WOMEN’S INHERITANCE RIGHTS, BARGAINING POWER
AND FERTILITY: EVIDENCE FROM KENYA
MSc Economics Thesis Supervised by Prof. dr. Pauline Rossi
Marielle Schweickart
University of Amsterdam
15 July 2018
ABSTRACT
Though millions of African women rely on land for their lives and livelihood, far fewer women than men enjoy secure rights to it. How do land inheritance legal reforms affect the health and education achievement of girls? This paper estimates the impact of the revised Kenyan
Constitution of 2010, which granted daughters equal shares in inheritance relative to sons regardless of marital status, by examining fertility indicators of women. As the policy reform reduced litigants’ rights for Muslim women differently than non-Muslim women, I employ a differences-in-differences strategy, exploiting variation in pre-reform inheritance rights across religious groups. It also investigates changes in birth spacing following an earlier, 1990 reform, which decreased Muslim women’s inheritance rights. Estimates suggest that the fertility of women living in urban settings is more quickly impacted by the 2010 policy change than rural women. The analysis also suggests that Muslim women spaced their children less following the 1990 policy change. Evidence also suggests that barriers to policy enforcement and gender discrimination remain.
Schweickart - 2 Statement of Originality
This document is written by Student Marielle Schweickart who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
TABLE OF CONTENTS
1. INTRODUCTION & MOTIVATION ... 3
2. BACKGROUND & CONTEXT ... 6
3. CONCEPTUAL FRAMEWORK ... 9
4. DATA ... 11
4.1 Descriptive Statistics ... 12
5. METHODOLOGY ... 14
5.1 Main Estimation Strategy ... 15
5.2 Duration Model ... 16
6. RESULTS ... 19
6.1 Age at First Birth ... 19
6.2 Age at First Wed ... 19
6.3 Births Last Year ... 20
6.4 Contraceptive Use ... 20
6.5 Birth Spacing ... 21
7. CONCLUSION ... 22
Schweickart - 3 1. INTRODUCTION & MOTIVATION
Legal rights are often at the root of broad patterns of inequality. The discrepancy between the legal rights of men and women is much greater in developing countries than it is in developed countries (Doepke et al. 2012). In poor contexts, inheritance is one of the main ways to acquire and control property, yet women are often discriminated against in their ability to inherit land and obtain property rights. According to the World Bank, this legal inability can deprive women of bargaining power within the household, significantly undermining their economic security and independence, as well as their access to opportunities, social mobility and wealth creation
(World Bank 2012) This is particularly relevant in rural societies, in which a significant amount of households’ endowment of physical capital is in the form of land (Agarwal 1994).
Additionally, equal rights to girls and women inheriting property has broader human capital implications. Women with greater inheritance rights in Sub-Saharan African countries have safer sex practices and significantly lower HIV rates than those living in countries with less equal laws (Anderson 2017). Furthermore, the benefits of inheritance laws go beyond that of women; they can change the human capital outcomes of children as well. An increased likelihood of inheriting land in India was found to be associated with increased age at marriage for girls, higher
education attainment of girls, and also increased household investment in daughters ((Deininger et al. 2012)
Leading development economist and philosopher Amartya Sen posits that women’s
empowerment, including property rights, is a key instrument for reducing fertility rates (Sen 2001). Fertility rates, in turn, have high economic and health consequences. Family planning and birth spacing leads to positive health, social and economic outcomes for women, families and society. The United Nations has stated that low income households with many children often find it more difficult to get out of poverty than those with less children, and high fertility societies face greater demands for services from their youthful populations (UNPD 2009). If increased bargaining within the household due to access to land inheritance influences fertility, the significance of equal inheritance laws would be unquestionable.
Over the last two decades, extensive statistical evidence has tied women’s education and labor force participation with demographic transition in many developing countries. However, there is
Schweickart - 4
far less empirical evidence on the influence of property, and in particular inheritance rights, on fertility. This paper investigates the fertility impacts of a statutory legal reforms that granted Kenyan women equal inheritance rights ranging the years 1981-2010, exploiting variations prior to the various reforms across religious groups. Building off of Harari’s (2017) study of human capital effects in Kenya, this paper’s contribution is to specifically examine the role of
inheritance laws on fertility outcomes to indicate broader development impacts. It complements other studies of the effects of inheritance rights, such as fertility in Nigeria and HIV infection rates in India (Godefroy 2018, Roy 2015).
The 1981 Kenyan LSA (henceforth: LSA) was a landmark piece of legislation allowing for equal inheritance rights for Kenyan men and women. Prior to 1981, inheritance rights were dictated by customary ethnic laws, which vary from tribe to tribe, but a commonality of these laws is that married daughters are not allowed to inherit the estate of their deceased father. Where customary laws are followed, Muslims in Kenya and elsewhere follow Koranic law, which allots women one-half of the inheritance share that her brothers assume. The 1981 Law repealed all pre-existing laws on succession and made women’s inheritance rights a statutory law in Kenya, regardless of religious affiliation. Following a period of pushback from the Muslim population, in 1990 the government acquiesced and amended the Law, invalidating the LSA for persons who, at the time of their death, were Muslims. Muslim women again were only entitled to half of the inheritance of their brothers. It was not until 2010, with the new Kenyan Constitution, that the law was again changed to create equal inheritance rights for women and men.
I investigate the effects of these reforms in three series of estimations. The first series estimates the impact of the reform on fertility. Firstly, I exploit the timing of the 2010 reform and the variation in religious affiliation to estimate the causal impact of the reform on fertility timing. Secondly, I use the timing of the 1990 reform to estimate the change in birth spacing. The second series of estimations uses the 2010 reform and aims to identify the mechanisms that explain these changes, beginning with an investigation into the changes in ideal family size of couple, and also includes a study of contraceptive use and marriage timing. In each of these analyses, I am
estimating the impact of the incremental reform, wherein Muslim women are transitioning from equal shares of their brothers’ inheritance to 0.5 shares in 1990 and vice-versa due to the 2010
Schweickart - 5
reform. All estimations draw from Kenyan Demographic and Health Survey data, which is rich with fertility, health and education outcomes at a household level.
My estimates suggest that the fertility of women living in urban settings is more quickly impacted by the 2010 policy change than rural women. It also strengthens work by Harari (2017), suggesting that Kenyan women’s fertility was altered by the 1981 reform in the form of fertility and nuptial postponement; my analysis suggests that Muslim women spaced their children less following the 1990 policy change, which had negatively impacted their inheritance shares. Legal evidence suggests that barriers to policy enforcement and gender discrimination remain, particularly in the form of constitutional dualities.
This paper adds to literature on intra-household bargaining and decision-making. In a non-unitary model of household behavior, a household consists of two decision-makers. The allocation of resources between those two decision-makers – generally a husband and wife – influences bargaining and associated socio-economic outcomes. Relative income is not the only possible variable that may affect the intra-household decision making; the process can change due to a range of variables that shift members’ respective bargaining positions (Chiappori et al 2002). Factors, such as family law, that affect opportunities of spouses outside of the marriage can influence the intra-household balance of power, ultimately shifting the allocation of resources.
Recently, several papers have investigated the link between family law - particularly inheritance laws – and the allocation of household resources in developing contexts, demonstrating that a woman’s ability to control household resources is associated in increased investments in
education for girls, elevated labor supply for women, and safer childbirth (Deininger et al. 2012, Health and Tan 2016, Harari 2017). However, analysis of the human capital effects of the Indian Hindu Succession Act, which strengthened women’s inheritance rights between 1986 and 1994, demonstrated that shifts in external factors can potentially be detrimental to women. For
example, the reform was found to influence more suicides and wife beatings, interpreted as the result of greater conflict over household property, and it was found to affect female mortality, due to the dowry system and therefore increased cost of having a daughter (Anderson and
Schweickart - 6
Genicot 2015, Rosenblum 2015). The net effect of this type of legislative reform has not been thoroughly explored.
The rest of the paper is organized as follows Section 2 describes the demographic, socio-political and legislative context in Kenya, as well as a timeline of statutory inheritance reforms. Section 3 discusses the conceptual framework for interpreting the impacts of the reform. Section 4 presents the data used in the study and Section 5 examines the specifications and empirical strategy I utilize. In Section 6, I discuss the results of the investigation and in Section 7, I conclude.
2. BACKGROUND & CONTEXT
Kenya, a former British colony that borders the Indian Ocean in Eastern Africa, won independence in 1963. Since then, ethnic, religious and political fractionalization has been prevalent. Fertility rates in Kenya are some of the highest in the world, with approximately four children born for every woman. However, Kenya’s government has not been able to deliver the public services necessary to support with this growth: unemployment rates are among the highest in the world, and infrastructure is poor.
Today, the country’s population exceeds 48 million; three out of four Kenyans in rural areas, and more than half live below the poverty line. According to the 2009 census, Christians account for 83% of the population, made up of Protestants (48%), Catholics (23%), and other Christians (12%). Over 11%, or approximately 4.3 million people, of the population is Muslim, and the remaining 6% includes, Hindus, and those who identify with traditional religions or are non-religious (Kenya NBS 2009).
Muslims in Kenya predominantly inhabit the Northeastern Province, a sparsely populated area largely occupied by nomadic groups, and the coastal region, which includes the major port city Mombasa. Islam has remained the dominant religion in these areas, along with traditional tribal institutions and values, but pockets of Muslim populations are also concentrated in the interior of the country in various urban such as Nairobi, Nakuru, Eldoret, and Kisumu. Notably, Kenyan Muslims, like their Christian compatriots, demonstrates a sociocultural heterogeneity that cuts across the various racial and ethnic groups in the country, including Somalis, Arabs, and people of mixed Arab-African descent (Ndzovu 2017).
Schweickart - 7
As a former British colony, Kenya adopted a system of common law, while countries that were colonized by continental European countries retained civil codes of law (Anderson 2017). Under traditional common law, married women have virtually no legal rights; husbands were
considered the sole owners of all marital property. This strict interpretation has made it such that women have little or no protection in the event of marital dissolution, and widowhood is
generally associated with a vital loss in use and control over assets, as a husband’s assets are returned to his family.
Inheritance laws have evolved over the last four decades, starting in 1981 with the LSA; the Law was in fact passed in 1972 but not put into effect for nine years. Prior to this act, property rights were defined by a “complex interplay of customary law, statutory law and Islamic Law” as directed in the 1969 constitution (Harari, 2017). For most Kenyans, inheritance laws were dictated by the customary law, which, for virtually all ethnic groups, meant women were denied any right of in heritance. Some tribal customary laws recognized unmarried daughters may as a son and gave them equal rights. Others only granted “spinster” daughters equal inheritance rights. For Muslims, Koranic law was followed, entitling women to one-half of the inheritance share that goes to each of her brothers. The 1981 LSA consolidated the different systems of inheritance into one uniform law, establishing equal inheritance rights for female or male children to their parent’s property (Cooper, 2011). It is applied automatically in the case of intestate succession – wherein the deceased person has not made a will – and in cases in which is a will but not reasonable support for any dependents (Section 38). Although it is their right
granted by the 1961 African Wills Ordinance, most Kenyans die without a will. (Mutongi, 2007). The 1981 LSA also provided some protection a spouse; in an effort to discourage land grabbing from widows, the Law states that “where an intestate has left one surviving spouse and a child or children, the surviving spouse shall be entitled to (a) the personal and household effects of the deceased absolutely; and (b) a life interest in the whole residue of the net intestate estate.” (Section 35) However, if the surviving spouse is a woman, and she were to remarry, her inheritance rights would be invalidated (whereas a surviving husband would retain the interest upon remarriage). In the case of a widow’s remarriage, her children would then inherit her estate. The Law also provides guidance in the case of polygamy, stating that “personal and household
Schweickart - 8
effects and the residue of the net intestate estate shall, in the first instance, be divided among the houses according to the number of children in each house.” (Section 48)
Between 1981 and 1990, there was continuous outcry by Kenya’s Muslim population, which regarded the LSA as a repudiation of the government’s assurance given to the group at
independence. The 1897 Native Courts Regulations Ordinance had proclaimed that for Muslims would follow any laws of succession that were written in the Koran. They sought to be exempted from the Law, which conflicted with the Koranic principles. The text clearly states that a widow with no children should receive one-fourth of her husband’s estate, and if there are children, she takes on-eighth; women in polygamous marriages should share the quarter or the eighth,
depending on if they have children (Kameri-Mbote 1995). Regarding children, the Koran states “Allah ordains concerning your children are that the male shall have a share equivalent to that of two females. If the children are females numbering two or more, their proportion is two thirds of the inheritance.”1 In, 1990, the Kenyan government amended the LSA to exclude application
to Muslims.
In her study, Harari (2017) argues that the contentiousness and continued discussion of women’s inheritance rights in Kenya, as well as a 2008 Kenya Law Reform Commission study inviting civil society organization to provide feedback on the LSA, suggests that knowledge of its existence is widespread. There is little evidence to prove this, however in 2002, UN-HABITAT reported that the rights of women under the LSA have largely been upheld2, though some
incorrect interpretations of the law have been made, on occasion disinheriting married daughters. Others have reported that Kenya’s 1969 Constitution has discouraged full obedience of the LSA; Section 82(4) recognized customary law, which heavily opposes women’s inheritance, to be applicable in matters of personal law (Kameri-Mbote 1995, Cooper 2011). However, there is no sufficient data to be able to analyze if asset ownership changed after the reform.
1 Note: The Koran dictates that only one-third of a deceased person’s estate can be organized by a Will. The
remaining two-thirds is distributed under intestacy rules, which fix shares allocated to the widow or widower, father and children (IELRC 1995).
2 One notable example is a 2005 High Court Case, Manunzyu v. Musyoka, in which a deceased man’s sons and wife
sought to invalidate a married daughter’s inheritance claim through customary law. The judge ruled that the customary law in question was discriminatory as it prevented the daughter from inheriting a portion her father’s estate, as is permitted under the LSA (Cooper 2011).
Schweickart - 9
In 2010, after a long period of deliberation, the Kenyan government ratified a new Constitution, which was intended to clarify any ambiguity regarding customary law that remained with the previous version. Article 60 of the new Constitution provides for the elimination of gender discrimination in law, customs and practices related to land and property (Cooper 2011). This would in effect allow Muslim women to again inherit an equal share of assets as their brother(s) from a deceased parent. However, legal pluralism persists, particularly in Section 82(4), which recognizes customary systems of law to govern inheritance. As a result, Kenya’s courts have ruled inconsistently on cases related to a women’s rights to inherit.
3. CONCEPTUAL FRAMEWORK
In this study, I explore the fertility outcomes of allowing women to inherit parental property.3 The primary channel by which this question will be investigated is through bargaining, and the changes in a woman’s intra-household bargaining power. There has been wide consensus that a unitary household model is insufficient in reflecting reality; with heterogeneous preferences, the distribution of resources within the household affects intra-household bargaining and is
associated with socioeconomic outcomes at the individual level (Anderson and Eswaran 2009). Furthermore, women and men consistently devote proportions of their income to family needs at different levels, affecting the health and education outcomes of children, as well as fertility (Strauss et al. 2000, Dyson and Moore 1983).
Therefore, this study utilizes a non-unitary model of household behavior, wherein a household consists of two decision-makers. These individuals both have a distinct preferences over the number of children that should be in the household. Evidence has suggested that outcomes in the household depend on which decision-maker in the household has an income stream or ownership of an income-generating asset (Chiappori et al 2002). In a model in which the woman has little or zero bargaining power, the number of children in the family will be determined by the male. However, if the woman has an outside option – an income stream or asset apart from her husband – her bargaining power within the household will rise.
3 As with Harari (2017), I do not focus on the ability of widows to inherit from deceased husbands, which was also
provided in the 1981 LSA; there is little economic reasoning to give evidence to how the bargaining power of a wife would change while her husband is still living.
Schweickart - 10
Following Becker and Tomes (1979) work on wealth transfers, this investigation suggests that the optimal mix of human and physical capital given to sons and daughters will shift with a change in inheritance policy. Several studies have suggested that women’s asset ownership and socioeconomic outcomes have a strong positive correlation, indicating a systematic difference in the preferences of males and females; females attach higher needs to children’s welfare and family needs (Strauss et al. 2000, Qian 2008). The channel suggested in this paper is that of mothers’ bargaining power; as it increases, intra-household decisions regarding human capital investment will shift to align more with the preferences of women. This generally involves a smaller number of children at optimum, and, as concluded by the many works demonstrating that the preferences of women are skewed toward the well-being of children, a greater investment in girls.
As suggested by Harari (2017), with an inheritance reform, “the optimal amount of human capital bequeathed to sons and daughters will change, in a direction which depends on whether human and physical capital are complements or substitutes in the income-generating process.” I hypothesize that for mothers, they are complements. With a fixed household income and a policy shift that ensures assets for girl children, mothers will place equal weight on each gender of child, thereby incentivizing a smaller family size. Additionally, as suggested by French historians as the cause of a marked decline in birth rate in the 19th century, a change in
inheritance to be equitable among all children may drive couples to reduce their target family size in order to prevent the fragmentation of family assets (Garner 1914).
Thus, couples’ fertility choices are affected to the degree that the relative bargaining weight of wives increases due to these outside shocks. The key underlying assumption is that a change in bargaining power does not rely on women receiving their inheritance rights immediately
following the policy change, but instead that the opportunity and legal backing is enough to shift their standing in the household. This was shown in human capital changes due to the Indian Succession Law; though women were found to not inherit more land in practice, the Law had large and significant effects on girls’ education (Deininger et al. 2013). Therefore, though there is not enough data to prove that women’s inheritance was realized, it is plausible that a shift in intra-household bargaining could be detected.
Schweickart - 11
This paper hypothesizes that Kenya’s inheritance policy changes in 1981 and 2010 acted as positive shocks to women’s potential asset ownership. This, in effect, increases the bargaining power of women in a household. It also has the potential to reduce the assets, and therefore intra-household bargaining power, of men in a intra-household, who now need to share their parental inheritance allotment with one or many sisters. In 1981, all wives in Kenya experiences the positive shock. In 1990, the shock was a negative one for Muslim women and in 2010, the shock was again positive for Muslim women.
4. DATA
This study utilizes data from six Kenyan Demographic and Health Surveys (DHS) spanning the years 1989-2014: 1989 (DHS-I), 1993 (DHS-II), 1998 (DHS-III), 2003 (DHS-IV), 2008-2009 (DHS-V), and 2014 (DHS-VI). Funded by the United States Agency for International
Development (USAID), these nationally-representative household surveys provide a range of health and fertility indicators of women aged 15-49, and indicators related to their children. The surveys used are Standard DHS Surveys, meaning they have large sample sizes, usually between 5,000 and 30,000 households, and typically are conducted every five years to study and compare indicators over time. Basic demographic data and information on educational attainment and income is also collected, and in each round, a sub-sample of households is selected for an additional questionnaire to be answered by male household heads (15-54).
Utilization of DHS datasets is advantageous for several reasons. Firstly, DHS surveys include information on religious and tribal affiliation, key indicators that otherwise Kenya has been careful to report on due to internal political risks. Furthermore, the relatively large sample sizes of DHS surveys allows for fairly precise estimates, though the empirical strategy relies on a minority subset of the population. Another benefit of DHS surveys is that the general amount of comparability across waves is such that measurement error problems due to pooling together different waves are eased.
However, this collection of DHS Surveys also includes several disadvantages. In 2003, the DHS surveys began including the North Eastern province of Kenya; excluded in the first three
Schweickart - 12
Muslim populations. After further investigation, the assumption that the degree of comparability across waves is high was brought into question. According to the DHS 1989 report, despite an emphasis on obtaining district-level data for planning purposes, it was decided that reliable estimates could not be produced from the Kenyan DHS for all 32 districts in the National Sample Survey and Evaluation Program, unless the sample were expanded to an unmanageable size. However, it was felt that reliable estimates of certain variables could be produced for the rural areas in 13 districts. This caveat was consistent across the DHS 1993 and DHS 1998 surveys, but not in subsequent years, creating a dichotomy in the sample data used for analysis: DHS surveys I-III (1989-1998) and DHS surveys IV-VI (2003-2014).4
Another key limitation of the data is their timing; all DHS surveys were administered following the initial 1981 reform, and only the first occurred prior to the 1990 amendment. I was unable to gain access to the Kenyan Fertility Survey (1977-1978), a subset of the World Fertility Survey program, established by the International Statistical Institute and funded by the United Nations Population Fund (UNFPA), USAID and the UK Overseas Development Administration due to a lack of online formatting and time constraints. Moreover, only one survey has been conducted following the 2010 reform. Therefore, and due to the nature of a woman’s fertility, this analysis focuses on past or cumulative outcomes, such as the timing of fertility onset and marriage – rather than concurrent health outcomes.
4.1 Descriptive Statistics
I split the sample data is split into two cohorts, Cohort 1 including DHS surveys 1989, 1993 and 1998 and Cohort 2 including DHS surveys 2003, 2008-2009, and 2014. Based on sample issues reported in final DHS reports, each year’s sample has been reduced to only include 15 rural districts and two urban districts.5 As mentioned previously, Cohort 1 does not include the Northeastern region.
As demonstrated from Table 1 below, the differences between Muslim and non-Muslim women in Cohort 1 and 2 contrast greatly. In Cohort 1, Muslim women are much more likely to live in
4 Whereas Harari (2017) argues that her results are “qualitatively unchanged if the North Eastern province is
included,” I reject the assumption that the Muslim cohorts are the same due to the sampling variances across Kenyan DHS surveys.
5 It is noted by DHS that reliable estimates for certain variables could be produced for the rural areas in 15 districts:
Bungoma, Kakamega, Kericho, Kilifi, Kisii, Machakos, Meru, Murang’a, Nakuru, Nandi, Nyeri, Siaya, South Nyanza, Taita-Taveta, and Uasin Gishu. The sample also includes the two largest Kenyan urban centers, Nairobi and Mombasa.
Schweickart - 13
urban areas, are much more likely to have no education, and are less likely to fall into the lowest wealth quintile. They are more likely to have had more than one union (marriage) in their
lifetime, more likely to be in a polygamous marriage, and are more likely to be using a modern contraceptive method. However, in Cohort 2, Muslim women are much more likely to be in the lowest wealth quintile than non-Muslim women and are much less likely to have a job. They are again more likely to have had more than one union, and be in a polygamous relationship, but in Cohort 2, they are much less likely to be using a modern contraceptive method.
Based on these differences, this investigation uses the Cohort sample data separately. Cohort 1 is used to investigate the 1990 Referendum, reversing inheritance rights gains made for Muslim women. Cohort 2 is used to investigate the fertility impacts of the 2010 Constitution on Muslim women, regaining their inheritance to again equivalent to that of their brothers.
Schweickart - 14 5. METHODOLOGY
Three legal changes affecting women’s inheritance rights have occurred in Kenya over the last four decades: the 1981 LSA, which granted an equal share of parental inheritance to female and male children, regardless of ethnic or religious affiliation; the subsequent 1990 Amendment to the LSA, following backlash from the Muslim community, which exempted Muslim women from the Law; and finally, the 2010 Kenyan Constitution, which overrides the 1981 LSA, thereby creating equal inheritance rights for males and females in the country. This generates four different inheritance periods for women, as exhibited in Table 2 below.
Data limitations, and the newness of the 2010, reform make it impossible to observe total fertility; women exposed to the form haven’t completed their child-bearing years at the time of the survey. Instead, I use two key channels to investigate whether fertility has changed due to one of both of the inheritance reforms. Using the 2010 reform, I am able to analyze fertility onset, restricting my sample to those between 15 and 24, as well as the likelihood of having a child in the last year. I use this same specification to test the mechanisms that may be at play, such as delayed onset of first marriage, type of contraceptive use, likelihood of having a child last year, and impacts on women’s number of ideal children. Due to violation of the parallel trend assumption, I am unable to test changes in ideal number of children directly.6 Instead, I use a proxy; determining changes in use of joint family planning methods – those that require both women and men’s involvement such as withdrawal and condoms - to investigate intra-household bargaining power. Though data limitations disallow me from employing my main empirical strategy to investigate the 1990 Reform, I utilize a duration model of birth intervals to gain insights regarding changes in birth spacing.
Schweickart - 15
These empirical strategies follow an intent-to-treat analysis, including every subject who is “randomized” by religion, and thus not factoring noncompliance. A weakness of using this design with DHS datasets is that it is impossible to know if a woman’s parents were already deceased before the inheritance policy reform; those women will not experience any increased bargaining power. Kenyan DHS survey data does not have information on the timing of parental death. Thus, it is likely that I will have included women with deceased parents in the “treatment” group. However, while this will likely diminish the estimates reported here, it is unlikely that this timing or pattern of parental death differs severely between religious groups enough to invalidate the identification strategy.
5.1 Main Estimation Strategy
The first investigation follows Godefoy (2018) and Harari (2017) in exploiting within-country variation in customary inheritance laws across religious groups. The specification relies on a difference-in differences model between cohorts exposed and not exposed to reforms, comparing Muslim women (experiment) and non-Muslim women (control), over two time periods. Based on a key study by Duflo (2001), the model’s identifying assumption is that without the inheritance policy changes, the outcomes of interest would have evolved over time with an equivalent time trend across the different groups. This design is robust to variations in time-invariant
characteristics of, in this study, the different religious groups.
The main empirical specification is represented by the following (Equation 1):
𝑌𝑖𝑟𝑒𝑤= 𝛼 + 𝛽1𝑃𝑜𝑠𝑡 + 𝛽2𝑀𝑢𝑠𝑙𝑖𝑚 + 𝛽3𝑃𝑜𝑠𝑡 × 𝑀𝑢𝑠𝑙𝑖𝑚 + 𝛾𝑟𝑡 + 𝛿𝑙𝑡 + 𝜃𝑋𝑖𝑟𝑙𝑠𝑎𝑤𝑖𝑒+ 𝜇𝑙+ 𝜎𝑤 + 𝜑𝑟+ 𝜀𝑖𝑟𝑒𝑤
𝑌𝑖𝑟𝑒𝑤 is an outcome of interest for individual i, residing in region r, belonging to ethnicity e, surveyed in wave w. 𝑋𝑖𝑟𝑒𝑤 represents a set of individual level controls (age, education, wealth
quintile); 𝜇𝑙 , 𝜎𝑤 , and 𝜑𝑟 are location(rural/urban), wave, and region fixed-effects, respectively; 𝛾𝑟𝑡 is a region-specific time rend and 𝛿𝑙𝑡 is a location-specific time trend employed to capture region and/or locational effects that may be correlated with the error term such as an expansion of health clinics into rural areas. As these factors are related to ethnicity, I do not include ethnicity fixed effects. Ethnicity dummies capture time-invariant characteristics of different
Schweickart - 16
ethnic groups, the basis of customary law, to which Kenyans still strongly identify.7 Post is a
binary variable, with a value of 0 if prior to the 2010 reform and 1 if 2010 or later. Muslim is also a binary indicator, which takes on the value 1 if the individual is Muslim and the value 0
otherwise. I estimate the coefficients of the equation in OLS and cluster standard errors at the religious group times birth cohort level, following Harari (2017) and La Ferrara and Milazzo (2017).
The key coefficient of interest is 𝛽3, which measures the change in the dependent variable for Muslin women – the treatment group – relative to other, non-Muslim women – the control group. The identification assumption is that no factor that is not controlled for and that could influence fertility decisions changed at the same time specifically for Muslim women. Kenya’s historical context supports this assumption; the changes in question did not stem from in shifts in fertility preferences, but from sudden policy changes that affected women’s legal rights. While the 1981 LSA did not apply to several so-called “gazetted” districts located in the semi-desertic part of the country, the 2010 new Constitution was effective across the country.8
5.2 Duration Model
Due to data constraints described in the previous section, I was not able to investigate fertility changes due to the 1981 LSA. Furthermore, a lack of information prior to 1989 did not allow for confirming the common trend assumption required for the methodology above. I therefore alter my estimation strategy to investigate birth spacing, another key fertility outcome; if a woman prefers fewer children, she will likely wait longer to become pregnant between births. Following Lambert and Rossi (2016), the idea underlying this empirical strategy is to compare fertility choices of Muslim women and non-Muslim women before and after the 1990 Addendum to the 1981 LSA using a duration model of birth intervals. The existence of a change in fertility due to the 1990 Referendum is confirmed when the length of the birth interval before the next child is shorter for Muslim women. Based on a key study by Leung (1998), arguing that the
7 Ethnic affiliation may also serve as a strong proxy for area of birth (Ferré 2009).
8 The main reason for this exemption was that the semi-desertic region of Kenya in the Northeastern part of the
country are areas of communal ownership and would thus be difficult to attribute to individuals (UN HABITAT 2002). The Duration model, which is the only empirical analysis related to this policy, uses a sample of DHS surveys that excludes this Northeastern region. Furthermore, according to the 2009 Kenyan census, though these areas encompass an estimated 60% of Kenya’s land, only 15% of the population resides within their boundaries (Kenya Bureau of Statistics 2010).
Schweickart - 17
duration model is more powerful than traditional the methods of parity progression ratio and OLS regression, this model has been utilized in studies related to sex preference across Asia and, more recently, Africa (Lambert and Rossi 2016).9
As explained in the textbook Handbook of Transport Modelling, “Hazard-based duration models represent a class of analytical methods which are appropriate for modeling data that have as their focus an end-of-duration occurrence, given that the duration has lasted to some specified time. This concept of conditional probability of termination of duration recognizes the dynamics of duration; i.e., it recognizes that the likelihood of ending the duration depends on the length of elapsed time since start of the duration.” (Hensher and Button 2007, p. 105) This model’s key strength, as outlined by Leung (1998), is its ability to account for right-censored observations: women that had not completed their fertility at the time of the survey. This feature is applicable as long as the duration variable and the right-censoring variables are independent. As the DHS survey administration is unrelated to latest births, this requirement is satisfied for these purposes. To test whether birth interval is related to the inheritance policies, I use a model, similar to that in Lambert and Rossi (2016), with proportional hazard. The variable interest R is the duration between successive births, measured in months. My coefficients of interest measure the effect of the 1990 Addendum on the subsequent birth interval, among Muslim and Non-Muslim women. My model considers intervals between births n and (n+1) for n = {1, 2,3} and my main
specification is a pooled regression of all intervals.
In the specification with proportional hazard, the hazard function is the instantaneous probability of having another child at date r. It is modeled as the following:
𝛿(𝑟) = 𝛿0(𝑟) × exp (𝑋′ 𝑛𝛽)
In the model, 𝛿0(𝑟) is the baseline hazard function, common to all women, while 𝑋𝑛 is a vector
of individual covariates at birth n that may affect the hazard function. The probability note to have an additional child until at least t is:
9 Hazard-based models are not new; they have been used in biometrics and industrial engineering for decades. Due
to the association with time until failure, whether related to human body functioning or industrial components, they have also been labeled as “failure-time” models, though “duration” models more generally reflects the scope of application to any duration occurrence (Hensher and Button 2007).
Schweickart - 18
𝑆(𝑟) = Pr(𝑅 > 𝑟) = exp (− ∫ 𝛿0(𝑢)𝑑𝑢 × exp (𝑋′ 𝑛𝛽)) 𝑟
0
In this analysis, 𝑋𝑛 includes Treated, which, as above, is a dummy variable equaling 1 if the woman is a Muslim and 0 if the woman is a non-Muslim.
It should be noted that this model does not take into account child deaths, which may lead to an overestimation of the effect; the model assumes that a woman is having another birth to have another child, not replace a lost child. It also includes birth order to control for potential difference between parities, and age at birth n. In a simple difference in difference framework, the estimation amounts to computing the empirical counterpart of the difference in difference term of this theoretical framework:
𝛽𝐷𝐼𝐷 = (𝔼[𝑟]𝑃𝑜𝑠𝑡90𝑀𝑢𝑠𝑙𝑖𝑚− 𝔼[𝑟]𝑃𝑟𝑒90𝑀𝑢𝑠𝑙𝑖𝑚) − (𝔼[𝑟]𝑃𝑜𝑠𝑡90𝑁𝑜𝑛−𝑀𝑢𝑠𝑙𝑖𝑚− 𝔼[𝑟]𝑃𝑟𝑒90𝑁𝑜𝑛−𝑀𝑢𝑠𝑙𝑖𝑚)
In regards to the proportional hazard model, 𝑒𝐵𝑘is the hazard ratio between two women that
differ only by one unit of 𝑋𝑘. In this model, 𝛽1, 𝛽2 and 𝛽3 are respectively the coefficients on Post, Muslim and Muslim x Post. Thus, 𝑒𝐵3 measures the hazard ratio of interest: the odd-ratio
that captures the difference in birth spacing between Muslim women and Non-Muslim women following the 1990 Addendum. If the hypothesis described above is true, then one would expect 𝑒𝐵3 < 1, i.e. that a decrease in inheritance rights in 1990 would lead to a decrease in spacing
between births for Muslim women. With lesser intra-household bargaining, and no outside option, Muslim women would want to bear more children.
I investigate birth intervals by using a semi-parametric estimation of the Cox proportional hazard model; a partial likelihood method of estimation makes it feasible to estimate the vector of coefficients (𝛽) without imposing a function form on 𝛿0(𝑟) (see Cox 1972, Lambert and Rossi
2016). I employ robust standard errors clustered at the woman level to correct for the correlation between the error terms related to the different birth intervals for the same woman. As the exact marginal-likelihood method is not possible when clustering standard errors, I employ the Breslow method to address ties among non-censored duration. A key limitation of this model it would be noted, is that this method is not designed to model individual-level unobserved heterogeneity, such as a woman’s potential for reproduction, which could be helpful in calculating more precise estimates.
Schweickart - 19 6. RESULTS
6.1 Age at First Birth
Table 3 presents the results of the OLS model defined in Equation 1, where 𝑌𝑖𝑟𝑒𝑤 is the age of female i when she gave birth for the first time. A graph using descriptive data from DHS surveys IV, V and VI demonstrate that the parallel trends assumption is met. In this analysis, the sample size is reduced to girls aged 15-24 and younger; the average age of first birth is approximately 18, so the inclusion of the entire cohort (aged 15-49) would likely include trends unrelated to the 2010 Constitutional amendment. Note the inclusion of controls, time trends and clustering are individually included as the columns progress from left to right in the results table.
Theoretically, inheritance rules affect fertility. As stated above, I hypothesized that the 2010 Constitution, and subsequent shift from Muslim women’s inheritance allotment from half of that of their brothers’ to equal inheritance would reduce women’s fertility. Because total fertility is not observed, I examine fertility onset to investigate whether there has been a shift in timing of entry to motherhood for Muslim women due to the reform. Though positive, the results are not significant in any of the specifications, and lower in significance as controls and other measures are included.
6.2 Age at First Wed
Relatedly, another key test to examine fertility is to investigate whether there has been a shift in the timing of entry into marriage. The two assumptions here are that firstly, fertility onset and number of births is correlated with being in a union, and secondly, that a postponement in nuptuality would therefore lead to a decreased total fertility rate. This rationale also relates to findings from India that inheritance reforms increase schooling in girls and thus could be related to the impact of inheritance reforms on mothers’ resources. In other words, if a wife has greater control of household resources, she will invest them in her children and girls will stay in school longer; if they are in school longer they are less likely to get married at a very young age. However, the empirical results of this potential shift are not significant, as with the previous analysis. As discussed in the next section, this may be due to an alternative channel at work, and would likely have a lagged impact.
T able 3: Age of First B irth Y ears (of Age) Lo w e st W ealth Quar- tiles Highest W ealth Quar- tiles (1) (2) (3) (4) (5) (6) (7) P ost 0.096 (0.078) 0.034 (0.093) -0.41 (0.257) -0.413 (0.363) -0.674 (0.389) 5.75 (1.26) -0.824* (.435) Muslim -0.0617 (0.166) -0.702 (.238) -0.593 (.259) -0.591 (0.299) -0.458 (.0.264) -0.199 (.418) -0.753* (.423) Muslim x P ost 0.212 (0.197) 0.472 (.267) 0.425 (0.297) 0.425 (0.272) 0.397 (0.315) -0.425 (.500) 0.924** (.436) V ector of Con trols No Y es Y es Y es Y es Y es Y es Region x Y ear No No Y es Y es Y es Y es Y es Lo cation x Y ear No No Y es Y es Y es Y es Y es Cluster (RegionxY ear) No No No Y e s Y es Y es Y es Surv ey W eigh ts No No No No Y es Y es Y es R-Squared 0.005 0.229 0.23 0.231 0.242 0.192 0.263 No. Observ ations 6,017 5183 5183 5177 5177 2190 1926 Mean Dep enden t V ar. 17.79 17.79 17.79 17.79 17.97 17.46 18.38 Notes: This table rep orts estimations of Equation 1, using data from DHS 2 008-2009 and 2013. The whole sample is c ompri sed of one observ ation p er individual w oman i aged b et w een 15 and 24 at the time of the surv ey . Muslim is an ind ic ator v ariable equal to 1 if i iden tifies as Muslim; Post is a dumm y v ariable equal to 1 if t > 2010. Standard errors are clustered at the region and y ear lev el. Standard errors in paren theses. *p < 0.05, **p < 0.01, ***p < 0.001. Column (6) res tri c ts the sample to the lo w est tw o quartiles. Column (7) restricts the sample to th e highest tw o quartiles.
T able 4: Age at First M arriage Y ears (of Age) Lo w e st W ealth Quartiles Highest W ealth Quartiles (1) (2) (3) (4) (5) (6) (7) P ost 0.059 (0.089) 0.109 (0.105) -0.327 (0.276) -0.331 (0.260) -0.667* (0.288) 2.55 (2.24) -0.664** (0.310) Muslim -1.14*** (0.170) -0.600** (0.249) -0.581** (0.269) -0.581*** (0.208) -0.24 (0.298) -0.474 (0.530) 0.126 (0.528) Muslim x P ost 0.052 (0.202) 0.187 (0.278) 0.214 (0.308) 0.213 (0.250) -0.114 (0.350) -0.111 (0.610) -0.323 (0.559) V ector of Con trols No Y es Y es Y es Y es Y es Y es Region x Y ear No No Y es Y es Y es Y es Y es Lo cation x Y ear No No Y es Y es Y es Y es Y es Cluster (RegionxY ear) No No No Y es Y es Y es Y es Surv ey W eigh ts No No No No Y es Y es Y es R-Squared 0.026 0. 213 0.214 0.214 0.217 0.143 0.209 No. Observ ations 5,743 4783 4783 4777 4777 1974 1887 Mean Dep enden t V ar. 17.51 17. 51 17.51 17.51 17.84 17 18.41 Notes: This table rep orts estimations of Equation 1, u sin g data from DHS 2008-2009 and 2013. The comprise ones one observ ation p er individual w oman i aged b et w een 15 and 24 at the time of the surv ey . Muslim is an indicator v ariable equal to 1 if i iden tifies as Muslim; Post is a dumm y v ariable equal to 1 if t > 2010. Standard errors are clustered at the region and y ear lev el. Stan dard errors in paren theses. *p < 0.05, **p < 0.01, ***p < 0.001. Column (6) restricts the sam p le to the lo w est tw o quartiles. Column (7) restricts the sample to th e highest tw o quartiles.
T able 5: Lik eliho o d of Ha ving a Birth in the P ast 1 Y ear Lo w e st W ealth Quartiles Highest W ealth Quartiles (1) (2) (3) (4) (5) (6) (7) P ost -0.19*** (0.05) -0.22*** (.006) -0.013 (0.015) -0.013 (0.018) -0.017 (0.022) 0.211* (0.119) -0.032 (0.021) Muslim 0.06*** (0.010) 0.045*** (0.017) 0.043** (0.018) 0.043 (0.032) 0.004 (0.029) 0.107 (0.074) -0.025 (0.031) Muslim x P ost -0.016 (0.012) -0.039** (0.018) 0.039* (0.020) -0.039 (0.040) -0.01 (0.037) -0.157* (0.084) 0.056 (0.037) V ector of Con trols No Y es Y es Y es Y es Y es Y es Region x Y ear No No Y es Y es Y es Y es Y es Lo cation x Y ear No No Y es Y es Y es Y es Y es Cluster (RegionxY ear) No No No Y es Y es Y es Y es Surv ey W eigh ts No No No No Y es Y es Y es R-Squared 0.003 0 .023 0.023 0.023 0.024 0.023 0.016 Observ ations 39,523 34,093 34,093 34,059 34,059 11,836 15,182 Mean Dep enden t V ar. 14.40% 14.40% 14.40% 14. 40% 13.30% 18.11% 10.25% Notes: This table rep orts estimations of Equation 1, u sin g data from DHS 2008-2009 and 2013. The comprise ones one observ ation p er individual w oman i aged b et w een 15 and 49 at the time of the surv ey . Muslim is an indicator v ariable equal to 1 if i iden tifies as Muslim; Post is a dumm y v ariable equal to 1 if t > 2010. Standard errors are clustered at the region and y ear lev el. Stan dard errors in paren theses. *p < 0.05, **p < 0.01, ***p < 0.001. Column (6) restricts the sam p le to the lo w est tw o quartiles. Column (7) restricts the sample to th e highest tw o quartiles.
T able 6: Join t F amily Planning and Mo dern F amily Planning Join t F amily Planning Mo dern F amily Planning Lo w es t Quar- tiles Highest Quar- tiles
Rural Urban Lo w e st Quar- tiles Highest Quar- tiles
Rural Urban (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) P ost -0.019 (0.042) -0.012 (0.106) -0.014 (0.047) -0.116* (0.042) 0.011 (0.046) -0.046 (0.036) 0.004 (0.096) -0.042 (0.036) 0.1 (0.029) -0.077** (0.038) Muslim -0.083* (0.040) -0.116 (0.079) -0.108* (0.056) 0.002 (0.072) -0.214*** (0.067) 0.049 (0.034) 0.062 (0.063) 0.047 (0.045) -0.002 (0.057) 0.109* (0.056) Muslim x P ost 0.071 (0.051) 0.107 (0.092) 0.093 (0.076) 0.014 (0.086) 0.154* (0.084) -0.043 (0.043) -0.057 (0.082) -0.041 (0.055) -0.008 (0.076) -0.082 (0.065) V ector of Con trols Y es Y es Y es Y es Y es Y es Y es Y es Y es Y es Region x Y ear Y es Y es Y es Y es Y es Y es Y es Y es Y es Y es Lo cation x Y ear Y es Y es Y es Y es Y es Y es Y es Y es Y es Y es Cluster (Re-gionxY ear) Y es Y es Y es Y es Y es Y es Y es Y es Y es Y es Surv ey W eigh ts Y es Y es Y es Y es Y es Y es Y es Y es Y es Y es R-Squared 0.039 0.045 0.046 0.038 0.054 0.032 0.068 0.029 0.042 0.033 No. Observ ations 13976 4188 6668 8377 5599 13976 4188 6668 8377 5599 Notes: This table rep orts estimations of Equation 1, u sin g data from DHS 2008-2009 and 2013. The comprise ones one observ ation p er individual w oman i aged b et w een 15 and 49 at the time of the surv ey , at y ear t. Muslim is an indicator v ariable equal to 1 if i iden tifies as M uslim; Post is a dumm y v ariable equal to 1 if t > 2010. Standard errors are clustered at the region and y ear lev el. Standard errors in paren th e ses. *p < 0.05, **p < 0.01, ***p < 0.001. Columns (2) and (7) restrict the sample to the lo w est tw o quartiles. Colu m n 2 (3) and (9) restrict the sample to the highest tw o quartiles. Columns (4) and (9) restricts the resp onden ts li ving in rural lo cations. Column s (5) and (10) restrict the sample to resp onden ts living in urban p opulations.
T able 7: B ir th Spacing (1981-1997) Whole Sample Whole Sample Rural Urban No Radio Has Radio (1) (2) (3) (4) (5) (6) P ost 0.459*** (0.013) 0.806*** (0.036) 0.792*** (0.035) 0.835 (0.152) 0.788*** (0.053) 0.827*** (0.048) Muslim 1.48 (0.412) 0.473* (0.206) 0.335*** (0.123) 2.36 (3.98) 0.284* (0.190) 1.24 (2.46) Muslim x P ost 0.738* (0.122) 0.69** (0.131) 0.683* (0.164) 0.975 (0.344) .666 (.217) .755 (.190) Con trols No Y es Y es Y es Y es Y es Log-lik eliho o d -14572 -13499 -11546 -1922 -5123 -8091 Observ ations 36871 36871 31820 5051 14886 21842 Clusters 12738 12738 10460 2278 4993 7776 Notes: Robust standard errors are in paren theses (clustered at the w oman lev el). This table re p orts estimations using data from DHS 1989, 1993 an d 199 8. The sample consists of ones one observ ation p er birth after 1981 p er individ ual w oman i aged b et w een 15 and 49 at the time of the surv ey . Muslim is an indicator v ariable equal to 1 if i iden tifies as Muslim; Post is a dumm y v ariable equal to 1 if t > 1990. . Standard errors in paren theses. *p < 0.05, **p < 0.01, ***p < 0.001. Column (3) restricts the sample to resp onden ts living in rural lo c ation s; Column (4) restricts the sample to resp onden ts living in urban lo cations; Column (5) restricts the sample resp onden ts that do not ha v e a radio (a p ro xy for lo w er income); and Column (6) restricts the sample to re sp onden ts that ha v e a radio (a pro xy for h igher income).
Schweickart - 20 6.3 Births Last Year
As previously discussed, data limitations and the recentness of the 2010 Constitutional adoption do not allow me to observe total fertility. However, the DHS survey asks a question to each women regarding how many births they gave in the past year. For each individual, this value is either 0, 1 or 2, which I convert into a likelihood of giving birth in the last year (12 month) period; the variable of interest is equal to 1 if the woman gave one or two births last year and is equal to zero if she did not . I am then able to analyze whether that value changed for Muslim women following the reform using the main difference-in-difference specification. The results are as follows:
Similarly to Table 3 and 4, Table 5 begins, with Column (1), using a model with the least amount of specifications, working up to the full model including controls, time trends, clustering and weighting in Column (5). Overall, these estimations have much lower R values than the previous analysis. The coefficient of interest – Muslim x Post – is negligible, and not significant.
However, when the sample is narrowed, analyzing individuals in the two lowest quartiles of wealth and the two highest quartiles of wealth separately (Columns 6 and 7, respectively), the value for the lowest two wealth quartiles is significant at a 10% level. The negative value (-0.157) indicates that following the Constitutional reform, Muslim women in the lowest quartiles were less likely to have had a birth in the last year following the reform. This could indicate that the most immediate effect of increased inheritance women for women, and particularly poorest women, is to slow having children. However, the R-squared value on this makes the estimate imprecise.
6.4 Contraceptive Use
Although I am unable to use the main specification to evaluate the intra-bargaining channel directly, i.e. how women’s number of ideal children has changed across religious groups and between couples, I can use a proxy to investigate the channel. The DHS survey asks women if they are using a contraceptive method and if so, what kind. Overall, Non-Muslim women are much more likely to use a method of contraception over the period 2008-2009 to 2014.10
Following Anderson (2017), I organize these into groups: those that require the consent of her male partner (henceforth: joint family planning), such as condom, abstinence or the withdrawal 10 See Appendix 3 for descriptive data on their use by religion.
Schweickart - 21
method, and those that do not. In order to understand if any change in the drop or increase in joint family planning is related to a shift to modern contraceptive methods.11
The right column in Table 6 reports the results from an estimation of the main estimation strategy, where the dependent variable is instead “Joint Family Planning,” which is equal to one if a woman uses a condom, abstinence or the withdrawal method. The variable is equal to zero if the woman if she does not use contraception, or is using any other modern or traditional method. The left column in Table 6 reports the results from an estimation of the main estimation strategy, where instead the dependent variable is “using a modern method of contraceptive.” It is equal to 1 if she is using a modern method of contraception and equal to 0 otherwise.
In Table 6, Columns (1) and (6) use the entire sample, while Columns (2), (3) and likewise (7) and (8) use restricted samples based on wealth quintile. Columns (4) and (5) and (9) and (10) also use restricted sample, but this time based on urban/rural differences.
We see from Table 6 that again, the explanatory power of our model is low – the R-squared remains under 10 percent in all specifications. As expected per my graphs [Appendix 1], the coefficients on joint family planning indicators are positive, but only significant (at the 10% level) for those living in urban areas. This may indicate that women in urban areas are quicker to change their behavior; they are closer to media and have larger networks, so information on legal changes and rights would reach them sooner. Also as expected, the coefficients on modern contraceptive methods is negative, but not significant. More research is needed, particularly on the healthcare system of Kenya and trends related to the availability of contraceptive methods in urban and rural regions, to better understand this trend.
6.5 Birth Spacing
Finally, the results of the Cox estimation are presented in Table 7. These results are the most significant of the methods used here. When the entire sample of births between 1981 and 1997 is considered for a sample size of 12,738 women, the coefficient of interest – Muslim x Post- is significantly different from zero at the 5% level. In order to investigate what segments of the Muslim population are driving this in particular, Columns (3) and (4) split the sample into urban
11 Modern contraceptive methods include: contraceptive pill, intrauterine device, injection, condoms, female
Schweickart - 22
and rural groups, and Columns (5) and (6) divide the sample by whether the respondent owns a radio or not, a proxy indicator for wealth.12
These results from Column (2) indicate that Muslim women spaced their children 30% less than Non-Muslim women following the reform. Column (3) indicates that this change was most likely driven by women residing in rural locations, but the insignificance of results in Columns (5) and (6) demonstrate the result was not due to a particular change in poor or wealthy households.
7. CONCLUSION
In this study, I provide estimates of the impact of two inheritance law reform changes, altering women’s inheritance rights in Kenya. I exploit variation in inheritance rights across religious groups and survey cohorts to evaluate how improved statutory inheritance rights affects fertility. This analysis assumes a theoretical model of a collective household: spouses bargain over the Pareto efficient frontier, and as the wife’s outside option goes up, she is able to have greater power in household decision-making. Together, parents must decide on how much capital to invest in their children. If we are to assume parents are altruistic, they will maximize a collective utility function that includes both their own consumption and their children’s future incomes (Becker and Tomes 1979). This model posits that parents will negotiate the optimal mix of human and physical capital to give to their offspring, given their relative comparative advantages in income-generating activities, and hypothesizes that relatively, wives prefer to have fewer children and to invest more in the children they have.
Overall, my results in regards to the 2010 adoption of a new Constitution do not follow a linear pattern. I start by considering fertility onset and compare cohorts of women between the age of 15 and 24 before and after the reform. I find a positive but insignificant result. When I break the sample into wealth quintiles, I find that the wealthiest 40% of Muslim women delay their births by nearly a year following the 2010 reform. Using the same specification, I then consider nuptial onset and find similar – positive yet insignificant – results. Even when breaking the sample into related to wealth and urban/rural location, the no significant results are found. As mentioned previously, this may be due to a lag; later marriage is generally correlated with girls staying in
12 Though some recodes of DHS surveys include income and wealth quintile information, Kenyan DHS birth
Schweickart - 23
school longer. However, it is therefore the girl’s mother who is pivotal in this indicator, ensuring her daughter has the resources to do so. If this is true, there would likely be a delay in the change of nuptial timing, which would not be captured in the 2014 DHS survey.
Next, I used the model to investigate fertility changes directly, using a question in the DHS survey that asks how many births the respondent had in the past year. I created a binary variable with this question to test the change in the likelihood of Muslim women having a child in the previous year. The sample as a whole produces negative but significant results, however when broken into population segments, the poorest 40% of the population showed significantly negative results, indicating that Muslim women were 17% less likely to have a child in the previous year when asked indicate 2014 than they were in 2008/2009. There are unfortunately no other results that justify this result; it may indicate a change in fertility for rural woman, or may be capturing a trend among Muslim women that I was unable to control for. I also use the specification to investigate changes in joint family planning method use, as a proxy for a shift in intra-household bargaining, and find that similar to age at first birth analysis, only an increase in these types of contraceptive methods are only significant for urban women. This strengthens my hypothesis that the impact of inheritance policy changes has a quicker effect in urban locations. Finally, I employ a duration model to test whether the 1990 Addendum to the 1981 LSA
impacted birth spacing for Muslim women. I find a clear negative impact on Muslim women’s birth spacing, i.e. fewer months between births before and after the policy reversal. These results strengthen Harari (2017), who suggests that women who were exposed to the 1981 tend to
postpone marriage and childbirth.
Further recent investigation of legal proceedings and in media suggests that though the 2010 Constitution was intended to supersede all other previous laws, thereby clarifying laws that preceded it, it is clear that dualities remain.13 Based on this, it is likely that many Muslim
women are still barred from equal inheritance rights, which is likely driving the non-linear results of my analysis around the 2010 Constitution. These indicators should continue to be monitored following the completion of successive DHS surveys. However, my secondary analysis shows that birth spacing was reduced in Muslim women following the 1990
13 For instance, Islamic Sharia courts are still exist, and are allowed in the Constitution. These alternative courts, as
well as language that still recognizes customary law as one of the sources of law in Kenya’s new Constitution, demonstrate clear dualities still existing in Kenyan laws. (Kiriminy 2017)
Schweickart - 24
Referendum. This is a clear negative impact that should not be overlooked. The evidence of these changes suggest that equal inheritance policies at the statutory level can have an impact for women and communities, even in societies where poor legal enforcement and male-centered social norms persists.14 Based on this evidence from Kenya and a growing body of related work,
inheritance legislation in Sub-Saharan Africa can be a strategic starting place for social change, particularly in contexts with deep-rooted biases or persisting customs.
14 “Most Kenyans have faced difficulty in pursuing their land rights through the official court system due to
Schweickart - 25 8. REFERENCES
Agarwal, B. (1994). A field of one's own: Gender and land rights in South Asia (Vol. 58). Cambridge University Press.
Anderson, S. (2018). Legal Origins and Female HIV. American Economic Review, 108(6), 1407-39.
Anderson, S., & Eswaran, M. (2009). What determines female autonomy? Evidence from Bangladesh. Journal of Development Economics, 90(2), 179-191.
Anderson, S., & Genicot, G. (2015). Suicide and property rights in India. Journal of Development Economics, 114, 64-78.
Becker, G. S., & Tomes, N. (1979). An equilibrium theory of the distribution of income and intergenerational mobility. Journal of political Economy, 87(6), 1153-1189.
Chiappori, P. A., Fortin, B., & Lacroix, G. (2002). Marriage market, divorce legislation, and household labor supply. Journal of political Economy, 110(1), 37-72.
Cooper, E. (2011). Inheritance in Kenya. Chronic Poverty Research Centre (CPRC), Policy Note 1.
Deininger, K., Goyal, A., & Nagarajan, H. (2012). Do changes in inheritance legislation improve women's access to physical and human capital?: evidence from India's Hindu succession act. NCAER working papers on decentralisation and rural governance in India; no. 2, March 2012. Doepke, M., Tertilt, M., & Voena, A. (2012). The economics and politics of women's rights. Annu. Rev. Econ., 4(1), 339-372.
Duflo, E. (2001). Schooling and labor market consequences of school construction in Indonesia: Evidence from an unusual policy experiment. American economic review, 91(4), 795-813. Dyson, T., & Moore, M. (1983). On kinship structure, female autonomy, and demographic behavior in India. Population and development review, 35-60.
Ferré, C. (2009). Age at first child: does education delay fertility timing? The case of Kenya. The World Bank.
Garner, J. W. (1914). The Decreasing Population of France. The Popular Sceince Monthly, 85(15), 247-259.
Godefroy, R. (2016). How women's rights affect fertility evidence from Nigeria. The Economic Journal.
Harari, M. (2014). Women’s inheritance rights and bargaining power: Evidence from Kenya. Department of Economics of MIT, 1-47.
Heath, R., & Tan, X. (2014). Intrahousehold bargaining, female autonomy, and labor supply: Theory and evidence from India. University of Washington.
Schweickart - 26
Hensher, D. A., & Button, K. J. (Eds.). (2007). Handbook of transport modelling. Emerald Group Publishing Limited.
Kameri-Mbote, P. (1995). The Law of Succession Act in Kenya: Gender perspectives in property management and control.
Kimiriny, D. S. (2017). Kenya’s laws of succession and the bill of rights of the Kenya constitution (2010): a study in legal reform (Doctoral dissertation, Strathmore University). Kenya Law Resource Center, http://www.kenyalawresourcecenter.org/search/label/succession. Kenya National Bureau of Statistics (2010), Census 2009.
Lambert, S., & Rossi, P. (2016). Sons as widowhood insurance: Evidence from Senegal. Journal of Development Economics, 120, 113-127.
Leung, S. F. (1988). On tests for sex preferences. Journal of Population Economics, 1(2), 95-114. Ndzovu, H. (2017). Muslims in Kenyan Politics: Political Involvement, Marginalization, and Minority Status. Northwestern University Press.
Qian, N. (2008). Missing women and the price of tea in China: The effect of sex-specific earnings on sex imbalance. The Quarterly Journal of Economics, 123(3), 1251-1285. Rosenblum, D. (2014). Unintended consequences of women’s inheritance rights on female mortality in india. Economic Development and Cultural Change, 63(2), 223-248.
Roy, S. (2015). Empowering women? Inheritance rights, female education and dowry payments in India. Journal of Development Economics, 114, 233-251.
Sen, A. K. (2001). Development as freedom. Oxford University Press.
Strauss, J., Mwabu, G., & Beegle, K. (2000). Intrahousehold allocations: a review of theories and empirical evidence. Journal of African Economies, 9(Supplement_1), 83-143.
UN-HABITAT (2002). Rights and Reality: Are women’s equal rights to land, housing and property implemented in East Africa?.
United Nations Population Division (2009). Policy Brief: What would it take to accelerate fertility decline in the least developed countries?”.
Walsh, J. (2003). Kenya, double standards: women's property rights violations in Kenya (Vol. 15). Human Rights Watch.
World Bank (2012). Women, Business and the Law.
Schweickart - 27
Schweickart - 29
Appendix 2: Ideal Number of Children (2003-2014)
