In recent years, there has been a dramatic increase in the number of women in positions of political power. This shift has largely been driven by the fact that over 100 countries, to date, have adopted some sort of gender quota in the form of new electoral laws or by requiring that women occupy a certain share of legislative seats.Footnote 1 This change in descriptive representation has been welcomed because of its perceived implications for substantive representation: it is commonly argued that increasing the inclusiveness of elected bodies provides a voice for previously under-represented groups.Footnote 2 Two key assumptions underlie such reasoning: first, that women have different policy preferences than men; and secondly, that female elected officials are more willing or able to represent the preferences of female constituents.Footnote 3 If both assumptions hold, then it follows that low levels of participation by women, compounded by the reluctance of (some) men to vote for female candidates,Footnote 4 could reproduce gender inequalities in various social and economic domains. This conclusion is what drives advocacy related to adopting policies to promote women’s formal representation.
Surprisingly, even though the above assumptions are commonly invoked in both academic and policy circles, with few exceptions they have not been subjected to rigorous empirical testing by scholars of the developing world. Little is known about whether women in developing countries do, in fact, have different political priorities and preferences than men, and if so, why.Footnote 5 We begin addressing this gap in the literature by studying the first assumption (regarding gender differences in policy preferences) in Sub-Saharan Africa, where descriptive representation of women is among the highest in the world and debates about enacting policies to promote women’s representation have been especially salient.Footnote 6 Thus it is imperative to evaluate the assumption of gender divergence in preferences within this context.
To do so, we first identify the presence of gender-based differences in political preferences across twenty-seven African countries using data culled from nationally representative surveys collected by Afrobarometer in 2008–09 and 2010–12 (n=69,640).Footnote 7 Specifically, we focus on whether female and male constituents differ in their policy priorities across ten meaningful policy areas . While we find that the effect of gender on the prioritization of many of these policy domains is statistically significant, we cannot as robustly claim that these differences are substantively significant, given that the size of coefficients is relatively small (compared to findings from advanced industrialized democracies) and because we find only minor differences in the aggregated priority rankings of male and female constituents.
Classifying Policy Domains
These relatively small average gender gaps in policy prioritization, however, belie considerable variation in the size of the gap both within and across countries in our sample. For example, we observe larger gender gaps across policy domains in Swaziland and Mali, but far fewer (and much smaller) gaps in Guinea and Sierra Leone (see Appendix Figure A.1). Such a pooled analysis also treats all policy areas as equivalent, ignoring important differences across policy domains, especially in terms of the direction of the gender gap (that is, whether men or women are more likely to prioritize the domain) and the factors that are likely to be driving them. We thus leverage substantial variation both between and within countries to better understand the factors that exacerbate gender divergence in different policy domains.
We argue that for many policy domains, the size of the gender gap in policy prioritization is decreasing in women’s empowerment. Importantly, our goal in this article is to both tease out and identify the relationship between women’s empowerment and the size of the gender gap for a number of theoretically relevant policy domains, not to explain all the variation in the size of the gender gap across all policy priorities. We hope that the framework we present in this article will encourage students of developing countries to think carefully about the political implications of the interaction between socially constructed gender roles and contextual features of the polity.
We begin by testing the above argument on two key policy domains, one typically favored by men (infrastructure investment) and one typically prioritized more by women (access to clean water). Specifically, we examine gender-based policy prioritization in these domains as a function of two key factors underlying female empowerment: financial independence and social vulnerability. Financial independence is strengthened when women are employed or have greater prospects of being employed, and this should in turn affect women’s policy prioritization. For example, in advanced industrialized democracies, a closing of the gender gap occurred within several policy domains as women entered the workforce and increased their financial independence from men.Footnote 8 In fact, women even began to prioritize the expansion of state services more than men as their social and economic independence increased.Footnote 9 We similarly expect women who enter the labor market in African states to take a greater interest in policies such as infrastructure investment compared to women whose livelihoods primarily depend on their spouse or extended family, consequently closing the gender gap.Footnote 10
In contrast, female empowerment is weakened when women are more socially vulnerable; that is, when cultural, traditional and religious norms place women at a systematic disadvantage. For example, women who are constrained by traditional gender roles are more likely to prioritize access to clean water relative to other policy domains because norms prescribe fetching water as a role for women.Footnote 11 Thus we expect that the greater social vulnerability of women will result in a greater divergence of preferences within some gendered policy domains.
Financial independence and social vulnerability can operate at both the individual and country levels. At the individual level, a change of one’s status – for example, gaining formal employment – affects policy priorities via changes in one’s needs, interests or direct experience. Country-level factors – for example, the overall share of female employment or the presence of institutions that perpetuate social practices that limit women socially – instead affect policy prioritization because they shape individuals’ expectations and future opportunities.
We also examine one domain – poverty alleviation – in which we expect the gender gap (but not necessarily levels) to be relatively stable; that is, less sensitive to women’s economic and social empowerment. Here we draw on the ‘ethics of care’ literature,Footnote 12 which stipulates that women are more likely to internalize a responsibility to care for others and to protect the most vulnerable in society due to differential patterns of socialization. Gendered socialization processes in turn encourage a general disposition that leads women to be more likely than their male counterparts to prioritize poverty alleviation policies.Footnote 13
Our results are largely consistent with these expectations. Greater financial independence, measured as participation in the labor force, increases women’s prioritization of infrastructure relative to men’s and decreases women’s prioritization of water relative to men’s, ultimately eliminating the gender gap in both policy domains. Social vulnerability widens the gap in infrastructure prioritization, but not in water prioritization. Consistent with the ethics of care literature, we find a large gender gap in prioritizing poverty alleviation that is relatively stable across measures of social venerability, female employment and the respondents’ own level of poverty. Taken together, these findings provide a set of analytical tools and benchmarks against which to better examine the substantive representation of women’s needs and preferences.
Finding that gender gaps in policy prioritization indeed exist in Africa, and in the policies and places we would expect, we conclude our analysis by briefly interrogating the political implications of such gaps. Specifically, we examine the relationship between gender gaps in policy priorities and gender gaps in political representation and participation. We find a strong negative correlation between gender gaps in policy priorities and the share of women in the national parliament. Similarly, we find a large positive correlation between gender gaps in policy priorities and gender gaps in political participation. Put simply, it is exactly in places where women and men have the most divergent policy preferences – and thus female representation and participation is most needed – that we also observe the least female participation in politics. While we make no claims that the relationship between gender gaps in policy prioritization and those in political participation are causally related, it is nevertheless worrisome that African countries that have lagged in addressing gender inequality in descriptive representation are precisely those in which such representation is needed most, due to the divergent policy priorities of male and female constituents.
This article’s main contribution is to add new insight and empirics from the developing world to the growing literature on the relationship between gender and political preferences. Until recently, the study of gender gaps in preferences has concentrated on identifyingFootnote 14 and explainingFootnote 15 gender-based preferences in a small number of advanced, industrialized democracies. By contrast, the study of gender-based preferences in the developing world is limited and, for the most part, focuses on identifying and explaining gender gaps in discrete domains such as support for democracyFootnote 16 and women’s health.Footnote 17 We thus expand the geographic coverage, as well as deepen the treatment of policy domains, for which gender plays an important role in the construction of political preferences. Unifying findings from the developed world with findings from Africa enables us to develop a more comprehensive theoretical framework for understanding the dynamic relationship between gender and political preferences.
The article also contributes to the nascent literature on the relationship between descriptive and substantive representation of women by female parliamentarians in developing countries. Mostly qualitativeFootnote 18 and conceptual,Footnote 19 with a few notable exceptions,Footnote 20 this literature has had difficulty substantiating the key assumption that women’s policy priorities are distinct. We contribute to this literature by identifying the domains in which the assumption of women’s distinct preferences is indeed supported (or not) by data from a key developing region. By so doing, we aim to contribute to both theoretical and policy debates regarding the urgency of increasing women’s descriptive representation.
THEORIZING GENDER DIFFERENCES IN POLITICAL PREFERENCES
In advanced, industrialized democracies, existing scholarship has documented that men and women tend to hold distinct political preferences. These differences translate into women’s greater support for policiesFootnote 21 and parties associated with the left.Footnote 22 As a result, a rise in women’s political participation has historically been associated with increased social spending.Footnote 23
While some explanations for this gender gap in political preferences focus on innate differences between the sexes,Footnote 24 such explanations receive less traction today given that divergent preferences have emerged over time,Footnote 25 with women previously holding more conservative preferences than men.Footnote 26 Thus more recent research instead highlights the importance of women’s changing socioeconomic status relative to men,Footnote 27 as well as the socialization of gender roles.Footnote 28 In particular, declining marriage rates, increased risk of divorce and subsequent increased female labor participation are thought to encourage women to support policies that alleviate traditional female responsibilities within the home, such as caring for children, the elderly and the ill.Footnote 29 In short, as women become less reliant on their husbands’ income,Footnote 30 and as they increase their valuation of self-sufficiency,Footnote 31 they also increase their support for state services that are able to free up women’s time.
Thus some gender gaps in political preferences in advanced industrialized democracies, such as women’s greater support for social services, are a sign of the decreasing vulnerability of women or an increase in women’s liberation from constraints placed upon them by family and social norms. As Iversen and Rosenbluth explain in their deconstruction of Becker’s seminal efficiency model,Footnote 32 when the division of labor between husband and wife is maximized and the family constitutes the smallest organizing unit of the economy, women should share their spouses’ policy preferences – for example on taxes and spending.Footnote 33 In this sense, increasing gender gaps in policy preferences can be considered a normatively positive feature, as it marks the erosion of assigning women to traditional family roles. Under such conditions, a growing gender gap in political preferences can be consistent with democratic ideals of freedom, growing equality and the ability to formulate independent preferences.
However, we should take pause before directly applying these lessons to low-income countries. There are key social and economic distinctions in the largely non-industrial societies of the developing world that produce a more theoretically varied set of policy domains. In non-industrial economic activities such as subsistence or small-farm agriculture, men and women may both be engaged in revenue-generating activities that, because of differential access to production and distinct roles in the household, require different inputs. Thus while women will not necessarily share their husbands’ preferences because they are still economically productive, the gender divergences in policy preferences do not imply economic independence, since a woman’s production function is socially constrained, along with control over her own revenue.
The more that gendered social and economic roles constrain women’s choices, the more likely it is that differences in policy preferences signal women’s vulnerability rather than their liberation. There is scant work on gender gaps in political preference in the developing world that allows us to investigate such a claim. Two prominent exceptions, Chattopadhyay and Duflo, as well as Olken, document that women in rural India and Indonesia, respectively, prioritize access to drinking water more than men since they are more likely to incur the opportunity cost of fetching water.Footnote 34 Similarly, Olken finds that men are more likely to prioritize roads and bridges, which attests to their greater mobility.Footnote 35 Gender differences in the prioritization of infrastructure and clean water thus demonstrate gender gaps in policy prioritization that are borne out of constraints on women’s economic opportunities and/or social norms that disadvantage women.
In Africa, the systematic study of gender-based political preferences has focused primarily on attitudes toward regime types rather than policy prioritization.Footnote 36 Using data from fifteen African countries, Logan and Bratton find that women are less supportive of democratic institutions but have relatively similar attitudes towards the role of the state in economic development.Footnote 37 They do not, however, look at gender gaps in other types of policy prioritization. Recently, García-Peñalosa and KonteFootnote 38 replicated the finding that women are less likely than men to support democratic institutions, and attribute this difference to women’s greater risk aversion, which stems from the fact that elections in many developing countries are associated with instabilityFootnote 39 and women bear a higher cost of conflict than men.Footnote 40 In the course of their study, García-Peñalosa and Konte also examine the possibility that gender differences in policy preferences could account for the divergence in support for democracy, which they ultimately rule out. As a result, they report (but do not attempt to explain) gender gaps in policy prioritization.Footnote 41 We thus conclude that it is still unknown whether men and women have divergent policy prioritization in the African context.
We address this gap in the literature by investigating three policy domains for which we have clear expectations regarding how the gender gap varies with women’s economic and social position relative to men. We focus on these domains because they represent ideal types in our framework: where financial dependence and social vulnerability drive women to prioritize a policy domain less than men (infrastructure investment), where these factors drive women to prioritize a policy domain more than men (access to clean water), and where we expect the gender gap in policy prioritization to be less affected by financial independence and social vulnerability (poverty alleviation). Through this focused examination, we demonstrate the utility of our framework as an analytical tool for studying gender gaps in preferences in other policy domains or regions.
Theory and evidence suggest that the social and economic empowerment of women are distinct processes that, while interrelated, operate on different time scales and respond differentially to shocks. While increasing a woman’s financial independence can increase her bargaining power in household decision making,Footnote 42 gendered institutions or asymmetric social normsFootnote 43 ‘may neutralize the bargaining power that women derive from individual resources, by affecting their exit options, their bargaining agency, for example, accepting male authority when they have formally equal rights, their preferences, through adapting these to what is deemed proper for women, and their roles in the household, limiting what can and what cannot be bargained over.’Footnote 44
Women’s social position relative to men tends to be stickier than their relative economic position. For example, the global trend in economic development in recent decades has yielded important gains in female labor force participation, but far fewer improvements in women’s political voice or the incidence of domestic violence against women.Footnote 45 Green et al. similarly find that while increasing access to income improves the economic well-being of women, it does not immediately improve their social standing in the home or their perceptions about women’s rights more generally.Footnote 46 This resistance to social change may be attributed to the self-enforcing nature of norms or the lag between de jure and de facto change, which is particularly relevant in weak states. Within our three policy domains, we will thus examine the impact on the gender gap of both economic empowerment (greater financial independence) and social empowerment (lesser social vulnerability).
Importantly, both forms of empowerment could operate at both individual and country levels. A woman may gain access to assets or income, but remain frustrated by barriers to formal employment. Some studies find that the latter matters even more than the former: women’s absolute level of earnings has a relatively small effect on her bargaining power, while a lower gender wage gap in the local labor market significantly lowers women’s unpaid workload and reduces domestic violence.Footnote 47 Similarly, a woman may increase her formal status by accruing years of education, but remain limited by social norms that circumscribe her role in society. We thus exploit the multilevel nature of our data, which varies within and between countries, to explore both individual- and country-level determinants of the gender gap in policy prioritization.
We start by examining the gender gap in the prioritization of infrastructure investment. Better roads and bridges reduce the cost of trade and commuting time. Similarly, productivity is enhanced when electric power is more reliable. It is reasonable to expect that employed citizens, and those who have greater expectations of future employment, will be more likely to prioritize infrastructure investment. To the extent that prioritizing infrastructure investment results from access to employment, women should care less about infrastructure than men if they are unemployed or have unequal access to employment outside the home. By contrast, the more women that are employed, or expect to be employed, the more likely they are to close the gap in infrastructure prioritization.
Central to our framework is the idea that prioritizing the quality of, for example, roads and bridges, is not simply an economic question but also has a key social component. For example, women who are socially vulnerable are less likely to travel outside their village and thus to prioritize infrastructure investment. If, on average, male constituents start off with higher preferences for infrastructure investment than their female counterparts, it follows that the narrowing of the gender gap would evidence a favorable shift in female empowerment – greater employment opportunities and a reduction in vulnerability. These expectations are presented graphically: Figure 1 shows that the negative gap in prioritization of infrastructure will narrow with economic empowerment, and Figure 2 demonstrates that it will widen with women’s social vulnerability.
Predicted effects of economic empowerment
Predicted effects of social vulnerability
We next evaluate the gendered prioritization of access to clean water, with the expectation that women prioritize it more than men, on average, as has been shown in other developing countries.Footnote 48 Where water is unavailable inside the home – as is the case in 65 percent of our sample (80 percent in rural areas) – accessing water for essential household activities is time and energy consuming. Fetching water is more likely to be a woman’s role in cultures and religions that emphasize traditional divisions of labor between men and women – by one estimate, water for household use is fetched by women in 71 percent of households in Sub-Saharan Africa.Footnote 49 In this domain, women’s greater prioritization of water likely reflects both their social vulnerability and fewer employment opportunities. If, on average, women start off with a greater prioritization of access to water, it follows that an unfavorable shift in female empowerment – fewer employment opportunities and an increase in vulnerability – would be evidenced by the widening of the gender gap, as depicted in Figures 1 and 2.
While women’s economic and social empowerment are expected to condition the gender gap in the prioritization of government investment in infrastructure and water, we expect that these factors will have weaker explanatory power in the prioritization of poverty alleviation. Note that women’s greater support for government welfare in the developed world is focused on providing services, such as early child care and elderly care, that allow women to pursue employment outside the home.Footnote 50 By contrast, our expectation that African women’s greater prioritization of poverty alleviation is insensitive to social and economic empowerment is rooted in the fact that in Sub-Saharan Africa, welfare policies are related instead to the alleviation of suffering (see Afrobarometer responses in Table 1, second panel).
This expectation builds on the social psychological understanding of how the differential patterns of socialization experienced by boys and girls impact their ethical dispositions. For example, Gilligan argues that morality is socially constructed differently for men and women.Footnote 51 Through socialization processes that are reinforced from childhood, women internalize an expectation to care for others and to help protect society’s most vulnerable members. Men, on the other hand, are encouraged to focus on self-fulfillment, which is ostensibly consistent with their roles as breadwinners and protectors of the family.Footnote 52 Thus while we expect that richer and employed women will prioritize poverty alleviation less than poorer unemployed women, the gap between men’s and women’s prioritization should be relatively stable across economic and social strata, as demonstrated in Figures 1 and 2. Our expectation here focuses explicitly on the relative stability of the marginal effect of gender rather than levels of domain prioritization, which are influenced by many additional factors, such as the level of development and state capacity.
In sum, we expect that both financial dependence and social vulnerability will exacerbate gender differences in the prioritization of infrastructure investment and access to clear water, but have little effect on gender differences in prioritizing poverty alleviation. While we ultimately focus on testing theoretical expectations for these three policy areas, we first measure gender gaps in policy prioritization for a wider array of domains in order to observe more general trends in gender-based political preferences.
MEASURING GENDER DIFFERENCES IN POLICY PRIORITIZATION IN AFRICA
We locate our study in Sub-Saharan Africa, where evidence on the gender gap in policy preferences is wanting and yet assumptions about such a gap fuel debates surrounding policies to increase female political representation. Studying this region has several advantages. First, there is substantial variation in the status of women across Africa, which allows us to explore how local conditions might affect gender-based political preferences. Secondly, in exploiting this variation, we are still able to minimize the effects of factors that are relatively common across the continent, such as similar colonial experiences and countries’ relations to the West. Finally, the Afrobarometer is a standardized survey that allows us to compare a representative sample of men and women from a relatively large set of countries.
Using data from the fourth (2008–09) and fifth (2010–12) rounds of the Afrobarometer survey, we evaluate respondents’ policy prioritizations as a function of their gender. The dataset includes 69,640 respondents across the following twenty-seven countries: Benin, Botswana, Burkina Faso, Burundi, Cameroon, Cote d’Ivoire, Ghana, Guinea, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mozambique, Namibia, Niger, Nigeria, Senegal, Sierra Leone, South Africa, Swaziland, Tanzania, Togo, Uganda, Zambia and Zimbabwe.Footnote 53 Samples are randomly drawn to be nationally representative, and stratified by gender to ensure gender parity. Respondents’ policy priorities are captured by the question ‘In your opinion, what are the most important problems facing this country that government should address?’
Each respondent was asked to give up to three open-ended responses, which were coded by the Afrobarometer into one of thirty-three policy domains. We collapse those thirty-three response categories into ten: economy, poverty, infrastructure, health, agriculture, water, education, violence, social and political rights, and other. An additional category (‘none’) identifies respondents who did not indicate any policy priorities. Each category is treated as a binary variable, such that a respondent is coded as prioritizing a policy domain (variable value equals 1) if he or she mentions it in any of the three responses.Footnote 54 This classification scheme is elaborated in Table 1, and summary statistics for all variables discussed below are reported in Appendix Table A.1.
To explore whether policy preferences are structured along gender lines, we run seemingly unrelated OLS pooled regressions, with country fixed effects, separately for each of the ten categories, controlling only for the total number of priority responses given.Footnote 55 We deliberately do not control for covariates on which men and women are likely to differ – such as wealth, education, urbanization and labor market participation – since we are interested in policy prioritizations that reflect such differences in women’s social roles and cultural and economic position. If we control for, say, education, we are unable to capture gaps in preferences that stem from the fact that women have, on average, unequal access to education. The estimated effects of gender (female) are presented in Figure 3 and in Appendix Tables A.2 and A.3.
Impact of gender (female) on policy domain prioritization (pooled analysis)
We find that women are significantly more likely to prioritize government investment in poverty alleviation, health and access to clean water. In contrast, men are more likely to prioritize the management of the economy, infrastructure, agriculture, violence, and social and political rights. Education is the only policy domain for which there is no detectable difference in men’s and women’s prioritizations. While eight of the nine substantive policy categories are prioritized by one gender over the other, these marginal differences are not large, with the exception of poverty (8 percentage points). For all other significant differences, the magnitude is less than 4 percentage points, which is smaller than gender gaps identified in advanced, industrialized countries.Footnote 56
Additionally, these differences are not large enough to significantly change the aggregated rank order of preferences by gender. Using the pooled estimates, Figure 4 plots the predicted probability of policy prioritization by domain for men and women separately. With the exception of infrastructure (second for men but third for women), the overall prioritization of the most important policy domains is quite similar for men and women. In short, we identify relatively small (but statistically significant) gender gaps for most policy domains, but those gaps do not result in meaningful differences in the ranking of government policy priorities.
Predicted probabilities of policy domain prioritization by gender (pooled analysis)
EXPLAINING VARIATION IN POLICY PRIORITIZATION GENDER GAPS
While the results presented above do not suggest large gender differences in policy prioritization, the pooled analyses on which they are based may mask significant variation in gender gaps both across and within countries. Figure 5 plots the distribution of country-level gender gaps (the proportion of women that prioritizes a policy domain minus the proportion of men) by policy domain. The figure demonstrates that the gender gap diverges quite a bit between countries, sometimes even changing sign. This suggests that, for at least some domains, the sources of variation can be traced to country-level factors.
Country-level gender gap distribution by policy prioritization Notes: diamonds refer to the mean gap across the nineteen countries.
Figure 5 also demonstrates that while some domains are much more likely to be prioritized by women in almost all countries (for example, poverty and water) and others by men (for example, economy, infrastructure and agriculture), several policy domains are arguably more gender neutral (such as health and education). The importance of these findings is that the extent to which policy priorities exhibit a gender gap is very much a function of the nature of the policy domain itself. Importantly, the fact that the attributes of a policy domain mediate the extent to which female constituents exhibit different preferences than their male counterparts informed our decision to refrain from attempting to locate a root cause of a single overall ‘total-gap’ measure.Footnote 57
The pooled results may also belie variation within countries in the degree of gender divergence over policy prioritization. For example, if there are gender gaps in different directions across different segments of a society, this would result in a small average gender gap despite large differences among most of the population. We therefore exploit variation both within and across African countries as leverage to explore the relationship between the size of the gender gap and women’s economic and social empowerment. As previously discussed, we focus on three key policy domains – infrastructure, water and poverty – which serve as ideal types in our theoretical framework of how social and economic conditions, at both the individual and country levels, relate to variations in the size of the gender gap.
Measurement of Key Explanatory Variables
We have argued that the existence and size of gender gaps in policy prioritization are a function of financial independence and social vulnerability that operate at both the individual (for example, employment status) and country levels (for example, expectation of employment). In this section we describe the data and measures we use to evaluate empirically the relationship between the size of the gender gap and our explanatory variables in three key policy domains: infrastructure, water and poverty.
Financial independence (economic empowerment)
Following our theoretical framework, we proxy financial independence (economic empowerment) at the individual level with a measure of employment status. Employed is a binary variable that has a value of 1 if the respondent indicates that he or she has a full- or part-time paid job.Footnote 58 At the country level, we create the variable Share Female Employment that measures the proportion of women that reports being employed (taking survey weights into account), which is calculated separately for each country in the sample.Footnote 59 The share of female employment ranges from less than 10 percent in Guinea, Togo, Burundi, Malawi, Niger and Burkina Faso to over 40 per cent in Kenya, Madagascar, Uganda and Ghana. This measure of employment is admittedly imperfect in the context of many African countries, as informal labor and subsistence agriculture are common substitutes for income-paying jobs. However, this sort of measurement error would lead us to under-report rates of female labor participation, biasing us against finding that economic empowerment closes the gender gap in policy prioritization.
Social vulnerability (social (dis)empowerment)
We define social vulnerability as a set of cultural, traditional and religious norms that place individuals at a systematic disadvantage, which further affects their prospects, expectations and coping abilities. Since we are interested in explaining gender-based differences in political attitudes, we focus explicitly on gendered sources of vulnerability. As vulnerability is complex and multi-dimensional, we measure gendered (female) vulnerability for each country as a standardized index that includes the extent and legality of polygamy,Footnote 60 adolescent fertility rateFootnote 61 and average age of first marriage for women.Footnote 62 Following Anderson, these different factors are adjusted such that higher values reflect greater vulnerability, standardized and then combined into a single index using an inverse covariate weighted average.Footnote 63 The resulting Vulnerability Index suggests that women are least vulnerable in Namibia (−1.59), South Africa (−1.56) and Botswana (−1.51), and most vulnerable in Mozambique (0.95), Mali (1.27) and Niger (1.62).
Unfortunately the Afrobarometer does not collect data on the above factors at the respondent level (that is, the age of marriage and of first born, or respondents’ marital status). We therefore proxy social vulnerability at the individual level using a measure of the gender gap in education. We create a binary variable, Education Gap with Avg. Male, coded 1 (that is, vulnerable) if a respondent’s highest level of education is below the median education level for male respondents in his or her age cohort in that respondent’s country, and 0 otherwise. This measure is not ideal, but reasonable given that the Afrobarometer does not ask for personal information that would have allowed for a less noisy measure.Footnote 64
Appendix Figure A.2 shows the relationship between the key explanatory variables at the individual level, while Figure A.3 shows the relationship between the country-level correlates. The relatively weak correlation between female employment and social vulnerability at the country level suggests that these measures indeed capture different underlying constructs of female (dis)empowerment. Appendix Figure A.4 shows a scatterplot of the (un-modeled) relationship between the prioritization of each of the three policy domains of interest and the two key county-level explanatory variables: share of female employment and the vulnerability index. Similarly, Figure A.5 shows the (weighted) mean share of people in country j that prioritizes infrastructure, water and poverty, broken down by all combinations of gender, employment status and vulnerability. It is not easy to use these figures to assess gender gaps rather than levels of prioritization. In the next section, we thus move beyond plotting the raw data to using regression analysis, which allows us to estimate the association between the size of the gap and our explanatory variables of interest.
Estimation
Because we have data at both the individual and country levels, and are interested in the effects of variables at each level, we use a series of mixed-effects linear (multilevel) regressions, one for each of the three policy domains. Because we are interested in the correlates of the gap in policy prioritization and not levels, we interact all variables of interest with an indicator for gender. The coefficient of the interaction term will indicate the degree to which an explanatory variable conditions the extent to which men and women prioritize policies differentially. Because the effect of individual-level characteristics may be conditional on country-level ones (for example, unemployed women may behave differently where prospects for employment are higher or lower), we include a triple interaction term for each set of explanatory variables. Formally, we fit the following model:
$$\eqalignno{ y_{{ij}} =& \alpha {\plus}\beta _{1} F_{{ij}} \!*\!\beta _{2} E_{{ij}} {\plus}\beta _{3} G_{{ij}} {\plus}\beta _{4} F_{{ij}} \!*\!E_{{ij}} {\plus}\beta _{4} F_{{ij}} \!*\!G_{{ij}} {\plus}\beta _{6} S_{j} {\plus} \beta _{7} V_{{ij}} {\plus}\beta _{8} F_{{ij}} \!*\!S_{j} {\plus}\beta _{9} F_{{ij}} \cr &\!*\!V_{j} {\plus}\beta _{{10}} F_{{ij}} \!*\!E_{{ij}} \!*\!S_{j} {\plus}\beta _{{11}} F_{{ij}} \!*\!G_{{ij}} \!*\!V_{j} {\plus}F_{{ij}} \!*\!\gamma _{{ij}} {\plus}F_{{ij}} \!*\!\lambda _{j} {\plus}R_{j} {\plus}\upsilon _{j} {\plus}\epsilon _{{ij}} $$
where y ij is a binary outcome that takes a value of 1 if respondent i from country j reports the policy domain as one of his or her top three priorities for government action,Footnote 65 F ij is an indicator for female, E ij is the respondent’s employment status and G ij is the individual-level measure of education gap. As for country-level input variables, S j is the share of female employment and V j is the continuous vulnerability index, while γ ij and λ j are individual- and group-level controls. Again, because we are interested in the relationship between these control variables and the gender gap rather than just the level of prioritization, we interact each control variable with the female indicator. R j is an indicator variable corresponding to the round of the Afrobarometer (4 or 5). To account for the nested nature of the data we include υ j , a random intercept for country j, and ε ij , which is the individual error term.
We control for individual- and country-level correlates that likely affect the size of the gender gap in policy prioritization. We include both individual- and country-level indicators for Islam following the literature, which finds that Muslim countries are, on average, more conservative, and women are more likely to be assigned, on average, traditional gendered roles.Footnote 66 At the individual level, we use an indicator for whether the respondent identified as being Muslim. The country-level variable simply aggregates responses to produce a continuous variable for the share of respondents identifying as Muslim. We also control for GDP to explicitly separate social vulnerability and relative economic empowerment from a country’s level of development. Finally, we control for the age of respondents using a continuous measure as well as an indicator for whether the respondent resides in an urban community.
RESULTS
In this section we describe the study’s empirical findings, starting with infrastructure and access to water, where we expect the gender gap to close with greater economic and social empowerment. For all regressions, we show results in graphical form in the main text, and in tabular form in the Appendix. In graphing the key results, we focus on the marginal effects of gender conditional on our key explanatory variables, while holding all other variables to their means.
Infrastructure
Figure 6 shows the conditional marginal effects of gender on the probability of prioritizing infrastructure (see also Appendix Table A.4). These results are very consistent with our theoretical framework. Starting with individual-level factors (top panels), we see that both greater financial dependence (unemployed) and social vulnerability (relatively less educated) are associated with large and significant negative gaps (that is, women are almost 5 percentage points less likely to prioritize infrastructure investment). By contrast, both greater financial independence through employment and the reduction of vulnerability through education not only reduce the gap, but eliminate it.
Marginal effect of gender (female) on prioritization of infrastructure
Moving to country-level factors (bottom panels), we find strikingly consistent results. Where the share of female employment is low, women have little expectation of entering the labor market, which can explain their low prioritization of infrastructure investment relative to men. This pattern is reversed with increases in the share of female employment, even controlling for one’s own employment status. Similarly, at high values of social vulnerability, we find that women prioritize infrastructure at significantly lower rates than men – up to 6–7 percentage points. However, this negative gender gap in infrastructure prioritization closes completely at low levels of vulnerability.
Access to Clean Water
Women tend to prioritize access to clean water more than men, arguably due to their greater responsibility to fetch water for the household.Footnote 67 Figure 7 and Table A.5 show the conditional marginal effects of gender on the probability of prioritizing access to clean water. Consistent with our theoretical framework, we find that economic empowerment mediates the relationship between gender and water prioritization – though this relationship is more evident at the country level than at the individual level. For the latter, moving from those employed at a paid job to those unemployed increases the gender gap in water prioritization (Figure 7, top left panel), although this difference is not statistically significant. Similarly, where the share of women employed is high, we do not find any gender gap in water prioritization. Where the share of female employment is low, the gender gap in water prioritization is estimated to be around 3 percentage points (Figure 7, bottom left panel). Contrary to expectations, we do not find a strong association between our vulnerability proxy measures and water prioritization at either the individual or country level.
Marginal effect of gender (female) on prioritization of water
Note that when we add to our base multilevel model a proxy measure of individual wealth, the gender gap in water prioritization slightly closes but is far from eliminated (Table A.5, Column 2). This suggests that our findings do not simply capture the fact that women are likely to be, on average, poorer than men. Instead the gap is produced by a combination of social norms and the relative opportunity costs of fetching water in contexts where women have very few opportunities to work outside the home.
Poverty Alleviation
On average, women prioritize government investment in poverty alleviation measures significantly more than men (Figure 3). In fact, as Figures 5 and A.1 show, this is the case in all twenty-seven African countries in our sample. Moreover, the bottom panel of Figure A.4 further reveals that, though the share of citizens prioritizing government investment in poverty alleviation decreases in female employment and increases in female vulnerability, the size of the gap appears constant across the entire range of values of our key country-level inputs. Similarly, looking at the raw individual data, Figure A.5 shows a relatively unchanged gender gap in all combinations of employment and vulnerability.
Indeed, we have argued above that the gender gap in (rather than levels of) poverty alleviation prioritization has more to do with gendered socialization processes than with women’s structural position in the economy (for example, her employment status) and society (for example, her vulnerability). Testing this proposition more rigorously using a multilevel model, Figure 8 shows that the conditional marginal effect of gender is not substantively different across employment status, or for more and less vulnerable respondents. These findings are also evidenced in the country-level factors: the marginal effect of gender does not change across the range of female employment values, or across the entire range of our social vulnerability index.
Marginal effect of gender (female) on prioritization of poverty
Together, these findings suggest that, across country- and individual-level conditions, women in the twenty-seven Sub-Saharan African countries in our sample feel more responsible for demanding that the state alleviate the burden of poverty. Importantly, as Table A.6, Column 2 shows, this gap cannot be explained away by the fact that women are, on average, poorer than men. When adding an individual-level control for respondents’ wealth, the marginal effect of being female hardly changes. In fact, Figure A.8 suggests there is a threshold of poverty prioritization that women never fall below. While unemployed women’s prioritization of poverty starts highest where female employment share is low and converges with employed women’s level of prioritization where the rate of female employment is high, the level for employed women stays constant across this range, never falling below about 0.42.
The consistency of these results with findings from advanced industrialized democraciesFootnote 68 suggests that there are important determinants of the gender gap in poverty alleviation prioritization that are not captured by economic factors. Indeed, the extent to which this variation stems from gender-based divergent socialization processesFootnote 69 offers exciting opportunities for future empirical research.
DISCUSSION
This study is motivated by the need to better understand a key assumption underlying the call for greater descriptive representation: that there are meaningful gender gaps in policy priorities, and that the current state of politics does not sufficiently represent these divergent preferences. We have demonstrated that important gender gaps do exist, and that economic and social factors affecting female empowerment help shape those differences. In this section, we first discuss the relative importance of individual- and country-level indicators of women’s empowerment. While the discussion and empirical exploration are preliminary, we think it is important to distinguish theoretically between gender differences in policy prioritization that are driven by functional differences in gender roles at the individual level and those that are driven by future expectations of change based on country-level attributes. We then discuss the relationship between gender differences in political prioritization and the rate of women’s representation and participation across countries. We present only simple correlations, and thus make no causal claims about why participation and representation are related to gender gaps in policy priorities. Nevertheless, we think it is useful to explore how these factors are related in order to motivate future research on women’s representation and participation in Africa.
Individual- vs. Country-Level Characteristics
The inclusion of both individual- and country-level predictors in our analysis of domain-based gender gaps is motivated by distinct theoretical expectations of relationships at each level. A country-level characteristic such as the share of women employed might drive the gender gap in policy prioritization if it changes the underlying expectations of either male or female citizens regarding the potential for women to be employed in the future. This effect could be independent of an individual-level characteristic such as whether a particular woman is employed. Conversely, a gender gap may instead be driven only by the actual employment status of women, where the general state of opportunities for women has no moderating effect.Footnote 70
While we do not evaluate the relative importance of individual- and country-level factors empirically, our main results suggest that it may differ for water and infrastructure. In the prioritization of infrastructure, we see that indicators of economic and social empowerment at the individual level have strong effects even when controlling for country-level indicators. While we also see substantive effects at the country level, these could simply be driven by the mechanical aggregation of individual-level attributes. The predicted probabilities of prioritizing infrastructure by gender (Appendix Figure A.6) indeed show that employed females look more like employed males than unemployed females, even at low rates of female employment. The closing of the gap where females are employed at higher rates thus appears to be driven by the underlying share of employed women increasing and the aligning of preferences between employed men and women. This is suggestive of gendered preferences for infrastructure representing a more functional gap, driven by individual-level preferences rather than expectations set by country-level characteristics.
For the prioritization of water, we find no significant effect of individual-level employment, but do see that the gender gap closes when more women are employed. This suggests that water prioritization may instead represent a gap driven more by expectations about female employment (which may, in turn, affect norms about gender roles) than by individuals’ employment status. At low rates of female employment, the gender gap is driven by employed men (very low prioritization) and unemployed women (very high prioritization), as demonstrated in Figure A.7. This gap closes as more women participate in the labor force, but this occurs because employed men care more about water, not because employed women care less. Men, in this case, are likely re-optimizing the value they place on household responsibilities now that women have a higher opportunity cost for fetching water, or even outsourcing domestic tasks affecting both income earners.Footnote 71
Whether a gender gap in preferences is purely functionally driven (by being employed) or expectation driven may have different implications for policy. In the former case, we would not expect elite women to sufficiently represent the preferences of women in functionally different situations who had distinct preferences. In the latter case, the very fact that some women in society have better opportunities changes the preferences of women in more precarious situations, such that women of all economic status have more similar preferences. It also seems reasonable to expect greater mobilization and policy coordination on the basis of expectations-driven preferences compared to functional-based preferences – a possibility that could serve as an interesting avenue for future research.
Gender Gaps in Political Representation
Though women are formally represented in politics at relatively high rates in Africa, there still exists significant variation across countries. Past research has found that patriarchal cultures and clientelism reduce female access to positions of power, while electoral institutions such as proportional representation increase such access.Footnote 72 In this section we examine the relationship between formal descriptive female representation and the gender gap in policy prioritization. This relationship is important to document, because the argument for increasing formal female representation is stronger the more political preferences diverge between male and female constituents.
We measure the degree of female political representation using the share of female members of parliament – in the lower or single chamber – in the year prior to the Afrobarometer data collection.Footnote 73 Within our sample of twenty-seven countries, South Africa has the highest proportion of women (45 percent), while Nigeria has the lowest (7 percent).
In Figure 9, we present the relationship between female representation and the aggregated (total) gender gap in policy preferences, summing the absolute difference between the share of female and male constituents that prioritizes each policy domain. The picture that emerges in the aggregate is very clear: places that exhibit a large divergence of gender-based preferences also have the smallest share of female members of parliament.Footnote 74 In Appendix Figure A.24, we present the relationship with each policy domain separately, which shows the strongest correlations for poverty, education and violence.
Relationship between gender gap in (aggregated) policy prioritization and female (descriptive) representation
These patterns suggest that where the representation of women is most needed due to gender-based differences in political prioritization, women are the least likely to hold political office. Of course, the structure of our data does not permit claims about the causal relationship between these two gender gaps. Admittedly, omitted features of countries (such as overall gender inequality) may be driving gender gaps along both dimensions. Regardless of why they are correlated, our goal is to point out the unfortunate fact that successes in gender parity in representation are not occurring in the places with the greatest need for more gender-based substantive representation – a reality that may be of interest to policy makers and practitioners alike.
Gender Gaps in Political Participation
While the share of women in parliament is one indication of women’s voice in politics, female political participation is another. Here, we replicate the above analysis to evaluate whether there are larger gender gaps in political participation where gender gaps in policy prioritization are greatest. Existing research has documented a sizable gender gap in political participation acrossFootnote 75 and within African states,Footnote 76 though again, this gap varies substantially across countries.Footnote 77 In the context of our study, a gender gap in political participation has the most negative implications for democratic representation when men and women have divergent preferences.
We measure political participation by combining the following set of variables using a standardized weighted summary index: membership in voluntary association or community group, attendance at community meetings in the past year, joining others to raise an issue, attending a demonstration or protest march, voting in the most recent general elections, and contacting one’s local government councilor, MP or other government official.Footnote 78 We calculate the average individual-level measure of political participation for men and women separately in each of the twenty-seven countries, and then take the difference between the average for men and women as our measure of the size of the gender gap in political participation. The mean gender gap in political participation is equal to 0.18 standard deviations, with small differences in Namibia (0.04), South Africa (0.04) and Botswana (0.6) and much large gender gaps in Cameroon, Burkina Faso, Liberia, Mali and Ghana (all above 0.26 standard deviations).
Summing over all domains, Figure 10 shows that larger divergences in policy prioritization are associated with larger gender gaps in political participation.Footnote 79 Appendix Figure A.23 shows how this relationship holds across different policy domains. For domains typically preferred by men – especially infrastructure, agriculture and violence – the larger the gap between men and women (more negative values on the y-axis), the larger the gap in political participation. For domains typically preferred by women, such as poverty and water, the larger the gap (more positive values on the y-axis), the larger the gap in political participation.
Relationship between gender gap in (aggregated) policy prioritization and gender-gap in political participation
These findings, like those above, suggest that the greatest barriers to female participation are present exactly where their removal is most needed due to women’s distinct preferences. While we again stress that our data do not allow us to make causal claims, the pattern is nonetheless worrisome. For this reason, future work should evaluate whether closing the gender gap in political participation results in less gendered policy divergence or in better representation of those divergent preferences.
CONCLUSION
This article explores the size and direction of gender gaps in the prioritization of government action on a broad set of policy domains in Sub-Saharan Africa. Overall, we find modest differences between men and women in terms of policy prioritization when pooling across a relatively large number of countries, but more meaningful differences when disaggregating the analysis between and within countries. Drawing from the existing literature on gender gaps in industrialized countries, we estimate gender gaps across policy domains and then evaluate how economic and social factors shape those differences. Using variation in the size of the gender gap both across and within twenty-seven African countries, we report a set of patterns that is consistent across both levels.
Measures of women’s financial independence and social vulnerability correlate as predicted with the size of the gender gap for a policy domain typically favored by men (infrastructure investment) and one typically favored by women (access to clean water). Specifically, we find that greater financial independence of women, proxied by female labor participation, closes the gaps between men’s and women’s prioritization in both policy domains. These results are consistent with earlier research that attributed shifts in women’s preferences in response to changes in their economic standing within the home and within society at large. In addition, women’s vulnerability widens the differences between genders on prioritizing infrastructure. Further, we find that individual- and country-level demographic variables have little explanatory power with respect to a gender gap in the prioritization of poverty alleviation. We have argued that this finding is consistent with theories focusing on differences in gender-based socialization processes in early childhood.
We also observe some evidence that individual- and country-level correlates matter differently in different policy domains. The gender gap for infrastructure appears to be more sensitive to individual-level correlates, pointing to a more functional gender gap, while country-level correlates seem to be more important for the gender gap in water prioritization, evidence of a more expectations-driven gender gap. While further research is required to validate this distinction, it points to an interesting extension of our theoretical framework. Rather than just differentiating the direction of the gender gap across policy domains, it may also be that gaps that are based on functional differences between men and women (preferences over infrastructure, which depend on interactions with the state) are driven by individual-level determinants (whether or not one is employed) more than gender norms. However, gaps that are based on sociocultural or expectations-driven differences between men and women (who fetches water, which depends on intra-household bargaining and resource allocation) are driven more by socially defined expectations of the role of women than functional differences. In the former case, the gap is more likely to close when more women change their status, for example by becoming employed. In the latter case, the gap is more likely to close when men (or women) change their expectations in reaction to a new distribution of power or resources within the household – for example, if men and women share responsibility for accessing water.
Finally, we explored the relationship between gender-based divergence of policy priorities and the barriers that women face in aggregating those preferences. We find that there is a strong, positive correlation between the gender gap in preferences and barriers on women’s preference aggregation in the form of women’s participation and women’s descriptive representation. In other words, it is exactly in countries where women and men have the most divergent preferences that we also find that women face the greatest difficulties influencing government policies through political participation or by electing women into public office.
Given that there are meaningful gender differences in preferences, at least in some contexts, an important avenue for future research is to examine whether greater descriptive representation actually leads to better substantive representation. Evidence from India suggests that female policy makers do better serve the interests of female constituents because of aligned preferences,Footnote 80 but whether this is the case in Africa is an open question. Many policy interventions on the continent are aimed at increasing the number of female representatives (for example, via quotas), under the assumption that such policies will lead to the better representation of ‘women’s issues.’ However, the results of this article suggest that at least some such interests, such as improved access to safe water, are only women’s issues in contexts where women are vulnerable and dependent. Thus in addition to improving representation for women, perhaps policy should also seek to alleviate the causes of divergent gender preferences.
