community more generally. Settling armed conflict through comprehensive peace agreements also can be an important instrument in addressing many of the Sustainable Development Goals in conflict-affected regions, including on gender equity (SDG 5). The second significant take-away message from this research is the importance of mainstreaming gender in all aspects of conflict management and peace building, as called for by, e.g., the UN Women Peace and Security agenda. Including women in peace processes is the right thing to do from an equity perspective, but it also is likely to have positive and far-reaching societal implications, not only in securing gender provisions in formal peace agreements but also in sustaining social development and political liberalization and reducing the risk of civil war recurrence (Melander 2005;Schaftenaar 2017;Rosenfield and Wood 2017).
While this research has been able to respond to some pertinent questions about gendered implications of armed conflict, other puzzles remain unaddressed and new questions have emerged as a consequence of this work. Future research should seek to evaluate the relative importance of the three proposed mechanisms linking societal transformation during civil war with post-conflict empowerment for women. Quite likely that work will require a different analytical framework than the aggregated and generalizable approach taken here. Likewise, there is a need for more knowledge on long-term implications of conflict resolution and peacebuilding processes for gender equity. The results presented here demonstrate that the termination of civil war can serve as a critical juncture for women's empowerment, but less is known about whether and how this affects the risk of resurgence of violence, as well as the sustainability of such improvements in the medium and long term. Moreover, recent evidence suggests that countries with lower levels of women's participation in public life are less likely to engage in negotiations (Nagel 2020). The nexus between war-time female mobilization, peace agreements, post-conflict female empowerment, and peace duration deserves further exploration.
Although the results of this article show that women's political empowerment improves in the wake of a civil war, it should not be taken to mean countries escaping civil war score better on these indicators that they would have done in a counterfactual world without conflict. Nor does it mean that post-war improvements in women's status come naturally and effortlessly-quite the contrary, the increase in empowerment comes at a high cost and is only possible with the continuous and purposeful work of discriminated women.