Women’s Parliamentary Behavior

As women’s political representation has risen across the globe, scholars of gender and politics have devoted increasing attention to understanding dynamics of women’s election and representation. A significant body of research has studied women’s behavior in elected political office, often focusing on the conditions under which they will represent women’s interests, linking their descriptive representation to substantive representation (Childs and Krook, Reference Childs and Krook2006; Franceschet and Piscopo, Reference Franceschet and Piscopo2008; Phillips, Reference Phillips1995). This work has largely supported the proposition that women MPs will tend to support policies relevant to women (Childs, Reference Childs2004; Dodson, Reference Dodson2006; Piscopo, Reference Piscopo2011). This may be because of genuine interest, where women MPs have shared experiences with women as a group and thus form similar policy interests (Phillips, Reference Phillips1995), as well as a feeling of duty or obligation among women MPs to represent women’s issues (Childs, Reference Childs2004; Franceschet and Piscopo, Reference Franceschet and Piscopo2008; Schwindt-Bayer, Reference Schwindt-Bayer2010).

Researchers have also offered consistent evidence that gender norms and socialization shape women’s behavior and policy interests within legislatures. Using Eagly and Karau’s (Reference Eagly and Karau2002) role congruity theory as a framework, this body of work argues that social norms dictate certain characteristics that men and women should display within legislatures, influencing how they behave and what policies they support. Eagly and Karau (Reference Eagly and Karau2002) argue that women are expected to have more ‘communal’ characteristics, including being more nurturing, honest, and helpful, while men are expected to have ‘agentic’ qualities, such as being more assertive and dominant. These expectations can provide prescriptions for behavior in parliament, as well as backlash for defying such norms (Bäck et al., Reference Bäck, Debus and Müller2014; Brescoll, Reference Brescoll2011).

Experimental research has offered ample evidence in support of the role congruity theory. In various workplaces and social settings, researchers have found that women speak less, are interrupted more, and use less assertive language than men do (Brescoll, Reference Brescoll2011; Karakowsky, et al., Reference Karakowsky, McBey and Miller2004). Many studies have similarly found that gender affects speech behavior in parliaments. Bäck and Debus (Reference Bäck and Debus2019) and Clayton and Zetterberg (Reference Clayton and Zetterberg2021) examine European and African parliaments, respectively, and both find that women speak less than men in parliamentary debates. These patterns are also impacted by constructs of power. For example, in Uganda, women in reserved quota seats speak less often than men and women in open seats (Clayton, Josefsson et al., Reference Clayton, Josefsson and Wang2014).

Scholars have also used the role congruity theory to explain how social norms can influence what policies women MPs prioritize. Women MPs are expected to focus more on communal policy issues due to their stereotyped ‘communal’ traits (Eagly and Karau, Reference Eagly and Karau2002; Huddy et al., Reference Huddy, Terkildsen, Huddie and Terkildsen1993; Koch, Reference Koch2000). This includes topics categorized as ‘soft’ or ‘feminine’, like women, gender, family, education, health, social affairs, culture, and youth.Footnote ¹ Across different levels of democracy and regime type, research shows that women MPs tend to speak on topics that relate to women and gender (Clayton, Josefsson et al., Reference Clayton, Josefsson and Wang2017; Piscopo, Reference Piscopo2011), as well as other ‘soft’ issues that are linked to communal stereotypes, such as education and healthcare (Bäck and Debus, Reference Bäck and Debus2019; Baumann et al., Reference Baumann, Bäck and Davidsson2018; Vet and Devroe, Reference de Vet and Devroe2022). Women politicians are also more likely to receive appointments that are relevant to ‘soft’ issues than ‘masculine’ or ‘hard’ issues, such as defense or the economy (Barnes and O’Brien, Reference Barnes and O’Brien2018; Krook and O’Brien, Reference Krook and O.’Brien2012).

These dynamics create double standards for women. They face limitations in voicing their opinions and interests within these institutions, as well as possible barriers to certain policy portfolios. When they defy such gendered expectations, however, they can face backlash from both their party and voters. For instance, women who hold masculine policy portfolios and who are more active on the parliament floor are less likely to be reappointed or promoted (Baumann et al,. Reference Baumann, Bäck and Davidsson2018; Yildirim et al., Reference Yildirim, Kocapınar and Ecevit2021). This creates a policing mechanism in which women may be unlikely to defy these norms even if they have policy interests beyond ‘soft’ issues because they must strategically consider how it may impact their personal career (Bäck et al., Reference Bäck, Debus and Müller2014).

Scholars have found, however, that these patterns are tempered when women MPs are introduced to some kind of intervention in gendered norms and socialization. Wahman et al. (Reference Wahman, Frantzeskakis and Yildirim2021) demonstrate that, in Malawi, women MPs were more active in parliament after Joyce Banda ascended to the presidency following the unexpected death of Bingu Mutharika. They argue that Banda was able to ‘normalize female political power, redefine gendered norms about appropriate female political behavior and competences, and create momentum for more assertiveness among female MPs’ (1). Similarly, in the United Kingdom’s House of Commons, Blumenau (Reference Blumenau2021) shows that female MPs are more engaged in debates led by women ministers, suggesting that women in leadership positions act as role models for other women in politics.

These studies offer evidence of gendered patterns of behavior that can be further improved by conditions that empower MPs to be more engaged. While this demonstrates the possibility of changes in group-level socialization, it does not explore variation in individual-level socialization. The following section discusses the potential micro-level effects of women’s wartime experiences as one type of socialization that can alter gendered behavior.

Women’s Socialization in Militarized Organizations

The role congruity theory suggests women and men have different social expectations, leading to unique socialization and behavior. We build on this to examine variation within groups of legislators, particularly women legislators, based on their background and their respective socialization regarding feminine and masculine behaviors. In this section, we discuss the socialization of women within non-state armed groups and then build our theory on the legislative behavior of such rebel women in the following section.

Women are motivated to join rebel groups, like men, for various reasons, such as ideology, political and economic grievances, security and protection, upward mobility, and social networks (Henshaw, Reference Henshaw2016; Thomas and Bond, Reference Thomas and Bond2015; Viterna, Reference Viterna2013). Many women are also motivated by the political ideals of the organization (Henshaw, Reference Henshaw2016; Sjoberg and Gentry, Reference Sjoberg and Gentry2007). For others, the decision to join may be a matter of pure survival and protection (Coulter, Reference Coulter2008; Mazurana et al., Reference Mazurana, McKay, Carlson and Kasper2002). However, interpretations of women’s participation in rebel groups are often colored by gendered stereotypes. Because participation in a violent rebellion defies ideas about women as passive and peaceful, when they participate, Sjoberg and Gentry (Reference Sjoberg and Gentry2007) argue their ‘womanhood, their cognitive ability, and their sanity’ are all questioned, which in turn ‘negates any cause for which violent women may be acting’ (50). Yet, their choice to join represents their agency and is a ‘vast departure’ from gender expectations (Azmi, Reference Azmi2015). This could suggest these women may have a less common characteristic based on their choice to escape the mold. By joining, they demonstrate a willingness to buck the traditional gender socialization and norms they have known in their lives.

As women are integrated into these organizations, they are further socialized in a manner that contradicts traditionally feminine norms. Like formal state militaries, non-state armed groups are hyper-masculine organizations in which typical masculine characteristics are amplified to glorify aggression and strength (Enloe, Reference Enloe1989, Reference Enloe2000). For men, these spaces reinforce and augment existing masculine socialization. However, for most women, who likely were socialized in a feminine manner, these spaces problematize feminine characteristics as weaknesses and push women to act in a more masculine manner (Cohn, Reference Cohn2000; MacKenzie, Reference MacKenzie2015). Beliefs that effective militaries must amplify masculine characteristics lead to a glorification of masculine behavior and a rejection of the feminine as anything other than the innocent ‘beautiful souls’ to be rescued (Elshtain, Reference Elshtain1982; Enloe, Reference Enloe1989). Conceptualizations of a ‘good soldier’ and a ‘masculine soldier’ are indistinguishable – to be a good soldier, one must be masculine (MacKenzie and Foster, Reference MacKenzie and Foster2017).

Socialization of these beliefs is likely pervasive across most levels of the organization. Rebels are socialized according to the organization’s political and tactical goals through group institutions such as recruitment, military training, political training, and disciplinary regimes (Green, Reference Green2017). Even for those who may have held non-gender-conforming characteristics before joining, this socialization likely pushes a deeper and more extreme form of masculinity that contradicts any pre-war upbringing. This socialization is reinforced as women in both non-combat and combat roles are trained and educated similarly to men.Footnote ² Alison (Reference Alison2004) quotes a female combatant of the Irish National Army, recalling, ‘First and foremost, you were a soldier, and that was it’.

Throughout their training and socialization, ‘sameness’ is emphasized between women and men in a way that glorifies masculine behavior in all members. In these hyper-masculine spaces, women’s ‘sameness’ comes from adopting ‘masculine and militarized attitudes and values’ (Jordan and Denov, Reference Jordan and Denov2007, 47). In Colombia, Schmidt (Reference Schmidt2020) recalls Fuerzas Armadas Revolucionarias de Colombia – Ejército del Pueblo (FARC–EP) women frequently claimed that there were ‘no women in FARC, because they were all treated like men’ (6). Similarly, in Eritrea, Barth (Reference Barth2002) found that ‘Rather than describing women as acting out a new gender role, female fighters in Eritrea say that they were acting as men’ (17). For rebel women, masculine behavior is the currency for respect (Jordan and Denov, Reference Jordan and Denov2007; Marks, Reference Marks2017; Schmidt, Reference Schmidt2020). By acting the ‘same’ as men, rebel women may believe they will be valued and rewarded. In South Africa, a former female combatant of the Umkhonto we Sizwe (MK) argued that through the ‘equality practiced’ in the organization, women were ‘recognized as equals’ (Cock, Reference Cock1992, 163).Footnote ³

This sense of ‘sameness’, along with the skill-based training that rebel women receive, can affect how they view themselves and their place in society (Barth, Reference Barth2002; Lyons, Reference Lyons2004; Moral, Reference Moral2014; Yadav, Reference Yadav2016). Steenbergen (Reference Steenbergen2022) finds among former female combatants in Liberia and Nepal that joining the rebellion increased their overall self-reliance, self-esteem, self-confidence, and awareness that they can and should speak out regarding injustices (134). One female combatant told Steenbergen (Reference Steenbergen2022), ‘Before joining the party, there was not much courage within me as I was female and I cannot do everything, but after joining and surviving on my own, I became self-reliant and [feeling] I am not less than men’ (135). When women return from the war, this socialization and confidence remain with them, impacting how they feel and act in post-conflict environments (Barth, Reference Barth2002; Bernal, Reference Bernal2001).

In the next section, we explore how wartime gender socialization may be associated with how these select women behave once they enter political roles.

Theoretical Expectations: Rebel Women’s Post-Conflict Political Behavior

Following civil conflicts, a high number of women opt into political roles. Their choice of career confers information about their personality and behavior, as Karpowitz and Mendelberg (Reference Karpowitz and Mendelberg2014) argue that women who select into politics do not represent the average woman – instead, they are a more ‘unusual’ and ‘masculine’ type of woman (335). Yet, gender expectations still present powerful effects on how women participate in politics, resulting in them speaking less than men and focusing more on softer, feminine topics (Bäck and Debus, Reference Bäck and Debus2019; Bäck et al., Reference Bäck, Debus and Müller2014; Baumann et al., Reference Baumann, Bäck and Davidsson2018; Clayton, Josefsson et al., Reference Clayton, Josefsson and Wang2014; Clayton and Zetterberg, Reference Clayton and Zetterberg2021). Rebel women also represent an ‘unusual’ type of woman. Their choice to join the rebellion and engage politically signals a willingness and comfort with defying gender norms (Azmi, Reference Azmi2015). Yet, through their participation and subsequent socialization, their defiance of gender norms, as well as their confidence and political interest, likely deepens over the war. We argue that behavior among women MPs after war may vary based on rebel background.

We assume that women MPs have more feminine gender socialization based on prescribed societal gender norms, which often emerge even more forcefully after conflict (Handrahan, Reference Handrahan2004; Pankhurst, Reference Pankhurst2002). Following the role congruity theory, women MPs may adhere to traditional gender norms that dictate that they should be less assertive and agentic than their male colleagues, meaning they will speak less frequently in parliament, especially on ‘hard’ or ‘masculine’ topics. Additionally, they likely have genuine preferences that also influence their behavior.

In contrast, rebel women are likely to display more masculine behaviors. Based on their selection into rebel groups, it is likely that these women already possess non-gender-conforming characteristics. In cases where they do not, such characteristics emerge as they are socialized into these hyper-masculine organizations. Those with non-gender-conforming traits will likely see these characteristics deepened and reinforced over the course of the war. Moreover, they will become more confident through practice as they demonstrate this ‘sameness’ in behavior and agency with men.

Rebel women MPs also may not face the same pressures for role congruence, particularly when working within their former organizations. In fact, they may have little incentive to change their behavior, as they have already learned that it is accepted and encouraged within their group. Further, more masculine behavior is congruent with their identities as former rebels, which may trump other identity groups. As a result, when they step into political roles within the same organization, they can continue their wartime masculine behavior.Footnote ⁴

This may have implications for several facets of women MPs’ behavior in government, including how often they speak and what they speak about. While women MPs tend to be less assertive in parliament, speaking less than men (Bäck and Debus, Reference Bäck and Debus2019; Clayton, Josefsson et al., Reference Clayton, Josefsson and Wang2014; Clayton and Zetterberg, Reference Clayton and Zetterberg2021), rebel women may be more assertive than the average woman MP due to factors associated with their wartime participation. This may include their pre-war levels of gender incongruity, their wartime socialization, and the confidence gained through their participation. For example, in Ethiopia, a woman shared with Veale (Reference Veale2003) that non-rebel and rebel women acted differently in their local council, saying:

In politics, those (women) who were fighters are better than those who were not. If there is a meeting at the Kebele, these fighter women ask and answer questions. They participate actively, but this activity is not common to those who were not fighters or other women because they feel frightened or afraid of anybody. (47)

Rebel women’s wartime experiences may also impact their fear of backlash for non-gender-conforming behavior. Scholars have emphasized that women’s gender-conforming behavior in parliament may also result from a fear of punishment – rather than a true individual adherence to these norms – as women who are perceived as more assertive and dominating in parliament tend to be punished by party leaders and voters (Baumann et al., Reference Baumann, Bäck and Davidsson2018; Yildirim et al., Reference Yildirim, Kocapınar and Ecevit2021). If rebel women are primarily elected to their former rebel parties, these same incentive structures may not be in place, as rebel women may have experienced both the expectation and reward of acting in a more masculine manner. In turn, they may be less likely to experience some of the double standards and backlash that other women MPs may experience from demonstrating more masculine behavior in parliament. Thus, they may be more likely to demonstrate more assertive behavior, including speaking more frequently on the parliament floor.

This may similarly impact the content of rebel women’s speeches. Although average women MPs are also less likely to speak on masculine or ‘hard’ topics (Baumann et al., Reference Baumann, Bäck and Davidsson2018; Vet and Devroe, Reference de Vet and Devroe2022), rebel women may not avoid these topics because they do not view them as incongruous to their roles as rebel MPs. Due to their wartime experiences, they may be accustomed to speaking on more masculine topics, especially in military, defense, and security matters. Moreover, their feelings of ‘sameness’ developed during the war may have instilled confidence to speak on issues often assumed to be beyond women’s purview. Again, they also may be less likely to fear backlash from speaking on these topics, as their colleagues have proven accepting of more masculinized behavior. As such, we argue, they will be more likely than other women to speak on the same topics as men rather than avoiding ‘hard’ topics.

Alternatively, rebel women may be more inclined to speak on topics related to wartime interests and experiences. Just as women may be more likely to speak on communal issues because they are more relevant to their experiences and interests (Phillips, Reference Phillips1995), rebel women’s policy interests in government may reflect the factors that drove them to join the rebellion or issues that became especially salient to them over the course of the conflict as they were further socialized into the group’s political goals. This includes certain hard topics like defense or security but also may extend to topics that are considered soft, such as civil rights, social welfare, and inequality, that are related to the sort of grievances that drive individuals to rebel. They may be more likely to speak on such issues due to an interest related to or borne from their wartime participation. Likewise, they may have become more confident speaking on these issues during the war, increasing their likelihood to continue in parliament after the war. In contrast, non-rebel women are not expected to prioritize this collection of issues, as they do not share direct experiences of conflict or socialization into the political goals of a rebel group that typically characterizes rebel women.

Case Selection

Uganda and Zimbabwe offer a useful test of our theory due to the dynamics of their civil conflicts and the subsequent high levels of women’s post-war political representation. The Ugandan Bush War (1981–6) concluded with the rebel forces, the National Resistance Army (NRA), overthrowing the government. Scholars have argued that the war transformed Uganda from a country with limited opportunities for women to one in which women were accepted and promoted in politics (Boyd, Reference Boyd1989; Byanyima, Reference Byanyima1992; Goetz, 1998; Tripp, Reference Tripp2000). Women were active within NRA forces and held a variety of combat and non-combat roles (Boyd, Reference Boyd1989; Byanyima, Reference Byanyima1992; Kasfir, Reference Kasfir2005; Mugabme, Reference Mugabme2000). The women who participated have emphasized how their role in the NRA was transformational, arguing that they ‘acquired a sense of competence, self-confidence, and opportunity for self-determination’ (Byanyima, Reference Byanyima1992, 129). Women’s participation and their tactical success also led to changing attitudes among their male comrades and outside observers (Tripp, Reference Tripp2015). Uganda became one of the first countries to adopt gender quotas in 1989, leading to high levels of women’s representation over the subsequent decades. Further, the continued role of the NRA (now the National Resistance Movement/NRM) has ensured that former combatants are influential within formal politics (Reuss and Titeca, Reference Reuss and Titeca2017). This makes it an ideal case to compare former rebel and non-rebel women’s behavior. In our sample, although their number is quite small (less than five per cent of the non-rebel women), rebel women are still about thirty per cent of the ex-rebels in the chamber (also see Appendix, Table 2). Moreover, scholars have argued that though ex-rebels’ representation in Uganda has dwindled in the last few decades, their influence is perceived to still be strong (Brannon, Reference Brannon2023; Reuss, Reference Reuss2020; Reuss and Titeca, Reference Reuss and Titeca2017).

The Zimbabwean case offers similar and contrasting dynamics. The Zimbabwean war (1964 to 1979) ended with the removal of colonial forces. After the war, the rebels, the Zimbabwe African National Liberation Army and the Zimbabwe People’s Revolutionary Army, transitioned into political parties and eventually merged into one party, the Zimbabwe African National Union – Patriotic Front (ZANU-PF), in 1987. Similar to the NRA, women in the guerilla forces served in support, combat, and leadership roles and have expressed that these experiences offered them ‘new vision for their role in society’ (Seidman, Reference Seidman1984, 427). Yet, many scholars have questioned the veracity of claims that women’s roles in the field were emancipatory, emphasizing that they only had a ‘limited voice’ (Seidman, Reference Seidman1984, 420) and that gendered hierarchies were reinforced, limiting women’s access to decision-making and training and frequently subjecting women to sexual abuse and harassment (Lyons, Reference Lyons2004). This is reinforced by the lack of feminist progress after the war (Law, Reference Law2021; Ranchod-Nilsson, Reference Ranchod-Nilsson2006; Sadomba and Dzinesa, Reference Sadomba and Dzinesa2004). It was not until 2013 that the country adopted gender quotas, and large gains in women’s representation were achieved. Nonetheless, due to the dominant position of the former rebels in government, ex-rebels have long cemented their place in politics, though the representation of non-rebel women did not rise significantly until several decades after the war. This is evident in our sample, where almost twenty-five per cent of the women MPs are ex-rebels. This figure goes up to forty-seven per cent for women MPs of the government party after the 2013 election (also see Appendix, Table 2). As such, Zimbabwe offers a useful comparison point to Uganda.

Beyond these dynamics, three other critical factors could impact conclusions drawn from these cases. First, each is a semi-authoritarian country with a dominant political party – both of which are former rebel parties (the NRM in Uganda and the ZANU-PF in Zimbabwe). While the closed nature of the authoritarian regime may hinder the degree to which all representatives, including women, are outspoken within parliament, the prominent nature of the dominant rebel parties may also offer some perceptions of protection to representatives – mitigating some of the effects of the regime type. There is evidence that ruling party MPs tend to be fairly strong-minded in the chamber in both Uganda (Collord, Reference Collord2016, Reference Collord2021; Doro and Kufakurinani, Reference Doro and Kufakurinani2018) and Zimbabwe (Doro and Kufakurinani, Reference Doro and Kufakurinani2018). In addition, since our research design compares the behavior of sub-groups of MPs that are all active within the same political context, these regime qualities should not present a validity threat. Furthermore, while the regime type and party system do limit the scope conditions of any findings, it does not necessarily limit generalizability given that rebel parties – especially victorious rebel parties – are more likely to be authoritarian and dominant political parties (Lyons, Reference Lyons2016).

Second, the rebel parties included in the sample each held a leftist ideology during the war. While leftist ideology is associated with an increased likelihood of recruiting women to rebel groups (Wood and Thomas, Reference Wood and Thomas2017), previous work has emphasized that leftist groups frequently abandon a commitment to ideological commitments (Burihabwa and Curtis, Reference Burihabwa and Curtis2019), but also specifically to gender equality after the war is over (Henderson and Jeydel, Reference Henderson and Jeydel2013). In fact, Brannon (Reference Brannon2025) finds that leftist rebel parties tend to elect far fewer women on average than rebel parties with other ideological stances. Thus, due to the dominant status and ideology of these rebel parties, it is possible that any findings understate the true relationship more generally.

Third, these cases are suitable regarding institutional setup. Both countries are comprised of first-past-the-post single-member constituencies. However, in both cases, constituencies are part of electoral districts, each of which elects a woman representative. In terms of speechmaking, MPs have considerable flexibility. There are rules in place to ensure that members of both the government and the opposition have the opportunity to take the floor, but to do so, an MP only needs to be recognized by the Speaker. In both cases, ministers, committee chairpersons, leaders of the opposition, and government representatives are given priority, but other than that, the Speaker is responsible for distributing time on the floor (Collord, Reference Collord2016, Reference Collord2021; Doro and Kufakurinani, Reference Doro and Kufakurinani2018; Kriger, Reference Kriger2006; Parliament, Reference Parliament2021, Reference Parliament2005). The similarity of the systems and the limited institutional constraints make these cases suitable for testing our theory.

We expect that the findings presented should be applicable, particularly to contexts with other rebel parties. Many rebel parties, especially in Africa, are dominant parties and hold a majority of seats within their legislatures (Lyons, Reference Lyons2016; Söderberg Kovacs and Hatz, Reference Söderberg Kovacs and Hatz2016). Further, the level of representation of female rebels in both cases is similar to global patterns, where rebel women hold an average of five per cent of rebel party seats (Brannon, Reference Brannon2025). While this may seem like a low percentage of seats, women’s numerical representation varies according to party legislature size. Further, though rebel women only hold a small percentage of seats in the cases analyzed in this paper (and in other cases), existing research on female rebel veterans in politics in Uganda and Zimbabwe suggests that they have an outsized impact on politics (Boyd, Reference Boyd1989; Goredema and Chigora, Reference Goredema and Chigora2011; Mugabme, Reference Mugabme2000; Seidman, Reference Seidman1984; Tamale, Reference Tamale1999), making them an important group to study.Footnote ⁵

Data on Parliamentary Speaking and Research Design

To test our hypotheses, we leverage a novel dataset of parliamentary speeches including data from MPs in Uganda (2011–21) and Zimbabwe (2008–18). In total, this amounts to 247,000 speeches made by 1,093 MPs. To compile this dataset, we collected the transcripts of parliamentary discussions (Hansards), which were available on the website of each country’s parliament. The speeches were categorized based on the Comparative Agendas Project (CAP) coding scheme (Baumgartner, et al., Reference Baumgartner, Green-Pedersen and Jones2006). Specifically, each speech was classified into one of the major 23 CAP topics, which ranged from macroeconomics to cultural policy. An additional category was included for procedural speeches.

To identify speeches on ‘hard’ topics, we follow the categorization of Bäck et al. (Reference Bäck, Debus and Müller2014). Based on their classification, we categorize the following as hard topics: law and crime, defense, macroeconomics, domestic commerce, foreign trade, and international affairs. We base our classifications of topics related to wartime experiences on common factors associated with rebel group recruitment and the political and military priorities of rebels in Uganda and Zimbabwe. This includes defense, international affairs, law and crime, civil rights, and social welfare.Footnote ⁶

To facilitate coding this expansive corpus of speeches, we employed a neural network. The effectiveness of this approach for text classification has been widely documented (see, for example, Mironczuk and Protasiewicz (Reference Mironczuk and Protasiewicz2018); Popov (Reference Popov2018)), and it is increasingly utilized within the context of political science (see, for example, Ballard et al. (Reference Ballard, DeTamble, Dorsey, Heseltine and Johnson2022); Wahman et al. (Reference Wahman, Frantzeskakis and Yildirim2021)). For our application, we employed BERT (Devlin et al., Reference Devlin, Chang, Lee and Toutanova2018), a neural network developed by Google. This data generation process is part of a larger data-building project, so the network was fine-tuned on data from Ghana, Malawi, Zambia, and Zimbabwe. The data from Ghana were used to train the model, and the data from the other countries were used to validate that the trained net could code speeches across countries with sufficient accuracy. The network coded the policy topic of test data with an average f1 score of 82.6 per cent (precision: 83.2 per cent, recall: 82.9 per cent).

We combined the speech data with data collected on the individual backgrounds of MPs elected during this time period. This includes data on their education, professional background, gender, and, most pertinently, connections to the former rebel organizations: the National Resistance Army in Uganda, the Zimbabwe African National Liberation Army, and the Zimbabwe People’s Revolutionary Army. These data were coded using various sources, including parliamentary profiles, media interviews, policy reports, academic research, and biographical profiles. Their participation had to be clearly specified to code an MP as a former rebel. While in some cases it is conceivable that individuals may want to conceal their wartime ties, given the prominence of these organizations within Ugandan and Zimbabwean politics, this is unlikely in these cases.

For the purpose of the analysis, the dataset was collapsed into a country–MP–year format, and it includes every MP elected in office for the years covered in each country. The key dependent variable, which is used to test H1, is operationalized as the number of speeches per year. To test H2, we rely on two variables. The first is the number of speeches on ‘hard’ topics, and the second is the ratio of ‘hard’ speeches to the total number of speeches made by the MP over the course of a year. This two-step approach allows us to better understand both the engagement with specific topics in parliament and the existence of a relative focus in particular topic areas. Similarly, to test H3, we use the number of speeches made on wartime experience-related topics and the ratio of wartime experience-related speeches over the total number of speeches made in a year. As we are primarily interested in differences in the level of activity between rebel and non-rebel women MPs, we introduce an interaction term with the constituent terms being gender and rebel group membership. Both independent variables are coded as binary.

We also include several control variables. We control for government membership, ministerial appointments, committee chairmanships, independent status, and the number of years that the MP has been in office.Footnote ⁷ In addition, we control for two categories of special elections. First, both Zimbabwe and Uganda have reserved women’s seats at the district level. Second, in Uganda, there are reserved seats for representatives of the military who are appointed by the Army Council. Further, we control for whether an MP was a rebel leader during their country’s liberation struggle. This helps us differentiate between effects driven by socialization within rebel groups versus individuals with specific traits self-selecting into both rebel groups and the parliament. We provide a summary of the dataset in Table 1 of the Appendix.

To test H1, we use zero-inflated negative binomial models. This accounts for the nature of the dependent variable – the number of speeches – which is an overdispersed and zero-inflated count variable. To test H2 and H3, we use zero-inflated negative binomial models for the count-type dependent variables and OLS regressions when the dependent variable is a ratio. To account for the structure of our data, we add country-level fixed effects. To account for time dependencies, we include linear splines with knots at the beginning and end of each parliament.Footnote ⁸ Also, we cluster standard errors at the MP level.

Analysis

First, we explore the relationship between gender, rebel legacies, and parliamentary speech-making descriptively. Figure 1 depicts the number of speeches per year made by MPs elected to political parties with rebel legacies. The barplot on the left compares speeches per year by men and women who were not rebels themselves. In the right-hand panel, we plot the number of speeches per year for MPs who participated in the rebellion. Based on these plots, non-rebel men speak more than rebel men on a yearly basis. Conversely, non-rebel women MPs make 19.77 speeches per year, while rebel women MPs make 28.67 speeches. This is a substantial difference of thirty-one per cent. These descriptive relationships suggest that women’s wartime roles may be related to their active engagement in debates, as they may feel more comfortable and confident speaking up than average women MPs.

Figure 1. Speeches per year by rebels and non-rebels.

In Figure 2, we examine the same relationship but focus on speeches made regarding ‘hard’ topics. The patterns are similar. While rebel men make substantially fewer speeches than non-rebel men, rebel women make 2.9 speeches per year more than non-rebel women. This corresponds to a difference of about thirty-three per cent. Similarly, in Figure 3, which graphs the number of speeches on wartime experience-related topics, we observe that rebel women are considerably more active than all other groups of MPs. Compared to non-rebel women, they make forty-six per cent more speeches per year. Overall, Figures 1, 2, and 3 show some level of support for all our hypotheses, indicating that rebel women are overall more active on the floor of parliament than non-rebel women generally and specifically on ‘hard’ topics and topics that relate to wartime experiences. This offers further evidence that some aspects of rebel women’s identity set them apart from other women, likely their propensity to display more masculine behaviors.

Figure 2. Speeches on hard topics per year by rebels and non-rebels.

Figure 3. Speeches on wartime experience-related topics per year by rebels and non-rebels.

To further explore the relationship between rebel ties and behavior in the chamber, we present a number of regression models. Since we are interested in the marginal effect of rebel background on women’s behavior, we base our discussion of the results on substantively meaningful simulations (see Ai and Norton (Reference Ai and Norton2003); Brambor, et al. (Reference Brambor, Clark and Golder2006)) following the observed values approach (Hanmer and Kalkan, Reference Hanmer and Kalkan2013). The results of the simulations are presented in Figures 4, 5, and 6. In each figure, each row has three graphs. The left-hand panel shows the predicted values for each group of MPs. The middle panel evaluates whether the difference between these predicted values are statistically significant between the groups of MPs (rebels and non-rebels) sharing the same gender (men and women). The right-hand panel evaluates the statistical significance of the difference in the predicted values between genders (men and women) with the same group (rebels and non-rebels). The full models are presented in Table 3 of the Appendix (Models 1–5).

Figure 4. Marginal effects between gender and rebel group membership for all speeches.

Note: Model results can be found in model 1 of Table 3 in the Appendix.

Figure 5. Marginal effects between gender and rebel group membership for speeches on hard topics.

Note: Model results can be found in Models 2 and 3 of Table 3 in the Appendix.

Figure 6. Marginal effects between gender and rebel group membership for speeches on wartime experience-related topics.

Note: Model results can be found in Models 4 and 5 of Table 3 in the Appendix.

Support for H1 is evaluated in Figure 4. The left-hand panel shows the expected number of speeches per year by rebel and non-rebel men and women MPs. Non-rebel women MPs are projected to make 22.7 speeches each year (95 per cent CIs: 18.4–27.1). On the other hand, Rebel women MPs are projected to make 41.3 speeches per year (95 per cent CIs: 27.3–55.4). This is a drastic increase of about fifty-five per cent. On the contrary, non-rebel men are projected to make 35.7 speeches per year (95 per cent CIs: 30.8–40.6), while rebel men are projected to make 28.9 speeches per year (95 per cent CIs: 15.7–42). The middle graph depicts the estimated change in the expected number of speeches for men and women when they have a rebel legacy. We find that the increase of 18.6 speeches per year for rebel women is statistically significant (95 per cent CIs: 4.5–32.7). Finally, in the right-hand panel, we compare the number of speeches made by men and women who either have or lack a rebel legacy. We find that non-rebel women make thirteen fewer speeches than non-rebel men each year, which is statistically significant (95 per cent CIs: 5.8–20.1). On the other hand, rebel women make 12.5 speeches per year more than rebel men, but this relationship is not statistically significant (95 per cent CIs: −6–30.9). Therefore, after comparing all possible groups based on gender and rebel group affiliation, we find that rebel women make more speeches than non-rebel women, but their behavior is not significantly different compared to rebel men. On the other hand, non-rebel women take the floor less frequently than both rebel women and their male colleagues without rebel experience. Overall, this provides sound support for H1.

To evaluate support for H2, we examine the graphs in Figure 5. The top row of graphs shows simulations based on the count dependent variable (DV) (hard speeches per year), while the bottom row of graphs shows simulations based on the ratio DV (hard to total speeches). Starting with the top-left panel, we see that rebel women are expected to make 12.9 speeches on hard topics per year (95 per cent CIs: 7.2–18.6), while non-rebel women are expected to make 6.9 speeches (95 per cent CIs: 5.4–8.4). This is an increase of fifty-three per cent. In regards to men, non-rebels are expected to make 12 speeches on hard topics (95 per cent CIs: 9.9–14.1), while rebels are expected to make 9.9 speeches (4.7–15.1). The middle graph shows that the change of six speeches between rebel and non-rebel women is statistically significant (95 per cent CIs: 0.27–11.7), but this is not the case for the difference between male MPs. Finally, the top right panel shows that non-rebel men make 5.1 more speeches on hard topics than non-rebel women (95 per cent CIs 2.5–7.8), but there is no statistically significant difference between rebel men and women MPs when it comes to this category of speeches.

We then proceed to the bottom row of graphs in Figure 5, where we investigate the differences in the ratio of speeches on hard topics. The bottom left panel shows the expected ratio of hard speeches by rebel and non-rebel men and women MPs. For non-rebel women MPs, the projection is for 28 per cent of their speeches to be on hard topics (95 per cent CIs: 25.8–30.2). On the other hand, when rebel women MPs take the floor, they are expected to speak about a hard topic 29.2 per cent of the time (95 per cent CIs: 24.8–33.6). In comparison, this ratio is projected to be 30.3 per cent for non-rebel men (95 per cent CIs: 29.2–31.5) and 30.5 per cent for rebel men (95 per cent CIs: 25.5–35.6). In the middle bottom-row panel, we show the expected ratio change for men and women when they have a rebel legacy. We find that the change for rebel women is not statistically significant (95 per cent CIs: −3.2–5.7). Finally, in the bottom-right panel, we compare the ratios of hard speeches made by men and women who either have or lack a rebel legacy. The ratio of hard speeches to total speeches is lower by 1.3 per cent for rebel women compared to their male colleagues, but the relationship is not statistically significant (95 per cent CIs: −7.8–5.2). Similarly, the gap of 2.3 per cent between non-rebel women and men is not statistically significant (95 per cent CIs: −5.2–0.1). Taken together, these simulations provide mixed support for H2. We find that rebel women make more hard speeches than non-rebel women, which is consistent with the fact that they make more speeches overall (H1), but we do not find evidence of an increased focus on hard topics compared to other topics.

We evaluate support for H3 through the simulations presented in Figure 6. The top-level graphs contain simulations from the count model and the bottom-level graphs from the ratio model. Starting from the top-left panel, rebel-women MPs appear considerably more active on topics related to wartime experiences. Specifically, they make 10.7 speeches per year (95 per cent CIs: 5.6–15.8), while their non-rebel counterparts make 4.7 speeches (95 per cent CIs: 3.6–5.8). This is an increase of sixty-four per cent. In regard to men, non-rebels are expected to make 6.3 speeches on these topics (95 per cent CIs: 5.5–7.1), while rebels are expected to make 7.2 speeches (3.2–11.1). The middle panel of the top row indicates that the difference of six speeches between rebel and non-rebel women is statistically significant (95 per cent CIs: 0.9–11.2). Finally, the top right panel shows that there is no statistically significant difference between rebel men and women.

The graphs in the bottom row of Figure 6 show that rebel women also devote a greater share of their speeches to topics related to wartime experiences. These topics make up 27 per cent of the speeches given by rebel women (95 per cent CIs: 22.4–31.5). This number drops to 20.1 per cent for non-rebel women (95 per cent CIs: 18.2–22.1). Rebel men also appear to focus more on these topics as they make up 22.6 per cent of their speeches (95 per cent CIs: 17.4–27.7) compared to 17.7 for non-rebel men (95 per cent CIs: 16.7–18.7). The middle panel of the row indicates that the difference of 6.8 per cent between rebel and non-rebel women is statistically significant (95 per cent CIs: 2.2–11.5), but the difference between their men colleagues is not. Finally, the bottom-right panel shows that there is no statistically significant difference between rebel men and rebel women. Overall, these simulations offer strong support for H3, with women MPs making more speeches on topics related to wartime experiences and also focusing on these topics more than non-rebel women.

To test the robustness of these findings, we run a series of additional tests. First, we restrict the sample to women MPs so that the relationship can be tested linearly (Appendix: Table 4, Models 6–10). We also ran the main models with an alternative operationalization of the dependent variable. Parliamentary speeches vary tremendously in length. It is possible for some MPs to make fewer but longer speeches compared to others but still occupy the floor for long stretches of time. To account for this possibility, we ran our models with the number of spoken words as the DV (Appendix: Table 5, Models 11–15, Figures 1– 3). In both cases, we find support for hypotheses H1 and H3 but not for H2.

Speaking on Soft Topics

While we theorize that rebel women will be more active on ‘hard’ topics and topics related to wartime experience, there is strong evidence in the existing literature of women MPs focusing on ‘soft’ topics (Baumann et al., Reference Baumann, Bäck and Davidsson2018; Clayton, Josefsson et al., Reference Clayton, Josefsson and Wang2017; Eagly and Karau, Reference Eagly and Karau2002; Vet and Devroe, Reference de Vet and Devroe2022). Therefore, it would be possible for rebel women to be more active overall, as we find (H1), but to focus especially on ‘soft’ topics out of genuine interest. To test this possibility, we run our main models with a count DV for soft topics and a ratio DV for the share of soft to total speeches. As ‘soft’ topics, we code civil rights, health, education, community development, and social welfare.

We find that rebel women make more speeches on soft topics, but the degree to which they focus on them is identical to non-rebel women. In addition, we find that rebel women speak significantly more on those issues than rebel men, but they do not make up a significantly higher percentage of their total speeches. However, the finding that stands out the most in part of the analysis is the relationship between non-rebel men and women. While in other simulations non-rebel women speak significantly less than men, they make virtually the same number of speeches with them on soft topics. In addition, speeches on soft topics comprise a significantly higher proportion of their speeches compared to their men colleagues (Appendix: Table 10, Figure 14). These findings corroborate the conventional wisdom regarding women MPs focusing on particular policy domains while avoiding others.

Exploring Government Party Effects

An important limitation of our research design is that the overwhelming majority of rebel women are elected on the ticket of the government party, which is also a rebel party. As a result, an alternative explanation could be that women MPs are empowered due to being part of the government (ruling) party or that, due to their past, rebel parties provide a more egalitarian environment in which women MPs may get more opportunities to take the floor. To test this alternative explanation, we ran models where government membership replaces rebel group membership in the interaction term (Appendix: Table 6, Figures 4–6). The results show that there is no evidence that membership in a ruling party is driving the results.

Background Differences Between Women MPs

Another limitation concerns the possibility that systematic differences in the backgrounds of rebel and non-rebel women MPs affect their behavior in parliament. While women with non-gender-conforming backgrounds and preferences may be more likely to join rebel groups (Azmi, Reference Azmi2015; Thomas and Wood, Reference Thomas and Wood2018; Yadav, Reference Yadav2016) and be active in parliament, a plethora of research at the individual level suggests that women who join these groups commonly have low levels of education and are of a lower socio-economic status (Barth, Reference Barth2002; Byanyima, Reference Byanyima1992; Meintjes et al., Reference Meintjes, Pillay and Turshen2002; Mugabme, Reference Mugabme2000). In either case, it is possible that these women possess personalities or experiences that deem them more likely to challenge gender norms, though the narratives of women’s experiences in rebel groups still stress the changes in socialization they undergo nonetheless. While it is not possible to directly test for pre-war socialization, to address this concern, we utilize the level of education and type of career as proxies.Footnote ^{9 ,} Footnote ¹⁰

Preliminary analysis shows that both variables are associated with our key independent variables. A t-test indicates that non-rebel women tend to have significantly higher levels of education compared to rebel women MPs (p < 0.001). Also, simulations based on a logit regression indicate a statistically significant association between certain career paths and rebel group legacies for women MPs (Appendix: Table 7, Figure 7). Adding these controls leaves our results largely unaffected (Appendix: Table 8, Figures 8–10).Footnote ¹¹

Time in Office and Speechmaking.

The results presented to this point indicate that rebel backgrounds among women are associated with greater speechmaking activity. However, it is not clear when this relationship develops temporally. If, as we suggest, rebel women are more active due to their prior socialization in environments dominated by men, we should expect them to be more active from the beginning of their careers in parliament. Furthermore, if this is indeed a matter of socialization, the literature would suggest that non-rebel women also start participating more as they become more experienced (see, for example, Pfafman and McEwan (Reference Pfafman and McEwan2014)).

To test this, we ran additional models where the number of years that the MP has been in office is added as a third interaction term (Appendix: Table 9, Figures 11–13). We observe that in terms of absolute numbers, the difference between rebel and non-rebel women is statistically significant across the board during their first years in office, but not after year 6. However, the lack of statistical significance in later years is due to a widening of the estimated confidence intervals as opposed to the gap between the means becoming narrower. The exception to this rule is the rate of speeches related to wartime experiences (H3), for which the difference remains statistically significant for all simulated years.