The politics of presence

The expectation that female politicians will better represent the concerns of female constituents requires men and women to diverge in terms of their policy preferences (Phillips, Reference Phillips1995). Empirical evidence indeed suggests women tend to favor redistribution more than men, even controlling for political ideology (Alesina and Giuliano, Reference Alesina, Giuliano, Benhabib, Bisin and Jackson2011; Iversen and Rosenbluth, Reference Iversen and Rosenbluth2006; Finseraas et al., Reference Finseraas, Jakobsson and Kotsadam2012). Scholars have also identified gender gaps with respect to partisan alignment and voting behavior (Inglehart and Norris, Reference Inglehart and Norris2000; Corder and Wolbrecht, Reference Corder and Wolbrecht2016). In low-income countries, recent scholarship suggests that divergent preferences may be even more common (Gottlieb et al., Reference Gottlieb, Grossman and Robinson2016; Clayton et al., Reference Clayton, Josefsson, Mattes and Mozaffar2019).

To explain these divergent preferences, scholars highlight differences in the social structures in which men and women live (Khan, Reference Khan2017; Sapiro, Reference Sapiro1981; Mansbridge, Reference Mansbridge1999), including the fact that women are socialized to be caretakers of others (Hutchings et al., Reference Hutchings, Valentino, Philpot and White2004). Thus, they tend to be more directly concerned with child health and nutrition (Box-Steffensmeier et al., Reference Box-Steffensmeier, De Boef and Lin2004; Sapiro, Reference Sapiro1998). Therefore, women disproportionately suffer as a result of weak reproductive and child health services (Bhalotra and Clots-Figueras, Reference Bhalotra and Clots-Figueras2014). Such concerns often lead women to favor policies that improve these outcomes (Duflo, Reference Duflo2012). Importantly, Gottlieb et al. (Reference Gottlieb, Grossman and Robinson2016) show that in contexts where women are more vulnerable, the gender gap in preferences widens. Further, the changing roles of women in societies as a result of declining fertility rates, higher divorce rates, and increasing independence of women from their husbands as they enter the workforce also motivate women to support policies that alleviate their traditional roles of taking care of others (Edlund and Pande, Reference Edlund and Pande2002).

Divergent preferences can have real consequences for the enactment and implementation of policies. For instance, the introduction of female voting rights in early 20th century United States qualitatively changed the the political agenda, leading to a sharp increase in health expenditure and a drop of 8–15% in child mortality (Miller, Reference Miller2008). In Africa, Gottlieb et al. (Reference Gottlieb, Grossman and Robinson2016) find that women are significantly more likely to say that health is a top priority the government should address and invest in improving.

Gendered policy preferences can manifest in campaigns and party platforms, with female politicians campaigning more on issues that matter to women, and their election helping to ensure that such concerns are represented in government. Indeed, empirical studies find female candidates talk more frequently about social issues during campaigns, including healthcare, when compared to their male counterparts (Kahn, Reference Kahn1993; Evans and Clark, Reference Evans and Clark2016; Mechkova and Wilson, Reference Mechkova and Wilson2019).

Further, we can still expect female politicians to better represent the interests of female voters even if they campaign on platforms that are not particularly gendered or on party tickets that promote broader interests. As Phillips (Reference Phillips1995: p. 44) points out, “representatives… have considerable autonomy, which is part of why it matters who those representatives are.” As she argues, under a strict definition of representation, politicians only truly represent their constituents on the issues that were explicitly debated during the election campaign. On all other issues, they necessarily rely on their own judgement or prejudice. If we expect that women and men tend to differ on average in terms of their judgments and prejudices, female constituents should be better represented by female politicians even if those politicians do not campaign on “women’s” issues.

Female legislators can act to promote women’s interests by putting women’s issues on the political agenda (Devlin and Elgie, Reference Devlin and Elgie2008), bringing them up during parliamentary debates (Yoon, Reference Yoon2011), and building coalitions across party lines, with grass-root movements and with their male counterparts (Johnson and Josefsson, Reference Johnson and Josefsson2016). Ultimately, this concentrated effort on part of female parliamentarians helps for the adoption of policies that address the issues women face through adopting specific legislation (Wang, Reference Wang2013) and allocating more money from the budget for specific programs that benefit women (Clayton and Zetterberg, Reference Clayton and Zetterberg2018). Importantly, this evidence for the different actions women parliamentarians take comes from case studies of democratic and non-democratic contexts.

Considerably fewer studies have examined how female representation affects the outcomes of public policy – particularly in less developed settings. Notable exceptions include Bhalotra and Clots-Figueras (Reference Bhalotra and Clots-Figueras2014), who find increased women’s political representation is associated with improved public provision of antenatal and childhood health services in India. Also in India, Chattopadhyay and Duflo (Reference Chattopadhyay and Duflo2004) show that female local leaders tend to initiate projects that reflect their female constituents’ concerns. Looking across 102 developing countries, Swiss et al. (Reference Swiss, Fallon and Burgos2012) find that descriptive representation is associated with increased immunization rates and infant and child survival.

Diagonal accountability and substantive representation

While shared preferences help to explain why female politicians may be more likely to implement policies that benefit female citizens, there are also various instances in which descriptive representation does not beget substantive representation. We argue that female politicians will be more likely to implement policies that benefit women when they feel the pressure from their female constituents – that is, when representatives are held accountable to those they purportedly represent. We follow Goetz et al. (Reference Goetz, Cueva-Beteta, Eddon, Sandler, Doraid, Anwar, Bhandarkar and Dayal2008: p. 2) and understand accountability as involving “assessment of the adequacy of performance, and the imposition of a corrective action or remedy in cases of performance failure.”

Such a conceptualization is perhaps most easily understood in the context of a democracy, where citizens-as-voters assess the adequacy of their elected politicians’ performance and then act to correct performance failure by voting them out of office. However, citizens can still impose meaningful corrective actions even when politicians are not elected through free and fair elections by taking various civic actions (Lindberg, Reference Lindberg2013). In addition, civil society and media actors play an equally if not more important role in many settings. Through what has been termed “diagonal accountability,” such actors may constrain governments indirectly by enhancing the effectiveness of other accountability actors (political parties, judicial branches, supreme audit institutions, etc.), or directly by pressurizing them through media or advocacy campaigns (Goetz and Jenkins, Reference Goetz and Jenkins2005; Mechkova et al., Reference Mechkova, Lührmann and Lindberg2019). Indeed, the presence and actions of women’s groups have been shown to play an important role when it comes to the enactment of policies advancing women’s priorities (Htun and Weldon, Reference Htun and Weldon2010; Shiffman and Smith, Reference Shiffman and Smith2007). Relatedly, Dovi (Reference Dovi2002) posits that descriptive representation will confer greater benefits when formal and informal interactions exist between representatives and those being represented.

In many settings, we expect that female citizens – and the diagonal accountability actors who represent their interests – will be more likely to hold female politicians to account. This relates to Pitkin’s notion of symbolic representation, which implies shifts in attitudes and habits. Having more women in positions of political power can break down widely held perceptions that politics is a man’s game, which in turn can improve women’s awareness and participation.

Formal representation and institutional rules

We also expect that formal representation will condition the nature and degree of answerability and sanctions when it comes to achieving gendered accountability. In particular, we expect the following rules and procedures (institutions) to matter: gender quotas, type of electoral system, and democracy.

Quotas have been argued to create a “mandate effect,” obligating female legislators elected under such institutions to act on behalf of women (Franceschet and Piscopo, Reference Franceschet and Piscopo2008). Furthermore, by ensuring that a non-negligible number of women are present in the legislature, quotas can improve the likelihood that the interests of “women-as-group and women-as-differentiated-individuals” will be represented (Franceschet and Piscopo, Reference Franceschet and Piscopo2008: p. 396). The importance of “critical mass” has been embraced by influential international institutions such as the United Nations (UN Women, 2020) though scholars caution against the notion of a “tipping point” at which change automatically occurs (Childs and Krook, Reference Childs and Krook2009). That said, we may still expect that legislatures featuring greater proportions of women should on average promote outcomes more favorable to them, following a logic similar to that outlined in Section 2.1. Indeed Clayton and Zetterberg (Reference Clayton and Zetterberg2018) find that when quota adoption results in a substantial increase in the number of women in parliament, countries see a shift in government priorities – increasing health budgets at the expense of military spending. Similarly, Swiss et al. (Reference Swiss, Fallon and Burgos2012) find that countries where women make up 20% or more of the legislature are more likely to experience improvements in child health.

Previous research has also shown that electoral systems with some degree of proportionality elect more women to parliament (Matland and Studlar, Reference Matland and Studlar1996). Whereas in majoritarian systems, parties can only nominate one candidate per district, making female candidates compete against existing interests within a party that may be dominated by men, proportional systems allow more flexibility to nominate women (Matland, Reference Matland1998). These systems also provide less of an advantage to incumbents, reducing the barriers to entry for female politicians (Thames and Williams, Reference Thames and Williams2010). Proportionality should matter for substantive outcomes as well, since such systems tend to generate higher levels of policy congruence between voters and governments and promote policy responsiveness (Schwindt-Bayer and Mishler, Reference Schwindt-Bayer and Mishler2005; Miller et al., Reference Miller, Pierce, Thomassen, Herrera, Esaisson, Holmberg and Wessels1999; Stokes, Reference Stokes2001; Powell, Reference Powell2000).

Finally, democracies have more channels through which citizens can pressure their political representatives. As suffrage expands and poorer people are allowed voting rights, the preferences of the median voter shift toward introducing higher taxes and redistribution (Acemoglu and Robinson, Reference Acemoglu and Robinson2005; Besley and Kudamatsu, Reference Besley and Kudamatsu2006). This shift in preferences incentivises governments to provide public goods to a larger pool of people or else they face the possibility of not being re-elected (Olson, Reference Olson2000). Furthermore, in democracies, state officials are more likely to be selected based on meritocratic reasons (Besley and Kudamatsu, Reference Besley and Kudamatsu2006). This in turn makes it more likely that qualified and impartial officials will implement the designed policies by the government. Empirically, the literature has found a strong relationship between democracy and development (Gerring et al., Reference Gerring, Thacker and Alfaro2012; Wang et al., Reference Wang, Mechkova and Andersson2018), with competitive elections as the main mechanism. Thus, extending the argument made above, we expect that in democracies women parliamentarians will also be more likely to implement policies in the interest of their (female) constituents. ^{Footnote 2}

Gendered accountability in Africa

Our theoretical framework is informed primarily by studies of Western Europe and the United States. However, sub-Saharan Africa is characterized by a distinct political history and institutions. That said, there are reasons to believe that certain assumptions are potentially more credible in the African context than in settings with more institutionalized politics. Our theory currently has no role for political parties, which can be important when it comes to selecting women to run for political office and supporting their successful campaigns (Kunovich and Paxton, Reference Kunovich and Paxton2005; Cowell-Meyers, Reference Cowell-Meyers2016). However, political parties in Africa tend to be less important when it comes to aggregating societal interests and serving as a vehicle for candidates to advance those interests. Rather, African parties frequently represent particular ethnic identities or personal interests (Elischer, Reference Elischer2013). ^{Footnote 3} This grants representatives in most African political systems considerably more autonomy than their counterparts in more consolidated party systems. As noted above, it is precisely this autonomy that makes representatives’ personal characteristics (e.g., gender) important when it comes to promoting policies that serve the interests of their constituents.

Furthermore, non-electoral forms of political participation are particularly important in many African countries, where elections often fail to serve as a source of accountability due to a lack of programmatic parties and open and informed public debate (Joshi and Houtzager, Reference Joshi and Houtzager2012). Rather, what often matters more in such contexts is the extent to which civil society and media actors constrain government, either by enhancing the efforts of other accountability actors or by exerting direct pressure (Goetz and Jenkins, Reference Goetz and Jenkins2005; Mechkova et al., Reference Mechkova, Lührmann and Lindberg2018).

Moreover, gender roles in many African societies are highly circumscribed. For example, Kes and Swaminathan (Reference Kes, Swaminathan, Blackden and Wodon2006) analyze time-use surveys in Benin, South Africa, Madagascar, Mauritius, and Ghana and find that women devote substantially more time to care and domestic activities than men and considerably less time to market activities.

Dependent variables

Our analysis focuses on the outcomes of two policy areas that we argue women disproportionately prioritize in our study setting: infant and child mortality. ^{Footnote 4} We use data for infant mortality from Gapminder, spanning 1950 until 2015, and for child mortality from the World Bank’s World Development Indicators, available from 1960 to 2014. Instances of infant and child deaths have been substantially reduced in recent years. However, progress has varied. The lowest rate of infant mortality observed in our data is 11.7 deaths per 1,000 cases in the Seychelles in 2015, and the highest for that year is 96 per 1,000 cases in Angola.

Main explanatory variable: descriptive representation

We operationalize descriptive representation as the percentage of women in parliament, for which data are available from 1900 to 2017 from the V-Dem data set. Rwanda leads the region with a 63.8% of women in parliament in 2015. At the other end, the share of women in Comoros’s parliament was 3% in 2015. As an alternative measure of women’s descriptive representation, we also examine the proportion of women in ministerial-level positions.

Women’s participation

To measure women’s participation, we use V-Dem’s Women civil society participation index, which captures three related aspects: the extent to which women are free to openly discuss political issues, can form and participate in civil society organizations, and are represented in the ranks of journalists (Coppedge et al., Reference Coppedge, Gerring, Knutsen, Lindberg, Teorell, Altman, Bernhard, Fish, Glynn, Hicken, Lührmann, Marquardt, McMann, Paxton, Pemstein, Seim, Sigman, Skaaning, Staton, Wilson, Cornell, Gastaldi, Gjerløw, Ilchenko, Krusell, Maxwell, Mechkova, Medzihorsky, Pernes, von Römer, Stepanova, Sundström, Tzelgov, Wang, Wig and Ziblatt2019a, Reference Coppedge, Gerring, Knutsen, Lindberg, Teorell, Altman, Bernhard, Fish, Glynn, Hicken, Lührmann, Marquardt, McMann, Paxton, Pemstein, Seim, Sigman, Skaaning, Staton, Cornell, Gastaldi, Gjerløw, Mechkov, von Römer, Sundtröm, Tzelgov, Uberti, Wang, Wig and Ziblatt2019b; Pemstein et al., Reference Pemstein, Marquardt, Tzelgov, Wang, Medzihorsky, Krusell, Miri and von Römer2019).

Institutional rules and procedures

Our theoretical framework also anticipates that certain institutional conditions should affect the realization of gendered accountability. Specifically, we consider the existence of gender quotas and whether there is a mechanism to enforce these in practice. We also examine how our results vary given the type of electoral system that a country has in place. Finally, we test whether the influence of having more women in parliament depends on the country being a democracy. To capture how democratic a country is, we use a measure of how free and fair elections are. ^{Footnote 5}

Endogeneity and other potential determinants

Due to the nature of the data, it is difficult to disentangle causality in the relationship we are studying. We follow best practices from previous research to address the two main endogeneity concerns: reversed causality and omitted variable bias.

Reversed causality could be an issue in our analysis to the extent that high infant and child mortality rates could drive voters to choose female candidates in elections, as citizens might believe that female candidates are in better position to address such issues. To mitigate that, our main models use 5-year lags of the independent variables. This lag structure corresponds to the election cycle in many countries and also reflects the fact that our outcome variables are likely to change slowly as the implementation of any policy is time consuming. We also estimate system-GMM models with 5-year average panels, which are designed to address endogeneity concerns and to analyze relationships with slow-changing variables. ^{Footnote 6}

We expect variation in progress toward achieving women’s policy priorities to reflect a number of other country-specific factors. We estimate models with women in cabinet and gender quotas as main explanatory variables to rule out the possibility that changes in women’s standing in society are driving the results. We also include in our models several variables which could potentially influence both the dependent and independent variables. The rationale for including these variables is presented below, and the operationalization is provided in the Appendix.