The relationship between women’s representation and
gender norms
Master Thesis
15-06-2020
Word count: 9854
Pages: 31
Dorine klein Gunnewiek, s1812351
Master Political Science: Nationalism, Ethnic Conflict and Development
Department of Political Science
Leiden University
Instructor: Dr. S.P.A. Chauchard
Second reader: Dr. O.B.R.C. van Cranenburgh
Abstract
Over the last two decades, Latin-America and Africa have seen an increase in the number of women in parliament. This thesis will analyse whether an increase in descriptive representation leads to a change in gender norms regarding political equality. An increase in the percentage of women in parliament might impact people’s opinions on the capabilities of female leaders. People might view female leadership more favourably if they witness an increase in female members of parliament. This hypothesis will be addressed by running logistic regression analyses, using data from the Afrobarometer and the Americasbarometer. Surprisingly, this thesis will show that an increase in women’s representation in Africa is associated with an increase in the number of people who think that men are better leaders than women. This effect is small but significant. In Latin-America, this relationship seems to be similar, however, these results are not significant. These findings show that just increasing the number of women in parliament is not enough to substantially influence gender norms regarding political equality positively.
1 1. Introduction
Throughout the last two decades, there have been many development aid policies aimed at increasing women’s representation in the developing world, ranging from promoting the adoption of gender quotas to introducing education programs aimed at women (Krook & Norris, 2014). This seems to have worked: the percentage of women in Sub Saharan African parliaments increased from an average of 11.6 per cent in 2000 to an average of 24 per cent in 2019. In addition, the percentage of women in Latin American parliaments increased from an average of 15.2 per cent in 2000 to an average of 31.6 in 2019 (Worldbank, 2020).
It is important to understand the fundamental assumption behind the urge to increase women’s representation. These policies are not implemented just because they are assumed to increase women’s representation; they are implemented precisely because this increase in women’s representation is argued to achieve significant effects, such as an increased focus on women’s issues. This thesis will assess whether an increase in women’s representation influences gender norms regarding political equality in society. If more women are elected to national legislatures, does this change gender norms in society regarding the political position of women? Therefore, my research question is: “What is the influence of a change in women’s representation on gender norms regarding political equality?”
Research shows that an increase in descriptive representation impacts both behaviour and attitudes of women (Beaman, Duflo, Pande, & Topalova, 2012; Campbell & Wolbrecht, 2006; Paola, Scoppa, & Benedetto, 2014; Mansbridge, 1999). An increase in percentage of women in parliament might impact people’s opinions on the capabilities of female leaders. People might view female leadership more favourably if they witness an increase in female members of parliament. To test this hypothesis, this thesis employs several logistic regression analyses, drawing on data from the Afrobarometer and the Americasbarometer. Contrary to the expectations, the research indicates that, in Africa, an increase in the number of women in parliament is associated with an increase in the number of people who think that men are better leaders than women. In Africa, an increase in women’s representation thus negatively impacts gender norms regarding political equality. In Latin-America, a similar relationship emerges, however, this is not statistically significant. In addition, an increase in the percentage that indicates a change in representation is associated with an increase in the number of people who think that men are better leaders than women as well. Again, these findings are statistically significant for the African data, but not for the Latin American data. These findings could have
2
important policy recommendations. They show that just increasing the number of women in parliament is not enough to substantially influence gender norms regarding political equality positively. Therefore, policies should not be aimed solely at increasing the number of women in parliament.
This thesis is organised as follows: The next section provides a theoretical framework containing already existing research on gender norms, women’s representation, and the relationship between the two. This section also displays the three different hypotheses that follow from the research question and the literature. Section 3 will discuss the research methods and in section 4 the results will be displayed. Lastly, section 5 summarises the main findings, provides possible explanations for these findings, examines potential policy implications and features suggested areas for further research.
3 2. Theoretical framework
2.1. Gender norms
Gender norms are ideas people form about how men and women in society should act and interact (United Nations, 2015). Gender norms are not fixed; they are fluid and subject to change by our observations of the day to day behaviour and attitudes of men and women in society (Carli, 2015). Thus, gender norms are a two-way street: They influence how we think the world should look, yet they are also influenced by what we observe in this world.
Gender norms can be very harmful; they can perpetuate stereotypes about the role of men and women in society, which often leads to unequal treatment, for example when women are not able to live up to their full potential because they are expected to stay at home (Chandra-Mouli, Plesons, & Amin, 2018). However, because gender norms are influenced by our observations of the world around us, this means that if we are able to change this world, our observations change and consequently, our gender norms might change as well. There is a broad literature that looks into the effect of gender norms on various dependent variables, mainly associated with health-related factors, such as the influence on HIV testing (Nanda et al., 2018; Okigbo, et al., 2018; Weber, 2019). In this research, however, gender norms are the dependent variable. Because gender norms are such a broad concept, there is a lot of research on how (different types of) gender norms are affected in society (Burns, Gallagher, & Katherine, 2010). For example, gender norms might be affected by access to the Internet (Shamaileh, 2016); visibility of traditional gender roles (Scharrer & Blackburn, 2018), but also by the way people identify themselves and to which social group they belong (Conover, 1988). However, no large cross-regional research on the influence of women’s representation on gender norms has been conducted. Therefore, this thesis will address that question.
2.2. Women’s representation
A lot of research on the consequences of an increase in women’s representation has already been conducted. In political science, there is a wide array of theories that assess these consequences. According to Anne Phillips (1998, p. 228), these arguments are divided into four groups: Arguments that equal representation is only fair based on principles of justice; arguments that equal representation will boost democracy; arguments that female politicians could act as role models and arguments that female politicians will represent people and interests that would be overlooked otherwise. The first theory argues that equal representation should be strived for because of justice principles. This is a normative argument; it is not
4
concerned with the empirical results of an increase in women’s representation, which is what is addressed in this thesis. Thus, although the relevance of this argument cannot be overstated, it is not what this thesis will address.
The second theory argues that equal representation will boost democracy, because, at its core, democracy means ‘rule by the people’ (Dahl, 2020). According to this theory, equal representation will bridge “the gap between representation and participation” (Phillips, 1998, p. 228). The main assumption behind this theory is that if more women are elected to national legislatures, this might increase behaviour that is associated with a strong democracy. For example, it could influence voter turnout or support for democracy, both characteristics that are associated with a strong democracy (Hajnal, 2010; Grassi, 2011). Turnout of women might be influenced when women feel they are actually able to vote for someone they feel represents them. Evidence from Italian municipalities corroborates this assumption. Paola, Scoppa, and de Benedetto (2014) show how voter turnout under women decreased less in places where gender quotas increased the number of women in the political body than in places where no gender quotas were implemented.
The third theory argues that female politicians could be role models for other women in society. The main assumption behind this is that when women are exposed to female leaders, these female leaders will act as role models, which manifests itself in two ways. First, the presence of women in national legislatures might lead to a change in the political behaviour of women. For example, Beaman et al (2009) show how women are more likely to run for office when the leader of their council is female. Other studies show that exposure to female leaders also increases the likeliness that adolescent girls and women will discuss politics with friends or family (Campbell & Wolbrecht, 2006; Wolbrecht & Campbell, 2007). However, evidence from the US shows there might be mitigating factors. Mariani, Marshall & Mathews-Schultz (2015) show how the increased visibility of Hillary Clinton and Nancy Pelosi in the US led to increased levels of political involvement among young democratic women, while the increased visibility of Sarah Palin did not have the same effect on young republican women. Ideology and partisanship might thus affect how role models influence the behaviour of other women.
There is another way in which female leaders might act as role models: The presence of women in national legislatures might not only affect behaviour, but also attitudes of women. Mansbridge (1999) argues that descriptive representation of disadvantaged groups influences
5
the way these groups view themselves. She argues that an increase in female descriptive representation will lead to more women who perceive themselves to be just as capable of leading the country as men. Beaman et al (2009) provide evidence for this claim, showing how perceptions of the effectiveness of female leaders change when people are exposed to a female leader. In later research Beaman et al (2012) show how this exposure also leads to a rise of political aspirations, in both parents and adolescents. This role model effect on attitudes is not only manifested in political attitudes but in other spheres of life as well. For example, Rosenthal et al (2013) show how women who were exposed to successful female physicians felt a higher sense of belonging in the medical field and reported that it was easier for them to identify themselves with the female physician. This shows how role models not only affect behaviour, but also attitudes.
It is also interesting to think about how a change in the composition of parliament influences the opposite group. For example, research from the US shows that the election of Obama as the first black President accelerated the trend in which white Americans’ attitudes towards black Americans became more positive (Welch & Sigelman, 2011). Thus, while Obama might have acted as a role model for black people, his election also seems to have affected white people’s attitudes. Female leadership might have a similar effect. It probably not only affects the behaviour of women, but also the behaviour of men. Gangadharan et al. (2016) show how male behaviour is significantly altered after women are elected as leaders: They are less likely to cooperate and contribute to the public good because of “social norms relating to the role of men and women as leaders” (Gangadharan, Jain, Maitra, & Vecci, 2016, p. 33) However, the researchers also show how this effect disappears after greater exposure to female leaders. This indicates that while at first the election of female leaders might create backlash under men, long-lasting exposure will eventually influence social norms. In addition, by studying the results of a policy experiment in Lesotho, Clayton (2014) shows how an increase in the number of women in the local electoral divisions not only had a symbolical effect on the local communities, it also decreased the perceived influence of traditional leaders: the traditional male elites. It thus changed people’s attitudes towards existing patriarchal structures.
Lastly, the fourth theory argues that an increase in the number of women in national legislatures will lead to a shift in policy decision-making. The main assumption behind this is that descriptive representation (increasing the actual number of women) will lead to substantive representation (the actual incorporation of women’s interests in policies) (Bauer & Britton,
6
2006). Chen (2010) provides evidence for this claim, showing that an increase in the share of female legislators increases government spending on social welfare. Wängnerud (2009) assesses the broader literature on this question of the relationship between descriptive and substantive representation. When comparing several different studies, she concludes that an increase in female representation does strengthen the position of women’s preferences when it comes to policy decisions.
There could also be some reasons why an increase in women’s representation might not automatically lead to a change in gender norms regarding political equality. For example, Broockman (2014) shows, employing a regression analysis with data on US state legislative elections, that the role model-effect does not seem to apply to the US: political participation of women did not increase when additional women were elected. This discrepancy in results, as Broockman (2014) himself argues, could be attributed to the different stages of women’s inclusion. When female leaders are first elected, they might act as role models towards other women, but when additional women are elected, the role model effect is already in effect, so there is no discernible effect anymore. Gilardi (2015) actually provides evidence for this claim by showing how the relationship between the election of a woman in a Swiss municipality and the election of a woman in a neighbouring municipality in the next election was strongest when there were no female incumbents. This could also be the case for the influence on gender norms: gender norms might be influenced just after the election of the first women to national legislatures but might not be further influenced when additional women are elected.
Furthermore, there is also the question of visibility. This thesis will assess whether an increase in women’s representation in the national legislature influences gender norms. It could be the case however, that an increase is not really visible in society. For example, in countries with a presidential system, stories about the president might dominate the news. If there is almost no mention of the presence or the work of female legislators, then it will be hard for them to be able to serve as role models or to communicate to the public the changes in policy they have brought about. Gender norms about the capabilities of female leaders might therefore not change, because people are not aware of the presence of these women.
There could be another consequence of an increase in women’s representation: It could be the case that an increase might even lead to a negative change in gender norms regarding political equality. For example, Foos & Gilardi (2016), using experimental evidence, even show how
7
exposure to female role models led to a decrease in political ambition under female students at the University of Zürich. They theorise that this could be because students were put off by the experiences of the female role models. Though this does not exactly work in the same way for gender norms; gender norms are only attitudes, while political ambition implies some sort of behaviour, it does raise questions about the possible negative consequences of an increase in women’s representation. For example, if women are only elected to increase descriptive representation, then people might feel as if they are not getting anything done, which might increase the number of people who believe that men make better political leaders than women, instead of the other way around. However, as shown above, in general, an increase in women’s representation not only increases descriptive representation but substantive representation as well (Wängnerund, 2009). The expectation in this thesis is, therefore, that there will be a positive change in gender norms regarding political equality as a result of an increase in women’s representation. Accordingly, the first hypothesis in this thesis is: H1 = When the
percentage of women in parliament increases, the number of people who believe that men are better leaders than women decreases.
As mentioned above, the visibility of the change in representation could be a key factor. After all, if the change is minimal or barely visible then chances are that the effects are minimal as well. To take this into account, this thesis also analyses whether a difference in change in representation influences gender norms. In conventional political science, the idea exists that the percentage of women in parliament needs to pass a certain threshold to have any discernible effects (Dahlerup, 2006). In recent years, research on this topic has increased and so has the discussion on this topic; some research shows that a higher number of women in parliament positively affects policies geared towards women, while other research shows that it is easier for a small group of women to form a front to advance these policies (Childs & Krook, 2006). Still, the main narrative in political science is that an increase in women’s representation will lead to more inclusion of women’s issues in policy. For example, Grey (2006) suggests that although just reaching a threshold often is not enough to incorporate female interests into policies, these thresholds can act as inhibitors. Reaching these thresholds can thus often pave the way for more inclusive policies. The chance that such a threshold is reached increases when the change in representation increases as well. For example, if representation increases with fifteen percentage points, the chance that the threshold is reached is higher than if representation increases with five percentage points. Therefore, the second hypothesis in this thesis is: H2 =
8
When the change in percentage of women in parliament increases, the number of people who believe that men are better leaders than women decreases.
Lastly, this thesis will also test whether a change in representation influences men and women differently. Evidence from some Western countries shows that “women tend to evaluate women leaders more favorably” (Banducci & Karp, 2000, p. 840). This could extend to their assessment of the capabilities of women leaders as well, leading to an increase in the number of women who think women are just as capable as men at leading the country when the number of women in parliament increases. Furthermore, as the literature on the role-model effect shows, the exposure to more women in parliament might influence women substantially more than men. Therefore, the third hypothesis in this thesis is: H3 = When the percentage of women in
parliament increases, the effect on the number of people who believe that men are better leaders than women should be smaller for men than for women.
9 3. Research Methods
To test these hypotheses, data from the Afrobarometer, the Americasbarometer, the Interparliamentary Union (IPU), the International Institute for Democracy and Electoral Assistance (IDEA), and the World Bank is used.
In this thesis, two main independent variables are used. The first independent variable is the level of representation. This variable indicates the percentage of women in parliament a year before the date of the interview, by using data from the IPU (2020). This time frame has been chosen to consider the fact that the effects of a change in representation will take some time to manifest themselves. This variable is used to assess the influence of the level of representation on gender norms regarding political equality across countries. For example, does a higher degree of representation lead to more equal gender norms?
This thesis also assesses whether different levels of change in representation influence gender norms. This is the second independent variable. This variable indicates the difference in representation in percentage points between 5 years before the date of the interview and 1 year before the date of the interview, again by using data from the IPU (2020). This time frame thus takes the election cycle into account. This variable is used to look at the relationship between representation and gender norms more closely and takes the time factor into account; If people witness a higher change in representation, does this influence their gender norms regarding political equality?
The dependent variable is gender norms regarding political equality. This is operationalised using questions asked in the Afrobarometer and the Americasbarometer. In the Afrobarometer, people were asked the following question: “Which of the following statements is closest to your view? Choose Statement 1 or Statement 2. Statement 1: Men make better political leaders than women and should be elected rather than women. Statement 2: Women should have the same chance of being elected to political office as men” (Afrobarometer, 2015). This variable is recoded into a dichotomous variable with 0 meaning people believe women should have the same chance of being elected as men and 1 meaning people believe men make better political leaders than women. People who answered they agree with neither or do not know are coded as missing. In the Americasbarometer people were asked whether they agreed that “Men are better political leaders than women.” People could answer with either “strongly agree,” “Agree,” “Disagree,” or “Strongly disagree.” This is recoded into a dichotomous variable as well, with 0 meaning people who disagree that men are better political leaders than women and
10
1 meaning people who agree that men are better leaders than women. This dependent variable might seem a little narrow, but in both barometers, this is the only variable that has been consistently asked throughout the different survey rounds and to a large part of the sampled population. Therefore, for generalisability, this is the only dependent variable included in the analysis.
The supposed relationship between women’s representation and gender norms regarding political equality can be modelled as follows:
𝑃(𝑌𝑖 = 1) = 1
1 + ℯ−(𝛽0+𝛽1𝑋1𝑖)+ 𝜀𝑖
𝑦𝑖 gender norms regarding political equality
𝛽0 intercept
𝛽1 coefficient
𝑋1𝑖 percentage of women in parliament
𝜀𝑖 error
This thesis also controls for other factors that could influence gender norms regarding political equality. Individual control variables include age, gender, education level, media use, political interest, religion, and access to basic necessities. Age is included as a control variable because different generations might hold different views on gender norms. Gender is included as a control variable because there might be a difference between men and women in the way they view the capabilities of female leaders. In addition, education level is included because there might be a difference between highly educated people and lower-educated people in the way they regard gender norms. Media use is also included, because, as mentioned before, gender norms are influenced by what we observe in the world. Therefore, people who use media might be exposed to different views and opinions than people who do not use media and thus, their views on gender norms might differ as well. This is operationalised in two ways: How often people get their news through the Internet and how often people get their news through newspapers. In this way, both new and old ways of using media are controlled for. Religion is also included because there is a large body of feminist literature that asserts that religion has perpetuated a system of patriarchy (Sharma & Young, 1999). Because of this, it is necessary to control for religion because this could influence gender norms as well. Access to basic necessities is also included as a control variable, because people who do not have access to
11
basic necessities might regard gender norms differently than people who do have access to basic necessities, for example, due to the fact that they do not think instating female leaders should be a priority when they do not have access to basic necessities. This is operationalised by using the variable that measures how often people have gone without food. In the Afrobarometer, it measures how often people have gone without food during the last twelve months; in the Americasbarometer it measures how often people have gone without meals in the last three months. Lastly, political interest is included as a control variable because people who are interested in politics might have a different opinion on the capabilities of female leaders than people who are not interested in politics. For example, people who are not interested in politics might not be interested in changing gender norms as well. The Americasbarometer measures political interest directly, but unfortunately, there is no equivalent variable in the Afrobarometer, so a closely related variable, whether people discuss politics, is used. The appendix provides an extensive overview of all variables used in the analyses and shows how they are coded.
Control variables at the country level are also included. These include the presence of gender quotas, the level of democracy, GDP, and the type of political system. The presence of gender quotas is included because there could be a huge difference between countries with and countries without gender quotas. Gender quotas are an artificial instrument to boost women’s representation, so even though they might influence women’s representation, people might feel as though the women elected through these quotas might not be as deserving as the men, which might influence their perceptions of the capabilities of female leaders. The level of democracy is included because people in authoritarian states might regard the capabilities of female leaders differently than people in democracies because they are less exposed to the democratic process. GDP is included because people in countries with a higher GDP might perceive the capabilities of female leaders differently than people in countries with a lower GDP. The reason for this is tied in with the reason why people who do not have access to basic necessities might have different gender norms than people who do: people in countries with a low GDP might be less interested in changing gender norms, simply because they have to worry about putting food on the table. Lastly, the type of political system is also included. This is important because it could be the case that the political system influences the way people perceive gender norms regarding political equality. As mentioned above, the increase in women’s representation, as well as the effects of this increase, needs to be visible to the people in order to possibly change their opinions on the capabilities of female leaders. It could be the case that an increase in women’s
12
representation in a presidential system is a lot less visible than an increase in a parliamentary system and therefore, it is important to control for this fact.
In addition, an interaction effect is also included in the analysis, because the relationship between women’s representation and gender norms regarding political equality could be different for women than for men. To test this, an interaction effect variable was created by multiplying the variable gender and the variable that measures women’s representation. The relationship between women’s representation and gender norms regarding political equality including the numerous control variables and interaction effect can be modelled in the following way.
𝑃(𝑌𝑖 = 1) = 1
1 + ℯ−(𝛽0+(𝛽1𝑋1𝑖 +⋯+ 𝛽12𝑋12𝑖) +𝛾0𝑋1𝑖+𝛾1(𝑋1𝑖𝑋2𝑖)+ 𝛾2𝑋2𝑖)+ 𝜀𝑖
𝑦𝑖 gender norms regarding political equality
𝛽0 intercept
𝛽1 … 𝛽12 coefficients
𝛾0… 𝛾2 coefficients for interaction effect
𝑋1𝑖 percentage of women in parliament
𝑋2𝑖 gender
𝑋3𝑖… 𝑋12𝑖 all other control variables
𝜀𝑖 error
Below is a visual representation of the relationship between the independent variables, dependent variables, control variables, and the interaction effect variable.
13
Because the dependent variable is dichotomous, logistic regressions have been run. However, as a robustness test, OLS regressions, as well as ordinal regressions have been run as well to see whether the results hold up independent of the type of regression analysis. To do so, the original ordinal variables have been used. The output of these regressions can be found in the Appendix.
This thesis uses data from Africa and Latin America because, in the last two decades, the percentage of women in parliament in Latin American and African countries has risen starkly, often due to the implementation of gender quotas (IPU, 2020; Worldbank, 2020). Across Latin America and Africa, there are hundreds of development programs aimed at increasing women’s representation (Krook & Norris, 2014). It is, therefore, interesting to see whether an increase in women’s representation has actually had a tangible effect that creates a sustainable foundation for advancing gender equality. In addition, data availability is also a key factor in this case selection. Data on gender norms regarding political equality is readily available from the Afrobarometer and the Americasbarometer. These barometers also allow for research that looks at effects over time, because these surveys are repeated every few years with roughly the same questions. Other data sources, like the World Values Survey, for example, do not allow for the same research, because each wave includes different countries. From the Americasbarometer, a merged dataset that already exists is used. This dataset includes 7 waves, for which data was collected in 2004, 2008, 2010, 2012, 2014, 2016/2017, and 2018/2019. Data from 2004 and 2010 was deleted because no data on gender norms regarding political equality was collected during these waves. The merged dataset for the Afrobarometer was created manually using data
14
from the fifth, sixth, and seventh wave which was collected in respectively 2011/2013, 2016, and 2019. Unfortunately, earlier waves had to be excluded because the question concerning gender norms regarding political equality was not present in these waves. In the interest of being able to compare the data over time, only data from countries that are present in all three waves is included. In terms of data collection, the Afrobarometer uses a “clustered, stratified, multi-stage, area probability sample” (Afrobarometer, 2020). Likewise, the Americasbarometer uses “stratified, multi-stage cluster sampling” (Americasbarometer, 2020). This means that there is no random sampling at the national level. Rather, the population is divided into subgroups, from which a random sample is drawn. This way, the sampling error is reduced, and the sample is a better representation of the entire population.
In both datasets, the observations are not all independent from each other because of geographical clustering as well as time dependency. To control for geographical clustering, country-fixed effects are added, and the standard errors are clustered at the country-level. To control for time dependency, survey round controls are added. Lastly, a weight is added that combines the weight which ensures every country is weighted equally and the weight which ensures that the data presents a representative sample of the population.
15 4. Results
The results from the Afrobarometer will be displayed first, before displaying the results from the Americasbarometer. In the conclusion, the results will be compared, and possible differences will be explained.
Results of Afrobarometer
First, a regression analysis was run to see whether the level of representation influences gender norms regarding political equality. The dependent variable is the dichotomous variable that indicates whether people think men are better leaders than women. The independent variable is the percentage of women in parliament. The results of this analysis can be found in Model 1 in Table 1. As can be seen in the table, the level of representation does influence gender norms regarding political equality, when controlling for the other individual and country-level factors described above. The odds that people will agree with the notion that men are better leaders than women increase with 1,018 when the percentage of women in parliament increases with 1 percentage point. Thus, if the number of women in parliament increases, the number of people who agree with the notion that men are better leaders than women increases as well. This effect is significant at the 99% confidence level, p = 0,001. This is the opposite of the expected results. Thus, H1 is not supported in Africa. However, this is a small result. An increase in the level of
representation from 20 per cent to 25 per cent increases the predicted probability from 0,28 to 0,30, so 2 percentage points. This means that if the level of representation is 20 per cent, approximately 28 people will agree with the notion that men are better leaders than women. If the level of representation is 25 per cent, then approximately 30 people will agree with this notion. All predicted probabilities can be found in the Appendix.
Then, a regression analysis was run to see whether different levels of a change in representation influence gender norms regarding political equality. The dependent variable is the dichotomous variable that measures whether people think men are better leaders than women. The independent variable is the change in percentage points of the percentage of women in parliament between five years before the interview and one year before the interview. The results of this analysis can be found in Model 2 of Table 1. As can be seen in the table, different levels of a change in representation do influence gender norms regarding political equality, when controlling for the individual and country-level factors described above. The odds that people will agree with the notion that men are better leaders than women increase with 1,013 when the change in representation increases with one percentage point. Thus, if the change in
16
representation increases, the number of people who believe men are better leaders than women increases as well. This effect is significant at the 99% confidence level, p = 0,002. This is the opposite of the expected results. Thus, H2 is not supported in Africa. Similar to the effect of the
level of representation, this effect is quite small. An increase in the change of representation from 5 per cent to 10 per cent increases the predicted probabilities from 0,29 to 0,30, so 1 percentage point. This means that if the change in representation is 5 per cent, 29 people will agree with the statement that men are better leaders than women. If the change in representation is 10 per cent, 30 people will agree with this statement. All predicted probabilities can be found in the Appendix.
In the third model, both the independent variable that indicates the level of representation as well as the independent variable that indicates the level of the change in representation over the last election cycle are included. In this model, the dependent variable is the dichotomous variable that measures whether people think men are better leaders than women. The results of this analysis can be found in Model 3 of Table 1. This model was run to assess whether different levels of the change in representation would influence gender norms controlled for the level of representation. It could be the case that a difference in the level of representation influences the effects of a change in representation because these effects might be different for a low level of representation as opposed to a high level of representation. For example, an increase from five per cent women in parliament to ten per cent women in parliament might seem more significant to people than an increase from thirty per cent to 35 per cent. In this model, both the change in representation as well as representation itself are not significant: p = 0,129 and p = 0,239, respectively. Therefore, when controlled for the level of representation, different levels of change in representation do not influence gender norms.
The fourth model includes an interaction effect to assess whether the effect of representation on gender norms differs between women and men. As can be seen in Model 4 of Table 1, an increase of 1 percentage point of representation increases the odds that a man would agree with the notion that men are better leaders than women with 1,020 and it increases the odds that a woman would agree with this notion with 1,016. However, these results are not significant, p = 0,283. This means that H3 is not supported in Africa.
The fifth model also includes an interaction effect. This effect assesses whether the effect of a difference in change of representation differs between women and men. As can be seen in
17
Model 5 of Table1, an increase of 1 percentage point in the change of representation increases the odds that a man would agree with the notion that men are better leaders than women with 1,016 and it increases the odds that a woman would agree with this notion with 1,009. However, these results are not significant, p = 0,241. This means that H3 is not supported in Africa.
In these analyses, the control variable type of political system has been omitted. Dummies for the different categories were automatically omitted from the analysis because they were too closely correlated with the other variables. However, by including the country dummies, differences in the type of political system between the countries are accounted for.
As can be seen in Table 1 below, the variables that controlled for age, gender, education, and media use are significant in every model. Furthermore, the category Muslim for the variable that controlled for religion and the category voluntary gender quotas for the variable that controlled for type of gender quotas are significant as well. As can be seen in Model 1 through 5, the odds that women would agree that men are better leaders than women are lower than the odds that men would agree. For every year that people are older, the odds that people agree that men are better leaders are lower. The odds that someone with either secondary or post-secondary education would agree that men are better leaders are lower than someone with primary education, while the odds that someone with no formal education would agree are higher than someone with primary education. The odds that people who have access to media, whether it is newspapers or the internet, would agree with this statement are lower than people who do not have access to media. Lastly, the odds that Muslims would agree with this statement are higher than the odds that Christians would agree.
18
Table 1. Afrobarometer – logistic regression analysis of the probability of a change in gender norms
Model 1 Model 2 Model 3 Model 4 Model 5
(Constant) 0,125** (0,667) 0,318 (0,657) 0,198* (0,725) 0,120** (0,677) 0,314 (0,658) Representation 1,018** (0,005) 1,009 (0,008) 1,020 (0,005) Change in representation 1,013** (0,004) 1,010 (0,006) 1,016** (0,006)
Interaction effect female and representation 0,996
(0,004) Interaction effect female and change in
representation 0,993 (0,006) Age 0,995*** (0,001) 0,995*** (0,001) 0,995*** (0,001) 0,995*** (0,001) 0,995*** (0,001) Gender (Ref. = male).
Female 0,501*** (0,048) 0,501*** (0,048) 0,501*** (0,048) 0,545*** (0,093) 0,512*** (0,048) GDP (PPP) per capita / 1000 0,996 (0,046) 0,984 (0,047) 0,991 (0,046) 0,996 (0,046) 0,984 (0,046) Education (Ref. = primary education)
No formal education Secondary education Post-secondary education 1,196*** (0,034) 0,794*** (0,029) 0,653*** (0,033) 1,198*** (0,034) 0,793** (0,030) 0,653*** (0,033) 1,197*** (0,034) 0,794*** (0,030) 0,653*** (0,033) 1,194*** (0,046) 0,795*** (0,034) 0,653*** (0,033) 1,198*** (0,033) 0,793*** (0,030) 0,653*** (0,033)
19
Religion (Ref. = Christian) Muslim Other religion 1,322*** (0,056) 1,082 (0,043) 1,321*** (0,056) 1,085 (0,043) 1,321*** (0,056) 1,084 (0,043) 1,323*** (0,056) 1,082 (0,043) 1,321*** (0,056) 1,085 (0,043)
Gone without food 1,008
(0,026) 1,008 (0,026) 1,008 (0,026) 1,009 (0,026) 1,008 (0,026) Used newspaper 0,951* (0,023) 0,949* (0,023) 0,950* (0,032) 0,950* (0,023) 0,949* (0,023) Used Internet 0,857*** (0,031) 0,860*** (0,032) 0,858*** (0,032) 0,857*** (0,031) 0,860*** (0,032) Discussed politics 0,954 (0,024) 0,952* (0,024) 0,953 (0,024) 0,955 (0,023) 0,953* (0,024) Presence of gender quotas (Ref. = no presence
of quotas)
Presence of voluntary gender quotas Presence of mandatory gender quotas 2,993*** (0,061) 1,027 (0,067) 2,834*** (0,057) 1,054 (0,138) 2,881*** (0,061) 0,994 (0,063) 2,987 (0,061) 1,027 (0,066) 2,830*** (0,057) 1,041 (0,068) Level of democracy (Ref. = not free)
Partly free Free 1,098 (0,117) 0,954 (0,136) 1,054 (0,138) 0,937 (0,162) 1,073 (0,130) 0,937 (0,154) 1,098 (0,117) 0,954 (0,136) 1,054 (0,138) 0,937 (0,163) Cox and Snell’s R2
0,075 0,075 0,076 0,075 0,076
Nagelkerke’s R2
0,107 0,107 0,108 0,107 0,108
N 127683 127683 127683 127683 127683
Note: odds ratios with standard errors (clustered at the country level) between brackets.
20
Results of Americasbarometer
First, a regression analysis was run to see whether the level of representation influences gender norms regarding political equality. The dependent variable is the dichotomous variable that indicates whether people think men are better leaders than women. The independent variable is the percentage of women in parliament. The results of this analysis can be found in Model 1 of Table 2. As can be seen in the table, an increase in the level of representation does seem to increase the number of people who agree with the statement that men are better leaders than women. However, this effect is not statistically significant, p = 0,148. This means that H1 is not
supported in Latin-America.
Then, a regression analysis was run to see whether different levels of a change in representation influence gender norms regarding political equality. The dependent variable is the dichotomous variable that measures whether people think men are better leaders than women. The independent variable is the change in percentage points of the percentage of women in parliament between five years before the interview and one year before the interview. The results of this analysis can be found in Model 2 of Table 2. As can be seen in the table, an increase in the change in percentage does seem to decrease the number of people who agree with the statement that men are better leaders than women. However, this effect is not statistically significant, p = 0,170. This means that H2 is not supported in Latin-America.
In the third model, both the independent variable that indicates the level of representation as well as the independent variable that indicates the level of the change in representation over the last election cycle are included. In this model, the dependent variable is the dichotomous variable that measures whether people think men are better leaders than women. The results of this analysis can be found in Model 3 of Table 2. As in the Afrobarometer, this model was run to assess whether different levels of the change in representation would influence gender norms controlled for the level of representation. Interestingly, in contrast to the Afrobarometer, the variable representation is significant when including both variables at the same time. As can be seen in the model, this means that, when controlling for change in representation, an increase of 1 percentage point of the number of women in parliament increases the odds that people agree with the notion that men are better leaders than women with 1,008. This is significant at the 95% confidence level, p = 0,033.
21
The fourth model includes an interaction effect to assess whether the effect of representation on gender norms differs between women and men. As can be seen in Model 4 of Table 2, an increase of 1 percentage point of representation increases the odds that a man would agree with the notion that men are better leaders than women with 1,006 and it increases the odds that a woman would agree with this notion with 1,004. However, these results are not significant, p = 0,593. This means that H3 is not supported in Latin-America.
The fifth model also includes an interaction effect. This effect assesses whether the effect of a difference in change of representation differs between women and men. As can be seen in Model 5 of Table 2, an increase of 1 percentage point in the change of representation decreases the odds that a man would agree with the notion that men are better leaders than women with 0.998 and it decreases the odds that a woman would agree with this notion with 0,996. However, these results are not significant, p = 0,731. This means that H3 is not supported in
Latin-America.
In these analyses, the control variables type of political system and level of democracy have been omitted. Dummies for the different categories were automatically omitted from the analysis because they were too closely correlated with the other variables. However, by including the country dummies, differences in the type of political system and level of democracy between the countries are accounted for. Unfortunately, the individual control variables for media use, access to basic necessities, and religion had to be left out as well because data for these variables was only available for less than 5 per cent of all cases in the sample. Adding these variables would, therefore, result in unrepresentative results.
As can be seen in Table 2 below, the variables that control for gender, education, and political interest are significant. The odds that a woman would agree with the notion that men are better leaders than men are lower than the odds that a man would agree. With every additional year of education, the odds that someone would agree with the statement decrease. Lastly, the odds that someone interested in politics would agree with the statement are higher than the odds that someone who is not interested in politics would agree.
22
Table 2. Americasbarometer - logistic regression analysis of the probability of a change in gender norms
Model 1 Model 2 Model 3 Model 4 Model 5
(Constant) 1,174 (0,514) 1,434 (0,490) 1,057 (0,525) 1,152 (0,505) 1,431 (0,487) Representation 1,005 (0,003) 1,008* (0,004) 1,006 (0,004) Change in representation 0,997 (0,003) 0,994 (0,004) 0,998 (0,003)
Interaction effect female and representation 0,998
(0,004) Interaction effect female and change in
representation 0,998 (0,006) Age 1,002 (0,001) 1,002 (0,001) 1,002 (0,001) 1,002 (0,001) 1,002 (0,001) Gender (Ref. = male).
Female 0,482*** (0,046) 0,482*** (0,046) 0,482*** (0,046) 0,504*** (0,103) 0,484*** (0,052) GDP (PPP) per capita / 1000 0,973 (0,030) 0,970 (0,030) 0,976 (0,031) 0,973 (0,030) 0,970 (0,030) Education 0,938*** (0,007) 0,938*** (0,007) 0,938*** (0,007) 0,938*** (0,007) 0,938*** (0,007) Political interest 1,172** (0,040) 1,172** (0,039) 1,172** (0,039) 1,172** (0,040) 1,172** (0,039) Presence of gender quotas (Ref. = no presence
of quotas)
Presence of voluntary gender quotas Presence of mandatory gender quotas
0,827 (0,174) 0,768 (0,147) 0,839 (0,169) 0,797 (0,125) 0,830 (0,162) 0,777 (0,131) 0,827 (0,173) 0,768 (0,147) 0,839 (0,169) 0,797 (0,125)
23
Cox and Snell’s R2 0,080 0,080 0,080 0,080 0,080
Nagelkerke’s R2
0,115 0,115 0,115 0,115 0,115
N 97595 97595 97597 97595 97595
Note: odds ratios with standard errors (clustered at the country level) between brackets.
24 5. Discussion and conclusion
This thesis analysed the relationship between women’s representation and gender norms regarding political equality. This was done by running five logistic regression analyses for the Afrobarometer and the Americasbarometer each. The results show that in Africa, an increase in both level of representation as well as the change in representation will lead to an increase in the number of people who agree that men are better leaders than women. This effect is small, yet significant and directly contrasts the hypotheses. The results from Latin-America also show a small positive relationship between women’s representation and whether people believe men are better leaders than women. However, these results are not statistically significant.
There could be several reasons for the difference between Africa and Latin-America. First, the Afrobarometer and the Americasbarometer are cross-regional comparative surveys that are held independently from each other. Because of this, the list of questions differs a lot. The dependent variable in this research, gender norms regarding political equality, was measured by looking at the number of people who agree that men are better leaders than women. In the Americasbarometer, people could agree or disagree with this statement. However, in the Afrobarometer, people could agree with this statement or they could agree with the statement that women should have the same chance of being elected as men. Although these questions are largely alike, the subtle difference could have caused the difference between the barometers. A second reason could be the differences between the two regions. For example, most of the sampled countries in Latin-America have presidential systems, while most of the sampled countries in Africa have presidential-parliamentary systems. As mentioned before, members of parliament in presidential systems could be less visible than those in other systems. Another consequence of this difference in political systems is that in a presidential system, it could be the case that people do not regard members of parliament as leaders. Perhaps they only see their president as the leader. If this is the case, then their opinion on the leadership capabilities of women does not change when more women enter parliament, because they do not regard these women as leaders.
There could also be several reasons for the observed positive effect between women’s representation and whether people think men are better leaders than women. One possible explanation for the positive relationship is in line with Grey’s (2006) theory that just reaching a certain threshold is not enough. Women might be elected, but if they do not feel supported, they might not be able to act in the interest of women. As a result, they might not be visible
25
enough. People who elected them might feel as though their representatives did not accomplish any of their promised goals, which in turn might disappoint them. In some cases, people might be so deeply disappointed that it changes their opinion on women’s capabilities.
Secondly, unfortunately, data on gender norms from the Afrobarometer and the Americasbarometer was only available from 2011 onwards and 2008 onwards respectively. The percentage of women in parliament had already risen starkly in the decade before, so it could be the case that in the early stages, an increase in women’s representation does decrease the number of people who believe men are better leaders than women, while it achieves the opposite results further down the line.
Thirdly, the positive relationship could also be the result of backlash. In this case, backlash is the attempt to reverse women’s gains. Sanbonmatsu (2008) describes how an increase in the number of women in parliament might lead to an increase in perceived threat by the people who benefit from the status quo. She explains how an increase in women’s representation could pose a threat to people’s power or status, but also their personal identity. This backlash could thus result in a higher number of people who think that men are better leaders than women. Additionally, even though his research is focused on LGBT-legislators and not specifically female legislators, Haider-Markel (2010) provides statistical evidence that an increase in LGBT-legislators in the US led to an increase in anti-LGBT legislation. This type of backlash could apply to female legislators too. More female members of parliament could have led to more anti-gender equality legislation, which in turn could have influenced gender norms.
These findings could have important policy recommendations. They show that just increasing the number of women in parliament is not enough to substantially influence gender norms regarding political equality positively. Rather, solely increasing women’s representation could even have negative effects. It is therefore important that policies aimed at increasing women’s representation, such as gender quotas, are not solely implemented to fulfil requirements, please international donors, or just as a symbolic measure. Rather, they should be accompanied by other policies aimed at educating both the female candidates as well as the population.
Regrettably, the results from the Americasbarometer might be less robust than the results from the Afrobarometer because it was not possible to control for individual factors as thoroughly as in the Afrobarometer due to the limited data availability. Unfortunately, information on certain control variables like media use and access to basic necessities was only available for a very
26
limited number of cases in the sample and could therefore not be taken into account in the analysis. This might have made the results from the Americasbarometer less reliable.
Furthermore, this thesis assumes that the change in gender norms is a consequence of an increase in women’s representation. However, this could also work the other way around: People’s gender norms regarding political equality could shift, which might cause them to vote differently. To some extent this is a chicken and egg situation; it is impossible to say for sure which factor caused the other. However, at first glance, there do not seem to be many strong theoretical arguments that support the theory that a negative change in gender norms regarding political equality would increase women’s representation. If the number of people who agree that men are better leaders than women increases, why would the number of people who vote for women increase as well? Yet some arguments warrant taking a closer look at this theory. For instance, more regressive gender norms could lead to a higher level of political mobilisation. A recent example of this were the US midterm elections in 2018. In response to Trump’s election as President in 2016, a record number of, mostly Democratic, female representatives were chosen in 2018 (Pew Research Center, 2018). For many, Trump’s election signified a step back in the fight for gender equality. Yet his election spurred on a lot of political mobilisation, which in 2018 resulted in the highest number of women in Congress ever (Chira, 2020). A similar mechanism could explain the results in this thesis; In Africa, an increase in the number of people who think men are better leaders than women could have led to a higher number of women in parliament, because an increase in regressive gender norms triggered political mobilisation. Some evidence for this claim might be the fact that in the last decade, Africa experienced “an overall deterioration in the quality of political and economic transformation and governance” (Cheeseman, 2018). This decline could have led to more repressive gender norms since research shows how a lower quality of democracy is associated with fewer advances in women’s rights (Walsh, 2012). However, this theory needs further research.
This thesis assumes that a change in women’s representation influences gender norms. Further research could investigate the reverse relationship already briefly outlined above. For example, research could look into whether this is a partisan effect or whether it manifests itself across the population equally. Additionally, this research has mostly been descriptive. Further research could focus on explaining why the relationship between women’s representation and gender norms regarding political equality seems to be negative. In this conclusion, some possible
27
explanations were outlined, but further research should examine these explanations more carefully. Furthermore, this is a large cross-regional research. It could be helpful to study one or a few particular countries, to be able to look at the negative relationship more closely.
28
Bibliography
Afrobarometer. (2015). Data Codebook for a Round 5 Afrobarometer survey in 34 African
countries.
Afrobarometer. (2020). Sampling principles and weighting. Retrieved March 4, 2020, from Afrobarometer: https://www.afrobarometer.org/surveys-and-methods/sampling-principles
Americasbarometer. (2020). Design Questionnaires and Sample Designs. Retrieved March 4, 2020, from Americasbarometer: https://www.vanderbilt.edu/lapop/core-surveys.php Banducci, S. A., & Karp, J. A. (2000). Gender, Leadership and Choice in Multiparty Systems.
Political Research Quarterly, 53(4), 815-848.
Bauer, G., & Britton, H. E. (2006). Women in African Parliaments: A Continental Shift? In G. Bauer, & H. E. Britton, Women in African Parliaments (pp. 1-30). London: Lynne Rienner Publisher.
Beaman, L., Chattopadhay, R., Duflo, E., Pande, R., & Topalova, P. (2009). Powerful Women: Does Exposure Reduce Bias. The Quarterly Journal of Economics, 124(4), 1497-1540.
Beaman, L., Duflo, E., Pande, R., & Topalova, P. (2012). Female Leadership Raises Aspirations and Educational Attainment for Girls: A Policy Experiment in India.
Science, 335(6068), 582-586.
Broockman, D. E. (2014). Do female politicians empower women to vote or run for office? A regression discontinuity approach. Electoral Studies, 34, 190-204.
Burns, N., Gallagher, & Katherine. (2010). Public Opinion on Gender Issues: The Politics of Equity and Roles. Annual Review of Political Science, 425-443.
Campbell, D. E., & Wolbrecht, C. (2006). See Jane Run: Women Politicians as Role Models for Adolescents. The Journal of Politics, 68(2), 233-247.
Carli, L. L. (2015). Social Influence and Gender. In S. G. Harkins, Kipling, Williams, & J. Burger, The Oxford Handbook of Social Influence. Oxford: Oxford University Press. Chandra-Mouli, V., Plesons, M., & Amin, A. (2018). Addressing harmful and unequal gender
29
Cheeseman, N. (2018). A Divided Continent: BTI 2018 Regional Report Africa. Gütersloh: Bertelsmann Stiftung 2019.
Chen, L.-J. (2010). Do Gender Quotas Influence Women's Representation and Policies? The
European Journal of Comparative Economics, 7(1), 13-60.
Childs, S., & Krook, M. L. (2006). Should Feminists Give Up on Critical Mass? A Contingent Yes. Politics & Gender, 522-530.
Chira, S. (2020). Donald Trump's Gift to Feminism: The Resistance. Daedalus, 149(1), 72-85. Clayton, A. (2014). Electoral gender quotas and attitudes toward traditional leaders: A policy
experiment in Lesotho. Journal of Policy Analysis and Management, 33(4), 1007-1026.
Conover, P. J. (1988). The Role of Social Groups in Political Thinking. British Journal of
Political Science, 18(1), 51-76.
Dahl, R. A. (2020, February 19). Democracy. Retrieved February 26, 2020, from Britannica: https://www.britannica.com/topic/democracy
Dahlerup, D. (2006). The Story of the Theory of Critical Mass. Politics & Gender, 2(4), 511-522.
Foos, F., & Gilardi, F. (2016). Bridging the Gender Gap in Political Ambition: Experimental Evidence. Working Paper, 1-16.
Gangadharan, L., Jain, T., Maitra, P., & Vecci, J. (2016). Social identity and governance: The behavioral response to female leaders. European Economic Review, 302-325.
Gilardi, F. (2015). The Temporary Importance of Role Models for Women's Political Representation. American Journal of Politics, 59(4), 957-970.
Grassi, D. (2011). Democratic consolidation. In B. Badie, D. Berg-Schlosser, & L. Morlino,
International Encyclopedia of Political Science, Vol. 1 (pp. 614-620). Thousand Oaks:
SAGE Publications.
Grey, S. (2006). Numbers and Beyond: The Relevance of Critical Mass in Gender Research.
Politics & Gender, 492-502.
Haider-Markel, D. P. (2010). Out and running: Gay and Lesbian candidates, elections, and
30
Hajnal, Z. L. (2010). America's Uneven Democracy: Race, Turnout and Representation in
City Politics. Cambridge: Cambridge University Press.
IPU. (2020). Electoral quota for women. Retrieved February 16, 2020, from Inter-Parliamentary Union:
https://data.ipu.org/compare?field=chamber%3A%3Afield_is_electoral_quota_women &structure=any__lower_chamber#map
IPU. (2020, April). Percentage of women. Retrieved April 10, 2020, from IPU:
https://data.ipu.org/compare?field=chamber%3A%3Acurrent_women_percent&struct ure=any__lower_chamber#map
Krook, M. L., & Norris, P. (2014). Beyond Quotas: Strategies to Promote Gender Equality in Elected Offices. Political Studies, Vol. 62, 2-20.
Mansbridge, J. (1999). Should Blacks Represent Blacks and Women Represent Women? A Contingent "Yes". The Journal of Politics, 61(3), 628-657.
Mariana, M., Marshall, B. W., & Mathews-Schultz, A. L. (2015). See Hillary Clinton, Nancy Pelosi, and Sarah Palin Run? Party, Ideology, and the Influence of Female Role Models on Young Women. Political Research Quarterly, 68(4), 716-731. Paola, M. d., Scoppa, V., & Benedetto, M. A. (2014). The impact of gender quotas on
electoral participation: Evidence from Italian municipalities. European Journal of
Political Science, 141-157.
Pew Research Center. (2018, December 18). A record number of women will be serving in the
new Congress. Retrieved May 16, 2020, from Pew Research Center:
https://www.pewresearch.org/fact-tank/2018/12/18/record-number-women-in-congress/
Phillips, A. (1998). Democracy and Representation: Or, Why Should It Matter Who Our Representatives Are? In A. Phillips, Feminism and Politics. Oxford: Oxford University Press.
Rosenthal, L., Levy, S. R., London, B., Lobel, M., & Bazile, C. (2013). In Pursuit of the MD: The Impact of Role Models, Identity Compatibility, and Belonging Among
Undergraduate Women. Sex Roles, Vol. 68, 464-473.
Sanbonmatsu, K. (2008). Gender Backlash in American Politics. Politics & Gender, 4(4), 634-642.
31
Scharrer, E., & Blackburn, G. (2018). Cultivating Conceptions of Masculinity: Television and Perceptions of Masculine Gender Role Norms. Mass Communication and Society, 149-177.
Shamaileh, A. (2016). Am I equal? Internet access and perceptions of female political leadership ability in the Arab world. Journal of Information Technology & Politics, 257-271.
Sharma, A., & Young, K. K. (1999). Feminism and world religions. Albany, NY: State University of New York Press.
United Nations. (2015, May 21). Gender Statistics Manual. Retrieved February 20, 2020, from UNStats: https://unstats.un.org/unsd/genderstatmanual/Glossary.ashx
Walsh, D. M. (2012). Does the Quality of Democracy Matter for Women's Rights? Just Debate and Democratic Transition in Chile and South Africa. Comparative Political
Studies, 1323-1350.
Wängnerund, L. (2009). Women in Parliaments: Descriptive and Substantive Representation.
Annual Review of Political Science, 51-69.
Welch, S., & Sigelman, L. (2011). The "Obama Effect" and White Racial Attitudes. The
Annals of the American Academy of Political and Social Science, 207-220.
Wolbrecht, C., & Campbell, D. E. (2007). Leading by Example: Female Members of Parliament as Political Role Models. American Journal of Political Science, 51(4), 921-939.
Worldbank. (2020). Proportion of seats held by women in national parliaments (%) - Latin
America & Caribbean, Sub-Saharan Africa. Retrieved March 17, 2020, from World
32
APPENDIX
Table 1. List of countries in datasets
Afrobarometer Americasbarometer
Benin Mexico
Botswana Guatemala
Burkina Faso El Salvador
Cameroon Honduras
Cape Verde Nicaragua
Côte d’Ivoire Costa Rica
Ghana Panama Guinea Colombia Kenya Ecuador Lesotho Bolivia Liberia Peru Madagascar Paraguay Malawi Chile Mali Uruguay Mauritius Brazil Morocco Venezuela Mozambique Argentina
Namibia Dominican Republic
Niger Haiti Nigeria Jamaica Senegal Guyana Sierra Leone South Africa Sudan Swaziland (eSwatini) Tanzania Togo Tunisia Uganda Zambia Zimbabwe
Table 2. List of survey rounds in datasets
Afrobarometer Americasbarometer Round 5 (2011-2013) Wave 2008 (2008-2009) Round 6 (2014-2015) Wave 2012 (2012) Round 7 (2016-2018) Wave 2014 (2014) Wave 2016/2017 (2016-2017) Wave 2018/2019 (2018-2019)
33
Table 3. List of variables in the dataset with data from Afrobarometer
Name Explanation Values
AGE Indicates age Interval-ratio variable
ACCESS_FOOD Indicates how often people
have gone without food in the last 12 months
0 = never
1 = just once or twice 2 = several times 3 = many times 4 = always
NEWSPAPER_USE Indicates how often people
get their news from newspapers
0 = never
1 = less than once a month 2 = a few times a month 3 = a few times a week 4 = every day
INTERNET_USE Indicates how often people
get their news from the Internet
0 = never
1 = less than once a month 2 = a few times a month 3 = a few times a week 4 = every day
DISC_POLITICS Indicates how often people
discuss politics
0 = never 1 = occasionally 2 = frequently
NORMS_LEADER Indicates whether people
agree with statement 1: Men make better political leaders than women and should be elected rather than women or statement 2: Women should have the same chance of being elected to political office as men.
1 = agree very strongly with 1 2 = agree with 1
3 = agree with 2
4 = agree very strongly with 2 5 = agree with neither
GENDER Indicates gender 1 = male
2 = female
EDUCATION Indicates level of education 0 = no formal education
1 = primary education 2 = secondary education 3 = post-secondary education
COUNTRIES Indicates the country of the
respondent
See table for countries
COUNTRIES_TIME Indicates the country of the
respondent and the survey round of respondent
See table for survey rounds and countries
RELIGION Indicates religion 1 = Christian
2 = Muslim 3 = Other
FEMALE Is a dummy recoded from
gender
0 = male 1 = female
SURVEY_ROUND Indicates survey round 5 = Round 5
6 = Round 6 7 = Round 7
