Gender Analysis of the Ethiopian National Household Surveys

Transcription

1 ETHIOPIA POVERTY AND GENDER UPDATE GROWTH AND GENDER INEQUALITIES IN ETHIOPIA Gender Analysis of the Ethiopian National Household Surveys David Lawson 1 1 October

2 Contents Page Number List of Figures 3 List of Tables 4 Executive Summary 6 1. Background and Rationale 9 2. Typology and Monetary Based Poverty Health, Education and HIV/AIDS Awareness Assets/Household Public Goods Occupation/Activity Types Summary of Findings 41 Appendix One: Detailed Descriptive Tables 43 Appendix Two: References 59 2

3 List of Figures Figure 1 Proportion of Male and Female Headed Households Disaggregated By Quintile and National/Urban/Rural 14 Figure 2: Distribution of FHH Disaggregated by Expenditure Quintile and 15 Region Figure 3: Proportion of Female Headed Households (FHH) for 1995 and 2005, Disaggregated by Quintiles 15 Figure 4: Percent Stunting from 1996 to 2004 at National level, and 21 Rural and Urban Figure 5: Figure 6 Percent Wasting from 1996 to 2004 at National level, and Rural and Urban 21 Percentage Change in Illiteracy Disaggregated by Rural/Urban Male/Female Across Quintiles 25 Figure 7: Illiteracy levels disaggregated by Region and Gender 26 Figure 8: Percentage of Individual Household Members Who Have Completed Upper Levels of Primary Education and Beyond 28 Figure 9: Distribution of Household Head Engaged in Economic Activity 37 Figure 10: Main Types of Economic Activity of Household Members, Disaggregated by Region, and M:F Ratio 39 3

4 List of Tables Table 1: Table 2: Table 3: Table 4: Table 5: Percentage Distribution of Household Members by Household Size, Expenditure Quintile and All/Rural/Urban (2004/5) 11 Percentage Distribution of Population by Broad Age Group, Dependency Ratio and Survey Year 12 Percentage Distribution of Population by Broad Age Group, Disaggregated by Gender of Household Head 13 Distribution of Household Heads for 2005 Disaggregated by Expenditure Quintile and National/Urban/Rural 13 Distribution of Household Heads for 2005 Disaggregated by Expenditure Quintile and Region 14 Table 6: Distribution of Individuals Across Quintiles 16 Table 7: Table 8: Expenditure Quintile of Entire Population Disaggregated by Gender and Age 16 Headcount Poverty (%) Disaggregated by Gender of Household Head 18 Table 9: OLS Regression - Expenditure Per Capita Coefficients 19 Table 10: Prevalence of Wasting, Stunting, and Underweight by Gender, Place of Residence and Over Time 20 Table 11: Determinants of Health of Pre-School Girls and Boys (Wasting) 22 Table 12: Determinants of Health of Pre-School Girls and Boys (Stunting) 23 Table 13: Literacy Rates (%) , Disaggregated by Gender 24 Table 14: Table 15: Table 16: Percentage Illiterate 2005 Household Members Disaggregated by Expenditure Quintile and Gender 24 Illiteracy Levels and Percentage Change Across Expenditure Quintiles Disaggregated by Region and Gender 26 Percentage of All Household Heads 2005 Who Have Completed Up To Final Primary School Levels, Or Above, Disaggregated Expenditure Quintile and Gender 27 4

5 Table 17: Table 18: Percentage of Population Completing Up To 7-8 Years of Primary School or Beyond, Disaggregated by Region and Gender 28 Reason for Drop-Out by Gender - Primary& Secondary School 29 Table 19: Women's participation in decision making 30 Table 20: Table 21: Percentage of Women Who Believe That a Wife Is Justified In Refusing To Have Sexual Intercourse With Her Husband In Specific Circumstances 30 Empowerment - Attitudes Towards Decision Making and Violence (%) 31 Table 22: Employment, Cash Earnings and Decision Power Over Cash Earnings Among Women of Childbearing Age In Ethiopia 32 ` Table 23: Percentage of Households That Own Land and Cattle 34 Table 24: Table 25: Table 26: Household Public Goods (%) Disaggregated by Gender and Marital Status 36 Individual Members Main Activities and Male:Female Ratio for Individual Members Disaggregated Region 38 Type of Employment and Male Female Ratio of Workers, Disaggregated by Quintiles (1995 HICE) 39 Appendix Tables Appendix Table 1: Appendix Table 2: Distribution of Total Expenditure by Source, Type of Indicator and Household Expenditure Quintiles Country, Rural + Urban (Male Headed) 43 Multinomial Logit Regression - Household Expenditure P.C. Disaggregated by Quintile - Marginal Effects 56 Appendix Table 3: Determinants of Breastfeeding 58 5

6 Executive Summary The centrality of addressing gender issues in poverty reduction and development in Africa is being increasingly recognised, with growing evidence showing that the nature, causes and impacts of poverty are different for men and for women. Gender inequality persists in access to and control of a range of productive, human, and social capital asset - consequently, the core components of poverty capability, opportunity, security, and empowerment differ along gender lines. Poverty reduction strategies, therefore, need to address these differences in understanding the constituent elements and dynamics poverty and vulnerability, in prioritising actions, including setting and monitoring performance indicators, and in monitoring implementation. This study undertakes a gender analysis of Ethiopia s main, recent, national household data sets surveys, extracting the most appropriate data from the main sources, including Welfare Monitoring Surveys (WMS) Household Income Consumption Expenditure (HICE) and Demographic Health Surveys (DHS),. We address the gender dimensions of income/consumption-based poverty measures and trends in other key variables. The main focus of the report provides a detailed analysis of the household surveys, looking at the key gender trends both within households and across certain household types. Specifically the section is disaggregated into the following categories: Household Structures, Expenditure Quintiles, Poverty and Expenditure Type Health, Education and HIV/AIDS awareness Household Assets and Public Goods Occupation/Activity Types Household Structures, Expenditure Quintiles, Poverty and Expenditure Type The most commonly observed household size was 6 members, with Urban households being slightly smaller than rural households. Households in the highest quintiles appear to have a higher average household size. Gender differences appear to increase with household expenditure. For example, of the top two quintiles approximately 85.4% of the households are MHH compared with less than 15% being FHH. From a regional perspective, beyond the lowest quintile, there is a fairly dramatic fall in the proportion of households that are FHH. For example in the case of Harari, 68.5% of quintile 1 households are FHH compared with 33.1% of those households in quintile 2. Hence FHH appear to be particularly biased towards being in the lowest quintile of all regions. There appears to be smaller proportions of FHH s in the upper expenditure range in 2005 than 10 years earlier. The poverty profile found suggested there to be no statistical difference between MHH and FHH poverty. 6

7 Health, Education and HIV/AIDS Awareness There is a clear decline in levels of stunting, since 1996, and it is evident that urban children faired better than rural children Wasting has slighting increased for both girls and boys Although increased age of the household head does appear to be associated with less wasting and stunting, this is only statistically significant for boys On average 44.8% (80.9%) of the urban (rural) lowest quintile households are illiterate Relatively large gender differences also appear to exist with 89.3% of rurally based female heads, in the lowest quintile, being illiterate compared with 68.5% of men. Unsurprisingly, illiteracy levels decrease as expenditure quintiles increase. On average 5.8% (46.6%) of MHH rural (urban) based households have completed upper levels of primary education and secondary, compared with 1/7% (24.4%) of FHH s.. For males, one of the main reasons for primary and secondary school drop out is the need to work. With approximately ¼ of all primary aged students dropping out of school because of this reason (30.2% for boys and 18.9% for girls). Increased wealth appears to be associated with significantly more control over the decision making process. For example, approximately 60^ of women in the lowest quintile decide over their own health care compared with 78% of women in the highest quintile. Almost 30% of the poorest women in Ethiopia believe that they do not have the right to refuse sex if their husband has sex with another women. This compares with approximately 10% of the richest. Almost ¾ s of the poorest think it is justified for the beating of wife if she goes out without telling him, compared with 40% of the wealthiest. Household Assets and Public Goods Larger proportions of MHHs than FHH s own farm land, although there are some regional variations to this trend. The use of private pit latrine is more common for MHH s than for FHH s Although very small proportions, a greater percentage of FHH s have access to taps inside their houses or compound. Occupation/Activity Types The ratio of male:females employed dramatically increases, as expenditure quintiles increase, especially in the public sector (self employed formal sector) where the ratio of males to females increases from 1.29 (1.91) to 4.5 (5.27) men to every women employed Of the individuals involved in agriculture the proportion of men and women in the lowest quintile are roughly equal, but the ratio of men to women involved in this type of activity, as the highest expenditure quintile is reached increases to approximately 2/3rds men to one third women. 7

8 Perhaps of greater interest is the male:female activity ratio in the main sectors. The overall average proportion of male to females in agriculture is 1.47 male for every female employed, but this varies significantly across some regions. In Harari for example, the ratio of men to women employed in agriculture (hotels) sector is almost 5 (8) times greater with 7.13 (1.57) males employed for every female. In contrast regions such as Amahara have below average number of male to female workers for manufacturing etc, and hotels and restaurant industry. Overall, although the household analysis has added substantially to our understanding of gender issues, for example, it is one of the first pieces of research which uses statistical household data to look at key gender issues and poverty related dynamics, there have been limitations. In particular the household surveys do not contain information how income and expenditure is allocated within the household, nor who control assets. Futhermore, access to Ethiopian household panel data would also provide opportunities to provide a dynamic poverty analysis 8

9 Gender Analysis of the Ethiopian National Household Surveys 1. Background and Rationale 1.1 Introduction The centrality of addressing gender issues in poverty reduction and development in Africa is being increasingly recognised. For example a recently published World Bank policy research report confirms that gender-based inequality limits economic growth in Africa, and that it is essential for Africa to unleash the enormous productive potential of its women if it is to make impressive strides toward pro-poor growth. It is increasingly recognised that poverty in Africa has strong gender dimensions which need to be addressed explicitly by African countries as they formulate and implement their national poverty reduction strategies and the associated Poverty Reduction Strategy Papers (PRSPs). Developing country evidence shows that the nature, causes and impacts of poverty are different for men and for women. Gender inequality persists in access to and control of a range of productive, human, and social capital asset - consequently, the core components of poverty capability, opportunity, security, and empowerment differ along gender lines. Poverty reduction strategies, therefore, need to address these differences in understanding the constituent elements and dynamics poverty and vulnerability, in prioritising actions, including setting and monitoring performance indicators, and in monitoring implementation. From an Ethiopian specific perspective there is evidence that over the period of implementation of Ethiopia s first PRSP, poverty declined by 6 percentage points to 38 percent. Poverty remains concentrated in rural areas where the majority of the population lives, though progress in poverty reduction has been slower in urban areas. To date there has been little analysis of these trends, and no direct assessment of the differential impact of growth and poverty reduction on women and men. To understand the challenge that gender inequality poses to poverty reduction and identify policies to address them, it is key to assess how women have fared compared to men, given different asset holdings (human and physical), norms and expectations, regional and urban-rural differences (for example fertility rates are much lower in Addis), the segmented nature of labor markets, the sectoral specialization by gender, and so on. Despite the difficulties and data limitations associated with this analysis, much can be learnt by combining insights from the rich sources of nationally based information that are available in Ethiopia (such as consumption and expenditure, and demographic health surveys). 1.2 Ethiopian National Household Surveys Ethiopia has a relatively rich source of data upon which economic analysis can be based. From a purely micro perspective there have been two Demographic Household Surveys (DHS), in the last 7 years, each comprising 5,000 households. Furthermore, there have 9

10 been a series of household surveys since 1981, the most useful of which, for robust data analysis, are the two nationally representative Welfare Monitoring Surveys (WMS) and 2004, combined with information from the Household Income Surveys which are based on representative sub-samples of the aforementioned. The HICE survey has been conducted together with the WM survey every four-five years since 1995/96. The latest of these HICE surveys is for 2004/5 and covered a representative sample of 21,600 households. Previous HICE were similarly representative, covered 17,332 households for 1999/00. All of the Ethiopian nationally representative surveys WMS and HICE adopted two-stage stratified random sampling and provide a rich source of information on socio economic, consumption data and community data. 1.3 This Report - Outline In order to address the main gender based questions highlighted above, ideally a report of this nature would adopt a main focus of investigating intra-household gender relations. Therefore, whenever the household data allows, we adopt an intra household approach but also complement this by looking at male and female headed households (i.e. across household types). From a methodological perspective, when disaggregated, categories such as these form groups which are of interest and are sensible from sociological perspective i.e. widows. By using data contained in the surveys, we get insights that households headed by, for example, widowed and divorced women are poorer in a number of dimensions. As such a more sophisticated and differentiated analysis of the households by household head types gives us some information about gender relations within society. The desired outputs will principally focus around the following areas, with a summary of the desired outputs being provided at the start of each section: 1) Typology and Monetary Based Poverty, 2) Health, Education and HIV/AIDS Awareness, 3) Consumption Data, 4) Household Assets and Public Goods 5) Occupation/Activity Types 6) Income Sources. By using a combination of all the household data sets, and by reviewing previously published (household data based) work on gender issues in Ethiopia, this report adds substantially to the understanding of gender based issues, such as poverty and vulnerability. 10

11 2. Household Structures, Expenditure Quintiles, Poverty and Expenditure Type 2.1 Household Structures Introducing this gender based analysis, we first focus on some simple demographic trends before reviewing household welfare level; by considering household expenditure, and latterly some poverty incidence considerations. All three of the major data sources - WMS, HICE, DHS series of surveys contain data on personal characteristics, such as marital status, sex, education, health information. This data will be used to produce cross tabulated output, which will be disaggregated by gender of household head and across household members. 2 Specifically, in this section, descriptive data will also be produced which ranks households by income levels (e.g. by quintiles). We can then see if, for example, Female Headed Households (FHH), are over represented in the lowest income categories, or below the poverty line, relative to Male Headed Households (MHH). We start the analysis with population/typology descriptive data. From the HICE 2004/5 dat we know that the total population of Ethiopia, increased from 56 million in 1999/00 to 64.5 million. 3 There are roughly equally proportions of males and females across both periods with the rural/urban population split being 55.3/9.1 million. The most commonly observed household size was 6 members, with Urban households being slightly smaller than rural households. Prior HICE reports /96 and 1999/00 identified the national average household size have decreased from 5 (1995/96) to 4.9 (1999/00), figures in Table 1 indicate that this has now reduced to 4.8. Considering the urban/rural dimension the average household size for urban (rural) based households has decreased from 4.7 (5.1) in 1995 to 4.3 (4.9) in Table 1: Percentage Distribution of Household Members by Household Size, Expenditure Quintile and All/Rural/Urban (2004/5) Exp. Place of Household Size (No. of Individuals) Qui. Residence All Rural Urban All Rural Urban All Rural Urban All Rural Urban It is important to recognise the weakness of using household head as a gender based instrument of analysis. It is for this reason that wherever possible we complement such analysis with individual based descriptive/econometric data. 3 Excluding Gambella figures based on HICE 1995 report. 11

12 5 All Rural Urban TOTAL All Rural Urban (Source: Authors calculations, for HICE Report 2007). We can also see form Table 1, that households in the highest quintiles appear to have a higher average household size. For example, the most common household size in the lowest quintile is 4 persons [21.7% of individuals in the first quintile live in households with 4 individuals, compared with 10.9% of individuals in the highest quintile]. This associated increase in household size, with increase in expenditure, appears to be relatively smooth across the quintiles. For example, by looking at the middle income quintiles (3 to 4) the median household size increases, from 4 to 6 individuals. From a regional perspective the HICE (2007) report highlights that Oromiya has the highest average household size with 5.2 persons per household, while the average household size observed in the rest of the regions range from 4.3 persons (in Harari) to 4.9 persons per household observed in Addis Ababa. Table 2 shows the composition of Ethiopian households and how this varies by national/rural/urban and changed across the last decade. Interestingly we see that the proportion of children that comprise rural household has gradually increased, although marginally, from 47.5% to 49%. This has resulted an increase in the dependency ratio from to i.e. there are more dependabts in the household than working aged individuals. 5 Contrary to this the urban dependency ratio has fallen. Table 2: Percentage Distribution of Population by Broad Age Group, Dependency Ratio and Survey Year 1995/6 1999/0 2004/5 All Urban Rural All Urban Rural All Urban Rural Children (0-14 years) Working Age (15-64) Old Age (>64 years) Dependency Ratio (Source: Authors Calculations abridged from HICE, various yrs.) If we disaggregated the composition by gender of household head, perhaps surprisingly we find relatively little difference across Male Headed Households (MHH) and Female Headed Household (FHH), with only a slight higher dependency ratio for MHH. This is in contrast to many other SSA countries, particularly those affected by HIV/AIDS where FHH s tend to be more vulnerable to lower welfare levels and poverty, because of heightened responsibility for orphaned children. 5 Dependency ratio is calculated by dividing the number of working aged by non-working aged individuals. 12

13 Table 3: Percentage Distribution of Population by Broad Age Group, Disaggregated by Gender of Household Head Male Headed Households Female Headed Households All Urban Rural All Urban Rural Children (0-14 years) 47.5% 37.0% 48.9% 43.9% 34.3% 47.2% Working Age (15-64) 49.6% 60.4% 48.2% 48.8% 61.3% 47.7% Old Age (>64 years) 2.9% 2.7% 2.9% 4.9% 5.1% 5.1% Dependency Ratio (Source: Authors Calculations HICE 2005) From a gender perspective we can also see from Table 4, and Figure 1, below that there appears to be significant gender household head differences across the expenditure quintiles. 6 For the population in quintile 1 roughly equal proportions are MHH and FHH. However the gender differences appear to increase with household expenditure. For example, of the top two quintiles approximately 85.4% of the households are MHH compared with less than 15% being FHH. The trends across the urban and rural areas tend to corroborate the aforementioned observation, although but with further accentuation of the gender based transition across quintiles. In the case of urban based households more than 60% of the lowest quintile households are FHH compared with one quarter of households in the upper quintile. The differences are even more extreme for rural households with only 11% of the households in the top quintile being FHH s, and 47.4% in the bottom quintile. Table 4: Distribution of Household Heads for 2005 Disaggregated by Expenditure Quintile and National/Urban/Rural Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Country MHH 50.5% 72.1% 79.2% 85.4% 85.4% FHH 49.5% 27.9% 20.8% 14.6% 14.6% Urban MHH 39.3% 55.0% 63.4% 67.0% 74.7% FHH 60.7% 45.0% 36.6% 33.0% 25.3% Rural MHH 52.6% 74.6% 81.5% 88.2% 89.0% FHH 47.4% 25.4% 18.5% 11.8% 11.0% (Source: Authors calculations, for HICE Report 2007). 6 The authors analysis formed the basis of Ethiopia s HICE 2004/5 and as such some tables are reproduced for this report. 13

14 Figure 1: Proportion of MHH and FHH Disaggregated by Welfare Q1&Q5 and National/Urban/Rural 100.0% 90.0% 80.0% 70.0% % 60.0% 50.0% 40.0% Quintile 1 Quintile % 20.0% 10.0% 0.0% national national Urban Urban rural rural MHH FHH MHH FHH MHH FHH National/Urban/Rural From a regional perspective, we c an see in Table 4 that F igure 2, beyond the lowest quintile, t here is a fairly dramatic fall in the propo rtion of households that are FHH. For example in t he case of Ha rari, 68.5% of quintile 1 households are FHH co mpared with 33.1% of those h ouseholds in quintile 2. Hence FHH appear to be particularly biased toward s being in the low est quintile of all regions. Table 5: Distribution of Individuals Across Quintiles Quintile 1 Quintile 2 Qu intile 3 Quintile 4 Quintile 5 Tigray MHH 40.4% 72.4% 75.8% 82.6% 75.7% FHH 59.6% 27.6% 24.2% 17.4% 24.3% Affar MHH 61.3% 70.4% 80.7% 82.2% 87.3% FHH 38.7% 29.6% 19.3% 17.8% 12.7% Amhara MHH 51.6% 78.5% 86.3% 89.1% 90.9% FHH 48.4% 21.5% 13.7% 10.9% 9.1% Orimiya MHH 50.5% 67.4% 75.7% 86.5% 86.3% FHH 49.5% 32.6% 24.3% 13.5% 13.7% Somale MHH 45.4% 76.2% 81.7% 79.5% 77.6% FHH 54.6% 23.8% 18.3% 20.5% 22.4% B-Gumuz MHH 51.4% 78.4% 80.1% 86.3% 92.4% FHH 48.6% 21.6% 19.9% 13.7% 7.6% SNNP MHH 53.5% 72.0% 80.0% 84.0% 88.2% FHH 46.5% 28.0% 20.0% 16.0% 11.8% Harari MHH 31.5% 66.9% 59.9% 81.3% 77.4% FHH 68.5% 33.1% 40.1% 18.7% 22.6% Addis MHH 42.9% 55.1% 63.9% 62.2% 69.7% FHH 57.1% 44.9% 36.1% 37.8% 30.3% Dirre Dawa MHH 44.6% 65.4% 71.6% 65.4% 71.0% FHH 55.4% 34.6% 28.4% 34.6% 29.0% (Source: Authors calculations, for HICE Report 2007). 14

15 Figure 2 Distribution of FHH Disaggregated By Expenditure Quintile and Region 80.0% Quintile 1 Quintile % Quintile 3 Quintile % Quintile 5 %FHH 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Tigray Affar Amhara Orimiya Somale B-Gumuz SNNP Harari Addis Dirre Dawa Region Comparing the proportion of FHH s across the quintiles and time, specifically with the 1995 HICE data we find that: On average the proportion of quintile 1 FHH s has increased from 43.5% and the proportion of FHH s in quintile 5 has remained roughly constant (Figure 3.1). In the middle expenditure range (quintile 3) the proportion of FHH s in 2000 decreases below the 1995 level - i.e. there appears to be smaller proportions of FHH s in the upper expenditure range in 2005 than 10 years earlier. Figure 3 Proportion of Female Headed Households (FHH) for 1995 and 2000, Disaggregated By Quintiles % 55% 50% 45% 40% 35% 30% 25% 20% 15% 10% Quintile 1 Quintile 2 Quintile 3 Quint ile 4 Quintile 5 Quintile

16 Table 6 considers household size and age groupings of households members and, overall, indicate a relatively equal distribution of both males and females, across the expenditure quintiles, calculated by column. Ta ble 6: Distribution of Individuals Across Quintiles Quintile Quintile Quintile Quintile Quintile All Male s 46.3 % 48.1% 50.0% 51.7% 52. 0% 49.7% Female s 56.5 % 51.9% 50.0% 48.3% 48. 0% 50.3% ( Source: Authors calculations, for HICE Report 2007). It is worth noting that of the individ uals in the lowest expenditure quintile, 56.5% are females compared with the highest expenditure quintile where there are 48% women. When we disaggregate this even further, Tab le 7, also fin d further interest ing trends su ch as the proportion of c hildren a ged 0-9 year s in the lowest quintile is approximately 11% compared with one quarter in the upper q uintile. T able 7: Expenditure Qui ntile of Entire Populat ion Disaggregated by Gend er and Age Quintile Quintile Quintile Quintile Quintile ALL Age 0-9 M+ F 2,428, ,815, ,621, ,246, ,461, ,573, M 1,224, ,936, ,351, ,671, ,757, ,940, F 1,203, ,879, ,269, ,575, ,704, ,632, M+ F 815, ,369, ,698, ,182, ,786, ,853, M 400, , , ,163, ,401, ,511,950 7 F 415, , , ,018, ,385, ,341, M+ F 737, ,131, ,355, ,687, ,283, ,195, M 277, , , , ,232, ,540, F 460, , , , ,051, ,654, M+ F 706, , , ,057, ,270, ,961, M 277, , , , , ,286, F 429, , , , , ,675, M+ F 715, , ,032, , ,053, ,793, M 331, , , , , ,157, F 384, , , , , ,635, M+ F 520, , , , , ,533, M 249, , , , , ,702, F 270, , , , , ,830, M+ F 388, , , , , ,155, M 160, , , , , ,469, F 227, , , , , ,685,

17 40-44 M+ F 247, , , , , ,327, M 86, , , , , ,046, F 160, , , , , ,281, M+F 242, , , , , ,046, M 79, , , , , , F 162, , , , , ,059, M+F 220, , , , , ,641, M 63, , , , , , F 157, , , , , , M+F 185, , , , , ,138, M 36, , , , , , F 149, , , , , , M+F 237, , , , , ,150, M 54, , , , , , F 183, , , , , , & M+F 578, , , , , ,112, above M 219, , , , , ,142, F 358, , , , , , Total M+F 8,025, , 290, ,066, , 967, ,132, , 482, M 3,461, , 383, ,482, , 660, ,738, ,725, F 4,564, , 907, ,583, , 307, ,394, ,756, ( Authors Calculations, HICE 2005) From a poverty perspective, we can see fr om the 1999/00 poverty data (Table 8), that there appears top be no real clear relati onship betw een pover ty and gender of the head of the household at country level. The poverty profile found suggested there to be no statistical difference between MHH and FHH poverty, however there appears to be interesting trend urban/rural differentiation - that is suggestive of significantly greater proportions of Urban based FHH s below the poverty line. 7 One primary reason could be the prevalence of divorce, with 74% of Urban FHH s divorced (Federal Republic of Ethiopia 2002: 73). Table 8: Headcount Poverty (%) Disaggregated by Gender of Household Head 1995/ /00 All Urban Rural All Urban Rural All Households Male Headed Household (MHH) Female Headed Households (FHH) (Source: Federal Repub lic of Ethiopia 2002) 7 Figures for the latest HICE are yet to corroborate or negate this. 17

18 If we consider a slightly more technical analysis of the expenditur e related data Table 9 provides an Ordinary Least Squares (OLS) Regression of Exp enditure per. capita, indicating that relative to be being widowed, households are likely to have significantly larger welfare if cohabiting or unmarried, and a signific antly increased level of welfare the higher the level of education for the household head. Table 9: OLS Regression - Expenditure per capita - Coefficients Variable Coefficient T-ratio Constant (8.696)*** Age of head (4.725)*** Female Head (-6.433)*** Age of head, squared (-3.987)*** Heads Marital Status Unmarried (14.916)*** Married (0.259) Cohabiting (1.835)* Divorced (0.925) Heads E ducation Some Primary (6.522)*** Primary (7.434)*** Some Sec ondary (7.309)*** Secondary (15.453)*** University (22.773)*** Heads Activity Type Agriculture/ Mining (-1.986)** Manufacturing/Construction (-1.346) Trade (-1.49) Motor Vehicle (2.435)** Hotel (2.162)** Transport (-0.126) Financial/Real Estate (4.027)*** Public Administration (-0.597) Education, Health and Social Works (-2.152)** Region Other Community and Social Activities (-1.819)* Amhara (-2.062)** Orimiya (0.144) Tigray (1.76)* Affar (1.604) Somali (0.442) B-Gumuz (2.933)*** SNNP (-0.817) Harari (3.465)*** Addis (0.255) Defaults: Missed education, Widowed, Private households and other organisations activity, Dire Dawa * Significant at 10% level, ** Significant at 5% level, *** Significant at 1% level (Sourc e: Authors Calculations) 18

19 Key Findings The most commonly observed household size was 6 members, with Urban households being slightly smaller than rural households. Households in the highest quintiles appear to have a higher average household size. Gender differences appear to increase with household expenditure. For example, of the top two quintiles approximately 85.4% of the households are MHH compared with less than 15% being FHH. From a regional perspective, beyond the lowest quintile, there is a fairly dramatic fall in the proportion of households that are FHH. For example in the case of Harari, 68.5% of quintile 1 households are FHH compared with 33.1% of those households in quintile 2. Hence FHH appear to be particularly biased towards being in the lowest quintile of all regions. There appears to be smaller proportions of FHH s in the upper expenditure range in 2005 than 10 years earlier. The poverty profile found suggested there to be no statistical difference between MHH and FHH poverty. 19

20 3. Health, Education and HIV/AIDS Awareness This section will focus on the key human capital components of education and health. Specifically we will analyse how health status, usage of health care facilities and education attendance/attainment varies across males and females within (and across) households. For instance, we will look at whether girls and boys receive disproportionately less (or more) education or health care, relative to boys within households, and see if this varies by income group (i.e. are women in the top income quartile more likely to seek health care than those in lower income quartiles). For the HIV/AIDS awareness section, we are able to focus on HIV/AIDS knowledge/awareness by gender of the household members, and investigate, for instance, if married men know how HIV/AIDS is prevented but their wives do not. 3.1 Health Child health, and more specifically, stunting and underweight prevalence have improved significantly since 1996, especially in urban area (Table 10). Rural areas are slower to respond, but both stunting and underweight have dropped significantly in the last decade. Wasting has not had the same fate, with a rise in rates in the late nineties that has not corrected it self to pre-1996 levels as of There is a clear decline in levels of stunting, and it is evident that urban children faired better. Stunting has fallen from 65.7% in 1996 to 46.9% in Wasting has slighting increased for both girls and boys. Table 10: Prevalence of Wasting, Stunting, and Underweight by Gender, Place of Residence and Over time 8 Wasting Stunting Underweight National Level Boys Girls Both Boys Girls Both Boys Girls Both Rural Urban (Source: CSA, WMS 2004) 8 CSA. WMS

21 Figure 4.: Percent Stunting from 1996 to 2004 at National level, and Rural and Urban % Stunting Boys Girls Both National Rural Urban Figure 4 shows the drastic rise in wasting from 1996 to Though decreasing since 1998, pre-1996 levels have not been attained. Note the stagnant and slight increase in stunting among urban children between 2000 and 2004, while the rate for rural children declined further. This could be because the majority of interventions are targeted to rural areas. This is, in turn, due to the higher malnutrition rates in rural areas, but this graph shows us that urban children should not be ignored. Figure 5: Percent Wasting from 1996 to 2004 at National level, and Rural and Urban 12 % Wasting Boys Girls Both National Rural Urban 21

22 Table 11: Determinants of Health of Pre-School Girls and Boys (Wasting) Regressions (Obs. 4206) Weight For Height Z-Score (Girls) Weight For Height Z- Score (Boys) R-squared= R-squared= Variable Coefficient T-ratio Coefficient T-ratio Constant (-3.138)*** (-2.651)*** Child Age (months) (2.269)** (2.094)** (4.989)*** (2.175)** (2.290)** (0.419) (1.143) (1.103) (0.064) (0.412) (0.051) (0.506) (0.150) (0.360) (0.403) (0.554) (0.176) (0.427) Age of Head (-0.891) (-2.274)** Female Head (0.686) (1.531) Household size (1.657)* (2.151)** Parental Education Primary (-0.094) (-0.166) Secondary (1.386) (1.088) Higher (1.273) (0.783) Region Tigray (-0.271) (-0.286) Afar (-4.478)*** (-4.164)*** Amhara (-1.766)* (-0.842) Ormiya (-1.718)* (-1.481) Somali (-3.298)*** (-2.247)** Bensh-Z (-1.348) (-1.339) SNNPR (-2.13)** (-1.235) Gambella (-3.316)*** (-1.828)* Dire dawa (-0.17) (-0.344) Addis (-0.978) (-0.092) Iodine (-1.045) (-1.066) Iodine (-1.121) (-1.424) Iodine (-1.284) (-1.181) Afterbirth care rec d (1.692)* (-0.276) Two month check up post birth (-.735) (-.901) Toilet Type Flush Toilet (2.062)** (1.825)* Pit Latrine (1.648)* (1.469) No Flush (1.284) (1.418) Wealth (1.553) (1.988)** * Significant at 10% level; ** Significant at 5% level; *** Significant at 1% level Defaults Missed Education (for all education variables), Toilet bush, Iodine 30, No 2 month check up post birth, region Harari (Source: Authors Calculations) 22

23 Table 12: Determinants of Health of Pre-School Girls and Boys (Stunting) Regressions (Obs. 4206) Height for Age Z-Score Height for Age Z-Score (Girls) (Boys) R-squared= , R-squared= Variable Coefficient T-ratio Coefficient T-ratio Constant (-3.178)*** (-2.685)*** Child Age (months) (4.908)*** (3.797)*** (4.857)*** (3.573)*** (1.405) (1.639) (5.759)*** (3.647)*** (3.734)*** (2.689)*** (5.708)*** (3.982)*** (4.351)*** (2.221)** (7.364)*** (5.834)*** (5.361)*** (4.057)*** Age of Head (-0.907) (-2.279)** Female Head (0.649) (1.497) Household size (1.729)* (2.213)** Parental Education Primary (-0.167) (-0.218) Secondary (1.001) (0.788) Higher (1.07) (0.612) Region Tigray (-0.157) (-0.206) Afar (-4.365)*** (-4.085)*** Amhara (-1.603) (-0.718) Ormiya (-1.578) (-1.378) Somali (-3.184)*** (-2.172)** Bensh-Z (-1.229) (-1.238) SNNPR (-1.913)* (-1.062) Gambella (-3.262)*** (-1.763)* Dire dawa (-0.138) (-0.321) Addis (-1.083) (-0.196) Iodine (-1.052) (-1.088) Iodine (-1.12) (-1.446) Iodine (-1.28) (-1.196) Afterbirth care rec d (1.768)* (-0.287) Two month check up post birth (-0.897) (-0.733) Toilet Type Flush Toilet (2.06)** (1.833)* Pit Latrine (1.626) (1.458) No Flush (1.295) (1.433) Wealth (1.445) (1.874)* * Significant at 10% level; ** Significant at 5% level; *** Significant at 1% level Defaults Missed Education (for all education variables), Toilet bush, Iodine 30, No 2 month check up post birth, region - harari (Source: Authors Calculations) 23

24 If we consider the determinants of wasting and stunting for children under the age of 5 years, though Ordinary Least Squares (OLS) regression approach. We potentially further explain some of the characteristics associated with stunting and wasting in Ethiopia. For example, we can see from Tables 11 and 12, that: Household size appears to have a strongly positive association with both stunting and wasting, i.e. the larger the household size the more like the probability that a child is either wasted and stunted. Although increased age of the household head does appear to be associated with less wasting and stunting, this is only statistically significant for boys. Wasting, in particular, positively associated with the weaning period. 3.2 Education Focusing on literacy Table 13 provides support for prior Ethiopian household survey descriptive data (see for example WMS 2004) that has identified approximately 1/3 rd (32.1%) of the country to be literate. Using the HICE data to analyse this further we can see there are significant rural/urban/quintile differences in literacy rates. For example, on average 48.2% (83.3%) of the urban (rural) lowest quintile households are illiterate (Table 3.1). Table 13: Literacy Rates (%) , Disaggregated by Gender 1995/6 1999/ /5 All Urban Rural All Urban Rural All Urban Rural Total Male Female (Source: Abridged: Federal Republic of Ethiopia 2002; WMS 1995, Dercon 1997, Authors calculations for HICE Report 2008) 9. Table 14 focuses on literacy household heads and as highlighted in prior household survey descriptive data - approximately 37.6% of the country are literate. Using the HICE data to analyse this further significant rural/urban/quintile differences in literacy rates appear to be present. Table 14: Percentage Illiterate 2005 Household Members Disaggregated by Expenditure Quintile and Gender National Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 All All 75.0% 70.6% 65.0% 61.2% 50.1% 62.4% Male 61.5% 59.9% 54.0% 50.1% 39.5% 50.6% 9 Note: Sampling and calculation differences do not make the 1995 and 1999 figures directly comparable with 2005, although it is reasonable to draw trend analysis conclusions. 24

25 Female 84.0% 79.6% 75.5% 72.9% 61.3% 73.3% Rural All 80.9% 76.0% 70.6% 66.8% 60.1% 69.4% Male 68.5% 65.3% 59.3% 54.7% 47.9% 57.2% Female 89.3% 85.3% 81.6% 80.0% 73.7% 81.3% Urban All 44.8% 33.7% 27.4% 23.6% 18.5% 26.3% Male 26.6% 21.4% 15.6% 14.1% 8.9% 14.7% Female 56.7% 43.4% 37.2% 31.8% 26.8% 35.8% (Source: Authors calculations, for HICE Report 2008). On average 44.8% (80.9%) of the urban (rural) lowest quintile households are illiterate Relatively large gender differences also appear to exist with 89.3% of rurally based female heads, in the lowest quintile, being illiterate compared with 68.5% of men. Analysing this across expenditure classification reveals that, unsurprisingly, illiteracy levels decrease as expenditure quintiles increase. For example, the proportion of illiterate, rurally based, individuals decreases to 73.7% and 60.1% for female and males, respectively, as expenditure increase to the highest quintile. Considering the proportion change in illiteracy status across quintiles (Figure 6) and disaggregating by gender we can see that: Notably, moving from to the 4 th the 5 th quintile provides some the most distinct reductions of illiteracy status - particularly in rural areas. For males and females illiteracy rates fall by 12.4% and 7.8%, respectively, when comparing the 4 th and 5 th quintile. Across most quintiles the reduction in female illiteracy is less than that for males (a notable exception to this is for the move from quintile 1-2 an and urban based males and females) Figure 6 Percentage Change in Illiteracy Disaggregated by Rural/Urban Male/Female Across Quintiles % Change in Illiteracy 14.0% 12.0% (Rural) 10.0% 8.0% 6.0% 4.0% 2.0% 0.0% Rural Male Rural Female Urban Male Urban Female Quintile 1-2 Quinile 2-3 Quintile 3-4 Quintile 4-5 Quintile % Change in Illiiteracy 20.0% 15.0% (Urban) 10.0% 5.0% 0.0% 25

26 Figure 7 highlights some of the interesting regional differences of illiteracy, across both quintiles and gender. Across regions, the average level of female illiteracy (73.3%) is higher than that for males (50.6%). The largest inequalities between males and females being in Benishangul-Gumuz and Orimiya where 28.3% and 26.5% more males are literate than females. The least gender inequality appears to exist in Addis, where 9% of males are illiterate, compared with 26.2% of females. Table 15: Illiteracy levels and Percentage Change Across Expenditure Quintiles Disaggregated By Region and Gender Tigray Affar Amhara Orimiya Somali B- Gumuz SNNP Harari Addis Dirre Dawa All Male Female (Source: Authors Calculations) Figure 7: Illiteracy Levels Disaggregated By Region and Gender 90.0% 80.0% 70.0% % Male Illiterate % Female Illiterate 60.0% % Illiterate 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% All Urban Rural Tigray Affar AmharaOrimiyaSomali B-Gumuz SNNP Harari Addis Dirre Dawa Region 26

27 Now we consider the educational level of household head and individual member and disaggregated by expenditure quintiles and gender. Considering household heads, Table 16, indicates that approximately 10.2% of all household heads have not completed the final levels of primary education or any secondary education. In the lowest (highest) quintile only 4.9% (18.9%) have completed up to the final primary educational levels or above. Major educational attainment differences exist across rural and urban areas: 4.9% (38%) of the rural (urban) based population heads having completed the upper levels of primary education, or above. On average 5.8% (46.6%) of MHH rural (urban) based households have completed upper levels of primary education and secondary, compared with 1/7% (24.4%) of FHH s.. The data indicate that in all cases smaller proportions of FHH, relative to males. Perhaps surprisingly this is particularly acute in the higher quintile rural (urban) households where for example only 1.8% (36.8%) of FHH have completed upper levels of primary education compared with 8.9% (56.5%) of MHH. Table 16: Percentage of All Household Heads 2005 Who Have Completed up to Final Primary School Levels, or above, Disaggregated Expenditure Quintile and Gender National Urban Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Overall All 4.9% 6.5% 9.4% 11.1% 18.9% 10.2% Male 6.1% 6.4% 9.7% 11.4% 19.2% 11.3% Female 3.6% 6.8% 7.7% 10.3% 17.1% 7.1% All 18.5% 29.6% 36.5% 44.4% 51.4% 38.0% Male 25.7% 34.9% 42.3% 52.0% 56.5% 46.6% Female 13.7% 23.1% 26.8% 29.6% 36.8% 24.4% Rural All 2.3% 3.0% 5.3% 6.1% 8.1% 4.9% Male 3.4% 3.5% 6.1% 6.6% 8.9% 5.8% Female 1.1% 1.6% 2.2% 2.5% 1.8% 1.7% (Source: Authors calculations, for HICE Report 2008). When we consider educational attainment beyond secondary, interestingly the inequality gap between the highest and lowest quintile household heads seems to grow further than what appeared with primary education. For example, and as already noted, the average proportion of lowest quintile heads not completing primary education is 5.2% compared with 20.2% of the highest quintile. If we consider educational attainment across all household members (Figure 8) are the 27

28 rural/urban/quintile differences for individual members who have reached the upper levels of primary education or beyond (i.e. secondary education, university). As would be expected the proportion of individuals in the highest quintile who have completed these educational levels is higher than that of the lower quintiles, as is the case for more men than women. Figure 8: Percentage of Individual Household Members Who Have Completed Upper Levels of Primary Education, or Beyond 50.0% 40.0% % 30.0% 20.0% Quintle 1 Quintile 5 Overall 10.0% 0.0% National All Naqtional Male National Female Rural All Rural Male Rural Female Urban All Urban Male Urban Female National/Urban/Male/Female For regional comparisons of educational attainment (Table 17) the higher proportions of individuals completing primary educational level of 7-8 and beyond is dominated by the relatively high performers of Harari, Addis Ababa and Dire Dawa, with an average of 31.2%, 49.7% and 30.4%, of individuals, respectively completed up to this level of education. Throughout all regions there appears to be a disparity across gender. Amhara has relatively small differences across male/female school completion, although this is considering the very low level of overall educational attainment. Table 17: Percentage of Population Completing Up to 7-8 years of Primary School or beyond, Disaggregated By Region and Gender All Tigray Afar Amhara Orimiya Somali B- Gumuz SNNP Harari Addis Ababa Dire Dawa All 10.3% 11.0% 12.0% 5.6% 8.9% 8.5% 9.8% 8.7% 31.2% 49.7% 30.4% Men 13.0% 13.0% 16.2% 6.4% 12.0% 12.0% 14.3% 12.7% 35.5% 57.1% 38.8% Women 7.8% 9.1% 7.9% 4.9% 5.9% 5.0% 5.0% 5.0% 27.3% 43.5% 22.6% (Source: Authors calculations, for HICE Report 2008). Going a little further in the analysis and considering the reasons for educational drop out (Table 18), we can see that: For males, one of the main reasons for primary and secondary school drop out is the need to work. With approximately ¼ of all primary aged students dropping out of school because of this reason (30.2% for boys and 18.9% for girls). Pregnancy results in 10% of girls dropping out of Secondary School Interestingly being unable to afford material is a relatively uncommon reason for secondary school drop out for girls (8.0% overall Male 10.4%, Girls 1.6%). This 28