Tanzania - The challenge of reforms : growth, incomes and welfare (Vol. 2 of 3) : Appendices and statistical annexes

Report No. 14982-TA Tanzania The Challenge of Reforms: Growth, Incomes and Welfare (In Three Volunies) Volume Il: Appendices and Statistical Annexes May 31, 1996 Countrv Operations Division Eastern Africa De[partment Africa Region Document of the World Bank  Government Fiscal Year FY94=July 1, 1993 to June 30, 1994 Currency Equivalents Currency Unit: Tanzania Shilling (T Sh) Bureau Selling Rate: US$1.00 = T Sh 572 (March, 1996) Bureau Buying Rate: US$ 1.00 = T Sh 552 (March, 1996)  Tanzania The Challenge of Reforms: Growth, Incomes and Welfare Volume 2: Appendices and Statistical Annexes Contents Appendix to Chapter 1 ...............1 Appendix to Chapter 3 ...............3 Appendix to Chapter 4 .............. 35 Appendix to Chapter 5 .............. 37 Statistical Annexes   APPENDIX TO CHAPTER 1 DATA ISSUES 1. The limitations of the data need to be noted at the outset. In particular, the official estimates of the national income accounts grossly distort the level of economic activity. The official estimate of per capita income is about US$100; however, various other estimates, including separate household surveys undertaken in 1991 and 1993 indicate a per capita expenditure in the range of about US$200. A review of the national income accounts undertaken by the IMF in 1992 indicates a per capita income of US$142, which is also higher than the official estimate of about US$100. The revision involved updating various components of the national income accounts, including investments and exports. Despite these qualifications, the data are useful in indicating general trend in performance. 2. A major reason for the poor quality of the data is that the capacity to collect and process information is lagging behind the changes taking place in the economy. For example, the base for the national income accounts is still the 1976/77 household budget survey, which is outdated. Also, with the recent expansion of the private sector, data currently provided by parastatals, which used to serve as data collection points, no longer reflect overall economic activities. Similarly, informal activities, which emerged since the mid-eighties are not adequately documented. Moreover, the collection of data on private construction, for example, was disrupted following the abolition in 1972 of town and city councils, which kept such records; and though the councils were reconstituted in 1982, the data collection was not fully revived. 3. Various agencies, including the Bank of Tanzania (BOT), have been attempting to fill the data gap through alternative data estimates. For example, BOT estimates import figures are based mainly on Import Declaration Forms. While these efforts need to be strengthened, the central statistics departments need more capacity and resources to improve data collection and processing. A technical assistance program to computerize the Customs Department is being implemented. However, more is required in the form of far reaching institutional reforms to provide qualified and adequate personnel, equipment, legal authority and other necessary resources. Currently, the Government is making efforts to help the Central Bureau of Statistics improve the national income accounts. It will: (i) integrate the recent household and other surveys into the national income accounts, (ii) identify gaps in the existing data, and (iii) design an approach to fill the gap and provide more reliable national income accounts.   APPENDIX TO CHAPTER 3 APPENDIX 3.1 HOW TO RANK HOUSEHOLDS "We would [..] claim it to be uncontroversial 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)' 1. Economic theory and common sense suggest the use of consumption of goods and services (proxied by expenditure) to measure household welfare. This measure of welfare includes the value of home-produced consumption. For example, for consumers occupying their own houses consumption includes the rent they would have to pay for a similar house given the current market conditions. There are several advantages to the choice of consumption as an indicator:2 (i) The incomes of the poor are very unstable, and there are usually consumption smoothing opportunities available. (ii) Consumption is likely to involve a higher degree of accuracy in its measurement than income; and (iii) Utility and well-being depend on consumption, not on income. 2. However, households with different sizes and compositions have different needs. Thus, as an indicator of household welfare, consumption expenditure still has to be adjusted for size and composition. For example, children have different nutritional needs from adults. In addition, larger households may need less resources per person, because to cook a meal for two does not require double the energy needed to cook a meal for one (i.e., there are economies of scale). M. Lipton, and M. Ravallion, "Poverty and Policy" in Handbook of Development Economics, Vol. 111, Chapter 42, forthcoming (preliminary draft), 1992. 2 For example, see E.K. Boateng, Ewusi, R. Kanbur, and A. McKay, "A Policy Profile for Ghana, 1987-88. Social Dimensions of Adjustment in Sub-Saharan Africa, ". Working Paper No. 5, World Bank, Washington D.C., 1990.  4 Appendix to Chapter 3 3. Differences in family composition (i.e., the gender and age of family members) and family size are accounted for by estimating the number of adult equivalents in the household. In the absence of estimated equivalence scales for Tanzania, this study follows the approach used by Collier, et al.3 Economies of scale are accounted for by using the Engel food scales estimated by Deaton.4 Differences in family composition and different nutritional needs are accounted for by using the caloric requirements by age and gender for East Africa.5 Total household expenditure is then divided by the number of adult equivalents to arrive at expenditure per adult equivalent, which can then be used to rank households in welfare terms.6 For instance, consider two households-A and B- with the same annual expenditure (of T Sh 70,000) and different sizes. Family A is composed of 2 adults and 2 children (comprising 3.5 adult equivalents), and family B is composed of 4 adults (comprising 4 adult equivalents). Given that family B has greater nutritional needs, family A is considered to be better off. Family A has an adult equivalent expenditure of T Sh 20,000 while family B has a lower adult equivalent expenditure of T Sh 17,500. By contrast, comparing families on the level of household income would have led to the conclusion that the two were at the same level of welfare. 4. Once household expenditure is adjusted so that welfare comparisons are possible, households are ordered by adult equivalent expenditure and divided into five different groups comprising an equal proportion of the population-or quintiles. Thus, each quintile contains 20 percent of the population, ordered from those with the lowest expenditure per adult equivalent (quintile 1) to those with the highest expenditure per adult equivalent (quintile 5). However, when analyzing people in rural locations, urban locations outside Dar es Salaam, and Dar es Salaam, by each quintile, then the quintiles used are those defined at the national level. For instance, the poorest 20 percent of the population, who are in the bottom quintile for the nation, all have an expenditure level less than T Sh 79,335. However, because there are many more poor households in rural areas, more than 20 percent of the rural population will fall below this expenditure level (in fact, about 25 percent of the rural population is in the bottom quintile). Although this results in quintiles of unequal population sizes this approach is considered preferable to alternatives, because it compares people with comparable expenditure levels across rural and urban locations. P. Collier, S. Radwan, and S. Wangwe, with A. Wagner, "Labour 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" (Oxford: Clarendon Press), 1986. 4 A. Deaton, "Inequality and Needs," Living Standards Measurement Survey No. 1, (Washington, D.C., The World Bank, 1980. 5 M.C. Latham, Human Nutrition in Tropical Africa, (Rome: FAO), 1965. 6 J. Muellbauer, "Inequality Measures, Prices, and Household Composition, " Review of Economic Studies, 1974, 41: 493-504.  Appendix to Chapter 3 5 MEASURES OF POVERTY 5. The first requirement for measuring poverty is to select and justify a poverty line i.e., the level of expenditure or income below which people or households are considered poor. Currently, no official poverty line exists for Tanzania, but this study uses one of the poverty lines employed in the 1993 Poverty Profile for Tanzania updated for inflation. 6. Once a poverty line has been selected, it can be used to compute measures of poverty. Two standard measures presented in the Poverty Assessment are as follows.7 7. Head Count Measure. The head count measure is the simplest index of poverty. The number of people or households below the poverty line is divided by the total number of people or households, to give the proportion living in poverty. This measure reports the prevalence or incidence of poverty, but says nothing about how far poor households are below the poverty line. It is sometimes referred to as the P0 index. 8. Poverty Gap. This measure indicates how far, on average, a poor household lies below the poverty line. It is therefore said to measure the depth of poverty, and is referred to as the P1 index. The sum of all poverty gaps is interpreted as the minimum amount of transfers required to bring all people or households up to the poverty line, if perfect targeting were possible. APPENDIX 3.2 TARGETING ON HOUSING CHARACTERISTICS 9. This section shows the usefulness of a composite index of housing quality as an indicator of welfare, and it compares this to the simpler approach of looking directly for lower quality housing materials. The composite index was built up from three dummy variables which showed whether or not a households was characterized by at least one of three lower-quality housing features: an earthen floor, a straw roof or a roof which was permeable to rain, and a crude, uncovered window or a lack of windows. 10. The performance of indicators is judged on the basis of efficiency and effectiveness. Efficiency is defined as the extent to which one can distinguish between poor households and nonpoor households on the basis of a characteristic or index. An efficient indicator will help to identify the poor, with few errors. Effectiveness is defined as the likelihood that poor households will actually display a given characteristic and will thus be identified as poor. The two terms have a natural interpretation in terms of conditional probabilities. Efficiency implies that the conditional probability of a household being poor, given that it shows some characteristic, is high. Effectiveness A far more complete description of these measures is given in Tanzania: A Poverty Profile, World Bank, Washington, D.C., 1993.  6 Appendix to Chapter 3 implies that the conditional probability of having a certain characteristic, given that one is poor, is high. 11. Table 3.1 below shows the conditional probabilities that will allow one to rank the performance of indicators based on the simple housing characteristics and the composite index. A poor person is defined in this section as someone living in a household whose adult equivalent expenditure falls in the poorest two quintiles. Table 3.1: Effectiveness and Efficiency of Housing Indicators Mainland Tanzania (in percentage) Probability Composite Earth Permeable Low Quality Index Floor Roof Window P(Poor/Characteristic) 44.1 44.3 47.9 54.4 P(Characteristic/Poor) 93.5 92.0 74.3 33.6 12. The upper row of the table shows that the composite index of housing quality is not much more efficient as a means of identifying the poor than picking people at random. The likelihood of finding someone in the bottom two quintiles, by picking at random, is 40 percent, while the likelihood of finding someone in the bottom two quintiles using the composite index is 44.1 percent. Looking for simple evidence of low- quality housing materials is more efficient, with the presence of an uncovered window, or no window at all, being the most efficient indicator of poverty. However, these increases in efficiency would be won at the cost of effectiveness, since some poor households would not be identified using these basic indicators of housing quality. The bottom row of the table shows that a poor household is less likely to be identified on the basis of their window type than on the basis of their roofing, their flooring, or the composite index. APPENDIX 3.3 DATA SOURCES AND CAVEATS 13. The primary sources of household data used are the 1991 Cornell/ERB survey and the 1993/94 Human Resource Development (HRD) Survey.8 Both surveys are nationally representative and based on the National Master Sample, created by the Bureau of Statistics in the Planning Commission. The 1991 survey interviewed 1,047 households, while the 1993 survey covered 5,000 households in Mainland Tanzania. Additional data sources are the 1983 Rural Household survey, the published results of the 1976 Household Budget survey conducted by the Tanzania Bureau of Statistics, and the 1995 See World Bank, Tanzania: A Poverty Profile, Washington, D.C. for details on the Cornell/ERB Survey, 1993.  Appendix to Chapter 3 7 Rural Participatory Poverty Assessment.9 Some data caveats may be in order. Only the 1993 and 1995 surveys were designed to be directly comparable. There are differences in the survey design, namely, whether income or expenditure should be measured, and how to measure it. Sample size also differs between surveys. Both the 1991 and 1993 surveys were undertaken by the same institution under the supervision of Prof. Amani from the Department of Economics at the University of Dar es Salaam. Furthermore, several results, such as the shares of expenditure on health, education, and food, the ownership of assets, and educational attainment, indicate consistency of results across surveys. 14. However, comparison of some results reveals large differences and the reader should be aware of important factors that may be driving these differences. * Price variation between 1991 and 1993. It is not easy to obtain an accurate estimate of how much prices increased over the two years. Depending on the sources used, one finds a price variation of 43 percent (GDP deflator- International Financial Statistics) or 67 percent (Private Consumption-Official Statistics). Between these extremes there is a wide range of values for all possible tastes. For example, the GDP deflator presented in the Agricultural Sector Memorandum (p. 208) yields a price variation for the period of 58.9 percent, and the document shows that the GDP deflator has generally given a lower inflation figure than the consumer price indices. The choice of inflation rate is very important when estimating the 1991 poverty line in 1993 prices. * Price increases on basic goods between 1991 and 1993 are clearly higher than one would expect using the Consumer Price Index.'0 According to van den Brink and Cotlear's calculations: "Food prices in Tanzania may have been risingfaster than was implied by the National Consumer Price Index [(NCPI)] since November 1990. Over the November 1990-November 1991 period, the open market index suggests that the main food prices rose by approximately 75 percent, whereas the food component of the NCPI indicates an increase of only 22 percent. To obtain an aggregate NCPI inflation rate of 22 percent in the presence of a food price increase of 75 percent, one would need a deflation of the nonfood component of 75 percent." * Food price estimates from the Ministry of Agriculture also show that maize prices rose by much more than the National Consumer Price Index. This is of special For details on the 1983 rural household survey see D. Bevan, P. Collier, and J. Gunning, with A. Bigsten and P. Horsnell, Peasants and Governments: An Economic Analysis, (Oxford: Clarendon Press), 1989. 10 The following conclusions draw on World Bank Office Memorandum prepared by van den Brink and Cotlear on May 8, 1992.  8 Appendix to Chapter 3 importance to the poor because of the greater share of basic foods in their consumption. For example, maize represents 25 percent of the food share for the lowest quintile. The open market producer prices for maize, paddy, and wheat rose 97.6 percent, 174 percent, and 83 percent, respectively, between 1990/91 and 1992/93. If one looks at the evolution of retail prices in 28 urban locations, one finds that in July 1991 a debe of maize costs T Sh 365.9. In July 1992, the same debe was 172 percent more expensive at T Sh 996.6 per debe. The retail price for one kilogram of rice was T Sh 73 in July 1990 but cost twice as much two years later in 1992. However, the price of one kilogram of wheat flour only increased 40 percent in the same period. From the evidence of basic commodities, the NCPI may be underestimating the price increases that consumers face in daily life. Survey design issues. The 1991 survey and the 1993 survey gathered information on expenditures in slightly different ways. For instance, the collection of information was more thorough in 1993, with more questions used about expenditures on education and health, and more detailed questions asked about other expenditures. There is empirical evidence that average expenditures obtained from more detailed questioning will indeed be higher than estimates drawn from more general questions. However, recorded food expenditures were also much higher in 1993, and expenditure shares are comparable for the two years. It may be that 1991 was a very bad year. Another possibility is that the 'Lofficial" inflation rate does not reflect the true increase in the costs of basic products. Given all of these different factors, it will not be possible to infer how much of the increase in expenditures is due to differences in survey design. Seasonality issues should not be important since both surveys were undertaken at about the same time of the year. Also, the relatively small size of the 1991 survey sample should be kept in mind. The results from the 1991 survey are more likely to be influenced by "extreme" observations, than the 1993 survey. This may explain the results from the 1991 survey on the temporal evolution of inequality. SURVEY RESPONDENTS 15. A unique ID code was given to each household member and was used to verify which household member responded to each section of the survey. This was done both as a quality check and to enable users to examine whether replies differed systematically by type of respondent. Instructions on the preferred respondent were given at the start of each section to guide enumerators. Table 3.2 indicates the characteristics of the respondents for each section of the survey. 16. One concern was that males would dominate survey responses. However, for most sections of the survey, the percentage of respondents who are male is hardly different from the percentage of males in the mainland population. The percentage rises for the section for livestock, and declines for the sections on prenatal and delivery services, and child spacing.  Appendix to Chapter 3 9 17. The percentage of respondents who are heads of households is higher than the percentage of respondents who are male, indicating the presence of female heads of households among survey respondents. For many sections, the head of household was the preferred respondent. The fact that in many cases the household head did not respond may indicate absence from the household for employment or other reasons, and the possibility of de facto female headed households. 18. Section 1E (prenatal care) of the questionnaire was intended to be addressed to each woman in the household and, if a woman was absent, the enumerator was requested to ask the "main female in the household" for information on absent women. However, over 27 percent of respondents were male, indicating either that no female was present to give such information or that males did not allow women in the household to respond directly. 19. Section 2C on child spacing may seem as though it should be addressed primarily to women. However, the intent of this section was to solicit views on willingness to pay or be paid for child spacing from both men and women. The list of preferred respondents was: the head of household, the spouse of the household head, a man from 15 to 25, and women from 15 to 25. The actual percentage of male respondents, at 39.7 percent is, thus, within the desired range.  10 Appendix to Chapter 3 Table 3.2: Characteristics of Respondents Description of Section Gender % Average Schooling (% Head Age (% Male) Completed Primary)* Household Roster 51.9 62.0 38.5 55.6 IB: Parents of children 7 - 15 51.8 61.0 40.4 50.1 IC: Schooling: children and young adults 7 50.8 59.6 38.1 55.6 - 25 and everyone in school ID: Acute illness in past 4 weeks 51.0 60.7 38.6 57.1 IE: Pre-natal and delivery services last 27.6 37.1 34.5 58.7 year and family planning 3A: Land 57.9 66.4 40.5 47.4 3A: Livestock 61.9 67.4 41.1 47.1 3B: Economic Activities 53.0 63.2 38.5 55.6 3C/D/E: Household Expenditures 53.6 64.0 38.7 55.5 3F: Housing Characteristics 51.8 62.0 38.4 55.8 3F: Water, Fuel, Roads, Markets 51.8 61.9 38.4 55.7 3G: Mortality 55.5 64.1 41.2 42.4 2A: Desired features of primary health 51.9 61.2 38.1 56.3 facilities 2A: Willingness to pay for desired features 52.3 61.9 38.3 56.3 in primary health facilities 2A: Actual features of local primary health 52.1 61.8 38.3 56.2 facility 2B: Desired features of primary school 52.1 61.5 38.2 56.6 2B: Willingness to pay for desired features 52.4 62.0 38.3 56.1 of primary school 2B: Desired features of primary school 52.4 61.9 38.2 56.4 curriculum 2B: Actual features of local primary school 52.3 62 38.2 56.3 2D: Perceived required income level 52 62.4 38.4 55.7 2C: Willingness to pay or be paid for child 39.7 39.6 34.5 57.8 spacing * Includes those who achieved primary education and higher levels of education.  Appendix to Chapter 3 11 APPENDIX 3.4 RURAL URBAN PRICE DIFFERENTIALS Table 3.3: Prices by Location Commodity Rural Urban Differential (Outside Urban vs. Dar es Rural Salaam) Maize Grain (T Sh /kg.) 55.7 64.8 16.3 Maize flour (T Sh /kg.) 104.9 110.9 5.7 Sorghum (T Sh /kg.) 114.7 76.6 -33.2 Cassava (T Shb/kg.) 70.2 76.7 9.3 Millet (T Shb/kg.) 78.8 90.0 14.2 Fresh Fish (T Sh /kg.) 268.7 427.9 59.3 Dried Fish (T Shb/kg.) 440.0 596.3 35.5 Sugar (T Sh /kg.) 294.5 270.7 -8.1 Beans (TSh/kg.) 153.8 179.6 16.8 Cattle meat (T Sh /kg.) 351.6 426.1 21.2 Paraffin/kerosene (T Sh/liter) 231.9 178.9 -22.9 Cooking oil (T Sh/liter) 526.5 472.2 -10.3 Bicycles (each) 41,618 38,700 -7.1 Source: HRDS 1993/94 Community Price questionnaire. 20. The 1993 HRD Survey gathered information in each community on prices that were observable for a set of goods. Using this information, average prices were computed to study whether the cost of living is higher in urban areas or in rural areas. Some goods are more expensive in rural areas than in urban areas, but, with the exception of fish, the differences in food prices never exceeded 15 percent. This result seems in line with van Ginneken's estimate of a Fisher Food Price Index (with rural areas as the base)." He found that, for 1970, food prices in urban areas were just 9 percent higher than in rural areas. However, if one looks at imported or processed food and manufactured products-e.g., sugar, cooking oil, and kerosene-prices in urban areas appear to be lower than in rural areas. Lugalla writes "prices of various commodities are higher in the countryside than in the urban areas due to poor infrastructure and other problems".12 Evidence from an ILO study indicates that consumer goods were about 50 percent more expensive in rural than in urban areas. Given this, it is presumed that the W. van Ginneken, Rural and Urban Inequalities in Indonesia, Mexico, Pakistan, Tanzania, and Tunisia, (Geneva: International Labor Organization), 1976. 12 J. Lugalla, "Poverty and Adjustment in Tanzania: Grappling with Poverty Issues in the Adjustment Process", mimeo, University of Dar es Salaam, 1993.  12 Appendix to Chapter 3 composite cost of living in rural areas is not likely to be much higher than in urban areas. In other words, the price differences are likely to cancel out. APPENDIX 3.5 A TOBIT MODEL OF RURAL POVERTY 21. A model of rural poverty was developed to generate simulations of the effects of key policy variables on poverty. The model explains how well households do in terms of a "welfare ranking" based on the expenditure per adult equivalent of each household. Different household characteristics are then used to explain where a household lies on the ranking. The household characteristics fall into four categories: the demographic make- up of the household; the assets which the household has at its command; the extent to which the household is involved in market activity; and the agro-ecological characteristics of the area. 22. The magnitude of the effect of each factor, and whether or not it is significant, was estimated by regressing selected variables for household characteristics on expenditure per adult equivalent. A Tobit procedure was used to control for the truncation of expenditure values at zero. Table 3.4 lists the household characteristics and shows the population average for each and the effect on poverty. For instance, the first row, with the variable "Female Head", indicates that female-headed households make up 12.2 percent of the population. However, whether or not a household is headed by a woman is not found to have a significant effect on its expenditure per adult equivalent. By contrast, the third row indicates that the average family size for Tanzania is 5.91 people, and that increasing the household's size is found to depress expenditure per adult equivalent.  Appendix to Chapter 3 13 Table 3.4: Determinants of Rural Poverty-A Summary of the Results Variable Average for Effect on the Rural Welfare Population Demographic Female Head: I if the head of household. is a 12.2% 0 Characteristics of woman. the Household Dependency Ratio: Number of dependents divided 1.15 0 by number of adults. Family Size. 5.91 Women's Education: Average years of schooling 3.73 + for each woman in the household over 20 years of age. Men's Education: Average years of schooling for 4.94 + each man in the household over 20 years of age. Land Ownership/ Land: Acres of land per family member 16 years or 2.04 + Agricultural older. Production Cashcrop: 1 if household grows a traditional export 36.9% + crop; 0 otherwise. Sources of Labor Wage Earnings: I if any member of the household 16.2% 0 Income receives wages from a public organization or private business; otherwise 0. Farmwage: I if any member of the household 10.9% 0 received a farmwage in the last 12 months; otherwise 0. Self-Employment: I if any member of the 8.1% 0 household describes themselves as primarily self- employed; otherwise 0. Market Distance to local crop market: the average distance 6.39 km Integration cited by households in that cluster who traveled to the market to sell. Road Quality Index: Combines four aspects of road 23.75 + quality - the maximum value is 39, the minimum is 0. Interaction term for distance and education. 10.2 + Note: 0 stands for no effect; + for positive effects, and -for negative impact.  14 Appendix to Chapter 3 APPENDIX 3.6: ADDITIONAL TABLES EXPENDITURE LEVELS AND COMPOSITION Table 3.5: Expenditure Levels (T Sh per year) Tanzania Rural Urban Dar es Salaam Poorest Better- All Poorest All off Per -Adult Equivalent 61,145 381,674 183,162 60,959 146,297 230,667 416,387 Percapita 38,010 267,355 123,351 37,709 95,328 161,773 290,631 Averages computed at the household level. Table 3.6: Distribution of People by Adult equivalent and Per Capita Annual Expenditures (averages computed at the household level)'3 Annual Expenditure per Adult Equivalent Annual Expenditure Per Capita (mean in T Sh per year) (mean in T Sh per year) Quintile All Rural Urban Dar es All Rural Urban Dar es Tanzania Salaam Tanzania Salaam Poorest 1 61,145 60,959 62,666 N/A 38,011 37,709 40,479 N/A (I obs) 2 95,367 94,975 97,081 101,230 60,963 60,182 64,535 66,520 3 129,270 129,053 129,574 133,895 83,012 81,745 86,452 87,621 4 180,099 178,945 181,484 182,273 118,583 117,002 121,145 124,380 Richest 5 381,674 319,842 398,062 500,299 267,355 216,258 287,607 351,266 Richest/Poorest 6.24 5.2 6.3 4.94 7.03 5.73 7.10 5.28a All (Household 183,162 146,297 230,667 416,387 123,352 95,328 161,773 290,631 level) All in US) 370 296 466 841 249 193 327 587 a Estimated as the ratio of top quintile to second quintile, since no observationsfor the bottom quintile were available in Dar es Salaam. 'Exchange rate: TSh 495 per dollar. 3 It should be noted that these values were computed at the household level (i.e., annual expenditure per adult equivalent, or per capita expenditures at the household level is averaged over all households.) Therefore, they are not directly comparable with GNP per capita, which is computed at the national level (i.e., total expenditure per annum is averaged over all individuals). One is the ratio of the averages, and the other one is the average of the ratios. Unless the function mapping households to national population is linear-which in our case would require that all families to be of equal size- these values will be different.  Appendix to Chapter 3 15 Table 3.7: Expenditures Patterns (Cash and Kind) by Household Type (T Sh per year) Tanzania Rural Urban Dar es Salaam Poorest Better-off All Poorest All All All Food 189,202 703,838 415,811 75477 379,596 454730 678,503 Health 3,063 25,534 11,300 8,279 8,607 15,338 25,900 Education 3,495 15,406 8,269 3,506 5,846 13,543 14,277 Housing 13,982 66,570 33,190 14,085 24,784 49,783 61,400 Water 98. 3,184 1,068 15.78 64.98 1,919 9,333 Clothing 9,452 46,466 23,802 9,593 19797 28,954 49,180 Other Total 255,847 1,139152 614,150 258,820 523,393 732,635 1,181,651 Household Expenditures Table 3.8: Composition of Food Expenditures (T Sh per week) Tanzania Rural First Highest All Rural Rural Urban Dar Quintile Quintile Poor Food expenditures 3,304 12,573 7,379 3,361.4 6,742 8,026 12,171 of which: Maize 24 10 15 24 17 13 8 Roots and Tubers 12 7 9 12 10 7 5 Pulse and Seeds 16 11 13 16 14 11 11 Fruits and 11 16 14 11 14 13 15 vegetables Meat and Dairy 15 25 22 16 20 24 24  16 Appendix to Chapter 3 Table 3.9: Composition of Food Expenditures (Shares)-Percent Of Expenditures in Each Item on Total Food Expenditures Quintile Quintile All Poor Rural Urban DSM 1 5 Rural Maize 0.24 0.10 0.15 0.24 0.17 0.13 0.08 Rice 0.05 0.09 0.09 0.05 0.07 0.11 0.10 Other Grains 0.07 0.03 0.04 0.07 0.05 0.03 0.02 Roots and Tubers 0.12 0.07 0.09 0.12 0.10 0.07 0.05 PulseandSeeds 0.16 0.11 0.13 0.16 0.14 0.11 0.11 Fruits and Vegetables 0.11 0.16 0.14 0.11 0.14 0.13 0.15 Meat and Dairy 0.15 0.25 0.22 0.16 0.20 0.24 0.24 Beer 0.03 0.07 0.05 0.03 0.04 0.05 0.09 OtherFood 0.07 0.11 0.10 0.07 0.08 0.13 0.15 Total Food 1.00 1.00 1.00 1.00 1.00 1.00 1.00 Table 3.10: Composition of Food Expenditures (T Sh per week)* Expendi- Tanzania Rural Urban Dar es ture, Per Salaam Item Poorest Better- All Poorest All All All off Maize Cash 179.9 690.0 432.8 153.8 333.7 576.2 993.5 Kind 618.1 615.8 662.3 648.9 795.2 440.6 39.3 Rice Cash 82.9 790.8 386.0 74.9 242.7 605.8 1142.6 Kind 90.6 336.5 245.8 90.9 246.8 283.7 70.0 Other Cash 41.9 286.8 126.5 44.3 107.1 158.8 217.4 Grains Kind 178.6 127.9 170.3 193.1 230.0 44.9 1.6 Rootsand Cash 99.8 487.3 263.1 95.4 180.9 413.2 593.1 Tubers Kind 281.3 422.6 395.5 292.4 507.9 170.2 31.6 Pulses and Cash 205.0 915.7 523.0 195.6 399.0 682.3 1312.2 Seeds Kind 333.2 514.9 443.5 350.3 565.2 203.4 32.9 Fruitsand Cash 189.8 1278.2 615.8 185.6 412.0 913.3 1757.1 Vegetables Kind 159.4 676.9 405.6 170.3 534.5 132.7 51.7 Meat and Cash 354.6 2531.9 1246.0 357.0 938.8 1700.4 2940.0 Dairy Kind 153.9 643.2 368.7 169.8 435.8 260.2 39.7 Beer Cash 89.9 828.0 349.2 98.8 276.2 376.7 1102.04 Kind 6.4 44.3 22.6 6.8 22.1 26.8 9.4 Other Food Cash 217.6 1320.9 687.4 215.2 477.3 1003.0 1825.0 Kind 20.4 61.2 34.6 18.2 37 33.5 11.8 Total Food Cash 1461.4 9129.5 4629.9 1420.5 3367.5 6429.6 11,883.0 Kind 1842.1 3443.2 2748.9 1940.9 3374.3 1596.1 288.0 Total Food 3,304.0 12,573.0 7,379.0 3,361.4 6,742.0 8,026.0 12,171.0 * Averages were computed at the household level.  Appendix to Chapter 3 17 Table 3.11: Percentage of Total Income Accruing to Each Quintile All Rural Urban Dar es Tanzania Salaam Poorest 1 6.8 7.7 7.3 7.4 2 11.0 12.0 11.4 11.0 3 15.1 16.3 15.6 14.9 4 21.6 22.4 21.2 21.4 Richest 5 45.4 41.6 44.5 45.4 All 100.0 100.0 100.0 100.0 APPENDIX 3.7 INEQUALITY AND POVERTY Table 3.12: Headcount Measure by Socio-Economic Group (percent of population) Higher Poverty Lower Poverty Line Line Rural Farmer 23.1 6.4 self-employed 12.5 .9 government employee 12.3 1.5 Urban Private employee 3.4 0.02 Business person 0 0 Government or parastatal 1.2 0.1 employee Other (rural and urban) 12.9 4.1  18 Appendix to Chapter 3 Table 3.13: Poverty Measures by Location and Poverty Line (percent of population) Location Poverty Line Head Count Poverty Gap FGT Mainland Tanzania Line 1 4.4 0.8 0.3 Line 2 16.9 3.8 1.62 Line 3 22.3 5.5 1.62 Rural Tanzania Line 1 5.7 1.0 0.4 Line 2 21.2 4.7 2.07 Line 3 27.8 6.9 2.5 Urban Tanzania Line 1 1.4 0.1 0.05 Line 2 6.9 1.45 0.56 Line 3 9.9 2.22 0.7 Dares Salaam Line 1 0 0 0 Line 2 0.21 0.04 0.01 Line 3 0.21 0.06 0.01 Line I = 49,600 T Sh per year and per adult equivalent. Line 2 = 83,111 TSh per year and per adult equivalent. Line 3 = 73,877 TShperyear and per adult equivalent. Table 3.14: Evolution of Poverty: 1983 and 1991 (T Sh per year) 1983 1991 Head Count (P.) 64.6 50.5 Depth (P,) 35.8 34.2 Average Shortfall Income (at 5,389.0 5,143.0 current prices) Total Poverty Gap in 1991 T 55.2 63.2 Sh (billions of T Sh) Total Poverty Gap in billions of 1.21 .314 US dollars Head Count = Percent of the population falling below the poverty line. Depth = Percent of poverty line income required to bring everyone below it up to the poverty line. *Average shortfall income is the poverty line minus the average income of those below the poverty line.  Appendix to Chapter 3 19 Table 3.15: Elasticities of Poverty Measures for Mean Income and Gini Index and Marginal Proportionate Rate of Substitution (MPPRS): Tanzania, 1994 Poverty Value of Elasticity Elasticity MPRS Line Poverty for mean for Gini (T Sh) Measure income Index (%) 128,109 50.0 -1.20 .346 3.468 73,877 16.85 -2.26 2.79 1.23 49600 4.4 -3.1 7.16 2.3 * Poverty line that would yield an incidence of poverty of 50 percent. Table 3.16: Elasticities Measuring Effects of Economic Growth and Changes in Inequality Within Location on Total Poverty in Tanzania, 1993 Poverty Rural Villages Other Towns Dar es Salaam line Incidence 0 'o Incidence 710 Eo Incidence le so of of of Poverty Poverty Poverty 73,877 21.16 -2.1 1.83 5.7 -.25 .44 .03 0 0 83,111 27.8 -1.78 1.18 9.5 -.3 .45 0.2 0 0.01 92,346 34.2 -1.57 .78 13.3 -.31 .37 .9 -.01 0.02 128,609 58.1 -.95 .07 34.1 -.27 0.16 5.0 -0.02 0.04 Table 3.17: Poverty and Neutral Growth Year Poverty Per capita 1993 2000 2005 2010 2015 Line Growth rate (%) (percent of population below the poverty line) 0.0 50.0 50.0 50.0 50.0 50.0 T Sh 128,109 1.5 50.0 44.02 35.39 25.98 17.41 3.0 50.0 38.67 24.89 13.33 5.94 5.0 50.0 32.4 15.4 5.37 1.37 0.0 16.9 16.9 16.9 16.9 16.9 T Sh 73,877 1.5 16.9 13.28 8.78 4.88 2.29 3.0 16.9 10.34 4.45 1.35 .29 5.0 16.9 7.30 1.73 .35 .02  20 Appendix to Chapter 3 Table 3.18: Poverty and Inequality Year Percentile 1993 2000 2005 2010 2015 Change in the Gini Coefficient (percent of population below the poverty line) -I 16.9 13.87 9.88 6.11 3.28 -0.5 16.9 15.32 12.95 10.20 7.49 -0.1 16.9 16.57 16.03 15.28 14.37 0 16.9 16.9 16.9 16.9 16.9 0.1 16.9 17.23 17.82 18.68 19.86 0.5 16.9 18.62 21.98 27.81 37.71 1 16.9 20.49 28.49 45.46 83.21 These estimates are based on a poverty line of TSh 73,877per year. APPENDIx 3.8 BASIC NEEDS AND INDICATORS OF HUMAN DEVELOPMENT Table 3.19: Social Indicators for Tanzania and Sub-Saharan Africa 1980 1985 1990/91 Tanzania Sub-Saharan Tanzania Sub-Saharan Tanzania Sub-Sabaran Africa Africa Africa Per Capita Income (1991 US$) 284 582 309 491 110 340 Life Expectancy at Birth (years) 47 47 48 49 48 51 Infant Mortality (per 1,000) 122 127 117 118 115 107 Average Daily Caloric Intake (Kilocalories per capita) 2244 2107 2229 2040 2206 2120 Primary School Gross Enrollment Rate (%) 93 70 75 75 69 70 Source: African Development Indicators.  Appendix to Chapter 3 21 Table 3.20: Evolution in Access to Services (percentage of households) Facility 1976/77 1993 Rural Urban Rural Urban Water Source less than 0.5 km 29 66 45 71 Primary School less than 0.5 km 29 31 36 43 Health Facility less than 5 km 18 57 61 87 Food Market less than 5 km 68 88 71 98 Table 3.21: Type of Toilet Facility (percentage of households) Source All Income Location Tanzania Bottom Top 20% Rural Urban 20% Flush Toilet 2.5 0.1 5.9 1.1 4.9 Pit Latrine 92.2 89.8 91.5 92.3 92.4 None 5.3 10.1 2.4 6.6 2.6 Table 3.22: Basic Housing Amenities (percentage of households) Tanzania Rural Poorest Better- All Poor All Urban Dar es off Salaam House with non-earth floor 5.5 54.9 27.1 4.3 11.2 55.9 91.6 House with iron roof 20.1 71.9 45.0 17.5 29.9 75.0 94.2 House with no windows 25.5 7.4 14.8 27.8 19.3 5.5 .8 House with glass windows 1.1 3.4 2.3 1.2 2.0 3.2 1.8 Electricity Supply ( households 1.7 30.8 11.2 .5 .8 30.3 51.8 with) Kerosene as the main source of 94.4 70.5 88.0 94.6 97.1 72.1 48.8 lighting fuel Wood as main source of cooking 96.8 53.2 79.6 99.0 96.6 50.0 5.0 fuel Households without toilet 10.2 2.4 5.3 10.9 6.6 2.6 1.3 facilities  22 Appendix to Chapter 3 APPENDIX 3.9 DEMOGRAPHIC CHARACTERISTICS Table 3.23: Demographic Characteristics of the Tanzania Household, 1993 All Income Group Location Tanzania Bottom Top Rural Urban DSM 20% 20% Household Size (mean) 5.91 7.05 4.89 6.16 5.48 4.84 Dependency Ratio (mean) 1.15 1.31 0.94 1.23 0.99 0.88 Average Age of Household Head 43.7 45.8 41.4 43.8 44.6 39.2 Female-headed Households % 12.2 10.9 15.3 10.0 17.6 14.8 Average Number of Children 2.8 3.5 2.1 3.0 2.4 2.1 less than 15 Average Number of Children 3.2 4.1 2.4 3.4 2.8 2.4 less than 18 Average Number of Adults 0.15 0.2 0.14 0.16 0.15 0.05 Older than 65 Notes: Dependents are defined as people younger than 15 and older than 64. The dependency ratio is the number of dependents over the number of others. Table 3.24: Percentage Distribution of Households by Family Size Distribution All Income Group Location of Families Tanzania by Size Bottom 20% Top 20% Rural Urban DSM 1-5 people 49.9 33.9 66.6 46.4 55.4 68.3 6-10 people 45.3 57.1 30.8 47.8 42.0 29.7 11+ people 4.8 9.0 2.7 5.8 2.6 2.0 TOTAL 100.0 100.0 100.0 100.0 100.0 100.0  Appendix to Chapter 3 23 Table 3.25: Percentage of Distribution of Households by Family Size Family All Income Group Location Type Tanzania Bottom Top Rural Urban DSM 20% 20% One adult, no children 2.2 0.3 6.4 1.6 3.2 6.2 One adult, 1-2 children 1.4 0.4 2.2 1.2 1.8 3.1 One adult, 3-4 children 1.0 0.7 1.0 0.9 1.1 1.3 One adult, 5+ children 0.4 0.7 0.2 0.5 0.4 0 Two adults, no children 5.4 1.2 10.8 4.6 6.7 8.7 Two adults, 1-2 children 15.4 7.5 20.5 14.6 15.6 23.8 Two adults, 3-4 children 14.5. 10.2 14.5 15.5 11.1 16.9 Two adults, 5+ children 4.9 6.8 2.9 6.0 2.7 1.9 Three plus adults, no children 5.7 4.1 5.8 4.7 8.8 5.3 Three plus adults, 1-2 17.2 22.0 16.1 17.3 17.5 14.5 children Three plus adults, 3-4 19.3 25.9 13.6 18.6 22.6 4.2 children Three plus adults, 5+ children 12.6 20.4 6.1 14.7 8.7 2.0 TOTAL 100.0 100.0 100.0 100.0 100.0 100.0 Notes: An adult is defined as any family member aged 15 or over. Table 3.26: Households and Marital Types (percentage of households) Marital Basis of All Income Location Household Tanzania Bottom Top Rural Urban DSM 20% 20% Polygynous 9.2 15.4 6.7 11.2 4.8 1.2 Monogamnous 90.8 84.6 93.3 88.8 95.2 98.8 Note: A polygynous household was definedfrom the survey document as one with more than one wife reportedfor the head ofhousehold  24 Appendix to Chapter 3 APPENDIX 3.10 LAND AND DISTANCE TO ROADS Table 3.27: Market Integration Distance to Road Percentage of Households Average Expenditure (T Sh a year) 0 - 500 meters 56.85 149,800 500 - 1000 meters 9.35 147,853 I km - 2.5 km 10.15 151,895 2.5 km - 5 km 6.44 148,275 5 km - 10 km 9.99 130,585 10 km plus 7.21 127,298 Table 3.28: Percentage of Land, by Area, from Different Sources All Income Location Land Tenure Tanzania Bottom Top Rural Rural Urban Dar 20% 20% Poor Inheritance 0.3 0.2 0.5 0.2 0.3 0.4 0.6 Public Authority 21.3 30.3 17.1 30.5 23.4 13.6 7.3 Clearing 19.8 26.9 10.8 28.2 22.0 11.7 11.9 Purchased 45.5 33.9 55.0 34.2 44.1 50.0 69.4 Other Sources 5.6 3.6 6.8 2.6 4.5 10.0 4.6 Reason not Given 7.4 5.0 9.9 4.5 5.7 14.3 6.0 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Table 3.29: Land - A Comparison over Time 1983 1991 1993 Owners (%) 99.7 96.4 98.0 Mean Landholding (in acres) 8.10 11.51 5.9 Median Landholding (in acres) 5.9 7.41 4.0 Gini Coefficient 0.45 0.5 0.46  Appendix to Chapter 3 25 Table 3.30: Identity of Land Owner (percentage of households) All Income Location Owner Tanzania Bottom Top Rural Rural Urban Dar 20% 20% Poor Specific Person 36.5 32.0 44.8 33.6 35.8 38.0 67.2 All Members of 55.9 63.1 45.3 62.1 58.3 47.5 26.8 Household Other Person 7.2 4.8 9.3 4.2 5.4 14.2 6.0 Joint Ownership 0.1 0.1 0.2 0.1 0.2 0.0 0.0 with other HH No Answer 0.3 0.0 0.4 0.0 0.3 0.3 0.0 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Figure 3.1: Frequency with which Household Head is Land Owner (percentage of households) Individual Ownership 70 60 50 t 40 30 ~~~~~~~~~~~~~~~IniitiOwner 20 Ir).j~ k.