Zambia - Poverty assessment (Vol. 2 of 5) : Appendices

Report No. 12985-ZA Zambia Poverty Assessment (In Two Volumes) Volume II: Appendices November 10, 1994 Human Resources Division Southern Africa Department Africa Regional Office Document of the World Bank  REVIEW OF POVERTY STUDIES APPENDIX - 1  1-2 I. REVIEW OF ZAMBZ POVERTY STlUDES 1.1 Ihe studies pretd in this summary provide a glimpse of the analytical research that has been taking place in Zambdi. hese studies include: Jensen and Lucket (1993); Central Sttcal Office Priority Survey I (1993); Chipwende, et al. (1993); World Bank (1993a, 1993b); Caldwell; Siandwazi (1993); and Cogill and Zaza (1991). Ibis section reviews some of these studies with the nte of eamining methodological approaches, common conclusions, and areas of disagreement In these studies. Results fiom the various studies will also be compared with our own analysis of the Social Dimensions of Adjustment Priority Survey (PSI) data tha follows. A. DESCRIPT1ON OF POVERTY STUDIES 1.2 In the reviewed reports, poverty is analyzed first by developing 3bsolute or rlative poverty lines or by describing the prevalence of undernutrion or mahnithtion. Once the poor are identified, each study proceeds to examine the relationship between demographic, income or expendiure, and n=ition variables at the household, district, or province levels and the poverty status. 1.3 The Jensen and Luckett report used the Household Expenditure and Incomes Survey (HEES) which was conducted in June 1991 by the Prices and Incomes Commission of the GRZ. The study used household expendiures to charactei poverty. Two relative poverty lines were used: The mean Zambian income, and one-half the mean income. Research has shown that in poor counties the men income is a good approximation of the subsistence poverty level (Jensen and Luckett). The authors described the geography of poverty, and created a profile of poor households. 1.4 The Central Statitics Office (CSO) report used the Social Dimensions of Adjustment (SDA) Priority Survey which was conducted in October and November of 1991 by the CSO. The report used official absolute poverty lines and employed the income secton of the PS to measure the prevalence and the depth and severity of poverty in rural areas. Poverty was broken down by urban and rural areas, by province, by socio-economic group (SEG), and by household demographics. The CSO report also used apometric data from the PSI to describe the geographic and socia paen of malurton. 1.5 Chipwende, et al. used both the HEIS and the PSI to create a geographic and social profte of Zambian povert. They used 2/3 of the mean expenditures from the HEIS as a household poverty cutoff. They then examined the prevalence, depth and seerity of poverty by chaactsics of the household and its head. They created two cutoffs with the income portion of the PSI: 2/3 of mean household income for the moderately poor cutoff, and 1/3 of the mean for extremely poor households. The authors analyzed the relationship between poverty and ditance and access to filities, among other things. 1.6 Siandwazi used secondary data on the prevalence of malnuition gathered from a number of sources, hospial and clinic admissions data, and several household consumption studies to pa a broad picture of regional and seaonal variations in poverty and malnurion in Zambia.  1-3 1.7 Caldwell used a number of data sources, including percent underourished by geographic area, data on access to medical sources and other infrastructure, historical crop production and curent food aid data to create a geographical profile of vulnerability in rural Zambia. With geographical information systems (GIS) technology, he overlayed and combined a number of these indicators to create scores for chronic and currnt vulnerability. 1.8 Cogil and Zaza reported on results of the Nutrition Status Module, a part of the Zambia National Household Survey Capability Programme, Implemented by CSO in December 1989 and January 1990. The Module represented a random sample of 2,133 children under 5 years old from 1,200 rural households from all provinces. The module, which coltected anthropometry data and clinic attendance, was linked to the Crop Forecasting Survey, and the combined data were used to examine the geographic distribution of malnutrition, along with the relationship between socioeconomic characteristics and nutritional status. B. GENERAL RE:SULTS 1.9 The results from the three income- or expenditure-based poverty studies are summaized in Table L IThere is a reasonable correspondence of results across the studies, even comparing the HEIS-based studies with the PS-based ones. Estimates of the prevalence of moderate poverty range from a high of 77 percent to a low of SE percent. The studies clearly show that rural areas suffer from a higher prevalence of poverty than do urban areas. Rural poverty is a widely-docmented phenomenon in Zambia. There is a bias, however, in favor of overestimating rural relative to urban poverty for all of these studies for two reasons. First, none of the studies cited using income or expenditures adjusted for cost of living differentials, and these differences are likely to be high, despite pan-territorial pricing of some key food commodities. Second, income and expenditures are more accurately measured in urban areas; systematic undermeasurement of these variables in rural areas is likely. 1.10 Despite these biases, the studies universally conclude that the prevalence, depth, and severity of poverty is greater in rural Zambia. Although absolute comparisons between rura and urban areas may not be possible because of the points presented above, it is reasonable to conclude that rural are relatively worse off in Zambia. C. THE GEOGRAPHICAL DISTRIBUlION OF POVERTY 1.11 Poverty in rural Zambia shows wide geographic variation. Rural poverty as measured by Jensen and Luckett varied from 89 percent in Eastern to 70 perc In Cent province. he CSO report showed slightly different orderings by province but the magnitudes of the estimates are sinil. Moderate rural poverty in the CSO report ranged from a high of 93 percent in Northern to a low of 63 percent in Lusaka province. 1.12 According to Caldwell measures, Luapula, Western, and Northwestern provhines arethe most subject to chroic food isecurty, while Lusaka, Copperbelt, and Central provinces are relatively most food secure. A comparison of thA ceral production per caia data wih the otier idca show tdat the former is a poor predictor of either poverty or mantrtion. Lapula, Northen, and  1-4 Eastern provinces have the highest prevalence of malnutrition, according to the Cogill and Zaza study, while in the CSO study Northem province is unambiguously the worst. 1.13 Comparison across studies allows us to mabk some generalizadons. Luapula suffers from the worst mainutrition and chronic food insecurity, but according to the CSO report, poverty is slgnificantiy worse in other provinces. Northern province is always found to suffer from high prevalence of malnutrition and poverty. Eastern, Northwesten, and Western provinces also show relatively high rates of poverty, food insecurity, and malnutiton. Central, Southern, Copperbelt, and Lusaka rural are all relatively well off, although in some cases they score relatively poorly. Table I Urban/Rural Comparisons of Poverty by Source Perent failing into class Measre Rurl Urban All Jensen and Luckett Moderate Povet1y .79 .43 .69 core Poverty .42 .08 .32 CSO Modert Poverty .85 .55 .71 core Poverty .78 .44 .61 Cbipwende et al. Moderate Poverty' .88 .26 .68 Modere Povert9 .67 .29 .56 Moderate Povery .86 .69 .77 core Poverty4 .74 .44 .57 'Usbs firn S utIuuAAbuk '&OA&i4d War ih 7 '2W- 'Udi SB mm lam has = .zpmIbm a COW. U9g IS, bw"hd= a deftap homa.. 'Wqag S bawds km d&M ldswm WOWi" 1.14 Some of the differences across studies may be attributable to the charistics and actual viations of the dfferent indicators; they may also be explained by systematic biases in measurement and survey coverage. Ihe HI3S, for example, had a small ural sample size outside of Eastern provice relative to the PSI. Income measures employed in the PS-based studies have flaws whose impact varies from province to province. The one firm generalization is that there are sigificant varutions in poverty and malnutrition, and virtually all rural areas suffer from significant poverty. 1.15 Even though certain provinces have lower percentages of poor than others, they may have a higher number of poor due to large population size. This is the case in the Copperbelt where the absolute number of poor is exceeded only by that from Eastern province (World Bank 1993a). Copperbelt and Eastern provinces are heavily populated, and Coppeatlt is highly urbanized. Poverty in Copperbelt is primarily an urban phenomenon. D. SOCIO-ECONOMIC CHARACTERISTCS AND POVERTY 1.16 To vaying degrees the different studies examined the relationship between the following socio- economic characteristics and poverty. Hoesehold SYie 1.17 Household size was shown by Jensen and Luckett to be larger for poorer households in both  1-5 rural and urban areas. For rural areas, household size for the lowest expenditure decie was 6.42 members, while for the top decile it was 2.92 members. Size of household increases significandy along with the depth of household poverty. The Jensen and Luckett study found that when using the one- half mean income poverty level, 42 percent of rural households are poor while for large rural families (those with more than 4 members) the prevalence of poverty increases to 50 percem. 1.18 terestingly, both the CSO and the Cogil1 and Zaza studies found a negative relatonship between household size and the prevalence of either stuting or undernutrition of children. There is an inconsistecoy between the findings using income/expenditues and those using nutitional outcomes (anthropometry). There are a number of plausible explanations for this phenomenon, such as increased attention to children by older siblings (nutitional status should be closely linked to birth order in such a case), rems to scale in feeding and other nutrition inputs, etc. No study examined this discrepancy in a systematic fashion. More research on this issue is obviously needed. Headship 1.19 Households headed by females and by elderly of any sex were generally found by the studies to be more iikely to be poor than households headed by younger men. Jensen and Luktt found tlud households headed by married women are as likely 'o be poor as the general populaion. However, other groups of households headed by women were significantly above the mean poverty level. Households headed by widowed and divorced women are more likely to be severely poor, and the contribution to totl rur poverty of these households is substantial. 1.20 The CSO study found that female-headed households are more likely to fall below the moderate and severe poverty lines than most male-headed households. The World Bank (1993b), also citing the Priority Survey, reported that 88 percent of female-headed rural households are modely poor as compared to 81 percent for male-headed households. The figures are more starik for the exremely poor. Nationally, 70 pcent of female-headed households are very poor (contributing to 23 percent of total poverty), while only 55 percent of male-headed households are extremely poor (76 perceat of total poverty). 1.21 Cogill and Zaza reported that children in households headed by females are more likely to be stunted than those from male-headed households, although rates of underuwtrition are equal across headship. One of the authors' explanations for this finding was the smaller agicultural holdings of female-headed households. Once land holding size is controlled for, the relaionship between headship and nutritional status disapears. The CSO study also found a strong negative relationship betwevi female headship and nutritional status. Siandwazi reported that females who head households are likely to be much less educated than males, although no breakdown of poverty by household headship was provided in the study. Bducton 1.22 Jensen and Luckett found a relationship between education of the household head and household income. Secondary education significanly reduces poverty in both rural and urban areas. The effeot is higher in urban areas and in rural areas primary education is not associated with a aignificant decline in poverty. Using the one-half mean income poverty level in rural areas, 23 percent of those with secondary educations are poor and 6 percent of those with higher educations are poor. These numbers are significantly below the average rural poverty oe 42 percent. 1.23 The CSO report did not examine relationships between education of any member of the household and poverty, but did break down nutritional status by mother's education (presumably based on the theory that increases in mother's education translate to better care for the children, or better  1-6 awareness of nutritious feeding practices). The study found that malnutrition (stunting and undernutrition, and to a lesser extent, wasting) declines dramatically with education of the mother. For all Zambian children whose mother has no formal education, the prevalence of suting is 44 percent This number declines to 41 and 30 percent as education of the mother increases to grades 1- 7 and 8-12, respectively.1' Undernutrition falls from 28 to 22 to 18 percent as the level of mother's education grows to more than secondary education. Employment 1.24 Jensen and Luckett also analyzed the relationship between type of employment and poverty. They found that the rural self employed -overwhelmingly farmers- have the highest poverty rates and contribute most to rural (and urban) poverty. Using the one-half mean income poverty level in rural areas, 46 percent of those self employed are poor. Such a level constitutes 85 percent of total rural poverty. Cogill and Zaza stated that off-farm income may be critical to some households, especially those headed by women. Unfortunately, the authors provided no hard data on this issue and only state that the income may be critical in creating household food security. Agneadl 1.25 There was surprisingly little effort in any report to link land-holding size with poverty. The 'Yorld Bank (1993b) found that poverty rates for small, medium, and large farms are, respectively, 77.9, 55.2, and 34.6 percent. The actual size of the farms was not reported in the World Bank sudy and land quality was not controlled for. Chipwende, et al., using the PS, reported average holdings for the extremely poor, the moderately poor, and the non-poor to be 2.62, 2.49, and 2.50, respectively. These data did not include rented land but the results stil must be considered suprising. If the numbers are to be believed, there is no association between landholding size and poverty, and maore analysis of this finding is clearly needed. 1.26 The CSO report showed that rural small-scale farmers are the socio-economic group most likely to be poor and exremely poor. The prevalence of moderate and extreme poverty fall for the larger-scale farmer groups. Rural non-agricultural households have the second-highest rate of moderate and extreme poverty behind the small-scale class. Rural non-agricultural households had a rate of 68 percent falling into the moderate or worse poverty group, with 59 percent being extremely poor. 1.27 Cogill and Zaza were unable to find a conclusive link between household landholding and malnutrition of children. Although height-for-age and weight-for-age increase significantly for girls whose families plant more than 10 hectares, there is no relationship between boys' nutrional stans and land in production. For households with fewer than 10 hectares in production, there is no relationship between girls' nutritional status and plaried land. Y Childre with mothes with 'A Level and collc educaton and bih have ades of *mnting of 39 percat, but th cd sizo is ctady sma ad t dffaec is not lky to be taitioaUy sinficuL  1-7 E. GENERAL OBSERVATIONS ON RURAL POVERTY AND FOOD INSECURITY 1.28 The studies produced some general observationss on the causes of nrur poverty and food insecurity. Some of these observations are summarized below. The Role of Maize and Agriculture 1.29 Zambians depend heavily on maize for their caloric requirements. Maize provides up to 70 percent of calories consumed (Siandwazi). Maize is pardcularly drought prone, and has historically had very unstable levels of production in Zambia. Hybrid maize is less tolerant of water stress than local varieties. Farmers typically plant hybrid maize only after local maize is planted since there are well-documented labor constraints in rural Zambia. Planting local maize first is a strategy to mi imize risk of suffering a drought loss. However, late planting increases hybrid maize yield risk, and lowers yields. 1.30 Several reports (World Bank 1993a; World Bank 1993b; Chipwende et a.; Siandwazi) stressed that Zambian agriculture is such that access to vital inputs, particularly traction power, is the main constraints to production. The World Bank (1993b) stated that a typical family can hand-hoe only 2 hectares while the output from 5 hectares is needed for food security. It was also widely reported that most small-holder do not have access to adequate traction power, creating reduced yields and poor timing of farming operations. 1.31 The inability to provide timely delivery of credit, seeds and fertilizer, and transport and payment for crops has been due, in large part, to policies that gave too much control !o parastatal organizations that were given unrealistic mandates. !!ubsidies for credit, seeds and fertilizer, and pan- territorial and pan-seasonal pricing for maize led to high budgetary outlays that bankrupted the government and led to an uneconomic spatW distribution of maize production. Furthermore, the focus of maize production by all support services (credit, input and output markets, research and extension) led to a virtual neglect of other crop and livestock activities that may have been warranted based on comparative advantage. 1.32 Siandwazi and Chipwende, et al. emphasized the importance of credit which is usually avaiable only to farmers who have title to their land. In rural Zambia only about 6 percent of the popuation has title to their land and among the extremely poor only one percent has title to land (World Bank, 1993b). The current process fur obtaining title to land is overwhelmed and backlogged. Gender and Poveily 1.33 Farms with female heads are more likely to be poor and have inadequate access to land. These households are less likely to have land title due to divorce and inheritance customs, and are frther discriminated against in credit markets (Siandwazi). They do not generally possess the physical strength necessary to hand-hoe an area sufficient for subsistence (World Bank, 1993b). Woman farmers typically achieve lower yields since they do not rotate crop land as frequently as male farmers because cultivating new land by hand is labor intensive and arduous work. 1.34 A fiurher problem for women and children in Zambia is their poor social stats and the consequent low priority their nutrition receives (World Bank, 1993a). Siandwazi stressed this low  1-8 social status in concert with the poor educational status of women as a factor in Zambia's high child malnutrition statistics. Poor feeding and weaning practices along with low status make malnutrition more prevalent for women and children. The conclusion that gender-based nutritional status differentals exist, however, is not universal. Both World Bank (1993a) and Cogill and Zaza fount that female children have better nutritional status than male children. Impad4s of Adjustnent 1.35 Structural adjustment that frees markets from government-induced distortions is likely to improve the terms of trade for farmers in the long run. However, in the short run, hardship can be expected in nural as well as urban areas from the adjustment program. The World Bank (1993b) stressed that there are several mechanisms by which this will occur. Ihe expected increase in rural terms of trade will not affect all farmers equally. All farmers will fice increased input costs as a result of adjustment. Real interest rates can be expected to rise making it harder for marginal farmers to get loans (World Bank, 1993b). At the same time, higher real rates of interest may increase access to formal sector credit by restoring viability to rural financial markets. 1.36 The World Bank (1993b) also noted that a coping strategy of the urban residents Is to send away family members and to reduce remittances to their home village. If large transfers are taking place then the rural poor will be hurt as urban incomes decline. 1.37 The current market liberalization policies will lessen many of the policy-induced biases discussed above, but a question remains as to the responsiveness of individual farmers to new economic realities. For example, the impact of market liberalization on outlying regions is unknown, but it is obvious that transport costs for previously subsidized outlying regions will increase. In many cases subsidies created rationing whereby support services were simply not available at the subsidized price. This phenomena was wide-spread in outlying areas of ruril Zambia. Ihis makes forecasts of the impacts of market liberalization so difficult. F. SUMMARY OF STUDIES 4ND RECOMMENDATIONS 1.38 The World Bank (1993b) asserted that Zambian poverty fits generic models of poverty quite well. They found low levels of human capital due to low levels of education and poor health care resulting in low incomes that perpetuate poverty. Poor access to real assets due to unfvorable land ownership laws, underdeveloped rural credit ins"tions and price discrimination against agricultural products have also resulted in low retrns on assets. The agricultural sector is further limited in its ability to respond to market signals by high transactions costs due to underdeveloped infrastructure and monopsonistic marketing channels. 1.39 Jensen and Luckett recommended direct programs to help Zambia's poor. Tey concluded that targeted assistance programs designed to improve living conditions will have a much greater impact in rural areas where poverty is most prevalent. They argued against direct food assistance since most of the poor produce their own food. They recommended programs that develop off-fam job opportunities in rural areas and that improve agricultural practices. They also recommended education above the primarily level as a priority.  1-9 1.40 Siandwazi stressed the value of govenment services and programs to Zambia's poor. She showed that nutrition education (particularly for women), health services, sanitation and clean drinking water supply decrease malnutrition; the relative effectiveness, however, of each measure in reducing poverty was not measured. She also noted that easing land title problems and land access for women would benefit Zambia's poor. Siandwazi noted features of successful programs to target assistance to Zambia's poor. She found that nutrition education programs work because they have a persona approach that has a powerful effect on participants. She stated that programs to enrich and fortify foods have been successful. Siandwazi recommended that the High Energy Protein Supplement program (HEPS, administered by the World Food Program) be extended to feed children in schools as well. 1.41 Siandwazi found food for work programs beneficial, particularly because women tend to be integrated into these programs. Like the World Bank (1993a), Siandwazi found that NGO- administered programs benefit from less bureaucracy and greater flexibility. She also recommended targeting of poor regions based on the seasonality of malnutrition. Chipwende, et al. also emphasized policies to protect Zambia's poor. They stressed the need to provide basic social services to the poor, including health care, education and safe drinking water. They advocated public employment schemes to alleviate unemployment. They stated that rural poverty should be treated with priority. 1.42 The authors stated that the rural poor should be provided title to land so they will have access to credit. Ihe report also stressed that rural credit access should, in general, be improved and that such improved access might be accomplished by a targeted credit scheme. In agricultural policy Chipwende et al. see the need to develop a free market system but still see a role for subsidizing crops that the poor consume. They recommended that farmers be provided subsidies to cultivate these "self- targeted foods. 1.43 Cogill and Zaza began their recommendExions by noting that increased food production in rural Zambia is not a sufficient condition for reduction of malnutrition. They the stressed that the market reforms might not raise production or incomes, especially for the most food-insecure groups. Poor sanitation, poor water quality, infection, and poor health services are problems that need to be addrewsed concurrendy with pricing, marketing, and macro-economic reforms. 1.44 The interventions that Cogill and Zaza recommended were: low cost technology for food production, storage, and processing (without being more specific); better availability of inputs; easier access to credit; minimization of marketing costs; and, improvements in infrastructure. Note that none of these recommendations for interventions res! ited directly from their analysis of malnutrition data. They also recommended a policy of expanding off-farm employment and income generadon in rural areas. Once again, they were not stecific about the form such a policy should take. Finally, improvements in health service delivery aad education were recommended. 1.45 The World Bank (1993a) strongly recommended that food security become -. major focus of the government and stressed that the Policy Analysis and Coordination unit should be used to make government food security policy coherert. Many of the recommendations of this report Involve reducing the role of government and predicting the positive effects of decreased government interference in markets.  1-10 1.46 nhe report (World Bank, 1993a) stressed agricultural policy as a vehicle for reducing rural poverty. They stated that hybrid maize production should still be encouraged due to its much greater productivity but with provisions that will make local processing and storage viable. The role of local millers is seen as being critical to agricultural development Without pan-seasonal pricing, local storage will become profitable and millers are most likely to become storage agents. The Bank also sees a role for the millers in supplying credit and inputs to farmers. This recommendation represents a call for government to stay out of markets and assigns a role to government in terms of extension of storage technology and milling technology. 1.47 The World Bank (1993a) also stressed that current hybrid maize varieties are not satisfactory. Varieties that mature in a ehorter period of time are needed to address labor shortages and drought relateai yield variability identified in the report. The authors stated that GRZ needs to take a role in diversifying agriculture away from maize production. The lifing of maize subsidization will decrease incentives to plant maize but the government needs to ensure that alternative crops such as sorghum, milet and cassava receive research and extension priority so that local farmers will have alternatives to planting maize. This process can also be facilitated by encouraging the development of markets for alternative grains. 1.48 The repurt found the government programs that monitor production levels for advanced warning of droughts to be critical. It recommended that the government establish a 3 month supply of maize and needs to restore its macro economic health so that it can import food dunng times of extreme need. Reliance on food aid would be a mistake since current pressures in advanced countries to limit agricultural surpluses will reduce future non-emergency food aid. 1.49 The World Bank (1993a) also recognized the need for direct assistance to Zambia's poor. A number of programs that were supported by the donor community (particularly through the World Food Program) but administered by local NGO's for drought relief were found to work very well. Such programs should be continued for targeted assistance to Zambia's poor and for in*astructure improvement. Some examples of the programs are Program Against Malnutrition (PAM), Program to Prevent Malnutrition (PPM) and Program for Urban Self Help (PUSH). While recognizing the need for targeted assistance to Zambia's poor, the study is pessimistic and not very specific about how aid should be targeted. It noted that households without able-bodied members and male heads should be ta-geted and that targeting should be undertaken at the local level. However, it was pessimistic about this process due to bureaucratic problems. 1.50 In the agricultural section of the World Bank(1993b) report, the need to utilize Zambia's plentiful land and restructure the land ownership system was emphasized. This restructring is vital since secure ownership will make agricultural capital improvements less risky. This report recognized that in tho L-t run Zambia's poor in both rural and urban areas will be adversely affected by the adjustment program. However, it did not provide detail on recommendations to alleviate these effects. The report stressed that donor assistance in the past was poorly coordinated and hence inefficient. Ibis assistance should be allocated not by donor interest but by Zambia's development needs. 1.51 The report recommended that health services should be provided closer to peoples home with tertiary facilities receiving a higher proportion of resources. Efforts to make social services more effective need to include decentralization of the Zambian bureaucracy. The report stressed that human capital needs to be improved through education.  1-11 G. SUMMARY OF RECOMMENDATIONS 1.52 A common theme of these reports s the need for poor farmers to gain access to credit. There is general acceptance that subsidies to alleviate povrty need to be targeted. However, the specific means of targeting is never presented In detail. It is genealy recognized that in mral Zambia, female headed households and households without able-bodied members should be targeted but the mechanics of this targeting are not spelled out. Ihe World Bank (1993b) was skeptical that targetg can be accomplished with much efficiency. II. POVERTY MEASUREMENT: METHODOLOGY 1.53 The objectve of developing a poverty measure is to identify vuerable groups of individuals who are unable to atain a standard of living that is consistent with social standards. Drawing a poverty line is the frst step in the development of a poverty measure that can be used as a tool to assess the geographic incidence of poverty, the variation of poverty across sub-groups of society. To maximize accuracy in the development of a poverty line, using the PSI data, the following issues need to be considered: 1.54 (1 Income and exnenditure data: This report used bousehold expenditure data to develop a variety of poverty measures. As the literature points out, household surveys are able to measure expenditure data more accurately than they de income data. This is mainly due to people's high variability of income throughout the year, particdularly for the poorest sub-groups of the population who often engage in either small-scale agriculture or subsistence agricultre, as well as the inability of terviewees to acately recall their incomesY There are obvious shortcomings in the PSI measurement of income, particularly in rural areas. For example, in the case of a pure subsistence household measured income would be zero, thus, as home produced consumption increases, as a share of total coneumption, an income measure becomes progressively more deficient. Evidence from other surveys (HEIS, IFPRI from Eastern province) indicate that the value of home produced consumption can often be as high as 90 peren of total food expenditures in rural areas. 1.55 OlB Homeproduced consumption: Expenditure data, as collected in the PSI and provided by CSO, did not allow for the creation of accurate poverty measures. That is because the PSI did not collect reliable inforw*tion on home-produced consumption or on the rental value of housing. II liUht of the importance of e two types of expenditures, two imputations were necessary to make the PSI expenditure data as consistent with actual household expenditures as possible. The values of home- produced consumption and housing rental were imputed. The expenditure variables were derived from section 7 of the PS survey instrumenL 1.56 To impute expenditures on home-produced consumption, cassava, hybrid and local maize, sales were subtracted from reported quantites produced and multiplied by the appropriate producer 1 Ravafio, Mati (192) Powr Cxqwbmow A sdde to CXncept e dMetP, Th World Bank, W&ahQVoa DC.  1-12 prices of these products.1' Tlere are two problems that stem from imputing the value of home- produced consumption. First, home-produced consumpdon of vegetables, groundnuts, other relishes and other grains such as millet and sorghum is not counted since vegetable and grain production or sales ;ere not available or coded in the data. Also, home-produced consumption of animal products, from milk and eggs to meats from livestock and fish is not accounted for. Omission of these home- produced consumptions items will cause problems in identifying poor households and measuring their consumption deficit. The problems will vary in severity depending on geographic location. In drier regions, were millet and sorghum tend to be substituted for maize, consumption will be undercounted. Information on geographical variation in livestock or vegetable consumption is not available; thus the geographical distribution of the bias is unknown. Second, the use of a single cassava price for computing the value of cassava of consumed from home production. While maize prices were fixed pan-erritorially at the time of the survey, cassava prices varied substantially. Since regional price information was not avaiable, a single cassava price was used. Once again, the bias introduced will depend on quantities consumed and actual price variation, both of which are unknown. Table I Regression for Rental Imputations 1.57 The second imputation was for the rental value of housing. Because only 3.2 percent of the Deedet Vafb: Ln Month Famiy ham rur households and 53 percent of the urban household reported paying housing rent it was Jdeaadet Cefafdn (Stadar necessary to impute the value of these expenditures. Vaitbiabl Er9or) To impute expenditure shares on rental housing, a Intfewp 2.829 (0.206) Rrwal -0.546 (0.194) simple linear model was developed in which luig .0.3S (0.057) repore monthly expenditures on housing were Mw 0.203 (0.092) regressed on a number of variables. The dependent Diastie 0.013 (0.00 variables were chosen because they represented both RDidance 4.010 (0.009) the supply and the demand side of the housing w 0-1.563 (0.020) market; thus the resulting model was a reduced- Cfppabel -1.892 (0.064) form representing the price of housing. The Etem -0.978 (0.1S6) following variables were selected based on their L4zaputI 1.192 (0.163) correspondence to price: an urban/rural dummy NWrsn -0.795 (0.203) (RURAL = I if from rnual area), a dummy variable Southern -0.939 (0.114) for the house's lighting source (MLIGHT = 1 if Westem *0.707 (0.18) house has electric lights). a dummy representing the house's water source (MWATER = 1 if the house N 3387 has access to a public or its own tap), distance in kiilometers to a post office, the distance interacted with RURAL to allow differential impact by place --- of residence, provincial dummy variables, and log household income (LINC taken from CSO's database). The regression showed no signs of heteroskadasticity, violations of normality, or problems with functional form. The results are presented in table 1. One problem with the imputation is the selectivity bias introduced by the truncation of the dependent variable, however correction for this bias was not possible given the limited number s of variables available to distinguish between zero and positive rental payments in the first case, and the value of the rental payment on the second. In other An a cwa= of 4 peae rntion of loal maim seeds for planting in the sbsequgt year was madeand sor ls wine nt ajuted for aow they any legjAdy be oonidWend as a conr=mption ear.  1-13 words, given the limited data, identification of the Inverse Mills Ratio would bave been impossible. 'he regressions were used to create imputed values of rental expenditures for the households for which rental expenditures were either missing or zero. 1.58 It is important to ackmowledge the possible differences in the cost of living among regins/provinces, for which we were unable to correct. Although pan-territorial pricing for the principal staple maize and maize meal were still in effect in 1991, other prices are reported to have been higher in remoter areas. For example, the lack of correction probably under es the prvence of poverty in remoter areas (the most poor areas). As the correcdons for local prices were not possible and as we could not fully account for own production, it is probably not relevant to compare -tween rural and urban areas. Table H Household Expenditures Avss oUweho Artstp Hasehold adrmptta Nan-meoola st Priority my 159.75 K 24,890.14 K3 2a" osehod .912.00 K 49,982.00 R Budge Suey, 1993 S Di&eme a0 arms ar Lim huded Povey Lim PMy Una Adjuaed Pove"ty +80% LUm +100% co POO _ 3 10. 66i.7 406 Poor 14.1 6.5 13.0 14.1 Non-Por C23 S2.6 20.3 4f3 Total ~~~ 00 ~ 1O0s 10oo IIIs & wuw Pioy survey H. 1993 and EWAma WoaWi ude uvy Meto& oana Uran cente of Luk, Ndola Utban, Kitwe, Cbiugo, Cbslabonbwe. Lanh MufiaIzKaiuhl Nabwo, ad LivIgton.. he emaining reas fall iio Non_mopol_tn aras. 1.59 Since one-time household sunreys such as the PS generally tend to underestimate household epditues ad home produce consumpdion, we have aKtempted to measure the exent to which household expenditres are underetimat. We compared the findings of the first quarter of the Household Budget Survey (HBS) (July-Septembe 1993) with PSUY and indeed we did find that MBS household expenditures estimates are much higer than the PSII ones (the factor of correction being 80% in metropolitan areas, and 100% elsewhere). If these corrections were made the prevalence of poverty would fall to 41 percent in rural areas and 11 percent in urban areas1 Cable 11). However, neither the regional ranking nor the correlations shown below change significantly, so we have remained with a profile based on the PSI data if DBo Suvy msu househld exnTur an aoount for hom produce onumpbon. I Unfoudy data by income group are stili no avalb from th HDS, this woud baw eabed a bdber orec  1-14 1.60 flhi) Absolute Povertv Une: An absolute poverty line was set at 1380 kwacba per adult male equivalent unit per month. This amount was the one obtained from the Prices and Income Commission/National Food and Nutrition Commission study which was close to the Priority Survey data collection and used by the Centd Statistics Office in Zambia. The cost of a food basket for an adult equivalent person was calculated at 962 kwacha per month at the prices of October/November 1991. The 1380 kwacha was arrived by dividing 962 kwacha by 70 percent which is the average percentage household expenditure on food. Other necessities such as housing, clothing, and educaton were also taken into account. Individuals with adult equivalent expenditures above 1380 kwacha per month are considered "Non-Poor', Individuals whose income is less than 1380 kwacha but above 962 kwacha per month are deemed to be 'Poor' while the individuals whose income is below 962 kwacha are deemed to be "Core Poor*.' 1.61 fiv) Relative Poverty Line: Since the absolute poverty line is by definition 'fixed in tem of the living standards indicator in use" as well as "fixed over the entire domain of the poverty comparison"' a relative poverty line was also developed. The relative poverty line is inended to assess the degree of inequality in the distribution of mean expenditures. For all Zambia, the relative poverty line was set at 70 percent of mean expenditures and the severe poverty line at 50 percent of mean expenditures, while the urban and rural parts use the mean of urban and rural expendiur (respectively). 1.62 (y) Adult Eaulvalency Scales: Equivalency scales were calculated to measure the relative incomes needed to enable families of different size, or under different e, to enjoy the same standard of living. This report follows the 'food energy method' where the cost of a bundle of goods necessary to attain some recommended food energy intake level is developed. The energy intake requirements used in this report are those recommended by the World Health Organizaon (WHO),V' which were also used by CSO. According to WHO energy intake requirements for moderate and very active levels are defined as: MWoder-ate Activit Level: Men - Most men in light industry, students, soldiers (not on active services, fishenren) Calorie requirements 3,000 day. Women - Light industry, housewives without mechanical housed appliances, students, department store workers. Calorie requirement 2,200/day. F Cenrl Sstical Oficee (19), Dhwou 4A4= w9t Ssew Rept, Lusaka Zambia t A m bogal point to raurboris tha im awe pobaby dltmm !a cm of livinS lag eIo for Wh& we were une to cooet, alhouhb 'lpan4wriwa priing for the pine4pl aple,e and fw maim saill in effct in 1991 diminish8the pmblem Pino, madom and seed daa sugt that do t of lving is Udy tDbe highe in offML-Aof-ral provnc. IV Ravaon, pp. 25. 1' Maurimce B Shbs andVenon R. Young (1988). Modex Ntrt in eal and D" La Migr- PI:r is. pp.M.  1-15 Yen Active Level: Men: some agricultura workers, unskilled laborers, foresaty workers, army recru and soldiers on adve services, mine workers, steel workers. Women: some farm workers (especially peasant agculture), dancers athletes (2,600 calories) Calorie requirements may also Increase for women who are pregna and lactating (2,900). 1.63 In Zambia the majority of women engage In small-scale or peasant agriculre which requir an energy intake level of 2,600. There is also a substantial proportion of women who are either pregnant or lactating and whose calorie requirements are 2,900 calorie/day. In light of this Issues, it was deemed appropriate to use higher calorie requirmens for Zamblan women and therefore adult female equivalency scales were set at unity (Table O). Table m Equivalency Scales Age Revised adlMt CSO's adult equivalency scales quivalency sl Child O yes 0 0 Cbild 1-3 years 0.36 0.36 Child 4.6 years 0.62 0.62 Child 7-9 years 0.78 0.78 Child 10-12 yeas 0.95 0.95 Adult feme 13+ yars 1.0 0.76 Adut male 13+ yeas 1.0 1.0 IIL MULTIVARL4TE ANALYSIS: RESuLMS 1.64 This section witt present multivariate analysis of poverty to analyze the correlation of household expenditur with the individual, household and community level c tcs thus far desribed In the chapter. For this purpose, two regression analysis were carried out, one for rural and for urban areas using the nattral logarithm of expendires per adult equivalent per month as the dependent variable. lbis type of muitivariate analysis is motivated by the notion of a household production process in which a typical household takes human capital endowmets and constraints and uses them to produce well-being, proxied in this case by expenditues. It could also be interpreted as a 'reduced form equation summaizing a sImdtaneous equation system. The resuts do not attempt to specify a caus reatonship, instead they attempt to show the independent efect of the various vaiables.  1-16 A. RURAL ZAMBIA& I~ ~ ~~~~~n ini~a aao ~~~~~M ,. DPC-' 1C)o .... . _......... -.- ............... .- ..--. ..UN:.....- ~pahlos IuXP - :nAO~lstAlOD~ SX, bWR , 1O , EDt, E, 1EIC, DIRC. ....,. : -H,we4 N N ;- S - ; X <~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~evor- b aO" ' DNOSP. ta . . .OVINCj YatlblesEXI' bqueloldt. 1 udeitu eA d'a-l?0 Iqan the AGfI h .o h oneodba,SXI b - Housol ads heade', Dby f s he aey pr- adl et nitursat Eontav es ag ; percentles thaeon mae-eaded fo usteholad ess Tha6ere is an)1. assciaio betawe frenwto household size mmand yexpsend OCm hduch thatger h h ausebold h-avel we expeni t uresi ix a dult e quvadlWen. iii *ven controi or depen. Denc DfSc, DHOn Sre tnt inrease in he usehood sieeis associatdi a .25 tihspita ls)i, epctv. VfIRU, WrMC WRAXO .......re dwuds fo ownenbp @y anyon .n thei feo0.9 peren decreaseminperuw cptrawr axpndzd~o eitureu4 By controang forab depentdency howm evsier,h wse see thatae much oal the i~ r cationshep betwesene hofusehol asiezte wand povsaedt foun Inaa other studiew fspuv may bes a ssets eeto dhendnc. Tho 's mal -. elasticy '- - ; - ' '. : . to the Boxl- Regression Equation for Rural Households Househol b Chctsics 1.65 Households headed by females have per adult equivalent eapenditures that are on average 18 percent less than malehoeaded houseolds. Ithere is an association between household size and expenditures in which larger h households have lower expenditures per adult equivalent. Even controlling for dependency effects, a one percent increasevin household size is associated with a .2S percent decrease in per capita expendituresxi'. By controlling for dependency, however, we see that much of the relationship between household size and poverty found in other studies may be atbutable to dependency. The smal elasticity associated with household size could not lead to the -degree of differences in household size by poverty class found in other studies. Eduaton 1.66 Education of the household head has a strong influence on household expenditures, even controlling for many of the other facors associated with income generation. The coefficients on education indicate that having a household head with some primary education is associated with 42 percent increase in per capita fmily expenditures while having more than 6 years of education is W Ihe LW! mode wosbeed omSpen ifiyan testsad tequstoan e various rstoidnot bamiessad to byDd. Alwang and Siegl for details). Ihe model satisfied all of the unelying statstical asujon. The influec of the provinces could have been modeed an a number of ways. nallly, full dcope and intercqsl diffeenes by prvic wes allowed. ollowin estimations of this full model, rsied verion were run where proica intq shifters were permited, and the slpe shfers were tese for iclson bloJck by block This lelt a modd with pm*inia dumm variables and slope shifters for thec primary schol vaial. This assoitin howevar, deeds criicaly on the household eqivlet deinition and shud not be oerIntel_td  1-17 associated with a 113 percent increase in per capita family expenditures. Returns to education in rural Zambia are substantial. Table IV: Regression Results for Household Employment !pdiWO in Rur.l A 1.67 Households headed by someone whose primary employment is in agriculture on average Dqm va have a 3 percent lower per capita expenditures than V d those headed by someone engaged in other activities. JaitsmqI Distanc to Services wt 6.943 ~~~.75 5 ht AaEH *0.190 (.2% Is A"E 0201 (0.2 1.68 Long distances from food maket, Su -0.201 ~~~~(0.046) DEPBATxQ *0.37g (D.06w transportation services, post office, and schools are LADBQ -0.2SS (0.0 associated with increaed poverty. The resuts BDI 0o34 @8) indicate that a 1 percent increase in distance from HAGUC .. (.043) food market, transportion services, and primary OVPO 0.497 (0.12Z school is associated with a .002, a .002, and a .054 DTRAN -0.249 (.115) percent decrease in per capita expenditures. D1'OC @ *6") Controlling for other influences, distance to services DM05? A0W4 (0.141) wm. -o.0039 (004 has the same wipact across all provinces except for wVRACr 0363 (0.0) distance to p schools. The only service in PRADIT 0.212 (0.141) which distance was not found to be significant was in UANUMBR 0.103 (0.0) hospital or health centers. Although Zambia's LAPU00.X42 ) retively good health coverage, a legacy of the NO mTHER .0.05 (0.089) 1970's and 1980's, may be a contrbutor factor, our 'ORThWBSTERN -026 (0.10 analysis does not measure the functionality, access SOUThERN o 0.177 (.0 or quality of health care. Obviously more research WESNTRN .0.147 (0.097 CO_POMT PSC O.QS6 AM.025) would be needed to account for such factors. EASTN * PSC 0.047 (00 CNTRAL * PSC 0.043 (0.02M LUSAK * NSC 0.062 (0.019) Ass NORThEN * PSC 0.044 (0.019) sOuuTHW*c 0E045 (0.0 1.69 The presence of a tractor in the ward of WER * PSC o.0S (0.017) residence is associated with higher expenditures. R2.3 This variable could either represent spill-over effects NG346 from having commercial farmers in the region or represent the existence of a rental market for hamster 100. plowing services. Alternatively, the positive sign on this coefficient might mean that for political or other reasons tracors were placed in wards with more wealthy residents. In any case, wards with tractors have, at the sample mean, a 44 percent higher household expenditures per capita as compared to wards without tractors. More investigation is needed to evaluate the impact of this variable; the data necessary to conduct such an analysis were wt available in the PSI. The information could, however, be used for targetng a program; there is unlikely to be significant poverty in wards with tractors. In addition, the coefficient on land ownehip shows that there is an association between land ownehip  1-18 and wealth even contolling for other factors. According to the PPA the cause goes from weat to tiding. i.e., you get a tile if you're wealthy. 1.70 Prownces: The ranking of provinces remain largely the same as with the blvara analysis above. B. URBAN ZMBMAW I~V1OR, I~]MMIA. ..9 'A fI .WO,d0B 1 HVlB vrw 1ItZx~L o ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-g RW=ion E"Mn for . Uan ou ds. 4 wwdaed ith ne-alf~ percewnteu ipr cait xpnitUure Th~Ui$s cresodst 4 the 'md . fm aatee s ld headdby o men an wot men Mpoore _m _who for atis dab *aubiw sgtW wa dm Isom .'.' ~PeI Wt I '-BCI,piu q svR omoaayi soee o rebio "modemVUMdRsWlI WU im4OhobambU moe 2 Regression Eaton for Urban Households Household Chwucteristkes 1.71 As for urban areas, the number of gadlt equivalents in the household was found to have a negative relationship with per capita expenlditures, a one percent increase in adult eqiaents is assoitd withi one-half percent decrease in per capita expenditre. This corresponds to the finding from Chawaena that eatended households headed by both men and women were poorer thanM non- W A Sfll modde which aoed ibr among the vabls was n and resri m eed mtl d d pretd abovewas idenified. Variables that wuertested bitdropped from the fin modd:wm the combined ifut of anp1oyumt, gne, and education; the efcsof distanes to bai svioe such as food maza, pos eM..e, trasporato serce, primar school, wate sonroe; the presence of 'modesn saIes nh Xh houselad (modm servic:es is defied as haig brtriid, ownl tap water, efse wcolion, and fuhtollet; in finll the typ of reidntalara hih hehoseol i lcaed(low, medium ar high cost). Nono of ths aibw undct to ho  1-19 extended households. Age of the household head did not appear to be a statistically significant determinant of household expenditure. Edawin 1.72 The effects of education and gender are significant and have a definite effect on household tenditures. For instance, male headed households with some education (7 years or less) or with more ta 7 years of education are, respectively, associated with a 21 and 48 percent increase in per capita household expenditure. Similarly, female headed households widt some education (7 years or less) and wth more than 7 years of education are, respectively, associated with a 25 and 67 pec increase in per capita household expenditure. These results indicate that the effect of pdmary school is similar for men and women, while the return to secondary education is higher for female headed households than for male headed households. Table V Regression Results for Household EmploYMent E h~~I Urbanl Areas Expendatures in rb 1.73 Having a household head who is employed is D,.v a* associated with higher per capita expditu, althoujb it varies according to *he type of employment Having a Va,ias COMi s_udaiMIbw household head whose primary employment is kftovt 3 (AM