Joint Land Titling and Female Empowerment:
Evidence from Ethiopia
Catherine Humphrey
Student number: 10635629
Master’s thesis under the supervision of prof. dr. Erik Plug
MSc Economics with a specialization in Development Economics
Department for Economics and Business
University of Amsterdam
14th August 2017
Statement of Originality
This document is written by Student Catherine Humphrey who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is 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.
Abstract
Using Ethiopia’s 21st century joint titling certification initiative, this research estimates how increased female land rights can impact the prevalence of female empowerment measured through female intra-household bargaining power in three of Ethiopia’s main rural areas: Amhara, Oromia, and Southern Nations, Nationalities and Peoples Region (SNNPR). Exploiting the DHS’ 2005 and 2011 and ERHS’ 1997 and 2009 data surveys, cross sectional OLS analysis and first-differences regression, respectively, this research estimates a positive impact of exposure to the titling initiative on female empowerment through increased intra-household decision making.
I. Introduction
Worldwide, land is recognized as an important asset in securing rural households the basic means for subsistence, market production and a safe base for which to shelter and develop families (FAO, 2014). In the rural economy, women provide an important resource through their roles as farmers, labourers and entrepreneurs (FAO, 2011). Researchers estimate that women make up 40 percent of the world’s agricultural labour force, predict between 60 to 80 percent of food in developing countries and half of the world’s food supply is produced by females (Momsen, 1991; Mehra and Rojas, 2008), yet globally only 20% of the land is owned by women. Moreover, with this figure standing at less than 10% in developing nations, women continue to face unequal and severe constraints in access to assets and productive resources, including land rights, compared to men. (FAO, 2014).
Achieving gender equality and empowering women, remains both a development outcome itself and “mechanism to stimulate growth, reduce poverty, and promote better governance” (Malhotra, et al. 2002). The socioeconomic value of women’s role in a nation’s development is still restricted by deep-rooted gender based discriminatory barriers often as a result of patriarchal attitudes and social norms (Tefera, 2013; Sida, 2013). Increasing female empowerment and advancing the status of women not only helps women reach their full potential in the household, workplace and community, but through “elevating women in isolation there is a multiplier effect on society as a whole” (UNFPA, 2005; Peterman, 2011 as cited in Tefera, 2013). Despite varying definitions of 3
female empowerment across literature, concepts commonly involve the expansion of choice, evidence of women exercising increased agency and evidence of more influence and control over their own lives. Intra-household bargaining power has become a dominant indicator in the measurement of female empowerment, particularly in developing nations where men and women are often bound to traditional gender roles (Argarwal, 1997).
In particular, the need for women’s equal access to land has recently gained momentum both on the international political agenda and by development practitioners alike (Allendorf, 2007). Female land rights are advocated as not only a necessary step in removing discrimination against women, but also as an important tool in the empowerment of women, increasing productivity, efficiency and welfare, and in the fight against hunger and poverty (Allendorf, 2009; Agarwal, 1994). In Ethiopia, due to women’s low socioeconomic status, cultural norms and discriminatory laws, rural women remain one the most in marginalized groups when it comes to access to assets, productive resources and decision making. In recent years, government interventions have made solid attempts to remove discriminatory laws and practices, with the 21stcentury joint land titling initiative being one of the
most prominent. (Sida, 2013)
Through exploiting this exogenously determined joint titling initiative, I estimate the impact of increased female land rights on female empowerment, indicated through household decision making and bargaining power through a randomized control trial setting.
This paper is structured as follows: Section II summarizes the literature identifying the research question and outlining the conceptual framework which motivates my hypothesis. Section III describes the data used to facilitate this research and describes the methodology used. Section IV presents the main results and discusses interesting findings. Section V Looks at other results and presents heterogamous effects and conducts a robustness check. Section VI concludes the findings and discussion for this paper.
II. Literature Overview
(i) The Ethiopian Context
Ethiopia is a largely rural and agricultural country located in East sub-Saharan Africa. With 83 percent of Ethiopia’s population living in rural areas, agriculture is the main source of livelihood for
80 percent of the population and employs 85 percent of the country’s economically active (FAO, 2015). Ethiopia’s rural economy is made up of approximately 12 million of small-hold farms, which are largely “subsistence, rain fed farming systems” producing 95 percent of the agricultural GDP (FAO, 2015). It is recorded that female farmers perform up to 75 percent of farm labour, which accounts to 70 percent of household food produced (USAID, 2015). The Ethiopian population is highly diverse, with over 80 ethnic groups and six religions, the majority following Christianity and Islam beliefs (Sida, 2013) Marriage is a “fluid state” in Ethiopia; divorce is not unusual and polygamy, although now banned, still persists in the SNNPR region. (Pankhurst, 1992 cited in Kumar 2015, p.408; Holden et al. 2008)
(a) Ethiopia’s Land Reform
In 1975 Ethiopia underwent a large egalitarian style land reform under the Marxist inspired Derg regime (Deininger, 2008). The reform transferred land ownership from absentee private landlords to the state level- redistributing land rights to households based on family size (Rahmato 1994 cited in Holden and Yohannes 2002; p.573). Land rights were non-transferable unless through inheritance and employing people to work on the land was forbidden (Holden and Tefera, 2008).
This land allocation scheme disadvantaged females from the outset. Female headed households were unequally allocated land (smaller and less fertile plots) compared to male headed households could not hire labour to undertake the traditionally male dominated farming activities; and the transfer of land was only permitted through inheritance, in which patrilineal inheritance was common (Tesfa, 2002. Kumar and Quisumbing, 2010, cited in Girma et. al 2014). Additionally, in areas where polygamy was common, tenure insecurity increased as it was only the traditional head of the household - men - (Gebrehiwot, 2007 cited in Girma et. al) who were registered and additional wives were excluded from the registration process and deprived of rights to the land: consequently multiple households’ land was left undocumented. Consequently, women’s weaker societal position meant female headed households from polygamous relationships could find it difficult to manage or rent out their land without losing it to to in-laws or blood relatives. (Holden et al. 2008)
The 1995 Constitution of the Federal Democratic Republic of Ethiopia gave landowners additional rights, to both male and female headed households with respect to use, transfer and control over their land which included permitting the employment of farm labour and the letting of land. In 5
1998, Ethiopia underwent a further land reform, issuing landholder’s land right certificates. It was not until 2002 however that joint titling of both husband and spouse was introduced (Holden and Tefera, 2008; Deininger et al. 2007 cited in Girma et. al 2014).
Land Administration Committees (LACs) were formed at village level (known as Kebelle) to distribute land registration certificates. Land holders are given a ‘Book of Holding’ which contains a certificate recording landholders names’ (both head and spouse if married), textual data, estimated plot size, passport photos (in some cases) and names of the landholders of the neighbouring plots (Abza, 2011 cited in Girma et al, 2014). In the case of polygamous relationships, only the first wife was registered with the husband, additional wives and land were given certificates with only the wife’s name present.
The four main rural areas of Ethiopia; Tigray, Amhara, Oromia and SNNPR saw the introduction of land certification in 1998, 2002, 2003 and 2005 respectively. Tigray was the first to implement the certification in 1998 however certifications contained the name of the household head only, whereas Amhara, Oromia and SNNPR required the names of both head and spouse. (Girma, et al. 2014) In addition, the revision of the Family Code in 2000 further strengthened women’s join/common land rights with their partner. The code ensured, inter-alia, joint decisions upon use and disposal of land; the sharing of income generated from the land (to protect women whose work mainly involves domestic activities and child rearing); and the equal division of land upon the death of spouse or divorce. (Kumar et. al 2015).
By 2010, the certification program had registered a majority of rural land in the regions of Amhara (87%), Oromia (85%), SNNPR (84%) and Tigray (97%) (Abza, 2011 cited in Girma et al, 2014) recording Ethiopia’s land titling initiative as the largest, fastest and cheapest land registration and certification reform in Sub-Saharan Africa (Deininger et al., 2008).
(b) Female Empowerment in Ethiopia
Despite Ethiopia now granting men and women equal rights through the new Federal Constitution (Mabsout et. al 2009), remnants of past patriarchal customs and laws, which were often disguised through religious beliefs and cultural norms, are still evident through women’s low status and gender disparities present in almost all levels of society (Sida, 2013).
It is recorded that 27 million Ethiopians (27% of the population) live in poverty with the majority of people living in absolute poverty being women Educational attainment is restricted by gender stereotypes, early marriage violence, sanitation facilities and health issues including pregnancy, contributing to a female literacy rate of only 26% compared to 46% of men. (Sida, 2013). Female employment is largely in the informal sectors due to low education, reflected in low earnings and increased job insecurity, with the majority of female agricultural labour input going unrecognized. Women and girls are also more vulnerable to health risks, such as HIV, caused by child marriages, rape and prostitution and dangerous practices, such as female genital mutilation, are still far from being eradicated (Sida, 2003; Mabsout et al, 2009).
At the societal level, women still occupy the lower status, marginalizing them from decision making opportunities both within the household and political spheres (Kassa, 2015). Furthermore, the recent gender equality laws and reforms are met with implementation problems caused by cultural norms and attitudes and limited capacity (Sida, 2013). Evidence blames “full state sovereignty, allowing certain regions to uphold previous discriminatory laws” (World Bank, 1998 cited in Mabsout et al. 2009, p.8). Women, particularly rural women, are still therefore reported as one of the most disadvantaged groups of society, faced with limited access to information technology, assets, productive resources and decision making opportunities (Sida, 2013).
At the household level, literature reveals women are still subject to regular domestic violence and lack control over most assets. However, evidence has been found by Lim, Winter-Nelson and Arends-Keunning, (2007) that “assets which women would retain upon divorce do appear to increase bargaining power” (as cited in Mabsout, R. 2009 p.9). Furthermore, Kumar and Quisumbing (2015) find that changes in the family code were positively supported by the extensive land titling process; “land registration process awareness is positively correlated with the shift in perceptions toward equal division of land upon divorce, particularly for wives in male-headed households”.
(c) Gender and Land Reform
Existing literature on the impact of Ethiopia’s land reform is relatively limited to outcomes focussing on the land rental market (Holden, 2011), investment and productivity (Holden, 2009) and tenure security (Deininger, 2008; 2011). Where attempts have been made to estimate the impact of Ethiopia’s land titling program on female outcomes, often these studies have been limited to one region or one data set. Kumar (2015) and Bezabih (2010) make attempts to investigate the impact of
the land reform on promoting female rights and welfare, and in reducing the gender gap in agricultural productivity respectively. Kumar (2015) does identify a positive effect of land certification on increasing female rights and welfare. Little research however has been conducted to estimate the direct effect of these increased female rights on the empowerment of women measured through intra-household decision making.
My research question is therefore, did Ethiopia’s joint titling land reform positively impact female empowerment?
(ii) Conceptual Framework and Evidence
Drawn from existing literature, the principal mechanisms, relevant to this study, through which female land rights can increase female empowerment, are discussed below.
(a) Intra-household Bargaining Power and Female Empowerment
The concept of female empowerment is defined by Kabeer (1999) and closely echoes Sen’s capability approach (1980). According to Sen’s approach, disadvantaged groups lack the capability to choose their preferred way of life. Kabeer extends this to define female empowerment as a process of change, which sees the extension of strategic and nonstrategic life choices for females (the disadvantaged group) (all cited in Seebens, 2011 p.3). Strategic choices refers to those choices critical for a woman to live her desired life (reproduction, marriage, health etc.), whereas non-strategic refers to economic choices (household purchases etc.) (Musonera et al. 2016) Empowerment therefore relates to the process of enhancing an individual's (or group's) capacity to make effective decisions which can be transformed into desired outcomes. (Alsop, Bertelsen and Holland, 2006, p.10 cited in Samman, 2009). Within this framework there are two main facets to empowerment - agency and opportunity, which work together. Increased agency includes taking action and effective resistance to actions, decisions and manipulation and is often linked to individual characteristics and assets (land, education, good health etc.) (Safilios- Rothshild, 1982; Sen, 1990 cited in Seebens, 2011 p.3). Opportunity refers to the “broader institutional, social, and political context of formal and informal rules and norms within which actors pursue their interests” (Samman et al. 2009). Kabeer outlines three stages to empowerment. First, agency – the process in which choice is made and transformed into effect. Second, resources (or opportunity)– the channel
through which agency is exercised (changes in laws and societal acceptance). Third, response and achievements associated with the increased agency. Hence, empowerment is only achieved when females utilize their increased agency to promote a change. (Cited in Seebens, 2011)
The measurement of the empowerment process depends heavily on the intent of the research and indicators available. Kabeer’s widely held definition results in empowerment measures most frequently being based on household decision making and/or self-esteem indexes (El Halawany, 2009 cited in Musonera et al. 2016). Studies by Musonera et al. (2016) and Allendorf (2007) both use intra- household decision making questions based on self-strategic and nonstrategic empowerment dimensions as their indicator. Indicators of bargaining power were developed from questions asked regarding income autonomy, household purchases and visits to relatives; this measure is known as the ‘direct’ method (through asking direct questions) as opposed to ‘indirect’ method (indicators inferred through age, education or earnings).
It is also important to outline the “sources and settings of empowerment” outlined by Kishor (2000) which include “socio economic characteristics and current environments that facilitate empowerment” (as cited in Allendorf, 2007). For example, evidence by Duflo, (2012) and Lundberg & Pollak, (1993 ) confirm a positive relationship between increased female education and female female empowerment through the mechanism of providing women “knowledge, skills, and resources to make life choices that improve their welfare” (as cited in Samarakoon et al. 2015). The
theory regarding older women and increased decision making power is also supported by the theoretical models of Aizer, (2007); Datta, (2006); Friedberg and Webb, (2006); Panda and Agarwal, (2005); Girikipati, (2008) (as cited in Mabsout, 2009). It is important to identify the evidence and motivation behind these settings of empowerment so I can identify covariates and control for any omitted variables which could bias my estimate.
(b) Land rights and Female Empowerment
The conceptual framework for female land rights leading to empowerment stems from, amongst others, the intra-household gender dynamics and bargaining power mechanisms. This specific channel identifies how female bargaining power can be boosted by land rights through changes which may increase female status without directly altering the household budget.
Within recent economic literature, the household bargaining power of a female has dominated female empowerment measures. This specific mechanism of female empowerment originates back 9
to when behavioural economists first started capturing the complex nature of gender relations within households. Models ranged from Becker’s (1965, 1981) unitary model, which did not distinguish between decision making agents within a household (Argawal, 1997), to the more recent cooperation- conflict collective models, distinguishing between household members and their preferences - with the state of pareto optimality at its core (Ballon, P, 2012). Ultimately, household members would cooperate with each other as long as cooperation makes them better off than non-cooperation. These asymmetric Nash-bargaining power models within households describe marriage as a cooperative game (Mader et. al 2013). Both the husband and wife have individual utility functions and outside options - often referred to as a threat point. Threat points refer the individual's utility in case of non-cooperation. This model is presupposes husband and wives to have different, and potentially conflicting, preferences and assumes partners to resolve conflicting preferences “in ways prescribed by the Nash or some other explicit bargaining solution”. The result of intra-household decision making relies on the strength of the household members’ bargaining power - “determined by individual’s access to extra-household resources” (Mader et. al 2013). The threat point in the cooperative game of marriage is the outcome in non-agreement - usually divorce. Additionally, the threat point is dependent on environmental variables such as the national divorce laws and customs (ibid).
It is this cooperation- conflict model of marriage, which determines a household member’s bargaining power, and, in the case of female members, their degree of empowerment. A female’s bargaining power can therefore be disaggregated into the strength of the her threat point. Thus to improve the empowerment of females within households, one needs to improve the female’s decision making ability/bargaining power by strengthening her fall-back position in case of non-cooperation. This theory has been tested by multiple authors, confirming women obtain bargaining power from securing resources such as income and assets (Agarwal, 1994; Kabeer, 1999; Quisumbing, 2003 as cited in Mabsout, 2009 p.3) and “compared to all male property rights, joint property ownership of land and houses improves women’s decision making power, their self-confidence, and reduces domestic violence” (Panda and Agarwal, 2005; Datta, 2006 cited in Mabsout 2009 p.3).
Based on this framework, the Ethiopian land joint titling initiative (combined with revised family code) should improve a female’s fall-back position through ensuring equal division of land upon
divorce, strengthening household bargaining power and hence result in increased female empowerment – this is the hypothesis I am testing in this research.
III. Data and Methodology
a) Data and Indicators
My research is facilitated by the Demographic and Health Surveys (DHS) for Ethiopia and the Ethiopian Rural Household Survey (ERHS) datasets. Comparable samples within each dataset have been created dependent on the specific features of each the datasets. In order to measure the impact of joint titling initiative on female empowerment, measured through the female intra-household decision making indicators, my sample contains couples who are currently married and living together; were married before their regions respective certification program was implemented; and are living in the Amhara, Oromia or SNNPR rural regions in households with land for agriculture. It is necessary to filter out couples married after the reform as my female empowerment measures are based on the relative position between husbands and wives - in incomplete couples these measures are therefore not defined. My sample selection means my analysis is on a sample of couples who have stayed together regardless of the land titling initiative, my results will therefore only hold for particularly stable couples.
Land titling took place in the four main rural areas of Ethiopia: Tigray, Amhara, Oromia and SNNPR in 1998, 2002, 2003 and 2005 respectively, however Tigray’s land certification process varied considerably from the other regions. With only one name registered, Tigray’s land titling initiative will not facilitate adequate comparisons to those implemented in the three additional regions. I have therefore eliminated Tigray from my samples and have focussed my estimated impact on the three remaining rural regions. In the DHS dataset, the samples in both years are limited to households which owned land for agriculture. In the ERHS data, due to nature of the survey, all households had land for agriculture.
Variables of interest for both samples are: treatment (exposure to the land titling measured in years); empowerment (measured in bargaining power); age of woman (years); education of women and men (measured on scale of 0-3); and regions (dummy variable).
The exposure to the joint titling initiative, (my independent variable) determines the treatment of the regions. As each region implemented the reform in different years, survey year minus the
regional titling year generates exposure to the treatment (in years). This ranged from 0 to 9 depending on the region and survey’s data.This variable is likely to be exogenous as introduction of the region to the reform was exogenously determined, independent of household and community characteristics and differs extensively between regions.
The female empowerment indicator, my dependant variable, is measured through female bargaining power on intra-household decisions based on past empirical literature and data availability. Four questions were selected in both the DHS and ERHS surveys based on both strategic and non-strategic choices. The empowerment score is calculated as the sum of four answers. Four questions asked - two points if decision made alone; one if made jointly; zero if made by someone else with a total possible score of eight. For the DHS these questions asked were: Who has the final say on: 1) health care for yourself; 2) major household purchases; 3) visits to family or relatives; 4) how income is spent. For the ERHS: 1) Do you and your husband have separate finances? ; Who 1
buys²: 2) cereals and grains; 3) meat and fish; 4) clothing for women. The choice of questions to generate this empowerment indicator relied on the consistency of the questions asked over the time periods (as some decision making questions were answered in one year and other were left blank or not asked) and comparability of questions between the two datasets. The empowerment indicator score is standardized to have a mean of zero and standard deviation of one for ease of analysis. In addition, the DHS 2011 survey asked women if they own land alone; jointly; alone and jointly or; do not own land. Including this variable in my analysis will provide an estimate on how many women were treated. The EHRS data does not include a question relating to titling of the household’s land, however in 2009, the survey asked households if they were aware of the land registration process. A dummy variable was created for answering ‘yes’. Additionally, a dummy variable indicating whether a woman understands their rights and the benefit the joint titling initiative is included in the ERHS 2009 sample. Through asking if she understood ‘ In case of divorce, what would happen to the land?’ and in answering ‘ shared equally between husband and wife’ this indicates female land rights understanding. These questions will be useful in analysing and testing for heterogeneous effects and mechanisms of my results and model.
1Scored 2 if they have separate finances, 0 if not.
² We know the term ‘buy’ refers to the decision maker rather than income earner when compared to additional questions on who ‘pays’ for certain things
Female age and education variables will allow me to control for individual characteristics which may cause bias unless controlled for. As discussed previously, education can be considered as a setting or source for empowerment and therefore needs to be controlled for. Additionally, there is potential for a reverse causality relationship as women with innate empowerment talents may undergo education - however this is not an issue in my setup assuming that women did not make their education choices in anticipation of the land reform. The education variable for both data sets was measured on a scale of zero to three. Zero indicating no education, one indicating primary education/literacy course, two indicating secondary education and three indicating university/tertiary education. This was a pre-existing scale used in the DHS data collection which I repeated in ERHS dataset, converting grade of schooling completed and referring to Ethiopian school grade system. The age variable is simply age in years.
Table 1 (below) provides the sample means and standard deviations for the statistics for the three samples. Table 6, 7, 8 in the appendix shows the same table segregated by region to analyse regional changes in decision making.
Table 1 - Summary Statistics: DHS 2005/2011; ERHS 1997/2009; EHRS 2009 Sample (1) DHS Repeated Cross
Section
(2) ERHS Panel (3) ERHS Cross Section Year 2005 2011 1997 2009 2009 R1= Amhara share / Exposure 33.7% / 3 34.5% / 9 42.0% / 0 42.0% / 7 28.5% / 7 R2= Oromia / Exposure 33.8% / 2 29.7% / 8 41.2% / 0 41.2% / 6 36.7% / 6 R3= SNNPR / Exposure 32.5% / 0 35.7% / 6 16.8% / 0 16.8% / 4 34.9% / 4 Treated Estimate Share - 89.4% 94.9% 85.4% Final say on at least one decision alone
or jointly
97.2% 89.4% 55.2% 78.1% 85.4%
Final say on all four decisions alone or jointly
47.5% 75.6% 26.9% 32.9% 29.9%
Final say on at least one decision alone 22.3% 23.1% 39.2% 56.3% 70.2%
Empowerment score/8 2.70 (1.32) 3.14 (1.58) 1.92 (2.10) 2.82 (1.90) 2.95 (1.74) Empowerment Standardized 0 (1) 0 (1) 0 (1) 0 (1) 0 (1) Female Age 31.29 (7.89) 34.22 (6.91) 35.83 (12.50) 40.41 (11.53) 41.99 (10.62) Female Education 0.18 (0.40) 0.21 (0.42) 0.43 (0.54) 0.43 (0.54) 0.39 (0.53) Male Age 38.72 (9.25) 41.39 (8.44) 43.95 (14.21) 49.18 (15.49) 51.74 (14.17) Male Education 0.47 (0.62) 0.52 (0.57) 0.69 (0.58) 0.69 (0.58) 0.66 (0.57) Observations 1185 1729 607 607 722
Notes:Data from ERHS 1997 and 2009 and DHS 2005 and 2011 surveys computed by author. Descriptive statistics for whole
sample and stratified by year of interview. Regional share indicates a region’s share of whole sample and exposure indicates how many years that region has been exposed to the treatment. Treated estimate share is an estimate of how many of the treated actually received treatment based on DHS question: female- do you currently own land? None; jointly; alone and jointly; alone. Or ERHS question: Are you aware of land registration process? First three figures represent percentage of women who have final say on decisions either alone; jointly or both. Remaining figures represent mean averages and standard deviations are reported in parentheses. Total observations: 2914 (DHS) 1214 (ERHS panel) 829 (ERHS cross section) , based on females and males currently married to each other.
Empowerment score out of total 8; 4 questions asked: 2 points if decision made alone; 1 if made jointly; 0 if made by someone else. Empowerment score standardized: Mean=0; SD=1. Education: 0= none; 1= primary/literacy programme/other ; 2=secondary; 3=university.
b) DHS Sample
The DHS surveys, from the DHS Programme, containnationally representative data on health and household demographics surveyed in 2000, 2005 and 2011. This dataset is a repeated cross-section rather than longitudinal and the challenges this presents will be discussed later.
The sample is limited to the years 2005 and 2011 due the lack of female decision making indicators in the 2000 survey. A total of 13,721 and 16,702 households were interviewed in 2005 and 2011 respectively. The number of couples used in this research is reduced to 1,185 and 1,729 in 2005 and 2011 respectively where respondents were limited to those married before their region’s land certification took place (and still married), living with their spouses, in Amhara, Oromia or SNNPR and answered yes to the whether or not the household owned any land for agriculture. The large attrition rate is large and is caused by selecting on couples who are married to each other and living together. However, I can assume this to be an accurate merge as the DHS programme created this couples file/dataset for ease of analysis for users. The DHS programme made this couples dataset based on married men and women who both declared to be married and living together to each other. This dataset is the result of matching the male and female individual surveys based on who they declared as partners. Attrition is mostly due to the male or female not accurately declaring who their partner is and occasionally due to their partner not living in the same households.
Table 1 indicates a relatively balanced regional share across the sample. The empowerment score increased between 2005 and 2011 of 0.44 points (5.5%), this can be loosely interpreted as one extra joint decision being made by every two women. Through analysing the breakdown of empowerment scale (using the three ‘final say’ percentages), this increase is driven by an increase of women having joint or alone final say on all four decisions rather than women having increased final say alone. Amhara is the largest contributor to this change (as seen in appendix table 5). Women and men are also slightly older and better educated in 2011 than 2005, with women being less educated and younger than men their male counterparts, controlling for these individual differences will be important to obtain a more accurate estimate. As the female land ownership question was only asked in the 2011 survey, the treatment estimate is only included for 2011, but standing at 89% I consider this a high treatment participation.
c) ERHS Sample
The ERHS surveys, from the International Food Policy Research Institute, is a longitudinal household data set covering households in a 24 villages in rural Ethiopia surveyed in 1987, 1994, 1995, 1997, 2000, 2004 and 2009 and contain rurally representative data on household, consumption, agriculture demographics. This dataset is a longitudinal survey (panel) with the same households surveyed over time.
The first (panel) sample for this dataset is limited to the years 1997 and 2009 due the lack of female decision making indicators in the 2000 and 2004 surveys. Totals of 1,425 and 1,577 households were interviewed in 1997 and 2009 respectively. The number of couples used in this research is reduced to 607 and 607 in both years, where respondents were limited those present in both the 1997 and 2009 survey, those married before their region’s land certification took place (and still married) and living with their spouses, in Amhara, Oromia or SNNPR. All of the households questioned in this survey had land. The second sample created from this dataset is for 2009 alone. This sample has the same characteristics as both the DHS and previous sample, however without the attrition from merging the 1997 and 2009 households, we have a larger sample of 722 couples.
Table 1 (above) indicates the panel sample and cross section sample to be very comparable in terms of demographics in 2009, yet there are differences in female decision making participation between the two data samples. This highlights that variations in regional share (and hence exposure) could be increasing empowerment score between 1997 and 2009 of 1.44 (18%). When segregated by region (table 6 in appendix) we can see SNNPR experienced a decline in empowerment over the time period. The regional share in the panel sample indicates a more unequal distribution across the regions, with higher regional shares in Amhara and Oromia, which have both been exposed to the treatment for longer. However, the 2009 cross section has a more equal distribution across all regions.
The share of the sample claiming they were aware of the land registration process is greater in the panel data sample than the cross section, but still high overall.
ii) Empirical strategy
To estimate the impact of Ethiopia’s joint titling land reform on female empowerment, I utilize both repeated cross sectional data and panel data, utilizing two different econometric approaches.
Due to being unable to measure who exactly was impacted through the titling initiative, through using ‘Intention to Treat’ analysis every subject who is randomized according to the regions (exposure) is analysed regardless of deviation that may occur after randomization. This will provide a more realistic estimate of the joint titling reform in Ethiopia on female empowerment, automatically taking into account measurement error.
a) Multiple linear regression model
The first model (1) (below) relates to the DHS repeated cross sectional sample, 2005 and 2011. Using ordinary least squares (OLS), this multiple regression model estimates the effect on Y - my dependent variable, female empowerment - of changing one unit of variable of my independent variable, ‘EXPOSURE’ – while holding the other regressors
X
F EM and REGION (where includes female age and education and REGION includes Oromia and SNNPR regionalX
F EMdummies, respectively) constant. As the dataset between 2005 and 2011 does not follow the same households over time, controlling for individual fixed effects is not possible. Through the inclusion of multiple regressors, female age, female education and regional dummies, variations are held constant over time, limiting omitted variable bias from socioeconomics characteristics which provide a setting for empowerment (as discussed previously) (Allendorf, 2007) and isolating the effect on empowerment of exposure to titling program. Note, Amhara dummy was omitted to prevent dummy variable trap which causes perfect multicollinearity. The coefficient β₁ is the estimated treatment effect (slope), β₀ is the intercept and u is the error term.
(1)
Y
irt=
β
0+ β
1EXP OSURE
rt+
β X
2 iF EM+ β
3REGION
r+ e
i = individual r = regiont = time
Regression model (2) relates to the DHS cross section and ERHS cross sectional samples. It involves the same principle as (1) above, however omits the regional time dummies as this cross
section is from one point in time. The coefficient ‘EXPOSURE’ once again is the estimated treatment effect of the titling program on female empowerment.
(2)
Y
irt=
β
0+ β
1EXP OSURE
rt+
β X
2 iF EM+ e
b) First- difference fixed effects model
The panel data set allows me to exploit a first- difference fixed effects model, removing unobserved effects. In this model, y is once again my female empowerment variable, EXPOSURE is my independent exposure variable.
The first-difference estimator eliminates time invariant omitted variables utilizing repeated observations over time. (3) The change in exposure over the time period is the same as the exposure as the data comes from two points in time - 1997 (before the reform) and 2009 (after the reform). Equation (4) therefore presents first differences model I run, regressing the differences in female empowerment on differences in exposure, using OLS.
(3) EXPOSURE = EXPOSURE
Δ
(4) Y
Δ
irt=
β
0+ β
1EXP OSURE
irt+ e
irtIV. Main Results
Table 2- Regression Results: Exposure Effects on Female Empowerment Outcome.
Dependant variable: Female Empowerment (standardized z-value) Regressor (1) DHS 2005/2011 repeated cross section (2) DHS 2011 cross section (3) ERHS 2009 cross section (4) ERHS 1997/2009 Panel Exposure 0.047*** (0.006) 0.131*** (0.020) -0.060* (0.031) 0.248*** (0.056) Female Age 0.006** (0.002) 0.007* (0.004) 0.002 (0.004) Female Education 0.057 (0.044) 0.110* (0.061) 0.096 (0.068) R1=Amhara 0 R2= Oromia -0.039 (0.446) R3=SNNPR -0.206*** (0.464) Intercept -0.3601*** (0.048) -1.141*** (0.197) 0.219 (0.230) -1.508*** (0.343) R² 0.044 0.027 0.007 0.030 Observations 2914 1729 722 607
Notes: Data from DHS and ERHS surveys, statistics computed by author. Robust standard errors are in the parentheses. *,**,*** indicate significance at the 90%, 95% and 99% level, respectively. Model (1)(2)(3) Linear regression with multiple regressors: controlling for female age, female education (classed as none=0, primary/other=1, secondary=2, university=3) and regional dummies, using DHS and ERHS cross sectional data. Model (4) First differences fixed effects regression, using ERHS 1997 2009 panel.
Columns (1) to (4) indicate the results from the four different regression models. Analysing model (1), the DHS repeated cross sectional sample, when we use a multiple linear regression model controlling for female age, education and regional dummy variables, the coefficient on exposure indicates that if we increase exposure (treatment) by one year the female empowerment indicator increases by 0.047 standard deviations, significant at the 99% level. This is in line with my hypothesis that increasing treatment through the land titling initiative increases the female empowerment indicator. The R-squared figure is low, yet with statistically significant predictors, this is common when estimating human behaviour.
The female age coefficient is also positively significant at the 95% level. This estimates that increasing female age by one year increases female empowerment standard deviation by 0.006. The magnitude of impact that one extra year of age gives, versus one extra year of exposure is however smaller for female age, due to differences in the range of each scale. The theory regarding older women and increased decision making power is supported by the research highlighted earlier. In addition, the SNNPR regional dummy coefficient is both negative and significant at the 99% level. This result could be explained by past literature indicating variation in ethnic groups between regions, creating diversity and hence social and cultural norms, manipulating current laws across regions, including those related to gender (Bjerén, 1985, as cited in Mabsout, 2009 p.9). Research by Fafchamps & Quisumbing, (2002) Bevan & Pankhurst (1996) Gopal & Salim, (1999) present a potential justification for this result. Their work identifies the relationship between ethnicity and gender norms in Ethiopia. They identify a geographical pattern, which still tend to favour men in property rights over assets in the southern regions of Ethiopia (including SNNPR). This has proven to have a detrimental on women’s status, and bargaining power. Without data on ethnicity, I am only able to make this link based on existing literature (cited in Kumar et al., 2015). Future research in this area should try to include this interesting dimension.
Column (2)’s regression model employs a similar multiple regression model, however on the DHS 2011 cross sectional sample and without regional dummies due to observations being taken from one point in time. I find similar results to before, with a positive and significant (at the 99% level) relationship between exposure to the reform and female empowerment. One extra year of land reform exposure increases female empowerment indicator by 0.131 standard deviations. There is also a significant (at the 90% level) and positive relationship between female age and female education and female empowerment.
Column (3) runs the identical multiple linear regression model as before, however on the ERHS cross sectional sample Interestingly, we have negative correlation, indicating an increase in exposure by one year decreases the standard deviation of female empowerment indicator by 0.06. This paradoxical result is not uncommon in literature, as Hossain, (1998) finds women’ who have a job, experience less decision making power than housewives in Bangladesh. Reasons for this result could be related to having a larger sample, compared to the other ERHS models (3) and (4), suggesting a different demographic of women being analysed and unobserved variations causing
bias. However as this result is less statistically significant than our other models and this model is not my preferred specification, I will not dwell on this outcome.
Although the estimates discussed above are significant, the cross sectional nature of these models present several sources of variability which are not easy to isolate (Almond, R. et al 2012). Through controlling for female characteristics, such as age and education and regional variation, we have aimed to control for sources of unobserved variation, however there will always be more unobserved individual variations, as indicated by some of our inconsistent results (omitted variable bias). Furthermore, as empowerment is a considered a process of change (Allendorf, 2007) and because our observational data is taken from either one point in time, or when we do observe change over time (DHS repeated cross section) we do not necessarily observe the same households over time, we cannot fully conclude whether a change has actually happened. Furthermore, if we presume the change did happen, there is the issue of reverse causality. This impacts the internal validity argument of our cross sectional models. According to past literature, cause and effect temporal ordering suggests that “first women secure the land rights and this then facilitates an increase in agency”. However, our cross sectional surveys do not include retrospective questions and this temporal ordering is subsequently lost. Technically, this causal relationship could run in the opposite direction or in both directions (Allendorf, 2007). Despite these limitations, these results are useful in confirming and contextualizing the results both between the DHS and ERHS datasets and from the preferred panel data set model (4).
By exploiting a panel data set using first-differences regression estimator, my final model with individual fixed effects and column (4) presents my preferred model and hence, set of results. The coefficient on exposure indicates a one year increase in exposure to the treatment results in a 0.248 standard deviation increase, significant to the 99% level. This result is likely to be most credible due to automatically controlling for unit specific characteristics and eliminating reverse causality problem due to the time ordering component of panel data sets. The magnitude of the first difference estimator is 5.3 and 2.6 times greater than those predicted by the repeated and cross sectional data respectively. A possible explanation for this magnitude difference could be down to controlling for unobserved effects in the first differences estimator, eliminating downward bias caused by omitted variables.
V. Other Results (i) Mechanisms
In order to assess if my positive impact found in the previous results truly comes from the land reform, table (3) presents findings from an OLS regression, exploring whether exposure to the reform is positively related to (1) female land ownership (2) reform awareness and (3) divorce related awareness, in both the DHS 2011 and ERHS 2009 cross section dataset respectively. Through using these variables as outcomes we can also assess the plausibility of mechanisms outlined in the conceptual framework.
Table 3 – OLS Regressions/ Mechanisms
Dependant variable: (1) Female land ownership (2) Land reform awareness (3) Divorce-related awareness
(all dummy variables)
Regressor (1) DHS 2011 Land ownership (2) ERHS 2009 Land reform awareness (3) ERHS 2009 Divorce related awareness Exposure -0.028*** (0.06) 0.005 (0.006) 0.056*** (0.011) Intercept 1.104*** (0.045) 0.932*** (0.354) 0.512*** (0.066) R² 0.013 0.001 0.032 Observations 1729 722 722
Notes: Data from DHS 2011 and ERHS 2009 survey, statistics computed by author. Results from 2011 DHS cross section data set and the 2009 ERHS cross section data set, OLS regressions. . Robust standard errors are in the parentheses. *,**,*** indicate significance at the 90%, 95% and 99% level, respectively. Land ownership dummy variable based on whether the woman owns land alone, jointly or alone and jointly. Land rights awareness dummy outcome variable based on survey question: are you aware of land registration process? Divorce rights understand dummy outcome variable based on survey question: In case of divorce, what would happen to the land? In answering ‘shared equally between husband and wife’ implies knowledge of divorce rights.
Table (3) column (1) highlights a negative and significant correlation between exposure to the reform and female land ownership for the DHS 2011 cross section. This questions whether my positive findings between female empowerment and exposure presented in the main results is truly
because of the land reform and questions whether the mechanism of female empowerment being positively impacted through greater female land ownership, is plausible and robust in this analysis. When analysing the ERHS 2009 cross section column (2) results, there is an insignificant result. Exposure to the land reform should ideally have had a positive impact on being aware of the land registration process, however there is no evidence of this in this analysis.
Column (3) offers more reassuring findings. There is a significant (at the 99% level) and positive relationship between exposure to the land reform. Through increasing exposure to the land reform by one more year, understanding your land- divorce related rights increases by 0.056. This offers some reassurance that knowing your land- related divorce rights in the case of non-cooperation can is a plausible driver in the positive exposure- female empowerment relationship previously found. It is important to note however, that these findings are from one point in time only due to the nature of the survey - asking these specific questions/ outcomes in 2011 (DHS) and 2009 (ERHS) only. Before and after data would be beneficial to strengthen this analysis.
(i) Robustness Analysis
Table 4 presents findings from robustness analysis, determining how robust the first-differences model is to choices of parameters.
Table 4 - Robustness Analysis: ERHS 1997 2009 Panel
Dependant variable: Female Empowerment (standardized z-value) Regressor (1) Women<45 years (2) Women>45 (3) Educated women (4) Uneducated women (5) Educated Spouse (6) Uneducated Spouse Exposure 0.273*** (0.073) 0.215** (0.093) 0.021 (0.780) 0.428*** (0.080) 0.242*** (0.071) 0.263*** (0.099) Intercept -1.458*** (0.459) -1.477** (0.577) -0.039 (0.494) -2.660*** (0.491) -1.487*** (0.443) -1.568*** (0.600) R² 0.038 0.029 0.0003 0.075 0.030 0.030 Observations 346 178 248 359 382 225 Notes: Data from ERHS survey, statistics computed by author. Results from 1997 2009 EHRS panel data set, fixed effects regression model. Robust standard errors are in the parentheses. *,**,*** indicate significance at the 90%, 95% and 99% level, respectively. 45 was median female age. Educated classed as received any education; uneducated classed as no education received.
Models (1)(2)(3)(4)(5)(6) present first differences fixed effects regression results, collapsed by age of woman, education status of woman and education status of spouse.
Estimates from columns (1) (2) (5) and (6) all indicate similar findings in terms of magnitude, sign and significance, to that estimated in the EHRS panel first differences estimate of the main results. Specifically, the first check examines whether the results change depending on age of the woman (1) (2). These estimates are similar to that of the main first-difference estimator (0.248). The estimate is larger in magnitude for women under 45.
I then change the parameters to account for changes in male education (5) and (6). An educated spouse could influence the results in various ways. First, an educated spouse may increase a female’s understanding or awareness of the land titling program and hence bargaining power. Second, and more commonly researched, an educated spouse is increases the likelihood of wives having an equal participation in decision making (Oropesa, 1997). These estimates however, also remain consistent with our main results suggesting robustness to changes in the parameters of the model with regard to both female age and male education.
Columns (3) and (4) test for robustness to changes in female education. Column (3) indicates an insignificant result, yet column (4) finds a significant estimate of larger magnitude than that seen in the main results. This implies that the model is only partially robust to the parameters of female education. Much literature declares there is a positive relationship between female education on household decision making (Sujatha et al, 2009) hence motivating the inclusion of this robustness check. As my findings suggest the estimate for uneducated women is both positive and significant, like in the main result, but 1.7 times as large. The insignificance of the educated women result implies the model is sensitive to the parameters of education of females. In particular, the inclusion of more uneducated women in my model will have a upward bias on my estimate. With five out of six robustness tests generating both significant and similar trends to those presented in my main results, I am convinced the method remains relatively effective under considerable change in parameters and different scenarios of the model. Extending this research to thoroughly investigate education impact on the model however, would be beneficial.
VI. Concluding Remarks
In the government’s attempt to eradicate gender discriminatory laws and practices in Ethiopia, the beginning of the 21 st century saw the implementation of a nation-wide joint land titling initiative.
The program successfully provided a certificate to rural households with land which included both the names of the husband and wife (the joint land holders). Amhara, Oromia and SNNPR saw this titling initiative be exogenously implemented in 2002, 2003 and 2005 respectively, independent of both household and regional characteristics. In conjunction with the newly revised family code, female land rights were strengthened by this titling initiative, providing women equal rights over joint land in case of dissolution of marriage or death of partner. Previous literature finds women who have greater access to assets and resources are more likely to have greater bargaining power and hence greater participation in household decision making. This is achieved through a women’s increased ‘threat point’ or strengthened fall-back position in case of non-cooperation (divorce). In Ethiopia’s case, land rights are therefore thought to strengthen married women’s fall-back position through inclusion of her name on the land certificate and legally being entitled to equal share of the land in case of divorce or death of partner.
Through running four different econometric models on two different datasets, I have found evidence of a positive relationship between exposure to the land titling initiative and female decision making in three out of the four models. My most credible result is derived from the ERHS panel data set, exploiting a first-differences estimate, controlling for all unobserved effects. The result estimates increasing exposure by one year causes the female empowerment indicator to increase by 0.248 standard deviations. This trend is confirmed by similar positive trends in my cross sectional sample models and strengthened further through a similar positive relationship between land reform exposure and divorce- rights related awareness- indicating the plausibility of mechanism that the land reform increases female empowerment through strengthening a women’s fall back position.
It is important to outline that these estimates can only be generalized to women who are in stable
marriages. Due to my sample selection being based on men and women who stayed married regardless of the reform, my samples do not take into account those who got divorced, partners who died or other reasons to not stay married.
Inconsistencies and surprising results in my estimates do point out an area for further research. Regional dummy coefficients in the repeated cross section estimate identify SNNPR’s exposure to
the reform linked to decreased female empowerment. This paradoxical result calls for further research on Ethiopia’s joint land titling and variations in implementation and uptake across ethnically diverse regions.
Bibliography Data
Ethiopia: Standard DHS, (2005) The DHS Program. Available at:
http://dhsprogram.com/data/dataset/Ethiopia_Standard-DHS_2005.cfm?flag=0 Ethiopia: Standard DHS, (2011) The DHS Program. Available at:
http://dhsprogram.com/data/dataset/Ethiopia_Standard-DHS_2011.cfm?flag=0
Ethiopian Rural Household Survey (1989 -2009) International Food Policy Research Institute. Hoddinott, J Yohannes, Y. Available at:
https://dataverse.harvard.edu/dataset.xhtml?persistentId=hdl:1902.1/15646 Literature
Ababa, (2008). ‘Gender Inequality and Women’s Empowerment’. Ethiopian Society of Population studies. Available at http://countryoffice.unfpa.org/ethiopia/drive/Gender.pdf Last accessed
6/5/2015 2.
Agarwal B. A Field of One's Own: Gender and Land Rights in South Asia. Cambridge: Cambridge University Press; 1994.
Agarwal, B. (1997). ''Bargaining''and Gender Relations: Within and Beyond the Household. Feminist economics, 3(1), 151.
Allendorf, K. (2007). Do women’s land rights promote empowerment and child health in Nepal?. World Development, 35(11), 19751988.
Almond, R, Simnharay, (2012) What can repeated cross-sectional studies tell us about student growth? Research Report ETS RR-12-17.
Ambza, 2011. Experience and Future Direction in Ethiopian Rural Land Administration. World Bank Conference. Washington, USA. Available at:
http://siteresources.worldbank.org/INTIE/Resources/475495-1302790806106/AbzaPres1.pdf Ballon, P. 2012. Structural Model of Female Empowerment. OPHI, University of Oxford, UK. Bezabih, M., & Holden, S. (2010). The Role of Land Certification in Reducing Gender Gaps in Productivity in Rural Ethiopia (No. dp1023efd).
Deininger, K., Ali, D. A., Holden, S., & Zevenbergen, J. (2008). Rural land certification in
Ethiopia: Process, initial impact, and implications for other African countries. World Development, 36(10), 17861812.
Doss et al. 2011. FAO ‘The Role of Women in Agriculture’ available at:
http://www.fao.org/docrep/013/am307e/am307e00.pdf Last accessed: 6/5/2015
Doss et al. 2014 IFPRI. Women's Individual and Joint Property Ownership Effects on Household Decision-making. Available at http://www.ifpri.org/sites/default/files/publications/ifpridp01347.pdf Last accessed 6/5/2015
FAO, 2011. The State of Food and Agriculture. Rome, 2011. Available at: http://www.fao.org/docrep/013/i2050e/i2050e.pdf Last accessed: 6/5/2015 12 FAO, 2015. Country Profile Ethiopia Avaiable at:
http://www.fao.org/ag/AGP/AGPC/doc/counprof/ethiopia/ethiopia.htm Last accessed at 6/52015 . Girma. Giovarelli, 2014. Brief: The Gender Implications of Joint Titling in Ethiopia. Available at: http://www.focusonland.com/fola/en/countries/briefthegenderimplicationsofjointtitlinginethiopia/ Last accessed 6/5/2017
Holden, Stein and Tewodros Tefera. (2008)“From Being Property of Men to Becoming Equal Owners?: Early Impacts of Land Registration and Certification on Women in Southern Ethiopia.” UNHABITAT
Holden, S. and Yohannes, H. 2002. Land Redistribution, Tenure Insecurity, and Intensity of Production: A Study of Farm Households in Southern Ethiopia, Land Economics, 78(4), 573590 Holden, Tefera . 2008. Land Certification in Ethiopia Early Impacts on Women Copyright. United Nations Human Settlements Programme (UNHABITAT), 2008
Holden, S. T., Deininger, K. and Ghebru, H. (2011). Tenure Insecurity, Gender, Lowcost Land Certification and Land Rental Market Participation. Journal of Development Studies 47(1): 3147. Kumar, N., & Quisumbing, A. R. (2015). Policy reform toward gender equality in Ethiopia: Little by little the egg begins to walk. World Development, 67, 406423.
Mabsout, van Staveren (2009) Disentangling Bargaining Power from Individual and Household Level to Institutions: Evidence on Women’s Position in Ethiopia. Paper presented at ISS
Development Research Seminar, The Hague, NL
Mader, K. Schneebaum, Al. 2013. The gendered nature of intra-household decision making in and across Europe. Vienna University of Economics and Business, Austria.
Malhotra, A. Schuler, S, Boender, C. 2002. Measuring Women’s Empowerment as a Variable in International Development. JSI Research and Training Institute Background Paper Prepared for the World Bank Workshop on Poverty and Gender: New Perspectives.
Momsen, J.H. 1991. Women and development in the Third World, Routledge, 1991
Musonera, A and Heshmati, A (2016) Measuring Women’s Empowerment in Rwanda. East Africa Research Papers in Economics and Finance. Sweden.
Oropesa, 1997. Development and marital power in Mexico. Social Forces, 75(4), 1291-1318.
Rahmato, Dessalegn, 1984. Agrarian Reform in Ethiopia, Uppsala: Scandinavian Institute of African Studies.
Samman, E. Santos M. 2009. Agency and Empowerment: A review of concepts, indicators and empirical evidence Prepared for the 2009 Human Development Report in Latin America and the Caribbean. University of Oxford
Seebens, H. 2011. Intra-household bargaining, gender roles in agriculture and how to promote welfare enhancing changes. ESA Working Paper No. 11-10. FAO.
Sida, 2013. A profile on Gender Relations: Towards Gender Equality in Ethiopia. Sweden. Retrieved from
http://www.sida.se/contentassets/54d269b9567243f0861deacae205859a/towards-gender-equality-in -ethiopia_470.pdf
Samarakoon, S. 2015. Does Education Empower Women? Evidence from Indonesia, World
Development Volume 66, Pages 428-442
Tefera, T (2013) Land ownership- the path towards rural women empowerment: A case from Southern Ethiopia, International Journal for Sociology and Anthropology. Vol. 5(8), pp. 330-338 USAID, 2015. Empowering women through agricultural development in Ethiopia. Reports and Data.
https://www.usaid.gov/results-data/success-stories/empowering-women-through-agricultural-develo pment-ethiopia [last accessed 15/07/2017]
Wiig, H. 2013 Joint Titling in Rural Peru: Impact on Women’s Participation in Household Decision-Making, World Development, Volume 52, 2013, Pages 104-119,
(http://www.sciencedirect.com/science/article/pii/S0305750X13001587)
Appendix
Table 5 - Summary Statistics: DHS Sample, 2005 2011 (repeated cross sectional data)
Whole Amhara Oromiya SNNP Year 2005 2011 2005 2011 2005 2011 2005 2011
Exposure - - 3 9 2 8 0 6
Final say on at least one decision alone or jointly
97.2% 89.4% 99.2% 93.6% 96.8% 88.9% 95.6% 85.8%
Final say on all four decisions alone or jointly
47.5% 75.6% 46.1% 80.7% 44.6% 75.9% 51.9% 70.4%
Final say on at least one decision alone 22.3% 23.1% 19.0% 24.1% 20.9% 22.4% 27.0% 22.7% Empowerment (standardized) -0.17 (0.88) 0.12 (1.06) -0.05 (0.72) 0.28 (0.92) -0.13 (0.99) 0.19 (1.10) -0.34 (0.89) -0.10 (1.11) Female Age 31.29 (7.89) 34.22 (6.91) 31.5 (8.3) 34.9 (7.17) 31.52 (7.90) 33.97 (6.64) 30.83 (7.44) 33.76 (6.82) Female Education 0.18 (0.40) 0.21 (0.42) 0.13 (0.34) 0.11 (0.32) 0.16 (0.38) 0.23 (0.43) 0.24 (0.47) 0.28 (0.47) Male Age 38.72 (9.25) 41.39 (8.44) 40.0 (9.64) 42.89 (8.62) 38.64 (9.01) 40.86 (8.00) 37.46 (8.91) 40.39 (8.47) Male Education 0.47 (0.62) (0.57) 0.52 (0.52) 0.26 (0.47) 0.28 (0.65) 0.54 (0.58) 0.58 (0.64) 0.61 (0.57) 0.69 Observations 1185 1729 399 597 401 514 385 618 Notes:Data from DHS 2005 and 2011 surveys and computed by author. Descriptive statistics for whole sample and collapsed across the
different treatment groups and years. First three figures represent percentage of women who have final say on decisions either alone; jointly or both. Remaining figures represent mean averages and standard deviations are reported in parentheses. Total observations: 2914, based on females and males currently married to each other.
Empowerment score out of total 8; 4 questions asked: 2 points if decision made alone; 1 if made jointly; 0 if made by someone else. Empowerment score standardized: Mean=0; SD=1. Education: 0= none; 1= primary; 2=secondary; 3=university.
Table 6 - Summary Statistics: ERHS Sample, 1997 2009 (panel data)
Whole sample Amhara Oromiya SNNP Year 1997 2009 1997 2009 1997 2009 1997 2009 Exposure - - 0 7 0 6 0 4 Final say on at least one
decision alone or jointly
55.2% 78.1% 62.0% 78.0% 36.0% 77.2% 90.2% 80.4%
Final say on all four decisions alone or jointly
26.9% 32.9% 38.8% 33.3% 14% 35.2% 35.3% 26.5%
Final say on at least one
decision alone 39.2% 56.3% 33.7% 51.0% 28.8% 54.4% 78.4% 75.5% Female Age 35.83 (12.50) 40.41 (11.53) 34.23 (12.73) 41.95 (12.15) 35.61 (12.28) 38.00 (11.15) 40.39 (11.47) 42.41 (9.82) Female Education 0.43 (0.54) 0.43 (0.54) 0.490 (0.57) 0.490 (0.57) 0.352 (0.50) 0.352 (0.50) 0.46 (0.54) 0.46 (0.54) Male Age 43.95 (14.21) 49.18 (15.49) 42.27 (12.28) 50.78 (14.40) 42.85 (14.14) 46.69 (16.99) 48.48 (13.38) 51.10 (13.62) Male Education 0.69 (0.58) 0.69 (0.58) 0.72 (0.50) (0.72) (0.50) 0.61 (0.55) 0.61 (0.55) 0.73 (0.75) 0.73 (0.75) Observations 607 607 255 255 250 250 102 102
Notes:Data from ERHS 1997 and 2009 surveys and computed by author. Descriptive statistics for whole sample and collapsed across the
different treatment groups and years. First three figures represent percentage of women who have final say on decisions either alone; jointly or both. Remaining figures represent mean averages and standard deviations are reported in parentheses. Total observations: 1214, based on females and males currently married to each other.
Empowerment score out of total 8; 4 questions asked: 2 points if decision made alone; 1 if made jointly; 0 if made by someone else. Empowerment score standardized: Mean=0; SD=1. Education: 0= none; 1= primary/literacy programme/other ; 2=secondary; 3=university.
Table 7 - Summary Statistics: ERHS Sample, 2009
Whole sample Amhara Oromiya SNNP
Year 2009 2009 2009 2009
Exposure - 7 6 4
Final say on at least one decision alone or jointly 85.4% 81.8% 89.1% 84.4% Final say on all four decisions alone or jointly 29.9% 34.7% 32.9% 22.8% Final say on at least one decision alone 70.2% 55.7% 72.7% 80.0% Empowerment score/8 2.95 (1.74) 2.59 (1.65) 3.19 (1.72) 2.99 (1.80) Female Age 41.99 (10.62) 44.03 (11.00) 41.05 (10.63) 41.44 (10.12) Female Education 0.39 (0.53) 0.44 (0.54) 0.40 (0.55) 0.34 (0.49) Male Age 51.74 (14.17) 52.94 (13.32) 51.69 (15.01) 50.80 (13.94) Male Education 0.66 (0.57) 0.73 (0.46) 0.66 (0.59) 0.62 (0.62) Observations 722 236 304 289
Note:Data from ERHS 2009 survey and computed by author. Descriptive statistics for whole sample and collapsed across the different treatment groups and years. First three figures represent percentage of women who have final say on decisions either alone; jointly or both. Remaining figures represent mean averages and standard deviations are reported in parentheses. Total observations: 829; based on females and males currently married to each other- larger sample due to lack of attrition across years. Education: 0= none ; 1= primary/literacy programme/other ; 2=secondary; 3=university.
Empowerment score out of total 8; 4 questions asked: 2 points if decision made alone; 1 if made jointly; 0 if made by someone else. Empowerment score standardized: Mean=0; SD=1. Education: 0= none ; 1= primary/literacy programme/other ; 2=secondary; 3=university.
