Is women’s empowerment associated with political knowledge and opinions? Evidence from rural Mali

World Development 106 (2018) 299–323

Contents lists available at ScienceDirect

World Development journal homepage: www.elsevier.com/locate/worlddev

Is women’s empowerment associated with political knowledge and opinions? Evidence from rural Mali Jaimie Bleck a,⇑, Kristin Michelitch b a b

University of Notre Dame, United States Vanderbilt University, United States

a r t i c l e

i n f o

Article history:

Keywords: Gender Inequality Empowerment Public opinion Political knowledge Family Code

a b s t r a c t According to many prominent theorists of democracy, citizens must be able to ‘‘formulate and signify preferences” to participate as ‘political equals’ for democracy to work (Dahl, 1971). However, a gender gap in political knowledge and opinions exists across the Global South, especially in rural areas. In this paper, we study the relationship between rural women’s socioeconomic empowerment (household agency and mobility outside the village) and political knowledge and opinions in Mali, a West African country with patriarchal gender norms. To reduce well-known difficulties of gaining access to rural women and reducing bystander effects, we use simultaneous co-gender interviews of one man and one woman per extended family household and a modified Audio-Self-Administered Questionnaire for illiterate populations. Further to reduce ‘‘satisficed” opinions, we elicit opinion justifications and measure ‘‘justified opinions.” Consistent with predictions, we find that women’s empowerment is positively associated with rural women’s political knowledge and opinions. We close by examining opinions towards one controversial policy area with redistributive consequences for men and women—the Family Code, which regulates rights of men and women in marriage, inheritance, and the family. More empowered women are more likely to support pro-woman changes. Ó 2018 Elsevier Ltd. All rights reserved.

Political knowledge and opinions are foundational building blocks for informed and autonomous political participation in a democracy (Dahl, 1971). Troublingly, a systematic gender gap pervades the Global South in levels of political knowledge and opinions, especially in rural areas (Chhibber, 2002; Logan & Bratton, 2006). Given men and women often have different policy priorities and preferences (Chattopadhyay & Duflo, 2004; Gottlieb, Grossman, & Robinson, 2016), such gender gaps in knowledge and opinions may hinder women’s ability to participate and influence the political process in line with their interests. It is therefore crucial to understand the determinants of improved levels of women’s political knowledge and opinions. A large literature emphasizes how women’s degree of political engagement depends critically on gender inequalities in socioeconomic resources that stem from differential access to education, formal employment, and participation in civil society (Burns, Schlozman, & Verba, 2001; Isaksson, 2014; Iversen and Rosenbluth, 2010, 2006, 2008; Kuenzi, 2006; Logan & Bratton, 2006). However, a large gender gap remains even when controlling ⇑ Corresponding author at: The University of Notre Dame, 2130C Jenkins-Nanovic Hall, Notre Dame, IN 46556, United States. E-mail address: [email protected] (J. Bleck). https://doi.org/10.1016/j.worlddev.2018.01.006 0305-750X/Ó 2018 Elsevier Ltd. All rights reserved.

for such socioeconomic resources, especially in the Global South (Coffe & Bolzendahl, 2011; Isaksson, Kotsadam, & Nerman, 2014; Jayachandran, 2015; Logan & Bratton, 2006). Many scholars (e.g., (Clayton, 2015; Gottlieb, 2016; Logan & Bratton, 2006)) suggest that this unexplained gender gap is due, at least in part, to pervasive patriarchal norms governing socioeconomic spheres of life. Indeed, qualitative studies have documented how patriarchal norms curtailing women’s socioeconomic empowerment inhibit women’s ability to develop political knowledge and opinions, and ultimately full and autonomous political participation (Armstrong et al., 1993; Beck, 2003; Callaway, 1984; Tripp, 2012). Of course, lack of women’s autonomous political participation mutually reinforces socioeconomic gender inequalities given that pro-woman policies are less likely to be enacted when women’s voices do not enter politics (Beath, Christia, & Enikolopov, 2013). In this study, we quantitatively examine whether the degree of rural women’s socioeconomic empowerment is positively associated with greater political knowledge and opinion formation in the lead up to the 2012 Malian presidential elections. We argue that such a correlation may exist due to higher levels of decisionmaking, discussion, and pro-woman household norms (‘‘household agency”) as well as travel outside the village. These factors mutually reinforce political efficacy, interest, and skills to learn about

300

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

and engage with the political world around them. In the study area, individuals have little access to schooling, formal employment, and information communication technology. Thus, we examine variation in the degree of women’s household agency and mobility outside the village holding such factors constant when they might ordinarily confound such a study. We conduct an original survey in 10 villages, employing new survey sampling and questionnaire techniques developed with villagers in an out-of-sample village. We employed co-gender enumerators and interviewed one man and one woman per du (extended family household) simultaneously to overcome inhibited access and common bystander effects to surveying female respondents (the latter of which we show in Afrobarometer data) (Lupu & Michelitch, 2018). Further, we adapt (Chauchard, 2013)’s Audio-Self-Administered Questionnaire for illiterate populations to further provide privacy on potentially sensitive questions. We measure political knowledge and opinions regarding three dimensions that were salient for the upcoming presidential elections1: presidential term limits, a local elected official’s identity and performance, and the Family Code (a policy issue with redistributive implications for each sex). To avoid falsely counting ‘‘satisficing”2 as an opinion, we use follow-up prompts in the questionnaire to differentiate between those offering an opinion from those who can offer an opinion and a justification of that opinion. Last, to measure women’s socioeconomic empowerment, we join international development scholars in collecting data on women’s household agency (Kabeer, 1999; Malhotra & Schuler, 2005) as well as mobility outside the village (Chhibber, 2002). The findings largely support the hypotheses. As predicted, women’s household agency and mobility outside the village are both positively associated with rural women’s political knowledge and opinions. We then examine the content of political opinions to see whether socioeconomic empowerment is associated with support for pro-woman changes in the Family Code, a proposed law that would regulate citizen’s rights and responsibilities within a marriage, inheritance, and the family. The changes were very contentious and there was mass mobilization against the code by the religious community in Mali. Changes like those in the Family Code (e.g., minimum woman’s age of marriage), have been hotly debated across Francophone West Africa (Wing, 2009). Encouragingly, we find that more empowered women are more likely to support the Family Code, which includes many pro-woman provisions that would feed back into women’s basic de jure (and de facto) empowerment if adopted and implemented. This paper contributes to existing substantive knowledge of gender and development in two main ways. First, while most studies examine the relationship between women’s socioeconomic empowerment with socioeconomic outcomes (e.g., Furuta & Salway, 2006; Kabeer, 1999; Malhotra & Schuler, 2005), our study joins the few other studies examining a systematic statistical link between women’s socioeconomic empowerment and political outcomes (e.g., Chhibber, 2002; Gottlieb & Robinson, 2016; Girard, 2014; Prillaman, 2016). These latter studies demonstrate how socioeconomic empowerment, access to economic networks, or matrilineal lineage is associated with higher levels of women’s political participation (e.g. candidacy, contacting an official, voter

1

The election was cancelled due to a military coup two months after the survey. ‘‘Satisficing,” is when respondents offer up responses that seem reasonable enough, but are not truthful or complete (Krosnick, Narayan, & Smith, 1996). This technique shows a large disparity in the number of people who report a response to an opinion question (e.g., how well the mayor is performing) and the number who report a response and state any reason at all when asked why (e.g., he does nothing for agriculture). 2

turnout). We complement such research by demonstrating a positive relationship between women’s socioeconomic empowerment and political knowledge and opinions. Women’s level of political knowledge and opinions should not be overlooked because these are the foundational tools women need for autonomous participation rather than simply mobilized participation (Collier, 1982). Indeed, high levels of rural women’s (and men’s) political participation in elections is often mobilized by chiefs or religious leaders looking to trade the maximum number of votes to political parties in return for patronage (Baldwin, 2015; Gottlieb & Larreguy, 2016; Koter, 2016), while otherwise excluding women from politics (Gottlieb, 2016). Further, the government and the international aid community often require women’s participation in political processes in exchange for public services, but such mobilized participation has shown mixed results in affecting women’s empowerment outside the mandated activities (Beath et al., 2013). More research is necessary to understand the conditions under which rural women can participate more and be truly autonomous. Second, many scholars have highlighted the large strides made by women in low-income countries. For example, scholars have investigated the enactment of legislative gender quotas (Tripp & Kang, 2008), the success of women’s movements in influencing policy (e.g., (Adams & Kang, 2007; Bauer, 2004; Kang, 2016)), factors associated with women’s appointment to high-level positions (Arriola & Johnson, 2014; Tripp, 2015), and the impact of women’s descriptive representation on broader participation (Barnes & Burchard, 2013). However, these narratives are by and large focused on urban and/or more elite women. Our study cautions that, amidst celebration of women’s recent successes in the political sphere, we should simultaneously acknowledge that rural women still face significant barriers to entering the political sphere with the basic building blocks of autonomous political participation—political knowledge and opinions. Encouragingly, however, we show that small variation in levels of household agency and mobility outside the village in a context of fairly strict patriarchal gender norms is associated with greater knowledge and opinion formation. Moreover, women with higher levels of empowerment are more likely to support pro-woman changes to formal political law that would act to further empower women in what may be the start of a virtuous cycle of women’s empowerment. Practitioners and scholars should put efforts into further discovery into how to begin and accelerate such a virtuous cycle.

1. Previous scholarship An extensive quantitative literature emphasizes how women’s degree of political engagement writ large depends critically on private sphere barriers, with emphasis on inequalities in socioeconomic resources (e.g., civic skills, money, political interest) that stem from schooling, formal employment, religion, civil society organizations, and family life (Burns et al., 2001; Isaksson, 2014; Iversen and Rosenbluth, 2010, 2006, 2008; Kuenzi, 2006; Logan & Bratton, 2006). However, a large gender gap in political engagement remains even when controlling for available socioeconomic resources, especially in low-income countries characterized by unequal gender norms (Coffe & Bolzendahl, 2011; Isaksson et al., 2014; Logan & Bratton, 2006; Jayachandran, 2015). Scholars (e.g., (Clayton, 2015; Gottlieb, 2016; Logan & Bratton, 2006)) suggest that this unexplained gender gap is due to pervasive patriarchal norms at the village and household level regarding the role of women in socioeconomic spheres of life, which inhibits women’s political engage-

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

ment. Such norms regarding women’s socioeconomic empowerment3 may be particularly powerful in rural areas in the Global South, given that women often live in extended family households in villages with hierarchical decision-making around scarce resources (Duflo, 2012; Jayachandran, 2015; Tripp, 2012). Indeed, qualitative studies (Armstrong et al., 1993; Beck, 2003; Callaway, 1984; Tripp, 2012) have demonstrated how patriarchal norms governing women’s socioeconomic empowerment curtail rural women’s de facto political engagement and participation in new democratic political institutions that de jure empower women to participate on an equal basis (e.g., universal suffrage and candidacy rights). Many underscore limited independent access to information or forums for political discussion. For example, where mass media and information communication technologies are limited, traditional authorities (e.g., village chiefs, religious leaders, du heads, or husbands) can act as ‘‘information brokers” to constrain villagers’ choices based on withholding or coloring information (Pye, 1963).4 For another example, mobility constraints to remain in the household or obligations to farm the family plot may limit opportunity for women to swap or discuss information or opinions (Janssens, 2010).5 Tripp (2012) underscores how informal patriarchal norms amongst local elites, who face few incentives to dilute their own power, limit women’s engagement. Few quantitative studies have examined the relationship between women’s socioeconomic empowerment and political outcomes.6 For example, Chhibber (2002) demonstrates the positive relationship between women’s household empowerment (e.g., ability to leave the home, ability to express themselves at home) and candidacy in India. Further, he shows a positive cross-national relationship between frequency of socializing with friends outside the home and political interest and participation (e.g. signing a petition, demonstrating) using data from the World Values Survey. Also in India, Prillaman (2016) demonstrates that the impact of women’s access to economic networks is positively related to political participation (e.g. contacting officials, attending party campaign events, protesting). Examining a gender quota embedded within an irrigation program in India, Girard (2014) shows that women who participate as decision-makers or laborers also demonstrate greater participation in meetings. Finally, using data from sub Saharan Africa, Gottlieb and Robinson (2016) show a positive relationship

3 Kabeer (1999) describes women’s empowerment as being constituted of agency, resources, and achievements, which enable women to make strategic life choices that they were previously denied. We recognize that women may exert their agency in different ways; they could actively chose not to participate in household decisions or forgo it as a part of a strategy to achieve a different set of goals. For instance, Callaway (1984) describes how Hausa women’s seclusion in purdah allows them to focus on income generating activities rather than household chores (e.g., fetching wood and water). 4 Pye (1963) relates that ‘‘since the communications process was generally so intimately related to the basic structure of the traditional society, the acts of evaluating, interpreting, and responding to all communications were usually strongly colored by considerations directly related to the status relationship between communicator and recipient” (p. 24). In low-information environments, a leader with status and resources to accumulate information can shape the way that information is disseminated to others in the village. 5 In a depiction of women living in Indian villages, Janssens (2010) stresses the damaging effect of social isolation within the household in limiting access to information. She relates that ‘‘the lack of access to knowledge and information combined with the social isolation keeps women in a vicious cycle, excluded from the decision-making processes within the family and within the broader community” (p. 976). 6 Scholars have extensively documented a strong positive statistical relationship between socioeconomic empowerment and socioeconomic outcomes (Furuta & Salway, 2006; Malhotra & Schuler, 2005; Mahmud, Shah, & Becker, 2012; Friedberg & Webb, 2006; Wiig, 2013; Kabeer, 2001).

301

between membership in a matrilineal (versus patrilineal) ethnic group and higher levels of political engagement (e.g., interest, discussion, and understanding of politics) and political participation (e.g. turnout, contacting an official, demonstrating, rally attendance, persuading others’ vote, campaigning) using Afrobarometer data.7 Of course, the positive association between women’s socioeconomic empowerment and political participation is likely mutually reinforcing. The present study complements these studies by empirically investigating whether a higher level of socioeconomic empowerment (household agency and mobility outside the village) are associated with greater political knowledge and opinion formation. Without political knowledge and opinions, citizens may not be able to fully ‘‘formulate and signify preferences” and to participate as ‘political equals’ in the political process (Dahl, 1971). Speaking directly about women, Carpini and Keeter (2005) refer to the notion of ‘‘enlightened political engagement” as an ideal in which women have independent opinions upon which to base their participation; this enables them to influence the political process to achieve outcomes in line with their interests. Given that women and men often have different political preferences (Chattopadhyay & Duflo, 2004; Gottlieb et al., 2016), it is imperative that women’s voices be able to enter the political process not only for intrinsic normative goals of democracy, but also to ensure that vital policy-making outcomes optimally reflect the citizenry as a whole. Especially in environments with low levels of human development and new democratic institutions, many scholars underscore that it is imperative that individuals gain the tools of political knowledge and opinions (O’Donnell, 2004; Sen, 1999). Such tools are particularly important in environments with elevated levels of citizen mobilization by electoral brokers. Collier (1982) makes an important distinction between mobilized participation, which ‘‘reflects the ability of elites to persuade the masses” and autonomous participation, which ‘‘reflects the interest and more spontaneously derived opinions of the participants” (p. 14). She argues that mobilized participation is a better measure of the capacity of political entrepreneurs to organize and influence constituencies rather than the articulation of certain voters’ interests and preferences. Problematically, much of women’s political participation is mobilized rather than autonomous. Perhaps one of the most salient areas of mobilized women’s participation by domestic elites is in voter turnout. A vast literature describes the role of traditional authorities, such as village chiefs, in coordinating voter turnout in rural areas in order to receive patronage from political candidates or parties (e.g., Gottlieb & Larreguy, 2016; Koter, 2016).8 Given that women’s and men’s votes count equally in elections, both women and men are mobilized in great numbers in rural areas for the purposes of block voting, even if such citizens have little political knowledge or opinions regarding the election. Fig. 1 below uses Afrobarometer data to show high reported turnout of over 70% for the last presidential election with a small gender gap within African countries (see Fig. 1 left panel). In contrast, there are stark gender differences in knowledge of presidential term limits across most countries—one of the most salient pieces of political knowledge in

7 Other scholars focus on the effect of political empowerment through quotas or mandated women’s participation on policy outcomes (e.g., Beath et al., 2013). 8 Koter (2016) describes this process as transforming voting ‘‘from an individual act in which voters make decisions independently to collective voting, or bloc voting, when voters follow the advice of their leader” (p. 132). While some have argued that villagers strategically vote with the chief to secure development goods for the village (Baldwin, 2015), others argue that such coordination can be also be coercive or exploit villagers’ deference to their traditional positions of authority (Beck, 2001, 2008; Lemarchand & Legg, 1972).

302

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

Fig. 1. Male/Female and Urban/Rural cleavages in voter turnout and political knowledge. Notes: Afrobarometer Data Round 3. Left Panel: With regard to the most recent, [20XX] national elections, which statement is true for you? Answers: 1 = voted, 0 = did not vote for any listed reason. Right Panel: Do you happen to know: How many times someone can legally be elected President/Prime Minister? Answers: 1 = correct, 0 = incorrect.

this context (Fig. 1 right panel): only 31% of rural women compared to 50% of rural men know term limits.9 Members of the international aid community often mandate women’s participation, in order for communities to receive development funds. For example, Beath et al. (2013) find that the introduction of gender-balanced village development councils in rural Afghanistan have mixed results, with positive effects of the program ‘‘limited to areas that are closely linked to direct interventions prescribed by the program” (p. 555). Even worse, some scholars have found that interventions to increase women’s participation have had adverse effects on women’s participation (Gottlieb, 2016). When social norms against women’s socioeconomic empowerment are so stark, it can be difficult if not impossible for the infusion of new and more egalitarian institutions to fully enable women to participate autonomously. While such interventions may (hopefully) lead to improved levels of women’s autonomous participation in the long-run, scholars should generally be careful not to equate such mobilized participation as autonomous participation moving forward. In sum, understanding the conditions under which women in rural contexts across the Global South are able to develop the building blocks of autonomous political participation—political knowledge and opinions – remains poorly understood. In this study, we examine whether small, incremental increases in socioeconomic empowerment are associated with holding political knowledge and opinions in rural Mali. We first describe the study area context and motivate our hypotheses linking socioeconomic empowerment and political knowledge and opinions, before turning to research design and results. 2. Study area: rural Mali Mali offers fertile ground to examine the building blocks of autonomous political participation amongst women in the rural areas of low-income countries. In Mali, women achieved universal 9 We use Round 3 because it is the last round containing knowledge questions. Estimating a logistic regression (see Appendix Table 3), we confirm that men and rural villagers are significantly more likely to vote by 6 percentage points respectively, versus women and urban counterparts. Men and urbanites are significantly more likely to know term limits than women and rural villagers, by 17 percentage points and 14 percentage points respectively. No conditional effects of gender and rural versus urban dwelling exist.

suffrage upon independence from France in 1960. After initial elections and an authoritarian turn for the subsequent two decades, Mali transitioned to multi-party democracy in 1992. Mali embarked on a decentralization campaign in the early 1990s, which created 702 municipalities and elected mayors. Mali had held four rounds of presidential and legislative elections and three rounds of municipal elections at the time of research in January 2012. Despite the introduction of elected officials at the local level, traditional authorities such as village chiefs still hold tremendous power and respect (Coulibaly & Diarra, 2004; Logan, 2013). Village chiefs obtain power via heredity and, due to the weakness of the state outside of the urban centers, end up wielding power over most areas of life such as taxation, redistribution, dispute arbitration, and inheritance law (Bleck & Michelitch, 2015). In 2014, Mali was ranked 150 of 180 countries listed on the Gender Inequality Index with a score of 0.677 (UN Human Development Report, 2015). Women in Mali often are subject to decision-making by their husbands. According to the 2006 Demographic and Health Survey, 72 percent of Malian women reported that their husbands make decisions concerning their health for them (Lackenbauer, Lindell, & Ingerstad, 2015). Relatedly, Mali exhibits one of the largest gender gaps in politics in sub Saharan Africa. At the citizen level, the gender gap for an index of participation in Mali is double that of the Afrobarometer mean (Logan & Bratton, 2006, p. 17). Some Malian women cite patriarchal norms as inhibiting their individual-level political participation or engagement with public sphere (Bleck, 2015, p,. 115). Most ethnolinguistic groups in Mali practice patrilocality, where women live in their husband’s extended family dwelling upon marriage. The extended family, or (du in Bamana and galê in Fulani) is comprised of a grandfather and grandmother, their sons and their son’s wives, their son’s sons and their wives as well as any unmarried women and girls. They may also harbor other relatives including adopted children and youth who are attending school in the area. Furthermore, polygamy is legal in Mali and allows men to take up to four wives. The present study takes place in rural Northern Mali, where women face constraints due to limited mobility, time and access to the public sphere. Women are responsible for the majority of the household workload, being primarily responsible for agricultural labor, food preparation, washing, transporting water to the house, childcare, and gathering fuel (Wooten, 2003). In these villages, men are typically seen as the household representatives to

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

the public sphere, and the primary decision-makers. Women vary in the degree to which they may take part in decision-making and household discussion. Domestic violence is relatively widespread; women who seek to think or act autonomously in the household or in politics may face physical punishment or other types of sanctions. While women are less mobile than men, there is variation in terms of freedom to leave the village to go to nearby towns, usually to visit a market. The area of study has little variation in access to education and limited opportunities for formal employment in the area. Adult villagers had no access to public school in these villages during their childhood. Most families were involved in agriculture, animal husbandry, or fishing if proximate to the Niger river. The information communication technology environment is very limited. The study area villages receive did not receive consistent cell phone reception. Access to national public radio was newly introduced through newly built emissions towers, but only the very few elites could afford a radio and batteries. To demonstrate the isolation of these villages, we asked respondents if they had ever heard of Michael Jackson (arguably one of the most famous public figures in Mali) and if so, if they knew that he had died. Given how popular Michael Jackson is, the news of his death in 2009 was very prominent in the media in the capital city of Bamako by the time of the survey in early 2012. In the village 90% of women knew who Michael Jackson was but not that he died, while only 4% of women didn’t know who he was and 6% knew who he was and that he had died (see Appendix B.1). These data demonstrate how slowly general current events news travel to women in these zones. In fact, women can often have quite limited independent access to information in the study area. Interviews with chiefs in the villages in our sample study revealed that women are often excluded from informational meetings in the village. For example, as one chief said, ‘‘Women are not authorized to participate in our meeting because women have nothing to do with political affairs. Concerning women, we think that every man present at the meeting has the duty to inform his wife. That is why women are not generally present at our meetings.” 10 However, despite the conservative context, there is variation in the degree that women in these villages have agency in their households and their mobility outside the village. We examine the relationship between these forms of small degrees of socioeconomic empowerment and women’s political knowledge and opinion formation. Because of limited education, employment, and access to mass media, we can examine the link between socioeconomic empowerment and political knowledge and opinions free of such confounds.

3. Socioeconomic empowerment, political knowledge, and justified political opinions We consider how patriarchal gender norms stymy women’s political knowledge and their ability and willingness to formulate opinions, and how socioeconomic empowerment in the private sphere may covary with burgeoning levels of political knowledge and opinions. We emphasize that the study area is perhaps one of the least developed and more conservative areas in subSaharan Africa with regards to gender norms. Thus, we are considering the variation in what may appear to be minute degrees of empowerment, but which are quite powerful for women in this context. We consider household agency and mobility outside the village, in turn.

10

Chief, Village 8, February 2012.

303

3.1. Household agency When women have burgeoning levels of agency in household decision-making and discussion, we anticipate that they may also be more likely to hold political knowledge and opinions due to increased efficacy, interest, and skills. First, when women have agency in their households, they may feel capable of affecting change, and gain greater levels internal efficacy—defined as a stronger belief in one’s competence to understand and participate in an activity (Niemi, Craig, & Mattei, 1991).11 By learning how one can affect change in the socioeconomic arena, such internal efficacy can translate and extend into efficacy in broader spheres. Second, women may be conditioned from childhood to be interested in traditional ‘‘women’s issues” such as small child care, not ‘‘men’s issues” such as household decision-making or politics. If women are responsible for making more household decisions, however, they might start to take interest in a broader class of phenomena in their environment including political phenomena. For example, when women have decision-making power over children’s education, what to buy in the market, or what business or agriculture to invest in, they may become interested in government education policy, markets, and agricultural extension services that affect such household issues. Interest in the political sphere could lead to interest in gaining greater political knowledge and formulating opinions about change in the political sphere. Last, women governed by patriarchal gender norms in the household may have developed few skills that aid in information search, processing, evaluation, and articulating their own ideas as related to their agency in the household. When women have expanded opportunities for decision-making and participate in household discussions more regularly, they may gain new skills that are also relevant for the political world. For example, in household decision-making, women may gain resources such as ability to search for information, evaluate and discern knowledge, as well as ability to articulate ideas, argumentation/reasoning, negotiating family members’ interests, and general problem-solving. Taken together, we postulate that a higher degree of household agency is positively associated with (a) political knowledge and (b) political opinions.

3.2. Mobility to leave the village Women have less mobility both within their village as well as outside their village. Within the village, women have fewer opportunities to directly access political information or forums for political discussion outside of their compound house. One primary example is the village mosques, where men gather and can exchange views on village matters or politics. Women, in contrast, are more likely to pray from home.12 Additionally, elder men in these villages have common space available where they regularly sit together to drink tea, listen to the radio, and discuss village affairs. While women have no such equivalent spaces, women can, to a limited extent, encounter information in the market or at the river (fetching water or washing with other women).13 Women thus have limited and irregular forums for information exchange and discussion within the village, as well as limited access to radio (given its cost). When women receive information, it may have been ‘‘filtered” by male household members or village leadership.14 11

Prillaman (2016) refers to this process as ‘‘gains in confidence.” Only 3.3% of women report exchanging information at the mosque (see Appendix B.1). 13 Indeed, 24% of women exchange information in the market and 27% exchange information at the river (see Appendix B.1). 14 The most of any category, 29% of women exchange information in the extended family household (see Appendix B.1). 12

304

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

Many women’s mobility is limited to their village. This phenomenon is in large part due to gender norms where dus control women from straying too far outside the extended family and husband’s purview. However, some women leave the village to visit relatives or to sell goods at the nearby town market. Towns, especially markets and transportation hubs are also well-known forums of informational exchange. People receive news from the regional capital from long-distance travelers. Additionally, market towns have greater access to cellular and radio signals. Mobility outside the village facilitates interactions with a broader network of people and access to information, and thus, a broader set of experiences with which to situate information and to draw on when they form opinions. Women who are able to leave the village to nearby towns can gain exposure to new ideas and information independent of hierarchical information dissemination the may exist in the village. Such exposure may increase women’s internal efficacy, political interest, or skills leading women to collect political knowledge and formulate opinions. Taken together, we postulate that a higher degree of mobility outside the village is positively associated with (a) political knowledge and (b) political opinions.

4. Research design 4.1. Survey sampling and enumeration We conducted fieldwork followed by survey research in ten mixed majority Bamana and Fulani villages in Northern Mopti in early 2012, following fieldwork in two ‘‘out-of-sample” villages. In the fall of 2011, the team leaders visited the villages twice to pay customary respects to village chiefs, build rapport, and gain permission for a research team to return to conduct research. From January to February 2012 the research team visited the villages to enumerate a survey. It is challenging to incorporate female voices in low-income, rural contexts (Lupu & Michelitch, 2018). First, at the sampling stage, male family members may resist a woman ‘‘representing the family” by being selected for the opportunity to participate in the survey. Stating opinions on public affairs is seen as the traditional roles of the (male) head of the extended household (dutigi in Bamana) or his sons. Further, women might be disallowed by male family members to speak openly with a male enumerator (a stranger). Second, when a woman does participate in a survey, it may be difficult to dissuade male family members from hovering around and attempting to influence a woman’s survey responses. Indeed, Afrobarometer data in Fig. 2 (left panel) show the percentage of interviews that were influenced by bystanders, as reported after each interview by enumerators. Dots represent women, while triangles represent men, and lighter versions of the symbols represent rural citizens, while darker versions represent urban citizens. In most countries, women are influenced by bystanders more often than men, and rural villagers more than urbanites. Illustratively, the highest percentage of influence occurs amongst rural Malian women. Estimating a logistic regression in the pooled sample (see Appendix Table 4), we confirm that men are significantly less likely to be influenced than women (2 percentage points) and urbanites significantly less than rural villagers (1 percentage point).15

15 The predicted probability that women are influenced by bystanders is 6 percentage points and men 4 percentage points. The predicted probability that rural villagers are influenced by bystanders is 5 percentage points and urbanites 4 percentage points. The relationship between gender and bystander effects is not conditional on urban/rural status (see Model (2)).

In order to gain women’s participation, we undertook two main initiatives based on focus groups in the two out-of-sample villages. First, after randomly sampling extended households, (du),16 in a public lottery, we randomly sampled one man and one woman from each du—rather than the standard procedure to sample one person per du and alternate the respondent’s gender. The sampled male and female were not restricted to be husband-wife pairs, but rather, could be a mother and a son, siblings, uncle and niece, etc. Second, we invested in recruiting and training female enumerators, many of whom had little to no previous survey experience.17 A team comprised of one male and one female enumerator interviewed respondents in co-gender interviews separately and simultaneously. In order to participate in the study, the head of the du needed to consent to the following sampling procedure or they would be ineligible and resampled. While sampling often took some discussion to get everyone on board, no dus opted out. We found that these methods, while increasing survey costs, ultimately improved access to female respondents. Commonly, such costs deter researchers from including female respondents (Chauchard, 2013). To further reduce bystander effects for sensitive questions, we incorporated a modified Audio-Self-Administered Questionnaire (Chauchard, 2013), where respondents listen to questions recorded on mp3s using headphones and mark their answer privately and silently on a device or paper. Such a method can be used to prevent bystanders within earshot from learning the questions or answers, and give the respondent a feeling of privacy and anonymity to respond truthfully. We used this technique for questions regarding household empowerment, which is not controversial, but perhaps subject to bystander effects and thus more sensitive. Because the population is mostly illiterate, we used pictures to illustrate agreement or disagreement with recorded statements. Last, another challenge to eliciting high quality survey data from this population is widespread ‘‘satisficing,” when respondents offer up responses that seem reasonable enough, but are not truthful or complete (Krosnick et al., 1996). Examples include frequent ‘‘I don’t know” responses, socially desirable responses, systematically biased responses (e.g. disproportionately first choices—‘‘pri macy”), or repeated acquiescence (e.g. agreeing with any assertion, regardless of content). Indeed, in the context of rural areas in lowincome countries, a standard survey environment can seem bewildering (Lupu & Michelitch, 2018). Individuals are not used to answering a ‘‘rapid fire” battery of political questions, much less in an environment where one after another question is asked in a close-ended format with no discussion or follow up. In the context of rural Mali, conversation partners typically acknowledge one another’s ideas by repeating them and prompting for more, allowing individuals’ opinions to emerge gradually. Indeed, the Afrobarometer data, which are collected in a more or less standard survey environment, show that many respondents have difficulty responding to survey questions. In Fig. 2 (right panel), we show the percentage of respondents that survey enumerators designated as having difficulty answering some, most, or all questions (rather than no difficulty or difficulty on a few questions). Indeed, rural respondents appear to have more difficulty than urban respondents, and women more than men. Estimating a logistic regression (see Appendix Table 4), we confirm

16 We subsequently refer to this extended household unit using the local Bamana term du. 17 We also had both Bamana and Fulani enumerators, such that respondents were interviewed by coethnics. Coethnic enumeration has been generally found to enhance survey data quality (Adida, Ferree, Posner, & Robinson, 2016). We spent time training enumerators on how to pleasantly but firmly remove any bystander presence or influence, should it occur.

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

305

Fig. 2. Male/Female and Urban/Rural cleavages in bystander influence and response difficulty. Notes: Afrobarometer Data Round 3. Left Panel: Do you think that anyone influenced the respondent’s answers during the interview? Answers: 0 = no, 1 = yes. Right Panel: What proportion of the questions do you feel the respondent had difficulty answering? Answers: 0 = none, 1 = few, 2 = some, 3 = most, 4 = all. Answers recoded as 0 = none or few, 1 = at least some (some, most, all).

that men and urbanites are significantly less likely to have difficulty (by 5 and 4 percentage points respectively).18 We therefore designed the survey questionnaire to elicit potentially higher quality data. First, we trained enumerators to ask questions multiple times and incorporated prompts to do so into the survey software. Doing so ensured, for example, that there was ample opportunity to understand whether a respondent had a response to a knowledge question, such as whether he or she knew presidential term limits. Second, for opinion questions, enumerators used follow-up prompts (e.g., ‘‘why do you think so?” and ‘‘any more reasons?”) and recorded justifications for the responses. The reason is that it is easy for respondents to satisfice by saying they like or don’t like, for example, term limits. However, it could be that the person does not know what term limits are, or had never considered whether he or she liked them prior to the survey.19 If, on the other hand, a respondent could state any reason for that opinion, then it indicates that the respondent had some justification and was not simply satisficing. 4.2. Measurement of key variables, descriptive statistics, and empirical strategy In this subsection, we explain how we measured key variables, provide descriptive statistics, and describe our empirical strategy. To generate the key variables, we conducted qualitative research and piloting in our out-of-sample village. We report descriptive statistics for key variables for women in Fig. 3, and list also their male family members for the dependent variables. Descriptive statistics on all other variables are listed in Appendix B.1. First, for outcome variables, we measure political knowledge and political opinions. We sought to examine political knowledge and opinions in three domains: institutions, elected official performance, and an ongoing policy debate relevant to gender. We first describe the political knowledge questions. First, we ask a question about political institutions, namely, the aspect of political institu18 The predicted probability that women have difficulty is 12 percentage points and men 7 percentage points. The predicted probability that rural villagers have difficulty is 11 percentage points and urbanites 7 percentage points. The relationship between gender difficulty answering questions is not conditional on urban/rural status (see in Model (4)). 19 Existing literature from a Western context shows that men are more likely to guess than women (Mondak & Anderson, 2004). We anticipate that this disparity in guessing might be even greater in environments where women are more vulnerable.

tions that respondents are most likely to know: can you name term limits for the president? Second, we ask a question about the performance of an elected official, in particular, an elected official that rural populations are most likely to know: can you name your current mayor?20 Last, we ask about a salient current events topic, reform to the Family Code, which regulates aspects of marriage and inheritance that has redistributive consequences for men and women: can you name any aspect of the Family Code or suggested reforms to the Family Code under debate? The first two questions came out of the Afrobarometer battery, but we wanted to make sure to include the Family Code question as research in the US context has found that these gender gaps around political knowledge cease to exist when women are asked about topics that are most relevant to them (Carpini & Keeter, 2005). Dolan (2011) emphasizes the importance of measuring ‘‘domain specific knowledge” as most typical political knowledge items ask about mostly male leaders as well as functions of government (2011: 99). For instance, women tend to perform comparatively well on questions about female leaders (Dolan, 2011) or local government (Carpini & Keeter, 2005, p. 33) – areas that they may find most relevant. The answers were coded as correct or incorrect (which also includes I don’t know answers) and summed to form an index. Fig. 3 (top right panel) shows that few women, only 9%, had a correct response to all three knowledge questions. However, 20% of women knew two of the knowledge questions, and 36% knew one with a remaining 35% knowing none. Women are significantly less likely to have political knowledge than men (t-test, p < 0:01) by almost a full category on average. For political opinions, we followed up after each of the three knowledge questions asking what is your opinion of [the presidential term limits, the current mayor, the Family Code]? These questions were followed up with why? prompts as described in the previous subsection. Enumerators coded whether or not the respondent was able to provide a response to this question (what we call a ‘‘myopic” opinion) and recorded any justifications for the response. To form a ‘‘justified” opinion variable, we coded respondents who did not provide a justification or who said ‘‘I don’t know” as 0, while simple restatements of an opinion were coded as 1,

20 In this context, a mayor indicates an elected official representing a local constituency. As explained above, the process of decentralization created mayors’ offices for all of Mali’s 702 communes. The process has succeeded in bringing political visibility to this position.

306

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

Fig. 3. Descriptive statistics.

responses with simple concrete reasons were coded as 2, and sophisticated concrete justifications were coded as 3. The responses were summed to create an index.21 Although we ultimately use our coding of an ‘‘justified” opinion in our analysis, it is informative to display the descriptive statistics on both the ‘‘myopic” opinion index and the ‘‘justified” opinion index. Turning to Fig. 3 (middle right panel), there is good variation in the number of ‘‘myopic” opinions women have out of three, with 20% stating responses to all items, 35% to two items, 26% to three

21 A list of the responses and the codings are available in the replication material. The three opinion variables have a Cronbach’s alpha of 0.64.

and 20% to none. Women are statistically significantly less likely to offer up responses to an opinion question than men (t-test p < 0:01) by a shade less than a full category on average. Interestingly, 54% of men stated opinions to all three items compared to 20% of women. However, there are far fewer individuals who hold justified opinions. For example looking at Fig. 3 (middle right panel), the number of women who had no response to any of the three questions was 20%, but when adding to that number women who responded but didn’t have any reason for that opinion, the number grows to 31% who do not have a justified opinion. Men are still much more likely to have an justified opinion versus women on average (t-test p < 0:01) by two and a half categories.

307

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

To measure key explanatory variables, we ask about how often the respondent leaves the village, and a battery of questions related to empowerment within the household. For mobility outside the village, we asked: How many times do you travel outside of the village each month? We coded 0 = never, 1 = 1 time, 2 = 2–3 times, 3 = weekly, and 4 = multiple times per week. As Fig. 3 (top left panel) shows, 14% of women never leave the village, 68% leave 1 time per month, while 11% leave 2–3 times per month and a small 7% leave weekly. No women left more than weekly. For the battery of household agency questions, respondents used the audio self administered questionnaire described in the previous section. We recognize the importance of capturing context-specific measures of decision-making (Kabeer, 2001; Malhotra & Schuler, 2005), thus, we adapt common indicators used in other studies (Mahmud et al., 2012), such as control of resources and role in decision-making to the Malian context. See Appendix D for questionnaire script. The first set of questions asked for each of the following situations: In your (plural you), marriages, do husbands and wife discuss: (1) choices about sending a child to school, (2) choosing a spouse for a son or daughter, (3) if the husband plans to take an additional wife, and (4) financial management in the household. The second set asked for each of the following situations: (5) if a woman earns money for herself, can she decide how to spend it, (6) do husbands and wives talk about whether husbands should bring home groceries for his wife (7) do husbands and wives discuss whether to use family planning, (8) do the women in the family discuss together if a female family members wants to start a business? The last questions asked about lingering patriarchal norms: (9) Do you think that husbands should be able to use beatings to discipline his wife? (10) Are the best parts of meals reserved for certain people (men) in your household? To construct the household empowerment index, women were given a point for each time that she expressed having a say in a decision or participating in discussion. For the final two questions, women’s resistance to domestic violence or reports of equal food distribution, were awarded a point. Answers to this battery of questions qualifies well for an index. Together, these variables receive a Cronbach’s alpha score of 0.8, qualifying them as sufficiently related to form an index. Fig. 3 (top right panel) depicts the distribution of the index. The majority of women have zero to two counts of decision-making power with 28% of all women having no empowerment in the household, at least according to the measure. We recognize that our battery of survey questions does not tap into all possible life dimensions upon which women have decision-making power. We emphasize that this study is correlational and associative rather than causal, given that we do not identify exogenous variation in household agency nor mobility in the study, for example, with an experimental intervention. In correlational studies, including control variables may improve statistical efficiency and allow explanatory variables to be better isolated (see Appendix B.1 for descriptive statistics of control variables). One such variable is ethnicity, as different ethnic groups often have different social norms. We create a Bamana indicator variable from the question: What is your ethnic group?22 Thirty-nine percent of women are Bamana. While public school was not available in the village when respondents were school-aged in these villages, they may have been sent to live with relatives in nearby towns to receive schooling. However, it is notoriously difficult to measure amount of education given that people often might go in and out of schooling. We thus control for education through french language ability, an indicator if someone had gone through school to become fluent in the language used in school. We create an indicator variable from the question: Do you 22 We code anyone with mixed Bamana parentage as Bamana given that this is the market language in most of Central Mali and speaking the language may facilitate access to information.

Table 1 Associations between socioeconomic empowerment and political engagement.

Leave Village HH Agency Index Wealth Proxy Understand French News Bamana Village FE Observations

(1) Knowledge Index OLS

(2) Opinion Index OLS

0.71⁄⁄⁄ (0.054) 0.06⁄⁄⁄ (0.013) 0.20⁄ (0.095) 0.78⁄⁄ (0.279) 0.03 (0.086)

0.78⁄⁄ (0.309) 0.13⁄⁄ (0.050) 0.25 (0.283) 0.15 (0.308) 0.00 (0.414)

Yes 281

Yes 226

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

speak French well enough to understand the news on the radio or someone who is reporting the news in person? Only 1.3% of women, however, qualify. We also include a proxy for wealth. In these villages, one of the biggest wealth indicators is how families manage healthcare, derived from our out of sample village as a good proxy for wealth in a context where people do not have regular cash income. We asked: Think about the last time you were sick, how did you take care of the illness? If respondents went to a health clinic, it was coded as 1, while other ways to deal with sickness were coded as 0 (consulting a traditional healer, asking a friend or family member to prepare medicine for you, nothing, called someone on a phone). About 58% of women went to a health clinic.

5. Results Turning to the results, we examine the association between of mobility outside the village and the household agency index and two types of political engagement amongst rural women: (1) political knowledge, and (2) political opinions. We restrict the sample to women only, since our hypotheses regard women - the survey was not designed to understand the very different processes that govern men’s political engagement. We estimate ordinary least squares regressions with the knowledge and opinion index. In all models we include the aforementioned control variables and village fixed effects. We cluster standard errors at the village level. Estimation results are listed in Table 1 and graphical predicted values can be examined in Fig. 4. Overall, we see a general pattern in Table 1 supporting our hypotheses. First, examining Model 1 on the knowledge index, we see that both leaving the village and achieving overall higher empowerment in the household are positively associated with a higher degree of political knowledge on the 0 to 3 scale, which are statistically significant effects with and without controls ðp < 0:01Þ. Changing from minimum mobility (never leaving the village) to maximum mobility (leaving more than once per week) is associated with a large substantive change in predicted value on the knowledge index: from 0.3 to 2.4. The Household Agency Index was also strongly associated with political knowledge. Moving from the lowest to the highest level of household empowerment (0 to 10) on this scale is associated with a change in predicted value on the knowledge index: from 0.9 to 1.5. Turning to Model 2 on the opinion index, we also see that mobility and the household agency index are significantly associated with holding political opinions on a 0 to 9 scale, with and without controls ðp < 0:05Þ. Changing from minimum to maximum mobility is associated with a large substantive change in pre-

308

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

Fig. 4. Graphical relationships between socioeconomic empowerment and political engagement. Notes: Predicted values.

dicted value on the opinion index: from 2.1 to 4.4. The Household Agency Index was also strongly associated with political opinion holding. Moving from the lowest to the highest level of household empowerment (0 to 10) on this scale is associated with a change in predicted value on the opinion index: from 2.6 to 3.9. Taking these results in hand, we extend the analysis in a few different ways.23 First, we check that the results are robust to nonclustered standard errors (see Appendix B.2). Second, one might question how related mobility and household agency are, and whether they might condition one another’s relationship with political engagement (see Appendix B.3). We find that mobility to leave the village and household agency are only weakly correlated at 0.16, suggesting that they capture different constructs. Combining the two measures into a single index strengthens results. Interacting the two variables shows that the two variables do not moderate one another’s relationship with political engagement. Third, one might suspect heterogeneous effects based on ethnic identity (see Appendix B.4). Fourth, while one might hypothesize that less mobile women in villages with overall higher average women’s mobility might be more informed or opinionated than such women in villages with overall lower average women’s mobility, we do not find evidence that this is the case (see Appendix B.5), perhaps because there is very little variation in average women’s mobility across villages (see Appendix B.1). Fifth, we examine whether we can detect any other demographic characteristics of women that might be associated with political engagement and/or socioeconomic empowerment (see Appendix B.1 for descriptive statistics). We examine civil status (see Appendix B.6), whereby 80% of women are first wives, 15% 23

We thank anonymous reviewers for these suggestions.

are second, third or fourth wives, and 5% are unmarried—single, divorced, or widowed. We find that being unmarried is positively associated with having more knowledge (but no relationship with opinions) and less household agency, while first wives were less likely to leave the village than other women. These results may reflect differences in civil status that reflect one’s role in the family – first wives may take more responsibilities to direct women’s work in the household, and therefore travel less. Given we have few unmarried women, we hesitate to make very strong claims about the role of this category, but generally unmarried women are somewhat marginalized within the household decision-making yet might have more free time for politics than women with husbands. We also examine the role of economic activities (see Appendix B.7). Women could name multiple economic activities in which they engage in addition to their household tasks, such as farming or fishing (29%) and small scale enterprise, such as hand crafts (48%). Participating in agricultural activities such as farming and fishing is associated with being less likely to hold political opinions, but not political knowledge, and has no relationship with mobility or household agency. Individuals engaged in such agriculture may perhaps be more isolated from others. Engaging in small scale enterprise was not associated with political engagement or mobility, but is positively associated with household agency. While we saw that education (proxied by French ability – only 1.3% of women could do so) was positively associated with political knowledge, but not opinions, we also examine the relationship of education with mobility and household agency, finding a strong positive relationship with the latter, as one might expect (see Appendix B.8). Sixth, we examined the role of spaces in which women engaged in information exchange in the village (see Appendix B.10). The

309

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

most frequent spaces where women cited that they exchanged information are the family at 29%, the river at 27%, the market at 24%, and the mosque at 3%. Interestingly, women who get information at the market, which is a more male-dominated public space, are more likely to hold political knowledge24 and opinions, but those women exchanging information at the river, a women-only space due to women’s traditional duty of fetching water and washing, are less likely. Exchanging information in the market is also positively associated with household agency. There is no relationship for those exchanging information in the family. In sum, it seems that discussions at the river, often considered a place where women can more freely talk, may be a forum for social rather than political exchange. The market on the other hand, which is traditionally a male dominated space, represents a place where women may more likely encounter politics. Given that the relationship between household agency and mobility are still strong and significant when including these variables, we can say that information exchange at the market is an additional factor that plays a role in women’s political engagement. Further, the non-finding related to the family space tells us that women are not simply getting politically informed or opinionated through discussion with family members. The private family sphere may not be seen as an appropriate forum for ‘‘public political discourse,” at least for women. The practice of patrilocality may play a role here, because married women live with husband’s families rather than their family of origin, the former of which may temper engagement in political discussion (as in Gottlieb & Robinson (2016)). Last, we examine whether there are any relationships between male family members’ level of knowledge or opinions, or other characteristics, and that of the female respondents. One might surmise that women who are empowered or mobile may simply come from houses where men are more politically knowledgable and opinionated, or more liberal regarding women’s status. We report descriptive statistics for men in Appendix C.1. We do not find evidence that male family members’ employment, civil status, wealth or mobility are associated with female family members’ mobility or household agency, with the exception that the male family members’ wealth proxy (getting treated for health problems in a clinic) is associated with women’s household agency.25 We note that mobility and the Household Agency Index remain significant with the inclusion of these variables. Interestingly, we do not find that male family members’ knowledge or opinion levels are associated with women’s knowledge or opinion levels, respectively. This additional analysis comes with the caveat that the male family members surveyed could be an uncle, brother, son, husband, du patriarch, son-in-law, etc., and there may be husband effects that are not picked up in the data. However, it may also reflect a reality that women may not be getting politically engaged through the family other than through family norms for women to have agency and mobility and that men and women have separate processes through which political engagement occurs. While these results shed light on the relationship between women’s political engagement and socioeconomic empowerment, we recognize certain limitations of the analysis. First, the analysis is correlational, not causal, since we have not identified exogenous variation in socioeconomic empowerment or political engagement, for example, through an experimental intervention. Political knowledge or opinions may affect socioeconomic empowerment and they may be mutually reinforcing. Second, the sample of villages is not large and the villages fairly isolated in a very conservative area (i.e., lacking opportunities for education, formal employment, and mass media). Due to the costs of gaining access 24 Exchanging information at the mosque is also positively associated with knowledge, but there are very few women who exchange information at the mosque. 25 We did not ask men about the agency men had in the household.

Table 2 Associations between socioeconomic empowerment and support for the Family Code. Support for Family Code Leave Village HH Agency Index Wealth Proxy Understand French News Bamana Village FE Observations

0.35⁄⁄ (0.172) 0.10⁄⁄ (0.048) 0.80⁄⁄⁄ (0.267) 0.50 (1.199) 0.35 (0.349) Yes 281

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

to women and eliciting high quality data discussed in the research design section, the budget only allowed for 10 villages of respondents.26 One might find that at higher levels of access to education, formal employment, and the mass media, the impact of socioeconomic empowerment may be different, perhaps smaller or larger. Furthermore, the study was conducted in the time period leading up to presidential elections, which is the time when we would most likely expect rural women to hold political knowledge and opinions. One might find that women’s political knowledge and opinions to be lower in the many years between elections, and perhaps the association or stronger between socioeconomic empowerment and political knowledge and opinion formation might differ. Last, we recognize that the study is not positioned to explain the origins of women’s mobility outside the village or household agency.

6. Policy with gender implications: The Family Code Given the finding that women’s empowerment is positively associated with political knowledge and opinions, we turn to examining whether women’s empowerment is associated with support of the controversial pro-woman changes proposed to the Family Code, a hotly debated topic across the Sahel.27 While term limits and mayor’s performance are salient general issues, the proposed changes to the Family Code have clear implications for men versus women. In particular, the Family Code reforms under debate included many provisions that would have arguably improved the empowerment of women, such as minimum age to marry, mandating a civil marriage ceremony, and greater inheritance rights for children born out of wedlock. In Mali, over 1000 such changes were passed in parliament in August 2009 with an nearly unanimous vote, sparking a series of protests organized by Islamic organizations. The protests culminated in an over 50,000 person rally, and were successful in inducing the president to revoke the code. During the study time period, pro-woman changes to the Family Code were hotly debated in the lead up to the 2012 election. Islamic associations had launched aggressive sensitization efforts to build resistance to the proposed code.28 We code dichotomously supporters of the pro-woman Family Code changes as 1 versus those that do not support the changes or had no opinion as 0. As might be suspected, more women—at 25%—supported the changes, than males at 11% (see Appendices B.1 and C.1). In Table 2, we see support for the hypothesis that women with higher levels of empowerment are more likely to 26 These cost tradeoffs in sample size and data quality, are, of course, well-known in survey methods in low-income countries (Lupu & Michelitch, 2018). 27 See Wing (2009) for a review of Family Code debates within the region. 28 These groups were successful in passing a highly revised, more conservative version of the code later in 2012

310

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

Fig. 5. Relationship between household agency, mobility, and Family Code opinion. Notes: Predicted probabilities for supporting pro-woman Family Code changes rather than not supporting or not having an opinion.

support pro-woman changes to the Family Code. Fig. 5 depicts the predicted probabilities of support for the pro-woman changes. Changing from minimum mobility (never leaving the village) to maximum mobility (leaving more than once per week) is associated with a large substantive change in predicted probability of supporting the Family Code: from 0.19 to 0.37. The Household Agency Index was also strongly associated with support for the code. Moving from the lowest to the highest level of household agency (0 to 10) on this scale is associated with a change in predicted probability of support of the code: from 0.2 to 0.38. Further, we perform the same extensions of the Family Code analysis as for knowledge and opinion levels. As for women’s characteristics, we find less support for family code changes among women who engage in farming or fishing above and beyond household tasks, as well as those that exchange information at the river or in the family, but otherwise find no further significant relationships (see Appendices B.4–B.10). Interestingly, the woman-only space at the river may be a place that reinforces policies supporting existing gender norms instead of challenging them. Likewise, women who tend to exchange information in the family may be receiving antiFamily Code messaging – which could be from men (or women) who may feel they stand to lose if existing practices within the family change. We also find that male family member’s support of family code changes is positively and significantly associated with women’s support for family code changes (see Appendix C.1). In sum, while we did not find a significant relationship between levels of knowledge and opinions between a woman’s male family member and herself, the substantive opinion (given that an opinion is held) is correlated within the extended family. 7. Conclusion Using a variety of cutting-edge survey enumeration and questionnaire techniques to elicit high quality data from rural women, we find wide support that greater household agency and mobility are associated with higher levels of women’s political knowledge and opinion formation. Moreover, we find that socioeconomic empowerment is positively related to women’s support for prowoman changes to the controversial Family Code, a law that regulates the rights and responsibilities between men and women in a marriage and family. The fact that increases in small degrees of empowerment are associated with higher levels of political knowledge and opinions are very encouraging given the high level of gender inequality in the study context. If more empowered women can become aware and supportive of policies that further empower them socioeconomically, it could create the beginnings of a virtu-

ous cycle of women’s overall empowerment in both socioeconomic and political spheres. Future research should further examine how women’s socioeconomic empowerment may be improved to accelerate political knowledge and opinions, and ultimately autonomous participation. Field experimental interventions show mixed results regarding empowerment interventions. While some ((Beath et al., 2013; Prillaman (2016) show some positive and neutral effects, others show that interventions designed to empower women to participate more can sometimes unintentionally backfire due to patriarchal norms at the village level (Clayton, 2015; Gottlieb, 2016). Those advantaged with power have few incentives to dilute their own power to extend it to the de facto disenfranchised, even with de jure enfranchisement. Scholars and policy makers should continue to understand how the juxtaposition of universal suffrage, with a ‘‘one person one vote” principle, and patriarchal socioeconomic norms affect rural women’s development of autonomous agency to develop political knowledge and opinions to politically participate in line with their interests. Conflict of interest None. Acknowledgments Support for this research was provided by the Spencer Foundation, the National Science Foundation, the US Agency for International Development, the Catt Prize, and the Helen Kellogg Institute for International Studies and the Institute for Scholarship in the Liberal Arts at the University of Notre Dame. We would like to thank our research team in Mali and the USA particularly Youssouf Konde, Mamadou Guindo, Amadou Guindo, Oumar Toure, Grace Watkins, Natalie Vellutini, and Liana Cramer. For research support we thank Idrissa Sidibe and the Kassambara family. We thank Jeffrey Conroy-Krutz, Tariq Thachil, Mike Hoffman, Karrie Koesel, Sean McGraw, Lauren Honig, Lakshmi Iyer, Atalia Omer, Amanda Clayton, Jessica Gottlieb, and Bruce Whitehouse for suggestions and discussion. We also thank our anonymous reviewers for their feedback. Previous versions of this manuscript were presented at the African Studies Association Conference, the Midwest Political Science Conference, and the University of Notre Dame. Appendix A. Afrobarometer regression estimations See Tables 3 and 4.

311

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323 Table 3 Relationships between urban/rural and male/female and political knowledge and turnout.

Male Urban

(1) Know Term Limits

(2) Know Term Limits

(3) Voted in Last Election

(4) Voted in Last Election

0.79⁄⁄⁄ (0.100) 0.62⁄⁄⁄ (0.067)

0.30⁄⁄⁄ (0.075) 0.31⁄⁄⁄ (0.053)

1.27⁄⁄⁄ (0.081)

0.84⁄⁄⁄ (0.111) 0.68⁄⁄⁄ (0.088) 0.13 (0.089) 1.29⁄⁄⁄ (0.086)

2.09⁄⁄⁄ (0.044)

0.31⁄⁄⁄ (0.091) 0.30⁄⁄⁄ (0.069) 0.01 (0.078) 2.09⁄⁄⁄ (0.051)

Yes 24254 0.101

Yes 24254 0.102

Yes 25278 0.036

Yes 25278 0.036

Male  Urban Constant Country FE Observations Pseudo R2 ⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Logistic regression estimation with standard errors clustered at the country level and standard survey weights inversely proportional.

Table 4 Relationships between urban/rural and male/female and survey environment.

Male Urban

(1) Influenced

(2) Influenced

(3) Difficulty

(4) Difficulty

0.39⁄⁄⁄ (0.114) 0.22⁄ (0.123)

0.68⁄⁄⁄ (0.059) 0.54⁄⁄⁄ (0.058)

3.07⁄⁄⁄ (0.062)

0.34⁄⁄⁄ (0.130) 0.17 (0.145) 0.14 (0.142) 3.09⁄⁄⁄ (0.072)

4.30⁄⁄⁄ (0.019)

0.68⁄⁄⁄ (0.065) 0.53⁄⁄⁄ (0.051) 0.02 (0.058) 4.30⁄⁄⁄ (0.019)

Yes 25377 0.047

Yes 25377 0.047

Yes 25391 0.108

Yes 25391 0.108

Male  Urban Constant Country Fixed Effects Observations Pseudo R2 ⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Logistic regression estimation with standard errors clustered at the country level and standard survey weights inversely proportional.

Appendix B. Main results robustness checks and interplay of other characteristics of women B.1. Additional descriptive statistics See Figs. 6–18.

Fig. 6. Support for the Family Code. Notes: women only.

Fig. 7. Able to Understand French News. Notes: women only.

312

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

Fig. 8. Where medical treatment is sought. Notes: women only.

Fig. 9. Ethnicity. Notes: Respondents could select more than one ethnic group. The graph above depicts those who identify as Bamana. Bamana is the dominant group and lingua franca in Southern and Central Mali, so being Bamana may allow individuals to access more news or information.

Fig. 10. Civil status. Notes: women only.

Fig. 11. Engages in farming or fishing. Notes: women only.

Fig. 12. Engages in small scale enterprise (e.g., Hand Crafts). Notes: women only.

Fig. 13. Michael Jackson. Notes: women only.

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

Fig. 14. Exchanges political information at the mosque. Notes: women only.

313

Fig. 17. Exchanges political information in the family. Notes: women only.

Fig. 18. Bar Chart of Mean Mobility to Leave the Village by Village. Notes: Village identity anonymized. Fig. 15. Exchanges political information at the market. Notes: women only.

B.2. Robustness checks standard errors not clustered See Fig. 19 and Tables 5 and 6.

Fig. 16. Exchanges political information at the river. Notes: women only. Fig. 19. Scatterplot of Mobility to Leave the Village and Household Agency. Notes: Correlation is 0.16.

314

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

Table 5 Associations between socioeconomic empowerment and political engagement – robust SEs (no cluster).

Leave Village HH Agency Index Wealth Proxy Understand French News Bamana Village FE Observations

(1) Knowledge Index OLS

(2) Opinion Index OLS

0.71⁄⁄⁄ (0.058) 0.06⁄⁄⁄ (0.014) 0.20⁄⁄ (0.094) 0.78⁄⁄⁄ (0.260) 0.03 (0.112)

0.78⁄⁄⁄ (0.241) 0.13⁄⁄ (0.060) 0.25 (0.341) 0.15 (0.593) 0.00 (0.455)

Yes 281

Yes 226

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Robust standard errors (no clustering at village level).

Table 8 Associations between socioeconomic empowerment and support for Family Code changes – interaction HH agency and mobility. Support for Family Code changes 0.64⁄ (0.328) 0.20⁄⁄ (0.099) 0.08 (0.068) 0.78⁄⁄ (0.332) 0.52 (1.082) 0.35 (0.400)

Leave Village HH Agency Index HH Agency Index ⁄ Leave Village Wealth Proxy Understand French News Bamana Village FE Observations

Yes 281

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Table 6 Associations between socioeconomic empowerment and support for Family Code changes – robust SEs (no cluster). Support for Family Code changes 0.35⁄ (0.205) 0.10⁄⁄ (0.050) 0.80⁄⁄ (0.330) 0.50 (1.045) 0.35 (0.396)

Leave Village HH Agency Index Wealth Proxy Understand French News Bamana Village FE Observations

Table 9 Associations between socioeconomic empowerment and political engagement – combined HH agency and mobility index.

HH Agency/Mobility Index Wealth Proxy Understand French News

Yes 281

Bamana

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Robust standard errors (no clustering at village level).

Village FE Observations

(1) Knowledge Index OLS

(2) Opinion Index OLS

0.09⁄⁄⁄ (0.017) 0.31⁄⁄⁄ (0.076) 0.88⁄⁄⁄ (0.180) 0.04 (0.108)

0.15⁄⁄ (0.063) 0.35 (0.229) 0.43 (0.419) 0.10 (0.477)

Yes 281

Yes 226

⁄

B.3. Possible relation between and alternative operationalization of mobility and household agency

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

See Tables 7–10. Table 7 Associations between socioeconomic empowerment and political engagement – interaction HH agency and mobility.

Table 10 Associations between socioeconomic empowerment and support for Family Code changes – combined HH agency and mobility index. Support for Family Code changes

Leave Village HH Agency Index HH Agency Index ⁄ Leave Village Wealth Proxy Understand French News Bamana Village FE Observations ⁄

(1) Knowledge Index OLS

(2) Opinion Index OLS

0.73⁄⁄⁄ (0.115) 0.07⁄⁄⁄ (0.019) 0.01 (0.018) 0.20⁄ (0.097) 0.78⁄⁄ (0.276) 0.03 (0.086)

0.81⁄ (0.441) 0.14 (0.107) 0.01 (0.071) 0.25 (0.295) 0.16 (0.335) 0.00 (0.436)

Yes 281

Yes 226

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

HH Agency/Mobility Index Wealth Proxy Understand French News Bamana Village FE Observations ⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

0.11⁄⁄ (0.048) 0.82⁄⁄ (0.324) 0.56 (1.090) 0.35 (0.394) Yes 281

315

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

B.4. Possible moderation of relationship: ethnicity

Table 13 (continued) (1) (2) (3) (4) Knowledge Knowledge Knowledge Knowledge Index Index Index Index OLS OLS OLS OLS

See Tables 11 and 12. Table 11 Associations between socioeconomic empowerment and political engagement – interaction ethnicity.

Bamana Bamana ⁄ Leave Village Bamana ⁄ HH Agency Index Leave Village HH Agency Index Wealth Proxy Understand French News Village FE Observations

(1) Knowledge Index OLS

(2) Opinion Index OLS

0.05 (0.128) 0.10 (0.083) 0.01 (0.025) 0.66⁄⁄⁄ (0.044) 0.07⁄⁄⁄ (0.017) 0.20⁄ (0.095) 0.79⁄⁄ (0.280)

0.20 (0.928) 0.18 (0.666) 0.00 (0.120) 0.68⁄ (0.346) 0.13⁄⁄ (0.056) 0.24 (0.288) 0.15 (0.658)

Yes 281

Yes 226

(0.016)

Understand French News Bamana

Observations ⁄

⁄⁄

303

Leave Village Village Mean Mobility Rank Leave Village ⁄ Village Mean Mobility Rank

Support for Family Code changes

Bamana ⁄ HH Agency Index Leave Village HH Agency Index Wealth Proxy Understand French News Village FE Observations

(3) Opinion Index OLS

(4) Opinion Index OLS

0.41⁄⁄⁄ (0.095) 0.06 (0.032) 0.02⁄

0.43⁄⁄⁄ (0.076) 0.03 (0.029) 0.02

0.42⁄⁄⁄ (0.094) 0.07⁄ (0.032) 0.02

0.43⁄⁄⁄ (0.076) 0.04 (0.031) 0.02

(0.013)

(0.010) 0.29⁄⁄⁄ (0.090) 0.44 (0.311) 0.30⁄ (0.153)

(0.014)

0.04⁄ (0.020)

(0.012) 0.32⁄⁄ (0.115) 0.31 (0.292) 0.33⁄ (0.165) 0.03 (0.018)

281

281

Bamana HH Agency Index Observations

303

303

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Yes 281

Table 15 Interaction of mean village women’s mobility on individual women’s support for Family Code changes.

B.5. Village level mobility interacted with individual level mobility

Leave Village

See Tables 13–15.

Village Mean Mobility Rank Leave Village ⁄ Village Mean Mobility Rank

Table 13 Interaction of mean village women’s mobility on individual women’s knowledge.

0.76⁄⁄⁄ (0.123) 0.03

0.77⁄⁄⁄ (0.109) 0.04

0.73⁄⁄⁄ (0.133) 0.01

0.75⁄⁄⁄ (0.116) 0.02

(0.031) 0.00

(0.030) 0.01

(0.027) 0.00

(0.027) 0.01

(1) Support Logistic

(2) Support Logistic

(3) Support Logistic

(4) Support Logistic

0.46 (0.395) 0.05 (0.111) 0.00

0.50 (0.357) 0.01 (0.090) 0.01

0.22 (0.398) 0.12 (0.103) 0.02

0.23 (0.355) 0.07 (0.080) 0.01

(0.058)

(0.053) 0.70⁄⁄⁄ (0.245) 0.89 (1.052) 0.74⁄⁄ (0.320)

(0.053)

0.11⁄⁄⁄ (0.041)

(0.048) 0.70⁄⁄⁄ (0.257) 0.55 (1.125) 0.66⁄⁄ (0.320) 0.10⁄⁄ (0.044)

281

281

Wealth Proxy

(1) (2) (3) (4) Knowledge Knowledge Knowledge Knowledge Index Index Index Index OLS OLS OLS OLS

Leave Village ⁄ Village Mean Mobility Rank

(2) Opinion Index OLS

Understand French News

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Village Mean Mobility Rank

281

(1) Opinion Index OLS

Wealth Proxy

⁄

Leave Village

281

303

⁄⁄⁄

Table 14 Interaction of mean village women’s mobility on individual women’s opinions.

Table 12 Associations between socioeconomic empowerment and support for Family Code changes – interaction ethnicity.

Bamana ⁄ Leave Village

0.08⁄⁄⁄ (0.014)

(0.014) 0.23⁄⁄ (0.082) 0.85⁄⁄ (0.285) 0.08 (0.074) 0.06⁄⁄⁄ (0.012)

p < 0.10, p < 0.05, p < 0.01. Notes: Standard errors clustered at the village level.

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

0.32 (0.676) 0.21 (0.421) 0.07 (0.101) 0.25 (0.288) 0.13⁄ (0.067) 0.81⁄⁄ (0.330) 0.53 (1.053)

(0.016)

HH Agency Index

⁄

Bamana

(0.015) 0.26⁄⁄ (0.080) 1.08⁄⁄ (0.336) 0.06 (0.076)

Wealth Proxy

Understand French News Bamana HH Agency Index Observations ⁄

303

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

303

316

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

B.6. Role of civil status

B.7. Role of employment status

See Tables 16–18.

See Tables 19–21.

Table 16 Associations between socioeconomic empowerment and political engagement – adding civil status. (1) Knowledge Index OLS

(2) Opinion Index OLS

⁄⁄⁄

Leave Village HH Agency Index Single/Divorced/Widowed First Wife Wealth Proxy Understand French News Bamana Village FE Observations

Table 19 Associations between socioeconomic empowerment and political engagement – adding employment status.

⁄⁄

0.71 (0.063) 0.07⁄⁄⁄ (0.014) 0.52⁄ (0.268) 0.04 (0.109) 0.21⁄ (0.103) 0.80⁄⁄ (0.298) 0.02 (0.093)

0.76 (0.324) 0.14⁄⁄ (0.052) 0.31 (0.987) 0.20 (0.303) 0.25 (0.280) 0.33 (0.410) 0.02 (0.415)

Yes 279

Yes 225

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level. Left out civil status is second/ third/fourth wives.

Leave Village HH Agency Index Farmer/Fisher Small Scale Enterprise Wealth Proxy Understand French News Bamana Village FE Observations

⁄

(1) Knowledge Index OLS

(2) Opinion Index OLS

0.71⁄⁄⁄ (0.055) 0.06⁄⁄⁄ (0.013) 0.05 (0.123) 0.03 (0.105) 0.20⁄ (0.096) 0.77⁄⁄ (0.279) 0.03 (0.088)

0.81⁄⁄ (0.311) 0.14⁄⁄ (0.048) 0.98⁄⁄ (0.379) 0.27 (0.393) 0.25 (0.262) 0.42 (0.459) 0.02 (0.426)

Yes 281

Yes 226

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Table 17 Associations between socioeconomic empowerment and support for Family Code changes – adding civil status. Support for Family Code changes 0.35⁄ (0.204) 0.10⁄⁄ (0.051) 0.08 (0.457) 0.40 (0.827) 0.81⁄⁄ (0.330) 0.54 (1.068) 0.34 (0.398)

Leave Village HH Agency Index First Wife Single/Divorced/Widowed Wealth Proxy Understand French News Bamana Village FE Observations ⁄

⁄⁄

Table 20 Associations between socioeconomic empowerment and support for Family Code changes – adding employment status. Support for Family Code changes

HH Agency Index Farmer/Fisher Small Scale Enterprise Wealth Proxy

Yes 279

Understand French News

⁄⁄⁄

p < 0.10, p < 0.05, p < 0.01. Notes: Standard errors clustered at the village level. Left out civil status is second/ third/fourth wives.

Table 18 Relationship between civil status and HH agency and mobility.

Single/Divorced/Widowed First Wife Wealth Proxy Understand French News Bamana Village FE Observations ⁄

Bamana Village FE Observations

Yes 281

⁄

(1) Mobility OLS

(2) HH Agency OLS

0.22 (0.257) 0.25** (0.106) 0.15 (0.111) 0.17 (0.235) 0.03 (0.096)

1.84⁄⁄ (0.782) 0.41 (0.434) 0.65 (0.410) 2.58⁄⁄ (0.963) 0.81⁄⁄ (0.329)

Yes 301

Yes 279

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Logistic 0.36⁄ (0.202) 0.11⁄⁄ (0.053) 1.26⁄⁄⁄ (0.387) 0.07 (0.350) 0.77⁄⁄ (0.339) 0.15 (1.049) 0.43 (0.414)

Leave Village

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Table 21 Relationship between employment status and HH agency and mobility.

Farmer/Fisher Small Scale Enterprise Wealth Proxy

(1) Mobility OLS

(2) HH Agency OLS

0.03 (0.028) 0.02 (0.107) 0.15 (0.118)

0.02 (0.413) 0.81⁄⁄⁄ (0.246) 0.69 (0.395)

317

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323 Table 21 (continued)

Understand French News Bamana Village FE Observations

(1) Mobility OLS

(2) HH Agency OLS

0.28 (0.213) 0.01 (0.094)

2.88⁄⁄ (0.929) 0.70⁄ (0.335)

Yes 303

Yes 281

Table 24 Associations between socioeconomic empowerment and support for Family Code changes – adding knowledge of MJ/death. Support for Family Code changes Logistic 0.35⁄ (0.205) 0.10⁄⁄ (0.050) 0.15 (0.605) 0.81⁄⁄ (0.332) 0.49 (1.066) 0.36 (0.393)

Leave Village HH Agency Index Knew Michael Jackson Died Wealth Proxy

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Understand French News Bamana

B.8. Role of education Education is proxied for by French language ability, given the difficulty of measuring educational attainment for folks that would have needed to leave the village to go to school and they may have been in and out of school (Table 22).

Village FE Observations

Yes 281

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Table 25 Relationship between knowledge of MJ/Death and HH agency and mobility. Table 22 Relationship between education and HH agency and mobility.

Wealth Proxy Understand French News Bamana Village FE Observations

(1) Mobility OLS

(2) HH Agency OLS

0.15 (0.117) 0.27 (0.211) 0.01 (0.092)

0.72 (0.402) 2.79⁄⁄ (1.003) 0.77⁄⁄ (0.323)

Yes 303

Yes 281

Knew Michael Jackson Died Wealth Proxy Understand French News Bamana Village FE Observations

(1) Mobility OLS

(2) HH Agency OLS

0.07 (0.328) 0.15 (0.121) 0.25 (0.246) 0.01 (0.095)

0.67 (0.933) 0.74 (0.416) 2.66⁄⁄⁄ (0.807) 0.79⁄⁄ (0.311)

Yes 303

Yes 281

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

B.9. Role of knowledge of non-political news – Michael Jackson and his death

B.10. Role of spaces where info is exchanged See Tables 26–28.

See Tables 23–25.

Table 23 Associations between socioeconomic empowerment and political engagement – adding knowledge of MJ/death.

Leave Village HH Agency Index Knew Michael Jackson Died Wealth Proxy Understand French News Bamana Village FE Observations ⁄

(1) Knowledge Index OLS

(2) Opinion Index OLS

0.71⁄⁄⁄ (0.053) 0.06⁄⁄⁄ (0.013) 0.43⁄ (0.219) 0.22⁄ (0.099) 0.70⁄⁄ (0.238) 0.04 (0.077)

0.76⁄⁄ (0.292) 0.13⁄⁄ (0.045) 0.79 (0.891) 0.29 (0.274) 0.88 (0.838) 0.02 (0.380)

Yes 281

Yes 226

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Table 26 Associations between socioeconomic empowerment and political engagement – adding exchanges info at different spaces.

Leave Village Agency Index Info at Mosque Info at Market Info at River Info in Family Wealth Proxy Understand French News Bamana

(1) Knowledge Index OLS

(2) Opinion Index OLS

0.70⁄⁄⁄ (0.069) 0.06⁄⁄⁄ (0.013) 0.40⁄⁄⁄ (0.067) 0.23⁄ (0.106) 0.24⁄⁄ (0.084) 0.21 (0.123) 0.20⁄⁄ (0.084) 0.77⁄⁄ (0.241) 0.06 (0.093)

0.79⁄⁄ (0.305) 0.10⁄ (0.055) 1.04 (1.350) 1.00⁄ (0.511) 1.75⁄⁄⁄ (0.227) 0.52 (0.435) 0.26 (0.228) 0.93⁄⁄ (0.366) 0.10 (0.410) (continued on next page)

318

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

Table 26 (continued)

Village FE Observations

(1) Knowledge Index OLS

(2) Opinion Index OLS

Yes 279

Yes 224

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Appendix C. Role of male family member characteristics in women’s political engagement and socioeconomic empowerment C.1. Descriptive statistics for male family members of women in sample See Figs. 20–28.

Table 27 Associations between socioeconomic empowerment and support for Family Code changes – adding exchanges info at different spaces. Support for Family Code changes Logistic 0.28 (0.218) 0.09⁄ (0.053) 0.15 (0.802) 0.51 (0.488) 1.46⁄⁄ (0.620) 1.91⁄⁄⁄ (0.628) 0.81⁄⁄ (0.354) 0.79 (1.137) 0.61 (0.419)

Leave Village Agency Index Info at Mosque Info at Market Info at River Info in Family Wealth Proxy Understand French News Bamana Village FE Observations

Fig. 20. Male family member able to Understand French News. Notes: Men only.

Yes 279

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Table 28 Relationship between exchanges info at different spaces and HH agency and mobility.

Info at Mosque Info at Market Info at River Info in Family Wealth Proxy Understand French News Bamana Village FE Observations

(1) Mobility OLS

(2) HH Agency OLS

0.31 (0.436) 0.03 (0.118) 0.06 (0.177) 0.19 (0.177) 0.14 (0.121) 0.32 (0.206) 0.03 (0.102)

0.63 (1.001) 0.81⁄⁄ (0.291) 0.61 (0.518) 0.28 (0.677) 0.69 (0.400) 2.61⁄⁄ (0.812) 0.78⁄⁄ (0.312)

Yes 301

Yes 279

Fig. 21. Male family member: where medical treatment is sought. Notes: Men only.

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Fig. 22. Male family member’s mobility. Notes: Men only.

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

Fig. 23. Male family member engages in farming or fishing. Notes: Men only.

319

Fig. 26. Male family member is religious leader. Notes: Men only.

Fig. 24. Male family member engages in small scale enterprise (e.g., Hand Crafts) Notes: Men only.

Fig. 27. Male family member’s civil status. Notes: Men only.

Fig. 25. Male family member engages in herding. Notes: Men only.

Fig. 28. Male family member’s support for the Family Code. Notes: Men only.

320

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

C.2. Role of male family members’ characteristics in women’s socioeconomic empowerment and political engagement

Table 31 (continued) (1) (2) (3) (4) (5) (6) Mobility Mobility Mobility Mobility Mobility Mobility OLS OLS OLS OLS OLS OLS

See Tables 29–32.

Table 29 Associations between socioeconomic empowerment and political engagement – adding male family member’s political engagement. (1) Knowledge Index OLS Male Family Member’s Knowledge Index

Leave Village Agency Index Wealth Proxy Understand French News Bamana

(0.195)

0.12 (0.078)

0.68 (0.063) 0.06⁄⁄⁄ (0.017) 0.26⁄⁄ (0.099) 0.74⁄⁄ (0.251) 0.01 (0.064)

0.16 (0.093) 0.64⁄ (0.314) 0.12⁄⁄ (0.042) 0.28 (0.391) 0.07 (0.304) 0.25 (0.465)

Yes 236

Yes 163

Village FE Observations

Male Family Member Herds Male Family Member Does Small Business Male Family Member Religious Leader Wealth Proxy Understand French News Bamana

⁄

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Support for Family Code changes Logistic 1.11⁄⁄ (0.479) 0.34⁄ (0.199) 0.12⁄ (0.061) 1.07⁄⁄⁄ (0.352) 0.38 (1.169) 0.55 (0.359)

Male Family Member’s Support Leave Village Agency Index Wealth Proxy Understand French News Bamana Village FE Observations

(0.215) 0.07

(0.202) 0.06

(0.246) 0.12 (0.150) 0.32

(0.249) 0.12 (0.157) 0.33

(0.211) 0.00 (0.104)

(0.222) 0.02 (0.098)

(0.203) 0.01 (0.101)

(0.214) 0.01 (0.100)

(0.199) 0.03 (0.106)

(0.190) 0.03 (0.112)

256

256

256

256

256

256

⁄⁄⁄

p < 0.10, p < 0.05, p < 0.01. Notes: Standard errors clustered at the village level.

Table 32 Relationship between women’s HH agency and male family member’s demographics.

0.04

0.03

(0.300) 0.75

0.74

(0.646)

(0.662) 0.64

0.56⁄ (0.299)

Male Family is Single/ Divorced/Widowed

(0.363) 0.52

0.52 (0.702)

Male Family Member Farms Male Family Member Herds Male Family Member Does Small Business

0.03 Male Family Member Religious Leader (0.070) 0.08

Wealth Proxy Understand French News

(0.147)

(0.199) 0.15

0.13 (0.151) 0.30

Male Family Member’s Wealth Proxy

(1) (2) (3) (4) (5) (6) Mobility Mobility Mobility Mobility Mobility Mobility OLS OLS OLS OLS OLS OLS

0.08

(0.205) 0.15

(0.287)

Table 31 Relationship between women’s mobility and male family member’s demographics.

(0.069)

(0.155) 0.25

0.13 (0.148) 0.28

Male Family Member Understands French News

p < 0.10, ⁄⁄p < 0.05, ⁄⁄⁄p < 0.01. Notes: Standard errors clustered at the village level.

Male Family Member Understands French News

(0.153) 0.26

0.13 (0.148) 0.31

Male Family Member’s Mobility

Yes 236

0.04

(0.184) 0.17

(1) (2) (3) (4) (5) (6) HH HH HH HH HH HH Agency Agency Agency Agency Agency Agency OLS OLS OLS OLS OLS OLS

⁄

Male Family Member’s Mobility

⁄⁄

0.17

0.12 (0.162) 0.29

Observations ⁄

Table 30 Associations between socioeconomic empowerment and support for Family Code changes – adding male family member’s support for Family Code changes.

(0.107) 0.07

0.07

Male Family Member Farms

OLS

⁄⁄⁄

0.05

(0.122) Male Family is Single/Divorced/ Widowed

(2) Opinion Index

Male Family Member’s Opinion Index

0.03

Male Family Member’s Wealth Proxy

(0.167)

0.42

(0.775) 0.40

(0.814) (0.810) 0.23 0.14 (0.525) (0.549) 0.40 0.51

(0.592) (0.551) 0.18 0.08 (1.125) (1.194) 0.78⁄ 0.75⁄ 0.79⁄ 0.76⁄ 0.82⁄ 0.78 (0.404) (0.384) (0.413) (0.387) (0.431) (0.443) ⁄⁄ ⁄⁄ ⁄⁄⁄ ⁄⁄ ⁄⁄ 2.73 2.59 2.83 2.69 2.68 2.61⁄⁄⁄ (0.983) (0.815) (0.848) (0.997) (0.990) (0.666)

321

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323 Table 32 (continued)

Appendix D (continued) (1) (2) (3) (4) (5) (6) HH HH HH HH HH HH Agency Agency Agency Agency Agency Agency OLS OLS OLS OLS OLS OLS 1.07⁄⁄ 1.04⁄⁄ 1.09⁄⁄ 1.09⁄⁄ 1.09⁄⁄ 1.06⁄⁄ (0.368) (0.322) (0.339) (0.349) (0.394) (0.404)

Bamana Observations ⁄

⁄⁄

236

236

236

236

236

236

⁄⁄⁄

p < 0.10, p < 0.05, p < 0.01. Notes: Standard errors clustered at the village level.

Appendix D. Survey questions Construct

Question Translation

Answer Choices

What is the name of your mayor?

a. correct b. incorrect

Political Engagement Political Knowledge 1: Politician Political Opinion 1: Politician Performance

Using a comparison, do you think that the current mayor is more understanding/recep tive/responsive than his/her predecessor? Justification for Why did you give this Opinion 1 assessment? What are presidential Political Knowledge term limits in Mali? 2: Institutions Political Do you think that term Opinions 2: limits in Mali are a good Institutions idea or that if a president is good, (s) he should be candidate in elections as many times as (s) he wants? Justification for Opinion 2 Political Knowledge 3: Policy Topic

Political Opinions 3: Policy Topic Justification for Opinion 3

Mobility Frequency of Leaving Village Women’s Household Agency

a. Much better b. Better c. Same d. Worse e. Much worse

(open response) a.correct b. incorrect

a. term limits are a good idea b. a. That if the president is good, he or she should have the ability to be a candidate in the elections as much as he or she wants For what reason? why? (open response write in) a.correct b. incorrect What do you know about the Family Code debates? Can you tell us one of the laws that was proposed by the Family Code? Do you think that the a. yes b. no Family Code Bill should be adopted? For what reason(s) do (open response write you think that it should in) be adopted (or not)?

How many times do you travel outside the village each month?

a. never b. 1 time c. 2–3 times d. every week

Construct

Question Translation

Answer Choices

Household Agency Index

In your (plural you) marriages, do husbands and wives discuss choices about sending to children to school? In your (plural you) marriages, do husbands and wives discuss choices about finding a spouse for their sons and daughters? In your (plural you) marriages, do husbands and wives discuss the family budget (financial management)? In your (plural you) marriages, do husbands and wives discuss the choice to bring a new (polygamous) wife into the family? In your (plural you) marriages, if a woman earns money herself, can she decide to how to spend it? In your (plural you) marriages, do the wife and the husband talk about if a husband should bring home condiments home (do shopping)? In your (plural you) marriages, do the husband and wife discuss family planning? In your (plural you) marriages, do the women in the family discuss together if a female family member wants to start a business? Think about food that is distributed in your household, is there special food reserved for certain people (implies husband) in your household? Should a husband be able to use beatings to discipline his wife?

yes (select chicken) no (select fish)

Demographics Ethnicity What are your ethnic groups?

yes (select chicken) no (select fish)

yes (select chicken) no (select fish)

yes (select chicken) no (select fish)

yes (select chicken) no (select fish)

yes (select chicken) no (select fish)

yes (select chicken) no (select fish)

yes (select chicken) no (select fish)

yes (select chicken) no (select fish)

yes (select chicken) no (select fish)

a. Bamana b. Peul c. Dogon d. Songhoy e. Bozo f. Tuareg e. Soninke f. other (continued on next page)

322

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323

References

Appendix D (continued)

Construct

Question Translation

Answer Choices

Education Proxy

Do you speak French well enough to understand news on the radio? Think about the last time you were sick, how did you address the sickness?

a. yes b. no

Wealth Proxy

Civil status of the respondent

What is your civil status?

Occupation

What types of jobs do you do? any others?

a. went to a medical center in a nearby village b. used a cellphone to call a health official from a nearby village c. consulted a traditional healer from the village d. asked a friend or family member to prepare medicine for you e. nothing a. 1st wife b. 2nd wife c. 3rd wife d. 4th wife e. husband f. single woman g. single man h. divorced woman i.divorced man j. widow k.widower a. fishing b. farmer/market gardening c. herding d. sewing/weaving e. repairperson f. trading g. transport h. AGR i. religious leadership (marabout/feticheur) k. household management k. dyeing. l. other

Other Supportive Questions Non-Political News Knowledge

Do you know how Michael Jackson is? (If yes) Were you aware that he died?

a. yes, know him and that he died b. know him, but don’t know he died c. don’t know Michael Jackson

Sources of News and Information Places where you exchange information

What are the different a. mosque b. market c.River d.Family places where you receive, share, discuss, and distribute information of regional, national, international nature? any others?

Notes: Non-response and I don’t know answer choices were available for all questions.

Adams, Melinda, & Kang, Alice (2007). Regional advocacy networks and the protocol on the rights of women in Africa. Politics and Gender, 3, 451–474. Adida, Claire L., Ferree, Karen E., Posner, Daniel N., & Robinson, Amanda Lea (2016). Who’s asking? Interviewer coethnicity effects in African survey data. Comparative Political Studies, 49(12), 1630–1660. Armstrong, A., Beyani, C., Himonga, C., Kabeberi-Macharia, J., Molokomme, A., Ncube, W., ... Stewart, J. (1993). Uncovering reality: Excavating women’s rights in African family law. International Journal of Law, Policy and the Family, 7(3), 314–369. Arriola, Leonardo R., & Johnson, Martha C. (2014). Ethnic politics and women’s empowerment in africa: Ministerial appointments to executive cabinets. American Journal of Political Science, 58(2), 495–510. Baldwin, Kate (2015). The paradox of traditional leaders in democratic Africa. Cambridge University Press. Barnes, Tiffany D., & Burchard, Stephanie M. (2013). Engendering politics: The impact of descriptive representation on women’s political engagement in SubSaharan Africa. Comparative Political Studies, 46(7), 767–790. Bauer, Gretchen (2004). ‘The hand that stirs the pot can also run the country’: Electing women to parliament in Namibia. The Journal of Modern African Studies, 42(4), 479–509. Beath, Andrew, Christia, Fotini, & Enikolopov, Ruben (2013). Empowering women through development aid: Evidence from a field experiment in Afghanistan. American Political Science Review, 107(03), 540–557. Beck, Linda (2008). Brokering democracy in Africa. The rise of clientelist democracy in Senegal. New York: Palgrave MacMillan. Beck, Linda J. (2001). Reining in the Marabouts? Democratization and local governance in Senegal. African Affairs, 100(401), 601–621. Beck, Linda J. (2003). Democratization and the hidden public: The impact of patronage networks on senegalese women. Comparative Politics, 35(2), 147–169. Bleck, Jaimie (2015). Education and empowered citizenship in Mali. Johns Hopkins University Press. Bleck, Jaimie, & Michelitch, Kristin (2015). The 2012 crisis in Mali: Ongoing empirical state failure. African Affairs, 114(457), 598–623. Burns, Nancy, Schlozman, Kay Lehman, & Verba, Sidney (2001). The private roots of public action. Harvard University Press. Callaway, Barbara (1984). Ambiguous consequences of socialization and seclusion of hausa women. The Journal of Modern African Studies, 22(3), 429–450. Carpini, Michael X. Delli, & Keeter, Scott (2005). Gender and political knowledge. Gender and American Politics: Women, Men, and the Political Process, 21–52. Chattopadhyay, R., & Duflo, Esther (2004). Women as policy makers: Evidence from a randomized policy experiment in India. Econometrica, 72(5), 1409–1443. Chauchard, Simon (2013). Using MP3 players in surveys: The impact of a low-tech self-administration mode on reporting of sensitive attitudes. Public Opinion Quarterly, 77(S1), 220–231. Chhibber, Pradeep (2002). Why are some women politically active? The household, public space, and political participation in India. International Journal of Comparative Sociology, 43(3–5), 409–429. Clayton, Amanda (2015). Women’s political engagement under quota-mandated female representation: Evidence from a randomized policy experiment. Comparative Political Studies, 48(3), 333–369. Coffe, Hilde, & Bolzendahl, Catherine (2011). Gender gaps in political participation across sub-Saharan African Nations. Social Indicators Research, 102(2), 245–264. Collier, Ruth Berins (1982). Regimes in tropical Africa: Changing forms of supremacy, 1945–1975. Berkeley, CA: University of California Press. Coulibaly, Mamadou, & Amadou Diarra. (2004). Democratie et Legitimation du Marche: Rapport d’ Enquete Afrobarometer au Mali. Afrobarometer Working Paper Series 35. Dahl, Robert (1971). Polyarchy: Participation and opposition. New Haven, CT: Yale University Press. Dolan, Kathleen (2011). Do women and men know different things? Measuring gender differences in political knowledge. The Journal of Politics, 73(01), 97–107. Duflo, Esther (2012). Women empowerment and economic development. Journal of Economic Literature, 50(4), 1051–1079. Friedberg, Leora, & Webb, Anthony. (2006). Determinants and Consequences of Bargaining Power in Households. Technical report National Bureau of Economic Research. Furuta, Marie, & Salway, Sarah. (2006). Women’s Position Within the Household as a Determinant of Maternal Health Care use in Nepal. International family planning perspectives pp. 17–27. Girard, Alexandra M. (2014). Stepping into formal politics: Women’s engagement in formal political processes in irrigation in rural India. World Development, 57, 1–18. Gottlieb, Jessica (2016). Why might information exacerbate the gender gap in civic participation? Evidence from Mali. World Development, 86, 95–110. Gottlieb, Jessica, & Amanda Lea Robinson. (2016). The effects of matrilineality on gender differences in political behavior across Africa. In April 2016 meeting of Working Group on African Political Economy in Washington, DC. Gottlieb, J., Grossman, G., & Robinson, A. L. (2016). Do Men and Women Have Different Policy Preferences in Africa? Determinants and Implications of Gender Gaps in Policy Prioritization. British Journal of Political Science, 1–26.

J. Bleck, K. Michelitch / World Development 106 (2018) 299–323 Gottlieb, Jessica, & Larreguy Horacio. (2016). An informational theory of electoral targeting in young-clientelistic democracies: Evidence from Senegal. Berkeley Center on the Politics of Development Working Paper 2016-15. Isaksson, Ann-Sofie (2014). Political participation in Africa: The role of individual resources. Electoral Studies, 34, 244–260. Isaksson, Ann-Sofie, Kotsadam, Andreas, & Nerman, Mans (2014). The gender gap in african political participation: Testing theories of individual and contextual determinants. Journal of Development Studies, 50(2), 302–318. Iversen, Torben, & McCall, Rosenbluth Frances (2010). Women, work, and politics: The political economy of gender inequality. Yale University Press. Iversen, Torben, & Rosenbluth, Frances (2006). The political economy of gender: Explaining cross-national variation in the gender division of labor and the gender voting gap. American Journal of Political Science, 50(1), 1–19. Iversen, Torben, & Rosenbluth, Frances (2008). Work and power: The connection between female labor force participation and female political representation. Annual Review of Political Science, 11, 479–495. Janssens, Wendy (2010). Women’s empowerment and the creation of social capital in Indian villages. World Development, 38(7), 974–988. Jayachandran, Seema (2015). The roots of gender inequality in developing countries. Economics, 7(1), 63–88. Kabeer, Naila (1999). Resources, agency, achievements: Reflections on the measurement of women’s empowerment. Development and Change, 30(3), 435–464. Kabeer, Naila (2001). Conflicts over credit: Re-evaluating the empowerment potential of loans to women in rural Bangladesh. World Development, 29(1), 63–84. Kang, Alice (2016). Bargaining for women’s rights: Activism in an aspiring muslim democracy. University of Minnesota Press. Koter, Dominika (2016). Beyond ethnic politics in Africa. Cambridge University Press. Krosnick, Jon A., Narayan, Sowmya, & Smith, Wendy R. (1996). Satisficing in surveys: Initial evidence. New Directions for Evaluation, 1996(70), 29–44. Kuenzi, Michelle T. (2006). Nonformal education, political participation, and democracy: Findings from Senegal. Political Behavior, 28(1), 1–31. Lackenbauer, Helene, Magdalena Tham Lindell, & Gabriella Ingerstad. (2015). If our men won’t fight, we will: A gendered analysis of the armed conflict in Mali. Technical report Swedish Defense Agency. Lemarchand, Rene, & Legg, Keith (1972). Political clientelism and development: A preliminary analysis. Comparative Politics, 4(2), 149–178. Logan, Carolyn (2013). The roots of resilience: Exploring popular support for African traditional authorities. African Affairs, 112(448), 535. 376. Logan, Carolyn, & Michael Bratton. (2006). The political gender gap in Africa: Similar attitudes, different behaviors. Afrobarometer Working Paper Series 58.

323

Lupu, Noam, & Michelitch, Kristin (2018). Survey research methods in developing countries. Annual Review of Political Science, 21, 11.1–11.20. Mahmud, Simeen, Shah, Nirali M, & Becker, Stan (2012). Measurement of women’s empowerment in rural Bangladesh. World Development, 40(3), 610–619. Malhotra, Anju, & Schuler, Sidney Ruth (2005). Women’s empowerment as a variable in international development. Measuring Empowerment: CrossDisciplinary Perspectives, 71–88. Mondak, Jeffery J., & Anderson, Mary R. (2004). The knowledge gap: A reexamination of gender-based differences in political knowledge. Journal of Politics, 66(2), 492–512. Niemi, Richard G., Craig, Stephen C., & Mattei, Franco (1991). Measuring internal political efficacy in the 1988 national election study. American Political Science Review, 85(04), 1407–1413. Guillermo, O’Donnell (2004). Human development, human rights, and democracy. In Guillermo O’Donnelll, Jorge Vargas Cullell, & Osvaldo M. Iazzetta (Eds.), The quality of democracy: Theory and applications (pp. 9–92). University of Notre Dame. Prillaman, Soledad. (2016). Strength in numbers: How women’s networks close India’s political gender gap. Presented at the American Political Science Association Conference, Philadelpha. Pye, Lucian (Ed.). (1963). Communications and political development. Princeton University Press. Sen, Amartya (1999). Development as freedom. New York: Alfred A. Knopf. Tripp, Aili Mari (2012). Women and politics in Uganda. University of Wisconsin Press. Tripp, Aili Mari (2015). Women and power in post-conflict Africa. Cambridge University Press. Tripp, Aili Mari, & Kang, Alice (2008). The global impact of quotas on the fast track to increased female legislative representation. Comparative Political Studies, 41 (3), 338–361. Wiig, Henrik (2013). Joint titling in rural Peru: Impact on women’s participation in household decision-making. World Development, 52, 104–119. Wing, Susanna D. (2009). Exploring democracy, women’s rights, and legal reform in francophone Africa. Rethinking Development: Societal Transformations and the Challenges of Governance in Africa and the Middle East, Yale University. Wooten, Stephen (2003). Women, men, and market gardens: Gender relations and income generation in rural Mali. Human Organization, 62(2), 166–177.

Further reading Krosnick, Jon (1999). Survey research. Annual Review of Psychology, 50, 537–567.