0 UNIVERSITY OF GHANA THE DETERMINANTS OF GENDER ASSET GAP IN GHANA BY CHARITY ACKUAKU (10242156) A THESIS SUBMITTED TO THE UNIVERSITY OF GHANA, LEGON IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE AWARD OF MASTER OF PHILOSOPHY (MPHIL) DEGREE IN ECONOMICS JULY, 2014. University of Ghana http://ugspace.ug.edu.gh i DECLARATION This is to certify that this thesis is the result of research undertaken by CHARITY ACKUAKU towards the award of the Masters of Philosophy (MPhil) degree in the Department of economics, University of Ghana. …………………………………… CHARITY ACKUAKU (STUDENT - 10242156) DATE: ………………… …………………………………… MS. ABENA D. ODURO (SUPERVISOR) DATE: ………………… …………………………………… DR. LOUIS BOAKYE-YIADOM (SUPERVISOR) DATE: ………………… University of Ghana http://ugspace.ug.edu.gh ii ABSTRACT Differences in economic outcomes and the analysis of their determinants are of utmost importance in various domains. Especially with respect to gender, there is massive (political) interest in whether these differences reflect discrimination or whether they simply arise from differences in relevant observable characteristics between men and women. This study focuses on estimating wealth differences between men and women in Ghana to comprehend and appreciate these differences and understand the determinants of gender wealth gap. The study utilized the standard Blinder-Oaxaca decomposition technique in examining and explaining the wealth differentials (in terms of the gross value of total physical assets owned) between men and women. The data for the analysis was obtained from the 2010 Ghana Household Asset Survey (GHAS). The results indicate that wealth levels of men were found to be significantly higher than those of women. The Blinder-Oaxaca decomposition result showed a mean difference of about 0.13 in favour of men. Approximately 21% of differences in the gender wealth were explained by the predictors of the model with the remaining 79% (approximately) representing the unexplained difference in gender wealth. Four factors (age, education, economic status, and location) were found to account for the explained difference. Out of the explained difference, these characteristic of asset owners were significant in explaining the gap at approximately 82%, 196%, 120%, and 40% respectively. The study recommends some constructive and feasible policies like policies that will empower women to own assets in their names, encourage women to reach higher levels in education and also push for more women in the public wage employment, in order to harness gender equality by ensuring a bridge in the gender wealth gap were recommended. University of Ghana http://ugspace.ug.edu.gh iii DEDICATION I hereby dedicate this thesis to God Almighty, the creator of man and by whose Grace, Strength and Wisdom made it possible for me to go through my course of study. To Him be the glory for how far He has brought me. My special gratitude and thanks to my loving and dedicated husband, David Papa Yaw Opoku, and my adorable daughter, Peyton Kafui Akua Opoku for their prayers and love. University of Ghana http://ugspace.ug.edu.gh iv AKNOWLEDGEMENT If one is not thankful to what he/she has got, he isn’t likely to be thankful for what he/she is going to get. To be grateful is to recognize the love of God in all things. I am very thankful to the Almighty God for my life and that of my family. My heartfelt thanks go to all the lecturers and stuff of the Department of Economics for their contribution in my education towards this end. I acknowledge with much appreciation the support and contribution of my supervisors; Ms. Abena D. Oduro and Dr. Louis Boakye- Yiadom, all of Department of Economics, University of Ghana, for their great devotion, attention, constructive suggestions, valuable guidance, and words of encouragement and advice. My indebtedness is unmeasurable in terms of the support and prayers of the Opoku and Ackuaku families. You forgot me not in your prayers, and I am grateful to also be a blessing to the family. Finally, to my friends Forster Shitsi Junior, Miss Millicent Awuku, and Miss Fatima Nkansah for your immense contribution in diverse ways towards the success of this work, I say thank you. To all loved ones not mentioned here I say thank you and that your contributions are well appreciated, and may the good Lord Almighty bless you and replenish your efforts in trillion folds. Amen. University of Ghana http://ugspace.ug.edu.gh v TABLE OF CONTENTS Page ABSTRACT ............................................................................................................................................ ii DEDICATION ....................................................................................................................................... iii AKNOWLEDGEMENT ........................................................................................................................ iv LIST OF TABLES ................................................................................................................................ vii LIST OF ABBREVIATIONS .............................................................................................................. viii CHAPTER ONE ..................................................................................................................................... 1 INTRODUCTION .................................................................................................................................. 1 1.1 Background of Study ........................................................................................................................ 1 1.1.1 The Ghanaian Context ............................................................................................................... 2 1.2 Problem Statement ............................................................................................................................ 5 1.3 Objectives of Study ........................................................................................................................... 7 1.4 Research Questions ........................................................................................................................... 7 1.5 Significance of Study ........................................................................................................................ 7 1.6 Organization of Study ....................................................................................................................... 9 CHAPTER TWO .................................................................................................................................. 10 LITERATURE REVIEW ..................................................................................................................... 10 2.0 Introduction ..................................................................................................................................... 10 2.1 Theoretical Literature ...................................................................................................................... 10 2.1.1 Sex and Gender ........................................................................................................................ 10 2.1.2 Sociological Concepts of Gender ............................................................................................. 11 2.1.3 Gender Differences, Asset and the Gender Asset Gap ............................................................ 12 2.1.4 Feminist Economics ................................................................................................................. 14 2.1.5 Determinants of gender differences in Asset ........................................................................... 15 2.2 Empirical Literature ........................................................................................................................ 17 2.3 Summary ......................................................................................................................................... 27 CHAPTER THREE .............................................................................................................................. 29 METHODOLOGY ............................................................................................................................... 29 3.0 Introduction ..................................................................................................................................... 29 3.1 Measuring the Gender Asset Gap ................................................................................................... 29 3.2. Model Formulation ........................................................................................................................ 30 3.2.1 Theoretical Framework ............................................................................................................ 31 3.2.2 Specification of Empirical Model ............................................................................................ 33 3.2.3 Description of variables of the model ...................................................................................... 35 3.2.4 Specification of Blinder-Oaxaca Decomposition Model ......................................................... 40 University of Ghana http://ugspace.ug.edu.gh vi 3.2.5 Blinder-Oaxaca Decomposition ............................................................................................... 40 3.2.5.1 Estimation Process ............................................................................................................ 43 3.2.5.2 Problems with Blinder-Oaxaca Decomposition ................................................................ 44 3.3 Source of Data ................................................................................................................................. 46 3.4 Summary ......................................................................................................................................... 47 CHAPTER FOUR ................................................................................................................................. 48 EMPIRICAL RESULTS AND DISCUSSION ..................................................................................... 48 4.0 Introduction ..................................................................................................................................... 48 4.1 Socioeconomic Characteristics of Asset Owners ............................................................................ 48 4.2 Empirical Analysis .......................................................................................................................... 54 4.2.1 Estimated Regression Models .................................................................................................. 54 4.2.2 Blinder-Oaxaca Decomposition Technique (Results) .............................................................. 61 4.3 Summary ......................................................................................................................................... 66 CHAPTER FIVE .................................................................................................................................. 67 SUMMARY AND RECOMMENDATIONS ....................................................................................... 67 5.0 Introduction ..................................................................................................................................... 67 5.1 Summary ......................................................................................................................................... 67 5.2 Recommendations ........................................................................................................................... 69 5.3 Suggested Area for Future Study .................................................................................................... 73 REFERENCES ..................................................................................................................................... 75 APPENDIX ........................................................................................................................................... 83 Appendix A: Tables of Summary Statistics from STATA ................................................................... 83 Appendix B: Regression Results using STATA ................................................................................... 86 Appendix C: Blinder-Oaxaca Decomposition Result for Gender Differences in Wealth ..................... 90 University of Ghana http://ugspace.ug.edu.gh vii LIST OF TABLES Page Table 3.1: Description of variables and their expected signs ………………………….…… 39 Table 4.1: Descriptive Statistics of Predictor Variables of the Model by Gender …….…… 51 Table 4.2: Results on Simple Linear Regression Model with ln Was Dependent Variable ... 54 Table 4.3: Results of Multiple Linear Regression for Male, Female, and Pooled Samples with lnW as Dependent Variable …………..…………………………………………………….. 58 Table 4.4: Blinder-Oaxaca Decomposition Result …………………………………….…… 64 University of Ghana http://ugspace.ug.edu.gh viii LIST OF ABBREVIATIONS SSA Sub-Saharan Africa UN United Nations GHAS Ghana Household Asset Survey UK United Kingdom FRS Family Resource Survey US United States NLSY National Longitudinal Survey of Youth BLS Bureau of Labour Statistics CPI Consumer Price Index HILDA Household, Income and Labour Dynamics in Australia GEM Global Entrepreneurship Monitor CAPMAS Central Agency for Public Mobilization RHS Right Hand Side lnW Log of Wealth ERF Economic Research Forum SOEP German Socio-Economic Panel ELMPS Egypt Labour Market Panel Survey HECS Higher Education Contribution Scheme WEF World Economic Forum PPP Purchasing Power Parity USD United States Dollars University of Ghana http://ugspace.ug.edu.gh 1 CHAPTER ONE INTRODUCTION 1.1 Background to Study Worldwide, gender relations play a significant role in the division of labour, distribution of income, wealth and decision making. There exists a recognized degree of disparity in each area. Gender inequality can significantly be found in the ownership of assets among males and females (Deere and Doss, 2006; Oduro et al., 2011; Swaminathan et al., 2012). For instance, approximately 70-90 percent of formal ownership of farmland is attributed to males in a large number of countries located in Sub-Saharan Africa (SSA) and Latin America (Quisumbing and Maluccio, 1999). In UK, it has been found that by large, there is significant gender gap in total wealth (Tracey Warren, 2006). In Kenya, Doss (2012) concluded that, not only did a gender asset gap exist in agricultural resources (especially in deeded land) but was distinguished by marital status. Çağatay (2001) explains that gender inequality is not only a development problem but also an important facet of poverty. Emphatically, gender equality plays a central role in the MDGs and therefore failure to deal with the gender inequality may cost weighty implications. As a result of the relevance and elusiveness of gender equality issues, UN adopted gender equality as part of the UN Millennium Development Goal (the third Millennium Development Goal). In various countries, research on gender gap in wages has been conducted, using it as a premise in understanding this inequality issue. Wages (income) is just but one form of outcomes that can explain the differences that reflects the disparity between men and women. This is to say, there exists other relevant outcomes such as assets. There is a massive growing University of Ghana http://ugspace.ug.edu.gh 2 awareness that ownership of and control over assets such as land and residence are crucial indicators of individuals’ well-being (Oduro et al., 2011). Assets can provide income, help access credit and provide economic security to individuals. Assets are reckoned to be more stable and less liquid than income and also serve as collateral for credit purposes. Oduro et al. (2011) citing Katz and Chamorro (2003) reported that women’s ability to gain access to and ownership of assets greatly influences household economic decisions. Women’s responsibility in the household is mainly centered on ensuring the well-being of family members and children. Therefore, more assets by women will significantly improve the well-being of families. It is profoundly acknowledged that for any form of women’s development (economic and social), prevalent issues on gender inequality and poverty must be tackled. World Bank explains that countries that invest in promoting the social and economic status of women, tend to have lower poverty rates. Also, asset poverty is increasingly becoming a vital and more reliable measure of poverty over time. Subsequently, drawing analysis from the asset perspective enables us extract some findings that would and recommend policies enhance help gender equality and marginalize poverty levels. 1.1.1 The Ghanaian Context In 2012, the Ghanaian economy recorded a growth rate of 7.9 percent. Growth rates have averaged over 5% in the last decade (Oduro et al., 2011) and this evidently indicates how rapid the economy is advancing. University of Ghana http://ugspace.ug.edu.gh 3 The incidence of poverty declined from 39.5 percent in 1998/99 to 28.5 percent in 2005/06. Poverty remains a serious issue in the northern sector of the country. In 2005/06, the Upper East and Upper West regions registered 70.4 percent and 87.9 percent incidence of poverty respectively. Although the nation’s incidence of poverty has declined over years, the Gini index unfortunately has not driven in the same direction. Ghana’s Gini index rose from 0.37 in 1992 to 0.40 in 1998 to 0.42 in 2006 (Coulombe and Wodon, 2007). There is a significant level of inequality in access to and ownership of assets between men and women in Ghana. For instance, recent findings available reveal that in Ghana, the principal residence as a form of ownership was majorly owned individually by men (that is 61%) with individual ownership by women accounting for 39% (Oduro et al., 2011). It has been found that women’s ability to gain access to and control of assets empowers them, highly influences decision-making in the household and may influence a decline in the exposure to domestic violence (Panda and Agarwal 2005; ICRW 2006). It suffices to say, the relevance of ensuring and improving women’s ownership of assets cannot be disputed. The various forms of asset holding can be categorized into the following: natural resource capital, human capital, physical, financial, social and political capital. The natural resource capital consist of land, water bodies, fertilizer, trees, genetic resources of the land, etc. It is the stock of natural ecosystem that yield a flow of valuable ecosystem goods and services into the future. It consists of valuable goods and services from the natural environment and all formations of the earth’s biosphere that provides us with ecosystem goods and services imperative for survival and over all well-being. In other words, it is the basis for all human economic activity. University of Ghana http://ugspace.ug.edu.gh 4 Human capital is defined as the distribution and skills of an individual, especially those acquired through investment in education and training, on the job training and experience that enhances potential income earning1. It is the collective skills, knowledge or other intangible asset of individuals that can be used to create economic value for the individuals, their employers, or their community. Education for instance is an investment in human capital that pays off in terms of higher productivity. Examples of human capital include skills, knowledge, health, nutrition, among others. Social capital refers to the institutions, relationships and norms that shape the quantity and quality of the society’s social interactions. Social coherence is linked to economic prosperity of a society and sustainable development. Social capital is not just the sum of the institutions that underpin a society; it is the glue that holds them together. Political capital on the other hand is essentially the opinion of another person, nation, or group of people about you or your government or organization2. It is primarily based on a public figure’s favourable image among the popularity and important actors in and or out of the government Physical capital which this study focuses on is the manufactured assets that are applied to production. It refers to any non-human effort or asset made by man to aid in the production process3. It is the type of asset that is used in further production or manufacturing process that allows a business to create goods and services for direct consumption or sale to consumers. Along with the plant that houses the actual production process, facilities such as warehouse that are owned by the company are also considered physical capital. Examples include plant and machinery, vehicles, buildings, etc. 1http://dictionary.reference.com/browse/human+capital 2 www.ask.com/question/what-is-political-capital 3http://www.investordictionary.com/definition/physical-capital University of Ghana http://ugspace.ug.edu.gh 5 The variable of focus in this study, wealth, was computed by Oduro et al (2011) as the gross value of all physical assets in the 2010 GHAS data. The physical assets included agricultural land, business, place of residence, livestock, real estates, and consumer durables. The gross values of these physical assets are summed up for each individual observation to compute the wealth of the individual. 1.2 Problem Statement Women contribute immensely to household welfare (education, health, feeding) through various economic and non-paid household activities. Even in situations where women are not gainfully employed, they still strive hard; going through great lengths to ensure the family is well taken care of. Women’s contribution is therefore very significant because they tend to focus more on the welfare of the household. Children’s welfare are the responsibilities of parents but often left in the care of women instead of men. As such, if women have more economic opportunities, welfare of children would necessarily be improved. Ibnouf (2009) citing Levin et al. (1999) established that women’s ability to access cash economy resulted in an improvement in household economic status where children are raised. Women’s share in asset ownership can influence the empowerment of women and improve their well-being (Agarwal, 1997). For instance, a wife economically empowered even at the household-level, will be able to be of good economic standing in the event of death of her husband. In Ghana, society views men as the heads of the households and hence assets and any form of property must therefore be controlled by men. Thus, making access to and ownership of assets by women difficult and marginal as compared to their male comparators. University of Ghana http://ugspace.ug.edu.gh 6 Women’s little ownership of asses has a detrimental impact on the welfare of the household as it reduces their ability to optimally cater for the nutritional needs of the household. Regardless of the upmost significance of asset ownership by women, very little research has been conducted in this field. Policy makers therefore do not have a road map to follow in order to make policy recommendations directed specifically to address the problem of asset ownership by women. The few studies conducted have not critically examined the factors of the gender asset gap. Research has also been conducted to focus on gender, household decision-making and its effect on wealth accumulation by women (Seguino and Floro, 2003). There has also been studies on gender and property rights (Deere et al., 2013). Several studies have been conducted in other parts of the world including developing countries, United States and United Kingdom, with little studies on Ghana (see Deere et al., 2012). Ghana, which is noted as one of the largest growing economies in Africa, cannot boast of much research in the area of women’s access to and ownership of asset irrespective of its recognized importance to women’s well-being. Until recently, data on net worth and asset composition at the individual-level were not readily available. However in 2010, a survey (the Ghana Household Asset Survey) was conducted to serve as a national individual-level asset data. This further enabled research to be conducted on gender inequality and asset ownership (Oduro et al., 2011). However, their study mainly focused on measuring the gender asset gap in Ghana. Issues on gender inequality in asset ownership arise because of some specific factors which could be socio-economic and/or institutional related. However, previous research has not University of Ghana http://ugspace.ug.edu.gh 7 clearly examined the specific determinants of gender asset gap and more essentially what patterns to Ghana. 1.3 Objectives of Study The main objective of the study is to examine the determinants of the gender asset gap in Ghana, using data from the 2010 Ghana Household Asset Survey (GHAS). The answers to the above research questions will achieve the objectives underlying this study. Specifically, the study seeks to:  Examine the socioeconomic characteristics of asset owners.  Identify factors underlying the gender wealth gap in Ghana, and also  To examine how the factors of gender wealth gap significantly impact on the difference in the wealth levels of men and women. 1.4 Research Questions The following research questions motivate this study: i. Is the Gender asset gap biased towards men or women? ii. What are the socioeconomic characteristics of the asset owner? iii. Do these socioeconomic characteristics significantly impact on the wealth levels of men and women? iv. What are the factors that influence the gender wealth gap in Ghana? 1.5 Significance of Study The issue of gender inequality and for that matter gender asset gap, is of great concern to the world economy, and the developing economies especially. The fight for gender equality has been on the mantra of most government policies, yet there exist a substantial gap. It is University of Ghana http://ugspace.ug.edu.gh 8 therefore imperative that this topic be explored to expose some of the deficiencies in previous gender policies in order to help direct the course of global development from gender perspective. Deere et al. (2006) argued that headship analysis into the study of development outcomes such as poverty gave only a partial outlook of gender inequality given that these analyses do not consider the position of women within male-headed households. Drawing on the individual-level data from the Living Standard Measurement Studies for Latin America and the Caribbean, their findings suggest that type of asset largely defined the degree of gender inequality in various countries. And just as numerous research have suggested, again the gender asset gap favours men and is considerably greater in relation to land. Interestingly however, Nicaragua and Panama observed gender equity with respect to the share of men and women homeowners. Further, review of literature reveals no indication of work done in examining the determinants of gender asset gap in Ghana. Study by Oduro et al. (2011) and other researchers utilized a more general approach to explore and measure the gender asset gap. However, this work is very specific and will focus on the determinants of gender asset gap. The study is significant because it explores and exposes the influence of these factors (age, marital status, education, religion, economic status, region, location, and ethnicity) on the issue of gender asset gap in Ghana, in which case will compliment previous study by Oduro et al. (2011). This study however adds to several related works and will serve as a reference for other researchers who would want to investigate gender inequality and asset related issues. The University of Ghana http://ugspace.ug.edu.gh 9 study will also serve as a road map for researchers who will desire to adopt the same technique in addressing similar issues in other geographical areas and would arouse some interest for further study in that regard. Again, findings from the study will help stakeholders, government officials and policy makers to gain a better understanding on the issues of gender gap in the ownership and control of assets. Essentially, identifying the factors that influence the gender asset gap in Ghana will enable government to fashion out policies that can empower women and marginalize the asset gap in Ghana. Identifying the factors that are associated to women’s ability to own assets in Ghana will also ensure that policy makers fashion out instrumental and social protection policies tailored to address these challenges. 1.6 Organization of Study The study which seeks to examine the determinants of the gender asset gap in Ghana will comprise five chapters. The introductory chapter gives a background to the study, the problem statement and the significance of the study. Chapter two will review existing empirical literature on the research area. It will also discuss the theoretical framework underpinning the study. Chapter three will present the methodology. This will comprise the model, estimation technique and data used in the study. Presentation and discussion of results will be discussed in chapter four whilst chapter five will present conclusions and policy recommendations. University of Ghana http://ugspace.ug.edu.gh 10 CHAPTER TWO LITERATURE REVIEW 2.0 Introduction This chapter reviews both theoretical and empirical literature on the gender asset gap and some related studies within the area of gender inequality from the economics perspective. The chapter goes on to examine the feminist strand of gender inequality and how to forge gender equity in the face of social norms and culture that tend to ascribe some and reserve some activities (both economic and social) to some group of society. Also, the issue of gender asset gap is critically analysed and the various measures of gender asset gap are examined. The empirical aspect of the chapter however, focuses on the several studies on gender differences in terms of asset ownership, as well as the wealth difference between men and women. It also reviews other studies in the area of gender differential that employed the Blinder-Oaxaca decomposition technique. 2.1 Theoretical Literature This section of the literature review focuses on sex and gender, looks at some areas of gender inequality (sociological, economical, psychological, etc.), gender asset and gender asset gap. It also introduces the feminist argument to the issue of gender disparity and how they sought to approach the issue of female delineation in asset distribution. 2.1.1 Sex and Gender The concept of sex and that of gender are used in the wrong context and approached in an inappropriate direction. Usually these two concepts are quite confusing, such that they are usually used incorrectly. According to SOFA (2011), sex is a biological concept while gender is a social concept. While sex describes the innate biological condition of being man or University of Ghana http://ugspace.ug.edu.gh 11 woman, and deals with the biological differences between a man and a woman, gender on the other hand is culture dependent and can be altered over time. Sex is biological and fixed whereas gender is social and can change with time. This issue of sex and gender is evident in the following citation from Quisumbing (1969): Sex differences are due to innate biological differences between man and woman. Gender on the other hand, arise from the socially constructed relationship between man and woman (Oakley, 1972). These differences affect the distribution of resources and responsibilities between men and women, and are shaped by ideological, religion, ethical, economic, and cultural determinants (Moser, 1989). Being socially determined, this distribution can thus be changed through conscious social action including public policy [Quisumbing, 1969: 34] According to Lindsey (2007), on “the sociology of gender”, sex is used to refer to the biological features distinguishing male and female. Gender on the other hand refers to those social, cultural, and psychological traits linked to males and females through particular social contexts. According to him, sex is an ascribed status because a person is born with it, and it makes us male or female; while gender is an achieved status which is learned, and which makes us masculine or feminine. 2.1.2 Sociological Concepts of Gender The social structure, statuses and roles allow us to manage our lives in a consistent and predictable manner. Along with social and institutionalized norms, they define our behavior and ensure smooth interaction with people who occupy different social statuses, whether we know them or not (Lindsey, 2007). University of Ghana http://ugspace.ug.edu.gh 12 However, rigidity of some of these norms results in stereotyping – overly conceptions that people within a particular status group share a common trait. These stereotypes usually take negative forms, which are used in justifying discrimination against members of a given status group. Biologically, the statuses of males and females are stereotyped. Women are perceived to be flighty and unreliable because of their uncontrollable raging hormones (Pearson, ND). This negative stereotyping could result in sexism, with the belief that the status of women is inferior to the status of men. Sexism is perpetrated by systems of male-dominated social structure (patriarchy) leading to the oppression of women (Lindsey, 2007). The system of patriarchy and androcentrism leads to the belief that gender roles are biologically assigned and therefore unalterable, and this reinforces sexism. An example in most developing countries is the belief that women are biologically suited for domestic roles which has restricted educational opportunities and achieving literacy. This condition has made men lords and guardians on the topic of gender and the role of men and women in society. Until recently, this history has been recorded from an androcentric stand that ignored the other half of humanity. 2.1.3 Gender Differences, Asset and the Gender Asset Gap The issue of gender differences can be analyzed from diverse perspectives. The issue could be analyzed from the socio-economic, psychologically and biologically context among others. In the labour market, these differences are seen in the percentage of men and women in the labour force, the type of occupation they choose or are engaged in, their relative incomes or hourly wages4. Generally, gender gap is used to refer to the differences between women and men, especially as reflected in social, political, intellectual, cultural, or economic 4http://www.econlib.org/library/Enc/GenderGap.html University of Ghana http://ugspace.ug.edu.gh 13 attainments or attitudes. It has since extended to other areas of knowledge used to refer to differences in major characteristics of men and women within various fields. This means that gender differences could result and be established from differences in employment opportunities, human capital endowment, access to productive resources (i.e. land), financial services, information, farm labour, culture, religion, sports, among others. However, one most important dimension of gender differential or gap and for which not much analysis has been done is the gender gap in asset ownership. In the developing countries especially, gender asset gap is a far reaching concept that is affecting all aspects of these countries. The concept of gender gap is very wide; the gender gap exists in enrolment, labour market participation, occupations, wages, access to financial resources etc. Gender gaps can be investigated across a wide range of dimensions. The livelihood of people is dependent on a wide range of factors. Humans depend on assets, income sources, products, and the interaction with the labour market as the basic sources of livelihood. According to Bebbington (1999), assets do not only provide people with the opportunity to earn a living, but also add meaning to people’s lives. Assets are owned by households and individuals in different forms and volume, both tangible (machinery, land, livestock, plants, etc.) and intangible (education, social relationship, among others). The gender asset gap is a measure of the difference in the ownership of assets by women and men, and it can be conceptualized in different ways. It explains the difference in the right to access, own and control assets between men and women. The issue of gender asset gap tend to mostly be in favour of men. The level or degree of access, ownership, and control of the University of Ghana http://ugspace.ug.edu.gh 14 various forms of assets (human, physical, natural, social, etc.) between men and women explains the disparity or how wide or narrow the gap is. 2.1.4 Feminist Economics According to Strassman (1999), an economics that serves the interest of large and diverse group of people has been developed by some feminist strand of economics. Some theories in traditional economics depict women as dependent on father, husband and male partners by considering the family as a basic economic unit which is evident in most labour market approaches like the Chauvinist model of labour supply. Feminist economists however disagree with the traditional economic models of gender which perceived women as dependent on father, husband and male partners. According to feminists, the idea of male chauvinist assumptions enforces women’s dependence on men, their secondary status within the family, the community and their less involvement or somewhat exclusion from household and general decision-making (Getachew, 2012). This however brings to bear the feminist theory of economics and decision making which serves the interest of large and different group of people. According to feminists, apart from the economic factors of gender inequality, cultural factors are also very important in explaining issues of gender difference in the area of agriculture in most developing economies. Marilyn (2004) categorized the feminist argument into domestic issues, economic success, human agency, ethical judgment and gender, class, race, power, among others. There are three well known feminist methodologies: domestic system methodology, a human agency, and the ethical judgment methodologies. University of Ghana http://ugspace.ug.edu.gh 15 The domestic system methodology argues that gender gap in productivity and for that matter agriculture, is heavily explained by the fact that majority of the productive efforts of the women are in domestic unpaid jobs or work (i.e. in the kitchen, washing, managing the house, caring for children, etc.). This strand therefore argues that the household should be regarded as an important and necessary institution in the productive sector, such that all unpaid jobs performed in terms of domestic chores in the domestic setting are accounted for in productivity, hence must be valued as such. The second methodology, human agency methodology, is concerned with the relationship between women and their environment as a whole. It focuses on women’s relationship with a given system, their relationship with people and the social institutions (Getachew, 2006). Their argument on the relationship of women to society stems from the fact that power is defined in a social context and within a confine (society). It therefore focuses on where power lies within the social system and how it is accessible. Women generally are said to have limited access to institutions in a given social or institutional setting and this constrains their involvement, and/or largely influences their authority in decision-making. Lastly, the ethical judgment approach to the feminist economics according to Folbre (1994) asserts that issues are better judged from the moral point of view. This approach stands and argues on specific moral positions than from the neutral observer’s point of view. 2.1.5 Determinants of gender differences in Asset Individuals, households, or economic units differ in all aspects of human nature. They differ in taste and preference, consumption patterns, savings behaviour, emotional and psychological make-up. University of Ghana http://ugspace.ug.edu.gh 16 Differences in economic outcomes and the analysis of their determinants are of utmost importance in various domains. Especially with respect to gender, there is massive (political) interest in whether these differences reflect discrimination or whether they simply arise from differences in relevant observable characteristics between men and women. In labour- economics however, most prominent studies in the areas of gender differences focus on the gender wage gap (See, e.g., Blau and Kahn, 2000; Baah-Boateng 2012; Biltagy, 2014). Any persistent differences between men and women in these observable characteristics (occupational status, education, experience, legal and social systems) will lead to gender differences in wealth accumulation. Empirical evidence indicates that potential gender differences may exist in many areas. Below are some discussions of these. Men and women differ in their attachment to the labour market. Warren et al. (2001) remarked that any disadvantage in net worth is partially attributable to lower female labour force participation. According to them, the normal trend or standard is a continuous full-time labour market attachment for male breadwinners, while women tend to have part-time work arrangements (including potential wage penalties), often resulting from diversified and perforated work histories due to child bearing and child rearing and more frequent job changes (Berger and Denton, 2004). Differences in earnings are yet another potential source of the wealth gap. Given a persistent gender gap in earnings, even when holding savings rates constant, women are expected to accumulate lower levels of wealth (Blau and Kahn, 1997, 2000; O’Neill, 2003). University of Ghana http://ugspace.ug.edu.gh 17 Empirical evidence from Jianakoplos and Bernasek (1998) indicates that women and men differ in their risk preference and hence, their returns to savings. Women tend to invest more conservatively than men. More conservative investment patterns in the past according to Brush et al. (2002) have led to lower returns to wealth, but at the same time may have protected women from higher risks associated with the stock-market at times of economic downturns. Also, they find that a relative lack of social networks reduces women’s access to venture capital, thus leaving them out of this particular avenue of wealth creation 2.2 Empirical Literature Warren (2006) hypothesized that an analysis of the levels of asset held by men and women has the likelihood of exposing a marked gender wealth advantage for men. Thus, other key social divisions are likely to result in distinct differences in typical wealth portfolios that divide men and women, and group some women with some men. She remarked that gender differences in asset ownership, decision making, pay, employment and other social, economic and psychological outcomes among others could be attributive to several socioeconomic, demographic, ethical, and natural forces like differences in employment, human capital, experience, religious affiliation, parenting and parenthood, inheritance, legal systems, stereotyping, etc. There are however different strands of literature that examine differences in the gender asset in terms of ownership, acquisition, levels and values of assets owned by men and women. Some studies try to identify the existence of asset gap between men and women, while others which focus on explaining the existence of this gap look at its causes or determinants and the impact of these factors on the gap (whether it bridges the gap or widens it). University of Ghana http://ugspace.ug.edu.gh 18 Review of literature reveals that there are gender disparities in asset ownership. Study by Warren (2006) to investigate the impact of gender, class, and ethnic divisions on inequality in wealth accumulation in the United Kingdom (UK) employed a sub-sample of individuals aged 18 to 59 from the 1996 data of the Family Resource Survey (FRS) and the distribution of individual-level pension wealth. The study looked at the distribution of wealth by gender, and was also concerned with the assets held by men and women from diverse class and ethnic groups with the main focus of the study being the size of gender wealth gap in the United Kingdom. The FRS is a large-scale survey that provides information from a sample of 26,000 households in Britain. The 18 to 59 age group was used because level of wealth was argued to have some correlation with age since the type and amount of asset held differs from the working-age class and the retired5. By family, warren referred to the “benefit unit”: either a single adult or couple living as married and any dependent children. Wealth in the study referred to a combination of financial, housing, and pension wealth. The study revealed that pension savings were much substantial for women and men who occupy a higher occupational class. From the gender perspective also, findings show that most advantaged group among employed women fared poorly in terms of pension. Gender and occupational class were found to significantly impact on pension. Analysis again showed that on the whole, men had not only accumulated pension assets, but also the levels of assets they had accrued were far greater, compared to their female counterpart. For instance in the managerial or professional levels, women held a median of £7,000 as compared to £22,000 held by their male counterparts. In general, the study finds that the gender distribution of 5 At age 59, women in UK become eligible for state pension. University of Ghana http://ugspace.ug.edu.gh 19 pension wealth is much more skewed than for total wealth, with women accounting for 29 per cent of pension wealth but 44 per cent of total wealth. Yamokoski and Keister (2008) also explored the differences in wealth in the United States (US) by gender, marital status, and parenting status. Using the National Longitudinal Survey of Youth (NLSY6-79) administered by the US Bureau of Labour Statistics (BLS) between 1979 and 2000, the study explored the effects of these factors on overall wealth for young baby boomers, born between the early 1950s and mid-1960s.Wealth estimation was carried out using data from 1985 through 2000 with assets and debt values adjusted by the Consumer Price Index (CPI). The study also used the likelihood-based general linear regression to model net worth, and in estimating the models they controlled for other individual and family attributes related to ownership of wealth. Net worth of the adult family of the respondents was estimated in three different equations. The first equation was modeled to examine the independent effects of gender and marital status, the second model looked at the combined effects (of gender and marital status) and the last model the assessed the joint impact of gender, marital status, and parenthood on net worth. Results showed that marital status is a very strong indicator of adult wealth. It also indicates that the gender gap in wealth accumulation between single men and single women is minimal. This was accounted for by delayed age at marriage and childbirth, as well as reduced fertility rates. 6 NLSY is a national representative longitudinal survey that was administered nineteen times between 1979 and 2000 by the BLS University of Ghana http://ugspace.ug.edu.gh 20 In a study by Deere and Leon (2003) using the various national rural household surveys in the early years of 2000 found that women are less likely to own land than their male counterparts. The study revealed that women represented 11 percent of the share of landowners in Brazil and 27 percent share in Paraguay. They focused specifically on the significance of the gender asset gap with respect to land in Latin America. They argued that gender disparity in land ownership is influenced by male preference in inheritance, the advantages men have in marriage, state programs and gender discrimination that exist in the community. Also, the gender stereotype found in the land market coupled with a weakness in the credit and labour market ultimately results in constraint of women’s access in the land market relative to men. A household survey in the Savelugu-Nanton District in Ghana by UNICEF/IFPRI/UDS in 2001 revealed that gender gap in landownership was far greater in favour of the men: men individually owned farms in about 72 percent of households surveyed whilst farms owned by women represents almost 48 percent of households. In Ghana, the study on the gender assets gap by Oduro et al. (2011) to measure the asset gap employed the 2010 Ghana Household Asset Survey in its analysis. Findings from the study indicate the existence of significant asset gap between men and women. On the ownership of household’s place of residence and agricultural land, the result was biased in favour of the men (with more than 70 percent of men) as against 27 percent of women owning household’s place of residence, and nearly 62 and 38 percent share of agricultural lands for men and women respectively. Again, although women were found to have a greater share in the ownership of businesses, the mean value of business assets for men far outweighed that of women. University of Ghana http://ugspace.ug.edu.gh 21 Study by Oduro et al. (2011) was however very specific on measuring the gender asset gap that existed in Ghana. It also looked at the evidence on gender gaps in assets ownership and some determinants of these gaps in the Ghanaian context. However, the study did not look at the significance of these factors and its effect on the gender asset gap. This study therefore intends to bridge the gap by looking at these determinants of gender asset gaps in Ghana and their significance on the subject matter. Also employing the 2006 Household, Income and Labour Dynamics in Australia (HILDA) in a survey Jefferson and Ong (2010) examined the gender differences in asset and debt portfolio in Australia where wealth was classified into 11 categories including wealth stored in primary home, business, other property, equity and cash investment, cash redeemable life insurance, vehicles, among others. They defined debt to also include debt secured against the primary home, other property, credit cards, business, and Higher Education Contribution Scheme (HECS). In the analysis, they categorized asset and debt portfolios by gender and household (i.e. couples, single men, and single women). The descriptive statistics indicated that regardless of gender and household type, primary home was the dominant asset owned by Australian household with 67 percent household representation owning a primary home (which constituted 42 percent of Australian wealth). The average wealth of an Australia household was also estimated to be $A727, 200. Out of all the 11 asset types used to represent wealth in the analysis, except for primary home, other property, and collectibles, the average single woman is reported to have a lower level of wealth as compared to an average single man, where in the primary home category of wealth, single women dominated with a share of 55 University of Ghana http://ugspace.ug.edu.gh 22 percent (with a primary home value of $A210,900) as against 40 percent ($A47,300) for couples. Empirically analyzing the data, Jefferson and Ong (2010) also revealed some gender dimensions to the composition and value of asset and debt portfolios. In terms of asset holdings, women had less in all respect as compared to single men household and that of couples, with individually owned assets biased in favour of men in all categories and across all types of asset provided by the data. In general, the asset portfolio of couples and single men were found to be comparably more diversified than that of single women who overly rely on the primary home type of wealth. The Herfindahl Index7 estimation for couples, single men and single women were calculated to be 0.53, 0.62, and 0.70 respectively. However, Jefferson and Ong (2010) could not investigate intra-household gender gap across the entire range of assets and debt portfolios. The Oaxaca decomposition technique is largely used to identify and quantify separate contribution of differences in groups in measurable characteristics. Study by Sherman (2007) employed the Oaxaca decomposition technique to analyse differences in ownership of high- risk assets using the 2004 Survey of consumer finances. The study however was not able to appropriately capture the relationship between the dependent and explanatory variables of the differentiated group. Research undertaken by Deere and Doss (2006) analyzed evidence available on gender wealth distribution across the globe and discovered if there is more to this. One of the 7Herfindahl index indicates the extent of diversification in asset portfolios and it is measured as the sum of the squared values of each asset’s share in the total wealth portfolio. University of Ghana http://ugspace.ug.edu.gh 23 conclusions that were drawn acknowledged the existence of gender wealth gap underscores marital status and parenthood as being significant determinants of wealth. Using the 2001 New Zealand Household Survey, Gibson et al. (2007) observed that the mean net wealth of households of married couples was considerably higher than that of single individuals. Interestingly, single women were also found to be better off than single men in terms of the median and mean net wealth. With the same data, John Gibson, Trinh Le, and Steven Stillman (2007) examined the wealth gap between immigrants and New Zealand- borns. They established that single migrants have a significant high wealth than New Zealand-born singles and to a large extent could explain by the age differences. Doss (2012) examined the gender asset gap in agricultural assets in Kenya. She discovered that there exists a significant gender asset gap in agricultural resources and this gap was mainly explained by marital status. The gap in deeded land was found to be more prominent in relation to the gender gap in agricultural land. Conley and Ryvicker (2005) stated that evidence had shown that there still existed a gender asset gap in the United States even after controlling for lower incomes of female heads. They therefore aimed to identify if some factors such as inheritance, savings rates or investment yield contribute to the gender asset gap. Their results established that differences in savings rates between men and women significantly explained this gap. Njuki and Mburu (2011) in an attempt to bridge the gap that exists in the literature concerning gender, livestock and asset ownership, examined the role of livestock as an asset for women in Kenya, Tanzania and Mozambique. The study results indicated that despite the University of Ghana http://ugspace.ug.edu.gh 24 high proportion of households where women owned livestock, women still owned far less livestock compared to men. For instance, men in Kenya and Tanzania owned 10 times and 18 times more cattle than women, respectively. Results also showed that the patterns of livestock ownership were highly differentiated across the three countries. Rustagi and Menon (2010) also examined the multitude of inheritance system and legal provisions existing in the Asia Pacific region with special reference to the challenges faced by women. The study looked at the evidences on gender asset gaps in respect to command over major assets (land, property, and housing) in the South and West Asia, North East and South East Asia, and the Pacific region. It employed the ownership and operation of enterprises, microfinance, SHG formation, and the role of women’s organization as the critical models or approaches for improving women’s authority over economic and financial resources in the Asia Pacific region. The two major sources employed in the analysis were the Global Entrepreneurship Monitor (GEM) on women and entrepreneurship and the Global Gender Gap report of the World Economic Forum (WEF) which were used to expatiate on the subject of gender gap. The Global Gender Gap report covered 130 countries (Hausmann et al., 2008) and provided the survey-based data on the ability of women to elevate to the level of spearheading enterprises in various countries, including 21 in the Asia Pacific region. Findings revealed that country level discriminatory laws concerning right to access, ownership, and control of lands and housing still existed although these rights are strongly upheld and affirmed in international laws, while efforts to improve property and inheritance rights are also constrained by lack of political will. Lack of awareness in terms of the level of University of Ghana http://ugspace.ug.edu.gh 25 literacy also accounted for this gap. The high level of illiteracy of women limits their rights on land (as in Cambodia) and affects their capabilities and earning potential as a result of lack of technical expertise which is otherwise available to men engaged in similar fields. Improper networking is also found to account for the gender gap in asset ownership. Findings pointed out that poor social networking among women acts as a constraint, hence serving as a blockade to women-led enterprises. In Germany, Sierminska et al. (2010) also examined the gender wealth gap using the German Socio-Economic Panel (SOEP) which collects wealth information for all adult household members at the individual level. The sample analysed by Sierminska et al. (2010) included 12,700 households with about 24,000 respondents surveyed in 2002. Using net worth as a dependent variable and a vector of four wealth determinants: labour market experience, educational level, intergenerational characteristics and demographic characteristics, they adopted the decomposition technique introduced by DiNardo, Fortin, Lemieux (1990) in order to avoid the risk of incorrect capturing of the relationship between the dependent and predictor variables. Findings from their study revealed that across most of the distribution the gap is explained by the differences in labour market experience (current income and experience) and is not greatly related to other characteristics like education. The study found the effect of labour market experience to be very strong in explaining the distribution of wealth between men and women. However, a very large proportion of the gap results from differences in how men and women accumulate their wealth, which depends on their characteristics. University of Ghana http://ugspace.ug.edu.gh 26 Also, the study revealed that among the married or cohabiting, men on average hold more wealth for all wealth composition with the difference being particularly wide for business assets. In owner-occupied homes however, the gap appears to be less pronounced for the married. According to Sierminska et al. (2010), there exist to be significant gender wealth gap in Germany of about 30,000 euros. They found that education has little impact in explaining this gap. However, the gap was largely associated with differences in current income and labour market experience. Their findings on the gender gap with regards to education confirm a decline in the rate of female illiteracy. Chauvin and Ash (1994) also employed the Blinder-Oaxaca to decompose the gender pay gap in base pay, contingent pay and total pay of a sample of American business school graduates. Including three indicator variables relating to occupation in the professional, technical and sales categories in the model, they found that 9% of the gap in total pay was unexplained or accounted for by the regressors of the model. There were no unexplained pay gap for base pay. By further decomposition, it was realized that gender differences in contingent par accounted for the unexplained pay gap in total pay; with this gap disappearing with the introduction of contingent pay in the model. This basically suggests that the type of pay data used as the dependent variable of the model strongly influences or determine the size of the gender pay gap. A study by Groshen (1991) also used the Blinder-Oaxaca decomposition on a data from five American Industry Occupational Wage Survey that ranged from 1974 to 1983. Analysis of data with industries which were decomposed separately (Gosse, 2002) found that occupation was highly segregated with the occupation variable accounting for more than 50% of the observed gender wage gap, where wages were strongly related to the proportion of females in University of Ghana http://ugspace.ug.edu.gh 27 the occupation. Males and females who worked in the same occupation were found to be segregated and among the employers, there was total segregation of most occupations. Contrary to Groshen (1991), findings from Bayard, Hellerstein, Neumark, and Troske (1999) indicated that females were segregated into lower paying industries and occupations, with the largest influence on the gender pay gap resulting from lower wages of females compared to males for the same occupation (job cell). This study matched employees’ record using the 1990 Worker-Establishment Characteristics Dataset constructed from the Australian Decennial Census to employers listed in the US Census Bureau’s American Standard Statistical Establishment List. In Egypt, Biltagy (2014), in estimating the gender wage differences employed the Oaxaca decomposition technique in its analysis. The study used data from the Egypt Labour Market Panel Survey (ELMPS 2006, 2012) which was presented by the Central Agency for Public Mobilization and Statistics (CAPMAS) in collaboration with Economic Research Forum (ERF). Estimation results from the study revealed a gender wage gap of 25% and 21% for 2006 and 2012 respectively. Education and experience were found to result in the explained difference in the model and differences due to discrimination against women to determine difference due to selectivity bias. However, the overall gap was attributed to discrimination against women in 2006 and 2012. 2.3 Summary In summary, even though most studies have been undertaken in the area of asset ownership and distribution between men and women in general, little has been done on the causes on gender asset gap. Review of literature indicates that the gender asset gap is usually in biased University of Ghana http://ugspace.ug.edu.gh 28 towards men, while the women suffer the incidence of constrained right to ownership of asset and wealth in general. The review looked at some classification of assets, the distinction between sex and gender, what gender gap is and how to measure this gap. This study however will go beyond the incidence of gender asset gap and try to examine the factors associated with the gender asset gap. University of Ghana http://ugspace.ug.edu.gh 29 CHAPTER THREE METHODOLOGY 3.0 Introduction This chapter discusses the methodology used and the data for the study. It looks at the theoretical framework of the study as well as the estimation procedure. It discusses the research design with emphasis on the variables of the model and the econometric technique employed in the study. The study also adopts the Blinder-Oaxaca wealth decomposition approach to show the proportion of gender wealth gap that can be explained by the regressors, and the proportion that is unexplained by it. The estimation of the gender wealth gap employs the 2010 Ghana Household Asset Survey (GHAS) data. 3.1 Measuring the Gender Asset Gap The gender asset gap can be measured in diverse ways using the various forms of asset under consideration. Oduro et al. (2011) in measuring gender asset gap, employed three approaches which they labeled Gap 1 and Gap 2 and the wealth gap. Gap 1 measured the distribution of asset owners by sex. This approach considered all individuals who are owners of an asset as owners of that particular asset, and also counted only once (as one) if the individual owned multiples of a particular type of asset. The Gap 1 is measured as the proportion of female or male owner to all owners of an asset, where the individual is treated as the unit of observation. Gap 1 is mathematically measured as follows: 𝑀𝑎𝑙𝑒 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 𝐴𝑙𝑙 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 ; 𝐹𝑒𝑚𝑎𝑙𝑒 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 𝐴𝑙𝑙 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 University of Ghana http://ugspace.ug.edu.gh 30 Where; 𝑀𝑎𝑙𝑒 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 𝐴𝑙𝑙 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 + 𝐹𝑒𝑚𝑎𝑙𝑒 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 𝐴𝑙𝑙 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 = 1 This implies that the proportion of assets owned by females could be written as: 𝐹𝑒𝑚𝑎𝑙𝑒 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 𝐴𝑙𝑙 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 = 1 − 𝑀𝑎𝑙𝑒 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 𝐴𝑙𝑙 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 The second approach, Gap 2, measured the asset gap by the incidence of assets ownership. It is measured by comparing the percentage of female owners of all women with that of male owners of all men. Gap 2 was basically calculated for adults (persons aged 18 years and beyond) as follows: 𝐴𝑑𝑢𝑙𝑡 𝑀𝑎𝑙𝑒 𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 𝐴𝑙𝑙 𝐴𝑑𝑢𝑙𝑡 𝑀𝑎𝑙𝑒𝑠 ; 𝐴𝑑𝑢𝑙𝑡 𝐹𝑒𝑚𝑎𝑙𝑒𝑂𝑤𝑛𝑒𝑟𝑠 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡 𝐴𝑙𝑙 𝐴𝑑𝑢𝑙𝑡 𝐹𝑒𝑚𝑎𝑙𝑒𝑠 Again, the gap could be measured using the gender wealth gap. Oduro et al. (2011) analyzed the gender wealth gap in three ways. First is by comparing the gross value of assets owned by men and women. For the second form of measurement, instead of using the gross value, the mean gross value is used instead for the comparison. Third is by estimating the share of gross value assets owned by women. 3.2. Model Formulation The model for the study is based on the theory of consumer behaviour which assumes a constrained utility maximization, where the rational consumer or individual thrives to maximize utility with limited resources. The study employs the multiple regression analysis and the Blinder-Oaxaca decomposition approach in examining the gender wealth gap in University of Ghana http://ugspace.ug.edu.gh 31 Ghana. The multiple regression analysis is used in determining the effect (both magnitude and direction) of the predictor variables of the model on level of wealth accumulation for both men and women. The Blinder-Oaxaca decomposition technique is however used to examine the difference in wealth between men and women (both the explained and unexplained differences). The Blinder-Oaxaca decomposition techniques decomposes the total difference in gender wealth into explained and unexplained differences (i.e. differences explained by the predictors of the model, and that due to probable discrimination or potential effects of unobserved variables). This study uses the technique to examine how the predictors of the model influences the differences in gender wealth. 3.2.1 Theoretical Framework Several economic and social factors largely influence the determinants of gender wealth gap. By large, the theoretical literature that explains the underlying principle behind asset and/or wealth accumulation is numerous. In economic theory, there is the general assumption that individuals arrive at decisions about assets and other economic variables by maximizing their utility subject to constraints. More specifically, both men and women have an expected utility for acquiring assets. Among the numerous outcomes that women can derive from asset ownership includes provision of economic security. Therefore, the expected utility theory will be employed basically due to the fact that an individual’s decision to acquire an asset is associated with the individual’s maximization of expected utility of acquiring that asset since individuals are assumed to be risk averse (Friedman and Savage, 1948). University of Ghana http://ugspace.ug.edu.gh 32 An empirical demand function can be shown to be consistent with the assumption that the potential consumer maximizes an expected indirect conditional utility function given by: 𝑉𝑖𝑗 = 𝑓(𝑥𝑖 , 𝑞𝑖𝑗 , 𝑎𝑖) (3.1) Where 𝑥𝑖 = a vector of observable socioeconomic attributes of individuals such as education, household size, sex and age. 𝑞𝑖𝑗 = vector of observable institutional characteristics j faced by individual i 𝑎𝑖= the price of other goods consumed by individual i Equation (3.1) denotes the standard expression of the indirect utility function in the demand theory of consumer behavior. The consumer then tries to maximize satisfaction (utility) subject to the given constraint. Again, adopting the difference in wealth accumulation used by Sierminska et al. (2010), we develop a simple model of accumulation, where asset at time t+1 (At+1) is expressed as: 𝐴𝑡+1 = (1 + 𝑟)(𝐴𝑡 + 𝑌𝑡 − 𝐶𝑡) (3.2) 𝑌𝑡 − 𝐶𝑡 = 𝑆𝑡 Where (r) is the gross rate of return on investment, (Yt) is Income at time t, (Ct) is Consumption at period t, and (St) represents Savings at time t. From equation (3.2), assets in the current period are influenced by the rate of return (r), asset investment in the previous period (At), and Savings (St). First of all, households differ in their preference for risk (Sierminska, 2010), whether they are risk averse, risk loving, or risk University of Ghana http://ugspace.ug.edu.gh 33 neutral; and this translates to different rates of returns (r) on their preferred portfolio allocations. Also, wealth may differ as households tend to enter into a period with different asset stock (A) from c years accumulation (At) or inheritance, which could result in wealth in subsequent time periods (e.g. At+1). Asset accumulation, and for that matter wealth in period t+1 could differ across units depending on the income and consumption pattern (i.e. savings). The amount saved (i.e. the excess of income over consumption), however, is dependent on the level of income and consumption (which is in turn determined by age, marital status, household size, geographical location, level of education, risk aversion, among others). 3.2.2 Specification of Empirical Model In addressing the direction of the wealth gap to determine which group the gap favours, we regress the log of wealth against the gender variable (i.e. lnwealth = f (Gender)) using the male dummy as the reference category. The coefficient of the dummy variable therefore gives the direction of the gap in relation to the reference category. A positive coefficient (i.e. for female) indicates that, in relation to the reference group (male), women have significantly higher mean wealth than men indicating that the wealth gap is biased towards the female group. The specific equation for the determination of the direction of the wealth gap is as below: 𝑙𝑛𝑊 = 𝑓 (𝐺𝑒𝑛𝑑𝑒𝑟𝑖); 𝑖 = 0,1 ∶ (0 = 𝑚𝑎𝑙𝑒, 1 = 𝑓𝑒𝑚𝑎𝑙𝑒) (3.3) From the theoretical framework, it was stated that an individual’s decision to acquire an asset is associated with the individual’s expected utility maximization with the assumption that the University of Ghana http://ugspace.ug.edu.gh 34 individual is risk averse. The study proposes to estimate the wealth equation (3.3) below based on the theoretical underpinning of the study as discussed. In addition to identifying the size and direction of the gender wealth gap Equation (3.4) will be run separately for male and female samples to examine the variables associated with wealth of women and men. 𝑙𝑛𝑊𝑖 = 𝛽0 + 𝛽1𝐴𝑔𝑒𝑖 + 𝛽2𝑀𝑎𝑟𝑖 + 𝛽3𝐸𝑑𝑢𝑖 + 𝛽4𝐸𝑐𝑖 + 𝛽5𝑅𝑒𝑙𝑖 + 𝛽6𝐸𝑡ℎ𝑛𝑖 + 𝛽7𝐿𝑜𝑐𝑖 + 𝜀 (3.4) i= (male, female,) Where ln𝑊 = Wealth Age = Age of respondent Mar = Marital Status of the individual Edu = Highest level of Education of the individual Ec = Economic status Rel = Religious affiliation of the individual Ethn = Ethnicity of the individual Loc = Location Characteristics of the individual ε = Error term of the model The 𝛽𝑖’s are the coefficients of the regressors in the linear model, this gives the marginal effect of the regressors on the dependent variable. The 𝜀𝑖 refers to the error component of the regression model. It accounts for the random effects in the variables of the model. It also accounts for the exclusion of important variables which could have a significant impact in the estimation of the model. University of Ghana http://ugspace.ug.edu.gh 35 3.2.3 Description of variables of the model The dependent variable of the model is wealth, which is computed as the total gross value of assets under consideration. The study however focuses on physical assets (which includes; agricultural land, agricultural equipment, place of residence, business, livestock, consumer durables, and other real estate) from the Ghana Household Asset Survey data (GHAS), 2010. Wealth in this study is captured by the gross value of total wealth. In the GHAS (2010) data, gross value of wealth was measured for all asset owners. In the computation of the wealth of the individual household (with the exception of businesses), the value of assets that were jointly owned was evenly distributed across owners. The value of businesses that were jointly owned were distributed according to an individual owner’s share of the business (Oduro et al., 2011). The unit of measurement of wealth in the Purchasing Power Parity (PPP) in US Dollars. The choice of these variables in the regression is influenced by the data collected by Oduro et al. (2010). The study seeks to go beyond “measuring the gender asset gap” to investigating the correlates of this gap, by employing the 2010 GHAS data. Potential gender differences exist in many areas and are influenced by several factors (Sierminska et al, 2010). The dependent variable of the model for this study is the mean wealth which is said to be influenced by variables like age, marital status, education, religion, ethnicity, economic status and location characteristics of respondents. Age is measured in years based on the individual’s last birthday. The age of the individual is introduced as a continuous variable in the model. University of Ghana http://ugspace.ug.edu.gh 36 The life-cycle hypothesis propounded by Ando and Modigliani suggests that an individual’s level of wealth accumulation changes with age young households are expected to have accumulated less wealth (Modigliani and Brumberg, 1954; Friedman, 1957), due to their different position in the life cycle compared to the elderly. It is expected that for both women and men there will be a positive relationship between age and wealth all things being equal. Education is defined in the context of formal education, and for that matter the highest educational level of the individual at the time of the survey. The education variable is categorised into five dummy variables, i.e. Never been to school (No education), Basic, Secondary, Tertiary education, and others, with No education as the reference class. The classification is derived from the 2010 GHAS data and it allows for the impact of the various levels of education on the wealth gap between males and females to be captured. Blau and Kaln (1997, 2000) and O’Neil (2003) argued that earnings are another likely source of wealth gap. According to them, women are expected to accumulate lower levels of wealth given the persistent difference in educational level between men and women. Since earnings is tied to human capital or education (all things being equal), we expect a positive relationship between education and wealth for both women and men. Marital status as a dummy: never married, currently married, and previously married assuming the values of 0, 1, and 2 respectively. The value of 1 represents the category of individuals who are married, 2 for those who have ever married and either divorced, separated, or widowed; and 0 represents those who have never married. According to Sierminska et al. (2010), differences in marriage pattern among women and men make a difference in wealth accumulation. University of Ghana http://ugspace.ug.edu.gh 37 In some jurisdictions where community property marital regime is practiced, marriage is more likely to increase a woman’s probability of owning an asset because property acquired during the marriage by the man is also the woman’s property. In the Ghanaian context however, the marital regime is the separation of property. It is not automatic that assets acquired during marriage are owned jointly by the couple (Oduro et al., 2010). Therefore the ability for a married woman to acquire wealth becomes dependent on her purchasing power, opportunities to work to earn an income, among others. This makes the association between marital status and wealth gap ambiguous, a priori. An instance in the Ghanaian setting is the case of a married man who owns an agricultural land. Although the wife cultivates on the land, it does not mean the farm land is jointly owned by the married couple. Also, it does not imply that the wife accumulates more wealth as a result of the income generated from the cultivation of the agriculture land. Economic status is yet another influential factor of the disparity in wealth between women and men as men and women differ in their attachment in the labour market (Sierminska et al., 2010). Warren et al. (2001) remarked that any disadvantage in net worth partially results from the low participation of women in the labour market. While men are said to generally have a standard pattern of continuous full-time attachment in the labour market, women tend to have part-time work arrangements (Bardari and Gornick, 2008) with several breaks within work as a result of child bearing, child rearing, and frequent job changes (Berger and Denton, 2004). Evidently, men and women differ in the kinds of occupation they find themselves engaged in. For many women in developing countries, the type of occupation they engage in provides them with minimal earnings and this influences the wealth gap as women are unable to have University of Ghana http://ugspace.ug.edu.gh 38 the means to purchase property that would eventually grow their assets. Relative to men, women in Ghana have fewer income-generating opportunities. According to Baah-Boateng (2012), in 2006 the labour force participation rate of women was 3 percentage points lower than that for men, recording an estimate of 68% and 71% for women and men respectively. The economic status of women can therefore be said to differ from men. For wage employment and self-employment, an estimate of about 9% and 57% respectively were accounted for by employed women and 27% and 53% respectively were accounted for by employed men (Baah-Boateng,2012). Economic status in this study refers to the type of employment engaged by the individual. Economic status is categorized into as a dummy in the model with0=Not working (reference category), 1=Not working (reference dummy), 2=wage employment, 3=self-employment, 4=casual labour, 5=domestic/family work, and 6= apprenticeship. The religion variable of the model is categorized into 5 dummies; 1= No religion (reference dummy), 2=Christian, 3=Muslim, 4=Traditional, and 5=Other religions. The directional effect of the various religious affiliations is unknown as the various religions do not share a common view on the issues of gender and asset ownership. Region is also introduced in the model is a dummy variable consisting of dummies of the ten regions of Ghana where 1=western, 2=central, 3=Greater Accra (reference category), 4=Volta, 5=Eastern, 6=Ashanti, 7=Brong Ahafo, 8=Northern, 9=Upper East, and 10=Upper West. As indicated in table 3.1, the effect of the region variable on the gender asset gap is ambiguous since the distribution of asset between men and women is expected to differ from one region to the other. University of Ghana http://ugspace.ug.edu.gh 39 The location characteristic of the individual is dummied into 1=urban and 2=rural. In the rural areas where traditions and culture that tend to delimit the activities of women is prevalent, the difference in wealth of men and women is expected to be relatively high. In some African countries and specifically Ghana, most traditions (ethnic group) promotes gender stereotypes (i.e. forbidding women from certain occupations, or tagging some occupations as a male or female-occupation), and this goes a long way to influence the working and for that matter, the earning potentials of the women especially. In Ghana for instance, some cultures or ethnic groups forbid women from engaging in some supposed male-occupations, and rather relegate them to domestic chores and these practices go a long way to influence future decisions of our women. The ethnicity variable is also expressed as a dummy with 1=Akan, 2=Ga-Dangme (reference category), 3=Ewe, 4=Guan, 5=Gurma, 6=Mole-Dagbani, 7=Grusi, 8=Mandi, and 9=other tribe originating from and outside Ghana. Table 3.1: Description of variables and their expected signs Variable Description Unit Apriori Expectation W Wealth (Gross value of Individual’s Physical Assets) USD Age Age of Respondents Years + Mar Marital Status Dummy +/- Edu Level of Education Years + Ec Economic status Dummy + Rel Religion Dummy +/- Reg Region Dummy +/- Ethn Ethnicity Dummy +/- Loc Location Characteristic of individual Dummy +/- Source: Author's expectations, 2014 University of Ghana http://ugspace.ug.edu.gh 40 3.2.4 Specification of Blinder-Oaxaca Decomposition Model Identifying the underlying causes or determinants of racial and gender differences in key economic and social characteristics like education, labour market, health, asset ownership, wealth, and other outcomes has been the goal of an enormous body of literature in the social sciences. The most common approach adopted by most studies in examining gender differences is the techniques of decomposing inter-group differences in mean levels of an outcome into those due to different observable attributes across groups (also known as the endowment effect) and those due to different effects of characteristics or “coefficients” of groups (Fairlie, 2013). This technique is commonly attributable to Blinder (1973) and Oaxaca (1973). The method has also been applied in measuring union wage premiums (e.g. Lewis, 1986), racial wage gap (e.g. Neal and Johnson, 1996), racial test score gaps (e.g. Fryer and Levitt, 2004) and the Ghanaian labour market wage discrimination (e.g. Baah-Boateng, 2012). 3.2.5 Blinder-Oaxaca Decomposition According to Baah-Boateng (2012), the Blinder-Oaxaca decomposition approach is especially used to measure the proportion of gender differentials (i.e. wage differences) that could be explained by differences in observable characteristics (endowment) such as education, experience, marital status and geographical location, to racial and gender gaps in outcomes (Fairlie, 2003) as well as the part that could be attributed to discrimination. The technique is easy to use and only requires coefficient estimates from linear regressions for the outcome of interest and sample means of the independent variables employed in the model. University of Ghana http://ugspace.ug.edu.gh 41 Also, according to Fairlie (2003), the Blinder-Oaxaca decomposition technique is widely employed in identifying and quantifying the separate contributions or share of group differences in measurable characteristics, such as education, marital status, experience and geographical differences, among others to racial and gender gaps in outcomes. The technique is the most commonly adopted in the past decade to examine inter-group difference in mean levels of an outcome into those due to differences in observable characteristics (endowments) across groups and those due to different effects of characteristic or “coefficients” of groups. To begin the decomposition approach, a wealth decomposition function is specified for both males and females as in equation (3.4) below: 𝑊𝑖 𝑚 = 𝑋𝑖 𝑚 ′𝛽𝑗 𝑚 + 𝜇𝑖 (3.4𝑎) 𝑊𝑖 𝑓 = 𝑋𝑖 𝑓 ′𝛽𝑗 𝑓 + 𝜇𝑖 (3.4𝑏) 𝜇𝑖 ~ 𝑁(0,1) Where X is a row vector of wealth determining characteristics which include age, marital status, education, occupation, religion, geographical location (i.e. where household is located) of an individual; 𝛽𝑗 denotes a vector of coefficients of these characteristics, and 𝜇 is a random term assumed to be normally distributed with zero mean and constant variance. The superscript m and f denote male and female respectively, while the subscript i denotes individuals. The dependent variable in the function is wealth. Based on the properties of OLS estimation technique, we write the wealth gap equation in terms of the mean (average) between the two sexes as indicated in equation 3.4 c below (mean wealth of males less that of the female counterparts): University of Ghana http://ugspace.ug.edu.gh 42 ln (?̅?𝑖 𝑚 − ?̅?𝑖 𝑓) = ?̅?𝑖 𝑚 ′𝛽𝑗 𝑚 − ?̅?𝑖 𝑓 ′𝛽𝑗 𝑓 (3.4𝑐) Where ?̅?𝑖 𝑚 𝑎𝑛𝑑 ?̅?𝑖 𝑓 are the mean asset wealth of males and females, ?̅?𝑖 𝑚 𝑎𝑛𝑑 ?̅?𝑖 𝑓 are vectors containing the means of independent variables for males and females respectively (they are the explanatory variables of the model that influence the gender wealth gap), and the 𝛽𝑗 𝑚 and 𝛽𝑗 𝑓 represents the parameter estimates obtained from the wealth functions of males and females respectively. For a linear regression, the standard Blinder-Oaxaca decomposition of the gap in the average value of the dependent (explained) variable, W, can be obtained by subtracting and adding ?̅?𝑖 𝑓 ′𝛽𝑗 𝑚 from equation (3.4c) as below: ln(?̅?𝑖 𝑚 − ?̅?𝑖 𝑓) = ?̅?𝑖 𝑚 ′𝛽𝑗 𝑚 + (−?̅?𝑖 𝑓 ′𝛽𝑗 𝑚 + ?̅?𝑖 𝑓 ′𝛽𝑗 𝑚) − ?̅?𝑖 𝑓 ′𝛽𝑗 𝑓 ; which can be expressed as: ln(?̅?𝑖 𝑚 − ?̅?𝑖 𝑓) = (?̅?𝑖 𝑚 ′− ?̅?𝑖 𝑓 ′) 𝛽𝑗 𝑚 + ?̅?𝑖 𝑓 ′(𝛽𝑗 𝑚 − 𝛽𝑗 𝑓) (3.4𝑑) This gives the Blinder-Oaxaca Wealth decomposition which expresses the wealth gap between males and females (gender wealth gap) as the differences in explained and unexplained components (Haider et al., 2009). The first term on the right hand side (RHS) of equation (3.4d) represents the differences in mean wealth as a result of the differences in observable characteristics between men and women, and the second term reflects the differences in the mean wealth due to the shift coefficient 𝛽. University of Ghana http://ugspace.ug.edu.gh 43 The Blinder-Oaxaca decomposition is however confronted with certain problems (Baah- Boateng, 2012), one key problem and of which much attention will be given is the inherent “index number problem”. This problem implies that equation (3.4d) could be indexed on 𝛽𝑗 𝑓 𝑎𝑛𝑑 ?̅?𝑖 𝑚 ′ (i.e. could be expressed in terms of the wealth of a man if faced with female wealth portfolio or structure) instead of 𝛽𝑗 𝑚 𝑎𝑛𝑑 ?̅?𝑖 𝑓 ′.This would have resulted in equation (3.4d*) as: ln(?̅?𝑖 𝑚 − ?̅?𝑖 𝑓) = (?̅?𝑖 𝑚 ′− ?̅?𝑖 𝑓 ′) 𝛽𝑗 𝑓 + ?̅?𝑖 𝑚 ′(𝛽𝑗 𝑚 − 𝛽𝑗 𝑓) (3.4𝑑 ∗) Equation (3.4d and 3.4d*) are what is technically referred in Blinder-Oaxaca decomposition procedure as the two-fold decomposition. In which case the decomposition is addressed in two-folds; the explained aspect of the gap and the unexplained portion which is usually attributed to discrimination (although it could also capture all potential effects of unobserved variables). 3.2.5.1 Estimation Process The study estimates the Blinder-Oaxaca wealth decomposition model presented in equation (3.4d). In the estimation process however, there is a possible effect of underreporting of wealth. Underreporting of wealth could result from the valuation of some physical assets like agricultural land, or place of residence which probably would not take into consideration the opportunity cost of the asset acquired. Due to this possibility of underreporting of wealth, the OLS estimation of wealth differential model or gap between males and females may produce some biasedness in parameter estimate. University of Ghana http://ugspace.ug.edu.gh 44 We therefore implement the Blinder-Oaxaca decomposition for equation (3.4d) above to determine the differences in the gender wealth in Ghana. The differences are however in two folds; the explained difference which is as a result of differences in observable characteristics or “endowments” and the unexplained difference resulting from differences in potential characteristics which is unaccounted for in the model or due to “discrimination”. 3.2.5.2 Problems with Blinder-Oaxaca Decomposition The Blinder-Oaxaca decomposition technique just like several other econometric techniques is faced with some challenges. Some of the weaknesses of the Blinder-Oaxaca decomposition are explained as follows: i. The Index Number Problem According to Gosse (2002), the most cited problem in the use of the Blinder-Oaxaca decomposition (e.g., Oaxaca, 1973) is the "index number problem". In other words, the choice of reference group in the model affects the results produced by the decomposition. The most common way to remove this problem however, has been to report the results based on the dominant group’s endowment, thereby standardising the literature. ii. The Use of Indicator Variables According to Jones (1983) the Blinder-Oaxaca decomposition has two main issues. Thus, the intercept and indicator variable coefficients are influenced by the reference group(s) used for the indicator variable(s) in the model; and the intercept is influenced by the choice of scale for continuous variables in the model. This he remarked, usually makes the interpretation of the intercepts meaningless. University of Ghana http://ugspace.ug.edu.gh 45 There is no estimate invariance for the coefficients and intercept when indicator variables are included in the model (Oaxaca and Ransom, 1997). Interpretable results occur when only one category of indicator variable is included in the regression, as the intercept result is simply added to the coefficient to produce the required estimate. When more than one category of indicator variable is included in the model, it becomes unclear how the intercept term should be applied which makes the individual effects of the indicator variables hard be determined. In both cases, however, the overall decomposition result and the coefficients of the non- indicator variables are invariant. Finally, Oaxaca and Ransom (1997) suggest that the continuous variable problem does not occur in practice as the scale of these variables is not as arbitrarily specified. For example, experience is always measured in years. Nielsen (1998) proposes that the 1 and 0 values of indicator variables should be replaced with the proportion of observations in each group. This ensures that the decomposition results are invariant to the choice of reference group. This means regardless of the number of groups, the overall decomposition result and the coefficients and intercept are invariant to reference group. iii. The Influence of Proxy Variables A proxy variable is the variable used as a "stand-in" or approximation for a variable that is harder to measure or collect. For example, age squared is often used to proxy work experience (Lambert, 1993). The use of any proxy measures may artificially inflate the unexplained residual in decomposition models (Gosse, 2002). This is evident is Swaffield (2000) who found that the use of potential labour market experience, rather than actual experience, increased the unexplained residual. University of Ghana http://ugspace.ug.edu.gh 46 Also, Lambert (1993) in examining the use of proxy measures of experience, proxies experience into four. The results of proxies for all four experience used (age, Mincer estimation of experience, Mincer estimation allowing for current dependent children and Mincer estimation allowing for both current and past dependent children) produced significant biases in the coefficient. The biasedness was not only on the experience variable and the constant but also, decreasing the effect of education in the same model. The first two specifications of experience (age and Mincer estimation without inclusion of children) produced the largest biases. This basically implies that the use of even one misspecified variable could significantly affect the entire model. According to Levi (1973) the result is that the ordinary least squares estimate of the misspecified variable will be biased towards zero, even if all other variables are measured without error. These highlights some of the weaknesses of the Blinder-Oaxaca decomposition in examining group differences in outcomes. Some extended version of the standard Blinder-Oaxaca has however been introduced (Neumann-Oaxaca decomposition) with some application to the Logit and Probit models. 3.3 Source of Data This study employs data from the 2010 Ghana Household Asset Survey (GHAS). The survey was conducted by a team of researchers from the Economics Department of the University of Ghana, Legon. The GHAS 2010 data consist of 7,979 observations containing household characteristics, demographics, wealth portfolios and ownership of various assets. The survey covered a total of 2,170 households from all regions of the country. The dataset is nationally representative and is the data used in measuring the gender asset gap in Ghana by Oduro et al. (2011). University of Ghana http://ugspace.ug.edu.gh 47 3.4 Summary This chapter has focused on the method and the specific econometric technique used for the study. It highlights the theoretical framework upon which the econometric method was based, the model of empirical estimation, the procedure of estimation, as well as the data employed in the study. The econometric model consisted of the multiple regression and Blinder-Oaxaca decomposition models that would be used in the data analysis and parameter estimation. The main aim is to be able to capture the factors of gender asset gap and how the determinants impact on the gap. The 2010 Ghana Household Asset Survey (GHAS) constitutes the dataset to be employed in the study. University of Ghana http://ugspace.ug.edu.gh 48 CHAPTER FOUR EMPIRICAL RESULTS AND DISCUSSION 4.0 Introduction This chapter deals with data analysis, empirical findings and discussion of the results obtained from the regression analysis. The first section uses descriptive analysis with the help of cross tabulations and graphs (bar charts) to analyze the socioeconomic characteristics of the asset owners, and the second part deals with the econometric analysis of data using both the multiple regression analysis and the Blinder-Oaxaca decomposition technique. 4.1 Socioeconomic Characteristics of Asset Owners As stated previously, the GHAS 2010 data consist of 7,979 observations containing household characteristics, demographics, wealth portfolios and ownership of various assets. Summary statistics on sex indicates that more women than men own assets in the GHAS 2010. In total, 3288 people were interviewed separately from 2170 households that had a total of 7970 members, out of which the majority, 4,164 (constituting an almost 52.2%) were female and the remaining 3,815 (47.81%) were male. However, in examining the gender wealth gap, the study focuses on observations with wealth. The sample basically consisted of all asset wealth owners (i.e. individuals with wealth). In the 2010 GHAS, not everyone is an asset owner; out of a total of 7,979 observations in the 2010 GHAS, only 4,936 of them have values for wealth. Hence the sample for the study is made up of 4,936 wealth owners. Table 4.1 presents the descriptive statistics of the characteristics of the selected sample for the study. As indicated in the table, the total number of observations of the sample for analysis is 4,936; out of which 2,343 (47.5%) are male and remaining 2,593 (52.5%) female. University of Ghana http://ugspace.ug.edu.gh 49 Also, summary analysis of the age variable reports a minimum and maximum age of 0 and 105 years respectively with a mean age of 27.69 for the sample data (see Appendix). The wealth of men also recorded a mean of $7,161.99 as against $5,409.65 for female, while the maximum value recorded $1,062,450 as against $747,289.90 respectively. The differences in mean values of wealth for male and female were tested for significance, using the t-test for independent means, and a null hypothesis of equal mean wealth for both men and women (see Appendix). The test was conducted on the hypothesis that the mean wealth of men is not statistically different from that of women (i.e. the difference in mean wealth between sexes is statistically zero (0)). With 4,934 degrees of freedom, the results shows a probability values of 0.0562 and 0.0281, hence we reject the null hypothesis of equal mean wealth between sexes, and conclude that the mean wealth of men is found to be statistically greater than that of women at 5%. The Stata result for the t-test of the mean wealth between men and women is presented in appendix. Descriptive analysis of the various predictor variables of the model is specified in table 4.1 below. Results indicates that out of the total 4,925 who responded to the question on marital status, 2,713 (54. 96%) representing the majority are not married. This was followed by 1,721 (34.87%) who are married (consesual union, polygamous, monogamous or betrothed marriage) and the remaining 491 (9.95%) are previously married (i.e. divorced, deserted, widowed or separated). University of Ghana http://ugspace.ug.edu.gh 50 The descriptives on marital status by sex as presented in table 4.1 indicates that there are more men (51.82%) as compared to 48.18% of women who are never married. The share of women who are married or previously married outweights that of men by about 4% and 58% points respectively. The results on education indicates that the majority of respondents 2,676 (54.21%) have basic level education, 319 (6.46%) and 263 (5.39%) have secondary and tertiary level eductaion respectively. A considerable number of asset owners (1411 representing 28.59%) have no eductaion. From the tables, it is evident that the proprotion of men with basic, secondary, and tertiary levels of eductaion outweights that for women by about 3% points for each fro the basic, secoondary, and tertairy categories. However, the percentage of women who have no education reports 33.59% as against 23.05% for the the proportion of men who are uneducated. Report on education in Ghana confirms that generally the enrollment rate for girls reduces as one goes up the educational ladder8. This confirms the result from the descriptive analysis on the less share of women at higher levels of education. With regards to the economic status of respondents, analysis revealed that about 30.88% of the total 4,936 sample are self-employed without employees and constitutes the majority of the economic status distribution. This was followed by about 4.46% who are contributing family workers, 3.28% private wage employees, 2.98% public wage employees, 2.19% self- employed with employee, 0.79% apprentice, 0.53% casual labourers and 0.20% domestic employees with pay (see appendix). However 2079 (42.06%) of the sample observation were 8www.sene.ghanadistrict.gov.gh University of Ghana http://ugspace.ug.edu.gh 51 not working. Out of this the share of men constituted 43.23% as against 40.92% for women who were not working. The “not working” observations consisted of full time students, the retired, too young, aged, the disabled, homemakers, and those looking for job. Table 4.1: Descriptive Statistics of Predictor Variables of the Model by Gender Predictor Variables MALE FEMALE TOTAL Act. Percent Act. Percent Act. Percent Asset Owners (N) 2343 47.47 2593 52.53 4936 100.00 Marital Status: Never Married 1406 60.01 1307 50.40 2713 54.96 Married 827 35.30 894 34.48 1721 34.87 Separated 104 4.44 387 14.92 491 9.95 Missing 6 0.26 5 0.19 11 0.22 Education: No education 540 23.05 871 33.59 1411 28.59 Basic 1316 56.17 1360 52.45 2676 54.21 Secondary 184 7.85 135 5.21 319 6.46 Tertiary 164 7.00 101 3.93 263 5.39 Missing 139 5.93 125 4.82 264 5.35 Economic Status: Not working 1015 43.23 1061 40.92 2076 42.06 Wage employee 221 9.43 88 3.39 309 6.26 Self-employed 719 30.69 913 35.21 1632 33.06 Casual labour 15 0.64 11 0.42 26 0.53 Domestic/family 54 2.30 176 6.79 230 4.66 Apprenticeship 12 0.51 27 1.04 39 0.79 Missing 307 13.10 317 12.23 624 12.64 Religion: No Religion 71 3.303 35 1.35 106 2.15 Christian 1664 71.02 1927 74.32 3591 72.75 Muslim 480 20.49 503 19.40 983 19.91 Traditional 118 5.04 120 4.63 238 4.82 University of Ghana http://ugspace.ug.edu.gh 52 Other Religion 3 0.13 5 0.19 8 0.16 Missing 7 0.30 3 0.12 10 0.20 Region: Western 239 10.20 248 9.56 487 9.87 Central 151 6.44 219 8.45 370 7.50 Greater Accra 228 9.73 276 10.64 504 10.21 Volta 229 9.77 275 10.61 504 10.21 Eastern 258 11.01 323 12.46 581 11.77 Ashanti 374 15.96 353 13.61 727 14.73 BrongAhafo 219 9.35 225 8.68 444 9.00 Northern 344 14.68 376 14.50 720 14.59 Upper East 171 7.30 182 7.02 353 7.15 Upper West 130 5.55 116 4.47 246 4.98 Location: Urban 756 32.27 896 34.55 1652 33.47 Rural 1587 67.73 1697 65.45 3284 66.53 Ethnicity: Akan 931 39.74 1,111 42.85 2,042 41.37 GA-Dangme 186 7.94 231 8.91 417 8.45 Ewe 297 12.68 294 11.34 591 11.97 Guan 75 3.20 74 2.85 149 3.02 Gurma 245 10.46 270 10.41 515 10.43 Mole-Dagbani 401 17.11 421 16.24 822 16.65 Grusi 117 4.99 99 3.82 216 4.38 Mandi 16 0.68 20 0.77 36 0.73 Other-Ethnic Groups 74 3.16 73 2.82 147 2.98 Missing 1 0.04 0 0.00 1 0.02 Source: Author's Computation, 2014 University of Ghana http://ugspace.ug.edu.gh 53 It is evident from the table that the proportion of men in wage employment, and casual labour outweighed that of their female comparators by about 6.04 and 022 percent points respectively. The dominance of the male counterparts in the wage employment could be attributed to the difference in share of male to female in the highest levels of education as more men than women tend to attain higher levels of education. The share of women is relatively larger in self-employment, domestic/family work, and apprenticeship by about 4.52, 4.49, and 0.53 percent points respectively. This distribution could also be attributed to the large number of women engaged in petty trading, retail business, and the market women, among others. Concerning the religious afiliations of the sample, data revealed that majority of them are “other christian” with a few of them being hindi. As evident in the descriptive statistics, it is clear tha although there are more women than men in the various forms of religion, the share of women is higher for “other christian”, and “oher non-chriatian” religions. The share of the men howeveris relatively higher than that of the women in the other religions, including “no religion”. The was however, an equal share of men and women in Hindu. The location characteristics in terms of rural-urban location of respondents is also reported in table 4.1 above. Exactly 3,284 (66.53%) of respondents are in the rural areas with the remaining 1,652 (33.47%) in the urban centres. Results show that the percentage of men in the rural areas is greater than thatof the women (67.73 as against 64.45 for the share women). Out of the 1.652 in the urban areas, the share of women (34.55%) outweighs thatof their male comparators(32.27%). University of Ghana http://ugspace.ug.edu.gh 54 The ethicity of the respondents also revealed that the majority of them (41.37%) are Akan, 16.65% Mole-Dagbani, 11.97% Ewe, 10.43% Gurma, and 8.45% were Ga-Dangme. There was however low represention of 3.02% for the Guan and 2.98 for other-ethnicities, and a 0.73% for the Mandi ethnic groups of the total of 4,935 respondents to the question on ethnicity. 4.2 Empirical Analysis 4.2.1 Estimated Regression Models9 As stated in the previous chapter, a simple linear regression will be employed in determining the direction of the gender wealth gap. In doing so, the log of wealth (lnW) was regressed against the sex variable, which takes a value of 1 for female and zero otherwise (male). Table 4.3 below shows the results. Clearly indicated in table 4.3, the coefficient of the female dummy which is significant at 5% shows a negative directional effect (i.e. -0.177). The negative sign of the coefficient of the female dummy implies that, relative to men (reference dummy), women’s wealth is significantly lower than that of men indicating that the wealth gap is biased against women. Table 4.2: Results on Simple Linear Regression Model with lnW as Dependent Variable lnW Coefficient Standard Error Sex female -0.1769542 0.0697436 constant 6.1302 0.0505496 N = 4936 R-Squared = 0.0013 Source: Author's Computation, 2014 9 Age squared (Age2) is not included in the regression model because it is insignicant in the model and does not affect the model. University of Ghana http://ugspace.ug.edu.gh 55 The summary statistic of wealth for men and women as indicated in table 4.2 above also indicates that the mean wealth of men is statistically greater than that of their female comparator at 5 percent. Again, in addressing the question of the significance of the predictor variables and how these variables determine the level of wealth of women and men, and the pooled samples, we analyze results from separate multiple linear regressions for women and men. Table 4.3 provides the results on the wealth determinants regression analysis for both men and women as well as the pooled sample. The table shows the relevant statistics and estimates from the separate multiple regression analysis by regressing the log of wealth on the observable characteristics (i.e. age, marital status, education, religion, economic status, region, ethnicity, and location) for each sex. The table reports all the significant variables for the three models of wealth determination. The F-statistics for the results indicates that the models are highly significant at 1 percent with an R-squared of 0.124 and 0.105 for the male and female regressions respectively and 0.103 for the pooled (overall) sample. The values for the R- squared means that the model explains approximately 12%, 11% and 10% of the male, female and total wealth respectively. For the pooled sample, the wealth of women is found to be significantly less than that of men (reference group) by about 16%. This clearly indicates that relative to the uncontrolled wealth model (which reports a wealth difference of about 18% in favour of the men), the wealth difference between men and women is higher for the controlled wealth model or function. This implies the importance of other variables in the determination of the wealth differences between men and women. University of Ghana http://ugspace.ug.edu.gh 56 From the table, it is also clear that age is significant in explaining the level of women’s wealth and the total sample but not significant for men. For women, there is a significantly positive relationship between age and wealth level, where an increase in age increases the level of wealth by about 2% for women. The wealth of an individual asset owner is also reported to increase by about 0.7% with an increase in age as indicated for the pooled sample. Marital status (married and previously married (separated)) were both significant in determining the wealth level for men and the overall sample. A coefficient of 0.988 and 0.924 indicates that, relative to the never married (reference class), the married man and the separated accumulates about 99% and 92% respectively more in wealth. This could probably be attributed to the many financial and social responsibilities associated with marriage. Intuitively, this could suggest that the married man with relatively high sense of responsibility develops the attitude of asset acquisition, savings, among others which shapes his attitude towards asset level, hence wealth. A separated man, once married before and already adapted to the idea of wealth accumulation, also has the potency of accumulating more wealth than the never married, all things being equal. Also for the pooled sample, the wealth level of the married is found to be about 38 percent significantly higher (at 1%) as compared to individual who is not married. Education is also found to significantly influence the wealth level of men, women, and the overall (pooled) sample. At secondary level of education, the wealth level of men is found to significantly (at 10%) higher than the uneducated man. A woman with tertiary level of education is found to have about 64% high in wealth level than the uneducated woman. This explains the significant difference in the asset accumulation for educated and uneducated women, and explains the importance of education for the woman. All things being equal, the University of Ghana http://ugspace.ug.edu.gh 57 highly educated woman tends to have significantly different attitude and approaches to issues (financial, decision making, social, etc.). Under education also, for the overall sample, relative to the not educated an individual with tertiary level education is expected to have 36% more in wealth. The economic status of the individual is also reported to be significant in determining the wealth level of women and the pooled sample. According to results, being a domestic/family worker with pay has reducing effect on asset acquisition (wealth level) for both the female and pooled samples compared to not working (reference category). Thus, relative to the “not working”, a domestic/family worker is reported to have lower wealth levels for the female and overall (pooled) sample at 1% level of significance. Also a female who is wage employed is reported to have more asset in wealth relative to the woman who is not working. Wage employment relatively increases the woman’s ability to acquire more assets (wealth) due to the regular labour income. With “no religion” as the reference dummy, all religious dummies except for “other religion” are insignificant in explaining the wealth level of both the female and overall sample. Relative to the reference category (no religion), the “other religion” dummy negatively impact on the wealth levels of women and the pooled sample at 1% level of significance. University of Ghana http://ugspace.ug.edu.gh 58 Table 4.3: Results of Multiple Linear Regression for Male, Female, and Pooled Samples with lnW as Dependent Variable Predictor variables MALE FEMALE POOLED SAMPLE Coeff. Std. Dev. Coeff. Std. Dev. Coeff. Std. Dev. Sex (Male): Female -0.158** 0.767 Age 0.012*** 0.004 0.007** 0.003 Marital Status (Never Married): Married 0.988*** 0.233 0.382*** 0.144 Separated 0.924*** 0.337 Education (No Education): Secondary 0.403* 0.227 Tertiary 0.636** 0.276 0.361* 0.178 Econ. Stat. (Not Working): Wage employee 0.552* 0.290 Domestic/family worker -0.603*** 0.224 -0.615*** 0.182 Religion (No Religion): Other Religion -3.661*** 1.120 -2.765*** 0.868 Region (Greater Accra): Western -0.622** 0.277 -0.987*** 0.251 -0.807*** 0.186 Central -0.481* 0.260 -0.446** 0.198 Volta -0.853*** 0.298 -0.677** 0.282 -0.761*** 0.204 Eastern -0.305* 0.166 Ashanti -0.807*** 0.252 -0.728*** 0.235 -0.774*** 0.172 University of Ghana http://ugspace.ug.edu.gh 59 Brong Ahafo Northern -0.876** 0.337 -1.385*** 0.309 -1.133*** 0.227 Upper East -0.866** 0.352 -0.806** 0.321 -0.828*** 0.237 Upper West -1.625*** 0.379 -1.585*** 0.361 -1.585*** 0.261 Location (Urban): Rural -0.796*** 0.140 -0.461*** 0.123 -0.594*** 0.092 Ethnicity (Ga-Dangme): Akan 0.416** 0.209 0.382** 0.156 Ewe 0.337* 0.194 Guan 0.760** 0.351 No. of Observations 1986 2231 4217 Prob > F 0.000 0.000 0.000 R-squaed 0.124 0.105 0.103 Adjusted R-squared 0.109 0.091 0.096 Note: ***, **, * signify 1, 5, and 10 percent significance respectively Items in bracket signifies reference dummy Source: Author's Computation University of Ghana http://ugspace.ug.edu.gh 60 Further, results on region indicated that for the male regression, being in the Western region, Volta, Ashanti, Northern, Upper East and Upper West regions have a reducing effect on the wealth level with reference to being in the Greater Accra region of Ghana. All the regional dummies for the male function were highly significant at 1% except for the Western, Northern, and Upper East, which were significant at 5%. Also, all the regional dummies except for the Eastern and Brong Ahafo regions reported significant impact (negatively) on women’s wealth level relative to the reference dummy. The results for the overall sample also showed reducing effect of the significant regional factor on the wealth level of asset owners. This could be attributed to the high economic nature of the Greater Accra region compared to the other region, with most major economic activities taking place in the region. The location (rural and urban) characteristics of the respondents revealed significant influence on the level of wealth for all 3 sample categories. The location variable was found to be highly significant for all samples (male, female and overall) at 1%. At 1% level of significance (for all samples) and with urban as the reference category, being in the rural area reduces one’s level of wealth as indicated in table 4.3 above. Also, some ethnic groups (the ethnicity of some individuals) tend to significantly impact on wealth levels with Ga-Dangme as the reference class, as indicated in table 4.3 above. For the ethnicity variable, with Ga-Dangme as the reference class, Akan was found to significantly influence (positively) the wealth levels for both the female and pooled sample. At 10% level of significance, an Akan and Guan woman is expected to have about 42% and 76% more in wealth compared to the Ga-Dangme woman all things being equal. In general, an Akan and Ewe is University of Ghana http://ugspace.ug.edu.gh 61 reported to hold approximately 38% and 34% more in wealth than the Ga-Dangme all things being equal. 4.2.2 Blinder-Oaxaca Decomposition Technique (Results) The empirical results of the Blinder-Oaxaca decomposition technique is analysed in examining the gender wealth gap in Ghana. The Blinder-Oaxaca model successfully identified some significant variables of the model. The implementation of the Blinder-Oaxaca decomposition technique using STATA (version 13) is presented in table 4.4 followed by critical analysis of the results. The total number of observations is recorded to be 4,217, out of which Group 1, representing the male sex and coded 0 for the sex variable account for 1,986, and the remaining 2,231 were women (Group 2, coded 1 for Sex in the analysis). The table presents the coefficients, the standard errors, the P-values as well as the minimum and maximum values at 95% confidence intervals. To begin with, the Blinder-Oaxaca decomposition results as reported in table 4.4 shows the potential wealth differences between men and women. The results show the predicted mean values of male and female respondents on wealth are all statistically significant at one percent (1%), with the result for men providing about 0.13 high mean value for wealth than women for the estimators. University of Ghana http://ugspace.ug.edu.gh 62 Evidently in the table, about 21.13% of the differences in mean wealth reported by men and women is significantly explained or accounted for by the observable characteristics (predictors of the model) – age, marital status, education, economic status, religion, region, location, and ethnicity. The remaining 78.72% of the variations in mean wealth between men and women are unaccounted for (unexplained) by the model, albeit insignificant. This means that while men provided higher mean values of assets (wealth) than women, the differences are primarily unaccounted for by the observable characteristics (predictors) of the model as outlined in equation (3.3). The result indicates that the marital status, religion, region, and ethnicity of the asset owners have a reducing effect on the gender wealth gap as indicated by the negative coefficients. However, these variables are all not significant (p-values greater than 10%) in explaining the 21.13% difference in the differences in the mean wealth between men and women. The age, educational levels, economic status, and location of the asset owners are however reported to be significant in explaining the gender wealth gap, with age and location having a reducing effect. The negative coefficients in the Blinder-Oaxaca Decomposition result is confirmed by the work of Jann (2008) which produced a negative result for the isco variable under both the explained and unexplained difference, and a negative for education under the unexplained differences in explaining the detailed decomposition. However, the educational level and the economic status of the individual asset owner reveals a positive influence (widening effect) on the wealth differences between men and women. University of Ghana http://ugspace.ug.edu.gh 63 In line with expectations, age plays a significant role in explaining the wealth gap between men and women. Age is found to be statistically significant (at 5 percent) in influencing the gender wealth gap. Age has the effect of bridging the wealth gap between men and women as it explains about 82% of the difference in gender mean wealth in the estimated model. Again, in line with apriori assumption, education is also found to be highly significant at 1 percent and explains approximately 196 percent of the mean difference in wealth between men and women. As indicative in the summary statistics on education above, there are more men than women in higher levels of education. The male sex dominates in the secondary and tertiary levels of education. Higher levels of education (increase in human capital) increases the chances of landing a decent and better paid job, all things being equal. This translates in the ability to acquire more assets, hence wealth level, ceteris paribus. In line with the findings of the study, Blinder (1973) in examining the differences in the gender pay gap which has a direct relationship with wealth all things being equal, also found that men are at a clear advantage with respect to education and local labour market conditions, hence enjoying much higher returns to these characteristic than their female comparators. The descriptive analysis clearly supports the findings on education as there are more men than women who are engaged in relatively high paying jobs (i.e. wage employment, large scale businesses, etc.). According to Langford (1995), women’s under investment in education (physical trade and sciences) severely acts to their disadvantage. Blinder (1973) also find that while men and women had the same average endowments of these factors, men received greater returns for education and were less affected by the local labour market conditions. University of Ghana http://ugspace.ug.edu.gh 64 10 ***, **, * signify 1, 5, and 10 percent significance Table 4.4: Blinder-Oaxaca Decomposition Result10 lnW Coef. Standard Error Overall Group 1=Men 6.1479*** 0.0565 Group 2=Women 6.0200*** 0.0508 Difference 0.1278* 0.0760 Explained 0.0272*** (21.13%) 0.0333 Unexplained 0.1006 (78.72%) 0.0782 Explained Age -0.0223* (82.10%) 0.0116 Marital Status -0.0020 0.0268 Education 0.0533*** (195.93%) 0.0180 Economic Status 0.0327*** (120.02%) 0.0120 Religon -0.0018 0.0027 Region -0.0064 0.0051 Location -0.0193* (40.25%) 0.0099 Ethnicity -0.0110 0.0070 Unexplained Age -0.3849** (376.09%) 0.1835 Marital Status 0.2703*** (277.13%) 0.1003 Education -0.0838 0.1110 Economic Status 0.1062 0.1439 Religon 0.1824 0.3414 Region -0.1405 0.1840 Location -0.4143 0.2695 Ethnicity -0.0500 0.1405 Constant 0.16276 0.4662 Model = Linear Number of obs. = 4217 Group 1: Sex = 0 Number of obs. 1 = 1986 Group 2: Sex = 1 Number of obs. 2 = 2231 Source: Author's Computation, 2014 University of Ghana http://ugspace.ug.edu.gh 65 Result also revealed that the economic status has a positive effect on the gender wealth gap with approximately 120% of the gap arising from differences in the economic status of women and men. The variable was also highly significant (1 percent) in explaining the differences in mean wealth between men and women as indicated in the table. This difference is primarily accounted for by the large difference in wage employment between men and women as depicted in figure 4.2 above. The differences in wage employment between the sexes could be linked to gender differences in education, especially at higher levels of education (i.e. tertiary). Women generally have little work experience, invest less in human capital and take intermittent breaks from paid work (due to marriage and childbearing) as compared to their male counterparts. According to Gorlich and de Grip (2008), women are more likely to engage in occupation that offer more flexibility and that do not require continual investment in skills unique to a firm, or occupation where skills do not depreciate significantly because of career interruption (intermittent breaks) due to child birth, child bearing and childcare. This explains the relatively less concentration of women in wage employment, hence the reason why women dominated occupations with lower wages. Contrary to the effect of economic status in widening the wealth gap, Biltagy (2014) in estimating the gender wage differential in Egypt finds that the gender gap in wage is due to discrimination against women in the labour market other than low levels of human capital characteristics like lower levels of education or less experience. Men were less affected by the local labour market conditions compared to women. University of Ghana http://ugspace.ug.edu.gh 66 Lastly, the location variable was also found to bridge the differences in wealth between men and women by approximately 40 percent. This could possibly be attributed to the share of women in the urban areas who own asset relative to their male comparators. All things being equal, an asset (particularly immobile assets like place of residence, agricultural land) in the urban centres is expected to command higher price than the same or similar type located in the rural areas. Therefore with more and more women than men owning asset in the urban centres, the probability of a reduced wealth gap is highly expected ceteris paribus. 4.3 Summary In the first part of this chapter, the socio-economic characteristics of the respondents were detailed. The second section looked at the empirical analysis of the model in determining the effects of the predictor variables of the model on the level of male and female wealth, the direction of the gender wealth gap, as well as the composition (explained and unexplained parts) of the wealth differences between men and women. The simple linear regression was employed in determining the direction of the wealth gap, using the male group as the reference category. In determining the directional effects also, separate multiple regression analyses were used in this regard. Lastly, the Blinder-Oaxaca decomposition technique was employed to determine the difference and the portion of the difference that is explained by the model and that which is unaccounted for by the predictors of the model. University of Ghana http://ugspace.ug.edu.gh 67 CHAPTER FIVE SUMMARY AND RECOMMENDATIONS 5.0 Introduction This chapter provides a summary of findings and makes constructive policy recommendations based on the findings from the data analysis as presented in the previous chapter. This chapter comprises three sections: summary, recommendation and area for future study. 5.1 Summary Employing the 2010 Ghana Household Asset Survey (GHAS) dataset, this study sought to examine the gender asset gap in Ghana. The study has examined the following research questions: i. Is the Gender wealth gap biased towards men or women? ii. What are the socioeconomic characteristics of the asset owner? iii. How significant are these socioeconomic characteristics in influencing the wealth levels of men and women? iv. What are the factors that influence the gender wealth gap in Ghana? Gender wealth gap in this study basically looks at the incidence of wealth; thus, the share of wealth owned by women (i.e. the value of women’s assets divided by the total value of assets) for particular assets (i.e. agricultural land, businesses, places of residence, consumer durables, etc.) compared to that owned by men (Oduro et al., 2011). This study however measures the gender wealth gap using the gross value of all physical assets of men and women captured in the 2010 GHAS data. University of Ghana http://ugspace.ug.edu.gh 68 In the empirical analysis, the study employed the simple linear regression analysis in determining the direction of the gender wealth gap which was reported to be in favour of men. This also implies from the analysis that men own more wealth in assets than women. Again, two separate multiple linear regression analyses were implemented to examine the significance of the observable characteristics in determining the wealth level of both men and women. The result from the analysis of the male sample (i.e. male asset owners) reported marital status, region, location (rural-urban) and some ethnicities to be significant in determining the wealth levels of men. For the female analysis, the age variable, economic status, region, location (rural-urban) and some ethnicities significantly impacted on the wealth level of women. The Blinder-Oaxaca decomposition technique was also implemented in examining the explained and unexplained difference in wealth for men and women. Results showed a difference of 0.1278 in mean wealth in favour of the men, out of which about 21.13% was explained by the predictors of the model and the remaining 78.72% was unexplained. As indicated in the Blinder-Oaxaca results, the age, education, economic status, and ethnicity variables were found to significantly (at 10%, 1%, 1%, and 10% respectively) explain the gender wealth gap. Age and ethnicity accounted for 82% and 40% respectively of the explained difference in the mean wealth, while the education and economic status variable accounted for about 196% and 120% respectively of the explained difference in the mean wealth between men and women. University of Ghana http://ugspace.ug.edu.gh 69 5.2 Recommendations Based on the findings, the following policies are recommended: In the case of asset acquisition and ownership, government could provide incentives to encourage women to acquire more property, especially by providing some subsidies or incentives for asset ownership by women. In some jurisdictions like Ecuador, when either a husband or a wife acquires an asset, it automatically belongs to the couple due to the practice of partial community property marital regime. The marital regime practiced in Ghana is the separation of property which treats all property solely acquired by a party regardless of when and how it was acquired, as individually owned. Under this regime, upon dissolution of marriage due to divorce or death, there is no joint property to be distributed unless it is registered as such. This puts women at a disadvantage in the event of separation as they stand to lose their right to claim of assets acquired during marriage due to possible encroachment by extended amily members who in most cases try to claim ownership of the assets of the deceased husband. To ensure equity and certainty in the sharing of property among spouses, government finally approved the “Spousal Bill” in the last quarter of 2013. It is therefore recommended that government ensure massive sensitization of the bill in order to harness its effective implementation through an Act of parliament. The findings from the study show that differences in the education of men and women account for about 30% of the explained differences (64.15%) in the gender wealth gap. This calls for University of Ghana http://ugspace.ug.edu.gh 70 more proactive measures from government, educational agencies, ministries of education and other related bodies in ensuring equality between males and females in education in all respects. The barriers that keep girls out of school are well known and solutions for lifting them exist. Achieving parity in enrollment remains a critical objective and is fundamental to gender equality. However, focusing on access as the primary issue for girls can undervalue the focus of quality and relevance. The focus should be on ensuring a balance between quality and quantity in female education. Government and donor agencies have focused primarily on increasing female access and enrollment, with insufficient attention paid to the quality or relevance of education for girls or their retention and achievement rates. A typical example is the increase in enrolment at the various levels of education, especially the basic level without commensurate infrastructural upgrade or adjustment (Akyeampong, ND). In addressing the issue of gender equality in education, there is the need for interventions to tackle the issue at different levels (micro, meso and macro levels): government, national, regional, sectorial, organizational and personal or individual levels. Again, the issue of gender cuts across all sectors, educational ministries, health sectors, employment ministries, among others. Education alone cannot tackle the major issue surrounding gender equality. It is therefore imperative for sectorial involvement (i.e. health, labour, etc.) in moving forward. Gender equality is not just about numbers. In order to go beyond the “numbers game” of gender parity, there is the need for government, other international bodies and other top stakeholders at the national levels of education to adopt measures to encourage more women to further their University of Ghana http://ugspace.ug.edu.gh 71 studies to enhance their professional careers and aim at attaining higher levels on the academic ladder. For instance, although there are some scholarship schemes aimed at providing opportunities for women to further their studies, there is the need to ensure effective monitoring of existing ones and also create more proactive schemes that would ensure the transition of more women to higher levels of education (especially from secondary to tertiary levels). At the individual level also, women must change their orientation within the schools, work environments and or households to challenge gender roles ascribed by society, while the men also learn to accept and adapt to such changes and providing the needed support in that regard. Education is supposed to be a mechanism of change for enlightenment, open mindset, supporting social, emotional and psychological development and a tool for self-empowerment especially for girls and women. Social mobilization campaigns which aim at sensitizing women’s and girls’ status and value in society, as well as the complementary role of men in the process should also be conducted at the individual levels. We found that more than 100% of the explained differences in the gender wealth gap is due to a difference in the economic status between men and women. This was attributed to differences between men and women in wage employments (public and private), where this gap favours men. Women tend to be highly concentrated in petty trading and engaged in small scale businesses that offer less reward as compared to the male-dominated occupation in the study. Also, career breaks for women that are associated with having children and home duties have been considered one of the main factors in explaining the differences between men and women in this study. According to the theory of human capital, parents invest less human capital in University of Ghana http://ugspace.ug.edu.gh 72 women since as compared to men, they concentrate more on the work done at home than on paid employment. Parents and guardians must be willing to undertake the investment at least until the girl child is 15 years old. In this regard, governments should ensure collaborative efforts between the educational system and the labour market to harness and bridge the pay gap. Governments should adopt policies, and/or enact and enforce existing labour laws that ensure opportunity and pay equity. Proactive and feasible policies that will ensure that women are given the equal opportunities to compete with their male counterparts in higher levels of employment (e.g. in the areas of management, supervision and administration) should be of top priority in bridging the gender inequality in employment. Although there are some employment policies like employment quota that is supposed to encourage increased participation of women in the public sector of employment, there has however not been significant improvement in that regard. There is therefore the need for the government to ensure strict and effective monitoring of these policies. In Ghana for instance, on the issue of women’s representation in parliament, Betty Mould-Iddrisu the former Education Minister remarked that there has not been any significant increase since 1992, in spite of their willingness to offer themselves as candidates. Government could also look at supplementing the economic activities of women by providing some subsidies to women in business. The policies could look at reducing the risks involved in some female-dominated occupations; policies that drive females in small-scale businesses to be University of Ghana http://ugspace.ug.edu.gh 73 able to thrive further into large-scale economic activities. Policies should also be geared towards the establishment of industry-specific courses to provide women and girls with the opportunity to acquire specific skills needed to engage in some occupations and also upgrade their professional skills in order to be able to compete with their male counterparts in most occupations. There is also the need to shy away from cultural norms and beliefs that limits the participation of women in some economic activities and limits the potentials of both men and women by way of ascribing certain activities to only men or women. The feasibility of these recommended policies is, however, dependent on the attitude of women towards education and their work life. Women are encouraged to cultivate the habit of independence by delineating from social norms and perceptions of society that try to relegate them to the confines of the home. Parents should also thrive hard to ensure that their daughters (women) acquire or attain higher levels of education; take the initiative to venture into risky but profitable economic activities, and also aspire for higher positions in life, while taking cognizance of the supplementary role of their male counterparts in development. 5.3 Suggested Area for Future Study Results from the Oaxaca decomposition technique revealed that the model explains only 11.60% of the wealth differences between men and women. This suggests that the predictor variables of the model do not highly account for the gender asset gap. In view of this, further studies could explore the area by looking at other determinants of the gender wealth gap, or even so look at the gender perspective of assets discrimination in Ghana. University of Ghana http://ugspace.ug.edu.gh 74 The study focused on the physical assets in its analysis. Future studies could however look at both physical and financial assets for a more comprehensive study. Also, future studies can focus on the use of other decomposition techniques like the extended Oaxaca (Newman-Oaxaca decompositions) in gender-wealth analysis. University of Ghana http://ugspace.ug.edu.gh 75 REFERENCES Agarwal, B. (1994). A Field of One’s Own: Gender and Land Rights in South Asia (Cambridge: Cambridge University Press. Agarwal, Bina (1997). “’Bargaining’ and gender relations: Within and beyond the household.” Feminist Economics 3(1): 1-51. Akyeampong, K. (no date). “Educational Expansion and Access in Ghana-A Review of 50years of Challenge and Progress.” Centre for International Education, University of Sussex, United Kingdom. Baah-Boateng, W. (2012). “Gender perspective of the Ghanaian labour market discrimination in Ghana.” LAP LAMBERT Academic Publishing. Bardasi, E. and Gornick, J.C. (2008). “Working for Less? Women’s Part-Time Wage Penalties across Countries.” Feminist Economics. 14(1): 37-72. Bebbington, A. (1999). Capitals and capabilities: A framework for analyzing peasant viability, rural livelihoods and poverty. World Development 27 (12): 2021–2044. Berger, E. and Denton, M. (2004). “The Interplay between Women’s Work Histories and Financial Planning for Later Life.” Canadian Journal on Aging, 23(Supplement1): 99- 113. Betty MouldI-ddrisu. “Towards increased women’s participation and representation in parliament”. Black, D.A. (1995). Discrimination in an equilibrium search model. Journal of Labour Economics, 13(2), 309-334. University of Ghana http://ugspace.ug.edu.gh 76 Blau, F.D. and Kahn, L.M. (1997). “Swimming Upstream: Trends in the Gender Wage Differential in the 1980s.” Journal of Labor Economics, 15: 1-42. Blau, F.D. and Kahn, L.M. (2000). “Gender differences in pay.” Journal of Economic Perspectives, 14(4), 75-99. Blinder, A.S. (1973). “Wage discrimination: Reduced form and structural estimates.” Journal of Human Resources, 8(4), 436-455. Çağatay N., (2001). Trade, Gender and Poverty. United Nation (UNDP). Cameron, C., & Trivedi, P. K. (2009). Microeconometrics using Stata. Stata Press. Chauvin, K.W. and Ash, R.A. (1994). “Gender earnings differentials in total pay, base pay, and contingent pay." Industrial and Labour Relations Review, 47(4), 634-648. Coulombe, H. and Q. Wodon (2007). “Poverty, Livelihoods, and Access to Basic Services in Ghana CEM: Meeting the Challenge of Accelerated and Shared Growth.” The World Bank, Washington D.C. Deere, C.D., Oduro, A.D., Swaminathan, H., and Doss, C. (2013). “Property Rights and the Gender Distribution of Wealth in Ecuador, Ghana and India." Working Paper Series: No. 13. Deere, C. D. and M. Leon. 2003. “The gender asset gap: Land in Latin America.” World Development 31 (6): 925–947. Deere, C. D., and Doss, C. (2006). “The gender asset gap: What do we know and why does it matter?” Feminist Economics 12(1–2): 1–50. Deere, C.D, and Doss, C.R. (2006). “Gender and Distribution of Wealth in Developing Countries.” Research Paper No. 2006/115. Katajanokanlaituri: UNU-WIDER. University of Ghana http://ugspace.ug.edu.gh 77 Dixon, S. (1996a). The distribution of earnings in New Zealand 1984-94.Labour Market Bulletin 1996:1, 45-100. Doss, C.R., Grown, C., and Deere, C. (2008). “Gender and Asset Ownership: A Guide to Collecting Individual Level Data.” Washington DC: World Bank. Fairlie, R.W. (2003). An Extension of the Blinder-Oaxaca Decomposition Technique to Logit and Probit Models. Yale University. Center Discussion Paper No. 873 FAO (Food and Agricultural Organization). 2004. “A Gender Perspective on Land Rights. Gender and Development Service Sustainable Development.” Department of Food and Agriculture Organization of the United Nations. Fortin, N.M. (2006). Greed, Altruism, and the Gender Wage Gap. Unpublished manuscript, University of British Columbia. Fortin, N.M. and Lemieux, T. (1998). “Rank regressions, wage distributions, and the gender gap.” Journal of Human Resources, 33(3), 610-643. Friedman, M. (1957): A Theory of the Consumption Function. Princeton University Press. Princeton. Friedman, M., & Savage, L.J. (1984). “The utility analysis of choice involving risk.” The Journal of Political Economy, Vol. 56, No. 4, pp. 270-304. Fryer, R., and Levitt, S. (2004). Understanding the black–white test score gap in the first two years of school. Review of Economics and Statistics 86 (2), 447–464. Fryer, Roland, and Levitt, Steven. (2004).Understanding the Black-White Test Score Gap in the First Two Years of School. Review of Economics and Statistics 86: 447.464. Groshen, E.L. (1991). “The structure of the female/male wage differential: Is it who you are, what you do, or where you work?” Journal of Human Resources, 26(3), 457-472. University of Ghana http://ugspace.ug.edu.gh 78 Gupta, N.D., Oaxaca, R.L., and Smith, N. (1998). “Wage dispersion, public sector wages and the stagnating Danish gender wage gap.” Working Paper 98-18, Centre for Labour Market and Social Research: Denmark. Hausmann, R., Tyson, L., and Zahidi, S. (2010). The global gender gap report. Geneva: World Economic Forum. htpp://www.investopedia.com http://dictionary.reference.com/browse/human+capital http://en.wikipedia.org/wiki/Natural_capital http://www.fao.org/docrep/007/y3495e/y3495e00.htm. Last accessed on 6 January2012. http://www.iisd.org/natres/agriculture/capital.asp http://www.investordictionary.com/definition/physical-capital http://www.wisegeek.com/what-is-physical-capital.htm Ibnouf F. O., (2009). The Role of Women in Providing and Improving Household Food Security in Sudan; Implications for Reducing Hunger and Malnutrition. 10(May): 144-167. ICRW (2006). “Property Ownership and Inheritance Rights of Women for Social Protection- The South Asia Experience.” Synthesis Report of Three Studies. Jann, B. (2008). The Blinder-Oaxaca decomposition for linear regression models. The Stata Journal 8(4): 453-479. Jefferson, T., and Alison, P. (2005). “Baby boomers and Australia’s other gender wage gap.” Feminist Economics, 11(2):79-100. Jianakoplos, N.A. and Bernasek, A. (1998). “Are Women More Risk Averse?” Economic Inquiry, 36: 620-30. University of Ghana http://ugspace.ug.edu.gh 79 Jones, F.L. (1983). “On decomposing the wage gap: A critical comment on Blinder's method.” Journal of Human Resources, 18(1), 126-130. Karnataka, R., & Swaminathan, H. (2012). Women’s Property, Mobility, and Decision making. (June). Lambert, S.F. (1993). “Labour market experience in female wage equations: does the experience measure matter?” Applied Economics, 25, 1439-1449. Langford, M.S. (1995) `The Gender Wage Gap in the 1990s', Australian Economic Papers, 34: 62-85. Levi, M.D. (1973). “Errors in the variables bias in the presence of correctly measured variables.” Econometrica, 41(5), 985-986. Lindsey, L.L. (2007). Gender Roles: A sociological perspective (5th edition). Pearson Marilyn, P. (2004). “Social provisioning as a starting point for Feminist Economics. Feminist Economics.” Vol. 10, No. 2, 2004, Routledge Modigliani, F. and Brumberg, R. (1954): “Utility analysis and the consumption function: An interpretation of cross-section data.” In Kurihara, K. (Ed.): Post Keynesian Economics, Rutgers Neal, D., Johnson, W. (1996). The role of premarket factors in black–white wage differences. Journal of Political Economy 104 (5), 869–895. Nielsen, H.S. (1998). “Two notes on discrimination and decomposition.” Working Paper 98-01, Centre for Labour Market and Social Research: Denmark. Njuki, J., Kaaria, S., Sanginga, P., Chamunorwa, A., and Chiuri, W. (2011). Linking smallholder farmers to markets, gender and intra-household dynamics: Does the choice of commodity matter? European Journal of Development Research 23: 426–443. University of Ghana http://ugspace.ug.edu.gh 80 O’Neill, J. (2003). “The Gender Gap in Wages, circa 2000.” American Economic Review, 93(2): 309-14. Oaxaca, R. (1973). “Male-female wage differentials in urban labour markets.” International Economic Review, 14(3), 693-709. Oaxaca, R.L. and Ransom, M.R. (1994). “On discrimination and the decomposition of wage differentials.” Journal of Econometrics, 61, 5-21. Oaxaca, R.L. and Ransom, M.R. (1997). “Identification in detailed wage decomposition.” Working Paper 97-12, Centre for Labour Market and Social Research: Denmark. Oduro et al. (2010). “Asset accumulation among women in Ghana: Understanding the process.” Working Paper Series: No. 4. Oduro et al. (2011). Measuring the gender asset gap in Ghana. Department of economics, University of Ghana: Legon. Oduro, A. D., Baah-Boateng, W. & Boakye-Yiadom, L. (2011). Measuring the gender asset gap in Ghana. Accra: Woeli Publishing House and University of Ghana. Panda, P. and B. Agarwal (2005). “Marital Violence, Human Development, and Women’s Property Status in India.” World Development, Vol.33, No.5, pp.823-50. Quisumbing, A.R., & Maluccio, J.A. (1999). “Policy Research Report on Intrahousehold Allocation and Gender Relations: New Empirical Evidence. Rustagi, P., and Menon, R. (2010). “Women’s command over assets: Addressing gender inequalities.” Asia-Pacific Human Development Report Background Papers Series Seguino, S. and M. Floro. 2003. “Does Gender Matter for Aggregate Saving? An Empirical Analysis. “International Review of Applied Economics 17(2): 147-66. University of Ghana http://ugspace.ug.edu.gh 81 Sibbons, M., Swamfield, D., Poulsen, H., Giggard, A., Norton, A., & Seel, A. (2000). Mainstreaming gender through sector wide approaches in education: Synthesis report. London: Overseas Development Institute/Cambridge Education Consultants. Sierminska, E., A. Brandolini and T.M. Smeeding (2006): The Luxembourg Wealth Study – A Cross-Country Database for Household Wealth Research. Journal of Economic Inequality4(3). Sierminska, Eva M.; Frick, Joachim R.; Grabka, Markus M. (2010). “Examining the gender wealth gap.” Oxford Economic Papers, Vol. 62, Iss. 4, pp. 669-690. SOFA Team (2011). Gender Differences in Assets. ESA Working Paper no? , Agricultural Development Economics Division. The Food and Agriculture Organization of the United Nations, Rome, Italy. SOFA Team and Cheryl Doss. 2011. The Role of Women in Agriculture. ESA Working Paper no? Agricultural Development Economics Division. The Food and Agriculture Organization of the United Nations, Rome, Italy. Strassman, D, (1999). Feminist Economics. The Elgar Companion to Feminist Economics (eds). Janice Peterson and Margaret Lewis, Northampton, Maine. Swaffield, J. (2000). Gender, motivation, experience and wages. Discussion Paper 0457, Centre for Economic Performance: London School of Economics and Political Science. United Nations, (2008). The Millennium Development Goals Report 2008. United Nations, New York. Waldfogel, J. (1998). Understanding the "family gap" in pay for women with children. Journal of Economic Perspectives, 12(1), 137-156. University of Ghana http://ugspace.ug.edu.gh 82 Warren, T. (2006).”Moving beyond the gender wealth gap: On gender, class, ethnicity, and wealth inequalities in the United Kingdom.” Feminist Economics, 12(1&2): 195 – 219. Warren, T.; Rowlandson, K. and Whyley, C. (2001).Female finances: Gender Wage Gaps and Gender Assets Gaps. Work. Employment and Society, 15: 465-488. World Bank (2003). Gender Equality and the Millennium Development Goals. Washington DC, World Bank. World Bank Policy Report (N.D). Engendering Development. Yamokoski, A., and Keister, L.A. (2006). “The wealth of single women: Marital status and parenthood in the asset accumulation of young baby boomers in the United States.” Feminist Economics, 12(1), 167-194 University of Ghana http://ugspace.ug.edu.gh 83 APPENDIX Appendix A: Tables of Summary Statistics from STATA Age 4929 27.68553 21.32777 0 105 Variable Obs Mean Std. Dev. Min Max . summarize Age Total 4,936 100.00 female 2,593 52.53 100.00 male 2,343 47.47 47.47 sex Freq. Percent Cum. . tab Sex lnW 4936 6.037241 2.448178 -.6865026 13.87609 W 4936 6241.421 32190.28 .5033333 1062450 Variable Obs Mean Std. Dev. Min Max . summarize W lnW University of Ghana http://ugspace.ug.edu.gh 84 Total 4,672 100.00 Tertiary 266 5.69 100.00 Secondary 319 6.83 94.31 Basic 2,676 57.28 87.48 None 1,411 30.20 30.20 Edu Freq. Percent Cum. . tab Edu Total 4,925 100.00 Separated 491 9.97 100.00 Married 1,721 34.94 90.03 Never_Married 2,713 55.09 55.09 Mar Freq. Percent Cum. . tab Mar Total 4,926 100.00 Other_religion 8 0.16 100.00 Tradition 238 4.83 99.84 Muslim 983 19.96 95.01 Christian 3,591 72.90 75.05 No_Religion 106 2.15 2.15 Rel Freq. Percent Cum. . tab Rel Total 4,312 100.00 Apprentice 39 0.90 100.00 Domestic/Family_worker 230 5.33 99.10 Casual Labour 26 0.60 93.76 Self-employed 1,632 37.85 93.16 Wage employee 309 7.17 55.31 No_Working 2,076 48.14 48.14 Ec Freq. Percent Cum. University of Ghana http://ugspace.ug.edu.gh 85 Total 4,936 100.00 Rural 3,284 66.53 100.00 Urban 1,652 33.47 33.47 status Freq. Percent Cum. rural Urban - . tab Loc Total 4,936 100.00 Upperwest 246 4.98 100.00 Uppereast 353 7.15 95.02 Northern 720 14.59 87.86 Brongahafo 444 9.00 73.28 Ashanti 727 14.73 64.28 Eastern 581 11.77 49.55 Volta 504 10.21 37.78 Gtccra 504 10.21 27.57 Central 370 7.50 17.36 Western 487 9.87 9.87 Region Freq. Percent Cum. . tab Reg Total 4,936 100.00 Rural 3,284 66.53 100.00 Urban 1,652 33.47 33.47 status Freq. Percent Cum. rural Urban - . tab Loc Total 4,935 100.00 Other-Tribes 147 2.98 100.00 Mande 36 0.73 97.02 Grusi 16 .38 96.29 Mole-Dagbani 822 16.66 91.91 Gurma 515 10.44 75.26 Guan 149 3.02 64.82 Ewe 591 11.98 61.80 Ga-Dangme 417 8.45 9.83 Akan 2,042 41.38 41.38 Ethn Freq. Percent Cum. . tab Ethn University of Ghana http://ugspace.ug.edu.gh 86 Appendix B: Regression Results using STATA i. Results on t-test for Significance of Differences in Mean Wealth between Men and Women (Male and Female) ii. Simple Linear Regression Result (lnW =f (Sex)) Pr(T < t) = 0.9719 Pr(|T| > |t|) = 0.0562 Pr(T > t) = 0.0281 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Ho: diff = 0 degrees of freedom = 4934 diff = mean(male) - mean(female) t = 1.9103 d 1752.296 917.2938 -46.00827 3550.6 combined 4936 6241.421 458.1812 32190.28 5343.182 7139.66 female 2593 5409.648 495.7162 25242.62 4437.608 6381.688 male 2343 7161.944 793.8793 38427.36 5605.165 8718.723 Group Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] Two-sample t test with equal variances . ttest W, by(Sex) _cons 6.1302 .0505496 121.27 0.000 6.0311 6.229299 Sex -.1769542 .0697436 -2.54 0.011 -.3136826 -.0402258 lnW Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 29578.3036 4935 5.99357723 Root MSE = 2.4468 Adj R-squared = 0.0011 Residual 29539.7628 4934 5.9869807 R-squared = 0.0013 Model 38.5408483 1 38.5408483 Prob > F = 0.0112 F( 1, 4934) = 6.44 Source SS df MS Number of obs = 4936 . reg lnW Sex University of Ghana http://ugspace.ug.edu.gh 87 iii. Multiple Linear Regression Results (Male Sample) _cons 7.070027 .4172485 16.94 0.000 6.251728 7.888326 _IEthn_9 .4935171 .4491672 1.10 0.272 -.3873806 1.374415 _IEthn_8 .0826618 .6900016 0.12 0.905 -1.270556 1.435879 _IEthn_7 -.276021 .3765936 -0.73 0.464 -1.014589 .4625469 _IEthn_6 -.4722158 .3434327 -1.37 0.169 -1.145749 .2013176 _IEthn_5 .276532 .3308465 0.84 0.403 -.3723176 .9253816 _IEthn_4 -.1228373 .3764111 -0.33 0.744 -.8610472 .6153726 _IEthn_3 .4435219 .2799707 1.58 0.113 -.1055511 .9925948 _IEthn_1 .356129 .2340257 1.52 0.128 -.1028374 .8150955 _ILoc_2 -.795603 .1394406 -5.71 0.000 -1.069071 -.5221347 _IReg_10 -1.625083 .3790095 -4.29 0.000 -2.368389 -.8817774 _IReg_9 -.86633 .3515158 -2.46 0.014 -1.555716 -.1769441 _IReg_8 -.875889 .3361753 -2.61 0.009 -1.535189 -.2165887 _IReg_7 -.0571372 .2937092 -0.19 0.846 -.6331537 .5188793 _IReg_6 -.8072098 .2519423 -3.20 0.001 -1.301314 -.3131056 _IReg_5 -.3759858 .2546698 -1.48 0.140 -.875439 .1234675 _IReg_4 -.8528478 .2974706 -2.87 0.004 -1.436241 -.2694544 _IReg_2 -.4013456 .3056451 -1.31 0.189 -1.000771 .1980795 _IReg_1 -.6218663 .2771789 -2.24 0.025 -1.165464 -.0782685 _IRel_5 -1.217229 1.414647 -0.86 0.390 -3.991606 1.557148 _IRel_4 -.4240899 .4026542 -1.05 0.292 -1.213767 .3655875 _IRel_3 -.1663201 .3663975 -0.45 0.650 -.8848917 .5522514 _IRel_2 -.1982609 .3093562 -0.64 0.522 -.8049641 .4084422 _IEc_6 .1616123 .72557 0.22 0.824 -1.261361 1.584586 _IEc_5 -.2852782 .3472712 -0.82 0.411 -.9663395 .3957832 _IEc_4 -.2987385 .6249034 -0.48 0.633 -1.524287 .9268096 _IEc_3 -.1755245 .1885423 -0.93 0.352 -.5452899 .1942409 _IEc_2 -.3264823 .2280987 -1.43 0.152 -.773825 .1208604 _IEdu_3 .1369556 .2434673 0.56 0.574 -.3405276 .6144387 _IEdu_2 .4029665 .2272884 1.77 0.076 -.042787 .84872 _IEdu_1 -.0998556 .154205 -0.65 0.517 -.4022794 .2025683 _IMar_2 .9238982 .3371423 2.74 0.006 .2627015 1.585095 _IMar_1 .9876186 .2322977 4.25 0.000 .532041 1.443196 Age -.0018059 .0046131 -0.39 0.695 -.010853 .0072412 lnW Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 12546.5855 1985 6.32069797 Root MSE = 2.3731 Adj R-squared = 0.1090 Residual 10993.2722 1952 5.63179926 R-squared = 0.1238 Model 1553.31331 33 47.0701002 Prob > F = 0.0000 F( 33, 1952) = 8.36 Source SS df MS Number of obs = 1986 i.Ethn _IEthn_1-9 (naturally coded; _IEthn_2 omitted) i.Loc _ILoc_1-2 (naturally coded; _ILoc_1 omitted) i.Reg _IReg_1-10 (naturally coded; _IReg_3 omitted) i.Rel _IRel_1-5 (naturally coded; _IRel_1 omitted) i.Ec _IEc_1-6 (naturally coded; _IEc_1 omitted) i.Edu _IEdu_0-3 (naturally coded; _IEdu_0 omitted) i.Mar _IMar_0-2 (naturally coded; _IMar_0 omitted) . xi: reg lnW Age i.Mar i.Edu i.Ec i.Rel i.Reg i.Loc i.Ethn University of Ghana http://ugspace.ug.edu.gh 88 iv. Multiple Linear Regression Results for Female Sample _cons 6.368508 .4901795 12.99 0.000 5.407244 7.329771 _IEthn_9 .2182315 .4017413 0.54 0.587 -.5696011 1.006064 _IEthn_8 -.8209158 .641243 -1.28 0.201 -2.078422 .4365901 _IEthn_7 .0680029 .3685537 0.18 0.854 -.6547471 .790753 _IEthn_6 .050554 .3152132 0.16 0.873 -.567593 .668701 _IEthn_5 -.0380755 .3118631 -0.12 0.903 -.6496529 .5735018 _IEthn_4 .7598439 .3508902 2.17 0.030 .0717326 1.447955 _IEthn_3 .2220838 .2726661 0.81 0.415 -.3126264 .7567941 _IEthn_1 .4156762 .2091163 1.99 0.047 .0055899 .8257625 _ILoc_2 -.4605148 .1225173 -3.76 0.000 -.7007767 -.2202529 _IReg_10 -1.584808 .3612067 -4.39 0.000 -2.29315 -.8764657 _IReg_9 -.8063908 .3212945 -2.51 0.012 -1.436464 -.176318 _IReg_8 -1.385126 .3091569 -4.48 0.000 -1.991396 -.7788554 _IReg_7 -.2445176 .2699931 -0.91 0.365 -.773986 .2849508 _IReg_6 -.7282192 .2349368 -3.10 0.002 -1.188941 -.2674977 _IReg_5 -.2810204 .2195363 -1.28 0.201 -.7115409 .1495001 _IReg_4 -.6769129 .2818093 -2.40 0.016 -1.229553 -.1242724 _IReg_2 -.480868 .2603867 -1.85 0.065 -.9914979 .0297618 _IReg_1 -.9871438 .2509853 -3.93 0.000 -1.479337 -.4949505 _IRel_5 -3.660579 1.120345 -3.27 0.001 -5.857625 -1.463534 _IRel_4 -.0120028 .5020128 -0.02 0.981 -.9964721 .9724665 _IRel_3 .3698187 .4715106 0.78 0.433 -.5548344 1.294472 _IRel_2 .117135 .4317228 0.27 0.786 -.7294927 .9637626 _IEc_6 -.1585668 .4585444 -0.35 0.730 -1.057793 .7406592 _IEc_5 -.60251 .2233818 -2.70 0.007 -1.040572 -.1644485 _IEc_4 -.7782021 .7010928 -1.11 0.267 -2.153076 .596672 _IEc_3 -.0644591 .1513324 -0.43 0.670 -.3612288 .2323105 _IEc_2 .5521771 .2902439 1.90 0.057 -.0170041 1.121358 _IEdu_3 .6357489 .2759422 2.30 0.021 .094614 1.176884 _IEdu_2 -.0674412 .2395779 -0.28 0.778 -.5372642 .4023817 _IEdu_1 -.1520432 .1260692 -1.21 0.228 -.3992706 .0951842 _IMar_2 -.1151049 .24738 -0.47 0.642 -.600228 .3700183 _IMar_1 .0103444 .1899569 0.05 0.957 -.3621695 .3828583 Age .0116565 .0041873 2.78 0.005 .0034449 .019868 lnW Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 12815.1339 2230 5.7466968 Root MSE = 2.2854 Adj R-squared = 0.0911 Residual 11474.9206 2197 5.22299527 R-squared = 0.1046 Model 1340.21327 33 40.6125235 Prob > F = 0.0000 F( 33, 2197) = 7.78 Source SS df MS Number of obs = 2231 i.Ethn _IEthn_1-9 (naturally coded; _IEthn_2 omitted) i.Loc _ILoc_1-2 (naturally coded; _ILoc_1 omitted) i.Reg _IReg_1-10 (naturally coded; _IReg_3 omitted) i.Rel _IRel_1-5 (naturally coded; _IRel_1 omitted) i.Ec _IEc_1-6 (naturally coded; _IEc_1 omitted) i.Edu _IEdu_0-3 (naturally coded; _IEdu_0 omitted) i.Mar _IMar_0-2 (naturally coded; _IMar_0 omitted) . xi: reg lnW Age i.Mar i.Edu i.Ec i.Rel i.Reg i.Loc i.Ethn University of Ghana http://ugspace.ug.edu.gh 89 v. Multiple Linear Regression Results for Pooled (Overall) Sample . _cons 6.849548 .3091852 22.15 0.000 6.24338 7.455715 _IEthn_9 .3294843 .2994179 1.10 0.271 -.2575338 .9165024 _IEthn_8 -.3839557 .4699048 -0.82 0.414 -1.305219 .5373074 _IEthn_7 -.1394148 .2619284 -0.53 0.595 -.6529337 .3741041 _IEthn_6 -.2064049 .2320001 -0.89 0.374 -.6612485 .2484386 _IEthn_5 .1137345 .2264491 0.50 0.616 -.330226 .5576951 _IEthn_4 .2878469 .2563753 1.12 0.262 -.2147849 .7904787 _IEthn_3 .3365295 .1941351 1.73 0.083 -.0440785 .7171375 _IEthn_1 .3822852 .1558822 2.45 0.014 .0766732 .6878972 _ILoc_2 -.5943768 .0919282 -6.47 0.000 -.7746049 -.4141486 _IReg_10 -1.58527 .2608189 -6.08 0.000 -2.096613 -1.073926 _IReg_9 -.8281692 .2364586 -3.50 0.000 -1.291754 -.3645847 _IReg_8 -1.133384 .2272385 -4.99 0.000 -1.578893 -.6878764 _IReg_7 -.144733 .1987346 -0.73 0.466 -.5343584 .2448923 _IReg_6 -.7740268 .17157 -4.51 0.000 -1.110395 -.4376585 _IReg_5 -.3051475 .1663758 -1.83 0.067 -.6313324 .0210374 _IReg_4 -.7611741 .2036413 -3.74 0.000 -1.160419 -.361929 _IReg_2 -.4461063 .1979869 -2.25 0.024 -.8342659 -.0579467 _IReg_1 -.8073446 .1858224 -4.34 0.000 -1.171655 -.443034 _IRel_5 -2.765258 .8681916 -3.19 0.001 -4.467375 -1.063141 _IRel_4 -.3190076 .3071619 -1.04 0.299 -.9212082 .2831931 _IRel_3 -.0068666 .2828261 -0.02 0.981 -.5613561 .547623 _IRel_2 -.1629673 .2478502 -0.66 0.511 -.6488854 .3229509 _IEc_6 -.0345351 .3896005 -0.09 0.929 -.7983591 .7292889 _IEc_5 -.6153315 .1823741 -3.37 0.001 -.9728816 -.2577814 _IEc_4 -.4746044 .4647072 -1.02 0.307 -1.385677 .4364686 _IEc_3 -.0888223 .1164441 -0.76 0.446 -.3171147 .1394701 _IEc_2 .0854273 .1700186 0.50 0.615 -.2478996 .4187542 _IEdu_3 .361026 .1779814 2.03 0.043 .0120879 .7099642 _IEdu_2 .2087187 .1621587 1.29 0.198 -.1091986 .526636 _IEdu_1 -.112412 .0970693 -1.16 0.247 -.3027195 .0778955 _IMar_2 .2684339 .1937697 1.39 0.166 -.1114577 .6483254 _IMar_1 .3821797 .1438233 2.66 0.008 .1002096 .6641499 Age .0066148 .0030576 2.16 0.031 .0006202 .0126093 Sex -.1577264 .0767583 -2.05 0.040 -.3082135 -.0072393 lnW Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 25378.892 4216 6.0196613 Root MSE = 2.3332 Adj R-squared = 0.0957 Residual 22765.3727 4182 5.4436568 R-squared = 0.1030 Model 2613.5193 34 76.8682148 Prob > F = 0.0000 F( 34, 4182) = 14.12 Source SS df MS Number of obs = 4217 i.Ethn _IEthn_1-9 (naturally coded; _IEthn_2 omitted) i.Loc _ILoc_1-2 (naturally coded; _ILoc_1 omitted) i.Reg _IReg_1-10 (naturally coded; _IReg_3 omitted) i.Rel _IRel_1-5 (naturally coded; _IRel_1 omitted) i.Ec _IEc_1-6 (naturally coded; _IEc_1 omitted) i.Edu _IEdu_0-3 (naturally coded; _IEdu_0 omitted) i.Mar _IMar_0-2 (naturally coded; _IMar_0 omitted) . xi: reg lnW Sex Age i.Mar i.Edu i.Ec i.Rel i.Reg i.Loc i.Ethn University of Ghana http://ugspace.ug.edu.gh 90 Appendix C: Blinder-Oaxaca Decomposition Result for Gender Differences in Wealth11 11 Group 1 = Male; Group 2 = Female _cons .7516368 .4662409 1.61 0.107 -.1621785 1.665452 Ethn -.0197602 .1404831 -0.14 0.888 -.295102 .2555815 Loc -.4424032 .2695126 -1.64 0.101 -.9706381 .0858317 Reg -.1297463 .1840401 -0.70 0.481 -.4904584 .2309657 Rel -.0659846 .3413989 -0.19 0.847 -.7351141 .6031449 Ec .2073693 .1438527 1.44 0.149 -.0745768 .4893154 Edu -.1008828 .1110184 -0.91 0.364 -.3184748 .1167093 Mar .2789018 .1003327 2.78 0.005 .0822533 .4755503 Age -.378492 .1835046 -2.06 0.039 -.7381545 -.0188295 unexplained Ethn -.0109496 .0070413 -1.56 0.120 -.0247503 .0028511 Loc -.019249 .0098524 -1.95 0.051 -.0385593 .0000614 Reg -.0063986 .005061 -1.26 0.206 -.0163181 .0035208 Rel -.0017724 .0027381 -0.65 0.517 -.0071391 .0035942 Ec .0326506 .0119709 2.73 0.006 .009188 .0561132 Edu .0533025 .0179414 2.97 0.003 .0181381 .0884669 Mar .0019567 .0268152 0.07 0.942 -.0506001 .0545136 Age -.0223347 .0116196 -1.92 0.055 -.0451086 .0004392 explained unexplained .1006389 .0782375 1.29 0.198 -.0527037 .2539816 explained .0272055 .0333164 0.82 0.414 -.0380935 .0925044 difference .1278444 .0760188 1.68 0.093 -.0211497 .2768385 group_2 6.020063 .050838 118.42 0.000 5.920423 6.119704 group_1 6.147908 .0565187 108.78 0.000 6.037133 6.258682 overall lnW Coef. Std. Err. z P>|z| [95% Conf. Interval] Group 2: Sex = 1 N of obs 2 = 2231 Group 1: Sex = 0 N of obs 1 = 1986 Model = linear Blinder-Oaxaca decomposition Number of obs = 4217 . oaxaca lnW Age Mar Edu Ec Rel Reg Loc Ethn, by(Sex) weight(0) University of Ghana http://ugspace.ug.edu.gh