Tanzania - A Poverty Profile

Document of The World Bank CONFIDENTIAL Report No. 12298-TA TANZANIA A Poverty Profile December 199 FILE COPY D Population and Human Resources Division Eastern Africa Department Africa Region FILE COPY CONFIDENTIAL Report No: 12298 TA Type: SEC CONTEN'IS Executive Sum m ary .............................................. 1 Background and Justification .................................... 1 Tracking National Income and Inequality ............................ 1 Summary of the Results ....................................... 3 Basic Needs Fulfillment ....................................... 5 Sources of Income and Holdings ofAssets............................ 9 Policy Im plications .......................................... 10 1 Background ................................................. 12 2 The Distribution of Income inTanzania ............................... 15 Measuring and Determining Poverty ............................... 15 Choice of Poverty Index ....................................... 16 Incom e Levels ................................... ...... ... . 17 Incom e Inequality ........................................... 20 3 Characteristics of the Poor inTanzania................................ 25 Levels of Expenditure ........................................ 25 Expenditure Shares .......................................... 26 Basic Demographic Characteristics ................................ 28 The Gender Dimension ofPoverty................................. 30 Depth and Severity .......................................... 31 Spatial M apping of thePoor .................................... 32 Socio-Economic Mapping of thePoor .............................. 37 Sensitivity Tests ............................................ 38 Poverty by Income Instead of Expenditure ....9...9................... . 39 Sum m ary ................................................ 39 4 Basic Needs and Assets of thePoor .................................. 40 Satisfaction of Basic Needs of the Poor .............................. 40 Education . . .. . .. . . .. .. . . .. . .. . . . . . .. . . . . . . . . . . . . . . . . . . . . . 40 Basic Housing Amenities/Housing Quality ........................... 46 Asset H oldings ............................................. 48 Sources of Income .......................................... 53 5 Conclusions and Policylmplications.................................. 56 The State of Poverty ......................................... 56 Strategies for Reducing Poverty .................................. 57 Targeting the Poor ........................................... 57 Bibliography . . . ... . . ... . . .. . ... . . .. . .. . .. . . . . . . . . . . ... . . . . . . . . . 60 ANNEX A Data ................................................ 64 Sam ple D esign ............................................. 64 Contents of the Cornell/ERB Survey ............................... 67 Sample Means for SelectedVariables ............................... 67 ANNEX B Methodological Background ................................. 71 A Poverty Profile ........................................... 71 How to Rank Households? ..................................... 71 Adult Equivalence Scales ....................................... 72 ExpenditureMeasurement...................................... 72 Poverty Line .............................................. 74 Poverty Indices ............................................. 76 The Decomposability Property ................................... 77 Inequality Indices and the LorenzCurve ............................. 78 ANNEX C AdditionalTables........................................ 81 TABLES Table 1.1: Social Indicators for Tanzania and Sub-Saharan Africa ............... 13 Table 2.1: Per Capita Income in Tanzania since 1969, in 1991 Prices ............. 19 Table 2.2: Some Indicators of Inequality in Tanzania, 1991 ................... 20 Table 2.3: Gini Coefficient for the Distribution of Income in Tanzania, 1969-91 ...... 21 Table 2.4: The Distribution of People by Adult Equivalent Household and Per-Capita Annual Expenditure ..................................... 22 Table 2.5: An International Comparison of Income Distribution Using the Gini Coeffi cient .......................................... 24 Table 3.1: Yearly Expenditures According to Type of Household, 1991 ........... 25 Table 3.2: Expenditure Patterns by Household Type ........................ 26 Table 3.3: Composition of Food Expenditures ........................... 27 Table 3.4: Demographic Characteristics by Type of Household ................. 28 Table 3.5: Three Different Poverty Indices and Two Poverty Lines, by Location of Household ..... 32 Table 3.6: Regional Decomposition of Poverty for Three Poverty Measures and Two Poverty Lines ......................................... 34 Table 3.7: Poverty by Agro-Climatic Zone ............................. 35 Table 3.8: Three Different Measures of Poverty and Two Poverty Lines, by Employment Group and Location ............................ 37 Table 3.9: Expenditure vs. Income as a Measure of Welfare - Soft Poverty Line ..... 39 Table 4.1: Literacy Among People Older Than 14 ......................... 40 Table 4.2: Highest Education Achieved (People Older than 14) ................. 42 Table 4.3: Reason to Stop AttendingSchool ............................ 44 Table 4.4: Fraction of Children Enrolled in School ........................ 45 Table 4.5: Basic Housing Amenities (Percent of Total Households in Each Category) . . 46 Table 4.6: The Main Water Source of Households ......................... 46 Table 4.7: Access to Source of Water - How Long Does It Take? (Percentages) ...... 47 Table 4.8: Ownership of Durables (Percent of Households in Each Category) ........ 48 Table 4.9: Percentage of Households owning Land and Livestock ............... 48 Table 4.10: Comparing the Inequality of the Income Distribution with the Inequality of Land Ownership ....................................... 49 Table 4.11: Livestock Ownership Distribution ............ ............... 51 Table 4.12: Structure of Physical Labor Use (Total Population) ................. 52 Table 4.13: Sources of Income ................................... . 54 Table 4.14: Income Differentiation by Income Source for Households with Positive Income from That Source ................................. 55 Table 5.1: Values of the Poverty Targeting Indicator PTIBxlO by Region and Socio- Econom ic Group ...................................... 59 Table A. 1: The NMS Frame and the Cornell/Economic Research Bureau Sub-Sample . .. 64 Table A.2: Urban and Dar es Salaam Sub-Sampling ........................ 65 Table A.3: Correction Factors for Regional Sub-Sampling .................... 65 Table A.4: Stratification for Economic Activity ........................... 66 Table A.5: Selected Characteristics of the Households Surveyed ................ 67 Table A.6: Regional Sampling ...................................... 68 Table A.7: Sample Size used in the computations of Table 4.14 ................ 69 Table A.8: Literacy in Tanzania (People older than 14) - Number of Cases ......... 70 Table B. 1: Index of Caloric Requirements by Age and Gender for East Africa ....... 72 Table B.2: Index of Household Economies of Scale ........................ 73 iii Table C. 1: Decomposition of P, Poverty (Soft) Index by Gender of the Head of the H ousehold ............................. ... ...... .. .. . 81 Table C.2: Numbers of LivestockOwned .............................. 82 Table C.3: Reason to Stop Attending Primary School ...................... 83 Table C.4: Reason to Stop Attending School, Forms 1 to 4 ................... 84 Table C.5: Reason to Stop Attending Higher Level Education (Forms 1-4 through U niversity) ....y..................................... . 85 Table C.6: Intensity of Work by Days Available ........................... 86 Table C.7: Land Use Patterns for Owners or Renters ....................... 87 Table C.8: Land Tenure for Land Owners or Renters ....................... 88- Table C.9: Landholdings per Household and Distribution of Landholdings .......... 89 Table C. 10: Regional Per Capita GDP forl991 .......................... 90 FIGURES Figure S.1: Changes in Per Capita Income Over Time ........................ 2 Figure S.2: Changes in Inequality overTime.............................. 3 Figure S.3: Percentage of Tanzanians Who Are Poor ......................... 4 Figure S.4: W ater Sources ......................................... 6 Figure S.5: Regional Map ofTanzania .................................. 7 Figure S.6: Proportion Literate by Age Group ............................. 8 Figure S.7: Sources of ncome ...................................... 9 Figure 2.1: Changes in Per Capita Income Over Time ....................... 19 Figure 2.2: Changes in Inequality overTime.............................. 22 Figure 2.3: Lorenz Curve for the Welfare Distribution ....................... 23 Figure 3.1: Expenditure Shares by Location ............................. 27 Figure 3.2: Gender and Age Distribution of the Population .................... 29 Figure 3.3: Age Distribution of the Population by Poverty Status ............ ... 29 Figure 3.4: Gender of the Head of the Household ......................... 30 Figure 3.5: Poverty Indicators by Location .............................. 33 Figure 3.6: Poverty Indicators for Poorest Regions, Soft-Core Poor ............... 35 Figure 3.7: Regional Map ofTanzania................................. 36 Figure 4.1: Proportion of Literate by Age Group and Poverty Status .............. 41 Figure 4.2: Proportion Literate by Age Group and Gender .................... 42 Figure 4.3: Percentage of Respondents Completing Primary or Secondary Education . . .. 43 Figure 4.4: Reasons for Abandoning Education (All Levels) ................... 44 Figure 4.5: Percent Walking More than 15 Minutes for Water .................. 47 Figure 4.6: Lorenz Curve for the Distribution of Land ...................... 50 Figure 4.7: Allocation of LaborTime ................................. 53 Figure 4.8: Sources oflncome ..................................... 54 Figure B. 1: Generalized Lorenz Curve for the Income Distribution ............... 79 BoXEs Box B. 1: The Absolute PovertyLine ................................ 75 Box B.2: Poverty Targetinglndicators................................. 80 Box B.3: Index of Livestock Values ................. ................ 80 iv ACKNOWLEDGMENTS This report was prepared by Maria Luisa Ferreira, consultant in the Population and Human Resources Division of the Eastern Africa Department with the collaboration of Lucy Goodhart (AF2PH), Charles Griffin (AF2PH), and Rogier van den Brink (AF6AE). The author is grateful for the collaboration received from Haidari Amani (Economics Department of the University of Dar es Salaam), Wilbald Maro (Economic Research Bureau of the University of Dar es Salaam), and Cleutius Mkai (Tanzania Bureau of Statistics). The Task Managers for the project were James Coates (AF2AE) and Charles Griffin (AF2PH). The report was prepared under the general supervision of Jacob van Lutsenburg Maas (AF2PH). Comments on an earlier draft were received from James Coates, Young Kimaro, and Bill Shaw. The Peer Reviewers were Ms. Margareth Grosh, Mr. Antoine Simonpietri, and Mr. Jack van Holst Pellekaan. The Lead Economist was Mr. Miovic. The data for this report were collected through a joint program between the Cornell University Food and Nutrition Policy Program and the Economic Research Bureau of the University of Dar es Salaam that was funded by the U.S. Agency of International Development. V ACRONYMS AND ABBREVIATIONS AIDS Acquired Immune Deficiency Syndrome ASM Agricultural Sector Memorandum CPI Consumer Price Index DSM Dar es Salaam ERP Economic Recovery Program ERB Economic Research Bureau of the University of Dar es Salaam FGT Foster, Greer, Thorbecke Index of Poverty Severity GDP Gross Domestic Product HBS Household Budget Survey of the Tanzania Bureau of Statistics ICP International Comparisons Project of the United Nations ILO International Labor Organization IMF International Monetary Fund NMS National Master Sample ODA Official Development Assistance PO Head Count Measure of Poverty P, Measure of Depth of Poverty P2 Measure of Severity of Poverty PPP Purchasing Power Parity SSA Sub-Saharan Africa Tsh Tanzanian Shillings UN United Nations USAID United States Agency for International Development US$ United States Dollar vi ExEcuTIVE SUMMARY BACKGROUND AND JUSTIFICATION 1. The standard dollar estimate of Tanzania's per capita income in 1990 was US$110, making- Tanzania one of the poorest countries in the world. Poverty reduction is the overriding objective of the Bank's lending activities, and it has also been a key objective of the Tanzanian Government since Independence. To help attack poverty, we need information. It is in this context that the Bank, the Government of Tanzania, and the University of Dar es Salaam collaborated in preparing this poverty profile. 2. The statistics presented in this profile come from one of the most recent, "post-adjustment," nationally representative household budget surveys in Sub-Saharan Africa. The survey was a joint effort undertaken by the Cornell Food and Nutrition Policy Program and the Economic Research Bureau of the University of Dar es Salaam, and was funded by USAID. The survey covered 1,046 households (out of 4.3 millions) in rural and urban areas of Tanzania drawn from the National Master Sample of the Bureau of Statistics.' 3. This poverty profile describes quantitatively the poor in Tanzania - who they are, where they live, what they consume, and where their incomes come from. It analyses the depth and nature of poverty, providing detailed information on the standards of living of poor households in terms of income, expenditures, and fulfillment of basic needs. TRACKING NATIONAL INCOME AND INEQUALITY Income Levels 4. The most recent trends in income and inequality were analyzed using past estimates from academic and official sources. The 1990 estimate of per capita income of US$1 10 must therefore be taken with caution, especially given the range of alternative estimates. For instance, the United Nations International Comparison Project (ICP) estimates per capita purchasing-power-parity- equivalent GDP for 1990 at US$540. Because of uncertainty regarding Dollar estimates, this profile will examine trends in per capita income mainly in Tanzanian Shillings, although these estimates must also be carefully interpreted. As Sarris and van den Brink (1993) point out, the official consumer price index (CPI) and the deflator implicit in estimates of real GDP are likely to be under estimated. Both use prices in official markets,which do not capture the reality of prices in extensive parallel markets in Tanzania. The result is that growth rates of real income in Shilling terms are over estimated. Putting aside these issues for the moment, our survey found that Tanzanian. annual per capita income was Tsh 55,369 in 1991 and per capita expenditures were Tsh 61.564. For the rural areas only, per capita income was Tsh 36,252 and for urban areas it was Tsh 100,222. Thus. per capita expenditure in Dar es Salaam was 129 percent greater than per capita expenditure in rural villages. The distinction betiween rural, urban areas outside Dar es Salaam and Dar es Salaam follows the classification adopted for the 1988 Population Census which is the basis for the NMS. 2 Executive Summary 5. Figure S. 1 shows the trend in per capita income, in 120 o S urban and rural areas, using past estimates as a comparison. All 1oo 0 per capita income data is expressed in real terms (in 1991 prices) so that the graph shows 80 real growth. Thus, there is a Urban strong argument to believe that so - both rural and urban incomes Pural improved substantially between the lows of the mid-seventies and . 1991. Looking at the more 0 recent data on the rural sector, it 20 seems as if the launching of the X ]K Economic Recovery Program in a 1 1 1986 is showing positive effects. 1969 1976/77 1979/80 1982/83 1991 Inequality Figure S.1: Changes in Per Capita Income Over Time 6. The main objective of Ujamaa - Tanzania's version of socialism - was to reduce inequality in Tanzania. A comparison' with earlier estimates, however, suggests an increase in income inequality between 1969 and 1991. In 1969, according to the Household Budget Survey, the Gini Coefficient' for Tanzania was 0.39. In spite of the implementation of Ujamaa in the 1970s, the coefficient rose to 0.44 in 1976/77, and rose again to 0.57 in 1991. 7. Figure S.2 highlights the trends in rural and urban inequality since 1969. Contrary to common wisdom, it suggests that Tanzania's rural inequality is now greater than its urban inequality. As the time lines show, rural inequality was much less than urban inequality in 1969. Since then, however, rural inequality has risen while urban inequality has dropped, so that the two lines cross over. The worsening rural inequality has outweighed the effect of improving urban equality, so that national inequality has gone up. Measuring Poverty 8. To measure poverty we need a standard: what is poverty and who counts as poor? Two poverty lines were chosen to conduct the analysis - one relative and one absolute. Household expenditure was then compared to the two poverty lines, but was first weighted to control for households' different adult and child composition. Households with less than 50 percent of the mean .adult equivalent" expenditure were classified as poor. This came out to a poverty line of Tsh 46.173 per adult equivalent. Households who were unable to afford even the basic needs recognized by the society were deemed to be very poor or hard-core poor. The basic needs criterion was built These compariA)ns should be taken with caution, since the different estimates cast the use of different methodologies. Iloever. the magnitude of the differences leads us to believe that in fact there was an increase of inequality overtime The Gini coefficient i% an index of inequality. When an income distribution is perfectly equal the Gini coefficient reaches the value of tert If the income is concentrated in the hands of a single agent (perfect inequality) then the Gini coefficient is equal to one Executive Summary 3 up from an essential food diet, Gini (Higher Means Greater inequality) related to local eating habits and 0.6- nutritional requirements, with the addition of a few non-food items necessary for survival. The cost of obtaining these basic needs - National was Tsh 31,000 for 1991, 0.4 Urban estimated using a market basket - ural of prices. Members of poor households are automatically 0.3 -- classified as poor, but the survey did not permit us to determine how income is distributed within 0.2 . 199 196/771991 households. The following 1969 1976/77 summary characteristics emerged Year from the analysis, using 1991 Figure S.2: Changes in Inequality over Time data. SUMMARY OF THE RESULIS4 Head Count 9. About 50 percent of all Tanzanians live in households classified as poor and 36 percent of all Tanzanians live in households classified as hard-core poor. Average expenditures per capita and per adult-equivalent for better-off households (with expenditures above the poverty line) are six to seven times higher than those of the poor, and eight to ten times higher than those of the hard-core poor (see Table 3.1 in the main text for data). Extremes of Poverty 10. The Foster-Greer-Thorbecke index of severity of poverty is 15.6 percent for the upper poverty line and 9.4 percent for the hard-core poverty line, indicating a greater severity of poverty in both distributions. Depth of Poverty 11. If perfect targeting were possible, the minimum amount of transfer payments required to raise all of the poor to the soft-core poverty line would be Tsh 11,497 annually per person' and Tsh 4,867 annually per person to raise people to the lowest, hard core poverty line of Tsh 31,000. The total national transfer required would be Tsh 306 billion and Tsh 130 billion, respectively. In dollar terms, the amounts would be US$1.5 billion and JS$0.64 billion, respectively. In general, the poor are poorer in the rural areas, so it would cost more in total to bring the rural poor up to the poverty line than the urban poor. All the results are based on a sample of less than 0.02% of the households. The 'generalization' of the results for Tanzania - even if subject to statistical error as it is the case with any estimate-- is made possible by the use of the National Master Sample of the Bureau of Statistics, which allows us to 'blow' the results from the sample to the population. These numbers refer to the cost per Tanzanian, not the cost per poor person. 4 Executive Summary Rural Poverty 12. As Figure S.3 shows, poverty in Tanzania is mainly a rural phenomenon - 59 percent Rural of people living in rural villages are poor (10.915 millions of tanzanians), 39 percent in the Urban urban areas excluding Dar es Salaam and only 9 percent in Dar es Salaam itself. Because the population is predominantly rural, Dar as Salaam this means that about 85 percent of the poor are found in rural 0% 26% 60% 75% 100% areas. The rural character of Poor M Better Off poverty is even more pronounced when looking at the hard-core Figure S.3: Percentage of Tanzanians Who Are Poor poor. In that case, rural villages account for 90 percent of hard-core poverty in Tanzania. Necessities of Life Dominate Budgets 13. The share of expenditures given over to food is high for all households, indicating (perhaps more conclusively than the income estimates) that Tanzania generally has a low standard of living. However, the poor and the hard-core poor spend more of their budgets on basic food items (76 percent and 77 percent respectively), than the in general (73 percent) and their diets contain more cereal, and less meat and vegetables, than the diet of better-off households. Thus, 54 percent of all food expenditures in poor households are for cereals, versus 44 percent for better-off households. Household Size 14. Poor and hard-core poor households are larger than other households and contain more children. On average, better-off households have 3.04 children under the age of 18, poor households have 3.88, and hard-core poor households have 4.03. Poverty seems to be closely related to a high dependency ratio. Gender 15. Almost 10 percent of households nationwide are headed by a women, but the percentage is higher for urban areas, where 18 percent of households are headed by a women. Comparative experience suggests that female headed households are more likely to be poor than male headed households. However, this does not seem to be the case for Tanzania where the two types of household face a similar likelihood of poverty. Executive Summary 5 Sources of Income 16. Farm households constitute the bulk of poverty in Tanzania: 59 percent of people living in households primarily engaged in agriculture are poor, and these households make up 83 percent of all poor households. Rural households headed by a business man or woman are even more likely to be poor (in 62 cases out of 100), but because these households are few in number, they make up just 0.9 percent of the total poor .6 Of all the urban groups, households who describe themselves as self- employed are the most affected by poverty: 34 percent of all such households are poor and they make up 14 percent of the poor . Regions' 17. The connection between poverty and low productivity in agriculture is confirmed by the regional mapping of poverty, which indicates that poor regions are characterized by low rainfall, poor soils, distance from markets, and minimal infrastructure. The four poorest regions are Kigoma, Shinyanga, Lindi, and Ruvuma (See Figure S.5). In both Shinyanga and Lindi, over 90 percent of all people live in poor households and approximately 80 percent are in households that are classified as hard-core poor. The depth of poverty in Lindi is 67.5 percent, indicating that households are far below the poverty line, and the Foster-Greer-Thorbecke index is 58.1 percent, indicating that the poor households are in severe poverty. BAsIc NEEDS FULFILLMENT 18. The poor are disadvantaged in terms of income and expenditure, but there are also stark differences between poor and better-off households in their access to water, housing conditions, durable goods, and education. There were also large differences between rural and urban households. Some of the findings are summarized below: Water 19. More than 40 percent of all households - both urban and rural - spend more than 15 minutes to reach the nearest water source. However, people living in rural households are far more likely to face a long walk: 26.1 percent of rural households must walk over 30 minutes to their main water source, versus 3.9 percent of urban households outside Dar es Salaam, and 2 percent of households in the capital.' Rural households are also distinguished from their urban counterparts 6 In our sample few households in the rural areas are headed by a business man or woman. For those living in poverty, the business activity that they were engaged in was related to food - processing of food or grains, and food trade (with and without a shop). No household whose head was involved with non food business activities was found to be living in poverty. The Cornell!ERB household survey was not designed to obtain estimates at the regional level. In some regions the sample size is too small. and the results should be taken with caution. We compared our estimates with the official regional per capita GDP estimates (see table C.10). Those results and the results in this survey are completely different. For example, according to the official estimates Kilimanjaro is the 5th poorest region in Tanzania, and Kagera the 2nd poorest region. According to the Cornell/ERB survey both Kilimanjaro and Kagera belong to the top 5 richest regions. This comparisons show how little we know about the regional distribution of poverty and income. Better targetting calls for further research in this area. ' While it would be interesting to determine who performs household activities, particularly activities that are typically carried out by women. there is no information in the survey that allows us to differentiate activity by gender. 6 Executive Summary Rural Urban Dar es Salaam 0% 25% 50% 75% 100% Surface E Spring Well Tap Figure S.4: Water Sources by a greater reliance on streams and wells for their water source. The particular sources used by each group are shown in Figure S.4. &eCU(.ve .fum&ry__7 俗 斤gure 5.5:Regional Map of Tanzania 8 Executive Summary Housing 20. Housing quality also reveals standards of living. Nat ionally, only 33 percent of households live in houses without mud walls, and 53 percent of households have no metal roof. FDr the poor, these values are 21 percent and 68 percent, respectively. Light 21. Only 25 percent of poor households obtain illumination from a non-kibatari light.' This increases to about 47 percent for better-off households. Consumer Durables 22. 73 percent of households in Dar es Salaam contained at least one person who owned a watch; this figure is 60 percent for households in other urban areas, 41 percent for rural households, and 37 percent for poor households. Literacy 23. Overall, the survey found Percent Literate a literacy rate of 68 percent, 100 which is below the government's most recent estimate. Among so-- ... ................. ........ women, the literacy rate (defined as those who can both read and 60 - ... .... ................ ...... write) is 61 percent, while for men it.is 76 percent. As many as 40-- ... ... ....... ....... .... 80 percent of all those who are illiterate live in rural arm, and 20- .... ... ... ... ...... 70 percent of the illiterate are I L women. However, looking at the 01 gender breakdown of literacy by 14-19 20-29 30-39 40-49 50-59 600 age range (illustrated in Men M women Figure S.6), one can see that the Figure S.6: Proportion Literate by Age Group gap between men and women's literacy has dropped substantially in recent years. Poor people also shoulder a disproportionate burden of illiteracy. Only 59 percent of poor people and 56.5 percent of the hard-core poor are able to read and write. Education 24. The poor have less education than other groups and thus a lower stock of human capital. 32.2 percent of all poor people, and 34.9 percent of the hard-core poor have received no education whatsoever, as compared to 19 percent of better-off people. One of the key distinguishing features of the poor is a much lower likelihood of reaching and completing secondary education. 11.3percent The original survey instrument asked respondents. "Do you have a non-kibatari light?-. A kibatari light is one using a wick i.e.. a candle or oil lamp. Executive Summary 9 of the better-off are in or have completed secondary education, versus just 2.2 percent of the poor and 1.3 percent of the hard-core poor. SOURCES OF INcoME AND HOLDINGS OF ASSETS 25. As expected, agriculture is the most important source of income in Tanzania, and more so for the poor and hard-core poor. Agriculture accounts for 55.3 percent of the average total income for survey households, but for the poor and hard-core poor agriculture accounts for 72.1 percent and 74.2 percent of income, respectively. The next most important category of income is business income, followed by non-agricultural labor income. Because so many of the poor live in rural areas and rely on agriculture for their income, it is appropriate to compare poor households with rural households more generally, to see if there are differences in their asset holdings and income sources. The three major assets in poor rural economies are land, livestock, and labor. Land 26. In Tanzania, unlike many other developing countries and especially South Asian countries, access to land is not the factor that divides poor households from better-off households. Nearly 95 percent of the poor own some land, and the position is the same for the hard-core poor.'0 The spread of farm sizes is substantial, but the average amount of land per household is quite similar for poor households and rural households. The mean land per household is 4.66 hectares for rural households generally, 4.43 for poor households and 4.44' for hard-core poor households. Even though the ownership of land is quite unequal, land does not explain poverty status or the inequality of income. However, the quality of land may differ. As Figure S.7 shows, the rural poor generate much less income from crop production (36 percent) than do rural residents in general (65 percent). .They also bought their land more recently, suggesting that their land is marginal." Livestock 27. Previous writers have found a close connection in Rura I Tanzania between livestock Ad- Rural Poor ownership and income inequality. Cops 62%,- That conclusion is not supported 6 by our survey. The poor and the hard-core poor are less likely o A LiveStock 9% than better-off households, or an- Ag 19% rural households, to own T Ag a% livestock, but many do own cattle Figure S.7: Sources of Income and other animals. 64 percent of rural households own livestock, as compared to 54 percent of poor households and 56 percent of hard-core poor households. There are some differences in the quality of livestock (see Table 4. I1). Poor households have a higher mean stock of improved bulls than rural households in general (3.23 versus 2.05) and a lower mean stock of improved cows (2.67 versus 6.87). These quality factors 30 Own land includes rented and borrowed land, which accounts just for a very low percentage of the total operated land. " The pies in Figure S.7 are proportional to the income accruing to the total rural population and the income accruing to the poor (soft-core poor). 10 Executive Summary show up in lower mean values of poor households' livestock holdings and may explain differences in income received from livestock. The poor rely on livestock for a higher percentage of their income than do other rural households, but the returns they receive are lower in absolute terms. However, there is large variation in the value of livestock ownership. Thus, there are poor households who possess many cattle, and better-off households, even in the rural areas, who have hardly any livestock to speak of. Livestock ownership is obviously a factor in welfare, and in generating income, but it is not a guarantee of prosperity. Labor 28. The patterns of labor use for the poor and hard-core poor are very similar to those for rural inhabitants in general. The amount of labor that households have at their disposal is not closely associated with poverty. The poor generate a higher proportion of their income than other rural households from activities outside their own crop and livestock production - in an apparent attempt to diversify their income sources. Yet, in many cases the amount they are able to raise through these alternatives is less than for other households. Another distinctive characteristic of labor use in poor households is that children are engaged much more heavily in animal husbandry. Nationwide, what distinguishes better-off households from others is that they are more often engaged in business and non-agricultural wage labor - activities connected to urban life. PoLIcY IMPLICATIONS 29. From 1971 to 1983, Tanzania had several years during which economic growth (at least rural economic growth) was not sufficient to match the increase. Since then, the economy seems to have rebounded, but continued growth is still necessary to raise average incomes and extract many people from poverty. However, economic growth is a necessary, but not sufficient condition for reducing poverty. For instance, the evidence on the Gini coefficient for rural areas shows that increasing income can be accompanied by rising inequality and continued substantial poverty. 30. The alleviation of poverty will also crucially depend on the government's ability to finance services for the poor, especially health, education, and public infrastructures, which have an important effect on the adoption of new technology, labor productivity, and agricultural output. Given the lack of reliable and regularly updated information on economic and social well-being, government agencies and others must strengthen monitoring systems, providing policy makers with updated information on the economic status of poor people. As this survey shows, poverty is so broad-based, and so widely spread in rural areas, that it is hard to distinguish the rural poor without resorting to direct measures of income or expenditure. :However, given a tight budget constraint on public resources, it will be crucial to work harder to identify the particular characteristics of the poor. This study has started that work, and reveals the following, initial ranking of needs. Target Rural Households 31. In terms of the spatial, or geographic, pattern of poverty, all measures indicate that rural areas should be favored before urban areas and Dar es Salaam itself. Target Farmers and a Few Other Pockets of Ibor 32. The exact poverty ranking of socio-economic groups changes according to the measure used. However, the three groups most clearly poor are farmers, rural businessmen/women, and resident rural government employees. Of these three, farmers make up the bulk of the poor . Executive Summary 11 Instruments to Reach the Hard Core Pbor 33. Given the rural tilt to poverty and the abundance of natural resources in Tanzania, one way to increase the income of the poor directly is to explore options to increase returns to factors - land, labor, and capital - used in agriculture. The key instrument for raising returns to factors used in agriculture is to liberalize agricultural prices further. In addition, investing in rural primary and secondary education will increase the human capital of farmers, which has been shown to be an effective way to increase productivity. Jamison and Lau (1982), for example, found that four years of education raises agricultural productivity by 9 percent relative to having no education. Effective extension work can further increase productivity through the dissemination and adoption of new agricultural techniques. Growth in agricultural incomes is also facilitated by development of agricultural markets, which may be important in Tanzania because the regional pattern of poverty appears to be associated with uneven access to domestic and international markets. Another strategy for increasing rural incomes may be to encourage alternative employment opportunities for poor households. A theme of a recent participatory study of selected rural areas in Tanzania is that liberalization of the economy has opened many. new income-generating opportunities for the poor (Bloom et al. 1993), indicating that this process has already begun. 34. We emphasize the tentative and hypothetical nature of these options and the need to investigate carefully the potential impact of alternative policies on reducing poverty before choosing interventions.'2 The Government of Tanzania and The World Bank are planning to follow this poverty profile with a poverty assessment in fiscal year 1995, which will assess the impact of current government policies on the poor as well as alternative approaches to reducing poverty. " For a more in-depth analysis of agricultural development strategy and its effect on poverty alleviation, readers are directed to the Agricultural Sector Memorandum (World Bank 1993). The Memorandum addresses the role of extension, education, access to roads and markets, and the liberalization of marketing and processing as potential measures to raise rural income. See Table 6. p. 27 for a regional analysis of land use and potential for expansion in annual crop area. 1 BACKGROUND 1. The World Development Report of 1992 estimated Tanzania's real per capita income at US$129 in 1989, and US$110 in 1990. This makes Tanzania one of the poorest countries, not only in Sub-Saharan Africa, but also in the world. All income estimates should be treated with caution, because of the very divergent estimates that exist, but under whatever measure is chosen, Tanzania is among the handful of the poorest countries in the world. That this poverty exists side by side with a tremendous natural resource endowment is alarming. The Tanzanian government has been recognized for decades as a leader among developing countries in pursuing poverty-alleviating strategies. However, there is mounting evidence that the impact of these policies may have been less positive than was previously assumed. 2. Between the mid 1970s and the early 1980s, external shocks contributed to macroeconomic imbalances, economic stagnation and a sharp decline in per capita income. The first major oil crisis, in 1973, caused prices to quadruple, and the second one, in 1978, caused prices to double. A severe drought in 1973-74 increased food import expenditures at the same time that declining international prices for traditional exports squeezed agricultural incomes. In 1977, the break up of the East African Community provoked a disruption of trade with Kenya, and the 1979 war with Uganda put a further drain on the country's resources. However, the most important cause of the economic decline was the failure of domestic policies to generate sustained growth in per capita incomes. Inordinate and inflexible state control over the economy resulted in a stifling of economic activity, widespread deterioration of the country's infrastructure, and a regression to barter trade on parallel markets at the height of the economic crisis in 1982 and 1983. 3. From 1981 to 1983, official GDP decreased and foreign aid flows slowed as donors became reluctant to continue support without a restructuring of government's basic economic policies. High inflation, the shortage of consumer goods, expanding parallel markets, and rapidly dwindling government resources led the government to introduce, in 1982, a "homegrown" structural adjustment program. However, international support for the government's adjustment policies was only obtained in 1986, with the launching of the Economic Recovery Program (ERP) supported by the IMF with an 18 month stand-by arrangement and by the World Bank with a Multi-Sector Rehabilitation Credit. The ERP included several crucial economic reforms, such as the liberalization of key imports, a more restrictive monetary policy, and an exchange rate policy that reduced the overvaluation of the shilling. 4. World Bank dollar estimates of per capita real income suggest that standards of living have fallen markedly since 1985. However, this index is affected by the sharp devaluation of the exchange rate. Per capita GDP in volume-terms has actually risen every year since 1985. Other. equally persuasive. dollar estimates of per capita GDP exist, and they also suggest an improvement in living standards in recent years. For instance, the United Nations International Comparisons Project (ICP) estimated per capita income for 1990 at US$540 (purchasing power parity equivalent), while our own estimates of per capita income, translated into US dollars at the 1991 official exchange rate, imply a per capita income of USS280. Given the enormous disparity in estimates, it might be preferable to examine past income trends in terms of Tanzanian Shillings. Chapter 2. which looks at income and inequality. takes this approach. The figures given in that chapter suggest a substantial increase Background 13 in income since the early eighties. However, even Tanzanian Shilling estimates are not immune to data problems (mainly to do with estimation of income from home production and the use of official prices). For this reason, it is useful to refer to basic social indicators of standard of living. 5. During the 1970s, Tanzania's social indicators were often equal to or even above the averages for the region (see Table 1.1). During the 1980s, however, conditions stagnated or declined compared to earlier years and compared to the rest of Sub-Saharan Africa - despite the apparent increase in income estimated in Shillings. For instance, life expectancy is the same as it was in 1980, average daily caloric intake is slightly below its former level, and primary enrollment rates have dropped sharply. Infant mortality in Tanzania is now over 110 per thousand - one of the highest rates in the world. Moreover, this is expected to increase if the AIDS virus is not contained.' As a recent World Bank study on AIDS suggests, the number infected with the AIDS virus will reach 5.8 to 17.4 percent of the by the year 2010, up from 1.4 to 5.3 percent currently.2 Table 1.1: Social Indicators for Tanzania and Sub-Saharan Africa 1980 1985 1990/91 Sub- Saharan Sub-Saharan Sub-Saharan Tanzania Africa Tanzania Africa Tanzania Africa Per Capita Income 284 582 309 491 110 340 (1991 US$) Life Expectancy at Birth 47 47 48 48 51 (Years) InfantMortality 122 127 117 118 115 107 (Per 1000) Average Daily caloric Intake 2244 2107 2229 2040 2206 2120 (Kilocalories per capita) Primary School Gross Enrollment 93 70 72 68 63 69 Rate Source: World Development Reports (various issues); African Development Indicators; Social Indicators of Development 6. The World Bank is determined to improve the impact of its lending activities on poverty. To assess the causes of poverty and to define an overall poverty-reduction strategy, the Eastern Africa Department of the World Bank prepares poverty assessmenty of all the countries in. the region, including Tanzania. The critical, empirical part of a poverty assessment is a poverty profile. In addition, work underway on the Agricultural Sector Memorandum (ASM) for Tanzania includes work on rural poverty. The Tanzanian Government is committed to reducing poverty in its country as fast as possible; this profile provides a statistical picture that can help in the formulation of strategies to reduce poverty. "Tanzania: AIDS Assessment and Planning Study," The World Bank, Washington D.C., 1992. 2 This will have major effects on the demographic composition of the population, and of the work force. "I...]The work force will become younger.[...] and less experienced, and will have less education and training.[...] The economy will be adversely affected. [....J GDP will grow more slowly[...], per capita GDP also will be impacted, but more moderately[...]." (World Bank 1992b). 14 Background 7. A poverty profile describes quantitatively people living in poverty - who they are, where they live, what they consume, and where their income comes from. A poverty profile analyzes the depth and nature of poverty, providing detailed information on the standards of living of poor households. A poverty profile is a necessary building block for an effective, cross-sectoral poverty- reducing strategy and supplies vital guidance for the targeting of programs. 8. The quantitative analysis for this poverty profile is based on a 1991 household survey conducted by the Cornell Food and Nutrition Policy Program and the Economic Research Bureau (ERB) of the University of Dar es Salaam. The Cornell/ERB Household Survey is one of the most recent, "post-adjustment," nationally representative household budget surveys in Sub-Saharan Africa. The survey covered 1,046 households in rural and urban areas, drawn from the National Master Sample of the Bureau of Statistics to ensure a sample that was nationally representative. The sampling scheme and questionnaire covered production, consumption, income, and assets for rural villages, urban areas outside Dar es Salaam, and Dar es Salaam itself. Actual surveying was undertaken between August 1991 and January 1992 and was funded by a USAID grant. 9. The report is organized as follows. Chapter 2 starts with a brief discussion of the approach taken to measuring poverty, then it explores income growth and changes in inequality over time. Chapter 3 presents the findings on the incidence, depth, and severity of poverty and discuss the special characteristics of poor people and households, in terms of demography, socio-economic group, location, and region. Chapter 4 presents further details about the lives of the poor in Tanzania: their sources of income, their assets, and their ability or inability to fulfill basic needs. Chapter 5 presents the major conclusions of the analysis and the associated policy implications. Attached to the report are three technical appendices. Annex A gives more detail on the sampling methodology and data collection. Annex B presents the methodological background to different measures of poverty, and Annex C includes additional tables. 2 THE DISrRIBUTION OF INCOME IN TANZANIA MEASURING AND DETERMINING POVERTY 1. Three major decisions have to be taken to construct a poverty profile. First, one has to choose some criterion by which to rank households. Second, one has to choose a poverty line by which to distinguish poor from better-off households. Third, one has to decide what aspect of households below the poverty line to look at: do we care about the number of households below the line (incidence), or how far below the poverty line their consumption lies (depth), or how many of them are extremely far below the poverty line (severity)? In other words, what poverty index should be used? 2. Economic theory, and common sense, suggest ranking household welfare by the level of their consumption of goods and services. However, households may differ in their needs because of family composition; for instance, families with more children may need less food to survive than an equal- sized household with more adults. To control for differences in household composition, this study ranks households by consumption per "adult equivalent." Annex B discusses how to construct the "adult equivalent" index, which is used to adjust consumption for different family composition. Consumption is taken to be equal to expenditure. Thus, households are ranked according to the level of their monetized expenditure, deflated by the number of adult equivalents in the household. 3. Total household expenditures include the value of home-produced consumption. Efforts were made to control for regional differences in prices.' The ideal standard of living criterion would also take into account the consumption of leisure, the consumption of public goods and services, and the present value of consumption across several periods. However, these adjustments are beyond the scope of the survey on which this poverty profile is based. 4. Information on expenditures was collected from the household survey. The questions (and recall periods) were designed so as to minimize the probability of inaccuracy in the measurement of total yearly expenditure. To compute the value of home production, we used the average price for the village or region as estimated by this survey. One gap in the expenditure data is the absence, for consumers occupying their own homes, of implicit rent or shelter expenditure. However, the major conclusions of the study are unlikely to be affected by this data omission, in part because rent is never more than 15 percent of expenditure and in part because the differences in expenditure between rich and poor are so great that they would not be removed by the inclusion of rent. 5. Once adult-equivalent household consumption has been measured, it can be compared against a poverty line which indicates the consumption level below which households are defined as poor. Currently. no official poverty line exists for Tanzania. There are several alternatives available for choosing such a line. First, one can choose a poverty line that cuts off a given percentage of the income distribution of the population (a relative poverty line). Alternatively, one can choose a basket See Annex B for a discussion of the construction of a Thornquist-Theil Index for food prices by region. The mdex showed relatively little difference in food prices between rural and urban areas. 16 The Distribution of Income in Tanzania that includes all the goods necessary to reach the standard of living that is regarded as essential by the customs of the country (an absolute poverty line). Because the choice of a given poverty line will affect the findings of any analysis, this study uses two poverty lines as an in-built sensitivity test. Using different poverty lines may highlight different characteristics of poor people; many studies distinguish the extremely poor within the broader category of poor people. Such extremely poor groups should constitute the primary target of poverty alleviating interventions. For all these reasons, two poverty lines were defined. a. Poor (soft-core poverty). The first poverty line (relative poverty line) was set at 50 percent of the mean adult-equivalent expenditure level (or 75 percent of mean per capita yearly expenditure). In other words, households whose members each consume less than half of the average expenditure level (adjusted for household composition) were defined as poor or soft-core poor. This corresponds to a value of Tsh 46,173 or approximately US$227 per capita per year in 1991 prices, at the average exchange rate of Tsh 203/US$.2 This is a relative poverty line, but well below the "one dollar a day" absolute line that is commonly used to define poverty. It actually amounts to about US$0.62 a day at the average exchange rate of Tsh 203/US$ at the time of the survey. b. Very Poor (hard-core poverty). The second poverty line (absolute poverty line) was taken from the poverty line computed by an International Labor Organization (ILO) study in 1982, evaluated at prices in 1991 for this poverty profile. The 1982 ELO poverty line was "based on an estimate of the minimum income essential to meet the basic need of food, and then topped up to take account of shelter, household, goods, etc." (ILO, 1982). Thus, this second poverty line is an absolute measure of poverty. The monetary value of the poverty line, estimated using a market basket of prices, is Tsh 31,000 or US$152 per adult-equivalent. This line can also be interpreted as a relative poverty line. The amount of 31,000 Tsh is approximately one-third of the mean adult-equivalent expenditure level (or half of mean per capita yearly expenditure). Households whose adult-equivalent expenditure is less than Tsh 31,000 per person constitute what we define as the very poor or hard-core poor. CHOICE OF POVERTY INDEX Incidence of Poverty 6. The simplest way to analyze poverty is the head count ratio. It counts the number of people in the country with expenditures below the poverty line. In other words, the head count shows the prevalence or incidence of poverty, but it does not say anything about the depth of poverty. However, it is easy to interpret and calculate and is therefore a popular measure. Most of the tables in the report give information on this measure. 2 The value of the purchasing power parity (PPP) is for 1991 equal to 40.621 TshIUSS. and the exchange rate computed by the Atla% mcthd as approximately 203 Tsh/USS (World Bank 1992). In 1991. the average official exchange rate wA& 230 TshfUSS at the time of the survey. The average official rate during the calendar year of 1991 was 219.16 Tsh't'SS. and the average official exchange rate during the fiscal year of 1991 was 201.8 TshlUSS. The Distribution of Income in Tanzania 17 Depth of Poverty 7. The poverty gap index measures the gap between the actual income of poor households and the poverty line. It indicates the depth of poverty, or how far below the line the poor fall, but it is not sensitive to the severity of poverty, in the sense that each additional Shilling that a person's income falls below the poverty line is given equal weights in terms of utility or welfare. The sum of all individual poverty gaps in a is the national poverty gap, or the minimum amount of transfers necessary to bring all households (or people) just up to the poverty line if perfect targeting were possible.' Severity of Poverty 8. The FGT index is commonly known as the P. index of poverty.' This index highlights the severity of poverty (when ci 2), here meaning the number of households or people in dire poverty, by weighting the shortfall of a household's income below the poverty line more heavily the further below the poverty line they fall. INCOME LEVELS 9. This study estimates 1991 per capita income and per capita expenditures at Tsh 55,369 and Tsh 61,564, respectively. At the average exchange rate of Bh 203/US$ in 1991 (during the field work for the survey), this comes to US$273 and 303, about 150 percent higher than the per capita GNP estimate for 1990 that was reported in the Rbrld Development Report. 10. The difference in the numbers may cast some doubt on our own estimates of per capita income and expenditure. However, there are several factors that argue in favor of the higher values calculated in this study. a. The ILO (1982) has reported that estimates for agricultural income in the 1970's were biased downwards by problems in estimating the value of crops sold through unofficial channels - a practice which was particularly prevalent in the seventies and early eighties. Thus, the real growth rates in income since the seventies, while still high, are probably less than the rates implied by the data in the table on the following page. For instance, calculated from the data in Table 2.1, the annual growth rate in rural income since 1976/77 has been approximately 12 percent. While growth has occurred, this rate seems exceptional and, by itself, would call into question the accuracy of the base year income, the end year income, or both. b. Our expenditure estimates are similar in magnitude (once in-kind expenditures are incorporated) to estimates presented independently in the article by Abel-Smith and Rawal (1992) on health costs and expenditures. Using their data, and assuming that in-kind expenditures are 30 to 40 percent of total expenditures, average per capita expenditure is estimated to be Tsh 67,626, very close to our figure. Perfect targeung means that each houschold or person would be given money just sufficient to bring their consumption up to the level of the poverty line and no one over the poverty line would receive a transfer. 'Named after its authors Foster, Greer,and Thorbecke (1984). The formula for the index is given in Annex B. 18 The Distribution of Income in Tanzania C. Overall averages for the income and expenditure calculated from the Cornell/ERB survey for this study are very close to each other, although there are many individual household differences. Average income exceeds average expenditure, with most of the resultant savings concentrated in Dar es Salaam, other cities, and among the better off, as would be expected. If low values for income are removed, income greatly exceeds expenditure (compare Table 3.1 and Table 4.14).. These consistency checks are encouraging for several reasons. First, these income and expenditure calculations are independent; that their aggregate values match fairly well is an indicator that expenditures are probably not overestimated. Second, in household surveys, income is typically underestimated; that our estimated of average income per capita exceeds average expenditure for some groups suggests that our apparently high estimates of expenditure (upon which most of the analysis is based) are probably conservative. d. The expenditure measure does not include imputed housing services and thus is known to be an underestimate of actual consumption. e. Finally, it is worth noting that whatever the true level of average per capita income and expenditure in Tanzania, this poverty profile is less concerned with absolute levels than with relative levels. Comparisons across the expenditure distribution and geographical areas make up the bulk of the work, and these comparisons are probably immune to changes that might be made to the overall averages. Especially for the soft-core poverty index, which is a completely relative measure, the identity of who is poor is unlikely to change much (if at all) even if the level of overall expenditure is adjusted.' 11. Comparing average per capita income estimates in this report with those presented in earlier studies, it appears that both rural and urban incomes improved substantially between the 1970s and 1991 and that there has been a very large growth in rural incomes since 1982/83 (see Table 2.1 and Figure 2.1). Apparently the launching of the Economic Recovery Program in 1986 is showing positive income effects, although this development is not yet reflected in basic social indicators, like infant mortality or basic education. 12. For the rural areas, we found a yearly per capita income for 1991 of.Tsh 36,250, suggesting large improvements in rural incomes over the last decade. Real rural per capita income shows the following broad trends: a. Rural incomes stagnated between 1969 and 1976. The Household Budget Survey (HBS) of 1969 found a per capita income of Tsh 6,620 per annum, while the HBS This long discussion is the result of a debate among researchers about the apparently high expenditure figures estimated in this report. Our inferences might be wrong due to the small sample size. the weighting scheme. judgements made when performing the calculations, and so on. To the extent possible. all assumptions, weights, and calculations have been checked and rechecked. A recent set of independent calculations of expenditures using the same data by Professor Alexander Sarris and Platon Tinios. University of Athens. for the Cornell Food and Nutrition Center. estimated average expenditure per adult equivalent at Tsh 47.217 for the rural areas. Tsh 107.221 for the urban areas. Tsh 151.394-for Dar es Salaam. and Tsh 49.620 for Tanzania (income was not estimated). Average adult equivalent expenditures calculated for this report are Tsh 77.246. Tsh 108.988. Tsh 158.965. and 61.564 respectively. The Distribution of Income in Tanzania 19 Table 2.1: Per Capita Income in Tanzania since 1969, in 1991 Prices 1969 1976/77 1979/80 1982/83 1991 Rural 6,620 6,860 8,570 10,280 36,250 Urban 74,980 23,890 n.a. n.a. 100,222 Tanzania n.a. n.a. n.a. n.a. 55,370 Note The values are rounded. To convert the figures to 1991 prices, the inflation rate implicit in the GDP deflator was used. Source: Tanzania Household Budget Survey, Collier et al. (1986), Bevan et al. (1990). of 1976 showed a per capita income of Tsh 6,860 per annum.6 Thousands of Shillings 120 100 - 0 80 - o Urban 60 . Rural 20 0- I I II 1969 1976/77 1979/80 1982/83 1991 Figure 2.1: Changes in Per Capita Income Over Time b. Rural incomes improved modestly between 1976 and 1982. Using only rural data. Collier et al. (1986) concluded that the 1980 per capita income was Tsh 8.570 (at 1991 prices), and Bevan et al. (1990) estimated that rural per capita income was Tsh 10.280 in 1982. * As noted in the abic. all income estimates are recalculated at 1991 prices to show real movements. These values were taken from the Internaional Labor Organization study (1982). 20 The Distribution of Income in Tanzania c. Rural incomes increased substantially between 1982 and 1991. 13. The following broad trends in real urban per capita income can be inferred from earlier estimates: a. Urban incomes declined between 1969 and 1976. The 1969 Household Budget Survey found an urban per capita income of Tsh 74,980, while the 1976/77 Household Budget Survey found an urban per capita income of Tsh 23,890. b. Urban incomes increased substantially between 1976 and 1991 and, by the end of the period, urban per capita income had reached Tsh 100,222. It is likely that urban income followed the same trend as rural income between 1976 and 1991, but we have no observations for the years 1980 and 1982/83. INCOME INEQUALITY 14. Since there is no consensus on the ideal inequality index, we calculated a number of different measures: the coefficient of variation, Theil's Index, and the Gini coefficient.' These are presented in Table 2.2. The result confirmed by all indicators is that inequality is most pronounced in rural areas. This finding awaits further analysis, but it may point to large regional (i.e., inter-village) differences in welfare. Table 2.2: Some Indicators of Inequality in Tanzania, 1991 Location Income Group All Tanzania Rural Urban DSM8 Poor Very Poor Median Household Adult 49,859 40,498 71,391 125,781 23,877 17,519 Equivalent Expenditure Coefficient of Variation 1 .89 2.41 1 .20 0.81 0.52 0.49 Gini Coefficient 0.57 0.60 0.48 0.39 0.30 0.28 Theil Coefficient 0.49 0.57 0.35 0.22 0.14 0.14 Note: a. DSM stands for Dar es Salaam ' See Annex B for a definition of several inequality indices and Lambert (1989) for a discussion of the pros and cons of these indices. The Distribution of Income in Tanzania 21 15. : An important objective of Ujamaa - Tanzania's version of socialism - was to reduce inequality. However, a time-series comparison (Table 2.3 and Figure 2.2) suggests that income inequality actually increased between 1969 and 1991. In 1969, according to the Household Budget Survey, the national Gini coefficient' was 0.39. In spite of the implementation of Ujamaa in the Table 2.3: Gini Coefficient for the Distribution of Income in Tanzania, 1969-91 1969 1976/77 1991 Change in Inequality Rural .30 .37 .60 Urban .51 .46 .46 DSM n.a. n.a. .39 n.a. Urban outside DSM n.a. n.a. .48 n.a. Tanzania .39 .44 .57 t 1970s, the Household Budget Survey found a higher Gini coefficient, of 0.44, in 1976/77.9 According to our survey, the Gini coefficient rose again, to 0.57 in 1991. However, income inequality within the poor and very poor strata is: lower than the national average. In other words, the income distribution among poor people is more equal than the income distribution over the whole 16. Rural income inequality has become worse than urban inequality only in recent years. In 1969, the Gini Coefficient was 0.51 for urban areas and 0.30 for rural areas but, by 1976/77, the gap had narrowed to 0.46 for urban areas and 0.37 for rural, and by 1991 the respective Gini coefficients were 0.60 and 0.46. The trend seen between 1969 and 1976 - of diminishing urban inequality and increasing rural inequality - continued between 1976 and 1991.10 However, the increase in rural inequality more than offset the decline in urban inequality, causing an increase in inequality for the country as a whole. 17. In Table 2.4 annual adult-equivalent expenditure and annual per capita expenditure are shown by quintile. In other words, quintile one for the column "all" is the poorest twenty percent of all households in Tanzania. The value of adult-equivalent expenditure shown is the mean adult- equivalent expenditure for the bottom twenty percent. The table shows that the average adult- equivalent expenditure of the richest group (quintile 5) is 24 times (278,968/11,887 = 24) greater than that of the poorest group (quintile 1) for the country as a whole, about 26 times greater for the rural areas, but only 7 times greater for the poor, and about 6 times greater for the very poor. These ratios are respectively 26, 27, 7, and 8 when per capita expenditure, rather than adult-equivalent expenditure, is used. Inequality is not just a matter of comparing the first and fifth quintiles of the . There is inequality across the entire distribution, as indicated earlier by the statistics in Table 2.2. Whatever the index used, the inequality ranking is the same and yields the same insight. The highest The Gini coefficient is an index of inequality. When an income distribution is perfectly equal the Gini coefficient reaches the value of zero. If all income is concentrated in the hands of one agent, the Gni coefficient is equal to one. o These comparisons should be taken with caution. since they cast the use of different methodology. However, the magnitude of the differences leads us to believe that in fact there was an increase of inequality over time. 1o The increase in rural inequality is in conflict with the past policies in Tanzania. but consistent with Kuznet's hypothesis that when income increases. it will be firstly accompanied by an increase in inequality. 22 The Distribution of Income in Tanzania Gini (Higher Means Greater Inequality) 0.6- 0.5- National 0.4-- Urban ---- Rural 0.3 0.2 1969 1976/77 1991 Year Figure 2.2: Changes in Inequality over Time Table 2.4: The Distribution of People by Adult Equivalent Household and Per-Capita Annual Expenditure Annual Expenditure Per Adult Equivalent Annual Expenditure Per Capita (Mean in TSH) (Mean in TSH) Quintile All Rural Poor Very Poor All Rural Poor Very Poor Poorest 1 11,887 9,613 6,350 4,347 7,362 5,942 4,045 2,679 2 29,300 23,576 16,234 12,544 18,068 14,463 10,149 7,781 3 50,054 39,742 23,831 17,610 31,848 24,483 14,626 10,965 4 86,380 64,598 32,056 22,891 56,422 41,952 19,824 14,156 Richest 5 278,968 247,516 41,474 28,208 193,750 158,103 27,294 20,065 Richest/ 24 26 7 6 26 27 7 8 Poorest All 91,500 77,200 24,000 17,100 61,600 49,620 15,200 11,000 inequality exists in rural areas, and the lowest degree of inequality is among the poor and very poor. The Distribution of Income in Tanzania 23 1.0 0.9 æ 0.8 (1)- Poor (2) - Dar es Salaom (3) - All Tanzania c0.7 (4) - Rural 0- x 0>0.6 0 0.5 -(1) 2) 0.4 c .2 (4) L- 0.3 . 0 0.2 - 0.1 0.0 - 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 09 1.0 Proportion of population Figure 2.3: Lorenz Curve for the Welfare Distribution 24 The Distribution of Income in Tanzania 18. Figure 2.3 displays the Lorenz Table 2.5: An International Comparison of curves" for Tanzania as a whole, rural Income Distribution Using the Gini Coefficient areas, Dar es Salaam (DSM), and poor Country/Area Gini Coefficient people, respectively." Since the Lorenz Tanzania .57 curve for the poor lies above the national nzania Kenyan.50-.65 Lorenz curve, the income distribution for Ya b the poor can be said to be less unequal Cugoslaia' .20 than the income distribution for the whole population. Nonetheless, because the Hungary' .23 mean income of the poor is lower than Bulgariab .22 the mean income of Tanzania, one cannot USAc (1979) .34 conclude that one distribution dominates Swedene (1981) .23 the other, in terms of the well-being of West Germany' (1981) .28 people within that income distribution." Japand (1985) . 35 On the other hand, given the position of Notes: a. Several studies, reporting values from late 60s the Lorenz curve for DSM vis-A-vis the through early 80s b. Milanovic (1992) rural areas, and the fact that mean income c. Reported in Bishop, Formby, and Smith (1991) in DSM is higher than in rural areas, the d. Oshima (1991) welfare distribution in DSM can be said to dominate the welfare distribution in rural areas, according to the Lorenz dominance criterion. The same holds for the comparison of other urban areas outside of DSM with the rural areas of Tanzania. 19. When comparing Tanzania with other countries, one can see that its income inequality is similar to other countries in the region, namely Kenya, but substantially higher than in the industrialized economies, as well as in the formerly socialist economies of Eastern Europe (see Table 2.5). " For the Lorenz Curve in Figure 2.3, the diagonal line indicates one possible limit: perfect equality in the distribution of expenditure, so that each person has exactly the same level of expenditure. Increasing distance from the diagonal indicates greater inequality. 92 For a discussion of Lorenz and the Generalized Lorenz Curves see Annex B, where the Generalized Lorenz Curve is also displayed (Figure B.1). The curves for urban areas, and for the very poor, are not presented here, because they are not qualitatively different from the curves for DSM, and the poor. respectively. 3 One says that a distribution of income/welfare, say A. dominates another distribution, say B. according to the Lorenz dominance criterion, when 'the bottom 100p [where p is any value between 0 and 1] percent of tncome recipients in distribution A have a greater share i of total income than do the corresponding group in distribution B. and this is true for every p between zero and unity." (Lambert 1989:34) 3 CHARACTERISTIcs OF THE POOR IN TANZANIA LEVEIS OF EXPENDITURE 1. Table 3.1 displays the mean per capita expenditure and the mean adult-equivalent expenditures for the Tanzanian population; subsets: the better-off, the poor, and the very poor; and geographic location: rural, urban excluding Dar es Salaam, and Dar es Salaam itself (DSM). The following characteristics emerge:I Table 3.1: Yearly Expenditures According to Type of Household, 1991 Income Group Location All Tanzania: Better off Poor Very Poor Rural Urban DSM Per Capita Household Expenditures (in 61,564 103,454 15,223 10,937 49,620 78,542 112,894 Tanzanian Shillings) Adult-Equivalent Household Expenditures 91,509 152,449 24,093 17,173 77,246 108,9889 158,695 (in Tanzanian Shillings) Population (Percent) 100 48.9 51.1 35.9 73.5 17.4 9.1 9.7 Households (Percent) 100 52.4 47.6 30.6 70.1 20.2 a. Over half of the population is living in poverty, and a third are living in hard core poverty. However, as Table 3.1 shows, just under half of all households are living in poverty. b. About seventy percent of Tanzanians live in rural villages, twenty percent in urban areas outside of Dar es Salaam, and ten percent in Dar es Salaam itself. c. Overall, average per capita expenditure in 1991 was Tsh 61,564 and average adult- equivalent expenditure was Tsh 91,509. d. For the poor, average per capita expenditure in 1991 was Tsh 15,223 and average adult-equivalent expenditure was Tsh 24,093. For the hard core poor, average per capita and equivalent expenditure was Tsh 10,937 and Tsh 17,173, respectively. e. Average expenditures per capita and per adult-equivalent for better-off households are about six to seven times higher than for the poor, and about nine to ten times higher than for the hard core poor. Table A.6. Table A.7. and Table A.8 display the sample size used in some of these calculations. 26 Characteristics of the Poor in Tanzania f. Average rural expenditures are Tsh 49,620 per capita and Tsh 77,246 per adult- equivalent; average urban (outside of DSM) expenditures are Tsh 78,542 and Tsh 108,988, and for DSM itself they are Tsh 112,894 and Tsh 158,695, respectively. g. Urban people (in and outside of Dar es Salaam) are about twice as rich as rural people, even when the difference in rural vis-i-vis urban family composition is taken into account. Households living in DSM have an estimated per capita expenditure that is approximately 1.5 times higher than in the other urban areas, and 2.3 times. higher than in rural areas of Tanzania. An estimated food price index (see Annex B) found little difference in food prices between rural areas, urban areas outside Dar es Salaam, and Dar es Salaam itself, implying that real urban income is unambiguously and substantially higher than rural income. EXPENDTURE SHARES 2. Patterns of food expenditures and expenditures on education and health reveal serious poverty related problems. Food Share of Expenditures 3. The share of food in total expenditures decreases from the very poor to the better-off, reflecting the normal propensity of the poor to spend more of their budgets on essential survival items. However, the share of food expenditures in total expenditures is quite high for both the better- off (70.8 percent) and the poor (75.6 percent), indicating a generally low standard of living. As can be seen from Table 3.2 and Figure 3.1, the share of food in the total budget is highest in rural areas, rising from 59.3 percent in the capital city to 77.4 percent in rural villages. Table 3.2: Expenditure Patterns by Household Type Income Group Location All Tanzania Better-Off Poor Very Poor Rural Urban DSM Percent Shares of Expenditures, by Item Food 73.1 70.8 75.6 77.2 77.4 63.6 59.3 Health 2.6 2.8 2.1 2.1 1.9 4.3 2.9 Education 0.66 0.70 0.61 0.62 0.57 0.90 0.80 Other 23.6 25.7 21.69 20.08 20.12 31.2 37.0 Sources of Food Expenditures (Percent) Market Purchases 66.8 65.9 67.8 64.8 58.3 82.1 97.2 Home Production 33.1 34.0 32.2 35.1 41.7 17.9 2.98 4. The share of food expenditure accounted for by consumption of own production is quite high (33.1 percent) for Tanzania on average, and is similar for both better-off (34.0 percent) and poor households (32.2 percent). The very poor record the highest share - 35.1 percent. The reduced importance of home production in the urban areas is consistent with expectations. In rural areas, 41.7 percent of total food consumption is from own production, while in the capital city it is as low as 3.0 percent. 5. Table 3.3 adds further insight into the food budget of the population. The main food item is cereals and, on average, 48 percent of all food expenditures go to cereals. The most striking point Characteristics of the Poor in Tanzania 27 Rural Urban DSM I 0 20 40 60 80 100 Percentage Share of Expenditure Food L J Health Education Other Figure 3.1: Expenditure Shares by Location Table 3.3: Composition of Food Expenditures Income Group Location Share All Tanzania Better-Off Poor Very Poor Rural Urban DSM Cereal 48.33 43.77 53.51 49.17 50.98 43.41 39.43 Vegetables 17.67 19.28 15.85 14.10 17.20 18.66 19.08 Meat and Fish 14.81 17.68 11.55 10.68 14.37 15.60 16.29 Other 19.19 19.27 19.09 26.05 17.45 22.33 25.20 Total 100 100 100 100 100 100 100 Percent of Consumption from own production Cereal 45.18 45.03 45.37 49.32 57.40 25.17 1.79 Vegetables 32.52 33.32 31.47 34.01 . 43.68 11.56 3.41 Meat 15.48 18.94 10.69 11.53 22.38 3.17 1.15 in the food patterns of the poor is the low percentage of meat consumption that comes from home production (11.5 percent), and the much lower share of vegetables in total food expenditure (about 62 percent of the vegetable share of the better-off).. The cereal share in the total food budget is 51 percent in rural villages, 43 percent in urban areas outside DSM, and 39 percent in DSM. In rural villages more than 55 percent of cereals consumption is from own production. For vegetables and meat, own production accounts for 43.7 percent and 22.4 percent of consumption, respectively. 28 Characteristics of the Poor in Tanzania Education and Health Shares of Expenditures 6. As Table 3.2 shows, the shares of education and health in total expenditure are consistently low across all the segments, with the share of income going to education never exceeding one percent. According to the African Development Indicators, the comparative Sub-Saharan expenditure shares for 1980-85 were: food (62 percent), clothing (12 percent), rent and utilities (8 percent), medical care (1 percent), education (5 percent). It is possible that expenditures related to education, such as uniforms and bus journeys, are included in other expenditure categories (clothing and transportation). Other explanations for the small expenditure shares include a low rate of attendance at school (see Table 4.4) or the substitution of public spending for private household spending on education. The share of spending on health care is high relative to the average for Sub-Saharan Africa and relative to the share of expenditures going to education. As a result, the ratio of education to health expenditures in Tanzania is 0.25 versus 5 for Sub-Saharan Africa. The implicit mix of private social spending reflects extremely low private investment in education. BASIC DEMOGRAPHIC CHARACIERISTICS Households 7. The following observations emerge from this study (see Table 3.4):2 Table 3.4: Demographic Characteristics by Type of Household Al Income Group Location Tanzania Better-Off Poor Very Poor Rural Urban DSM Household Size 6.31 5.88 6.79 7.06 6.61 5.42 5.96 Dependency Ratio (Percent) 1.15 . 1.01 1.31 1.35 1.25 1.03 0.75 Average Age of Household Head 45.68 42.86 48.85 51.31 47.33 42.45 40.41 Female Headed Households (Percent) 9.4 9.5 9.3 8.2 5.8 19.4 13.1 Average Number of Children Younger Than 15 2.76 2.46 3.09 3.19 2.97 2.25 2.35 Average Number of Children Younger Than 18 3.44 3.04 3.88 4.03 3.68 2.82 3.03 Average Number of Adults Older Than 64 0.199 0.167 0.235 0.259 0.222 0.185 0.057 a. The average family size is 6.31 members.' Poor households are somewhat larger than better-off households - the average family size of poor households is 6.79 people and the average family size of better-off households is 5.88 people. Poor households also tend to have more children - the average is 3.88 children for poor These results compare favorably with the preliminary unweighted results from the 1991 Household Budget Survey conducted by the Bureau of Statistics. They estimate the family size to be 6.6 people, and the distribution of the population by age and gender is very similar to the estimates reported here. I This is higher than the family size found in the 1988 census (with a different definition of household membership) but conforms to the findings presented by Abel-Smith and Rawal (1992). The 1991/92 Demographic and Health Survey reported an average family size of 4.9, but its sample was limited to households containing at least one woman of childbearing age. Characteristics of the Poor in Tarizania 29 and 4.03 children for very poor households versus an overall value for Tanzania of 3.44 children. b. Tanzania is. Percent of Population characterized by 3o a very young 25 (Figure 3.2). About 43 percent 20- of the population is younger than is- 15 years, and less than 5 10 percent of- the population is older than 64 o years.4 0-4 5-9 10-19 20-29 30-39 40-49 50-59 60+ = Men Women c. Poor households tend to be headed igure 3.2: Gender and Age Distribution of the by older people Population and have a higher average age of all members. These findings suggest that poverty is associated with a lower number of economically active people and large families (see Figure 3.3). As Lipton and Ravallion (1992:40) put it, "heavy.. .child [burdens and] poverty often go hand in hand." d. Most larger Percent of Population households are 30 found in rural 25 - villages - 84.5 percent of 20- families with 11 or more people 15- live in the rural a re as . In addition, larger I families account for a higher 0 percentage of 0-4 5-9 10-19 20-29 30-39 40-49 50-59 60* households living Better Of = Poor below the poverty line than Figure 3.3: Age Distribution of the Population by above it. Poverty Status 'The Social Indicators of Development 1990/1991 reports comparable figures for Tanzania: 46.7 percent of the population is aged 0 to 14. and people aged 65 or older account for 3.1 percent of the population. 30 Characteristics of the Poor in Tanzania Family Types 8. The two predominant family structures in Tanzania are the monogamous couple with children and the monogamous extended family. Together they account for 74.4 percent of all households. Most polygamous families with children (93 percent) live in rural villages. Whereas polygamous families account for more than 10 percent of total rural households, they account for just four percent of urban households living outside DSM, and also about four percent of households living in the capital city. However, more than 50 percent of the households in DSM are extended, which also allows them flexibility in labor supply. 9. Single adults account for 1.6 percent of all rural households, 8 percent of urban households, and 6.3 percent of households living in DSM. Single adults with children account for less than 6 percent of all households nationwide. These families live predominantly in rural areas (49 percent), and in the urban areas outside DSM (44 percent). While less than 10 percent of all households are headed by a female, more than 50 percent of these female-headed households contain only a single female adult, either with or without children. Thus, female headed households have relatively few adult earners and 72 percent of all single parent families are headed by a woman. THE GENDER DIMENSION OF POVERTY 10. Empirical evidence from other countries (including Ghana) suggests that female-headed households in rural areas are disproportionately poor.' However, the results in Table 3.4 suggest that female-headed households are fairly evenly distributed across the poor and the better-off households. a. In Tanzania, 9.4 percent of all households are headed by a female, but only 9.3 percent among the poor, and 8.2 among the very-poor households (See Table 3.4). b. The poverty indices calculated by the gender of the head of the household suggest that the pattern of poverty among female headed households and male headed households is very similar and is not sensitive to the poverty line chosen (see Table C.1). c. In rural areas, Urban Rural only 5.8 percent of households are Female female-headed, 18% as opposed to Female 19.4 percent in Male urban areas Male outside of DSM, e3% v and 13.1 percent in DSM. The Figure 3.4: Gender of the Head of the Household di ffe ren ce between rural and urban areas may be explained by a high divorce rate or a high migration rate of single mothers from rural to urban areas (see Figure 3.4, which includes Dar es Salaam together with other urban areas under "Urban"). 'See Due and GLadwn (1991). Haddad (1991). and Northrop (1990). Characteristics of the Poor in -Tanzania 31 11. Per capita household expenditures are higher in female-headed households than in male- headed households. However, given the different distribution of household types in rural and urban areas, the comparison between male and female heads of households is not straightforward. Urban incomes are higher, and female headed households are mainly found in urban areas. The locational differences may also account for differences with respect to access to land and the share of rent in total budget. 12. Although female-headed and male-headed households are almost equally likely to be poor, - we cannot come to any definite conclusion on the gender differentiation of poverty. The survey classified as "poor" anyone belonging to a poor household, but did not collect data on individual income shares within the household. However, some survey data was collected on an individual basis, to assess the level of education and literacy. These indicators suggest that females are doing slightly worse than males. The evidence from other studies is mixed. Appleton et al. (1991), found no difference between men and women with respect to people falling ill during a given period, or with respect to the length of the illness. On the other hand, Sender and Smith (1990) report that when surveying, "several households contained women and children without shoes, or any adequate clothing, while the husband/father had a watch, several sweaters, pairs of shoes, and a coat." DEPTH AND SEVERITY Measures 13. Because different measures may highlight different aspects of poverty, we next look at poverty in Tanzania according to the three different measures outlined at the beginning of this chapter. Table 3.5 displays the different poverty indices, where P0 is the standard head count measure, P, is the poverty gap measure (indicating depth), and P. is the FGT index (indicating severity). The table shows the extent of poverty under all three measures, in rural areas, urban areas outside DSM, and within DSM. 14. Let us look in more detail at the first line (rural areas of Tanzania) of Table 3.5. The first entry shows that, given a poverty line of approximately Tsh 46,173 a year, 59.1 percent of all people living in rural households were poor; in other words, the incidence of poverty (Po ) in rural Tanzania is 59.1 percent. Moving along the rural Tanzania row, we see that the value of the P, index is 29.9 percent. This value can be interpreted as a rural poverty gap. It tells us that if perfect targeting were possible one would require a minimum transfer payment of Tsh 13,806 (.299 multiplied by 46,173) per year and per person to eradicate rural poverty in Tanzania. The limitation of the P. index is that it does not show the severity of poverty. For instance, one could have two households with income of 100 units each, or another extreme situation where one household had no income and the other had an income of 200 units. Both situations yield the same value for P,, and the same average income. Nonetheless, poverty is clearly more severe in the second situation. The P2 index provides information on the severity of poverty by giving higher weight to people who have incomes further below the poverty line. 15. While the total number of people in poverty is large - 51.1 percent of the total population - the poverty gap is relatively small (24.9 percent). If perfect targeting were possible, a transfer of Tsh 11,497 Tsh per year and per person (46,173 multiplied by 0.249) would be required to remove poverty. This corresponds to 12.5 percent of the mean adult equivalent expenditure. If the lower poverty line is used, the minimum transfer necessary to eliminate poverty is Tsh 4,867, which corresponds to 5.3 percent of the mean adult equivalent expenditure. In nominal terms, the national poverty gap is Tsh 306 billion at the soft core poverty line, and Tsh 130 billion at the hard core 32 Characteristics of the Poor in Tanzania Table 3.5: Three Different Poverty Indices and Two Poverty Lines, by Location of Household Head Count (P) Depth (P,) Severity (P2) Population Soft Hard Soft Hard Soft Hard Share Villages 59.1 44.1 29.9 19.6 19.2 11.8 73.5 Urban Outside DSM 39.3 17.9 15.1 6.9 7.8 3.6 17.4 DSM 9.3 4.38 3.1 1.19 1.4 0.41 9.1 All Tanzania 51.1 35.9 24.9 15.7 15.6 9.4 100.0 Notes: Head Count = Counts the number failing below each poverty line Depth = Percent of poverty line income required to bring everyone below It up to the poverty line Seventy = More heavily weights the extremely poor: it would be smaller if most poor were near the poverty line and larger if many were much poorer than the poverty line. Soft = Soft core poverty line of TSH 46,173 per year Hard = Hard core poverty line of TSH 31,000 per year poverty line. These amounts correspond to approximately US$1.5 billion and US$0.64 billion respectively. In 1991 net Official Development Assistance (ODA) from all donors was US$1.076 billion (at 1991 prices and exchange rates).' Thus, ODA transfers in 1991 were more than enough to eliminate extreme (hard-core) poverty from Tanzania - again assuming that perfect targeting is possible and that the money is used as recurrent transfers rather than capital investment. SPATIAL MAPPING OF THE POOR Location of Residence 16. The division of households into three different zones - rural villages, urban areas outside DSM, and the capital city itself - allowed us to infer something about the spatial distribution of poverty:' a. The head count ratio (PO) indicates that 59 percent of people living in rural households are poor (10.6 millions), and 44 percent (7.9 millions) are very poor. For households living in the urban areas (excluding DSM), these values are 39.5 percent (1.7 millions), and 17.9 percent (.7 millions), respectively. Finally, "only" 9.3 percent (200,000 tanzanians) of the people living in DSM are poor, and 4.4 percent (less than 100,000 people) are in hard core poverty. b. Using the percentage shares of and the extent of poverty in each location, one can calculate how much each area contributes to poverty. 85 percent of the poor live in rural villages, 13 percent in urban areas outside the capital city, and just 2 percent in Dar es Salaam. The dissimilarity between the contributions of rural and urban poverty (whether in or out of Dar es Salaam) is even higher when using the "hard- 'Source of data is African Indicators of Development (1992). See Table A.5 (Appendix 1) for sample size in each location. Characteristics of the Poor in Tanzania 33 core" poverty line: 90 percent of the very poor live in rural areas, nine percent live in urban areas outside DSM, and one percent live in the capital city.' c. Not only is Percent poverty - more 60 concentrated in rural areas, but it 50 is also deeper and more severe. 40 . Using the hard- core poverty line, so .... and looking at the P2 index, we 20 -- . see that the s ev er i ty of 1o-- ..-- - poverty is about three times 0 higher in rural Head Count Depth Severity areas than in urban areas I Rural M Urban 0 Dar es Salaam M Tanzania outside Dar es Figure 3.5: Poverty Indicators by Location Salaam, and it is more than twenty times greater in rural areas than in the capital itself (see Table 3.5 and Figure 3.5). Regions 17. The sample of households in each region was small (between 16 and 50,. except for DSM and other large regions), so it is hard to make precise judgements about the regional levels of poverty.' Nonetheless, the regions showing higher incidence, depth, and severity of poverty tend to be regions that are commonly perceived as relatively poor. The very poor regions, defined as having an incidence of hard-core poverty over 40 percent are: Kigoma, Singida, Lindi, Dodoma, Rukwa, Shinyanga, Tabora, Mwanza and Ruvuma. The relatively rich regions are Dar es Salaam, Kilimanjaro, Tanga, and Mara (see Table 3.6). The poverty profile of the four poorest regions is highlighted in Figure 3.6. Lindi stands out as having the most extreme poverty by far (P2) 18. Agriculture is the main economic activity in rural areas and differences in the farming environment seem to be the key distinction between better-off, poor, and very poor regions. The very poor regions are generally characterized by infertile soils and unreliable rainfall of less than 750 Caution must be taken in the interpretation of the results since regional price indices were only available for food items. Nor were imputed housing costs available for homeowners, who mainly live in rural areas. However, it may be that the higher value of public goods available to urban residents offsets the rent differential. ' These finding% should be interpreted carefully since, with a small sample size, the standard deviation of estimates is high. The Cornell/ERB household survey was not designed to obtain estimates at the regional level. We compared our estimates with the official regional per capita GDP estimates (see table Table C. 10). Those results and the results in this survey are completely different. For example, according to the official estimates Kilimanjaro is the 5th poorest region in Tanzania. and Kagera the 2nd poorest region. According to the Cornell/ERB survey both Kilimanjaro and Kagera belong to the top 5 richest regions. This comparisons show how little we know about the regional distribution of poverty and income. Better targetting calls for further research in this area. 34 Characteristics of the Poor in Tanzania Table 3.6: Regional Decomposition of Poverty for Three Poverty Measures and Two Poverty Lines Head Count (Po) Depth (P) Severity (P2) Soft Hard Soft Hard Soft Hard Arusha 39.6 29.3 17.5 10.4 10.7 6.2 Coast 54.8 38.4 25.4 13.6 13.9 6.4 Dar es Salaam 9.3 4.4 3.1 1.2 1.4 0.41 Dodoma 57.7 52.4 33.5 22.4 21.5 12.6 Iringa 52.4 31.6 19,7 9.4 9.9 3.9 Kagera 36.5 17.9 14.4 9.9 9.6 6.8 Kigoma 76 73.3 47.2 34.4 32.4 20.7 Kilimanjaro 30.8 17.6 11.1 5.1 5.5 2.3 Lindi 91.2 81.1 67.5 60.5 58.1 50.2 Mare 39.2 6.8 11.7 3.5 4.8 1.9 Mbeya 55.0 41.0 24.8 13.3 13.6 6.2 Morogoro 59.6 24.8 22.0 10.5 11.4 5.2 Mtwara 56.9 29.8 22.0 9.9 10.9 4.5 Mwanza 58.2 40.6 27.5 17.9 17.7 10.7 Rukwa 56.2 44.7 34.3 26.1 24.8 17.6 Ruvuma 73.5 73.5 47.6 35 32.7 20.8 Shinyanga 91.5 79.1 50.0 32.1 31.9 19.1 Singida 56.5 56.2 34.5 23.8 22.4 13 Tabora 61.5 45.9 27.6 14.5 14.9 6.5 Tanga 45.1 21.9 16.4 5.3 7.0 1.6 All Tanzania 51.1 35.9 24.9 15.7 15.6 9.4 Notes: Head Count Counts the number felling below each poverty line Depth - Percent of poverty line income required to bring everyone below It up to the poverty line Severity More heavily weights the extremely poor: it would be smaller If most poor were near the poverty line and larger if many were much poorer than the poverty line. Soft - Soft core poverty line of TSH 46,173 per year Hard - Hard core poverty line of TSH 31,000 per year millimeters (the minimum amount required for plant growth) and/or by their remoteness, far from all-weather roads or markets, and often with negligible infrastructure. Differences in farming environment have dictated the type of agricultural activities undertaken. Arid and semi-arid conditions have forced rural households to concentrate on low-value drought resistant crops and/or livestock production. Only cotton and cashew nuts are grown as export crops. Some regions are so remote (Rukwa, Ruvuma) that their economies are affected by an almost total lack of market infrastructure. Apart from some coffee and tobacco, these remote regions produce non-tradeable crops such as maize and beans which are bulky and low in value. 19. The poor regions are characterized by a combination of semi-arid climates with moderately fertile soils and wet highland climates (see Figure 3.7). Unlike the very poor regions, they get adequate and reliable rainfall (800-1000 mm per annum). Farmers grow a variety of food crops, as well as the higher value coffee, tobacco, cashew nuts, and cotton. Arusha and Morogoro have relatively good physical infrastructure, but -the other three regions which fall into the category .poor." namely Mwanza. Tabora, and Mtwara. have poor infrastructure and hence poor connections to main domestic markets. Characteristics of the Poor in Tanzania 35 20. Apart from Dar es Value of the Poverty Index Salaam, which is the commercial 100 and industrial center of the country, the other "better-off 80 regions" are also dependent on agriculture. They receive 60 - adequate and reliable rainfall (800-2000 mm per annum) and 40 - they have highly fertile soils. A variety of food and export crops 20 - are grown; these include coffee, tea, cotton, cardamon, and cocoa. o H ct S i Off-farm employment Head Count Depth Severity opportunities are apparently Lindi M Shinyanga M Kigoma M Ruvuma higher because of the presence of plantations: sisal plantations in Figure 3.6: Poverty Indicators for Poorest Regions, Tanga, tea plantations in Mbeya, Soft-Core Poor and coffee plantations in Kilimanjaro. An additional advantage of the better-off regions is that they have good physical infrastructure. Furthermore, some of them share common borders with neighboring countries (Kilimanjaro with Kenya; Mbeya with Zambia and Malawi), and this gives farmers more market options than their counterparts in relatively inaccessible parts of the country. However, such links are more likely to be advantageous when the neighboring country's own economy is performing relatively better than Tanzania's, if farmgate prices are higher, or if the borders represent opportunities to avoid domestic price controls and taxes. Table 3.7: Poverty by Agro-Climatic Zone Agro-Climatic Head Count (Po) Depth (P) Severity (P2) Region Soft Hard Soft Hard Soft Hard Kigoma, Kagera, Kilimanjaro, and 40.2 27.1 18 11.5 11.3 7 Arusha Lindi, Mtwara, Coast, and 61.1 42.8 31.7 20.8 20.9 13.8 Tanga Tabora, Mwanza, and 61.2 43 28.9 17.8 17.8 10.3 Shinyanga Rukwa, Ruvuma, Mbeya, and 57.6 45 30 19.6 19 11.3 Iringa Morogoro. Dodoma. and 58.5 38.1 27.4 16.2 16.2 8.7 Singoda Dar es Salaam 9.3 4.4 3.1 1.2 1.4 .4 Notes Head Count = Counts the number falling below each poverty krne Depth - Percent of Poverty line income required to bring everyone below it up to the poverty line Seventy - More heavily weights the extremely poor it would be smaller if most poor were near the poverty line and larger if many were much poorer than the poverty line Soft * Soft core poverty line of TSH 46.173 per year. per adult equivalent Hard a Hard core poverty kine of TSH 31.000 per year. per adult equivalent 36 Characteristics of the Poor in Tanzania voUAeosA Í> 0o% '.'A>NWs . . ä3 ii •RA Figure 3.7: Regional Map of Tanzania Characteristics of the Poor in Tanzania 37 21. To overcome the small sample size problem, we also computed the indices of poverty by agro.;climatic regions (See World Bank 1993). This aggregation is not perfect, since there is no perfect overlap between the agro-climatic regions and the administrative division. The estimated poverty indices are displayed in Table 3.7. The results show that the Northern Highlands agro- climatic region (Kagera, Kigoma, Arusha, and Kilimanjaro) is doing better than the other regions. They have lower poverty, which is also less deep and severe. Soclo-EcoNoMIc MAPPING OF THE POOR 22. Another poverty-related classification is the primary occupation of the head of the household. This classification is important for policy design, since the way that adjustment policies influence households is closely connected with the economic role that they play. The following findings emerged from the survey and are summarized in Table 3.8: a. Farm households constitute the bulk of poverty in Tanzania: 59 (44) percent of people living in households whose primary occupation is agriculture are poor (very poor), and these households constitute 83 (88) percent of all poor households (very poor households). As the weight attached to the very poor in the FGT index increases, the contribution of farm households to national poverty also goes up, Table 3.8: Three Different Measures of Poverty and Two Poverty Lines, by Employment Group and Location Head Count (P) Depth (Pi) Severity (P2) POpulation Soft Hard Soft Hard Soft Hard Share Rural Farmers 59.1 44.1 30.1 19.8 19.4 12.0 71.9 Business men/women 61.6 49.2 23.1 8.5 10.6 2.9 1.4 Resident Govt Employee 28.7 26.3 15.5 9.8 9.1 4.1 0.24 Urban Private/Self-Employed 33.7 14.9 12.8 5.8 6.6 3.0 21.4 Business men/women 8.4 1.7 1.2 0.21 0.34 0.04 0.42 Govt Employee 9.7 7.1 3.5 1.4 1.6 0.4 4.8 All Tanzania 51.1 35.9 24.9 15.7 15.6 9.4 100 Notes: Head Count = Counts the number falling below each poverty line Depth = Percent of poverty line income required to bring everyone below it up to the poverty line Seventy = More heavily weights the extremely poor: it would be smaller if most poor were near the poverty line and larger if many were much poorer than the poverty line. Soft = Soft core poverty line of TSH 46,173 per year Hard = Hard core poverty line of TSH 31,000 per year indicating that farm households are often in severe poverty. b. Self-employed households living in urban areas are also prone to poverty, and these households constitute 14 (8) percent of the poor (very poor). In addition, rural households engaging predominantly in their own businesses are very likely to be poor - 61 (49) percent of rural households headed by a businessman or businesswoman are poor - but because there are so few of them in the such households represent just 0.9 (1.9) percent of the poor (very poor) in the country. Therefore, special 38 Characteristics of the Poor in Tanzania targeting of this group will make less of a dent in national poverty than will targeting other groups. SENSIVrrY TESIS 23. A battery of different poverty measures were used to check whether our results were sensitive to the methodology used. For instance, we measured poverty by looking at the number of households below the poverty line, rather than by counting the number of people in poor households. The resulting poverty measures are slightly smaller, as expected, because the average family size is larger for poor households (see Table 3.4). Thus, the proportion of poor people is larger than the proportion of large households. Nonetheless, the results are very similar to those gained from the earlier approach; the differences between the two results are not significantly different within a three percent confidence interval. Given the similarity of the results we have not presented the detailed findings from this alternative approach. 24. Sensitivity tests also showed that the extent of poverty would not have been very different if other, lower, poverty lines had been used for the analysis. If the poverty line were set at Tsh 29,000 a year (US$143), around 30 percent of households would still be classified as poor. If we had used per capita expenditure of household members, rather than adult-equivalent household expenditure, as the criterion for ranking households, about 50 percent of all households would be considered poor for a poverty line of Tsh 31,000 a year. With a poverty line of around Tsh 20,000 in per capita expenditure (corresponding to one-third of mean per capita expenditure) 35 percent of all households would be poor. In short, whichever criterion is used to rank hbuseholds, and for a quite wide spectrum of poverty lines, the incidence of poverty in Tanzania is high. If we had chosen the poverty line used in the World Development Report (1990) of US$370 a year to define soft-core poverty, and US$275 to define hard-core poverty, the equivalent poverty lines in Tanzanian Shillings (using a purchasing power parity estimate of the "real" exchange rate) would have been Tsh 14,800 and Tsh 11,000 respectively. Applying these poverty lines, and using per capita expenditure as the criterion to rank households, the incidence of poverty would have been around 25 percent and 18 percent of all households, respectively. 25. The general conclusion from the sensitivity tests is that the rural/urban/DSM poverty ranking is not at all sensitive to the choice of the poverty measures or poverty lines. Whichever poverty line or measure we choose, we always find a greater incidence, depth, and severity of poverty in ruml areas, among farmers.10 Given that the poverty profile of villagers generally and farmers in particular are very similar, and that these two groups are the main contributors to poverty in Tanzania, the remainder of the analysis in this assessment will concentrate on the distinctions between rural and urban areas and DSM. 'o The construction of confidence intervals and hypothesis testing (see Ravallion 1992) for the indices of incidence. depth, and severity of poverty tell us that the rural poverty is statistically significantly different from the urban poverty at the 5 percent level of significance. Characteristics of the Poor in Tanzania 39 POVERTY BY INCOME INSrEAD OF EXPENDrFURE 26. This study chose consumption, as proxied by current expenditure, as the indicator of welfare and standards of living. Several reasons guided this choice, including a concern with accuracy. Table 3.9 presents percentage of households living in poverty, if income, rather than expenditures had been chosen Table 3.9: Expenditure vs. Income as a Measure as the indicator of welfare. The data in of Welfare - Soft Poverty Line the table for expenditures matches the Expenditures household data on the percentage of Better-off Poor Total households in poverty presented in Better-off 25.1 3.2 28.3 Table 3.1. As the table shows, using Income Poor 27.4 44.3 71.7 income as the measure of the standard of T 2.5 47.5 100 living would have resulted in much higher measures of poverty, both for soft-core and for hard-core poverty. Using expenditure as the criterion, 47.5 percent of households were deemed to be poor. Using income, 71.7 percent of households were found to be poor. The two sets of poor households do not overlap, so that some households who were found to be poor on an income basis were not poor on an expenditure basis, and some households who were poor, based on their expenditure, were not poor based on their income. However, it was more frequently the case that households considered poor on an income basis were not classified as poor on an expenditure basis. 27.4 percent of all households have incomes below the poverty line, but expenditures higher than the poverty line. Reasons for the difference in results include inaccuracy in measurement, under- reporting of income, and transfers from richer to poorer households that compensate for falls in income, and/or the households own saving or dis-saving as consumption is smoothed over time. SUMMARY 27. The following summary characteristics of poor households emerge from the analysis, using 1991 data. a. About 50 percent of all Tanzanians live in poor households, i.e., with annual expenditures below Tsh 46,173 per adult-equivalent, and 36 percent of all Tanzanians live in very poor households, i.e., with annual expenditures below Tsh 31,000 per adult-equivalent. b. If perfect targeting were possible, the minimum amount of transfer payments required to eliminate poverty would be Tsh 11,497 per year and per person using a poverty line of Tsh 46,173, and Tsh 4,867 per year and per person using a poverty line of Tsh 31,000. c. Poverty in Tanzania is mainly a ruml phenomenon - 59 percent of people living in rural villages are poor, 39 percent in the urban areas excluding Dar es Salaam, and only 9 percent in Dar es Salaam itself. Around 85 percent of the national incidence of poverty is accounted for by rural areas. The rural character of poverty is even more pronounced when the hard core poverty line is used. In that case, rural villages account for 90 percent of overall hard core poverty in Tanzania. When analytical measures capturing the depth and severity of poverty are used, the rural-urban poverty gap looms even larger. 4 BAsIc NEEDS AND ASSETS OF THE POOR SATISFACTION OF BASIC NEEDS OF THE POOR 1. Analysis of basic needs provides a non-money measure of welfare, as opposed to the expenditure/income measures presented in Chapter Three. It can capture aspects of the standard of living, and details of people's lives and well-being that cannot be reduced to a monetary estimate. Basic needs include education, health, housing, and the ownership of small and large possessions. As Lipton and Ravallion (1992) stress, in poor countries a ten percent rise in non-monetary benefits will translate to a more than proportionate gain for the poor, because a given Shillings value of benefit will mean a larger percentage gain in their incomes. EDUCATION Literacy' 2. A very important aspect of basic needs fulfillment is education. In 1977, the official literacy rate was 73 percent (Collier et al. 1986), where literate was defined as being able to read and write. In 1978, according to the Tanzanian government, universal education was achieved, with a literacy rate of 85 percent (versus 10 percent in the 1960s). As of 1991, according to our survey, the literacy rate was about 68 percent, implying a drop in this indicator. It seems that either the situation has worsened in the last 14 years, or the earlier official statistics overestimated the literacy rate. However, in comparison to countries at a similar stage of development, Tanzania still has a relatively high rate of literacy. Literacy is much more extensive than in other eastern African countries, and Tanzania's illitemcy rate is well below the average for Sub-Saharan Africa, which is estimated by the World Bank (1992c) to be as high as 51.1 percent. 3. From Table 4.1 Table 4.1: Literacy Among People Older Than 14 one can see that people Income Group Location Gender living in rural areas and All women are more likely tO Tanza Better- Very nia off Poor Poor Rural Urban DSM Male Female be illiterate than other Read and groups. As many as 80 Write 67.8 75 59 56.5 61 79.4 88 76.2 60.8 percent of the people who ' Read Only 5.3 2.7 8.5 6.8 6.4 3.4 1.9 5.7 4.7 cannot read live in rural Neither 26.9 22.1 32.7 36.7 32.7 16.9 11.1 18.1 36.1 areas of Tanzania and 70 - - . - percent of the illiterate Total 100 100 100 100 100 100 100 100 100 are women. Furthermore. while only Note that Table 4 I through Table 4.3 are computed for all population older than 14. The age cut off is the same as that u%ed by the United Nations to facilitate comparisons. A person is said to be in poverty if he or she belongs to a famil) in pover" Also. people that can only read are classified as illiterate. Basic Needs and Assets of the Poor 41 61 percent of women are literate, 76 percent of the male older than 14 can both read and write. 4. The gender dimension of education is interesting because of its relevance for policy. There is empirical evidence from several studies (e.g., Meush, Leutzner, and Preston (1985)) that women's education is more closely associated with child mortality rates than other common indicators of development such as per capita income. Cochrane et al. (1980) found that the effect of women's schooling on child health is twice as strong as the effect of male schooling. The same study found that women's education is more effective in reducing child mortality, and increasing children's nutritional status, than the education of men. One possible explanation for this phenomenon is that child education and child rearing are the responsibility of women in their role as mothers. This responsibility also includes health and nutrition. In addition, empirical work shows that fecundity is negatively correlated with female education. In a country like Tanzania, where the growth rate is three percent, this is a factor of great significance for future welfare. The findings emerging from this report, that female literacy and female education is much higher for younger age cohorts, implies that there should already be a positive feedback effect on child welfare. 5. Figure 4.1 illustrates the Percent Literate link between literacy and poverty. 100 People who are poor are less likely to be able to read and so - - .-. write. Amongst poor people the literacy rate is 59 percent, lower o - than the national average of 68 percent. The literacy rate is even 40 ... ... - - lower (56.5 percent) among the very poor, whereas, the rate goes 20 - .. up to about 75 percent for the better-off. Establishing the causal 0 relationship between income and 14-19 .20-29 30-39 40-49 60-59 60* education would require more Better Off M Poor analysis: Are people poor because they are uneducated, or Figure 4.1: Proportion of Literate by Age Group and are they uneducated because they Poverty Status are poor? 6. The gender gap in illiteracy is even more pronounced for families with incomes lower than the poverty line. Among males living out of poverty 82 percent are classified as literate, while for their female counterparts the rate is about 67 percent. However, for people living below the poverty line, the rates are 49 percent and 68 percent, respectively. This amounts to a larger male/female literacy ratio for the poor: above the line the ratio is 1.22, while below the poverty line it is 1.38. 7. Figure 4.2 adds some more information on literacy by gender and age cohort for mainland Tanzania. Old people are more likely to be illiterate, and older women are particularly affected by illiteracy. The fact that female illiteracy is concentrated among the old suggests that factors limiting 2 These results are believed to be reliable because the interviews were conducted in the presence of all household members, hence significantly reducing the chances of "cheating" on the parts of the heads of households and other members. 42 Basic Needs and Assets of the Poor women's education have already Percent Literate started to change. Among people 100 below age 30, female and male literacy are about the same SO- nationally, reflecting policy changes under Ujamaa. 6o- 8. Because the head of the 40- household is the person taking major decisions, especially 20- production decisions, it is important to assess her/his degree 0 of literacy. Our findings show 14-19 20-29 30-39 40-49 50-69 60+ that the literacy rate for the head men women of household and the whole are similar. 67.7 percent of Figure 4.2: Proportion Literate by Age Group and households are headed by Gender someone who knows how to read and write, 5.4 percent are headed by someone who only knows how to read, and 26.9 percent of household heads can neither read nor write. 74 percent of the households above the poverty line are headed by literate person versus 60 percent for households below the poverty line. 9. There are several possible explanations for the gender gap in education. An ILO survey finds that females are less likely to participate in wage employment. Consequently, even if women receive similar wages, the expected returns to education are smaller for girls than for boys. This gives parents less incentive to support their daughters' education, and may even cause them to deny schooling to girls. Parents may also ignore the returns to education in self-employment, or they may be swayed by the gains from educating the sons who, by cultural norms, will ultimately be responsible for their parents' welfare in old age. Care for younger siblings and other factors specific to Moslem societies may also reduce girls' access to schooling. Table 4.2: Highest Education Achieved (People Older than 14) Income Location Gender All Tanzania Better Off Poor Very Poor Rural Urban DSM Male Female None 24.9 18.9 32.2 34.9 30.6 15.1 9.3 16.1 34.1 Primary 63.3 64.4 62.0 59.9 61.6 71.1 58.3 70.0 56.5 Secondary 7.2 11.3 2.2 1.3 3.6 10.1 24.1 9.4 4.9 University 0.5 0.9 0 0 0 1.4 4.6 0.8 0.2 Adult Literacy/Ot 4.0 4.5 3.6 3.9 4.2 2.8 5.6 3.7 4.3 her Total 100 100 100 100 100 100 100 100 100 Type and Level of Formal Education Achieved 10. As Table 4.2 shows, 88.2 percent of the population older than 14 does not have any education beyond the primary level. Thus, although Tanzania is performing relatively well in terms of the percentage of the population who can read and write, many of its literate population may have only Basic Needs and Assets of the Poor 43 a rudimentary education. Nonetheless, our estimates of educational achievement are more optimistic than those from Collier et al. (1986), who found that in rural areas only one percent of the population has received any secondary education. The small percentage is partly explained by a high rate of migration among the educated. 11. Very few of those who go Percent beyond a primary education live 1oo F in families with incomes below the poverty line.' Just 3.6 80- percent of the rural population older than 14 has a secondary 60- education, while the equivalent number is 10.1 percent in the 40 - urban areas outside DSM, and as much as 24 percent in DSM. 20 Amongst the poor and the very 1 1 poor, the chances of reaching 0 secondary education are even Better-oft Poor Very Poor Male Female smaller, as Figure 4.3 highlights. Primary E Secondary Eleven percent of the better-off go or have been to secondary Figure 4.3: Percentage of Respondents Completing school, while just 2.2 percent of Primary or Secondary Education poor people, and 1.3 peicent of very poor people have reached this level. Some of this bias may be due to financial constraints and limited capacity in public secondary schools - of the 7.2 percent of the total population over 14 that have reached secondary education, approximately one-half attended a private secondary school. There is also a gender bias in educational achievement - 9.4 percent of men go to or have been to secondary school vis-A-vis 4.8 percent for women. 12. Only one-half of a percent of the population older than 14 goes/went to a University, of whom none live in rural areas. Forty percent live in urban areas outside DSM, and the remaining 60 percent live in DSM. As expected, no one with a university degree is reported to live in a poor household. Men account for 80 percent of all people who have attended university. From the results one can see that higher education has not traditionally been open to women. Further analysis and data would be required to see whether the under-representation of people from poor or rural households among more highly educated respondents indicates either that the educational system is biased against poor people and rural people, or that higher education is the route out of poverty and into urban occupations. Reasons to Stop Attending School" The 1991 Household Budget Survey estimates that in the sample about 4% of the population completed secondary education. A high number of people responded to this question "Why did you leave school?" with the answer -completed and no desire to continue". Some of these answers may have been due to the way the question was formulated. For mistance. we may hypothesize that anyone who could not qualify for further education and felt embarrassed to say so could have responded this way. rather than with "unable to gain entry to higher level-. 44 Basic Needs and Assets of the Poor Table 4.3: Reason to Stop Attending School Income Group Location Gender All Tanzania Better off Poor Very Poor Rural Urban DSM Male Female Complete and no desire to continue 29.5 26.8 33.3 36.2 28.3 29.1 35.3 31.5 28.6 Incomplete but no desire to continue 12.8 9.9 16.8 18.1 14.2 11.8 7.5 6.3 15.6 Unable to gain entry into a higher level 2.9 3.2 2.2 1.1 3.4 3.2 0.0 2.7 3.3 Unable to meet expenses 38.6 44.0 30.6 29.5 36.2 37.7 51.9 47.0 30.6 Sick 5.2 4.4 6.4 5.2 4.8 7.7 3.0 4.6 6.2 Parents did not want me to continue 7.5 8.1 6.9 6.6 8.7 7.3 2.3 5.5 10.4 Other 3.6 3.5 3.7 3.3 4.5 3.2 0.0 2.3 5.3 Total 100 100 100 100 100 100 100 100 100 13. As Table 4.3 shows, an inability to meet expenses was the reason most often given in the sample overall for abandoning education (38.6 percent). About 61 percent of all respondents who said they dropped out for financial reasons live in rural areas, and 64 percent of them are men. One reason why financial constraints may affect boys more is the larger perceived opportunity cost on their work time. In rural areas both men and women will work in agriculture, and women also do household chores. Nonetheless, men are basically in charge of cash crops, in contrast to women's responsibility for food crops, and this may help to explain why their labor time is considered more valuable. Amongst the poor, 31 percent report having stopped school for financial reasons (30 percent of the very poor). However, the reason most frequently cited by the poor for abandoning education was a lack of desire to stay in school. Taking together those who answered "Incomplete and no desire to continue" and those who answered "Complete and no desire to continue" this accounts for 50.1 percent of the poor school leavers and 54.3 percent of the very poor school leavers respectively. This may be due to the low educational and professional expectations of poor children, which lowers their incentives to continue in school. Figure 4.4 gives some of the key breakdowns between different motivations for better-off, poor, and very poor respondents, as well as males and females. 14. At junior secondary school level (Form 1-4) the eetter-off inability to meet school expenses I accounts for 30.4 percent of those Poor who abandoned school before completion (see Annex C for a Very Poor detailed breakdown of motivations on leaving school bye education level). The majority Feme (83 percent) of the respondents in Female this category were men. located o 25% 50% 75% 100% mainly in rural areas and the city 1 Desir* r E,penses of Dar es Salaam. One of the m Parental Disapproval = Other possible explanations for this bias Figure 4.4: Reasons for Abandoning Education (All towards men is that there are fewer girls in secondary school in the first place. and they make up a small proportion of students. The proportion who said that they were unable to gain entry to higher Basic Needs and Assets of the Poor 45 level was very small, perhaps because in such cases people answered "completed and no desire to continue" as a face-saving mechanism.' 15. At levels of education higher than Form 4 (including university education), 29 percent of all respondents left school because of an inability to meet school expenses. The majority of the people citing financial constraints would have been attending private secondary schools, since education at public secondary school and University was officially free, at the time of the survey. Nonetheless, even with free education, the opportunity cost of being in school may have caused a financial burden as may have the ancillary costs of attending school. As with the Form 1-4 level, the majority of respondents referring to financial constraints were boys (84.5 percent). Another common feature of levels of schooling above primary is that the answer "Complete and no desire to continue" becomes more frequent than the answer "Inability to meet expenses." This may have to do both with the lack of the right exam results and with a real desire to leave school. 16. 7.5 percent of the population abandoned school, at whatever level, because their parents did not want them to continue. For the most part, people deprived of education because of parental denial are women (63 percent), and most of them live in rural areas (74 percent). The large representation of women in this category is consistent with the tradition that parents perceive education for their daughters as less attractive than education for their sons. "They may fear that education will harm their daughters' marriage prospects, subsequent domestic life, and even spiritual qualities." (Poverty and Human Development p. 18, 1980). 17. Many children start school after age nine. As Table 4.4 shows, only 29 percent of children between ages 7-9 are in school. This proportion is lower for rural areas than for urban areas, rising to 44 percent in Dar es Salaam. It is even lower among the poor, and the very poor. Apparently many children start school around age 10-13 and about 67 percent of children in this age group were found to be in school. Similarly to the younger age bracket, the proportion in school is lower for Table 4.4: Fraction of Children Enrolled in School All Income Group Location Gender Age Group Tanzania Better-off Poor Very Poor Rural Urban DSM Male Female 7-9 .29 .37 .21 .22 .27 .29 .44 .28 .29 10-13 .67 .72 .63 .66 .65 .71 .81 .63 .71 rural areas and among the poor and the very poor. As a consequence, the literacy rate may not rise in the near future (or may fall!). 18. There are basically no gender differences in the proportion of children in school at ages 7-9 (see Table 4.4). For ages 10-13, however, the proportion of girls is substantially higher than the proportion of boys. As explained above in paragraph 13, the absence of more boys in school may be attributed to their role in agricultural production. Respondents were asked questions about their educational achievement face to face, with other family members present. 46 Basic Needs and Assets of the Poor BAsic HOUSING AMENITIES/HOUSING QUALITY 19. The poor are identified not only by a lower expenditure level but also by distinct differences in amenities. Overall, just 47 percent of the population lives in a house with a metal roof, and 33 percent live in a house made of something other than mud (Table 4.5). But, of the poor population, only 32 percent live in a house with a metal roof and about 20 percent in houses without mud walls. Among the very poor these percentages are even lower, at 29 percent and 18 percent respectively. Table 4.5: Basic Housing Amenities (Percent of Total Households in Each Category) Income Group Location All Tanzania: Better-off Poor Very Poor Rural Urban DSM Houses with a metal roof 47.3 60.6 32.4 28.7 30.5 82.7 94.8 (Percentage) Houses without mud 33.4 44.9 20.5 17.8 19.4 56.4 86.6 walls (Percentage) Water Supply and Availability 20. One of the survey questions elicited information on the main water source of the household. As might be expected, the main source is very different depending on where the respondent lived and whether or not he or she was poor. The results, displayed in Table 4.6, indicate that about 38 percent of all households get their water from a tap, 30 percent from a stream, and 26.5 from a well. Among the poor, just 26.6 percent have a tap as their main water source, and only 22.7 percent of Table 4.6: The Main Water Source of Households Income Group Location All Tanzania Better-off Poor Very Poor Rural Urban DSM Stream/ Lake 31.0 24.0 38.9 41.1 41.3 8.4 3.1 Spring 4.9 5.7 4.0 4.7 6.7 4.9 1.0 Well 26.5 22.6 30.8 31.5 35.1 9.4 1.0 Tap 37.6 47.5 26.6 22.7 16.8 81.8 95.8 Total 100 100 100 100 100 100 100 the hard core poor. 21. Table 4.7 shows the length of time that people take to reach their main water source. A fortunate 10.6 percent of -all households have water service at the door. About 43.2 percent of households have to walk for 1-15 minutes in order to reach a source of water. Another 26.9 percent have to spend 16-30 minutes walking to the water source, and the remaining 19.3 percent spend more than 30 minutes to get to a source of water. The United Nations had set as a target for development that everybody would have access to safe water supplies within 400 meters of their household by the year 1991. However, as Table 4.7 shows, this objective has not yet been met in Tanzania. As a result many rural women spend excessive time and energy in fetching water. A characteristic of life for poor women in rural areas is prolonged and heavy manual labor in water collection, with consequent negative effects on crop production and family welfare, and especially young children's Basic Needs and Assets of the Poor 47 Table 4.7: Access to Source of Water - How Long Does It Take? (Percentages) Income Group Location All Tanzania: Better-off Poor Very Poor Rural Urban DSM 0 minutes 10.6 16.3 4.2 4.0 1.4 25.1 46.9 1-5 minutes 18.0 22.6 12.9 9.7 13.0 29.6 30.2 6-10 minutes 14.8 14.1 15.6 17.4 15.1 14.3 13.5 11-15 minutes 10.4 9.9 11.0 12.8 12.3 7.9 2.1 16-30 minutes 26.9 22.1 33.3 35.2 32.1 19.2 6.3 More than 30 minutes 19.3 15.9 23.0 22.7 26.1 3.9 2.0 Total 100 100 100 100 100 100 100 Percent 60 50- 40 30 - 20- 10 0 Better-off Poor Very Poor Figure 4.5: Percent Walking More than 15 Minutes for Water nutrition. 22. The reality of longer water collection time for poorer households is captured in Figure 4.5. Among poor households only four percent have water readily available, and about 56 percent need to spend more than 15 minutes to reach a water source. The situation is even worse for the very poor of Tanzania. 57.9 percent of the very poor need to spend more than 15 minutes to fulfill their water needs. On the other hand, about 16 percent of the better-off have readily available water, while about 38 percent of them spend more than 15 minutes to reach the main water source. 48 Basic Needs and Assets of the Poor Ownership of Durables and Semi-Durables 23. Table 4.8 presents some results on the ownership of durables and semi durables.' Patterns of ownership reflect strong differences in access to these type of goods across poor and better-off households, as well as between rural and urban households. 24. With the exception of bicycles, the ownership of durables is positively correlated with Table 4.8: Ownership of Durables (Percent of Households in Each Category) Income Group Location All Tanzania Better-off Poor Very Poor Rural Urban DSM Bicycle 26.6 26.8 26.3 24.6 29.5 20.2 18.8 Radio 45.7 54.6 35.8 34.6 37.1 61.6 76.1 Sewing Machine 9.1 13.5 4.2 3.1 5.1 12.9 31.3 Non-Kibatari Light 36.6 47.0 25.1 20.9 25.8 61.1 64.6 Watch 47.6 56.8 37.3 33.9 40.6 59.6 72.9 income. The two durables whose ownership is most closely connected with poverty status are radios and watches. While 45.7 percent of all households own a radio, and 47.6 percent include a member who owns a watch, only about 35 percent of the poor and very poor households have a radio. 37 percent of poor households and 34 percent of very poor households include someone who owns a watch. Against this, 55 percent of the better-off have radios, and 57 percent possess watches. ASSET HOLDINGS Land 25. Three of the major assets in poor rural economies are land, livestock, and labor. From Table 4.9 one can see that very few households in 'anzania, especially in rural areas, are excluded from owning land." Moreover, 96.6 percent of rural households own some land, with an average landholding per household of 3.94 hectares nationwide, and 4.66 hectares for rural areas. However, landholdings are very unequally distributed. More than 32 percent of households own less than one Table 4.9: Percentage of Households owning Land and Livestock Income Group Location Very Ownership All Tanzania Better-off Poor Poor Rural Urban DSM Land Ownership 82.5 72.7 93.4 95.2 96.6 60.3 27.5 Livestock Ownership 50.3 46.7 54.3 56.3 63.5 30.6 6.3 'According to the preliminary results (unweighted sample means) of the 1991 HBS 32 % of the rural households own a bicycle, and 47.83% own a radio. In Dar es Salaam these values are respectively 20.25% and 76.1%. ' Land ownership refers to land available to the household (cf. page 5 of the Cornell/ERB household survey). In this sense rented land also falls in the category of "owned" land. However, in overall Tanzania just 1.785 percent of the total land is estimated to be rented. Basic Needs and Assets of the Poor 49 hectare, and more than 60 percent own less than three hectares. However, there is no evidence in the data set that richer households own or operate more land than poor or very poor households. 26. The inequality indices computed and displayed in Table 4.10 confirm the inequality of access to land with a Gini coefficient,of 0.50 in the rural areas. The national Gini coefficient of inequality in access to land is 0.6 - quite high when compared with the average in 15 Afro-Asian countries of 0.53 (Nafziger 1988), but not excessively high when compared to countries like Kenya (0.55) or Ghana (0.64). Moreover, the fact that the national Gini coefficient is higher than the rural Gini Table 4.10: Comparing the Inequality of the Income Distribution with the Inequality of Land Ownership Income Group Location Gini coefficient Al Tanzania Poor Very Poor Rural Urban DSM Land Distribution: All Households .50 .60 .50 .77 .93 Land Distribution: Households with some Land .49 .47 .48 .62 .76 Income Distribution .57 .30 . .28 .60 .48 .39 coefficient implies that towns and cities (where many people hold no land) are the major contributors to land inequality. In addition, if larger farms are populated by bigger families, then some of the inequality in landholdings by household will cancel out at the individual level. Looking at the ownership of land in per capita terms, one sees that the value of the Gini coefficient is lower, indicating that household size and access to land are correlated. When the Gini coefficient is computed just for households that own some land, the values are much lower for the urban areas, indicating that inequality in these areas is linked to a high proportion of landless families. 27. In countries like India and Pakistan, access to land is the major determinant of poverty status and income distribution. But there are several other economic factors that may account for income inequality in rural areas. Analysis of Table 4.10 shows that income inequality is smaller than land inequality everywhere except for the rural areas of Tanzania, where income inequality exceeds inequality in landholdings by farm. This is one indication that the income distribution and the poverty breakdown for Tanzania are not solely caused by the quantity of land. 28. Figure 4.6 displays the Lorenz curves for the distribution of land in the rural areas and DSM, among poor households, and for the entire country. From the position of the curves it appears that the distribution of land is most "egalitarian" in rural areas. However, this result is driven by the different economic structure of cities. The large number of landless households in DSM is to be expected since far fewer urban households are engaged in agriculture. The important point to retain from the analysis of Figure 4.6 is that the Lorenz curve for the rural regions and the Lorenz curve for poor households almost overlap. This confirms the earlier observation that the distribution of landholdings for the poor is very similar to the distribution of landholdings for rural residents generally. 50 Basic Needs and Assets of the Poor 1 .0 I 0.9 0.8 (1) - Rural (2) - Poor (3) - All Tanzania 0.7 - (4) - Dor es Soloom - 0.6 - c 0.5 - o 0.4 0~ (4): 0.3 - (3) 0.2 - - 0.1 --- 0.0. 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Proportion of households Figure 4.6: Lorenz Curve for the Distribution of Land Basic Needs and Assets of the Poor 51 Livestock 29. Collier et al. (1986) argue "that livestock ownership is [...] a decisive factor in explaining income inequality [...] Among the settled farm population, ownership of livestock is relatively rare [...] In a sense, 'livestocklessness' is the Tanzanian counterpart to landlessness in rural Asia [...] and it is this which differentiates the rural into two sizeable groups." (pp. 54, 55) The next tables and figures investigate this issue. 30. It is true, as one can see from Table 4.9 that more than 49.7 percent of the population does not own any livestock. However, the percentage of those who own, and of those who do not is about the same for people above and below the poverty line. Further, the difference between rural residents and poor people in terms of livestock ownership is not great. 50.3 percent of all Tanzanians, 54.3 percent of poor people, and 63.5 percent of rural residents own livestock. Of course, it may be that poor households simply own fewer and less valuable cattle than their better-off counterparts. An index of livestock values (see Annex B, Box B.3) was used to produce a homogenous measure of livestock values. This is shown in Table 4.11. The average index for the Tanzanian population is 3.78, while for the poor (very-poor) the index is 3.54 (3.64), and for rural residents it is 5.06. When only those households that own some livestock are considered, the index values are 7.23, 6.3 (6.3), and 7.7 respectively for the nation, the poor (very-poor), the better-off, and rural dwellers. 31. Because mean values can mask information on the distribution of livestock values, several measures of inequality on the distribution of livestock ownership were computed.' These measures Table 4.11: Livestock Ownership Distribution Income Group Location All Tanzania Poor Very Poor Rural Urban DSM Mean 3.78 3.54 3.64 5.06 0.91 0.44 Mean (For households with 7.23 6.33 6.26 7.71 3.27 7.32 some livestock only) Standard Deviation 11.93 12.49 12.63 13.87 3.66 2.95 Coefficient of Variation 3.16 3.53 3.47 2.74 4.03 6.76 Gini Coefficient 0.86 0.87 0.85 0.82 0.93 0.98 Gini Coefficient * 0.73 0.76 0.73 0.72 0.74 0.64 Theil Coefficient 0.83 0.84 0.81 0.78 0.91 0.97 Note: a. For those with ownership of livestock strictly positive (see Table 4.11) show that the livestock holdings are quite unequal, with a Gini coefficient of 0.86. This is connected. in part, with a large number of households that do not own any cattle whatsoever. The inequality is even higher when we move from rural villages to the capital city, because of the many urban households that will not own any cattle. * See Annex B for a definition of the several inequality measures. 52 Basic Needs and Assets of the Poor Labor 32. The final asset covered by the survey was labor. Table 4.12 presents data on how many days of labor were put into production by different types of households. As one can see, the amount used changes according to location and the allocation of labor also varies by poverty status. 33. A characteristic of many poor economies is that people are engaged in multiple economic activities as a way of guaranteeing income. The Cornell/ERB survey included questions to all members of a given household about the number of days spent in "money generating" economic activities, but unfortunately time spent in fetching water, and other household chores, is not captured by this survey. Table 4.12: Structure of Physical Labor Use (Total Population) Income Group Location Labor Use All Tanzania Better-Off Poor Very Poor Rural Urban DSM Total Family Days (p) * 450.8 461.4 439.1 468.4 474.6 359.6 469.1 Total days (p) b 359.7 388.6 327.8 333.8 357.3 323.9 452.6 Total owncroprural 116.3 98.3 136.2 144.8 144.7 68.3 10.0 days (p) Total Crop Rural Wage 3.5 2.8 4.3 3.5 4.7 1.1 0.3 days (p) TotalAdultlivestock 113.8 110.6 117.4 126.9 149.2 37.0 26.9 days Total non farm wage 68.0 95.2 37.9 31.3 31.2 117.6 232.2 days (p) Total Business Days (p) 48.5 72.5 21.9 14.7 15.7 97.8 183.2 Total Communalwork 9.6 9.2 10.1 12.6 11.8 2.1 9.6 Total livestock child days 91.1 72.8 111.3 134.6 117.3 35.7 16.5 % of child livestock days out of total livestock 44.8 35.9 52.5 56.7 44.8 44.8 51.2 days Child livestock days as percentage of total days 9.8 7.8 11.9 13.8 12.1 4.4 2.5 that child available a/ Total Family Days include all members of the household b/ Total days exclude children working on livestock 34. Several remarks are in order. First, poor and very poor people spent more time in agricultural activities than better-off households. In this respect, poor households are a close match to rural households generally. Second, households that spend a lot of time in non-farm wage, or business activities are better-off. In both regards, the labor uses of the poor and better-off reflect the identification of poverty with rural life, and greater prosperity with urban life. The distinction between the labor use of poor and better-off households is put in graphic terms in Figure 4.7. 35. A very important aspect of labor use is child labor because of the role of children in animal husbandry. Table 4.12 also shows the total number of livestock child days and the percentage of livestock working days that are accounted for by children: Basic Needs and Assets of the Poor 53 a. On average, 44.8 percent of total labor time spent in animal husbandry is accounted for by child labor. For better-off households, the average is 35.9 percent versus 52.5 percent and 56.7 percent for poor, and very-poor households, respectively. Thus, child labor use by very poor households in animal husbandry is 57 percent higher than child labor use by better-off households. b. The percentage of total potential child work days Better-Off that is effectively used for livestock husbandry Poor increases with poverty status, as the bottom row V P Very Poor of Table 4.12 indicates. It is 7.8 percent for 0% 25% 60% 76% 100% b e t t e r - o ff Agricultural = Non-Farm households, and "Business M Commercial 11.9 and 13.8 Figure 4.7: Allocation of Labor Time percent for poor and very-poor households. 36. According to the World Bank librld Development Report of 1990 "A study of rural Tanzanian households in 1980 found that the poorest in twenty sampled villages did not possess significantly less land or labor resources than others. Differences in living standards were largely attributable to differences in human capital and in ownership of non-labor resources such as livestock. The poorer households were less likely to participate in market transactions than the non poor, since they lacked the resources to grow cash crops and could not take the chance of a bad harvest that would leave them dependent on the market for their food needs. The poor also had much lower rates of return on work away from the family farm." This study confirmed all the points made in the Wirld Development Report with the exception of the remarks on livestock. Given that the proportion of households without livestock is similar above and below the poverty line, we conclude that cattle ownership is not a strong indicator of poverty, although there are differences in the quality of assets (including livestock) held by poor and better-off households. SOURCES OF INCOME 37. A priori, one may expect a significant differential in the rates of return to assets, including labor, in different economic activities. According to an ILO Labor Force Survey (1992), the average monthly income for agricultural occupations is Tsh 4,440 in urban areas, and Tsh 2,750 in rural areas. A similar urban-rural differential exists for several other economic activities. 38. Table 4.13 presents the decomposition of income for location of residence, and for the higher and lower poverty lines. Clearly, agriculture is the most important source of income in Tanzania. On average, agricultural income accounts for 55.3 percent of total income. The next most important category is business income, followed by non-agricultural labor income. Agriculture is, without question, the most important source of income for the poor and the hard core poor and accounts for 71.6 percent and 74.4 percent of their average total income, respectively. For better-off households 54 Basic Needs and Assets of the Poor Table 4.13: Sources of Income All Income Group Location Income in Tsh (mean/year) Tanzania Better-off Poor Very Poor Rural Poor Rural Urban DSM Per-Capita Income (p) 55,369 96,225 10,171 9,334 10,153 36,252 69,544 164,947 Adult equivalent Income (y) 82,073 158,812 15,941 14,754 15,978.6 56,969 88,721 251,207 Agricultural Income 158,893 258,906 48,253 46,860 54,904 208,552 58,707 8,314 From Crop Production 143,575 239,587 37,361 35,848 42,181 188,231 52,217 10,493 From Livestock 9,377 13,758 4,530 3,758 5,336 12,883 3,299 .3,380 Other 5,941 5,561 6,362 7,154 7,387 7,388 3,191 1,200 Non Agricultural Income 128,891 228,152 19,082 16,150 13,029 17,017 141,951 915,574 Wage Labor 22,525 37,851 5,570 5,048 3,431 4,437 36,825 124,325 Business Income 99,827 181,369 9,620 7,824 .6,074 9,439 95,221 769,016 Transfer income 2,920 3,272 2,531 2,726 2,388 2,376 3,824 4,981 Other Income 3,619 5,660 1,361 552 1,137 765 6,081 17,252 Total Income 287,784 487,058 67,335 63,012 67,933 225,569 200,658 923,888 Share of Consumption from own production (for those with some 72.1 77.1 66.9 68.6 70.9 78.18 77.95 production) Value of Consumption of own 126,496 219,199 23,943 22,582 25,990 165,236 49,709 5,842 production (p - all households) this falls to 53.2 percent. Even in Dar es Salaam, households still participate in agriculture, but income from this source accounts for just 0.9 percent of their total income. 39. The pattern of income sources is very similar for the poor and the hard-core poor. In particular, they receive a very low percentage of average total income from business activity: 14.3 percent and 12.4 percent, respectively. By contrast, the percentage of income derived from business activity is 37.2 percent for better-off households. Communal income, and income from remittances, accounts for a very low percentage of the average total income of the poor and hard-core poor. 40. Because the poor are concentrated in rural areas, it is also illuminating to compare their sources of income with the sources of income of rural residents generally. Table 4.13 allows for a comparison of the rural poor with the entire rural . The two major sources of agricultural income are Rura I crop production and livestock activity. However, the division Rural Poor between livestock income and Croo6 83 C" 6?- crop income is very different for the poor and others. The rural )tf re Ag 3sesx,c poor derive 62 percent of total L""0 o6 income from crop production, I-- whereas, on average, the rural Figure 4.8: Sources of Income derives 83 percent of its income from this source. The rural poor also have a more diverse range of income sources - 19 percent of their income is derived from nonagricultural activities versus 8 percent for rural residents in general - but these other sources of income are not sufficient to raise them out of poverty. The difference in major income sources is brought out clearly in Figure 4.8. Basic Needs and Assets of the Poor 55 41. Table 4.14 shows the average income earned in different activities for households that do participate in that activity. The percent rows indicate the percentage of households participating in each activity, and the value (or IL) rows show their average income from that source. For instance, in Dar es Salaam, 36.3 percent of all households have positive business income, and that income averages more than Tsh 2 million per year. Further, 60 -percent of households in Dar es Salaam have positive nonagricultural labor income, with an average value of Tsh 204,540 per year. However, a far lower percentage of poor or very poor households have a positive income from either business Table 4.14: Income Differentiation by Income Source for Households with Positive Income from That Source Income Group Location All DSM Income Tanzania Better-off Poor Very Poor Rural Poor: Rural Urban Per Capita Income p 68,436 113,580 13,001 11,571 12,597 41,582 94,588 224,246 Adult Equivalent p 101,147 166,960 20,329 18,313 19,769 65,318 121,517 339,263 Income Agricultural J 222,440 368,863 64,658 66,878 68,524 244,519 120,360 66,068 Income percent 68.2 67.3 69.1 67.9 73.5 80.9 48.2 17.1 Crop p 229,196 383,888 60,491 58,110 63,622 252,030 122,218 71,682 Production percent 63.1 62.7 63.6 63.8 67.6 75.0 39.8 15.2 p 27,742 42,675 13,214 11,018 13,825 27,662 27,216 49,217 Livestock percent 39.1 36.7 41.7 42.8 45.8 50.3 17.9 2.4 p 35,208 33,449 37,095 46,038 39,060 34,692 40,976 31,415 Other percent 16.9 16.6 17.1 16.8 18.9 21.3 7.8 3.8 Non p 284,706 463,181 42,644 34,989 30,211 41,050 280,671 1,290,277 Agricultural Income percent 51.8 56.8 46.3 46.8 45.0 44.6 63.7 79.4 p 108,567 . 135,609 43,446 43,700 39,628 42,145 99,231 204,540 Wage Labor percent 20.8 27.9 12.8 11.5 8.7 10.5 37.1 60.8 Business p 734,493 1,078,091 88,685 78,514 58,000 92,131 596,880 2,435,654 Income percent 16.3 20.2 11.9 10.6 11.4 11.8 22.3 36.3 Transfer p 17,366 20,766 14,070 13,837 12,680 13,457 24,747 36,763 Income percent 20.7 15.8 18.0 19.7 18.8 17.7 15.5 13.6 Other p 34,291 61,815 11,250 4,310 10,088 7,652 50,866 164,466 Income percent 10.6 9.2 12.1 12.8 11.3 10 12 11.7 p 346,114 563,948 78,619 74,012 76,173 244,841 289,628 1,256,669 Total Income percent 85.9 90.2 84.0 82.5 83.3 87.9 80.8 81.9 p = Mean household income for households with strictly positive income from source (TshI. percent = Percentage of households with strictly positive income from the source. activity or nonagricultural wage labor.' ' Readers should note that the per capita and adult-equivalent income figures from Table 4.14 do not match those from Table 4.13 because Table 4.14 aggregates data only from households with a positive income. 5 CONCLUSIONS AND POLICY IMPLICATIONS THE STATE OF POVERTY 1. Poverty in Tanzania is connected to the low growth of the economy and the unequal distribution of income. The overall per capita income is Tsh 55,369 per annum, and per capita expenditure is Tsh 61,564. Per capita income in rural Tanzania is found to be less than half that of Dar es Salaam. 2. Poverty is particularly prevalent in rural areas. The incidence of poverty is 59 percent in villages, compared to 39 percent in urban areas other than Dar es Salaam, and 9 percent in Dar es Salaam itself. For the country as a whole, 85 percent of the incidence of poverty is rural and rural villages account for 90 percent of total hard core poverty in Tanzania. These results lead us to conclude that the rural areas of Tanzania should have priority in poverty alleviation programs. 3. Empirical evidence for other African countries shows that female headed households in rural areas are particularly affected by poverty. This does not seem to be the case in Tanzania. Nonetheless, it is important to remember first, that there are few female headed households in the poorer, rural areas, and second, that this does not conclude the debate on the gender differentiation of poverty. Individuals were classified in our survey depending on whether their households were poor, but different members of households may do better or worse in the distribution of income within the household. Additional data on nutritional status, health and education expenditures by member of the household is needed to shed light on this issue. 4. Poverty is put into concrete terms when we focus on the fulfillment of basic needs. Here again, the rural areas are doing worse than urban areas in the quality of their housing, access to water, and level of education. 32 percent of the is illiterate, which implies either a rise in illiteracy since 1978, when illiteracy was estimated at 15 percent, or a problem in official statistics. 88 percent of the older than 14 does not have any education beyond primary school, and this rises to 94 percent and 95 percent of the poor and hard core poor, respectively. Of the few who go beyond primary school, a negligible number live in rural areas. 5. Water is another basic need in short supply. Only 10.6 percent of households have water readily available, at the door. About 43 percent of households have to walk up to 15 minutes to fetch water and 19 percent spend at least 30 minutes walking to their nearest water source. Among the poor, 56 percent of households spend more than 15 minutes to reach water, and the proportion is even higher (58 percent) among the very poor. Given the different sources of water for rural residents, and the poor, it is likely that the poor will be more exposed to disease than urban residents. Conclusions and Policy Implications 57 STRATEGIES FOR REDUCING POVERTY 6. The Cornell/ERB survey covered a sample of 1,046 households, out of 4.3 million of households in Mainland Tanzania. Thus, the findings presented here should be taken with some caution. Where strategies are suggested, they are meant to be hypotheses to be weighted against other evidence and alternative approaches. 7. One important option is to accelerate economic growth, and hence per capita income. There is ample evidence from around the world, but especially from East Asia, to show that poverty can be effectively reduced through accelerating economic growth, especially if growth is generated in labor-intensive sectors of the economy. Income redistribution policies can also play a role in successful case studies of poverty reduction, and recent studies (e.g., Clarke 1992) have shown that economic growth and income redistributive policies are compatible. Neither income growth nor redistribution alone is sufficient to eliminate poverty. 8. Further research on asset holdings will clarify whether raising the return on assets, especially land, labor, and livestock, will be more important than increasing the access of the poor to productive assets per se. This Report shows that the rural poor own all of these assets, so raising the returns to agricultural production by reducing taxation on the rural sector and liberalizing markets should have a strong and immediate impact on the poor. Another strategy for increasing rural incomes would be to encourage options for non-farm employment for poor rural households. 9. Similarly, the quality and productivity of the assets of the poor can be increased. One of the key assets owned by the poor is their human capital, which can be increased through more and higher quality primary and secondary education. 10. Because agriculture is the dominant sector, particularly for the poor, poverty reduction also necessitates attention to the economic environment for agriculture. Options include accelerated introduction of better technology, via investment in agricultural research and extension, improved physical infrastructure, and better communicate. More research on the regional pattern of poverty will help isolate priority areas for investment in markets, infrastructure, and communications. 11. A standard strategy to alleviate poverty is for the government to target subsidies for social services to the poor. To ensure that targeting works effectively, the government and other agencies will need to strengthen poverty monitoring systems that update policy makers on the economic status of the poor. Effective poverty monitoring systems require not only that the data exist, but that it be promptly processed. A summary of selected indicators should be promptly calculated, allowing for quick inference of the effects of policy changes on welfare. TARGETING THE POOR "in poverty alleviation, targeting means trying to shift the benefits of public spending to the poor by selecting them as the direct beneficiaries of public programs." Ibn de Mlle, 1992 12. Targeting can be performed analytically, using the poverty measures outlined in Chapter Three. Assume that there are j sub-groups in the economy (where groups might be different regions or different socio-economic groups). P. is the national measure of poverty (either incidence, or depth. or severity). P,., is the poverty targeting indicator for the jth sub-group. For example. if 58 Conclusions and Policy Implications one were trying to minimize the depth of poverty for the nation as a whole, as measured by P,, then one should order the sub-groups in order of priority for targeting according to the value of P0, the head count poverty measure. In this case, a=1 and a- 1=0. 13. One can make two assumptions about targeting. The first is that all households will benefit equally from a poverty reduction policy. For instance,, each household could receive a value of Tsh 5,000 from nearer water sources. The other assumption is that households will benefit from a poverty reduction policy in proportion to the value of their incomes. For instance, one could assume- that a nearby water source was equal in value to five percent of each household's income. If the objective is to minimize P. at the national level, and assuming that each household benefits equally, then the different sub-groups should be ordered according to the values of P,,, (Poverty Targeting Indicator A or PTIA; see Grootaert and Kanbur 1990). If the households are not affected equally, then the j sub should be ranked according to the value:' P7D=[P..--PJJ where Y, is the mean income in the jth sub (see Box B.2 for a discussion of the poverty targeting indicators). 14. According to the poverty measures displayed in Table 3.5 and Table 3.8 for different socio- economic and regional groups, and assuming that all households within a given sub-population will be equally affected by the targeting policy, we arrive at the following priorities through this method: a. Location. If the objective is to minimize the depth and severity of poverty, then policy instruments should be directed to favor first rural villages, then urban areas outside the capital city, and lastly the capital city itself. b. Socio-economic groups. If the objective is to minimize the depth of poverty, then the priority group is rural businessmen and businesswomen, closely followed by farmers. If the objective is to minimize the severity of poverty, then farmers are the priority group, followed by resident rural government employees. 15. If we assume, however, that the benefits of an intervention will not be shared equally by all households within a given sub-population, a slightly different ordering, based on poverty targeting indicator B (PTIB) is appropriate. Such an ordering is shown in Table 5.1. a. Location. If the objective of the policy intervention is to minimize the depth and the severity of poverty, then policy should favor rural villages first, then urban areas outside the capital city, and lastly the capital city itself. b. Socio-economic groups. If the objective is to minimize the depth and severity of poverty. then the priority group is rural businessmen and businesswomen, followed by farmers.' For more on pwerty targeting indicators (FTI) see Kanbur 1987. 2 To minime the depth of poverty one should rank the groups according to the values of the head count index. Conclusions and Policy Implications 59 Table 5.1: Values of the Poverty Targeting Indicator PTIBx1 Os by Region and Socio- Economic Group a= 1 a=2 Target Group Poor Very Poor Poor Very Poor Region Villages 9.85 8.26 3.61 2.63 Urban 7.47 3.39 2.17 1.11 DSM 1.14 0.58 0.31 0.14 Socio-Economic Activity Rural Farmers 9.79 8.21 3.61 2.63 Businessmen/women (rural) 14.3 15.2 4.67 2.09 Resident Govt. Employee 2.5 3.13 1.21 1.08 Urban Private/ Self Employed 5.91 2.57 1.75 0.79 Businessmen/women (urban) 0.8 .095 0.16 .018 Govt. Employee 1.11 1.02 0.34 0.18 All Tanzania 8.03 6.19 2.85 1.93 16. If both PTIA and PTIB targeting indicators yield the same ranking, then the policy conclusion is clear cut - poverty alleviation should be targeted to the group indicated. Thus, we can say with confidence that the rural areas of Tanzania should have the highest priority in policy interventions. On the other hand, the ranking of socio-economic groups is not fixed. Instead it alters depending on the measure of poverty used and the assumptions regarding policy effects. However, what we do know is that under whatever pairing of measure and assumption, farmers are either the first or the second priority group. 17. Another category by which to rank different sub-groups of the poor is region. Although the poverty measures for regions are tentative, and subject to statistical error, the extent of poverty is so different by region, and poverty is so prevalent in some regions, that regional ranking could achieve a high level of efficiency in targeting. For instance, there are four regions (Kigoma, Shinyanga, Lindi, and Ruvuma) in which over 70 percent of all households are hard-core poor, and these regions could be prioritized for early intervention. BIBLIOGRAPHY Abel-Smith, B. and P. Rawal, 1992, "Report on the Potentiality of Cost Sharing: First Report - User Charges" (Health Sector Financing Study for the Government of Tanzania, mimeo) Appleton, S., D. Bevan, K. Burger, P. Collier, J. Gunning,' L. Haddad, and J. Hoddinotti 1991 "Public Services and Household Allocation in Africa: Does Gender Matter?" (University of Oxford, Centre for the Study of African Economies, mimeo) Appleton, S. and P. Collier 1992, "On Gender Targeting of Public Transfers," (University of Oxford, Centre for the Study of African Economies, mimeo). Bevan, D., P. Collier, and J. Gunning, with A. Bigsten and P. Horsnell 1989, Peasants and Governments: An Economic Analysis, (Clarendon Press, Oxford). Bidani, B. and M. Ravallion 1992, "Poverty in Indonesia 1990: A Consistent Regional Profile and its Implications for Regional Targeting," (The World Bank, Washington D.C., mimeo). Bishop, Formby, and Thistle 1991 Review of Income and Wkalth. "International Comparison of Tax and Transfer Progressivity: New Evidence from the Luxembourg Income Study." Boateng, E., K. Ewusi, R. Kanbur, and A. McKay 1990, "A Poverty Profile for Ghana, 1987-88: Social Dimensions of Adjustment in Sub-Saharan Africa," (Working Paper No. 5, World Bank, Washington D.C.) Booth, David, Flora Lugangira, Patrick Masanga, Abu Mvungi, Rosemarie Mwaipopo, Joaquim Mwami, and Alison Redmayne 1993, "Social, Economic and Cultural Change in Contemporary Tanzania: A People-Oriented Focus," Swedish International Development Agency, Stockholm. Clarke, R. G. 1992, "More Evidence on Income Distribution and Growth," Policy Research Working Paper, Transition and Macro-Adjustment (The World Bank, WPS No. 1064). Cochrane, S., J. Leslie, and D. J. O'Hara 1980, "The Effects of Education on Health," World Bank Staff Working Paper No. 405, (The World Bank, Washington D.C.) Collier, P., S. Radwan, and S. Wangwe, with A. Wagner 1986, Labor and Poverty in Rural Tanzania: Ujamaa and Rural Development in the United Republic of Tanzania, A Study prepared for the International Labor Office within the framework of the World Employment Program, (Clarendon Press, Oxford). Deaton, A. 1980. "Inequality and Needs," Living Standards Measurement Survey No. 1, (The World Bank, Washington D.C.) Deaton, A. and J. Muellbauer 1980, Economics of Consumer Behavior, (Cambridge University Press, Cambridge). Bibliography 61 Due, J. and C.Gladwin, 1991, "Impacts of Structural Adjustment: Progress on African Women Farmers and Female Headed Households," American Journal of Agricultural Economics, 73: 1431-39. Foster, J., J. Greer, and E. Thorbecke 1984, "A Class of Decomposable Poverty Measures," Econometrica, 52:761-765. Glewwe, P. and K. Twum-Baah 1991, "The Distribution of Welfare in Ghana, 1987-1988," Living_- Standards Measurement Study, Working Paper No. 75, (The World Bank, Washington D.C.) Grootaert, C. and R. Kanbur 1990, "Policy-Oriented Analysis of Poverty and the Social Dimensions of Adjustment: A Methodology and Proposed Application to Cote d'Ivoire, 1985-88," SDA Working Paper No. 3, (The World Bank, Washington D.C.) Haddad, L, 1991, "Gender and Poverty in Ghana: A Descriptive Analysis of Selected Outcomes and Processes," IDS Bulletin, University of Sussex, Institute of Development Studies, 22: 5-16. International Labor Organization 1982,"Rural-Urban Gap and Income Distribution," (Jobs and Skills Programme for Africa, Addis Ababa). International Labor Organization 1992, 1990/1991 Labor Force Survey Tabulations, to be published in a Labor Force Survey (draft). Kanbur, R. 1987 "Measurement and Alleviation of Poverty" IMF'staff Papers, Vol 34 (1). Jamison, Dean, and Lawrence J. Lau 1982, Farmer Education and Farm Efficiency, Baltimore, Md.: Johns Hopkins University. Johnson, M., A. McKay, and J. Round 1990, "Income and Expenditure in a System of Household Accounts," SDA Working Paper No. 10, (The World Bank, Washington D.C.) Lambert, P. 1989, The Distribution and Redistribution of Income, (Basil Blackwell Press, Oxford, England). Latham, M. C 1965, Human Nutrition in Tropical Africa, Rome, FAO. Lipton, M. and M. Ravallion 1992 "Poverty and Policy," in Handbook of Development Economics, Volume 3 Chapter 42, forthcoming (preliminary draft). Meush, B., H. Leutzner, and S. Preston, 1985, Socioeconomic Differentials in Child Mortality in Developing Countries, (New York: United Nations). Milanovic, B. 1992. -Distributional Impact of Cash and In-Kind Social Transfers in Eastern Europe and Russia." Policy Research Working Paper Series No. 1054 (The World Bank, Washington DC). Muellbauer, 1974. Inequality Measures, Prices, and Household Composition," Review of Economic Studies, 41:493-504 62 Bibliography Nafziger, E., 1988, Inequality in Africa: Political Elites, Pmletariat, Peasants nd the Poor, (Cambridge University Press, Cambridge). Northrop, E., 1990, "The Feminization of Poverty: The Demographic Factor and the Composition of Economic Growth," Journal of Economic Issues, 24: 145-60. Oshima 1992, "The Kuznets Curve and Asian Income Distribution Trends," Hitoisubashi Journal of Economics, 33:95-111. Politz, A., 1958, "Description of Operational Design and Procedures," Life Study of Consumer Expenditures, Vol 4 Ravallion, M., 1992, "Poverty Comparisons: A Guide to Concepts and Methods," Living Standards Measurement Study, Working Paper No. 88, (The World Bank, Washington, D.C.) Ravallion, M., 1992b, "Poverty Alleviation Through Regional Targeting: A Case Study from Indonesia," in The Economics of Rural Organization, eds. K. Hoff, A. Braverman, and J. Stiglitz, (Oxford University Press for the World Bank). Sarris, A. and R. van den Brink, 1993, Economic Policy and Household V1fare During Crisis and Adjustment in Tanzania, (Cornell University Food and Nutrition Policy Program). Scott and Amenuvegbe, 1990, "Effect of Recall Duration on Reporting of Household Expenditures: An Experimental Study in Ghana," Social Adjustment Working Paper No. 6, (The World Bank: Washington D.C.) Sen, A., 1992 "Political Economy of Targeting," (The World Bank, Washington D.C.) Sender, J. and Smith, 1990, Poverty, Class, and Gender in Rural Africa, (London and New York: Routledge.) United Nations Development Program and The World Bank, 1992, African Development Indicators. United Republic of Tanzania, 1978, Household Budget Survey 1976/77, Bureau of Statistics, Dar es Salaam. United Republic of Tanzania, 1992, Comprehensive Food Security Program, Dar es Salaam, October. United Republic of Tanzania, 1993, National Accounts of Tanzania 1976-1992, Dar es Salaam, August. van de Walle. D.. 1992 "Targeting," Outreach 6: Policy Views from the Country Economics Department. (The World Bank, Washington D.C.) van Ginneken. W., 1976. Rural and Urban Inequalities in Indonesia, Mexico, Pakistan, Tanzania, and Tunisia. (Genva, International Labor Organization) The World Bank. Various Years, Rbrld Development Report, (Washington, D.C.) The World Bank. 1992a. Aftican Development Indicators, (Washington D.C.) Bibliography 63 The World Bank 1992b, Tanzania: Aids Assessment and Planning Study, A World Bank Country : Study, (Washington D.C.) The World Bank 1992c, Poverty Reduction Handbook, (The Johns Hopkins University Press). The World Bank 1992d, Social Indicators of Development 1991-1992, (Washington D.C.) The World Bank 1993, Tanzania: Agriculture Sector Memorandum, White Cover Report - (Washington, D.C.), July 20. ANNEX A DATA SAMPLE DESIGN Geographic Sampling 1. The survey was made nationally representative by using the National Master Sample (NMS) of the Bureau of Statistics. This Cornell/ERB survey used a sub-sample of the NMS, covering 1,046 households in rural and urban areas of Tanzania." The NMS covers 100 rural villages (hereafter Rural), 70 clusters for the urban areas not including Dar-es-Salaam (Urban), and 52 clusters for Dar- es-Salaam itself (DSM).. The sample used in this project selected 30 units from the 100 NMS rural Table A. 1: The NMS Frame and the Cornell/Economic Research Bureau Sub-Sample Cornell/ Location NMS Frame ERB sample Rural .100 30 Urban areas outside DSM 70 20 Urban areas in DSM 52 52 villages, 20 units out of the 70 NMS urban areas outside Dar-es-Salaam, and all 52 NMS clusters for Dar-es-Salaam (see Table A. 1). 2. The 100 rural NMS villages were stratified to represent the following categories: (i) villages surrounding large towns (LTS); (ii) villages in low density districts (LDD); and normal villages (Normal). 3. In the urban areas not including Dar-es-Salaam the 20 clusters were selected according to the scheme displayed in Table A.2. In Dar-es-Salaam, the frame from the NMS was taken as the frame for the survey. The NMS clusters in the municipalities were stratified into high, middle, and low income neighborhoods. The small towns (other urban) were grouped with the low income neighborhoods to form a common frame from which 11 clusters were selected. See below for a summary of the survey format, and the sample means for some selected variables. Simple Random Sampling: -any of the possible subsets of n distinct elements from the population of N elements is equally likely to be in the chosen sample. Systematic (geographic) sampling (in this survey): the clusters were ordered serially according to the census coding system. In this way the selection ensured a geographical spread. Annex A: Data 65 The Weighting Mechanism 4. To compute the location weights we used the NMS weights, but these had to be adjusted because not all enumeration areas of the NMS were used (i.e.: the Cornell/ERB subsampled the NMS). Therefore, the location weight in the Cornell/ERB survey is given by: WJ* =Aj* W1, where f is the correction factor, w, is the NMS weight for the ith village, and w is the Cornell/ERB corrected weight for the ith village. Table A.3 displays, and summarizes, the values of the correction factor - fi* - for the several stratification regional areas. Household Sampling 5. A second level of stratification was adopted in the survey. For each geographical unit surveyed all households were first classified according to their main economic activity. The use of this supplementary information to obtain a stratified sample improved both the sample design, and . Table A.2: Urban and Dar es Salaam Sub-Sampling Locality NMS Frame Cornell/ERB Selection Sample Dar-es-Salaam 52 No selection 52 Municipality 70 Total 20 High Income 4 No selection 4 Middle Income 11 Systematic sampling 5 Low Income 25155 Systematic sampling 11 Other Urban 301 the sample estimators. Within each strata, obtained through a disproportionate stratification, the Table A.3: Correction Factors for Regional Sub-Sampling Locality f" Urban Dar-es-Salaam 1 Municipality ngwil iluUilt 1 Middle Income !.2 Low Income 5 Other Urban 5 Rural Large Town Surrounding 3 Low Density Districts 4 Normal 3.3 66 Annex A: Data Table A.4: Stratification for Economic Activity Locality Stratification Target Rural small farmer 10 large farmer 2 businessmen/women 2 government employee 2 Urban outside DSM self-employed/private employer 10 businessmen/women 3 government employee 3 DSM self-employed 3 businessmen/women 1 government employee 1 sample was generated by simple random sampling.3 Seven different economic activities were considered. Table A.4 gives the stratification of the households by economic activity. 6. The ERB/Cornell survey aimed to interview 60 households headed by large farmers, but only managed to survey five in 1,046 households because of insufficient size. Given that large farmers represent less than one percent of all households in the country, we decided to add them to the small farmers, and conduct the analysis in terms of farmers, rather than large farmers versus small farmers. 7. For a household belonging to the hth strata the sampling weight (factor from geographical unit to household level), depends on the total number and sampled number of households in each cluster. It is defined according to the following expression: Nk where Nh is the " size" of the hth strata, and n. is the "sample size" of the hth strata. 8. The final weight for the household takes account both of where the household lives and the economic strata to which it belongs. Therefore, the weight for a household living in village i and belonging to the hth strata - w, - is defined as wwl, i.e., the product of the household weight, and of the village weight. I Given simple random sampling in each stratum, it follows that the sample mean for each stratum is an unbiased estimator of the mean population of the stratum. Anner A: Data 67 CONTENTS OF THE CORNELL/ERB SURVEY Page 1: Household Roster and Education Level. Page 2: Non-household members - gifts and transfers information. Page 3: Household Ownership of Durables. Page 4-5: Land of the Household. Page 7: Farm Production - Masika Season. Page 8-9: Farm Inputs - Masika Season. Page 10: Farm Equipment - Masika Season. Page 11: Farm Assets - Renting Out, Masika Season. Page 12: Farm Production - Vuli Season. Page 13-14: Farm Inputs - Vuli Season. Page 15: Farm Equipment - Vuli Season. Page 16: Farm Assets - Renting Out, Vili Season. Page 17-19: Crop Sales, Marketing Season. Page 20-23: Livestock Assets and Income in the Last 12 Months. Page 24: Farm Labor Income in the Last 12 Months. Page 25: Other Agricultural Income in the Last 12 Months. Page 26: Farm Assets of the Household. Page 27: Nonfarm Wage Income in the Last 12 Months. Page 28-29: Business Income Main Activity in the last 12 Months. Page 30-31: Business Income Second Activity in the last 12 Months. Page 32-37: Business Assets and Expenditures-Sources of Finance. Page 38: Other Assets of the Household. Page 39: Income from Communal Activities. Page 40-41: Finance. Page 42: Loans Given and Loans Received in the last 12 Months. Page 45-48: Expenditures and Consumption During the Last Month. Page 49-50: Expenditures on Incentive Goods During the Last 12 Months. Page 51: Daily, Weekly; and Monthly Expenditures. . Page 52: Annual Expenditures. SAMPLE MEANS FOR SELECTED VARIABLES Table A.5: Selected Characteristics of the Households Surveyed Income Group Location Gender All Very Male Female Sample Mean Tanzania Poor Poor Rural Urban DSM Headed Headed Adult-Equivalent Household 115,957 25,513 18,535 80,140 129,607 165,582 116,814 109,342 Expenditures (Tsh) Households Interviewed ( 1046 344 224 477 310 259 926 120 Household Size 6.15 7.02 7.23 6.75 5.27 6.12 6.35 4.56 Land Ownership (ha/per capita) 0.54 0.57 0.54 0.72 0.61 0.11 0.56 0.41 Number of Plots (Per household) 1.55 1.73 1.74 2.04 1.23 1.04 1.59 1.20 Did HH buy land in the last decade? (Percent 9.3 6.9 8.5 8.4 8.4 11.9 10.0 0.3 68 Annex A: Data Table A.6: Regional Sampling Region Number of Households Kigoma 17 Tabora 32 Mwanza 80 Kagera 64 Rukwa 32 Shinyanga 32 Morogoro 32 Tanga 32 Kilimanjaro 80 Singida 16 Mara 17 Arusha 69 Undi 16 Mtwara 32 Ruvuma 16 Mbeya 46 Iringa 95 Dodoma 32 Coast 47 Dar-es-Salaam 259 Total 1,046 Annex A: Data 69 Table A.7: Sample Size used in the computations of Table 4.14 Income Group Location All Rural DSM Income Tanzania Better-off Poor Very Poor Poor Rural Urban 243 Agricultural Income 596 353 161 210 408 141 47 Crop 496 308 188 122 165 350 108 38 Production 341 190 151 102 137 275 57 9 Livestock Other 105 59 46 30 42 81 16 8 Non Agricultural 672 487 185 117 128 233 235 204 Income Wage Labor 388 311 77 46 41 83 155 150 Business Income 274 218 56 32 36 73 96 105 Transfer 163 108 55 38 43 75 53 35 Income Other 97 57 40 28 26 39 34 24 Income 910 609 301 197 235 437 261 212 Total Income 70 Annex A: Data Table A.8: Literacy in Tanzania (People older than 14) -- Number of Cases Literacy Illiteracy Male 14-19 359 45 20-29 432 45 30-39 305 45 40-49 196 44 50-59 127 50 60+ 71 67 Female 14-19 361 46 20-29 495 83 30-39 265 147 40-49 88 99 50-59 36 65 60+ 20 92 Total 2755 828 ANNEX B METHODOLOGICAL BACKGROUND A POVERTY PROFILE I. To construct a poverty profile three major decision have to be taken. First, one has to chose a criterion to rank households. Second, one has to chose a poverty line. Finally, one has to select a poverty index to measure poverty. These three issues will be discussed below. A recent paper by Ravallion (1992) gives an overview of the conceptual, methodological, and practical difficulties associated with poverty measurement. The following summary and discussion of the methodological approach used in this work relies heavily on his paper. How To RANK HOUSEHOLDS? "1%@ would also claim it to be un-controversial that inadequate command over commodities by prevailing social standards is the single most important dimension of poverty, and is also a key determinant along other dimensions, such as self respect." Lipton and Ravallion (1992) 2. Economic theory suggests the use of consumption of goods and services to measure household welfare. In this study, households are ranked by consumption per adult equivalent, as proxied by the sum of monetary expenditures of the household. Household income and expenditures will be analyzed in adult equivalent terms (see Deaton and Muellbauer 1980). Therefore, the selection criterion for the ranking of households will be the sum of monetized expenditures of the household (as a proxy for consumption), deflated by the number of equivalent adults in the household. Because it has been standard practice in empirical work, we will duplicate part of the analysis using per capita expenditure as the criterion to rank households. 3. Total household expenditures include the value of home-produced consumption, and take into account price differences across regions. Several advantages guide this choice (e.g., see Boateng et al. 1990). First, the incomes of the poor are very unstable, and there are usually consumption smoothing opportunities available. Second, consumption is likely to involve a higher degree of accuracy in its measurement than income. Third, utility and well-being depends on consumption, not on income. Nonetheless, it should be kept in mind that there are several shortcomings with this proxy (e.g., consumption varies over the life cycle and the borrowing opportunities that allow to smooth consumption are not as widely available for poor people as for non-poor). The ideal standard of living indicator also includes the consumption of leisure, the consumption of public goods and amenities, and intertemporal consumption. Nonetheless, unavailability of data and empirical feasibility preclude us from including these items. Further, for consumers occupying their own houses we have no measure of the implicit shelter expenditure. 4. Having chosen total expenditure as an indicator of household welfare, there is still the question of adjusting for household size and composition, as well as for regional price differences. Clearly. one wants the poverty lines to properly reflect differences in the cost of living across the regions, as well as to consider different family size and composition. Empirical evidence for 72 Annex B: Methodological Background Indonesia shows that ignoring spatial price differences is likely to produce an over estimate of aggregate poverty (Bidani and Ravallion 1992). ADULT EQUIVALENCE SCALEs Table B.1: Index of Caloric Requirements by Age and Gender for East Africa Gender Age group Male Female 0-2 0.40 0.40 3-4 0.48 0.48 5-6 0.56 0.56 7-8 0.64 0.64 9 -10 0.76 0.76 11-12 0.80 0.88 13-14 1.00 1.00 15-18 1.20 1.00 19-59 1.00 0.88 60 + 0.88 0.72 Source: Latham (1965) 5. Households have different sizes and compositions and therefore different needs. For example, children have different nutritional needs from adults. In the absence of estimated equivalence scales for Tanzania, we follow Collier et al.'s (1986) approach in this study. Economies of scale and size' are accounted for by using the Engel food scales estimated by Deaton (1980) and displayed in Table B.2. Differences in family composition - gender and age - are accounted for by using the caloric requirements by age and gender for East Africa (see Table B. 1). The household equivalence scale is computed as follows (see Collier et al. 1986 for more details): each member in the household is assigned, according to his or her gender and age, the respective caloric weight as in Table B. 1. These values are summed across all members of the household and multiplied by the respective average cost factor as displayed in Table B.2. This household equivalence scale is then used to deflate household expenditure. This "equivalized" expenditure (Lambert 1989) can then be used, unlike per capita expenditures, to compare the welfare of different households (Muellbauer 1974). 6. There are obviously some deficiencies in the use of these equivalence scales. Nevertheless, they are superior to the alternative of using per capita equivalence scales, which do not take into account economies of scale, or different nutritional requirements depending on family composition. EXPENDiTURE MEASUREMENT 7. The information to build the variable equivalent annual expenditure was taken from pages 45 through page 52 of the household survey (see Annex A). The questions on expenditures used different periods of recall. For food, the question "How much did you spend on food and drinks in the last..." was asked for the last 24 hours, the last two days, and the last seven days. There is empirical evidence for Ghana (Scott and Amenuvegbe 1990) that for frequently purchased items the reported average daily expenditure falls consistently as the recall period rises from one to seven days. ' For example. to cixik a meal for two does not require double amount of energy than cooking a meal for one. Annex B: Methodological Background 73 Table B.2: Index of Household Economies of Scale Cost Household Size Marginal Average (Number of Adults) 1 1.0 1.0 2 .892 .946 3 .798 .897 4 .713 .851 5 .632 .807 6 .632 .778 7 .632 .757 8 .632 .741 9 .632 .729 10 and up .632 .719 Source: Deaton (1980) Then it is constant from one to two weeks. For a one year recall period the behavior was erratic. Politz (1958) "spied" what people bought in the supermarket, and then immediately after asked them to report their expenditures. On average, reported expenditure was six percent smaller than actual. In conformity with the past experience, the average yearly expenditure reported by households within our survey, with one day, two days, and seven days recall was (excluding the zeros) Tsh 155,036, Tsh 133,024 and Tsh 106,461, respectively. 8. The information from the survey was used to maximize the probability of accuracy in the measurement of total yearly expenditure. For instance, total food expenditures were measured as an average of the daily, and bi-daily expenditures. In cases where the respondent answered "..don't know" for either one, or two, or both days, but had an answer for one week, this value was used to estimate expenditure on food. Each household was also asked, for a detailed list of food items, how much was bought in the market, in official stores, and/or how much was given as rent, or as a gift, and/or how much came from their own production. Whenever there was no answer on the value of the daily, bi-daily, or weekly expenditure, the monthly figure was used to estimate the expenditures on food. The consumption of own production, as well as rent, wages, and gifts of food items, was added to the value of expenditures in the market. For the rent and gift components of the food/non- food expenditure, the respondent was asked to evaluate their value. The value of consumption of own production was imputed by estimating an average open market price prevalent in the geographic unit to which the respondent belonged.2 9. The length of the recall period is positively associated with the magnitude of error for frequently purchased items. However, for infrequently purchased items, the "loss of memory" error is traded off against the error associated with infrequency of purchases. Some products are over-reported in shorter periods because of the infrequency of their purchase. So, to calculate expenditures on rent. water charges, medical expenses, and other occasional consumer items we asked "In the last 30 days how much did your household spend on the following (in Tsh)?". For expenditures on big appliances, house repairs, and improvements we used a one year period recall. 2 This implicitly assumes that the household is free to sell its product in the market 74 Annex B: Methodological Background 10. For homeowners, the implicit value of housing costs is not available. Two approaches were tried to predict implicit value. First, for renters, we regressed the rent amount on variables such as location, number of rooms, and some other housing characteristics (i.e.: the hedonic approach). We also tried to assess if there was any relationship between the total expenditures/income and the amount spent on rent. In both cases, and considering several alternative regressions, we never obtained a coefficient of determination over five percent. Such low values precluded us from using such estimates for implicit value. The other avenue considered (as suggested in Johnson et al. 1990) was to estimate a depreciation rate using the purchase price and the estimated value of the property, and then compute the value of services from the consumption of the housing durable. The average estimated depreciation rate was often negative, which is equivalent to a negative consumption. One problem with these approaches is that we had relatively few observations. In addition, in periods of high inflation people may retain unreliable perceptions of the present value of durables. Given these problems we decided not to impute any value. As it will be seen later, we deem that the major conclusions of this study are unlikely to be affected by this decision. 11. Economic theory, and empirical evidence suggests that the poverty line should properly reflect differences in the cost of living across the geographic areas to be compared. The Bureau of statistics does not have regional price indices available. Given the information available in the Cornell/ERB survey we constructed a price index for food items only. Since food expenditures are a big component of total expenditure (73 percent for Tanzania) we believe that this approach is worthwhile. The "regional" price index was constructed using the Th6rnqvist-Theil index. The Th6rnqvist-Theil index is a discrete approximation to a Divisia Index. Instead of using fixed weights (i.e. for only one region), the index uses as weights the consumption share of the various food products in both regions to be compared. In this way the patterns of consumption in both regions are considered, and thus the index is close to a true cost of living index. A different index was computed for the rural villages, the urban areas outside Dar-es-Salaam, and Dar-es-Salaam itself. Because the estimated food price indices were about the same in all three areas, no price deflation was considered when computing the equivalized income. This result seems in line with van Ginneken's estimation (1979) of a Fisher food price index (with rural areas as the base), and his findings that for 1970 food prices in urban areas were just nine percent higher than in rural areas. Given that the differential is not very high, and that all other product prices are likely to be much higher in the rural areas, we deem that using uncorrected expenditures (i.e.: not corrected for spatial differences in prices) may not be far from reality. In other words, the price differences are likely to cancel out.' POVERTY LINE 12. Currently, no official poverty line (division between poor and non-poor) exists for Tanzania. Several alternatives are available for choosing a poverty line. One can choose a poverty line that cuts off a given percentage of individuals in the income distribution. Under this approach poverty is never zero. Equivalently, one can analyze the standard of living of different income quintiles. Alternatively, and sometimes referred to in the literature as the "ideal approach," one can choose a basket that includes all the goods necessary for the standard of living that is seen as the minimum required by the customs of the country. We propose to take both a -pragmatic" approach, whereby the selected poverty line will be set at 50 percent of the mean equivalized household expenditures of According to the 1980 ILO study. prices of consumer goods were 50 percent higher in rural areas. Annex B: Methodological Background 75 the total , and an "ideal approach."" Measuring the poverty in relative terms makes sense in the context of a poor society, where the scarcity of resources does not allow one to help everybody. 13. It is well known that the choice of one specific definition of poverty has major consequences for the resulting found to be poor. It is therefore good practice to select different values for the poverty line, i.e.: perform a sensitivity analysis. Two poverty lines will be defined in this report. The first poverty line is set at 50 percent of mean, adult-equivalent expenditures (or 75 percent of the mean per capita expenditure) and its monetary value is Tsh 46,173 (approximately $US 230 per- capita at 1991 prices or less than a Dollar a day).' The second poverty line is based on the poverty line computed by an International Labor Organization (ILO) study in 1982, and evaluated at 1991 prices. This was an absolute poverty line, "based on an estimate of the minimum income essential to meet the basic need of food, and then topped up to take account of shelter, household goods, etc." (See Box B. 1 for a description of the computation). The monetary value of the second poverty line is Tsh 31,100, which is approximately equal to one-third of the mean adult equivalent expenditure (or one-half of the mean per capita expenditure. Tsh 31,000 a year corresponds to US$152, which is less than sixty cents a day at an exchange rate of 203 Tsh/$USD. 14. The households whose adult equivalent expenditure is less than Tsh 31,000 will constitute what we call the very poor (hard-core poverty), by contrast to the poor (soft-core poverty). The use Box B.1: The Absolute Poverty Line "We have taken account of three considerations in arriving at a basic food diet: (i) local eating habits, (ii) nutritional requirements, and (iii) cost. [.. . Thus local taste in food is for a diet consisting of a starchy staple topped up by a sauce of vegetables, or occasionally meat 1...) We shall choose an austere, and what might seem a monotonous diet. [They consider as foodstuffs maizemeal, beans, fats, and sugar.] In view of the arbitrariness of the costing of non-food needs we have eschewed this and instead shall assume that they cost half as much as food - or that food forms two-thirds of the poverty line. If pressed we would say that the TSH 600 poverty line (for a family of 1 adult male, 1 adult female, 2 children, and 1 male adolescent, and for the period of one month] so obtained would comprise 396 shillings for food, 54 shilling for clothing, 75 shilling for rent, plus 65 shilling for fuel, lighting, water, and 10 shilling for miscellaneous. The poverty line estimated by this study evaluated at 1991 prices corresponds to the value of 104,985 a year (in average, between 1981 and 1991 the prices are about 14 times higher), for a family of five. Considering the adult equivalence scales to be used in this study, and the type of family implicit in the estimation of this poverty line (3.3894 equivalent adults), we get, for a family of one, the value of TSH of 31,000 a year. Source: Rural-Urban Gap and Income Distribution, International Labor Organization, Addis Ababa 1982, and own calculations by the author. of more than one poverty line will add insight into whether the choice of poverty lines has a major impact on the socio-demographic characteristics of the poor. IS. Some studies available in the literature use different poverty lines for different sub-groups, with the most common differentiation occurring for urban and rural s. If the standard of living is " For example. see the Cote dIvoire (Boateng et al. 1990) poverty profile, where the cut offs were established at 50 percent of mean expenditure for the poor, and 25 percent for the very poor: or Ghana (Glewwe and Twum-Baah 1991). where the cui-off% were two-thirds and one-third respectively. ' The value of the purchasing power parity (PPP) is for 1991 equal to 40.621 Tsh/USD. and the exchange rate computed by the Atlas method Ls approximately 203 Tsh[USD. 76 Annex B: Methodological Background corrected for spatial differences in prices, and household size, setting different poverty lines for different groups will produce "inconsistent" poverty profiles (see Bidani and Ravallion (1992) for a discussion on this point). This study considers one poverty line to be used across all sub-groups. POVERTY INDICES 16. A good poverty index should satisfy three basic axioms: (Al) Monotonicity, (A2) the principle of Transfers, and (A3) the axiom of Decomposability or Additivity.6 The following poverty indices will be computed in this study: Head Count Measure The head count is the simplest measure of the incidence of poverty. It counts the number of poor in the country, i.e., measures the prevalence or incidence of poverty, but does not say anything about the depth of poverty. The head count is just a measure of the number of people/households living below the poverty line. It does not reveal if someone who is counted as poor has an income of nearly zero, or close to the poverty line. For instance, income transfers on one side of the poverty line will not be captured by this measure. This poverty index therefore violates both Al and A2. Nevertheless, its ease of interpretation and calculation contribute heavily for its popularity. The head count measure is given by: P-I N where q is the number of people living in households with income below the poverty line, and N is the total . Poverty Gap Index This index is the sum of the gaps between the income of each poor person and the poverty line. This is a good measure of the depth of poverty, but it is not sensitive to the severity of poverty. The sum of all poverty gaps in a is interpreted as the minimum amount of transfers necessary to bring all households/people up to the poverty line, if perfect targeting were possible.' This index violates A2, i.e., it is insensitive to redistribution of income among the poor. The poverty gap index is given by: 6 (1) Monotonicity: A reduction in income of a person below the poverty line should be reflected in an increase of poverty. (2) Transfers: A transfer of income from a person below the poverty line to someone who is richer must lead to an increase in poverty. (3) Additive Decomposability: The poverty index for a population can be written as a weighted average of the mutually exclusive and collectively exhaustive sub-group poverty indices. 'The maximum cost of eliminating the poverty in the country is z times the population of the country, i.e.. giving the poverty line to everybody in the country. The ratio of the minimum cost to the maximum cost is precisely the value of the poverty gap index (P,) (Ravallion 1992b). Annex B: Methodological Background 77 Nh-I z N z where y, is the average income of people below the poverty line, y. is the income of the hth individual, z stands for the poverty line, and I defines the income gap ratio. The value of NzP, is interpreted as the lower bound on the amount of transfers necessary to eliminate poverty. FGT Index of Poverty' 17. Commonly known as the P, index of poverty, the FGT index is given by: The value taken by alpha determines the relative weight given to the very poor in the index. Thus, a reflects concern about the severity of poverty. As a increases a higher weight is given to the poorest of the poor. The head count measure (a=0), the poverty gap measure (a= 1), and the P2 measure (a = 2) are particular cases of the P. measure of poverty. Usually the index is not calculated for values of alpha greater than two because of the difficulty with interpretation. The P, index measures the severity of poverty for values of alpha greater than one, i.e. is sensitive to the distribution of income amongst poor. The use of this class of indices is very attractive because it allows us to break down poverty by region, socio-economic group, or other categories, and to estimate the relative contribution of each group to total poverty. For a equal to two we have the standard FGT index. This index satisfies all three basic axioms Al, A2, and A3 (see Paragraph 16). Its relative absence from empirical work is due both to increased difficulty in computation, and a non straightforward interpretation. The FGT index of poverty for a = 2 is given by: N1 ., z THE DECOMPOSABILfTY PROPERTY 18. The decomposability property has important and interesting policy implications, since it enables analysis of the contribution to aggregate poverty of each socio-economic group, geographic region, or type of household according to the gender of the head. The contribution of each stratum to total/national poverty - Cj - is given by: ci =kjP. JP. where k, is the proportion of total located in the mutually exclusive subgroup/stratum j, P>, is the value of the index of group j, and P. is the value of the a index of poverty for the total. Note that P. can be written as EkPJJ. The value of C, will give a fairly clear idea of where the poverty in the nation is concentrated -and can be the basis for a policy targeting intervention. ' Named after its authors Foster. Greer. and Thorbecke (1984). 78 Annex B: Methodological Background 19. There are some other poverty indices available in the literature (e.g. the Sen measure of poverty), but for this study is was thought indispensable to have an index satisfying the decomposability property. The Sen (1976) index violates this axiom, as well as the transfer axiom. . INEQUALITY INDICES AND THE LORENZ CURVE 20. The Lorenz curve is a graphic representation of inequality, which displays the cumulative share of total income accruing to each cumulative share of the population, when incomes are ordered- from poorest to richest. However, if two income distributions are identical except that incomes in one are higher than the other, the two Lorenz curves will be identical. Nevertheless, the size of the cake in the two distributions is different. So, another representation is needed to assess the absolute differentials between poor and rich incomes. The Generalized Lorenz curve (Figure B. 1), unlike the Lorenz curve, incorporates both income and equity levels when comparing "income" distributions. In short, the Generalized Lorenz curve is, by construction, the mean income times the Lorenz curve. 21. The Gini Coefficient is an inequality index derived from the Lorenz curve. It measures how far a given distribution is from perfect equality, i.e. it measures the area between a given Lorenz curve and the Lorenz curve for a perfectly equal distribution (which is a 45 degree line). There are several indices available for measuring relative inequality. A satisfactory measure of relative inequality should satisfy the principle of transfers and of symmetry, while being scale invariant (Lambert 1990). Both Theil's scaled entropy coefficient, and the coefficient of variation are indices of relative inequality. Theil's scaled entropy coefficient is given by: The major advantage of the Theil coefficient, over the other indices mentioned, is that, like the P, measure, it is subgroup additive. Thus, overall inequality can be expressed as the sum of the inequality between, and within, groups. Annex B: Methodological Background 79 All Tanzania Rural Tanzania 1.59 1.59 .1.27 2r 1.27 E 0.95 E 0.95 0 0 C C 063 0.63 EE 0.32 0.32 0 00 0.00 00 0.1 02 0.3 04 0.5 0.6 0.7 08 09 10 00 01 02 0.3 04 05 0.6 07 08 09 10 Proportion 0f population Proportion of population Dar es Salaam Poor 1 59 .1.5 1.27 2 1.27 0.95 0.95 0 0 .C C 2 063 e 063 o o E E o0 32 L 0.32 000 000 O 0001 0.2 03 0' 05 06 0.7 0.8 0.9 1 0 0.0 0 1 02 0.3 04 05 06 07 08 09 1 0 P,oDortio, of pOUIllon Proportion of populolion Figure B.1: Generalized Lorenz Curve for the Income Distribution 80 Annex B: Methodological Background Box B.2: Poverty Targeting Indicators The idea here is that targeted policies will deliver poverty alleviation at much lower fiscal costs and are therefore preferred. [...] Faced with the high leakage costs of completely untargeted policies and the high administrative costs of very finely targeted policies, we are left with the option of an intermediate position where certain easily observable characteristics are used to target. [...) Suppose, then, we have chosen a small number of policy relevant disaggregation of the population - region is the most obvious candidate [...] Which region should have priority for a favorable policy stance in terms of investment or other income increasing measures? In posing this question it is understood that there will be leakage since both rich and poor in the category will benefit - this is unavoidable if fine targeting is not possible. However, by prioritizing categories we can minimize the leakage. Clearly the policy stance should favor the category whose improvement in income status will have the maximum impact on poverty at the national level. [...] The first set of targeting rules are derived under the [...] assumption that the fruits of the policy are shared equally by all in the targeted sub-group; that is, incremental income is distributed equally, so that the poor benefit proportionally more. In this case it can be shown that if the objective is to minimize P, at the national level then the categories should be prioritized according to the values of Pi.,, .1...] If it is felt that in fact the fruits of income increases are not shared equally then an alternative scenario can be considered - where all incomes increase in the same proportion, so that the rich benefit more than the poor in absolute terms. For this case it can be shown that if the objective is (to minimize] P, at the national level then the group should be ranked according to {P,,,, - P.,1]/Y, where Y, is the mean income in group j. Essentially, the targeting indicator above gives the impact on national poverty of an increase in category j's income which is absorbed by individuals in the category in proportion to their current income.( ...] If a category retains priority [according to both criteria], and for a range of a values, then in the policy context we can say with confidence that policy instruments should be directed to favor that category. Source: Adapted from Grootaert and Kanbur (1990) Box B.3: Index of Livestock Values Collier et al. (1986) assert that in a country where land is not a scarce factor livestock is likely to be the best indicator of wealth in a country. An index of livestock ownership was computed in the following manner: the weights were used to add up different kinds of livestock. This index (basis is 100 for local bulls) was computed as the average of the value (as indicated by the households) of each kind of livestock, and then converted to a common basis. Bulls (local) 100 Bulls (improved) 159 Oxen 117 Cows (local) 90 Cows (improved) 251 Sheep 14 Goats 16 Pigs 37 Chickens 3 Ducks and Geese 3 Donkeys 36 ANNEX C ADDITIONAL TABLES Table C.1: Decomposition of P. Poverty (Soft) Index by Gender of the Head of the Household For a Poverty Line of Tsh 46,173/year per adult equivalent Head Count (PO) Depth (P,) Severity (P2) Contri- Contri- Contri- PopulationS bution bution bution hare Female 55.6 6.9 24.4 6.27 14.4 5.9 6.4 Male 50.8 93.1 24.9 93.7 15.7 94.1 93.6 All Tanzania 51.1 100 24.9 100 15.6 100 100 For a Poverty Line of Tsh 31,000/year per adult equivalent Contri- Contri- Contri- Population bution bution bution Share Female 30.7 5.5 13.9 5.7 8.4 5.7 6.4 Male .36.3 94.5 15.8 94.3 9.4 94.3 93.6 All Tanzania 35.9 100 15.7 100 9.4 100 100 82 Annex C. Additional Tables Table C.2: Numbers of Livestock Owned Mean Number of Livestock Owned by Percent of Households Who Own Households with Some Livestock Livestock All Poor Very Poor Rural All Poor Very Poor Rural Bulls (local) 13 13.2 15.3 17.8 3.5 3.13 3.3 3.51 Bulls (improved) 1.3 0.7 0.4 1.6 2.17 3.23 3.66 2.05 Oxen 4.4 4.7 6.5 6.1 3.66 3.7 3.68 3.52 Cows (local) 20.4 18.5 20.9 27.6 10.59 11.5 9.21 10.59 Cows (improved) 1.9 0.7 0.7 2.1 6.21 2.67 2.0 6.87 Sheep 8.4 19.7 11.3 11.7 7.94 7.38 7.48 8.05 Goats 22.6 23.5 23.9 30.1 12.6 14.76 17.19 13.03 Pigs 3.1 2.7 2.8 3.7 3.26 2.59 2.7 2.84 Chickens 23.6 26.8 27.5 30.7 15.6 10.77 10.73 13.84 Ducks and Geese 4.4 4.9 5.9 4.6 5.5 5.54 5.56 5.88 Donkeys 1.7 2.6 3.8 2.4 2.51 2.79 2.89 2.51 Annex C: Additional Tables 83 Table C.3: Reason to Stop Attending Primary School Income Location Gender All Better Very Dar es Tanzania Off Poor Poor Rural Urban Salaam Male Female Complete and no desire 27.6 22.9 33.7 36.2 28 25.1 29.1 28.2 26.7 to continue incomplete but no desire 13.4 10.7 16.9 18.1 14.6 12.1 9.1 13.7 13.0 to continue Unable to gain entry into 3 3.3 2.3 1.1 3.4 3.3 0 2.5 3.6 a higher level Unable to meet expenses 39.4 46.1 30.7 29.6 36.6 40.0 55.5 44.4 33.3 Sick 5 4.7 5.6 5.2 4.4 7.9 3.6 3.6 6.7 Parents did not want me 8.2 9.1 7.0 6.6 9.0 7.9 2.7 5.6 11.5 to continue Other 3.4 3.2 3.8 3.1 4.0 3.7 0 2.0 5.2 Total 100 100 100 100 100 100 100 100 100 84 Annex C: Additional Tables Table C.4: Reason to Stop Attending School, Forms 1 to 4 Income Location Gender All Better Very Dar es Tanzania Off Poor Poor Rural Urban Salaam Male Female Complete and no desire 47.4 54.2 17.5 40.0 36.4 63.0 44.6 34.5 65.4 to continue Incomplete but no desire 7.5 5.3 16.9 32.0 7.0 9.7 6.0 4.7 11.5 to continue Unable to gain entry into 1.9 2.3 0.0 0.0 3.4 2.6 0.0 3.3 0.0 a higher level Unable to meet expenses 30.4 30.3 31.1 28.0 27.5 18.8 44.3 43.6 12.4 Sick 7.7 1.7 34.4 0.0 14.0 5.8 2.7 12.6 0.7 Parents did not want me 0.8 1.0 0.0 0.0 0.0 0.0 0.0 1.4 0.0 to continue Other 4.3 5.3 0.0 0.0 11.8 0.0 2.4 0.0 10.0 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Annex C: Additional Tables 85 Table C.5: Reason to Stop Attending Higher Level Education (Forms 1-4 through University) Income Location Gender All Better Very Dar es Tanzania Off Poor Poor Rural Urban Salaam Male Female Complete and no desire 50.2 55.4 22.8 48.6 33.9 63.4 53.2 40.8 66.0 to continue Incomplete but no desire 6.1 4.3 16.0 27.8 6.6 8.5 4.2 3.5 10.5 to continue Unable to gain entry into 1.6 1.9 0.0 0.0 3.2 2.1 0.0 2.6 0.0 a higher level Unable to meet expenses 29.0 29.0 29.0 0.0 25.6 19.7 38.3 39.1 12.3 Sick 6.8 1.9 32.1 23.6 13.0 6.0 2.5 9.6 1.9 Parents did not want me 0.7 0.8 0.0 0.0 0.0 0.0 1.7 1.1 0.0 to continue Other 5.6 6.7 0.0 0.0 17.7 0.4 0.0 3.3 9.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 86 Annex C: Additional Tables Table C.6: Intensity of Work by Days Available Income Group Location All Tanzania Better-off Poor Very Poor Rural Urban DSM Own crop Rural 11.0 10.0 12.1 12.7 13.7 6.6 0.63 Wage rural .37 .39 .34 .29 0.5 .06 .04 Total rural 22.3 21.7 22.9 23.3 28.5 10.6 1.5 Non-farm wage 7.20 10.3 3.7 2.7 3.19 13.9 22.1 Business 4.0 7.5 1.9 1.0 1.41 10.88 16.7 Communal 0.91 .92 .91 1.1 1.1 .36 .79 Child livestock days as percentage of total days 9.8 7.8 11.9 13.8 12.1 4.4 2.5 child available Annex C: Additional Tables 87 Table C.7: Land Use Patterns for Owners or Renters Income Group Location Gender All Male Female Use of Land Tanzania Better-off Poor Very Poor Rural Urban DSM Headed Headed Homestead 12.07 14.52 9.99 8.46 9.94 11.81 67.28 12.21 10.38 Crops 86.55 84.08 88.64 89.55 88.80 86.09 31.53 86.48 87.33 Grazing 0.49 0.10 0.82 1.2 0.28 1.76 0 0.53 0 Unused 0.42 0.54 0.32 0.46 0.40 0.34 1.18 0.27 2.28 Useless 0.48 0.77 0.23 0.33 0.59 0 0 0.52 0 Total 100.01 100.01 100.00 100.00 100.01 100.00 99.99 100.01 99.99 88 Annex C: Additional Tables Table C.8: Land Tenure for Land Owners or Renters Income Group Location Gender All Male Female Land Tenure Tanzania Better-off Poor Very Poor Rural Urban DSM Headed Headed Inheritance 36.89 40.70 33.62 32.64 38.33 29.19 36.28 36.46 42.12 Allocation by Village 32.26 31.68 32.75 34.69 33.80 30.84 2.37 31.80 37.81 Borrowed or Rented 1.78 1.91 1.66 1.28 0.48 9.12 0.30 1.90 0.24 Acquiredthrough 17.38 10.69 23.11 22.22 19.33 10.5 0.00 18.30 6.21 Clearing Purchased 9.93 13.00 7.30 6.92 6.78 16.74 57.60 9.79 11.59 Other 1.76 2.02 1.56 2.25 1.28 3.61 3.45 1.75 2.03 Total 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 Annex C: Additional Tables 89 Table C.9: Landholdings per Household and Distribution of Landholdings All Tanzania Income Group Location Poor Very Poor Rural Urban DSM Mean 3.94 A.43 4.44 4.66 3.1 0.55 Coefficient of Variation 1.70 1.51 1.46 1.41 2.48 5.49 Gini Coefficient 0.60 0.53 0.50 0.50 0.77 0.93 Gini Coefficient." 0.52 0.49 0.47 0.48 0.62 0.76 Theil Coefficient 0.52 0.44 0.41 0.39 0.73 0.93 P For those with some land. 90 Annex C: Additional Tables Table C.10: Regional Per Capita GDP for 1991 Region GDP per capita Arusha 24,716 Coast 6,595 Dar es Salaam 53,482 Dodoma 10,747 Iringa 44,331 Kagera 8,582 Kigoma 27,179 Kilimanjaro 12,612 Lindi 14,590 Mara 9,380 Mbeya 30,168 Morogoro 10,303 Mtwara 18,122 Mwanza 13,483 Rukwa 43,734 Ruvuma 44,543 Shinyanga 28,217 Singida 14,196 Tabora 38,256 Tanga 15,165 Tanzania 23,084 Source: National Accounts of Tanzania 1976-1992