RADBOUD UNIVERSITY
Nijmegen School of Management Master’s Thesis Research ProposalThe role of ethnicity in women’s labor force
participation
Name: Jikke Sieperda Student ID: S4575377
Supervisor: prof. dr. J.P.J.M. Smits Master: Economics
Specialization: International Economics & Development Faculty: Nijmegen School of Management
3
ABSTRACT
Female labor force participation is of great importance for the empowerment of women and the development of countries. However, in many sub-Saharan African countries, the FLFP is very low. This thesis contributes to the existing literature on the determinants of FLFP in sub-Saharan Africa by exploring the relationship between ethnicity and FLFP. Using data from the Global Data Lab, a multilevel logistic regression model is performed on 321115 women, living in 289 ethnic groups across 28 countries. The indicator of FLFP is whether women are working are not, classifying the farm work as not employed. The results confirm findings of previous studies that found that personal and household factors, such as the number of children, educational attainment, and marital status, play an essential role in determining the FLFP. Controlling for these individual and household factors, the analysis showed that ethnicity has a role in the FLFP. The more traditional the ethnic groups, the less likely it is that the women are employed.
5
Table of Contents
1. Introduction ... 6
2. Theoretical model ... 9
2.1 Labor force participation ... 9
2.2 Determinants ... 11
2.2.1 Individual characteristics ... 11
2.2.2 Household structure ... 12
2.2.3 Economic development ... 14
2.3 Ethnic inequality ... 15
3. Data and Method ... 18
3.1 Data ... 18 3.2 Method ... 18 3.3 Variables ... 19 3.3.1 Dependent Variable ... 19 3.3.2 Independent Variables ... 20 3.3.3 Missing Data ... 24 4. Results ... 25 4.1 Descriptive statistics ... 25 4.2 Analysis ... 27 4.2.1 Correlation Matrix ... 27
4.2.2 Multilevel Logistic Regression ... 29
5. Conclusion and discussion ... 37
6. References ... 40
7. Appendix ... 44
Appendix A. List of Countries and years of study ... 44
Appendix B. List of Countries and Ethnic groups ... 45
Appendix C. Correlation Matrix ... 49
Appendix D. Multilevel Analysis ... 51
6
1.
Introduction
Gender equality has been found essential for economic growth and sustainable development (ILOSTAT, n.d.; Klasen, 1999). The importance of gender equality is also recognized by the UNGA (2015) since one of the key goals of the 2030 Agenda for Sustainable Development is achieving gender equality and the empowerment of women. One of the goals (goal 5) is entirely devoted to this, but the topic is also of importance throughout the other goals (UNGA, 2015). The OECD (2011) stressed that it is crucial for women to participate in the economic process and to have access to economic resources to be able to gain more economic empowerment. By entering the labor force this can be achieved, but numbers show that women do not engage in the labor force as much as men do. The ratio of females that participate in the labor force, compared to males, is 67%, according to the International Labour Organization (ILO) (The World Bank, 2020b). In low and middle-income countries, the number of women that participate in the labor force is even lower. The statistics of women aged 15 and older show that in low and middle-income countries, only 46.5% of the women are actively involved in the labor force (The World Bank, 2020a). It is important to understand why more than 50% of the females in these countries are not active in the labor market because they can contribute to the positive economic growth of their economy.
In developing countries, women often work on farms, in family businesses, or have a job near their home. Women do not only work, but they also have family activities that they have to do during the day (The World Bank, 2020a). This mix of activities is not beneficial for their empowerment. The 'Modern Economy' that is illustrated by Boserup (1970) or off-farm employment, as defined in Van den Broek and Kilic (2019), could help to improve their empowerment, as well as economic growth and development of countries. This casual employment is of importance for the household and found to have a substantial share in the livelihood portfolios of households (Davis, Di Giuseppe, & Zezza, 2017). This relevance is reflected by a sharp decline of the labor force in Sub-Saharan Africa that is engaged in farming on the one hand, and an increase in off-farm employment on the other hand (Yeboah & Jayne, 2018).
The growing increase of the labor force that is engaged in off-farm employment does, however, not say anything about the female labor market participation in this region. Although the female labor market participation (FLMP) in the Western world has increased over the years, the labor market participation of women in developing countries is still highly varied. This variety can be explained by economic and social factors such as economic growth and
7 social norms (Verick, 2014). Improving the FLMP in developing countries is crucial since it is found to be critical for gender equality and women's empowerment (Gündüz-Hoşgör & Smits, 2008; de Jong, Smits, & Longwe, 2017; OECD, 2011).
Off-farm employment, however, is often harder to reach for women since they face more barriers for entry than men. One of the issues, for example, is the social norms and traditional responsibilities that women have in these areas, as well as the lower educational attainment of women (Slavcheska, Kaaria, & Taivalmaa, 2016). Not only personal characteristics are factors that influence the FLMP, but the context of their households are also of importance. Living in a less developed area and a rural village will make it less likely to be involved in the labor market (de Jong, Smits, & Longwe, 2017). Especially in Sub- Saharan African women face more barriers than men due to economic and social factors (Sher, 2014; Van den Broeck & Kilic, 2019).
In many low and middle-income countries (LMICs), ethnic groups are also of importance. Within countries, there is a differentiation of groups of women by life stage, class, but also ethnicity. There is no fixed division of labor; the division of labor by sex differs cross-culturally. Ethnicity is not, just like sex, biological determined, it is something that is constructed by society, and differs between societies (Momsen, 1991, p. 4). Ethnicity is a cultural identity that makes people feel like they are of being part of a group. People can identify themselves with a particular ethnicity because they share the same characteristics, it happens naturally and unconsciously (Ahmed, 2007). Ethnicity is found to be of greater importance than individual characteristics for the shaping of authority for women and the educational attainment and paid labor outside the household, according to Kritz and Makinwa-Adebusoye (1999). The difference in FLMP across different ethnicities in Africa is, however, have not been examined widely. Therefore this thesis aims to identify differences in female labor force participation of women between ethnic groups in Africa and try to explain why in some ethnic groups women work, while in others they do not.
Despite the amount of research that is done about the determinants of FLFP, there is still a lot unknown when it comes to the role of ethnicity. This study takes a new perspective in the search for determinants for the labor participation of women in Africa and included the ethnic identity of the women. Increasing the FLFP can help in the development of countries (Klasen, 1999), as well as increasing female empowerment (Gündüz-Hoşgör & Smits, 2008; de Jong, Smits, & Longwe, 2017), but differences within the countries have to be taken into account. This study will look at multiple African countries, taking into account personal, household, and
8 ethnic characteristics of women. The possible relationship between ethnicity and FMLP will be studied by answering the following question:
Research question: To what extend do ethnic factors influence the labor force participation of women in Sub-Saharan Africa?
This thesis continues as follows: Hereafter, in chapter 2, a theoretical framework will be presented, and hypotheses will be formed. Chapter 3 will explain the data that will be used in this study, and the methodology and chapter 4 will discuss the results, and explanations of the results will be given. In addition, the conclusion will outline the most important findings, the limitations, and recommendations for further research.
9
2.
Theoretical model
In this section, the literature on female labor force participation will be examined. First, the concept of labor force participation and the situation in Sub-Saharan Africa will be discussed. After that, the determinants that have been researched in previous literature will be elaborated on. Lastly, the possible role of ethnicity for the female labor force participation will be examined.
2.1 Labor force participation
The labor force participation of women is essential for economic growth and development and therefore studied extensively. Gender inequality is important to keep in mind when studying the FLFP since women do not get the same opportunities in the labor market, but also other human capital variables such as education (Jamali, 2009).
In several countries, especially western and Asian countries, the inequality between males and females is found to be a push factor women to become an entrepreneur according to Baughn, Chua, & Neupert (2006). Because of the limited opportunities in the labor market and the discrimination in labor markets, women get self-employed. Self-employment is not only a survival strategy for women, but it gives them also the flexibility that they need to fulfill the different roles that women have in their households.
In Sub-Saharan Africa, however, women have a unique role in the economy. In many SSA countries, women play a significant role in the agricultural output, and because in most regions, agriculture is still of great importance for the economy, this suggests that women have an essential role in the SSA economy (Benefo & Pillai, 2003; Boserup, 1970). The role of women in the African labor force and the history of it has been discussed since the 1970s, starting with the pioneering work of Boserup (1970). In her book, Woman’s Role in Economic Development, Boserup gives an overview of the role that women have in society and what they mean for the development process of countries. In SSA, women are mostly responsible for the tasks that are needed for the production of food. The only task that is done by men is clearing the land; all the other tasks are for the women.
Women in developing countries are involved in agricultural activities, family businesses, and other informal market sectors, which all can be classified as non-market activities. In recent years, however, it is also noticed that women also have an increased share in the modern sectors of employment. For example, Gundüz-Hosgör and Smits (2008) found that in Turkey, women were more involved in the modern sector employment. Evidence for
10 Sub-Saharan Africa, however, suggests that agriculture is still the primary driver behind the FLFP (Besamusca, Tijdens, Keune, & Steinmetz, 2015). Improvement in these countries thus still has to be made to increase female empowerment.
Women, however, have other duties as well. Besides the non-market or market activities, women also have a lot to do in the household. If you want to understand why some women do have a job outside the home, it is necessary to understand the dynamic between homework and paid labor. Women that for instance, feel that homework is a duty that they have to fulfill can have a view of more traditional family roles and are influenced by their husbands in the decision whether they should participate in the labor market (Spierings, Smits, & Verloo, 2010). In SSA, the role of women is ‘reflected in the social and cultural web of customs and legal rights that determine gender relations.’ (Svedberg, 1990, p.482). This could also mean that when there is not enough agricultural output for a family, it can decide to add more production capacity, which is female labor. It is prevalent in these areas that women marry at a very young age, and the households are polygamous (Boserup, 1970). The odds that women who are engaged in a polygamous household have a formal or paid job are very low, as shown by Spierings, Smits, and Verloo (2010). Not only traditionalism or polygamous societies can hinder labor force participation, but also other factors are found to be of influence, such as education of the women and her husband, or the development of countries. These factors will be discussed in the next section.
Albeit there is an essential role for women in SSA in agriculture, beliefs about the role of women in society remain. The idea of gender is a social construct, and the way people deal with the concept differs from one society to another, but there is one universal character, namely female subordination. Although there might be some different experiences and expressions of this idea, women are seen as subordinated. In low-income countries, this means that women are expected to have high fertility rates, but are also highly active in the agricultural sector. Subsistence farming is of great importance for women in Africa, and they need children to contribute to their duties. The level of development also influences fertility rates; the more developed, the more likely women are to get educated and have more opportunities for work outside their homes. In Africa, however, it is the men that are migrating from rural to urban areas, leaving women behind with agricultural duties (Momsen, 1991). To understand what the influence of ethnicity is on the division of labor, it is therefore essential to evaluate the role of women within households. To evaluate this role, personal characteristics of the women are taken into account, as well as characteristics of the household she lives in. These factors are found to be important determinants for FLFP, as the next section will show.
11
2.2 Determinants
Aforementioned, the labor force participation of women is of great importance for the economic growth and development of a country, as for the empowerment of women. Therefore the barriers and determinants of women for entering the labor market have been studied widely. Work is seen as an essential factor for the empowerment of women, albeit it is important to understand why some women work and others do not (Gündüz-Hoşgör & Smits, 2008). In this section, these existing explanations for female labor market participation are discussed, and the possible role of ethnicity as a determinant will be elaborated.
2.2.1 Individual characteristics
The age of the women is one of the individual characteristics that determine the female labor force participation of women. The relationship between age and labor force participation of women can be expected to be non-linear. Young women who are single and do not have children are more likely to be in the labor force. At the point they get married and get children, they get other duties and are less likely to participate, but on a later stage, when the children are older, they can return into the labor market (Momsen, 1991). The life-cycle women find themselves is thus of importance. Following this reasoning, marriage can also be of influence. The decision to engage in the labor market is highly dependent on the marital status of the women. According to Aguiar & Hurst (2007), married women have more responsibilities and thus less time to participate in the labor market. The higher the gains from work, the lower the probability that women will marry because the gains from marriage are lower compared to the gains from work (van der Klaauw, 1996). The general conception is that married women have the lowest activity rates compared to non-married women (Youssef, 1972).
Besides marriage, it is also found that fertility decisions are highly interrelated with the decisions to engage in the labor force. Findings show that within a group with married women, those with children work less than those who do not have children (Youssef, 1972). A theory proposed by Benefo and Pillai (2003) is that fertility can explain the FLFP. Young children are a barrier for women to engage in non-family work, and therefore the participation of women in the labor market is often explained by the presence of children (Benefo & Pillai, 2003; Besamuca et al., 2015; de Jong et al., 2017). In most cultures, women are considered the prime suppliers of household care needs, which increase with the presence of children. Women become more occupied with more household duties and, consequently, less likely to enter the labor market (Collever & Langlois, 1962; Semyonov, 1980; Youssef, 1972). Although the
12 evidence for developing countries is less persistent, De Jong et al. (2017) provide strong evidence that there is indeed a negative relationship between the number of children and FLFP. Using the presence of twins as an instrumental variable, they show that having young children reduces the odds of engaging in the labor market. Using twins as instrumental variables as well, Angrist and Evans (1998) show that children result in the reduction of the female labor supply. Especially women who have experienced low levels of education and do not have husbands with high incomes are confronted with the consequences of childbearing. Instead of working, women have to devote time to childcare, and the more children you have, the more time is needed (Angrist & Evans, 1998).
Another critical factor in explaining the female labor force participation is education. Benefo and Pillai (2003) propose that education is of importance since it can provide skills that are needed to be able to participate in the labor market. When the number of years of schooling an individual has received an increase, the more likely it is that she will participate in non-family work. Opportunities for women are thus strengthened by more investments in human capital (Besamusca, et al., 2015). Other researchers also note the role of education as a determinant of female labor force participation. Women in most cultures often fulfill household care needs. In developed countries, women are more educated, which increases their opportunities in the labor force. This creates higher opportunity costs for women to exit the labor market to take care of their children, which results in lower fertility rates and higher female labor force participation (Bloom, Canning, Fink, & Finlay, 2009). The traditional division of labor, in which women are responsible for household activities, discourages women from investing in their human capital and also reduces the amount of time they have available for participating in the labor force (Becker, 1985). When there is an investment in human capital, women are more inclined to work because of their earning potential and the increased opportunity cost of not working (Psacharopoulos & Tzannatos, 1989). Although it seems that education is of great importance for the female labor force participation, some hypotheses suggest that education might not be of such significant importance in some less developed, rural areas since there are already fewer possibilities for them (Pampel & Tanaka, 1986).
2.2.2 Household structure
Another argument proposed by Benefo and Pillai (2003) is the influence of the family organization. This theory suggests that female labor force participation depends on the family organization, in which the kinship system is the base. This system is determinantal for many
13 relations and obligations, such as property and political relations. It suggests that women are less likely to work outside the household when the household head is male. In traditional families, the first responsibility a married woman has is taking care of the home and family. Fulfilling these roles influences the way a woman can be involved in economic activities, as well as the extent to which she can participate in it. Economic progress not always benefits the work opportunities for women. Traditional agrarian economies, and economies in which handwork is still of major importance, the production is mainly done by members of the family. In such economies, women can fulfill their duties that are expected of her, and combine this with work. However, when work is not within the family, and the women have to travel for work, problems arise (Collver & Langlois, 1962).
Although differences in the level of the economy might explain differences in labor force participation of women, this alone cannot explain the participation rate of women. The family also plays an important role in the role of women, according to Collver and Langlois (1962). This is also acknowledged by Youssef (1972), but she argues that economic development is not of importance at all in explaining the female labor force participation. The most important explanatory factor is the organization of the family and the kinship systems of countries (Youssef, 1972). What Collver and Langlois (1962) and Youssef (1972) have in common, is that familial composition is found to be an important determinant for differences in female labor force participation. Women do not have to work in countries where there is a stable family system since family members support them, and there exists social control. In countries where this is not the case, and women are not married, or there is a high divorce rate, which leads to a gap in support of their well-being, they need to work to take care of themselves (Collver & Langlois, 1962; Youssef, 1972).
In stable family systems in which men the socioeconomic status of husbands, therefore, also is of influence on the FLFP since it can give opportunities for women. Through their husband, they can work on a social network which can give them access to jobs (Spierings, Smits, & Verloo, 2010). On the other hand, the social norms traditionally prescribe that husbands are the primary breadwinners, and this can negatively influence the possibilities of women (Smits, Ultee, & Lammers, 1996). When a husband earns enough to provide for the whole family, they can also decide that the women do not need to work but can spend more time on housework (de Jong, Smits, & Longwe, 2017).
In Africa, however, polygyny is widespread and common sense is that it related to economic conditions (Boserup, 1970). Especially in farming communities, polygyny is common, since an extra wife means more labor and thus more output. When a husband has
14 more wives, it often also means that it is expected of women to support themselves. It is usual for traditional African families that women support themselves and take care of their children. Especially in rural communities in which women do agricultural work have been found to have higher rates of polygamous households (Boserup, 1970).
Whether the household lives in an urban area or not might also influence whether she is involved in the labor market. As aforementioned, traveling for work might create problems. Living in an urban area can be less problematic since jobs are easier to access in those areas (Collver & Langlois, 1962). This view is also shared with Boserup (1970), she states that women living in rural areas are more involved in farming and there is a low possibility of having a non-farming job since the level of industrialization in such areas is often lower. Rural areas are characterized by a large agricultural sector, while in urban areas, most of the people are employed in the service sector (McCullough, 2017). Urbanization could also influence the labor force participation of women indirectly since living in a rural area is disadvantageous for the odds of staying in school, especially for girls (Huisman & Smits, 2015).
2.2.3 Economic development
One of the factors that are affecting female labor participation is the economic development of a country and the industrialization process. It is expected that a higher level of development increases the possibility of getting a job (Psacharopoulos & Tzannatos, 1989). The higher the level of industrialization, the more a country grows to modern sector employment. The relation between the level of industrialization and female market force participation can be explained with a U-shaped hypothesis. This relation suggests that when the underdeveloped economy becomes more industrialized, the female labor market participation will fall. In the early stages of development, there is a high number of women who participate in the labor force, especially in agriculture. In this stage, the incomes are meager, and the fertility rates high. When technology develops, markets expand, and agriculture gets modern, the division of labor changes, and women's participation in the labor market decline (Momsen, 1991). The economy focusses more on the formal sector and industrial production, while the subsistence sector is losing ground. This industrialization is often accompanied by urbanization, which restricts women even further for participating in the labor market. The low level of education of women and other social restrictions cause that women are cannot benefit from the industrialization. Especially married women face problems in this stage since there might be formal restrictions that they face, or a social stigma that exists (Boserup, 1970; Goldin, 1995). The
15 society keeps developing after industrialization has started. There is a higher education enrollment of women, which is also delaying their engagement in the labor market (Psacharopoulos & Tzannatos, 1989). A structural change in educational attainment, however, will increase the female labor market participation (Gaddis & Klasen, 2014). The development of the service sector increases the opportunities for women, which are often less subject to the stigmatization that is faced by industrial work. Another factor that benefits the labor force participation of women is the declining fertility rate, better child-care possibilities, and the possibility to work part-time, which makes it easier for women to combine work with raising children. Because there is more work available, wages rise, and women are called into duty (Gaddis & Klasen, 2014; Goldin, 1995; Psacharopoulos & Tzannatos, 1989).
The existence of this U-shaped relationship is widely studied in the literature (e.g., Cagatay & Ozler, 1995; Goldin, 1995; Mammen & Paxson, 2000). Some find evidence in favor of the hypotheses, while others deny that such a relation exists. Gaddis and Klasen (2014), for example, do not support the hypothesis and state that historical differences are of importance in the labor force participation of women. Simmalarly Gündüz-Hosgör and Smits (2008) found that Turkey there are areas with a comparable level of development; there are still differences in the labor force participation of women. They argue that is due to the patriarchal ideology of some regions that result in undervaluation and control of women by husbands. In a broader study of 70 countries, Pampel and Tanaka (1986) conclude that there is a U-shaped relationship between industrialization and FLFP, but note that it is also important to consider other factors with information about the family, social and demographic structures. The importance of such factors is also highlighted by Gaddis & Klasen (2014), who found that education and fertility might be better alternative mechanisms for explaining the U-shaped hypothesis. Combining multiple factors is necessary to explain the variation in FLFP since only looking at one theory will be misleading (Pampel & Tanaka, 1986).
2.3 Ethnic inequality
Existing research on female labor force participation is dominated by studies on the macrolevel, which do not give much insight into the drivers of decisions made by women. Studies that have examined the individual behavior of women often do not take into account the structures in which women live, such as their culture or ethnicity (Spierings, Smits & Verloo, 2010). Africa has a high level of ethnic diversity, and this has shaped the nations, regions, and cultures of Africa. When the European powers decided where the border of the African
16 countries should be, ethnic groups were split up, and it exacerbated the ethnic diversity in Africa. This diversity could influence the school attainment of children, but can also cause conflict among the groups that results in a power struggle and is unfavorable for the economic development of nations (Easterly & Levine, 1997). Ethnic diversity is seen as one of the most critical constraints for development in Africa. According to Bates (2000), Africa is poor because of the high instability it faces. People that are part of a smaller, less-advantaged group are being discriminated against and get fewer social opportunities. This feeling causes instability that hinders the development (Bates, 2000). People of disadvantaged ethnic groups, therefore, experience almost no improvement in their productivity. As a result, they cannot improve to a higher social class (Rothchild, 1969). This idea is also proposed by Alesina, Michalopoulos, and Papaioannou (2016), who argue that the wellbeing of people depends on their ethnic identity and that ethnic inequality can determine the social ties between different ethnic groups. Ethnicity thus might influence the FLFP in Sub-Saharan Africa, and it is thus important to investigate a possible relationship.
An example of existing literature that has researched heterogeneity in societies is the work of Mavisakalyan (2015). She found that there is a positive association between linguistic homogeneity and the female labor force participation. This is in line with the findings of Easterly and Levine (1997), who found that in countries with homogenous populations, there is a positive association with economic outcomes. About the influence of ethnicity on labor market outcomes is less known. There is evidence that countries with strong religious views about the role of women in society often have the lowest female labor force participation. Roman Catholic and Muslim countries report the lowest rates, according to Psacharopoulos and Tzannatos (1989). A similar effect is found by Feldmann (2007), who analyzed whether Protestantism affects outcomes in the labor market. He found that countries that are dominated by Protestantism have higher female employment rates than countries with other religions (Feldmann, 2007). The social norms and preferences of religion can hinder gender equality (Easterly & Levine, 1997; Feldmann, 2007).
Ethnic diversity does influence not only the economic development of countries but also the perceptions of gender roles that exist in society. This is examined by Awaworyi Churchill, Nuhu, and Lopez (2019), who argue that the social norms that exist help in the development of stereotypes in which women are not expected to participate in the labor market. Gender differences are rooted in the social and cultural norms that exist in countries. It can, for example, create stereotypes that view men as the primary income producer, which can hinder the engagement of women in the labor market. They found that ethnic heterogeneity expands the
17 gender gaps that exist in society. The ability of ethnic heterogeneity to conserve the social and cultural norms of the groups makes that it can the gender gaps can be preserved or widened. Conserving these norms is possible because ethnic groups are linked with less social cooperation resulting from their cultural drift and their tendency to isolate themselves (Awaworyi Chirchill, Nuhu, & Lopez, 2019). This can, however, not explain why, in some ethnic groups, women are participating more than in another ethnic group. This thesis will try to help in identifying what the causes of differences are by relying on the existing literature about the determinants of female labor market participation.
Traditional views about gender roles also play a role, according to Contreras and Plaza (2010). They find that culture and beliefs about gender roles are more important than other individual characteristics and found that there is a negative relation between culture and female labor market participation. Reimers (1985) found that in the United States, culture may influence the labor force participation directly and indirectly. By comparing home-country groups, she found that culture has a direct influence on the FLMP if there are differences between groups that cannot be observed. Indirectly culture can influence the labor market participation of women through things such as education, fertility, and experiences. This is an important fact since the ethnic identity is of importance when forming cultural norms, preferences, and values, and it is shown that there is an overlap between culture and ethnicity in Africa (Desmet, Ortuno-Ortin, & Wacziarg, 2017).
Following previous research, it thus can be expected that marriage, fertility, and living in a rural area will influence the labor force participation women in Sub-Saharan Africa negatively. At the same time, education will most likely have a positive effect on the labor force participation of women. The influence of ethnicity on FLFP is, however, not yet known. Based on the literature of ethnic inequality, it can be expected that some groups are more favored than others, and therefore get more opportunities. Besides, the more traditional the groups are, the more likely it seems that women have to take the role of the housewife and have to take care of homework and children and are therefore often not expected to work.
18
3.
Data and Method
This chapter consists of three parts. The first part discusses the data sources that are used for this study. The second part presents the methodology used, and lastly, the variables used to answer the research question are explained.
3.1 Data
Data in this thesis is obtained from 321115 women, living in 289 ethnic groups, distributed over 28 countries in Sub-Saharan Africa (Appendix A). The data of the thesis is based on household surveys of the Demographic and Health Survey (DHS; www.dhsprogram.com), which are obtained from the Global Data Lab (www.globaldatalab.org) in which the DHS surveys of multiple countries have been combined. The Global Data Lab has created databases and estimates to measure and analyze different topics for the developing world. The DHS are nationally representative household surveys that are being conducted in developing countries since the 1980s. The surveys are conducted in non-overlapping areas that are randomly selected. These areas are identified as communities, city quarters, or villages. Within these areas, households are randomly selected for the survey. The DHS has several surveys, of which the household survey and the women survey are used in this thesis. The household survey collects basic information of all the household members. The women's survey consists of women aged 15 to 49 that are being interviewed about demographic, socio-economic, and health-related issues (Schrijner & Smits, 2018).
3.2 Method
To examine whether there are differences between ethnic groups in SSA and their FLFP, a multilevel logistic regression analysis will be used. This thesis aims to learn whether females (individuals) are influenced by ethnicity in their labor market force participation. These individuals exist in households, which in turn are nested in other groups such as their village, community, or ethnic group. These groups can be grouped in even bigger unites such as regions or countries. For this thesis, it means that individuals are nested in households that are clustered in ethnic groups, which, in their turn, are again nested in countries. The social groups influence the women they belong to, and these social groups are, in turn, influenced by the people that are part of those groups (Hox, Moerbeek, & Van de Schoot, 2010).
Because of this hierarchical structure of the data, a multilevel approach is the most appropriate. Two levels are of importance, the household level, and ethnic level. The household level is based on the information obtained by surveys of individuals. Ethnicity is the second level, which reflects the values, beliefs, and customs of individuals of the same group. Including
19 the ethnicity is of importance for this thesis since the aim is to find out whether ethnicity also influences the FLFP of countries.
The explanatory variables that will be explained after are referring to individual characteristics as well as characteristics of the larger units. Some of the variables can change from individual to individual, while other variables only vary between ethnic groups. To obtain data on the ethnicity level, variables of interest will be aggregated. Since the dataset is large, it is possible to aggregate data from household level to ethnic (cluster) level. In the DHS, data is obtained on the individual level. Aggregating these to the ethnicity level gives a mean value of those variables which say something about the average behavior of the group. The mean of all individuals of one ethnic group is thus the value that is given to that variable on the ethnic level. To get a representative data sample, the aggregation takes into account the household weight factors that are provided by DHS. These weight factors adjust for the differences in the probability of the selection of households in a sample. Some groups, households, or people are underrepresented in the sample. Therefore they get a higher weight to correct for this uneven distribution and make sure that the sample is representative.
The countries all will get a fixed effect dummy to control for clustering at the national level and other contextual factors. Adding these country-specific dummies will reduce the omitted variable bias. It makes sure that there is controlled for the specific situations of countries in which women live. This corrects for the economic positions of countries since, as aforementioned, the economic development of a country might influence the FLFP.
3.3 Variables
3.3.1 Dependent Variable
The dependent variable will be whether women are working or not. A dichotomous variable measures this variable. A value of 1 indicates that the women are working, while a value of 0 indicates that the woman is unemployed. The upper age limit of 49 is chosen because the women that are interviewed for the women’s survey are questioned until that age. The lower age limit is 18 since younger children are more likely to be still in school and thus are not participating in the labor force.
The definition of working women is that women are active in the lower non-farm and higher non-farm sectors. Women who work in the lower non-farm sector are engaged in the domestic, sales, and service sector, while women in the upper non-farmers are people who have a role as manager, professional, technical, or work in the clerical sector. The ‘not employed’
20 category does not only consist out of women that are not working but also women that are employed in farming. The reason for this is that it is hard to differentiate between non-working women and women working at a farm. Women in Africa sometimes see their tasks at their farm as part of their housework and therefore categorize themselves as not employed. This makes it challenging to establish a difference between the two categories (Spierings, Smits, & Verloo, 2010). Moreover, non-agricultural labor is of more importance to the development of countries and the empowerment of women. Engaging in the non-farm sector is a more valuable step towards female empowerment and independence (Gündüz-Hoşgör & Smits, 2008; Spierings, Smits, & Verloo, 2010).
3.3.2 Independent Variables 3.3.2.1 Ethnic Characteristics
The main objective of this thesis is to observe whether there is a variation of FLFP depending on the ethnic group the women belong to. To capture ethnicity, different characteristics of these groups can be identified. As a proxy for these different characteristics, traditionalism will be used, following the approach used by Spierings, Smits, and Verloo (2010). In their study, several indicators are used to capture traditionalism. These indicators are also used in this theses by aggregating information on the variables from the household level to the ethnic level. This means that the data on these variables are compiled and summarized for each ethnic group. All information about the women and of her household are gathered per ethnic group and aggregated into one single value for the ethnic group. This results in an average for each variable that is allocated to the members of that group.
The first indicator is the age difference between spouses. This reflects the relative position of women in ethnic groups. This is calculated by subtracting the age of the husband of that of the women. The larger the difference in age between the spouses, the weaker her position is in that ethnic group. The education of the partner should also positively affect the FLFP. The educational attainment of husbands reflects the values that are important for them. It is found by Spierings, et al. (2010) that women who have a more educated husband are more often employed than women whose husband had lower education. This is in line with the findings of de Jong et al. (2017), who also found that women of highly educated men tend to work more in the non-agricultural sector. Another important factor is the family composition. Extended families are often found to be more traditional, and therefore this variable is included as a third indicator. When there is no other kin present besides the partner of the women, the family is classified as nuclear. The dummy variable gives a value of 1 if the family is extended, and a
21 value of 0 otherwise. The fourth indicator is whether (1) or not (0) at least one member of the household was involved in a polygynous relation. In regions of Sub-Sharan Africa, polygamy is common and is useful to increase wealth and agricultural output. The wealthier the men are, the more women they can obtain (Boserup, 1970).
The last indicator is the age of the women when giving birth to her first child. The older the woman is when she gets her first child, the more likely it is that she will be employed. Having no children saves time and the time she saves, she can use it to build up a career. At the same time, less is expected from her in the household. This might also have to do with the age at first marriage, and therefore this variable is added to the study as well. The younger the women get married, the more traditional the ethnic groups are expected to be. The woman gets confronted with the household duties that are expected from her and therefore has to spend more time at home.
The more traditional the outcomes of the variables, and thus the more traditional the ethnic groups are, the lower the tendency for women have to enter the labor market (Spierings, Smits, & Verloo, 2010). This means that the lower the age difference, the more educated the husbands and the higher the age at first childbirth and marriage, the better this should be for the FLFP. In ethnic groups where nuclear and monogamic families are the norm, the FLFP should be higher than in ethnic groups where polygamy and extended families are typical.
In the dataset, there are women included that are married, single, divorced, or widowed. For the women who are not married, values are missing for the characteristics of a partner. Those women get the mean scores of women with a partner. There are also women included that do not have children, and thus do not have a value for the age at first childbearing. For those women, the same is done as for the non-married women; they get the mean value of the women that do have children. As the next section will show, the model contains variables to indicate whether women have a partner or children. This makes sure that this procedure does not lead to biased estimates of variables (Allison, 2001).
These variables on the ethnic level could also apprehend the effects on the household level. Therefore all the variables that are obtained at the ethnic level will also be included at the household level variables. This makes it possible to see the effect of the traditional characteristics of the ethnic groups on the FLFP.
22 3.3.2.2 Household Characteristics
Next to the independent variables of interest, several other variables are included. In the existing literature, that has been discussed in the previous section, a large number of factors are found to be associated with female labor force participation. The factors on the personal level of the women that are included in this study are the years of education that she had, and her marital status. The factors that are associated with the women’s household are the occupation of the husband, whether they live in a rural area or not, the number of children of the household, and whether there is a child present that is younger than six years old. Since the household characteristics are associated with the women selected in this study, the household and individual characteristics are recognized as one level.
As aforementioned, the women’s DHS survey interviews women aged 14-49. However, the lower limit for the women included in this study is 18. The reason for this is that there is a possibility that younger women are still in school and thus do not work. It can be expected that this relationship is not linear. Young women, without children or a partner, usually are engaged in the labor market. The older they get, the more likely it is that they get married or get children, which decreases the likelihood that they work. However, at a certain point, they can return to the labor market because their children do not need as much care as in their early years (Momsen, 1991). The age of the women, as well as the age squared, is included to control for life-cycle differences of women. The age of the women is measured as the current age of the women in years.
The years of schooling is used as a proxy for the educational attainment of women. While Spierings et al. (2010) use the completion of primary, secondary, or tertiary schooling as a proxy for educational attainment, this study follows de Jong et al. (2017) and takes the years of schooling. The reason for this is that the average years of schooling for a similar grade can differ between countries, or the grades reflect different levels. This results in a range of 0 to 24 years of schooling. It is expected that higher educational attainment will result in access to better-classified jobs. Women then do not only are classified for farm jobs but can also participate in the non-farming sector (Boserup, 1970; Psacharopoulos & Tzannatos, 1989). The marital status of the women is measured as the current marital status, the value of 0 reflects that she is not married, while a value of 1 is reflecting that she is married or lives together. The original variable included categories that indicated whether women are married, divorced, widowed, or have other situations. Married women are often assigned household tasks, increasing their responsibilities compared to women who are not married. Many studies that have examined the labor force participation of women have included only married women,
23 but in this study also not married women are taken into account (e.g., De Jong, Smits & Longwe, 2016; Gündüz-Hoşgör & Smits, 2008; Mincer, 1962). This will give insight into whether ethnicity also influences women who are not married.
The presence of children in a household can also influence the FLFP. It is not only important to note whether there are children, but also how many children and whether there is a child present under the age of 6. On the one hand, the more children there are in a household, the more time needs to devote to childcare (Angrist & Evans, 1998). On the other hand, when the children get older, they can help with household duties. This, however, depends on the age of children. Young children need much more care than older children. According to Connelly (1992), the lower labor force participation of women with pre-school children was almost entirely devoted to the high amount of childcare needed for these children. Therefore a variable for young children is included in this study. The variable indicates whether there are children below the school-age of 6 present in the household.
The occupation of the husband is included because it indicates the social status of women, as well as the possibilities the women may get from the social network that she and her husband have developed. This variable is measured in four categories. The categories are: not working, farm work, lower non-farm work, and upper non-farm work. Lower non-farm workers reflecting people who are active in the domestic, sales, and service sector, while upper non-farmers are people who have a role as manager, professional, technical, or work in the clerical sector. For each of the categories, a dummy is made reflecting one if the man belongs to the group, and 0 if he does not. There is also a category for missing data for this variable to handle the missing data. This information, however, will not be shown in the analysis.
The urban location of women is also included as an independent variable. Rural areas are often characterized by lower industrialization and more subsistence farming (Boserup, 1970). This influences the FLFP negatively since women have fewer opportunities of getting a non-farming job. This opportunity might also indirectly be decreased due to lower access to education in rural areas (Huisman & Smits, 2015).
Since there is no data available about the income of the households, the International Wealth Index is used as a measure for household wealth. The IWI is an index that uses rating criteria that are the same for every household, independent of the countries or year. It reflects the economic status of households, based on basic assets, such as housing characteristics, consumer durables, and access to basic services. For example, if a household has access to electricity, it can use a refrigerator. This reduces the time that needs to be spent on shopping and therefore, more time is available to do other things. The index runs from 0 to 100, with a
24 value of 0 indicating that the household does have a low-quality home, has no consumer durables, and no access to basic services. On the other hand, a value of 100 indicates that the household had a good quality home, good access to services, and has all the consumer durables included in the index (Smits & Steendijk, 2015). This measure will also be controlled at the ethnic level by aggregating the data from the household to the ethnic level.
3.3.3 Missing Data
For this study, women in the age range from 18 to 49 years old have been selected. The data availability of the ethnic groups is limited, and therefore only countries with useful ethnicity data are selected. This means that countries with only two ethnic groups, mostly data of not specified groups, or a lot of missing values have been deleted from the dataset. Because of missing cases on the variables whether a woman is working or not, their marital status, educational attainment, and the International Wealth Index (IWI) in total 16059 (4.76%) women had to be removed from the initial dataset. For the missing data on the variables that are used for aggregation, another method is applied. These missing values have to do with marital status and whether the women have children. There are women without a husband or children, and thus without information on this variable. For these variables, the dummy variable adjustment method is used. A dummy is created, indicating whether or not an observation is missing for these variables. After that, the mean of the known variables is calculated, and this value is added instead of the missing. The women with missing on a variable thus get the mean scores of women with a partner or children. The model contains control variables indicating whether a woman has a husband or children, and therefore this procedure does not lead to biased estimates. This results in 321.115 women that are being studied, from 28 different countries in Africa with survey data that are used are held between 2000 and 2018. Appendix A presents a list of the countries the year of study used for this study and the number of ethnic groups in the country. Although selecting only countries with useful ethnicity data could lead to a selection bias, keeping countries with the only view ethnic groups would not be suitable since not much can be said about the influence of ethnicity. A list of all the countries in the sample and the ethnic groups per country can be found in Appendix B.
25
4.
Results
Having presented the variables included in this study, this section presents the findings of the analysis. Firstly, descriptive statistics are presented and discussed. Then, the regression results will be presented and explained.
4.1 Descriptive statistics
The descriptive statics of the dependent and independent variables, described in the previous section, are presented in table 1. The table presents the main, standard deviation, minimum and maximum values of the variable as well as the number of observations there are present. Of the women that are included in the dataset, 37 % were working at the time of the surveys. This is in line with what is introduced in the introduction that a majority of women in Sub-Saharan Africa are not involved in the labor market. This means that 63 percent of the women do not have a job, or are engaged in farm work.
Looking at the ethnic factors, it can be observed that the mean age differences between spouses are -9 years, which indicates that on average, husbands are nine years older than their wives. In all ethnic groups, it is common that the man is older than the wife. In some groups, the average is only 4.3 years, while in others, the age difference is, on average, 14 years old. The age of getting married also varies among groups; the average age of getting married for the first time is 18 years old. The average age of getting a first child seems to be related to the age of getting a first child, which is 19 years. This suggests that soon after women get married, they get their first child. The average education of the husband is 5.4 years, but in some groups, husbands almost did not have any education. The International Wealth Index also profoundly differs among groups. While some groups only have 2.51 points on average, others have 85.9 points. Of the women, 19.4 % is involved in a polygamous relationship, and 28.8% live in an extended household. For both variables, there are also ethnic groups in which there are no polygynous households or extended families.
Regarding the household factors, it can be observed that the average age of the women in the dataset was somewhat over 30 years old. In 73.2 % of the cases, the women are married or live together, and on average, the women have 3.4 children. In 61.1 percent of the cases, the women have a child at home that is under the age of 6. Looking at the age difference between spouses, the age of first married, and getting their first child, the mean values are comparable to the values at the ethnic level.
Looking at the husband's occupation, it can be observed that only 5.5 percent is not active in the labor market, while 33.6 % work at a farm and 33 % is active in the lower
non-26 farm sector. The high number of husbands active at farms could be explained by the fact that the majority of households (60.93%) live in rural areas. The mean International Wealth Index is 36.35, which is on a scale of 0-100 relatively low. What is also low is the years of schooling the women attain. Whereas some women have obtained more than 18 years of schooling, the average is 4.9. The years of education for husbands is a bit higher, 5.4 years. Most of the families are not extended with other family members, and almost 20 percent of the women have a polygynous relationship.
Table. 1 Descriptive Statistics: Percentages, means of characteristics of women aged 18-49
Mean(%) Min Max SD
Proportion employed women 37% 0 1 .482
Ethnic Factors
Age difference between Spouses (Ethnicity) -8.93 -14.01 -4.34 1.872
Age at first Marriage (Ethnicity) 18.25 15.08 21.67 1.177
Age of getting first Child (Ethnicity) 19.08 17.30 23.61 0.737 Years of education Husband (Ethnicity) 5.39 0.45 8.29 1.862
Polygynous Household (Ethnicity) 19.39% 0 0.49 0.127
Extended Family (Ethnicity) 28.78% 0 0.73 0.125
IWI (Ethnicity) 37.14 2.51 85.90 13.522
Demographic Factors
Age 30.22 18 49 8.658
Marital status 73.19% 0 1 0.443
Age difference between Spouses -8.96 -77 34 6.734
Age at first Marriage 18.25 13 49 3.873
Age of getting first child 19.07 12 48 3.478
Number of Children 3.35 0 18 2.748
Children <6 years at Home? 61.11% 0 1 0.487
Polygynous household 19.87% 0 1 0.399
Extended Family 29.38% 0 1 0.455
Socio-Economic Factors
Occupation Husband (ref. = Farm) 32.62% 0 1 0.469
Not Working 5.46% 0 1 0.227
Lower non-farm 33.03% 0 1 0.470
Upper non-farm 7.95% 0 1 0.271
Years of education Women 4.88 0 24 4.771
Years of education Husband 5.37 0 24 4.031
Living in Urban or Rural area 39.07% 0 1 0.488
IWI 36.35 0 100 24.746
27
4.2 Analysis
4.2.1 Correlation Matrix
Before running the analysis, a correlation analysis is done. It shows how strong the relationship is between FLFP and the independent variables. Regarding the variables at the ethnic level, it can be observed that there is a positive relationship between the age difference between spouses and FLFP. This means that the smaller or more positive the age difference is, the more women are involved in the labor market. This is in line with the expectations, the smaller the age difference, the better the position of the women. The age at which women get their first child, as well as the age at which they get married, also are positively associated with FLFP. The older the women are, the more they are involved in the labor market, according to the correlations. The education of the husband also has the expected sign, the more educated men in the ethnic group are, the better it is for women.
Furthermore, the correlations show that where there are polygynous households, there can be expected that fewer women are working. In agricultural societies, it is expected of women to support themselves. She has to work on the farm to gather food. On the other hand, if there are extended families, this relationship is positive. The wealth of ethnic groups is also positively related to working women. The wealthier the households, the more women work. On both the individual and ethnic levels, this relationship is positive and significant.
It can be observed that married women have significantly lower FLFP rates. This finding is in line with the expectations that women who have a husband or live together do not work as often as women who do not have a husband. The relationship is significant but not strong. The number of children also negatively influences the FLFP. The more children, the lower the FLFP, which is in line with the expectations. The same applies to whether there are children present under the age of 6 years. Both variables are significantly influencing the labor force participation; however, the relationship with FLFP is weak. Women living in urban areas show significantly higher numbers of FLFP.
Furthermore, older women tend to be more employed compared to younger women. The results also show that the more wealth a household has, the more women work. Regarding the occupation of the women’s partner, it can be observed that when the husband is not occupied or is working at a farm, there is a negative relationship. On the other hand, there is a positive relationship when the husband is working in the non-farming sector. Al, the four occupational groups have the expected sign and are significant.
28 Table 2 Coefficients of the correlations with the proportion of
employed women as the dependent variable
B
Proportion employed women
Ethnic Factors
Age difference between Spouses (Ethnicity) 0.035***
Age at first Marriage (Ethnicity) 0.121***
Age of getting first Child (Ethnicity) 0.128*** Years of education Husband (Ethnicity) 0.111***
Polygynous Household (Ethnicity) -0.012***
Extended Family (Ethnicity) 0.017***
IWI (Ethnicity) 0.181***
Demographic Factors
Age -0.115***
Marital status -0.043***
Age difference between Spouses 0.031***
Age at first Marriage 0.109***
Age of getting first Child 0.078***
Number of Children -0.021***
Children <6 years at Home? -0.073***
Extended Family -0.029***
Polygynous household -0.022***
Socio-Economic Factors
Occupation Husband (ref. = Farm) -0.233***
Not Working -0.021***
Lower non-farm 0.178***
Upper non-farm 0.113***
Years of education Women 0.190***
Years of education Husband 0.166***
Living in Urban or Rural area 0.264***
IWI 0.282***
Observations 321,115
29 In the correlation matrix in Appendix C it can be observed that there are high, and significant correlations between Age difference between spouses at the ethnic level and the following variables that are also aggregated to the ethnic level: Age at first marriage, years of education of the husband, Polygynous household, and extended families. Furthermore, the correlation between age at first marriage at the ethnic level and age of getting first-child, years of education of the husband, International Wealth Index, and polygynous households, all at the ethnic level, are also highly correlated. Other high significant correlations between ethnic factors can be found between polygynous households and years of education of the husband, between extended families and polygynous households. At the household level, the International Wealth Index and living in an urban area are highly correlated, as well as the age of the women and the number of children and the age of getting a first child and the age at first marriage.
These high correlations between the independent variables can cause multicollinearity. It can result in less precise estimates, and that different effects are not distinguished well form each other. Estimates might be too high or have signs that are the opposite of what was expected theoretically (Mansfield & Helms, 1982). To check whether the variables all can be in one model, the highly correlated variables will be removed from the dataset separately. If the variables change when deleting one of the highly correlated variables, it means that the others already capture the effects of that variable. This variable then can be deleted from the model since it gives biased results.
4.2.2 Multilevel Logistic Regression
A multilevel logistic regression analysis is used to test whether ethnicity influences the FLFP. It includes all the independent variables, which makes it able to control for underlying mechanisms. Tables 3 and 4 present the results of the multilevel logistic analyses. Model 1 only includes the ethnicity variables and model 2, only the demographic and socio-economic variables. Model 3 combines both the ethnic and the household variables to determine what the effect of ethnicity is on FLFP. All the three models include country-specific dummies to control for country-level differences.
Model 1 shows the log odds and odds ratios for the ethnic factors. Five of the seven factors show significant effects. Only the age of first childbirth and the dummy for the extended family are nonsignificant and therefore seem not to affect FLFP. However, it might be that due to the high correlations discussed earlier, the estimates are not correct. Theoretically, the age of first childbirth is the weakest factor. There is already a variable included for the age of first marriage, which is highly correlated to first childbirth. Women get married, and soon after that,
30 they start having children. Because of the high correlation, there will also be a model without the variable to check whether there estimates change and whether the variables become more accurate.
The education of the husband shows a strong significant positive effect, indicating that women with more educated women are more likely to be employed. This is in line whit previous literature, claiming that women with a more educated husband more often work in the non-agricultural sector (de Jong, et al., 2017; Spierings, et al., 2010). For the variables on family structure, it can be observed that only the variable for polygynous households has a significant positive effect. Women that are involved in a polygynous relation are more likely to be employed. This can be explained by the fact that they are expected to provide for themselves and thus need to work to earn money. The International Wealth Index for the ethnic groups also has a significant positive effect. The log odds of working are higher for women living in wealthier ethnic groups.
Model 2 in table 3 shows what the effect is of the household factors on FLFP. The results show that the log odds of age is larger than one, implying that the likelihood that older women are more likely to be employed. However, the squared term of age is negative, implying there a non-linear downward sloping effect on the likelihood of being employed. Young women are more likely to work then middle-aged women; however, this effect diminishes over time. This is in line with the expectations. Before the childbearing age, it can be expected that women work, but the likelihood decreases as they get older. However, when they have children, the likelihood might increase again.
The log odds for marital status are negative and significant, indicating that married women are less likely to be employed. This evidence is in favor of previous literature claiming that married women have more obligations and that their husbands provide for them and thus do not need to work (van der Klaauw, 1996; Youssef, 1972). The age difference between husband and wife also seems to be of importance, although the effect is small. The smaller or more positive the age difference between the spouses, the more likely it is that the women work. The relative position of women is stronger when the age difference is smaller (Schrijner & Smits, 2018), and therefore women have more freedom and possibilities to have a career. The age at first marriage is not significant, which is possibly due to the correlation with age at getting a first child. As aforementioned, model 4 will leave this variable out of the analysis.
Looking at the number of children and whether there are children under the age of six at home, it can be observed that the log odds of working are lower for women with more children. The more children, the more care needs there are present. Therefore it also was
31 expected that the number of children below the age of 6 would have a negative value; however, looking at model 2 shows that it has a positive value. This also could be because of the high correlation between the two variables. However, as table 6 in Appendix D shows, there is not much change when deleting the variable for children under the age of 6 present, and therefore this variable is kept in the model.
For the family structures at the household level, there are significant results. Living in an extended family decreases the log odds of being employed. This is surprising since it could also be reasoned that living in an extended household means that there are more adults present that can help with household duties and care needs. The results, however, are in line with Spierings, et al. (2010), who also found that employment rates for women were lower when they were living in an extended family. This negative relationship can be explained by the fact that such families are more often traditional, and thus the women expect to take care of children and stay at home (de Jong, et al., 2017). Polygynous households, on the other hand, positively influence the FLFP. This is in contrast with the findings of Spierings, et al. (2010), who found that the employment rates are lower for women that were living in a polygynous household.
The socio-economic factors show that when the husband is working at a farm, the log odds of being employed are the lowest. In line with de Jong, et al. (2017), it is found that women with a husband working in the non-agricultural sector have the highest odds of being employed. Both the year of education of the women and the years of education of the husband is positive. The higher the education of the husband, as well as that of the women, the higher the likelihood that the women will be employed. This confirms the findings of many previous studies that better educational attainment leads to higher FLFP ( (Buvinić & Furst-Nichols, 2016; Gündüz-Hoşgör & Smits, 2008; Spierings, Smits, & Verloo, 2010).
Lastly, the location and wealth of the household also are positively related to FLFP. Women in urban areas have higher log odds of being employed. In those areas, more non-agricultural jobs are available, and thus it is more accessible to being employed (Van den Broeck & Kilic, 2019).
