36546 v3 Republic of Malawi The World Bank Malawi Poverty and Vulnerability Assessment Investing in Our Future Appendices December 2007 TABLE OF CONTENTS Annex 1A: Source of Data available to Measure Poverty and Living Standards in Malawi.......................... 3 Annex 1B: Note on Construction of Expenditure Aggregate and Poverty Lines for IHS2 ........................... 5 Annex 1C: Evaluating Alternative Equivalence Scales................................................................................ 20 Annex 1D: Methodology use in preparation of the poverty maps, and Maps of Gini coefficient and Poverty Gap at Traditional Authority level................................................................................................................ 28 Annex 1E: Comparable Poverty Estimates for 1998 and 2005 .................................................................... 47 Annex 1F: Malawi's progress towards the Millennium Development Goals (as of 2005)........................... 53 Annex 2A: Gender inequities in Malawi...................................................................................................... 87 Annex 2B: Forests, Biomass Use and Poverty in Malawi.......................................................................... 137 Annex 2C: Distribution and breakdown of time use by location and gender............................................ 178 Annex 2D: Subjective wellbeing................................................................................................................ 185 Annex 2E: OLS Regression on Determinants of Household Log Per Capita Expenditure ....................... 196 Annex 3A: Shocks reported by households between 1999-2004 in IHS1 complementary panel surveys.. 199 Annex 3B: Probit Regression on Determinants of Ultra-Poor Status of Households................................. 200 Annex 3C: Characteristics of shocks and Association between specific household characteristics and likelihood of experiencing a shock............................................................................................................. 206 Annex 3D: Temporary withdraw from school, students 10-15 .................................................................. 216 Annex 4A: Details of the Estimation of Malnutrition Indicators and Comparison with 1998 IHS1.......... 218 Annex 4B: The Foster, Greer, and Thorbecke 1984 poverty measures...................................................... 223 Annex 4C: Different indices of dietary diversity: definitions and a full break down of results ................. 225 Annex 4D: The determinants of child malnutrition and household caloric availability ............................. 227 Annex 5A: Prevalence of households in which a child resides without either parent (either due to being orphaned or fostered).................................................................................................................................. 233 Annex 5B: Results of the model probit on likelihood of orphans attending school ................................... 237 Annex 6A: Tables on National Accounts Sectoral Composition of GDP, and Employment ..................... 243 Annex 7A: Land Holdings by household and per capita............................................................................ 246 Annex 7B: Proportion of farmers that received advice from field assistant by type of advice and by quintile .................................................................................................................................................................... 247 Annex 7C: Smallholder farmers' efficiency in hybrid maize and burley tobacco production.................... 248 Annex 7D: Weather-based Insurance......................................................................................................... 258 Annex 8A: Determinants of Commercialization (and adoption of tobacco production)............................ 267 Annex 9A: Weather-based Insurance: National-level drought insurance................................................... 281 Annex 10A: Details of the methodology for Benefit Incidence Analysis .................................................. 287 Annex 11A: National Data Sources for Monitoring the Millennium Development Goals......................... 290 2 ANNEX 1A: SOURCE OF DATA AVAILABLE TO MEASURE POVERTY AND LIVING STANDARDS IN MALAWI Integrated Household Surveys cover a wide array of subject matter, with the primary objective to provide comprehensive information on the socio-economic status of the population in Malawi. One of the main purposes of these surveys was to understand the conditions in which the population of Malawi is living. These surveys collect information that can be used to calculate both monetary and non-monetary indicators of welfare. Using this information, it is possible to derive a profile of the poor and to provide a summary of the determinants of poverty in the country. In addition, the surveys can be used to monitor indicators related to the Malawi Poverty Reduction Strategy and the Millennium Development Goals (MDGs). The First Integrated Household Survey (IHS1) is a nationally representative sample survey conducted by the National Statistical Office (NSO) and designed to provide information on various aspects of welfare for the population of Malawi.(*) The IHS-1 was administered between November 1997 and October 1998. There are 10,631 households and 45,859 individuals in the data set being used for these analyses. Only about 60 percent of households had complete expenditure information, with the major problems attributed to the use of a consumption-expenditure diary (in a population that is over 85 percent rural with low levels of education) as well as low quality of survey management and data entry. In fact, the IHS-1 was administered to 12,960 households, but the released data set consisted of 10,698 households of which 6,586 were judged to have reliable expenditure and consumption information for use in constructing a consumption-expenditure aggregate. For more information on the IHS1, see "Profile of poverty in Malawi, 1998". The Second Malawi Integrated Household Survey (IHS2) is the follow up nationally representative sample survey conducted by the NSO to update the information on various aspects of welfare for the population of Malawi (**). The IHS-2 was administered between March 2004 and March 2005 and collected information from a nationally representative sample of 11,280 households and 51,288 household members. In addition to the household survey, the IHS2 collected information from the 564 communities in which these households reside. (Seventy-three of the 564 communities included in the survey are urban and 491 are rural.) The survey was administered simultaneously throughout the entire country to avoid introducing temporal or spatial effects. The IHS2 is based on the Living Standards Measurement Study (LSMS) survey methodology which combines an integrated set of questionnaires to collect data on household living standards with extensive quality control features. The household questionnaire gathers information needed to generate a monetary measure of poverty (using consumption and/or income). It also collects information on key assets and other living standards indicators, including: labor, housing, land, basic services, transport, health, education, financial assets, household enterprises, agriculture, and nutritional status. The IHS2 also include a number of non-standard modules designed to collect information on the social dimensions of poverty including crime and violence, risks and shocks and subjective wellbeing. The community questionnaire seeks responses from community leaders about 3 infrastructure and basic services and a variety of qualitative issues such as perceptions of services, living standards, and key issues facing the community. A price questionnaire also collects necessary information for constructing detailed spatial and temporal price indices. For more information on the IHS2, see the Basic Information Document for the IHS2. This is available from the Malawi NSO (http://www.nso.malawi.net/). Additional data sources available for Malawi: Population Census of 1998 Demographic and Health Surveys (DHS) in 1992, 1996, 2000, and 2004. DHS education (DHS EdData) in 2002, with a focus on education. Malawi Core Welfare Indicators Questionnaire (CWIQ) in 2002, which includes some direct questions about access to services. Malawi National Child Labor Survey in 2002, which has detailed labor information for children 5-17 years. Malawi Complementary Panel Survey, includes 5 panel survey rounds of about 700 households from the IHS-1. Notes: (*) There is an earlier household survey that was conducted in Malawi by the NSO in 1990, and was called the Household Expenditure and Small Scale Economic Activities (HESSEA). (**)To provide interim statistics on well-being for a subset of indicators, the NSO has also fielded the Core Welfare Indicator Questionnaire (CWIQ) in 2002 and the Welfare Monitoring Survey (WMS) in 2005. 4 ANNEX 1B: NOTE ON CONSTRUCTION OF EXPENDITURE AGGREGATE AND POVERTY LINES FOR IHS2 Introduction The Malawi 2004-2005 Integrated Household Survey (IHS2) was a comprehensive socio- economic survey of the living standards of households in Malawi.1 The National Statistics Office administered the IHS2 household questionnaire to 11,280 households from March 2004-April 2005. The survey was designed so that information gathered could be used for, among other things, an assessment of the incidence of poverty in the population at the district level and above. Poverty is that condition in which basic needs of a household (or individual) are not met. Clearly it is a multidimensional concept. Nevertheless, for purposes of identifying the poor using international standards, a monetary measure is developed as a welfare indicator for each household. Using this monetary measure, households can be ranked from richest to poorest. By developing a poverty line (a monetary threshold, below which a household is labeled as poor), households can be further described as poor or non-poor. For more nuanced classification, multiple poverty lines can be developed, for example, to identify the ultra poor. This note describes the construction of this welfare indicator, referred to as the consumption expenditure aggregate, and the development of poverty lines which can be applied to the welfare indicator to label households as poor. The following sections 2-9 describe how the aggregate was constructed, supplemented by appendices. Consumption Expenditure Aggregate The data on the IHS2 consumption expenditure aggregate is contained in two data sets. The first data set (ihs2_exp.dta) includes real and nominal values of expenditure for each household by disaggregated expenditure categories. The second data (ihs2_pov.dta) collapses the 33 expenditure categories from the first data set into 12 groups and includes the poverty lines and indicators for poor/nonpoor status of the household. ihs2_exp.dta Following the UN statistical classification system called "Classification of Individual Consumption According to Purpose" (COICOP), consumption related expenditures were coded into 2-digit categories as described in Table A1.1 (see Appendix 1.1 for the 3 digit coding). Expenditures included in each category are only those related to household consumption. Expenditures related to business activities (such as expenditures on fertilizer) were not included. In addition, the list of durable goods for which implied use- 1 An earlier Integrated Household Survey was administered in 1997-98 using a different questionnaire and methodology than the IHS2. For details of the IHS1 and IHS2, see the Basic Information Documents for these surveys published by the National Statistics Office. 5 value is computed is restricted to those related to consumer durables, and excludes assets or durables related to income-generation (for more information see Section 5). Broadly speaking, the consumption expenditures fall into four categories: 1) food 2) non-food, non-consumer durables 3) consumer durable goods 4) actual or self-estimated rental cost of housing In the following sections 3 to 6 there is more information on the calculation of these components of the consumption expenditures. Some general issues to note are: - Outlier values. Outliers are identified based on a combination of graphical review and standard deviations from means for each subcomponent. Generally, outlier values are replaced by median values based on households in the enumeration area (or by median values at the district/ national level if less than 5 observations at the lower level). - Annualization. Recall periods for expenditure subcomponents vary, ranging from the last 7 days to the last 12 months. All values are annualized. - Real values. For each subcomponent, the Malawi Kwacha value is a nominal value. Since the data were collected over 13 months and across different districts, there are price differences which need to be taken into account. In order to compare the monetary values across households, the nominal values are converted to real values to take into account spatial and temporal price differences using a price index developed for this analysis. ihs2_pov.dta In order to use consumption expenditure for poverty measurement, the 33 subcomponents in ihs2_exp.dta were summed into 12 categories in ihs2.pov.dta. The relationship between ihs2_pov.dta and ihs2_exp.dta can be seen in Table A1.1. The following exceptions to expenditures in ihs2_pov.dta were made, following recommendations by Deaton and Zaidi (2000): - Expenditures related to repairs on dwellings were excluded in order to avoid double counting as these items are implicitly included in estimated and actual rental values of the dwelling - Night's lodging in rest house or hotel was excluded as a lumpy (rare) expenditure. It is work noting that sometimes health expenditures are excluded from consumption expenditure aggregates as: (i) they might not express welfare as higher expenditure do not necessary indicate more welfare; (ii) they tend to be lumpy, infrequent and even large expenditures that if generalized as normal expenditures can give a wrong indication of welfare. Deaton and Zaidi (2002) recommend "Health expenses should only be included 6 if they have high income elasticity in relation to their transitory variance or measurement error." Perhaps health should therefore be excluded from the aggregate, although typically it is included in consumption aggregates. A preliminary review indicates that health expenditures have relatively high income elasticity in Malawi. On this basis, it was therefore decide to include health expenditures in the consumption aggregate. In addition to the household real annual expenditures for each of the 12 categories, this data file also includes - poverty lines (total and ultra-poor poverty lines) - an indicator variable for whether the household is poor (household per capita consumption below the total poverty line) - an indicator variable for whether the household is ultra poor (household per capita consumption below the ultra poverty line) Table A1.1 - Components of Consumption Expenditures Component COICOP Code Description Food/Beverage 11 Food 12 Beverage Alcohol/Tobacco 21 Alcohol 22 Tobacco Clothing/Footwear 31 Clothing 32 Footwear Housing/Utilities 41 Actual rents for housing 42 Estimated rents for housing 43 Regular maintenance and repair of dwelling 45 Electricity, gas, other fuels Furnishing 51 Decorations, carpets 52 Household textiles 53 Appliances 54 Dishes 55 Tools/equipment for home 56 Routine Home maintenance Health 61 Health drugs 62 Health out-patient 63 Health hospitalization Transport 71 Vehicles 72 Operation of vehicles 73 Transport Communications 81 Communications Recreation 91 Audio-visual 92 Major durables for recreation and culture, 94 Recreational and cultural services 95 Newspapers, books, stationery 7 Education 101 Education Vendors/Cafes 111 Vendors/Cafes/Restaurants 112 Accommodation services Miscellaneous Goods & 121 Personal care Services 122 Personal effects 124 Insurance Food Food consumption is reported at the household level in Sections H and I of the IHS2 questionnaire. Section H has food consumption over the last 3 days for a limited number of products which were self-produced (not purchased). Section I has food consumed over the last 7 days for a much larger range of products. Section H, by definition, is a subset of Section I. Section H was included in the questionnaire to facilitate comparison with IHS1, while Section I was included to give richer and more detailed information on household consumption and thereby welfare. For computing food consumption in the expenditure aggregate, only Section I is included. Section I collects information on 80 different food categories which were all included. In addition interviewers could include "other" products consumed that were not specified in the list. Products reported under "other" were also included and given a value in the expenditure aggregate. Section I has information on food consumption from three different sources: 1) purchased food (MK expenditure and quantity) 2) consumption from own-production (quantity only) 3) food received as gifts or free from some other sources (quantity only). In addition, drinking water expenditures from Section G are later added to the beverage subaggregate. Unit prices and conversion of units to grams For purchased food, both quantity and expenditure in Malawi Kwacha (MK) were collected. For consumption of food from own-production and food from gifts, only quantities are recorded. In order to compute the MK value for own-production and food from gifts, a unit price for each purchased product was calculated. These unit prices were then multiplied by the quantities reported from own-production and food from gifts. In general, the unit prices are computed as the median product price over all households in that geographical area at that time of the year. If less than seven households in that area and time purchased a particular product, a larger time span and/or geographical area was used. The use of a minimum seven households to calculate the unit-price guards against high volatility in unit prices over different households. In the questionnaire, quantities can be reported in 20 different units of measurement. In order to calculate the unit prices and to multiply the quantities with the unit prices, the different units of measurements were all converted into grams. For example, a cup of maize was set equal to 160 grams of maize. Note here that three additional units (spoon, 8 sachet/tube/packet, basin/pot) of measurement were added to the list after the completion of the survey. These units of measurement were all very prominent in the "Other Units" category and it is recommended that these units of measurement are included in future surveys. On the same note was there little reason to have oxcart included in section I, since consumption in a week of any product is extremely unlikely to reach that large a quantity. Finally within each product group (vegetables, cereals etc...) a "other" products is included. Most of these responses have been recoded into already existing product codes, however some answers could not be recoded. The value for these observations was estimated as the median for that product group. Eg. A household reported "22" as an "other vegetable". This is obviously a mistake, but it is also likely that the household actually consumed something, therefore the value is estimated as the median value for all vegetables. Non-food, Non-consumer Durables Goods and Services Expenditure on non-food, non-consumer durable goods and services are collected in several sections of the IHS2 household questionnaire. Relevant sections of the questionnaire include sections J, K, L, and G (utilities). The recall period varies across items, depending on the general frequency of purchase. More frequently purchased items have shorter recall periods, while less frequent purchases have long recall periods. The recall periods are last 7 days, last month, last 3 months, and last 12 months. All values are annualized. Education expenditures were reported either by type of expenditure, for example, tuition, books, uniforms, etc., or as an overall total. The total expenditure for education was calculated as the sum of all the sub-categories if they were reported, or as the overall total, whichever is greatest. Consumer Durable Goods Section M of the IHS2 questionnaire collects information on household ownership of 36 durable consumer goods. Seventeen of the items were deemed consumer durables and are included in the consumption expenditure aggregate (see Table 2). They were coded into COICOP subcomponents 53, 55, 71 and 91. The other durable consumer goods are predominately related to income generation or enable higher consumption and, as such, are considered production durables. Therefore, these were excluded. Of course, there are some goods for which it is not clear how to classify them. The assignments used were based on best practices as well as the assignments used in the IHS1 analysis. As durable consumer goods last for several years, and because it is clearly not the purchase itself of durables that is the relevant component of welfare, they require special treatment when calculating total expenditure. It is the use of a durable good that contributes to welfare, but since the use is rarely observed directly the yearly use value is estimated (following Deaton & Zaidi, 2002) in the following way: 1) Assuming that each product is uniformly distributed, expected lifetime for each product is calculated as twice the mean age of each product. 9 2) Remaining lifetime is calculated as current age minus expected lifetime. If current age of product exceeds expected lifetime, remaining lifetime is replaced by two years. The two years were arbitrarily chosen, however other expenditure aggregates have used this value in the past. The yearly use value of each product is then calculated as current value divided by remaining lifetime. Table A1.2 shows the estimated lifetime used for IHS2 computation of user-values. It also shows the values used in IHS1 (1997/98) analysis for the subset of durables included in IHS1 with number of observations in parenthesis. Table A1.2: Estimated Lifetime for Consumer Durables in IHS1 and IHS2 Section M IHS2 Expected Lifetime (years) Item Code Item IHS2 (2004 ) IHS1 (1998) 501 Mortar/pestle/pounding mill 14.5 (5641) 16.3 (1457) 502 Bed 15.0 (3625) 17.0 (2498) 503 Table 13.8 (4030) 16.0 (2488) 504 Chair 12.3 (5060) 13.8 (3202) 505 Fan 7.8 (262) 506 Air conditioner 8.8 (28) 8.1 (153)* 507 Radio (wireless) 6.8 (6184) 4.1 (1199) 508 Tape or CD player, HiFi 8.4 (1851) 509 Television & VCR 6.9 (439) 7.5 (131) 510 Sewing machine 22.6 (326) 511 Kerosene/paraffin stove 10.4 (241) 512 Electric or gas stove, hot plate 7.6 (287) 4.8 (858) 513 Refrigerator 10.1 (227) 9.6 (238) 514 Washing machine 10.7 (19) 19.7 (14) 515 Bicycle 12.2 (4082) 9.7 (2156) 516 Motorcycle/scooter 14.0 (41) 13.3 (41) 517 Car/motor vehicle 12.9 (136) 8.1 (124) Note: Number of observations are in parentheses. Some of the item labels were slightly different across the two surveys. For example, fan and air conditioner were combined into one category in IHS1. Actual or Self-estimated Rental Cost of Housing IHS2 has information on housing expenditures for people renting as well as people owning their own dwelling. Renters are asked what they pay in rent which is recorded in COICOP subcomponent 41. People who own their dwelling or do not otherwise pay rent (reside for free) are asked to estimate the rental value of their dwelling if they were to 10 rent it out. The estimated rental value for non-renters is COICOP subcomponent 42. Rental value (actual or self-estimated) is missing for 108 household. For these households, a rental value was imputed based on housing characteristics of similar dwellings in the same areas (urban/rural and actual rent/estimated rent). Price Deflation There are three main sources of prices available for price indexes to correct for temporal and spatial differences in prices: 1) food prices from Section I in the IHS2: 2) food and non food prices from the community questionnaire of the IHS2; and 3) prices collected by the National Statistics Office used to compute the consumer price index (CPI) for Malawi. Due to few purchases of many products, including products that make up large parts of the total consumption basket over parts of the year, it was not possible to construct a spatial and temporal index based on the price information in IHS2 household data. Only 17 products out of the 104 products that have unit prices from Section I have prices available in all 52 combinations of time (13 months) and space (4 areas ­ urban, north rural, central rural and south rural). Among the products that don't have information for all 52 combinations are essential products such as all the maize products. The price index could be restricted to 2 areas (urban and rural); this provides 48 products with information for the complete 26 combinations (13 months, 2 areas). This still leaves out some important products such as maize bran flour, maize grain, green maize, and cassava flour. Secondly, some products have very few price observations over parts of the year and parts of the country. The few price observations can destabilize the index as they are more likely to be influenced by measurement errors and variation in quality etc. Restricting the number of price observations in each combination to at least 7 to ensure reliable price information reduces the number of available products to 26. Again important products such as maize are missing from the index. Given these constraints, the spatial and temporal index was calculated based on the price data series collected for the NSO's CPI. To construct the price index for each of the 13 months of the survey and each of 4 areas (North rural, Central rural, South rural, and Urban areas), the following steps were taken: 1) Develop a spatial price index. This was developed using price data collected by NSO for February/March 2004, along with the national basket weights for 42 food and non-food items, for the four regions. The spatial price index reports highest prices in urban areas, followed by north rural. Central rural and South rural have the lowest prices. 2) Apply the regional CPI of the NSO. The regional CPIs from April 2004-April 2005 were applied to the spatial price index developed above. The price differences in February/March in combination with the already existing temporal index from NSO give us the complete price index which encompasses price differences across time and space. The mean national prices in February/March 2004 are 11 the base for the price index. Figure A1.1 shows the price index that deflates to real values of expenditure. Figure A1.1: Price Index )4002ra 140 130 M b/ 120 Fela 110 onitaN:esaB( 100 90 80 x 70 ndeI ecirP rch ril May une July g pt v n Au Oct No Dec Ja Feb rch Ma Ap J Se Ma Urban North Rural Central Rural South Rural Poverty line The total poverty line has two principal components; a food component and a non-food component. The food poverty line is the amount of expenditures below which a person is unable to purchase enough food to meet caloric requirements, based on a set basket of food. It is also known as the ultra poor poverty line. Food Poverty Line The food poverty line is derived by estimating the cost of buying a sufficient amount of calories to meet a recommended daily calorie requirement. It is constructed in the following steps: 1) Set the daily calorie requirement. This is done using the WHO recommended calorie requirements for moderate activity levels as described in Table A1.3. These calorie requirements were applied to the IHS2 sample to yield a median calorie requirement, which was 2,400 calories per day per person. Table A1.3: Calorie Requirements Age Calorie requirements <1 820 1-2 1150 12 2-3 1350 3-5 1550 5-7 1800 7-10 1950 10-12 2075 12-14 2250 14-16 2400 16-18 2500 18+ 2464 Source: Adapted from the World Health Organization (1985) "Energy and Protein Requirements." WHO Technical Report Series 724. Geneva: World Health Organization. 2) Identify cost per calorie for a reference population. A set level of calories can be consumed through many different combinations of food. In order to price calories, a reference population needs to be identified. Ideally, the reference population would be households who are not extremely poor (thus resorting to eating extremely cheap foods) nor wealthy (consuming very expensive calories). Table A1.4 presents the mean and median cost per calorie by decile. The reference population was chosen to be the population in the 5th and 6th deciles of the consumption aggregate distribution. In fact, these are households that are close to/near the poverty line itself (as seen by the poverty rate which is described below The cost per calorie applied is 11.48 MK per 1000 calories. Table A1.4: MK Cost per 1000 Calories by Decile Decile Mean Median 1 9.02 8.67 2 10.18 9.61 3 10.97 10.36 4 11.65 10.83 5 12.21 11.57 6 13.17 12.16 7 14.35 13.22 8 15.49 14.44 9 17.20 15.64 10 23.86 21.13 3) Calculate the food poverty line. The food poverty line is calculated as the price per calorie multiplied (.01148 MK) by the recommended per capita daily calorie requirement (2,400). The food poverty line is 10,029 MK per person per year, or 27.5 MK per person per day 13 The food poverty line is also the Ultra Poverty Line. The ultra poor are those households whose total per capita expenditure levels are below the food poverty line. Non-Food Poverty Line Identifying basic needs non-food expenditures is more difficult as there is no concept like calories which can be applied. The non-food component of the total poverty line is based on the non-food consumption of those households whose food consumption is close to the food poverty line. The non-food component is calculated as the weighted average non-food expenditure for those close to the food poverty line. The average expenditure is kernel weighted so that that those that are very close to the food poverty line are given most weight and those further away are given less weight. Households with food expenditure per capita that was five percent below or above the food poverty line was included in the kernel weighted average. The non-food component of that total poverty line is 6,136 MK per person per year, or 16.8 MK per person per day. Poverty Line The total poverty line is simply the sum of the food and non-food poverty lines described above. The poverty line is 16,165MK per person per year, or 44.3 MK per person per day. Once the poverty line is established, all households can be categorized as poor or non- poor depending on whether their per capita expenditure (their welfare indicator adjusted for household size) is below or above the poverty line. The poverty headcount, then, can be computed, indicating the proportion of individuals living in poverty. The poverty rate for the population of Malawi is 52.4%. This is the percent of the population whose household per capita consumption is below 16,165MK per year. $1 per day Poverty Estimates International comparisons of poverty rates are difficult using poverty estimates based on national absolute poverty lines, since different countries set different subsistence minimum standards. Rather, comparisons tend to be made using a fixed poverty line, for example the well-known "$1 per day" poverty estimates. In this approach, the poverty line of $1 per day is converted into local currency units, using the purchasing power parity (PPP) conversion factor rather than exchange rates. To be more precise, the $1 per day poverty line is $32.74 per month (or approximately $1.08 per day). This conversion is defined as the number of units of a country's currency required to purchase a standard basket of goods and services collected in all countries. The 1993 PPP conversion factor (1.5221) was updated using Malawi CPI inflation rates from 1993 to 2004 (18.48). In 2004, one US dollar is equal to 28.13 Malawi Kwacha using PPP conversion rates. This is equivalent to a "$1 per day" poverty line of 11,051 Kwacha per person per year. The portion of the population living below the "1$ per day" PPP line is 28%. 14 . 15 APPENDIX 1.1 Table A1.1.1: COICOP coding for expenditure items in IHS2 IHS2 COICOP Sec Question Description code C C30A Tuition, including any extra tuition fees 101 C C30B School books and other materials 101 C C30C School uniform clothing 101 C C30D Boarding school fees 101 C C30E Contributions for school building or maintenance 101 C C30F Parent association and other school related fees 101 C C30G Other School expenses 101 D D12 Expenditures for illnesses and injuries (medicine, tests, consultation, & in-patient fees) 062 D D13 Expenditure not related to an illness (preventative health care, pre-natal visits, check-ups) 062 D D14 Expenditure for non-prescription medicines (Panadol, Fansidar, cough syrup) 061 D D16 Hospitalization(s) or overnight stay(s) in a medical facility 063 D D19 Stay(s) at the traditional healer or faith healer 062 G G03 Estimated the rent for non-renters 042 G G04 Actual rent payment 041 G G19 Value of the firewood used in the past week, whether gathered or purchased 045 G G22 Electricity 045 G G26 Landline telephone 081 G G29 Cell phone 081 G G35 Drinking water 012 I Food 011 I 901 Tea 012 I 902 Coffee 012 I 903 Squash 012 I 904 Fruit juice 012 I 905 Freezes 012 I 906 Soft drinks 012 I 907 Chibuku/Napolo (commercial traditional-style beer) 021 I 908 Bottled / canned beer (Carlsberg, etc.) 021 I 909 Local sweet beer (thobwa) 021 I 910 Traditional beer (masase) 021 I 911 Wine or commercial liquor 021 I 912 Locally brewed liquor (kachasu) 021 J 101 Charcoal 045 J 102 Paraffin or kerosene 045 J 103 Cigarettes or other tobacco 022 J 104 Matches 045 J 105 Newspapers or magazines 095 J 106 Public transport ­ bus fare, taxi fare 073 J 201 Milling fees, grain 056 J 202 Bar soap (body soap or clothes soap) 122 J 203 Clothes soap (powder) 122 J 204 Toothpaste, toothbrush 122 J 205 Toilet paper 122 J 206 Glycerine, Vaseline, skin creams 122 J 207 Other personal products (shampoo, razor blades, cosmetics, hair products, etc.) 122 J 208 Household cleaning products (dish soap, toilet cleansers, etc.) 056 J 209 Light bulbs 045 J 210 Postage stamps or other postal fees 081 16 IHS2 COICOP Sec Question Description code J 212 Petrol or diesel 072 J 213 Motor vehicle service, repair, or parts 072 J 214 Bicycle service, repair, or parts 072 J 215 Wages paid to servants 056 J 218 Repairs to household and personal items (radios, watches, etc.) 053 K 301 Infant clothing 031 K 302 Baby nappies/diapers 031 K 303 Boy's trousers 031 K 304 Boy's shirts 031 K 305 Boy's jackets 031 K 306 Boy's undergarments 031 K 307 Boy's other clothing 031 K 308 Men's trousers 031 K 309 Men's shirts 031 K 310 Men's jackets 031 K 311 Men's undergarments 031 K 312 Men's other clothing 031 K 313 Girl's blouse/shirt 031 K 314 Girl's dress/skirt 031 K 315 Girl's undergarments 031 K 316 Girl's other clothing 031 K 317 Lady's blouse/shirt 031 K 318 Chitenje cloth 031 K 319 Lady's dress/skirt 031 K 320 Lady's undergarments 031 K 321 Lady's other clothing 031 K 322 Boy's shoes 032 K 323 Men's shoes 032 K 324 Girl's shoes 032 K 325 Lady's shoes 032 K 326 Cloth, thread, other sewing material 031 K 327 Laundry, dry cleaning, tailoring fees 031 K 328 Bowls, glassware, plates, silverware, etc. 054 K 329 Cooking utensils (cookpots, stirring spoons and wisks, etc.) 054 K 330 Cleaning utensils (brooms, brushes, etc.) 056 K 331 Torch / flashlight 055 K 332 Umbrella 121 K 333 Paraffin lamp (hurricane or pressure) 055 K 334 Stationery items (not for school) 095 K 335 Books (not for school) 095 K 336 Music or video cassette or CD 091 K 337 Tickets for sports / entertainment events 094 K 338 House decorations 051 L 401 Carpet, rugs, drapes, curtains 051 L 402 Linen - towels, sheets, blankets 052 L 403 Mat - sleeping or for drying maize flour 051 L 404 Mosquito net 051 L 405 Mattress 051 L 406 Sports & hobby equipment, musical instruments, toys 092 L 407 Film, film processing, camera 091 L 410 Insurance - health (MASM, etc.), auto, home,life 124 M 501 Mortar/pestle (mtondo) 055 M 502 Bed 056 17 IHS2 COICOP Sec Question Description code M 503 Table 056 M 504 Chair 056 M 505 Fan 056 M 506 Air conditioner 053 M 507 Radio (wireless) 091 M 508 Tape or CD player; HiFi 091 M 509 Television & VCR 091 M 510 Sewing machine 053 M 511 Kerosene/paraffin stove 053 M 512 Electric or gas stove; hot plate 053 M 513 Refrigerator 053 M 514 Washing machine 053 M 515 Bicycle 071 M 516 Motorcycle/scooter 071 M 517 Car 071 18 IMPUTATIONS FOR DOWA DISTRICT The consumption expenditure aggregate for the Dowa district has been replaced by imputed values. Preliminary calculations using the IHS2 data indicate that Dowa district is the least poor district in Malawi. According to the NSO, this observation does not match reality, and it is more likely that the IHS2 data collection for Dowa district may have been affected by problems. The values for Dowa district have therefore been substituted by an imputed value. To impute values for the Dowa district the following three steps were followed. First, we developed a regression model that explains the relationship between real per capita expenditure consumption and non-expenditure variables as household composition, employment etc. The model used to impute values was based on regressions of per capita expenditure from the neighboring districts of Ntchisi, Lilongwe Rural and Kasungu. Secondly, we impute total expenditure aggregate in Dowa based on the relationship between total per capita expenditure consumption and non-expenditure variables established from the neighboring districts. The methodology applied builds on recent methodological developments in survey-to-survey imputations. It is superior to regular regression analysis as it builds on the entire distribution and not only the mean. 100 simulations of per capita expenditure were done for the Dowa district. The median value of the simulations was used for the final value. (See Elbers, Chris, Jean O. Lanjouw, and Peter Lanjouw. 2004. "Imputed Welfare Estimates in Regression Analysis". Policy Research Working Paper 3294. The World Bank, Development Research Group, Poverty Team, May.) We used the povmap program which applies the methodology noted above. (The povmap program is a program developed by the World Bank which is designed to make poverty maps based on a survey and a census, as well as survey-to-survey imputations. The program is still under development.) For the simulations, 100 simulations were run, using non-parametric distributions for both cluster draws and household draws. The following broad categories of variables were included in the model: household composition (size, number of dependents etc), education, assets (ownership of chair etc), employment (percentage on labor market, household has enterprise, engaged in agriculture etc.), and community characteristics (eg. trading market). Finally, imputed values of the consumption expenditure subcomponents were based on the imputed total expenditure aggregate and the proportion of each subcomponent in the actual (non-imputed) data. For example if a household spent 60 percent of their non- imputed total consumption expenditure on food, they also spent 60 percent of their imputed total consumption expenditure on food. 19 ANNEX 1C: EVALUATING ALTERNATIVE EQUIVALENCE SCALES Consumption expenditure levels are reported by households and are converted into a per capita expenditure measure by simply dividing the total household expenditure by the number of household members.2 More complicated equivalence scales consider that there may be economies of scale in expenditure and adult equivalent adjustments could be made to account for differing requirements across different household members. For example, a 2-person household does not imply double expenditures on housing, utilities or other non-food items for which expenditure can be shared (these are public goods whose cost does not vary whether one person or a number of people use the good). Larger households might also buy food or nonfood items in bulk, which can mean lower prices or discounts. The age structure of household members, where a child is assumed to not be equivalent to an adult in terms of needs, could be considered by using adult equivalent adjustments for composition. The choice of equivalence scale reflects judgments about differences in needs. In Malawi, food is a large share of household expenditure. Since food is generally not associated with economies of scale, using an equivalent scale adjustment that has no economies of scale seems plausible. The exception to the notion that food does not have economies of scale would be bulk-purchase discounts. It certain foods are perishable and cost of storage is high, then large households may be better able to take advantage of bulk-purchase discounts. Adjusting for household size and composition can be done in numerous ways, and there is not one clear dominant choice and, therefore, no widely accepted scale. As noted above, per capita adjustment is computed by dividing total expenditure by household size (total number of household members) regardless of age and sex. Household size can therefore be given as follows: N= A + K Where A is the number of adults and K is the number of children in the household. In this case, a house with one person and spending y kwacha may be regarded as having the same welfare as the one that has two members and spending y*2 kwachas. The appeal of the per capita measure is that it is easy and intuitive to comprehend. However, the assumptions embedded are important to consider. By this measure, every member of the household is given an equal weight thereby assuming that of the total, expenditure is evenly distributed among the members. As well, needs are assumed to be the same across household members and economies of scale are assumed to be zero. 2 A household member is any resident in the dwelling who had been present in the dwelling for 9 or more of the previous 12 months. The household head, guests who had visited more than 3 months, young infants younger than 9 months, new spouses, and members residing elsewhere but still dependent on the household were also considered members. Servants, hourly workers and lodgers were not members if they had their own family elsewhere. 20 Given that there may be economies of scale and that needs may vary across members, household size can be adjusted as follows (See Jenkins, S. P. and Cowell, F. A. (1994) "Parametric equivalence scales and scale relativities," Economic Journal, 104: 891-900): PAE_N = (A+ K) PAE_N is the per adult equivalent household size. The parameter reflects the weight to convert every child into an adult equivalent. In other words, it measures the "size" of children relative to adults. The parameter captures economies of scale in expenditure. Because of the difficulty in justifying the use of a particular equivalence scale, a preferred practice is to examine the sensitivity of poverty estimates to different plausible scales. It is important to ascertain that the general profile is robust to choice of scale, be it per capita or other scales. We compare the profiles of the poor based on four measures of consumption: (a) per capita consumption, (b) a measure of consumption adjusted by number of equivalent adults, (c) two EOS-adjusted measure of adult-equivalent consumption ­ one with =.7 and the other with =.8 . Table A2.1 shows per adult equivalent weights that have been used in this analysis which are drawn from the weights used in Zambia which are based on WHO estimates of caloric needs by age. Table A2.1:Per Adult Equivalent Scales in Malawi Age Adult Equivalent <1 0.33 1-2 0.47 2-3 0.55 3-5 0.63 5-7 0.73 7-10 0.79 10-12 0.84 12-14 0.91 14-16 0.97 16-18 1.00 18+ 1.00 Note: "Zambia Poverty And Vulnerability Assessment." World Bank. 2005. In order to examine the robustness of the per capita measure, we will set the poverty rate at a fixed level (40% of the population) and assess the profile for the alternative measures of adult equiavlent household size. We assess the scales along two dimensions using the fixed national poverty rate of 40%: 1) the poverty rate across groups and 2) the share of the poor across groups. The profile is examined with respect to several household 21 characteristics: location, demographics in terms of number of children, education of the household head, and gender of the household head. As shown in Figures A2.1-A2.9 below, the poverty profile (in terms of poverty rates or shares of the poor) is remarkably stable between the alternative scales. That is, the profile of the low income population is similar regardless of the use of per capita scales or scales based on equivalence weights in Table A2.1, or these adult equivalence weights further adjusted for economies of scale. 22 Figure A2.1: Poverty Rate by Location (Fixed National Poverty Rate of 40%) 1 1 or 0 Per capita po onitr Adult equiv. 0 EOS(.8) poorP 0 EOS(.7) 0 - Urban Rural North Rural Rural South Centre Figure A2.2: Share of the Poor by Location (Fixed National Poverty Rate of 40%) 60 50 roop 40 Per capita ehtfo Adult equiv. 30 EOS(.8) e arhS 20 EOS(.7) 10 0 Urban Rural North Rural Centre Rural South 23 Figure A2.3: Poverty Rate by Household Demographics (Fixed National Poverty Rate of 40%) 0.6 0.5 or 0.4 Per capita po noit Adult equiv. 0.3 or EOS(.8) oprP 0.2 EOS(.7) 0.1 0 NO KIDS 1-2 KIDS 3+ KIDS Figure A2.4: Share of the Poor by Household Demographics (Fixed National Poverty Rate of 40%) 120 100 roop 80 Per capita ehtfo Adult equiv. 60 EOS(.8) arehS 40 EOS(.7) 20 0 NO KIDS 1-2 KIDS 3+ KIDS Total 24 Figure A2.5: Poverty Rate by Education of the Household Head (Fixed National Poverty Rate of 40%) 0.7 0.6 0.5 Per capita poor 0.4 Adult equiv. onit 0.3 EOS(.8) oporrP EOS(.7) 0.2 0.1 0 None Jr. Primary Sr. Secondary University Primary Figure A2.6: Share of the Poor by Education of the Household Head (Fixed National Poverty Rate of 40%) 45 40 35 roop 30 Per capita ehtfo 25 Adult equiv. 20 EOS(.8) arehS 15 EOS(.7) 10 5 0 None Jr. Primary Sr. Primary Secondary University 25 Figure A2.7: Poverty Rate by Gender of the Household Head (Fixed National Poverty Rate of 40%) 0.6 0.5 0.4 Per capita poor Adult equiv. onit 0.3 EOS(.8) oporrP 0.2 EOS(.7) 0.1 0 Male Female Figure A2.8: Poverty Rate by Gender of the Household Head and Location (Fixed National Poverty Rate of 40%) 0.7 0.6 0.5 Urban poor 0.4 Rural North onit 0.3 Rural Centre oporrP0.2 Rural South 0.1 0 r Pe atip tlu .vui ca Ad eq )8.(S )7.(S r tlu Pe atip .vui ca Ad eq )8.(S )7.(S EO EO EO EO Male Female 26 Figure A2.9: Share of the Poor by Gender of the Household Head (Fixed National Poverty Rate of 40%) 90 80 70 roo 60 Per capita ephtfo 50 Adult equiv. 40 EOS(.8) e arhS30 EOS(.7) 20 10 0 Male Female 27 ANNEX 1D: METHODOLOGY USE IN PREPARATION OF THE POVERTY MAPS, AND MAPS OF GINI COEFFICIENT AND POVERTY GAP AT TRADITIONAL AUTHORITY LEVEL This note contains background information on the update of the 2002 Malawi poverty map. The idea of a poverty map is to combine census information and survey information to estimate poverty and inequality at a lower level than the survey is able to. The 2002 Malawi poverty map was based on the 1997/98 IHS1 survey and the 1998 census while this update of the poverty map is based on the 2004/05 IHS2 survey and the 1998 census. That is, the predictions of poverty in this update and the 2002 map build on the same census data and changes in poverty estimates between the two generally stems from changes in the survey information. Since the IHS2 2004/05 survey has improved data quality this notes explains the update based on this new survey. Methodology The methodology for this poverty map generally follows previous work in Malawi, Madagascar and other countries. The methodology is not described in this note, but for more information and description of the methodology see Mistiaen et al, 2002, Demombynes et all, 2002, and Elbers et al., 2002. For information on the first Malawi poverty map see Malawi Social Atlas and background papers by Benson, t, 2002. This note only describes the choices made for this Poverty Map. The following steps were taken: 1. Aligning of Data Potential explanatory variables from both IHS2 and the 1998 census are compared. The chief objective of the comparability assessment in this stage is to determine if the survey variable can reasonably be said to contain the same information as the corresponding census variable. This procedure is repeated for each of the four models used for the poverty map3. The set of common variables was taken from the previous poverty map and additional variables were added as IHS2 has information not available in IHS1. The general decision rule for being statistically similar was chosen to be variables with a census mean within the survey 95 percent CI interval. Exceptions were made to household size which is included even though it fell outside the 95 percent CI and a few variables in the northern model that were in the 99 CI interval. Household size is included due to its known relationship with poverty. Further there is no reason to doubt that household size variable should contain the same information across the survey and the census. The variables that were found comparable across the survey and census and found significant in at least one model are listed in appendix table 14. A description of the variables is found in appendix table 3. 2. Estimating household consumption. 3Most poverty maps estimate a model for each strata in the data. IHS2 has 30 strata which is a high number of strata considering the size of the country. 4The aligning of the data show that a number of variables are not found comparable across the two data sets, which might be due to the long time span between them. This also includes demographic variables. 28 For estimating correlates between log per capita consumption expenditure and explanatory variables stepwise regressions in STATA are used to select a subset of variables from the set of "comparable" variables from above. Estimations are done with household weights. All household survey variables that were significant at the 5% level were selected to be in the regressions. A portion of the error component in these regressions is attributable to the location effect. In order to reduce the magnitude of the location effect (and thus of the errors in our final welfare measure estimates) additional explanatory variables related to location is added. The variables used in the 2002 poverty map is also used in this poverty map, which include variables based on GIS information as nearest urban center and ea means of other variables. In addition district dummies were included. Location variables were included if significant at 5 percent level. The number of location variables was at times reduced even though they were significant at the 5 percent level in order not to over fit the model (the square of number of clusters in each model was used as indication of maximum number of cluster variables). 3. Estimating Household error. The household component of the error term is also estimated. This is also known as the heteroskedasticity or alpha model. For this model the pool of potential variables include all those available that were comparable between the census and survey and the predicted values of log per capita expenditure from the household consumption model. A significance level of 5 percent was also used here. 4. Estimation level. Poverty is generally estimated at Traditional Authority (TA) level. However, too few observations in an area will lead to very high standard errors. Therefore are TAs with less than 100 households not estimated (these were not estimated in the previous poverty map either). These TAs are typically parks, reserves and non developed urban areas. Further is some TAs - in particular in urban areas - merged with each other to achieve sufficient large estimation areas. Appendix table 4 that shows the final estimates also show, which TAs that have been merged with each other. Table one in section two shows a comparison of the resulting standard errors from the estimation and the standard errors in the survey. Finally institutions (households with more than 20 individuals) were also excluded. 5. Estimation with Povmap v1.2 The estimations were done using PovMap v1.2 a program developed by the research group of the World Bank (DECRG). The estimations were done with the following settings: The distributions for cluster and household draws were non parametric. Auto function were used for minimum and maximum xb-hat , and the estimated log consumption (y). The trimming fraction of "alpha" draws were set at 0.99. The trim for cluster and household error draws were set at 1.25. 29 Results Table 2 in the appendix shows the final four models and the variables included. The models perform well with adjusted R-squares between 0.41 and 0.67. The estimations of poverty and inequality (with standard errors) at TA level can be seen in table 4 in the appendix. Comparing estimation results from the povmap program with the survey results at stratum level also gives satisfying results. All stratum poverty estimates from the povmap program are within the 95 CI interval from the survey. Further 26 of 30 of the survey estimates are within the 95 CI of the poverty map5. This can be seen in appendix table 5. Standards errors of similar size as the survey at stratum level is often deemed an acceptable level for the poverty map too. Table 1 below shows that povmap errors are satisfactory, actually substantially below survey errors. Table 1 ­ Average Estimate to Standard Error Ratio Urban North Central South Survey 0.182 0.111 0.101 0.078 Poverty Map 0.040 0.079 0.045 0.044 Malawi Poverty Map in Comparison A direct comparison between the 2002 poverty map and this update is not possible for several reasons; 1) The consumption aggregation being regressed on is constructed differently in the two surveys (IHS1 and IHS2). This leads to two different assessments of poverty. For example is the estimate of urban poverty vastly different in IHS1 and IHS2. 2) The estimation areas are not always the same. 3) The pool of comparable variables has changed. A direct comparison over countries is also not possible. However it might be interesting to see if this poverty map performs vastly different compared to others, especially since it is based on four models and not one per stratum. Table 2 below shows r-squared levels for the regression models for different poverty maps. The level of explanatory power of the models is as high as the models for South Africa and Ecuador, while better than the Madagascar, Mozambique and the earlier one for Malawi. Table 2 ­ Comparison of Adjusted R2 over Countries Malawi* based on IHS2 0.41 to 0.67 Malawi based IHS1 0.25 to 0.59 Mozambique* 0.27 to 0.55 Ecuador 0.45 to 0.77 Madagascar 0.24 to 0.64 South Africa 0.47 to 0.72 Source: Demombynes , 2002. Benson, 2002. Simpler, 2003. 5The 95 CI for the poverty map estimates has been approximated as two times the standard error. 30 The poverty map also compares satisfactory, when looking at ability to replicate the survey results at stratum level. This can be seen in table 3 below. Table 3 ­ Comparison of Poverty Estimates between Survey and Census Malawi based on IHS2 All poverty map estimates within 95 percent CI of survey. 26 out of 30 survey estimates within povmap 95 percent CI.* Malawi based IHS1 7 out of 31 areas have statistically different estimates between the two estimates. Ecuador 6 out of 8 estimates within each others 95 percent CI Madagascar All estimates within each others 95 percent CI South Africa Can not reject equality of all at 5 percent significance level. Source: Demombynes , 2002. Benson, 2002. * Approximated by two times standard error. 31 32 mean C 22.74 4.15 0.33 2.15 0.49 0.21 0.40 0.53 0.40 0.49 0.24 0.40 0.62 0.41 0.14 0.03 0.07 0.04 0.07 3.19 0.04 0.00 0.13 0.24 0.14 0.16 2.49 0.98 0.01 0.73 0.39 h ut lim So L 22.45 4.21 0.28 2.15 0.54 0.22 0.33 0.50 0.41 0.53 0.25 0.33 0.58 0.42 0.13 0.05 0.05 0.03 0.04 3.66 0.06 0.00 0.13 0.32 0.13 0.12 2.45 0.94 0.00 0.68 0.37 Rural lim U 24.41 4.39 0.31 2.26 0.59 0.25 0.36 0.55 0.46 0.58 0.28 0.36 0.62 0.47 0.17 0.07 0.09 0.05 0.06 4.08 0.08 0.01 0.18 0.35 0.17 0.19 2.57 0.98 0.01 0.74 0.41 Incl yes no no yes no no no yes no no no no no no yes no yes yes no no no yes no no yes yes yes no yes yes yes mean 26.23 4.50 0.31 2.28 0.54 0.22 0.40 0.59 0.47 0.54 0.23 0.39 0.65 0.48 0.09 0.02 0.03 0.04 0.03 3.52 0.02 0.00 0.10 0.27 0.02 0.09 2.53 0.99 0.00 0.82 0.41 C_ Survey ntral Ce L_Lim 29.20 4.85 0.32 2.46 0.65 0.24 0.40 0.64 0.53 0.65 0.28 0.40 0.66 0.52 0.15 0.06 0.03 0.05 0.04 4.33 0.11 0.01 0.11 0.38 0.04 0.13 2.40 0.99 0.01 0.78 0.41 from Rural U_lim 26.87 4.66 0.28 2.35 0.60 0.21 0.36 0.59 0.48 0.59 0.25 0.36 0.61 0.48 0.11 0.04 0.01 0.03 0.02 3.92 0.08 0.00 0.07 0.34 0.01 0.07 2.25 0.97 0.00 0.72 0.37 Limits s s s s s erw Incl no no ye no no yes yes ye no no no yes yes yes no no yes yes yes no no yes yes no ye ye no no yes no ye Lo 51 63 15 07 07 53 13 mean and 33.92 5.08 0.43 2.61 0.63 0.25 0.41 0.67 0. 0. 0.26 0.45 0.75 0.52 0. 0.03 0. 0.07 0. 5. 0.05 0.00 0. 0.40 0.03 0.21 3.02 0.99 0.00 0.73 0.34 C_ Upper rthoN 04 44 L_Lim 31.44 5.00 0.35 2.62 0.63 0.27 0.42 0.63 0.56 0.71 0.30 0.43 0.69 0.57 0.11 0. 0.03 0.05 0.04 6.38 0.07 0.00 0.12 0. 0.02 0.15 2.93 0.99 0.01 0.63 0.34 thiw Rural n lim U 28.29 4.75 0.30 2.45 0.54 0.22 0.37 0.55 0.48 0.62 0.25 0.38 0.62 0.49 0.08 0.02 0.02 0.03 0.02 6.00 0.05 0.00 0.09 0.39 0.01 0.12 2.78 0.98 0.00 0.58 0.29 Mea Incl no no no yes no yes yes no yes yes yes no no yes no yes no no no no no yes no yes no no no yes yes no no Census of mean 25.39 4.33 0.40 2.07 0.46 0.11 0.48 0.82 0.39 0.51 0.07 0.33 0.77 0.39 0.68 0.22 0.26 0.25 0.23 7.96 0.09 0.01 0.57 0.31 0.20 0.81 2.85 0.64 0.13 0.22 0.23 C_ lim L 25.74 4.48 0.39 2.21 0.48 0.16 0.47 0.85 0.43 0.56 0.13 0.38 0.78 0.45 0.54 0.26 0.14 0.25 0.10 9.08 0.10 0.00 0.55 0.39 0.25 0.86 2.64 0.45 0.17 0.25 0.23 Urban Comparison-1 lim U 21.68 4.09 0.32 2.00 0.40 0.11 0.40 0.72 0.35 0.47 0.08 0.31 0.68 0.37 nte 0.42 0.18 0.08 0.15 0.06 7.71 0.05 0.00 0.45 0.33 0.11 0.68 2.28 0.31 0.06 0.14 0.17 my Table yes yes no yes yes yes no yes yes yes no yes yes yes no no no no no no no no no Incl Emplo yes yes yes d setssA yes yes yes yes yes hic an r 9 9 ol nda grap sqz ez iont sec nd cc d ng s odo ecl cndi Appendix Variable Demo hhsi hhsi nonfam female m6_14 m50up m30_4 m15_2 m0_05 f6_14 f50up f30_49 f15_29 f0_05 Educa tertind servi se prof othero hhhe fambus employe employee allscho Housi toilet taph2o room okwoc ecelk rts coo con bicy 12 228 33 SE 0.107 0.009 0.001 0.009 0.040 0.050 0.025 0.007 0.124 0.016 0.013 0.019 0.000 0.001 0.002 0.015 0.002 0.002 0.001 0.071 0.002 0.067 0.057 0.075 0.074 0.055 0.055 4559 0.507 Coef South -0.371** 0.019** 0.032** 0.298** 0.275** 0.038 0.089** 0.518** -0.181** 0.221** 0.060** 0.000 0.003* -0.005** 0.036* 0.002 0.002 0.001 0.007** -0.080** -0.178 -0.131* 0.156 0.204* -0.118** -0.235 -0.107 Rural Q T TR LECY D Variable HHSIZE HHSIZS M15_29 PROF TOILE TAPH2O ROOMS COOKELEC CONS BIC TERTIN POP EAPERMHS EAIMPTLT AVGMAXED BOMA ROAD URBAN POPDENS DIFF DIST302 DIST303 DIST305 DIST306 DIST307 DIST308 DIST311 9 192 SE 0.087 0.010 0.001 0.014 0.017 0.016 0.010 0.016 0.011 0.010 0.046 0.052 0.042 0.026 0.061 0.054 0.031 0.187 0.013 0.176 0.079 0.064 0.002 0.005 0.019 0.001 0.001 0.006 0.002 0.001 0.201 0.055 0.049 0.062 3840 0.452 ral Coef -0.309** 0.013** 0.017 0.064** 0.143** 0.101** 0.038* 0.066** -0.009 -0.037 -0.014 0.411** 0.007 -0.149* 0.174** 0.044 0.553** 0.290** 0.164 0.130 -0.313 0.002** -0.006 0.076** -0.001 0.004 0.020 -0.004 -0.003 0.004 0.079 -0.031 -0.159* -0.252** Cent percent. Rural 5 at Q EE ER G AMF ROCC T LECY T ificantn Variable HHSIZE HHSIZS NON M50UP M30_49 M15_29 F30_49 F15_29 F0_05 OTHE SECNDIND PROF EMPLOY EMPLOY TOILE TAPH2O COOKELEC BIC BOMA2 TRADIN FORES EAPERMHS EAIMPTLT AVGMAXED EANETENR BOMA HEALTHFA ROAD URBAN POPDENS DIST201 DIST208 DIST205 DIST207 Sig* 5 72 percent SE 0.367 0.024 0.002 0.025 0.218 0.135 0.263 0.023 0.022 0.026 0.023 0.103 0.021 0.025 0.030 0.023 0.048 0.020 0.170 0.127 0.126 0.095 0.316 0.010 0.005 0.162 1439 0.414 1 at th Coef Nor -0.287** 0.016** 0.017 0.348 -0.252 -0.285 -0.105** -0.078** -0.052* 0.011 0.454** -0.091** 0.074** -0.021 0.262** 0.169** -0.026 0.294 -0.297* -0.347** -0.297** -0.469 -0.020* -0.005 0.430* 0.318 significant ** te. Rural ma esti Q ER LECY MAO DLY OLS is Models Variable HHSIZE HHSIZS ALLSCHOOL COOKELEC COOKWOOD EMPLOY F0_05 F6_14 F50UP FEMALE FINTER M0_05 M30_49 M50UP BIC FAMBUS M6_14 TRADB DIST103 DIST104 DIST101 ROOMPCAP HEALTHFA POPDENS MEAN_ DIFF R^2 4 72 SE 0.373 0.034 0.002 0.056 0.034 0.028 0.035 0.049 0.046 0.048 0.046 0.053 0.048 0.055 0.062 0.038 0.034 0.004 0.073 0.000 0.002 0.037 0.000 0.233 1439 0.110 0.671 Adjusted Estimation­2 ept.cr inte Coef -0.185** 0.014** -0.055 -0.104** -0.079** -0.153** -0.076 -0.073 -0.007 -0.061 0.363** 0.068 0.118* 0.425** -0.261** -0.075* 0.048** 0.465* 0.000 0.004** 0.035 0.001* 0.002* -0.005 0.273** an thiw Urban Table holds s rs ata house str cluste Q of of of regressions T TR squared R GLS Appendix Variable HHSIZE HHSIZS FEMALE M6_14 M15_29 M0_05 F6_14 F0_05 F15_29 F30_49 TOILE TAPH2O FAMBUS COOKELEC CONS SERVICES HHHED ROOMPCAP POP EAPERMHS AVGMAXED POPDENS EANETENR DIST105 DIST206 Number number number Adj. All Appendix Table 3 ­ Description of Variables Household Variables Demographic hhsizsq Household size squared hhsize Household size nonfam* Household has member who are not of nuclear family nomarry* Household head not married femhhh* Female headed household female Number of females in household m6_14 Number of males aged 6 to 14 m50up Number of males above 50 m30_49 Number of males aged 30 to 49 m15_29 Number of males aged 15 to 29 m0_05 Number of males aged 0 to 05 f6_14 Number of females aged 6 to 14 f50up Number of females aged 50 and up f30_49 Number of females aged 30 to 49 f15_29 Number of females aged 15 to 29 f0_05 Number of females aged 0 to 5 Education and Employment tertind* Household members in tertiary industry services* Household members with services occupation secndind* Household members in secondary industry prof* Household members with professional, admin,or clerical occupation otherocc* Household members with other occupation(services, artisans, etc.) hhhed Educational level of head of hh in years fambus* Household has a family business employer* Head of household employer employee* Head of household employee allschool* All kids in primary shool age in school finter* A household member finished tertiary education Housing and Assets toilet* Dwelling has improved toilet facilities taph2o* Household gets water from tap rooms Number of rooms in household prfnlit* Household gets lighting from paraffin cookwood* Household cooks over firewood cookelec* Household cooks over electricity or gas Constr* Dwelling built by traditional materials bicycle* Household owns a bicycle EA Variables pop Population roompcap Avg. rooms per person in EA eapermhs Proportion of hhs in EA with houses of permanent materials eaimptlt Proportion of households in EA with improved toilets avgmaxed Average maximum education level in households in EA eanetenr Net enrollment rate in EA boma Straight-line distance (km) to nearest boma (district HQ) healthfa Straight-line distance (km) to nearest health facility market Straight-line distance (km) to nearest market centre 34 road Straight-line distance (km) to nearest primary or secondary road urban Straight-line distance (km) to nearest urban centre popdens Population density (persons/sq.km.) mean_yld Mean maize yield in EA over 20 years diff Difference maize yield in EA in '97 & '98 from long term mean *Dummy variables 35 Appendix Table 4 ­ Poverty and Inequality Estimates at TA Level TA code TA name FGT 0 SE FGT 0 Gini SE Gini 10101 TA Mwabulambya 0.693 0.059 0.320 0.014 10102 TA Mwenemisuku 0.729 0.060 0.307 0.015 10103 TA Mwenewenya 0.729 0.089 0.290 0.018 10104 TA Nthalire 0.814 0.055 0.333 0.023 10105 TA Kameme 0.748 0.061 0.318 0.018 10120 Chitipa Boma 0.342 0.127 0.337 0.021 10201 TA Kilupula 0.537 0.055 0.309 0.015 10202 SC Mwakaboko 0.459 0.080 0.313 0.022 10203 TA Kyungu 0.570 0.047 0.310 0.012 10204 TA Wasambo 0.536 0.064 0.340 0.018 10205 SC Mwirang'ombe 0.552 0.065 0.335 0.021 10220 Karonga Town 0.321 0.083 0.325 0.015 10301 TA Kabunduli 0.610 0.075 0.335 0.016 10302 TA Fukamapiri 0.546 0.075 0.329 0.016 10303 TA Malenga Mzoma 0.619 0.093 0.339 0.021 10304 SC Malanda 0.473 0.085 0.373 0.024 10305 SC Zilakoma 0.562 0.088 0.324 0.022 10306 TA Mankhambira 0.569 0.081 0.363 0.021 10307 SC Fukamalaza 0.573 0.087 0.424 0.034 10308 SC Mkumbira 0.673 0.079 0.325 0.030 10309, 10310, 10311, TA Musisya,SC Nyaluwanga, 10314 SC Mkondowe,TA Boghoyo 0.732 0.066 0.331 0.028 10312 TA Timbiri 0.645 0.080 0.311 0.019 10320 Nkhata Bay Boma 0.354 0.133 0.315 0.023 10401 TA Chikulamayembe 0.514 0.059 0.342 0.015 10402 TA Mwamlowe 0.680 0.087 0.354 0.032 10403 SC Mwahenga 0.536 0.087 0.333 0.023 10404 SC Mwalweni 0.760 0.074 0.339 0.026 10405 SC Kachulu 0.837 0.073 0.291 0.019 10106, 10407 SC Chapinduka, SC Mwankhunikira 0.608 0.072 0.329 0.020 10408, 10409 TA Katumbi, TA Zolokere 0.541 0.076 0.339 0.023 10420 Rumphi Boma 0.313 0.106 0.338 0.021 10501 TA M'Mbelwa 0.535 0.043 0.371 0.018 10502 TA Mtwalo 0.395 0.053 0.360 0.014 10503 SC Kampingo Sibande 0.546 0.044 0.354 0.018 10504 SC Jaravikuba Munthali 0.342 0.082 0.357 0.019 10505 TA Chindi 0.523 0.038 0.388 0.018 10506 TA Mzikubola 0.594 0.043 0.377 0.021 10507 TA Mabulabo 0.528 0.049 0.384 0.020 10508 SC Khosolo Gwaza Jere 0.574 0.059 0.360 0.017 10509 TA Mpherembe 0.321 0.051 0.363 0.014 10510 TA Mzukuzuku 0.470 0.046 0.376 0.018 10520 Mzimba Boma 0.265 0.072 0.369 0.022 10531, 10532, 10546 Nkhorongo Ward,Lupaso Ward, New Airport Site 0.436 0.068 0.381 0.016 10533 Zolozolo Ward 0.184 0.052 0.379 0.021 36 10134, 10535, 10338, Chiputula Ward,Chibanja Ward, 10339 Jombo Ward, Muzilawayingwe 0.305 0.055 0.364 0.013 Ward 10536 Mchengautuwa Ward 0.305 0.051 0.381 0.018 10537, 10544, 10545 Katoto Ward, Viphya Ward, Msongwe Ward 0.364 0.058 0.437 0.025 10140, 10541, 10542, Chasefu Ward, Katawa Ward, 10543 Masasa Ward, Kaning'ina Ward 0.076 0.033 0.4 0.018 10601 TA Mkumpha 0.413 0.107 0.319 0.023 20101 TA Kaluluma 0.328 0.04 0.337 0.01 20102 SC Simlemba 0.428 0.053 0.308 0.012 20103 SC M'nyanja 0.314 0.048 0.324 0.011 20104 SC Chisikwa 0.288 0.07 0.316 0.018 20105 TA Kaomba 0.394 0.051 0.325 0.008 20106 SC Lukwa 0.368 0.045 0.313 0.009 20107 SC Kawamba 0.385 0.048 0.304 0.01 20108 SC Njombwa 0.313 0.045 0.332 0.01 20109 SC Chilowamatambe 0.339 0.044 0.327 0.011 20110 TA Chulu 0.434 0.039 0.315 0.011 20111 TA Santhe 0.294 0.044 0.32 0.008 20112 TA Wimbe 0.372 0.038 0.313 0.009 20113 TA Kapelula 0.406 0.05 0.295 0.01 20114 TA Mwase 0.405 0.06 0.319 0.012 20120 Kasungu Boma 0.218 0.07 0.452 0.022 20201 TA Kanyenda 0.347 0.054 0.346 0.015 20202 SC Kafuzila 0.373 0.049 0.333 0.015 20203 TA Malenga Chanzi 0.582 0.037 0.313 0.007 20204 SC Mphonde 0.497 0.054 0.318 0.011 20205 TA Mwadzama 0.466 0.031 0.307 0.007 20206 SC Mwansambo 0.482 0.044 0.309 0.013 20220 Nkhotakota Boma 0.391 0.086 0.447 0.03 20301 TA Kasakula 0.561 0.052 0.29 0.011 20302 TA Chikho 0.536 0.041 0.295 0.009 20303 TA Kalumo 0.507 0.027 0.307 0.007 20304 SC Nthondo 0.537 0.046 0.287 0.008 20305 SC Chilooko 0.426 0.028 0.309 0.007 20320 Ntchisi Boma 0.28 0.071 0.454 0.026 20401 TA Dzoole 0.373 0.033 0.252 0.008 20402 SC Chakhaza 0.321 0.027 0.267 0.008 20403 SC Kayembe 0.335 0.041 0.271 0.01 20404 TA Chiwere 0.459 0.034 0.271 0.011 20405 SC Mkukula 0.445 0.034 0.256 0.007 20406 TA Msakambewa 0.478 0.033 0.254 0.008 20407 SC Mponela 0.366 0.042 0.261 0.01 20420 Dowa Boma 0.177 0.067 0.342 0.028 20421 Mponela Urban 0.109 0.043 0.344 0.02 20501 TA Maganga 0.586 0.05 0.344 0.02 20502 TA Karonga 0.581 0.046 0.306 0.008 20503 TA Pemba 0.692 0.054 0.292 0.008 20504 SC Kambwiri 0.594 0.058 0.3 0.009 20505 TA Ndindi 0.603 0.049 0.315 0.009 37 20506 SC Kambalame 0.633 0.056 0.337 0.013 20507 TA Khombedza 0.548 0.048 0.304 0.007 20508 SC Mwanza 0.607 0.056 0.298 0.009 20509 TA Kuluunda 0.615 0.059 0.313 0.017 20510 SC Msosa 0.492 0.081 0.305 0.018 20520 Salima Town 0.292 0.081 0.43 0.024 20521 Chipoka Urban 0.371 0.076 0.38 0.022 20601 TA Chadza 0.341 0.027 0.319 0.008 20602 TA Kalolo 0.317 0.022 0.319 0.007 20603 TA Chiseka 0.356 0.02 0.312 0.006 20604 TA Mazengera 0.374 0.029 0.305 0.008 20605 SC Chitekwele 0.449 0.041 0.282 0.008 20606 TA Khongoni 0.345 0.035 0.309 0.008 20607 TA Chimutu 0.347 0.027 0.316 0.007 20608 TA Chitukula 0.369 0.035 0.3 0.008 20609 SC Mtema 0.359 0.035 0.309 0.008 20610 TA Kalumbu 0.334 0.028 0.302 0.007 20611 SC Tsabango 0.38 0.032 0.315 0.011 20612 TA Kalumba 0.322 0.041 0.331 0.01 20613 SC Njewa 0.302 0.034 0.34 0.011 20614 TA Malili 0.321 0.024 0.32 0.009 20615 TA Kabudula 0.361 0.032 0.311 0.009 20631 Area 1 0.094 0.031 0.442 0.027 20632, 20633, 20639 Area 2, 3, 9 0.021 0.01 0.409 0.017 20637, 20663 Area 7,33 0.104 0.029 0.384 0.012 20638 Area 8 0.139 0.029 0.4 0.015 20640, 20641, 20642, 20644, 20660 Area 10, 11, 12, 14,15,18, 30 0.022 0.009 0.389 0.016 20651 Area 21 0.13 0.028 0.375 0.012 20652 Area 22 0.183 0.044 0.355 0.014 20653 Area 23 0.162 0.033 0.367 0.011 20654 Area 24 0.226 0.05 0.332 0.011 20655 Area 25 0.152 0.021 0.413 0.013 20656 Area 26 0.644 0.085 0.308 0.017 20657, 20659, 20669 Area 27, 28, 39 0.554 0.081 0.397 0.029 20665, 20673, 20674 Area 35,44 0.383 0.04 0.529 0.037 20666, 20668 Area 36,38 0.443 0.049 0.349 0.012 20676, 20687 Area 46, 57 0.259 0.035 0.357 0.011 20677 Area 47 0.015 0.008 0.362 0.018 20679 Area 49 0.126 0.026 0.463 0.021 20680 Area 50 0.6 0.062 0.309 0.011 20681 Area 51 0.267 0.044 0.352 0.024 20682, 20683, 20684 Area 52, 53, 54 0.396 0.041 0.54 0.021 20685 Area 55 0.692 0.062 0.297 0.012 20686 Area 56 0.393 0.05 0.336 0.014 20688 Area 58 0.301 0.036 0.439 0.019 20701 TA Mlonyeni 0.617 0.049 0.342 0.016 20702 SC Mavwere 0.526 0.047 0.344 0.016 20703 TA Zulu 0.601 0.044 0.375 0.017 20704 SC Mduwa 0.608 0.045 0.34 0.014 20705 TA Mkanda 0.573 0.047 0.344 0.015 38 20706 SC Dambe 0.529 0.052 0.348 0.016 20720 Mchinji Boma 0.405 0.079 0.473 0.025 20801 TA Pemba 0.445 0.036 0.306 0.006 20802 SC Chilikumwendo 0.364 0.039 0.3 0.007 20803 TA Kaphuka 0.444 0.035 0.309 0.005 20804 TA Tambala 0.543 0.041 0.298 0.008 20805 SC Chauma 0.477 0.052 0.294 0.007 20806 TA Kasumbu 0.611 0.039 0.307 0.008 20807 TA Kachindamoto 0.508 0.039 0.316 0.008 20808 SC Kamenya Gwaza 0.576 0.047 0.318 0.009 20820 Dedza Boma 0.329 0.075 0.452 0.026 20901 TA Phambala 0.456 0.027 0.313 0.007 20902 TA Mpando 0.599 0.042 0.303 0.008 20903 TA Kwataine 0.529 0.033 0.319 0.008 20904 SC Makwangwala 0.452 0.026 0.316 0.006 20905 SC Champiti 0.46 0.048 0.315 0.011 20906 TA Njolomole 0.53 0.034 0.318 0.007 20907 TA Chakhumbira 0.539 0.037 0.325 0.01 20908 SC Goodson Ganya 0.501 0.028 0.314 0.007 20909 TA Masasa 0.48 0.039 0.31 0.009 20920 Ntcheu Boma 0.312 0.078 0.466 0.027 30101 TA Mponda 0.641 0.028 0.347 0.008 30102 TA Chimwala 0.618 0.022 0.329 0.007 30103 TA Nankumba 0.544 0.038 0.355 0.017 30104 TA Jalasi 0.562 0.033 0.339 0.009 30105 SC Mbwana Nyambi 0.582 0.033 0.323 0.007 30106 SC Chowe 0.632 0.029 0.333 0.008 30107 TA Katuli 0.553 0.037 0.333 0.01 30108 TA Makanjila 0.613 0.078 0.321 0.008 30109 SC Namabvi 0.596 0.048 0.322 0.011 30120 Mangochi Town 0.343 0.043 0.406 0.016 30121 Monkey Bay Urban 0.318 0.061 0.422 0.023 30201 TA Liwonde 0.722 0.035 0.323 0.008 30202 SC Sitola 0.776 0.037 0.319 0.009 30203 TA Kawinga 0.674 0.034 0.322 0.007 30204 SC Chamba 0.767 0.042 0.33 0.013 30205 SC Mposa 0.714 0.046 0.319 0.01 30206 SC Mlomba 0.689 0.033 0.327 0.01 30207 SC Chikweo 0.66 0.044 0.318 0.01 30208 SC Ngokwe 0.652 0.05 0.314 0.012 30209 SC Chiwalo 0.672 0.048 0.327 0.014 30210 TA Nyambi 0.7 0.039 0.317 0.009 30220 Machinga Boma 0.478 0.127 0.356 0.022 30221 Liwonde Town 0.521 0.056 0.408 0.017 30301 TA Kuntumanji 0.684 0.033 0.321 0.006 30302 TA Mwambo 0.684 0.03 0.323 0.007 30303 SC Mkumbira 0.6 0.06 0.358 0.021 30304 TA Chikowi 0.677 0.037 0.326 0.008 30305 SC Mbiza 0.676 0.03 0.325 0.006 30306 TA Mlumbe 0.701 0.027 0.335 0.006 39 30307 TA Malemia 0.641 0.036 0.363 0.011 30331, 30332 Mbedza Ward, Mtiya Ward 0.251 0.042 0.462 0.024 30333, 30334, 30337 Masongola Ward, Chikamveka Ward, Chirunga Ward 0.063 0.022 0.412 0.018 30335 Chikamveka North Ward 0.402 0.047 0.362 0.016 30336 Chirunga East Ward 0.363 0.051 0.494 0.037 30338 Likangala Ward 0.31 0.044 0.345 0.011 30339, 30340, 30341 Zakazaka Ward, Zomba Central Ward, Chambo Ward 0.115 0.036 0.396 0.018 30342 Sadzi Ward 0.276 0.048 0.374 0.017 30343, 30344 Likangala Central Ward, Likangala South Ward 0.481 0.057 0.362 0.021 30401 TA Mpama 0.58 0.031 0.343 0.009 30402 TA Likoswe 0.604 0.028 0.346 0.009 30403 TA Kadewere 0.597 0.028 0.321 0.007 30404 TA Nkalo 0.586 0.03 0.327 0.008 30405 TA Chitera 0.631 0.038 0.322 0.01 30406 TA Nchema 0.6 0.034 0.327 0.008 30420 Chiradzulu Boma 0.388 0.084 0.381 0.021 30501 TA Kapeni 0.378 0.038 0.452 0.034 30502 TA Lundu 0.476 0.054 0.334 0.011 30503 TA Chigaru 0.499 0.051 0.335 0.01 30504 TA Kunthembwe 0.532 0.053 0.34 0.012 30505 TA Makata 0.528 0.06 0.323 0.012 30506 TA Kuntaja 0.453 0.045 0.369 0.011 30507 TA Machinjili 0.503 0.053 0.347 0.012 30508 TA Somba 0.564 0.047 0.343 0.01 30531 Michiru Ward 0.339 0.035 0.38 0.012 30532 South Lunzu Ward 0.334 0.039 0.343 0.011 30533 Mapanga Ward 0.373 0.055 0.363 0.016 30534 Nkolokoti Ward 0.385 0.041 0.359 0.012 30535 Ndirande North Ward 0.241 0.042 0.335 0.011 30536 Ndirande South Ward 0.181 0.032 0.361 0.011 30537 Ndirande West Ward 0.188 0.03 0.463 0.021 30538 Nyambadwe Ward 0.149 0.037 0.467 0.026 30539 Likhubula Ward 0.29 0.034 0.383 0.012 30540 Chilomoni Ward 0.164 0.035 0.365 0.014 30541 Blantyre West Ward 0.244 0.033 0.477 0.023 30542, 30543, 30544 Blantyre Central Ward, Blantyre East Ward, Chichiri Ward 0.048 0.015 0.4 0.017 30545 Mzedi Ward 0.471 0.047 0.398 0.022 30546 Bangwe Ward 0.241 0.052 0.367 0.015 30547 Namiyango Ward 0.334 0.038 0.379 0.018 30548 Limbe East Ward 0.19 0.024 0.462 0.018 Limbe Central Ward, Limbe West 30549, 30550 Ward 0.029 0.012 0.378 0.016 30551 Soche East Ward 0.016 0.009 0.358 0.017 30552 Soche West Ward 0.135 0.028 0.388 0.013 30553 Nancholi Ward 0.425 0.055 0.352 0.015 30554 Misesa Ward 0.501 0.05 0.345 0.011 30555, 30556 Chigumula Ward, Msamba Ward 0.44 0.048 0.404 0.012 40 30601 TA Dambe 0.577 0.057 0.319 0.012 30602 TA Mlauli 0.594 0.045 0.326 0.009 30603 TA Kanduku 0.61 0.051 0.32 0.011 30604 TA Nthache 0.571 0.047 0.322 0.01 30605 TA Symon 0.484 0.056 0.341 0.013 30606 TA Ngozi 0.474 0.058 0.336 0.011 30620 Mwanza Boma 0.286 0.056 0.387 0.016 30701 TA Nsabwe 0.693 0.041 0.336 0.009 30702 SC Thukuta 0.723 0.051 0.316 0.015 30703 SC Mbawela 0.741 0.035 0.327 0.012 30704 TA Changata 0.749 0.035 0.313 0.01 30705 SC Mphuka 0.743 0.031 0.325 0.009 30706 SC Kwethemule 0.655 0.033 0.355 0.01 30707 TA Kapichi 0.657 0.03 0.34 0.009 30708 TA Nchilamwela 0.598 0.03 0.365 0.012 30709 TA Chimaliro 0.673 0.028 0.336 0.007 30710 TA Bvumbwe 0.626 0.028 0.351 0.009 30711 TA Thomas 0.788 0.031 0.311 0.01 30720 Thyolo Boma 0.302 0.062 0.401 0.023 30721 Luchenza Town 0.409 0.054 0.396 0.02 30801 TA Mabuka 0.704 0.026 0.347 0.007 30802 SC Laston Njema 0.626 0.036 0.348 0.01 30803 TA Chikumbu 0.657 0.032 0.345 0.009 30804 TA Nthiramanja 0.674 0.039 0.331 0.009 30805 TA Nkanda 0.73 0.03 0.32 0.007 30806 SC Juma 0.745 0.031 0.32 0.008 30820 Mulanje Boma 0.397 0.051 0.394 0.02 30901 TA Mkhumba 0.604 0.022 0.34 0.011 30902 TA Nazombe 0.552 0.023 0.346 0.012 30920 Phalombe Boma 0.291 0.089 0.382 0.023 31001 TA Ngabu 0.675 0.038 0.338 0.008 31002 TA Lundu 0.477 0.03 0.39 0.016 31003 TA Chapananga 0.687 0.033 0.315 0.007 31004 TA Maseya 0.707 0.042 0.322 0.014 31005 TA Katunga 0.671 0.047 0.326 0.015 31006 TA Kasisi 0.738 0.036 0.341 0.015 31007 TA Makhwira 0.692 0.027 0.326 0.007 31020 Chikwawa Boma 0.477 0.06 0.412 0.021 31021 Ngabu Urban 0.367 0.067 0.413 0.02 31101 TA Ndamera 0.701 0.052 0.334 0.014 31102 TA Chimombo 0.749 0.055 0.332 0.015 31103 TA Nyachikadza 0.837 0.055 0.296 0.016 31104 TA Mlolo 0.668 0.046 0.344 0.009 31105 TA Tengani 0.734 0.045 0.327 0.011 31106 SC Mbenje 0.664 0.05 0.37 0.015 31107 TA Malemia 0.762 0.047 0.337 0.019 31108 TA Ngabu 0.724 0.059 0.325 0.012 31109 SC Makoka 0.825 0.051 0.327 0.024 31110 Mwabvi Game Reserve 0.796 0.063 0.335 0.029 31120 Nsanje Boma 0.596 0.071 0.398 0.019 41 31201 TA Nsamala 0.625 0.025 0.332 0.006 31202 TA Kalembo 0.608 0.026 0.33 0.007 31220 Balaka Town 0.255 0.047 0.396 0.017 All estimates are from the Povmap v1.2 program. Lines with one estimate for several TAs are merged areas for which only one estimate has been calculated. 42 Appendix Table 5 ­ Poverty Estimates at Stratum Level Survey and Povmap Code District Poverty CI of Poverty CI of estimate* Poverty IHS2 Estimate estimate Estimate Map mean IHS2 Povmap mean within Survey within Povmap survey CI CI Rural North 101 Chitipa 0.672 0.515 0.829 0.712 0.626 0.798 yes yes 102 Karonga 0.549 0.407 0.691 0.511 0.402 0.620 yes yes 103 Nkhata Bay 0.630 0.483 0.777 0.592 0.492 0.693 yes yes 104 Rumphi 0.616 0.477 0.755 0.550 0.443 0.656 yes yes 105 Mzimba 0.506 0.427 0.586 0.480 0.409 0.550 yes yes 106 0.436 0.191 0.681 Rural Center 201 Kasungu 0.449 0.349 0.550 0.356 0.296 0.415 yes no 202 Nkhotakota 0.480 0.365 0.595 0.441 0.388 0.494 yes yes 203 Ntchisi 0.473 0.336 0.611 0.487 0.446 0.528 yes yes 204 Dowa 0.366 0.305 0.427 0.381 0.346 0.417 yes yes 205 Salima 0.573 0.462 0.685 0.563 0.481 0.644 yes yes 206 Lilongwe 0.375 0.307 0.442 0.349 0.314 0.383 yes yes 207 Mchinji 0.596 0.469 0.723 0.568 0.483 0.653 yes yes 208 Dedza 0.546 0.469 0.623 0.479 0.415 0.543 yes no 209 Ntcheu 0.516 0.442 0.590 0.498 0.457 0.539 yes yes Rural South 301 Mangochi 0.607 0.527 0.688 0.580 0.533 0.627 yes yes 302 Machinga 0.737 0.670 0.803 0.691 0.634 0.748 yes yes 303 Zomba 0.700 0.599 0.802 0.681 0.622 0.739 yes yes 304 Chiradzulu 0.635 0.501 0.770 0.595 0.550 0.640 yes yes 305 Blantyre 0.465 0.325 0.605 0.474 0.389 0.560 yes yes 306 Mwanza 0.556 0.468 0.645 0.541 0.464 0.617 yes yes 307 Thyolo 0.649 0.556 0.743 0.669 0.622 0.717 yes yes 308 Mulanje 0.686 0.609 0.764 0.688 0.635 0.741 yes yes 309 Phalombe 0.619 0.525 0.714 0.582 0.546 0.619 yes no 310 Chikwawa 0.658 0.570 0.747 0.654 0.613 0.695 yes yes 311 Nsanje 0.760 0.692 0.828 0.700 0.626 0.773 yes yes 312 Balaka 0.668 0.557 0.778 0.598 0.554 0.642 yes no Urban 105 Mzuzu City 0.340 0.220 0.460 0.304 0.222 0.386 yes yes 206 Lilongwe City 0.246 0.137 0.355 0.233 0.190 0.275 yes yes 303 Zomba City 0.287 0.172 0.402 0.269 0.217 0.321 yes yes 305 Blantyre City 0.236 0.174 0.297 0.255 0.216 0.295 yes yes * CI approximated by two times the standard error 43 References Benson, Todd (2002). Malawi: An Atlas of Social Statistics . National Statistical Office, Government of Malawi and the International Food Policy Research Institute. Demombynes, Gabriel, Chris Elbers, Jenny Lanjouw, Peter Lanjouw, Johan Mistiaen, Berk Özler. "Producing an Improved Geographic Profile of Poverty: Methodology and Evidence from Three Developing Countries." World Institute for Development Economics Research, United Nations University discussion paper no. 2002/39. March 2002. Elbers, Chris, Jean O. Lanjouw and Peter Lanjouw (2002). "Micro-Level Estimation of Poverty and Inequality." Econometric 71:1, pages 355-364. Mistiaen, Johan, Berk Ozler (2002). "Putting Welfare on the Map in Madagascar." World Bank. Simler, Ken, Virgulino Nhate (2003). "Poverty, inequality, and geographic targeting: Evidence from small-area estimates in Mozambique." International Food Policy Research Institute, Ministry of Planning and Finance, Maputo, Mozambique. 44 MAP OF GINI COEFFICIENT AT TRADITIONAL AUTHORITY LEVEL (USING 1998 CENSUS AND 2005 IHS2) 45 MAP OF POVERTY GAP AT TRADITIONAL AUTHORITY LEVEL (USING 1998 CENSUS AND 2005 IHS2) 46 ANNEX 1E: COMPARABLE POVERTY ESTIMATES FOR 1998 AND 2005 This note describes the methodology used to develop comparable poverty indicators for Malawi from the two rounds of the Integrated Household Surveys. The first survey (IHS1) was administered in 1997-98. The second (IHS2) was administered in 2004-05. The issue of comparing welfare or poverty across time and surveys is not new. There is a large and expanding literature on how best to compare welfare indicators from two non- similar surveys. Changes in data collection methods, questions asked, etc., make direct comparison of poverty statistics problematic. This is the dilemma facing the analysis of the IHS data. There were some key differences in the surveys which are detailed in the Basic Information Document. As the poverty statistics rely on consumption expenditure data, one of the most critical differences in the IHS data rounds is how these data are collected. The IHS1 used a diary method to record consumption expenditure for food and frequently purchased items. The IHS2 was redesigned in part to take into account the experiences from IHS1 (both in terms of the fieldwork and subsequent analysis). A major redesign was to collect consumption expenditure information by using recall periods for all food and nonfood consumption expenditure (for example, 7 day recall for food) rather than a diary. Diaries are expensive and difficult to collect accurately especially among rural and illiterate populations. Thirty-eight percent of the 10,698 households in the IHS1 did not complete the diary (or the diary of expenditure was not consistently maintained). On the other hand, the food-recall method is, internationally, a common substitute to the diary for consumption measures to assess poverty. However, the levels of consumption between the two surveys are not directly comparable. Given these and other revisions to the survey methodology, the previous poverty estimates produced from the IHS1 should not be compared to the new poverty estimates from the IHS2. Rather, new poverty estimates must be computed from the IHS1 following a comparable approach used with the IHS2 data. There are numerous non-income dimensions which are directly comparable between IHS1 and IHS2, such as child malnutrition, school attendance, asset ownership, and employment activities. Nevertheless, income poverty is often taken as a primary, singular measure of progress in fighting poverty. Therefore, it is necessary to develop a poverty measure for IHS1 which can, in fact, be compared to the IHS2 poverty measure. The approach used in developing a revised 1998 poverty estimate, one which is comparable to the IHS2 estimate, follows recently-developed statistical techniques. We develop a comparable poverty measure in IHS1 by first imputing the consumption expenditures in IHS1 (described below). The methodology originated in Elbers et. al. (2002, 2003) and has since been widely applied in different countries, in particular for poverty maps, but also for survey-to-survey imputations (as used here). For example, this method is used to estimate population prevalence of HIV in Malawi, for lack of population-based data (Ivaschenko and Montana, 2005). An example closer to the situation in this note is from India where surveys were changed between rounds. Kijima and Lanjouw (2003) use the methodology to impute poverty at the regional level in India. Finally, recent work from Uganda is also very similar to the situation in Malawi in that 47 the surveys changed over time drawing into question the comparability of the poverty estimates (Luoto, 2005). Examples of a more simplified approach to the one used in these papers is the estimate of poverty status for households in the NSO's Core Welfare Indicator Questionnaire (CWIQ) 2002. Likewise, the IHS1 itself actually imputed per capita expenditure and poverty status for 38% of the 10,698 households in the survey (noted above). The imputation approach in IHS1 for these missing households is a simplified version of the approach described below. It was based on one regression model of per capita expenditure, whereas the approach here uses 4 region-specific models, 100 simulations, and estimates of otherwise omitted disturbance terms, for an improved fit. The steps involved in computing a comparable IHS1 poverty estimate are: 1) estimate per capita expenditure for IHS1 households based on a model of per capita expenditure developed from IHS2 using a set of household characteristics measured in both surveys, and 2) estimate poverty rates for households using the imputed per capita expenditure, applying the IHS2 poverty lines. The main assumption imbedded in this approach is that the correlation between poverty and the set of household characteristics has not changed significantly over time. To model this we first run a Generalized Least Squares (GLS) regression of the observed log per capita expenditure for household h as: (1) ln yh = xh+ uh, where xh is a vector of k parameters and uh is a disturbance term satisfying E[uh|xh] = 0. The set of household characteristics is in vector x. The vector includes demographic variables, household characteristics, and asset ownership, in addition to district indicator variables. The model in (1) is estimated separately for four regions (urban, north rural, central rural, south rural) using the IHS2 data. In a simple model of expenditure, we would impute values of per capita expenditure for IHS1 households, based on the set of x covariates and the estimated values. However, we can improve up on this estimate as follows. Because the disturbance term for households in the target population (IHS1 in this case) are always unknown, we estimate the expected value of the indicator given the IHS1 households' observable characteristics and the model of expenditure in (1). We denote this expectation as: (2) v = E[W | Xv , ], s s where is the vector of model parameters, including those which describe the distribution of the disturbances, and the superscript `s' indicates that the expectation is conditional on the sample of IHS1 households from district v rather than a census of households. In constructing an estimator of v we replace the unknown vector with consistent s estimators, ^ , from the IHS2 expenditure regression. This same approach is adopted out for the distribution of the residuals. In other words, in addition to assuming that the 48 distributions of the explanatory variables have not significantly changed over time, we also have to assume that the distribution of the household expenditures residuals in 1997 is determined by the same set of variables as in 2004. This yields ^v . This expectation s is generally analytically intractable so simulation is used to obtain our estimator, ~s . One hundred simulated draws are performed to derive our estimator for ^v in each of s the four models (urban, north rural, central rural, south rural). The four models use household weights and therefore serve as estimates for the population considered and not just the sample at hand. See Kijima and Lanjouw (2003) for more detail on the prediction error associated with our estimator ~s of the expected value of per capita expenditure for a given region. For the estimations we used the povmap program that builds on the methodology outlined. The povmap program is a program developed by the World Bank which is designed to make poverty maps based on a survey and a census. The program is still under development. As noted above, we estimate four models, which allows the estimated parameters to vary across the four areas and give a better fit of the models.6 For the simulations, we ran 100 simulations and used non-parametric distributions for both cluster draws and household draws. After estimating per capita expenditure 100 times for every IHS1 household, the povmap program then computes the poverty indicator (poor or non-poor) for each household's simulated per capita expenditure. Thus, every household has 100 imputed values for being poor. The median value is assigned to that household for the final computation of the poverty rate in IHS1. The detailed estimation models computed by povmap for each region are presented in Table 2 below. Since povmap follows a standard methodology in identifying these models, and that the models are not intended to provide a structural explanation of the data generation process, no explanation of these results is attempted here. Table 1 shows the imputed poverty rates for IHS1 and compares them with the IHS2 poverty estimates. Table 1.1: Poverty Rates for 1998 and 2005 1998 IHS1 2005 IHS2 Malawi 53.9 52.4 by Region Urban 19.6 25.4 North rural 55.9 56.3 Central rural 48.1 46.7 South Rural 67.2 64.4 6It is worth noting that Povmap also automatically models the distribution of the household error term: it picks a combination of predictors (from our list of explanatory variables, the y-hat as well as interaction terms) which captures most variance in the error term. 49 Table 2: Estimation models computed by povmap for each region Central Region Coefficient Std. Err. t |Prob|>t Label _intercept_ -16.8491 7.719826 -2.18257 2.91E-02 Intercept BED_0*_yhat_*_yhat_ 2.05E-03 9.26E-04 2.210517 2.71E-02 Dummy for (BED)=0*_yhat_*_yhat_ DIST20_0*_yhat_ 5.16E-02 1.20E-02 4.300302 1.75E-05 Dummy for (DIST20)=0*_yhat_ DIST21_0*_yhat_*_yhat_ 3.24E-03 0.001219 2.662161 7.80E-03 Dummy for (DIST21)=0*_yhat_*_yhat_ DIST25_0*_yhat_*_yhat_ 7.33E-03 1.64E-03 4.457506 8.53E-06 Dummy for (DIST25)=0*_yhat_*_yhat_ DIST27_1 -1.86991 0.115992 -16.1211 1.55E-15 Dummy for (DIST27)=1 DPEND*_yhat_*_yhat_ 2.62E-03 1.74E-03 1.508244 0.131575 dpend*_yhat_*_yhat_ HHENT_0*_yhat_*_yhat_ 1.65E-03 8.30E-04 1.990423 4.66E-02 Dummy for (HHENT)=0*_yhat_*_yhat_ HHSIZE -0.20362 4.69E-02 -4.34033 1.46E-05 hhsize HHSIZE2*_yhat_*_yhat_ 1.79E-04 4.09E-05 4.369129 1.28E-05 hhsize2*_yhat_*_yhat_ SICKLE_0*_yhat_*_yhat_ -1.74E-03 7.98E-04 -2.18223 2.92E-02 Dummy for (SICKLE)=0*_yhat_*_yhat_ SPOUSE_0*_yhat_*_yhat_ 1.65E-03 9.28E-04 1.773836 7.62E-02 Dummy for (SPOUSE)=0*_yhat_*_yhat_ STOVE_0 12.64419 7.718919 1.638078 0.101488 Dummy for (STOVE)=0 STOVE_0*_yhat_*_yhat_ 0.105591 6.13E-02 1.72172 0.085201 Dummy for (STOVE)=0*_yhat_*_yhat_ SST=22223.2711 SSR=2124.2702 F=31.1055 R2=0.0956 adjR2=0.0925 North Region Coefficient Std. Err. t |Prob|>t Label _intercept_ -3.7455 0.497725 -7.52522 9.37E-14 Intercept BED_0 0.32425 0.131817 2.459847 1.40E-02 Dummy for (BED)=0 DIST32_0 0.631398 0.179446 3.518589 4.47E-04 Dummy for (DIST32)=0 HHSIZE -0.35409 7.21E-02 -4.91025 1.01E-06 hhsize HHSIZE2*_yhat_*_yhat_ 3.12E-04 6.54E-05 4.777837 1.95E-06 hhsize2*_yhat_*_yhat_ MOTORCYCLE_0*_yhat_*_yhat _ -2.21E-02 1.13E-02 -1.95017 5.14E-02 Dummy for (MOTORCYCLE)=0*_yhat_*_yhat_ OX_0 -0.62055 0.412697 -1.50364 0.132895 Dummy for (OX)=0 PANGA_0*_yhat_*_yhat_ 2.21E-03 1.38E-03 1.606703 0.10834 Dummy for (PANGA)=0*_yhat_*_yhat_ REFRI_1 -1.65384 0.916048 -1.80541 7.12E-02 Dummy for (REFRI)=1 SST=8555.3504 SSR=326.8064 F=7.1042 R2=0.0382 adjR2=0.0328 South Region Coefficient Std. Err. t |Prob|>t Label _intercept_ -14.2867 3.7506 -3.8092 0.0001 Intercept _yhat_*_yhat_ 0.0305 0.0041 7.3834 0 _yhat_*_yhat_ DIST1_0 -0.3392 0.1538 -2.2061 0.0274 Dummy for (DIST1)=0 DIST10_0*_yhat_*_yhat_ 0.0026 0.0016 1.6394 0.1012 Dummy for (DIST10)=0*_yhat_*_yhat_ DIST11_0 -0.3288 0.1122 -2.9321 0.0034 Dummy for (DIST11)=0 DIST13_0 6.6562 3.7765 1.7625 0.078 Dummy for (DIST13)=0 DIST13_0*_yhat_ 0.6276 0.3895 1.6113 0.1072 Dummy for (DIST13)=0*_yhat_ DIST6_0 -0.3489 0.111 -3.144 0.0017 Dummy for (DIST6)=0 HHENT_0*_yhat_*_yhat_ -0.0011 0.0008 -1.4623 0.1437 Dummy for (HHENT)=0*_yhat_*_yhat_ HHSIZE2*_yhat_*_yhat_ 0.0001 0 4.3189 0 hhsize2*_yhat_*_yhat_ SICKLE_0 -0.1132 0.0675 -1.6781 0.0934 Dummy for (SICKLE)=0 SST=23261.8921 SSR=504.3870 F=10.0822 R2=0.0217 adjR2=0.0195 50 Urban Region Coefficient Std. Err. t |Prob|>t Label _intercept_ -10.429 2.069407 -5.0396 5.26E-07 Intercept _yhat_ 1.118614 0.225286 4.965315 7.69E-07 _yhat_ CHAIR_0*_yhat_*_yhat_ 2.29E-02 7.70E-03 2.979338 2.94E-03 Dummy for (CHAIR)=0*_yhat_*_yhat_ DIST24_0*_yhat_*_yhat_ -5.99E-02 0.03662 -1.63574 0.102115 Dummy for (DIST24)=0*_yhat_*_yhat_ DIST24_1 -2.59E-03 1.24E-03 -2.08198 3.75E-02 Dummy for (DIST24)=1 DIST31_0 -8.25E-03 2.60E-03 -3.16621 1.58E-03 Dummy for (DIST31)=0 DIST31_0*_yhat_*_yhat_ 8.589176 2.845808 3.018185 2.59E-03 Dummy for (DIST31)=0*_yhat_*_yhat_ DIST7_0*_yhat_ 3.06E-03 0.001549 1.975045 0.048456 Dummy for (DIST7)=0*_yhat_ DIST7_1 -0.77712 0.276856 -2.80694 5.07E-03 Dummy for (DIST7)=1 FIREWOOD_0*_yhat_*_yhat_ 2.649163 1.131794 2.340676 1.94E-02 Dummy for (FIREWOOD)=0*_yhat_*_yhat_ GENDER_H_0*_yhat_*_yhat_ -4.88854 1.557569 -3.13857 1.73E-03 Dummy for (GENDER_H)=0*_yhat_*_yhat_ HHSIZE*_yhat_*_yhat_ -4.07E-02 0.014629 -2.78526 5.42E-03 hhsize*_yhat_*_yhat_ HHSIZE2 -1.94E-02 1.02E-02 -1.89998 5.76E-02 hhsize2 PROPEMP*_yhat_*_yhat_ -5.16E-03 2.33E-03 -2.21127 2.72E-02 propemp*_yhat_*_yhat_ REFRI_0*_yhat_ 0.304808 0.183286 1.663024 9.65E-02 Dummy for (REFRI)=0*_yhat_ REFRI_0*_yhat_*_yhat_ -2.20E-03 9.56E-04 -2.29942 2.16E-02 Dummy for (REFRI)=0*_yhat_*_yhat_ SICKLE_1 0.651765 0.42185 1.545015 0.122565 Dummy for (SICKLE)=1 SST=7329.0933 SSR=348.4514 F=4.4395 R2=0.0475 adjR2=0.0368 51 52 ANNEX 1F: MALAWI'S PROGRESS TOWARDS THE MILLENNIUM DEVELOPMENT GOALS (AS OF 2005) Malawi is one of the 189 member countries that signed the Millennium Declaration adopted at the United Nations (UN) General Assembly in New York in September 2000. The Declaration outlines eight (8) goals and eighteen (18) measurable targets for enabling human beings to enjoy the minimum requirements of dignified life by 2015.7 The Millennium Development Goals (MDGs) commit countries to an expanded vision of development that promotes human development as key to sustaining social and economic progress, and recognizing the importance of creating a global partnership. The MDGs represent a state's obligation towards every individual regardless of their circumstances, and are therefore also human rights obligations. The eight Goals set by the summit are as follows: 1. Eradicate extreme poverty and hunger, 2. Achieve universal primary education, 3. Promote gender equality and empowerment of women, 4. Reduce child mortality, 5. Improve maternal health, 6. Combat HIV/AIDS, malaria and other diseases 7. Ensure environmental sustainability, and 8. Organize a global partnership for development. In Malawi, the MDGs are to be achieved through implementation of the Malawi Poverty Reduction Strategy (MPRS, 2002-2006) and the Malawi Growth and Development Strategy (MGDS, 2006-2010), which express the country's overarching poverty reduction strategies. The overall monitoring of the MDGs is expected to be aligned under the Monitoring & Evaluation Master Plan launched in November 2004, which lays the basis for the monitoring of the MPRS and MGDS. Malawi completed the first report on the progress towards the MDGs in 2003 (Government of Malawi and UNDP, 2003). The report highlighted that Malawi was falling short in a number of ways towards reducing poverty and advancing other human developments. Taking advantage of the availability of new data from the 2004 Malawi Demographic and Health Survey (MDHS 2004) and the 2005 Integrated Household Survey (2005 IHS), this note updates the information on Malawi's progress towards meeting the MDGs. This note is not comprehensive, in that it does not attempt to analyze the progress on every indicator. Only those indicators for which data is readily available have been covered. To this effect, only a few indicators will be discussed under goals number six and seven, while goal number eight will not be discussed at all. The report discusses progress towards the individual MDG targets and indicators. For each indicator, the report presents the most recent available data, as well as comparable 7Achievement of the targets will be monitored through a set of 48 indicators. The full list of goals, targets and indicators is available on the internet: http://www.un.org/millenniumgoals 53 earlier measurements, and provides an assessment of the progress achieved in recent years. In addition, for each indicator, the report presents a linear projection forward to the year 2015. The linear projections are derived by calculating the annual rate change using the two most recent available data points, and assuming that the rate of change will stay the same up to the year 2015. Clearly s linear projection is not an adequate methodology to forecast future evolution of these trends. However, given that the data points are so few, it would not be possible to model causation and verify probable non- linearities. It should be emphasized therefore that such linear projections do not provide a forecast of whether Malawi will achieve the MDGs. The projections presented in this report only provide an indication of whether, given recent trends, progress towards each of the MDGs has been satisfactory, and the country is `on-track' to meet the targets. For those MDGs where recent progress has been insufficient, the report highlights the need for the government to revise its policies and accelerate progress. Achieving each MDGs by 2015 will depend on the policies government will pursue to accelerate progress. In order to facilitate the presentation of the findings, for each indicator the report includes a graph which summarizes the progress made and the current trends. Each graph shows the changes between 1990 and 2005 (a dark blue line), and projections to 2015 (a light blue line). It also shows the path towards reaching the MDGs, taking into account the situation in 1990, and the goals that have been set for 2015 (a red dotted line). Malawi's Progress towards the Millennium Development Goals A summary of Malawi's progress towards the MDGs as of end-2005 is provided in Table 1. A complete list of progress towards each goal, target, and indicator is presented in Annex A. Malawi is unlikely to meet Goal 1 (to eradicate extreme poverty and hunger). Little progress has been made in reducing poverty and inequality over the past decade. Similarly, virtually no progress has been made in reducing ultra-poverty, indicating that the share of people who are unable to meet their minimum daily dietary requirements has not decreased. In fact, malnutrition rates remain very high; although the prevalence of underweight children under five appears to be decreasing, the levels of stunting remain some of the highest in the world. Good progress has been made towards Goal 2 (to achieve universal of primary education); nevertheless more progress is required to meet this MDG by 2015. While Malawi has improved its GER ratios in primary education, the NER has increased slowly and remains at around 80 percent, suggesting that Malawi will fall short of achieving universal primary education by the year 2015. According to the 2004 MDHS about 86 percent of children who start primary school reach grade 5. This is a substantial improvement from a completion rate of 79 percent in the year 2000. Malawi is well placed to reach Goal 3 (to promote gender equality and empower women). Notably, good progress has been made in reaching equality of enrollment in primary and secondary education and in reducing gender disparity in youth literacy. 54 More progress is needed in reducing the gender gap in tertiary education, however, and also in increasing women participation in the workforce and in position of authority. Good progress has also been made towards achieving Goal 4 (to reduce child mortality). While under-five mortality and infant mortality rates remain very high, they have been decreasing steadily over time and are projected to decrease by more than two-thirds, between 1990 and 2015. On a negative note, however, child immunization rates have worsened over the same period. Little progress has been made in achieving Goal 5 (to improve maternal health). Maternal mortality actually increased between 1992 and 2000, but has recently started to decrease. The current rate of reduction is not rapid enough to meet this MDG by 2015, however. Similarly the proportion of births attended by skilled personnel remains just above 50 percent, and has not improved significantly since 1990. Some progress has been made in reaching Goal 6 (to combat HIV/AIDS, malaria and other diseases). Notably, Malawi appears to have halted the spread of HIV/AIDS, as adult prevalence rates appear to have stabilized at around 14 percent. Accurate knowledge of the HIV/AIDS transmission remains weak, however, and only limited progress has been made in reducing risky behaviors. Good progress has been made in increasing the proportion of population using effective malaria prevention. Finally, mixed progress has been made in achieving Goal 7 (Ensure environmental sustainability). Deforestation is continuing at an alarming rate, and the proportion of population using solid fuels remains very high. On a more positive note, the proportion of people who have access to safe drinking water and improved sanitation appears to have increased significantly. In sum, Malawi appears well placed to achieve three of the MDGs by the year 2015, provided additional progress is made. Achieving the other MDGs looks unlikely unless substantial reforms are introduced to improve performance. This report does not discuss the likely impact of recent reforms, and/or the need for additional interventions. 55 Table 1: Summary of Malawi's progress towards the MDGs as of end-2005 Most MDG Target Baseline Intermediate Recent Target feasibility 1990-1992 1998-2000 2004-2005 for 2015 Goal 1: Eradicate extreme poverty and hunger Target 1: Halve, between 1990 and 2015, the proportion of people under the poverty line (Indicator 1) 54.0 53.9 52.4 27.0 Unlikely Target 2: Halve, between 1990 and 2015, the proportion of people who suffer from hunger (under the ultra-poverty line, Indicator 5) - 23.6 22.4 11.8 Goal 2: Achieve universal primary education Target 3: Ensure that, by 2015, children everywhere, boys and girls alike, will be able to complete a full course of primary schooling (Net enrollment ratio in primary education, Indicator 6) 58.0 78.2 82.0 100.0 Unlikely Target 3: Ensure that, by 2015, children everywhere, boys and girls alike, will be able to complete a full course of primary schooling (completion rate, Indicator 7) 64.4 79.4 85.9 100.0 Goal 3: Promote gender equality and empower women Potentially Target 4: Eliminate gender disparity in primary and secondary Feasible education, preferably by 2005, and to all levels of education (Gender ratio in primary, Indicator 9a) 0.87 0.92 0.95 1.0 Goal 4: Reduce child mortality Potentially Target 5: Reduce by two-thirds, between 1990 and 2015, the under- Feasible five mortality rate (Indicator 13) 234.0 189.0 133.0 78.0 Goal 5: Improve maternal health Target 6: Reduce by three-quarters, between 1990 and 2015, the maternal mortality ratio (Maternal mortality ration, Indicator 16) 620.0 1120.0 984.0 155.0 Unlikely Target 6: Reduce by three-quarters, between 1990 and 2015, the maternal mortality ratio (Proportion of birth attended by skilled health personnel, Indicator 17) 54.8 55.6 57.0 88.7 Goal 6: Combat HIV/AIDS, malaria and other diseases Target 7: Have halted by 2015 and begun to reverse the spread of Potentially HIV/AIDS (HIV prevalence among pregnant women, Indicator 18) 17.4 24.1 16.9 <17.4 Feasible Target 8: Have halted by 2015 and begun to reverse the incidence of malaria and other major diseases (Proportion of population in malaria risk areas using effective malaria prevention (percent of under five children using bednets, Indicator 22a) - 7.6 20.2 >7.6 Goal 7: Ensure environmental sustainability Target 9: Integrate principles of sustainable development into country policies and programs and reverse loss of environmental Unlikely resources (Proportion of forested land area, Indicator 25) 41.4 37.9 36.2 <41.4 Target 10: Halve, by 2015, the proportion of people without sustainable access to safe drinking water (Indicator 30) 72.3 81.4 83.9 86.2 56 GOAL 1: ERADICATING EXTREME POVERTY There are two main targets under this goal. The first target is to halve, between 1990 and 2015, the proportion of people whose income is less than one dollar a day. Three indicators were selected to monitor progress towards this target: the proportion of population below the national poverty line;8 the poverty gap ratio; and finally the share of the poorest quintile in national consumption. The second target set is to halve, between 1990 and 2015, the proportion of people who suffer from hunger. Two indicators have been selected: the prevalence of underweight children under five years of age; and the proportion of population below the minimum level of dietary energy consumption. Target 1: Halve, between 1990 and 2015, the proportion of people whose income is less than one dollar per day Indicator 1: Proportion of people below poverty line (Poverty Head Count) The first indicator under Target 1 of the MDGs is the proportion of population below the poverty line, as measured by the poverty headcount. In Malawi, the 1991 Household Expenditure and Small-Scale Economic Activity Survey (HESSEA) constitutes the first national household survey from which a national poverty estimate has been derived. The headcount poverty rate estimated from the 1991 HESSEA is 54 percent. The 1998 Integrated Household Survey (IHS) revealed a poverty headcount rate of 53.9 percent.9 Unfortunately, the IHS poverty measure is not comparable to the HESSEA, due to differences in survey methodologies. In this report, however, we ignore the issue on comparability.10 More recently, the second IHS revealed that about 52.4 percent of the population was living below the poverty line in 2005. The estimate from the 2005 IHS2 is comparable to the estimate derived from the 1998 IHS1. As indicated above, the target against this indicator is to reduce by half poverty level by 2015, from the 1990 poverty levels. Given a poverty rate of about 54 percent in 1990, the target implies reducing the headcount poverty rate to about 27 percent by the year 2015. The last two integrated household survey indicate that poverty in Malawi is declining at an average rate of 0.21 percentage points per year. Projecting this trend into the future, poverty headcount will be at 50.3 percent in 2015, implying that about half of the population in Malawi would still be living in poverty. The lack of progress in reducing poverty between 1990 and 2005 suggests that Malawi is unlikely to meet Target 1 of reducing by half the proportion of people living below poverty line by the year 2015. 8In this report we measure the share of population below the national poverty line. Ideally this indicator should be monitored using the $1 per day poverty line, however. 9The poverty rate from the 1998 IHS had originally been calculated at 65.3 percent (NEC, 2001). The rate has been recalculated at 53.9 percent in 2006 using a new methodology to ensure comparability with the results of the 2005 IHS2 (Government of Malawi and World Bank, 2006). 10In practice this is equivalent to assuming that the poverty headcount has not changed significantly between 1990 and 1998, and was stable around 54 percent. This assumption is consistent with the (lack of) changes in GDP per capita during the period. 57 Figure 1 depicts the changes in poverty rates between 1990 and 2005, and the linearly predicted poverty rate by the year 2015. The dotted line presents the path towards reaching the MDG goal, which shows the trend required if the country was to half the 1990 poverty rate. Figure 1: Poverty Headcount Poverty headcount 65 60 54.0% 53.9% 52.4% 55 50.3% 50 entc 45 40 per current path 35 Linearly projected value 30 27.0% 25 MDG Target 20 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1991 HESSEA, 1998, 2005 IHS Note: The poverty headcount estimate for 1991 (from the HESSEA) is not strictly comparable with the estimates for 1998 IHS and 2005 IHS. Only the latter are directly comparable. Indicator 2: Incidence of depth of poverty as measured by the poverty gap The second indicator under Target 1 is the extent of the poverty gap ratio. The poverty gap ratio is the average distance separating the poor from the poverty line, expressed as a percentage of the poverty line. Hence, the poverty gap accounts not only the number of poor, but it also considers how poor the people are. The 1991 HESSEA survey revealed a poverty gap of 16 percent. Although the two surveys are not directly comparable, the levels is not very different from the 18.6% estimated using the 1998 Integrated Household Survey. After seven years, the poverty gap was estimated at 17.8 percent using the 2005 Integrated Household Survey. The poverty gap ratio, therefore, has been declining at 0.11 percentage points per annum between 1998 and 2005. All other things being the same, if this rate of change continues, the poverty gap will have declined to around 16.7 percent by the year 2015. Hence, it looks unlikely that Malawi will meet the indicator of reducing by half the poverty gap ratio by the year 2015. Figure 2 depicts the trend in poverty gap ratio between 1998 and 2005. The dotted line depicts a linear projected line if poverty gap was to fall by half in 2015 from the 1991 level and achieve the MDG.11 11As discussed above, the poverty gap calculated using the HESSEA is not directly comparable with more recent estimates from the 1998 and 2005 Integrated Household Surveys. Hence the target of 8 percent by 58 Figure 2: Poverty gap (or depth of poverty) Poverty gap 30 25 18.6% 17.8% tne 20 16.0% 16.7% rc 15 pe 10 current path 8.0% Linearly projected value 5 MDG Target 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1991 HESSEA, 1998, 2005 IHS Note: The poverty gap estimate for 1991 (from the HESSEA) is not strictly comparable with the estimates for 1998 IHS and 2005 IHS. Only the latter are directly comparable. Indicator 3: Share of poorest quintile in national consumption. The third indicator under Target 1 is the share of poorest quintile in national consumption. Like the other two indicators above, the main sources of this indicator are household surveys. The target under this indicator is to double the share of the poorest quintile in the national consumption. Figure 3 illustrates the current and projected trend in the poorest quintile share in national consumption. Figure 3: Poorest quintile's share in national consumption Poorest quintile's share in total consumption 40 35 current path 30 Linearly projected value 25 MDG Target 20.0% ent rcep20 15 10.0% 10 5 10.1% 10.2% 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1998 and 2005 IHS. 2015 constitutes only an approximation. If we assumed that the poverty gap had not changed between 1991 and 1998, and was stable at 18.6 percent, then the target by 2015 would around 9.3 percent. 59 As indicated in the graph, the share of the poorest quintile in Malawi has not improved between 1998 and 2005. The 1998 Integrated Household Survey revealed that the poorest 20 percent of the population controlled only 10 percent share of national consumption. After 7 years, the share has remained around 10 percent implying that inequality is not decreasing. At this rate the indicator will not be achieved by 2015. Target 2: Halve, between 1990 and 2015, the proportion of people who suffer from hunger The second target under goal number one is to eradicate extreme poverty and hunger. As mentioned above, there are two indicators under this target. The current status and trends on these indicators is discussed below. Indicator 4: Prevalence of underweight children (under five years of age) The nutritional well-being of young children reflects the household's, community and national investments in family health, and contributes to the country's development both directly and indirectly. Prevalence of underweight children is taken as a proxy indicator of the proportion of population that is undernourished. This is the proportion of persons whose food intake falls below the minimum requirement or food intake that is insufficient to meet dietary energy requirements continuously. They are the people who suffer from hunger. Figure 4 shows the trend in the nutrition status of children under five years of age since the early 1990's. Figure 4: Prevalence of Underweight in under-five children Underweight prevalence 40 35 28.0% 30 25.4% tne 25 22.2% rc 20 current path 13.4% pe15 Linearly projected value 10 14.0% MDG Target 5 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, and 2004 MDHS The figure shows that prevalence of underweight in children 5 years of age has decreased steadily from 28 percent in 1992, to 25 percent in 2000, and 22 percent in 2004. If progress continues at the same rate, the underweight rate in children under five years of age will be just over 13 percent in 2015. The MDG target requires prevalence to be 60 reduced to 14 percent by 2015 implying that the country is on track to achieve the target of a reduction by half in the prevalence of underweight children by the year 2015. Indicator 5: Proportion of population below minimum level of dietary energy consumption The 1998 and 2005 poverty profiles of Malawi estimate the proportion of population below minimum dietary energy requirement by defining a food poverty line. All persons below this line were deemed ultra-poor. In 1998 about 23.6 percent of the population was deemed ultra-poor. After five years, the proportion of ultra-poor persons has stayed more or less constant at 22.4 percent. At this rate, the projected proportion of persons deemed ultra-poor will be about 20.7 percent by the year 2015. Therefore, based on the current trends, Malawi is not on track to reduce by half the proportion of persons below minimum level of dietary energy consumption by the year 2015. Figure 5: Proportion of population ultra-poor Proportion of ultra-poor 40 35 30 23.6% 22.4% tne 25 20.7% 20 perc15 current path 10 Linearly projected value 11.8% 5 MDG Target 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1998 and 2005 IHS 61 GOAL 2: ACHIEVE UNIVERSAL PRIMARY EDUCATION There is only one target under this goal, which is to ensure that by 2015, children everywhere, boys and girls alike, will be able to complete a full course of primary schooling.12 In order to monitor progress against this target, three indicators have been selected namely, net enrolment ratio in primary education, proportion of pupils starting grade 1 who reach grade 5, and literacy rate of 15-24 year olds. Target 3: Ensure that, by 2015, children everywhere, boys and girls alike, will be able to complete a full course of primary schooling. Indicator 6: Primary school net enrolment ratio Primary school net enrollment increased from 58 percent in 1992 to 78 percent in 2000, and 82 percent in 2004 (Figure 6). Assuming the recent rate of change continues, net enrolment rate will reach around 90 percent by 2015. Hence, while substantial progress is under way, Malawi is not on track to achieve universal primary enrolment by 2015. Figure 6: Primary school net enrolment Net enrollment in primary education 100 95 100% 90 82.0% 85 92.5% tne 78.2% 80 rc 75 pe 70 current path 65 58.0% 60 Linearly projected value 55 MDG Target 50 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, and 2004 MDHS Indicator 7: Proportion of pupils starting grade 1 who reach grade 5 This indicator measures the extent of dropout in primary education. The 2000 MDHS results indicate that 79 percent of pupils starting grade 1 were expected to reach grade 5. This proportion increased to 86 percent in the 2004 MDHS. Assuming the same rate of improvement in completion rates continues in the future, by 2015 all the pupils starting primary education would be expected to reach grade 5 (Figure 7).13 12Defined as the first four years of primary schooling (i.e. completing grades 1 to 4 inclusive). 13It should be noted that there are conflicting data on completion rates in Malawi. The MDHS EdData Education Profile reports a 69 percent completion rate in the year 2000 and a 60 percent in 2002 (ORC MACRO, 2004). The deterioration is largely due to the increase in drop out rates. 62 Figure 7: Proportion of pupils starting grade 1 reaching grade 5 Proportion of pupils reaching grade 5 100% 100 95 90 85.9% 85 79.4% 80 ent rcep 75 70 current path 65 60 64.4% Linearly projected value 55 MDG Target 50 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, and 2004 MDHS Indicator 8: Literacy rate of 15-24 year olds. The Education for All (EFA) initiative anticipated reaching 100 percent literacy for the 15-24 year-old population by the year 2015. Unfortunately, Malawi has not been able to make much progress towards this target to-date. Youth literacy increased from 68 percent in 1998 to about 71 percent in 2004.14 At this rate of improvement, youth literacy would reach 78 percent by 2015 (Figure 8). Figure 8: Youth literacy rate Youth (15-24 year olds) literacy rate 100 95 current path 100% 90 85 Linearly projected value tne 80 MDG Target rc 75 68.1% 77.5% pe 70 65 70.6% 60 55 50 1990-92 1995 1998-00 2004-05 2010 2015 Source: 2000 and 2004 MDHS 14Male youths literacy rate was 76 percent in 2004, compared to 65 percent for their female counterparts. 63 GOAL 3: PROMOTING GENDER EQUALITY AND EMPOWER WOMEN The third Millennium Development Goal is to promote gender equality and empower women. There is only one target under this goal and this is to eliminate gender disparity in primary and secondary education, preferably by 2005, and at all levels of education no later than 2015. Four indicators have been agreed to monitor this target, namely the ratio of girls to boys in primary and secondary schools; the ratio of literate females to males 15-24 years old; the share of women in wage employment in non-agricultural sector; and finally the proportion of seats held by women in the national parliament. Target 4: Eliminate gender disparities in primary and secondary education, preferably by 2005, and to all levels of education no later than 2015. Indicator 9a: Ratio of girls to boys in primary school The ratio of girls to boys is a measure of gender equity in the education sector. It is calculated by looking at the proportion of females to males enrolled in school. According to the 1992, 2000 and 2004 MDHS, the ratio of girls to boys in primary education has been steadily rising. The ratio was at 0.87 in 1992 and has increased to about 0.95 by 2004. At this rate of increase, by the year 2015 there would be no gender disparities in primary education. Figure 9a shows that Malawi is on track to achieve gender equality in primary education by 2015. Figure 9a: Ratio of Girls to Boys in Primary School Ratio of girls to boys in primary school 1.5 1.4 current path 1.3 1.2 Linearly projected value oit 1.1 MDG Target 1.0 Ra0.9 1 0.95 0.8 0.92 0.87 0.7 0.6 0.5 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, 2004 MDHS 64 Indicator 9b: Ratio of girls to boys in secondary education Although there are good prospects of achieving gender equality in primary education by 2015, it remains important to examine gender disparities at higher levels of education (i.e. whether girls are able to continue their post-primary education). According to the Demographic and Health Surveys, the ratio of girls to boys in secondary education has been increasing over the years. In 1992, the ratio was 0.5 implying there was one female to every two male students in secondary education. This ratio increased to 0.6 in 2000 and has further risen to 0.75 in 2004. If this rate of change continues, the ratio of girls to boys in secondary education would reach equality before 2015, implying that Malawi is on track to meet this target (Figure 9b). Figure 9b: Ratio of Girls to Boys in Secondary School Ratio of girls to boys in secondary school 1.5 1.4 current path 1.3 1.2 Linearly projected value oit 1.1 MDG Target 1 1.0 Ra0.9 1.00 0.8 0.7 0.6 0.50 0.75 0.5 0.60 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, 2004 MDHS Indicator 10: Ratio of literate females to males 15-24 years old In order to achieve gender equality, the disparities between girls and boys in all spheres of education need to be eliminated. The 2004 Demographic and Health Survey revealed that the ratio of literate females to males of age 15-24 was at 0.86 implying that there were about six female youths who were literate for every seven male youths.15 This rate has been improving over time, from 0.82 in the year 2000 (to 0.86 in 2004). Assuming the same rate of improvement continues, the literacy rate ratio between male and female youths is projected to reach about 0.97 by 2015 (Figure 10). Hence, unless additional action is undertaken, Malawi is unlikely to completely eliminate gender disparity in youth literacy by 2015. 15Considering that the ratio of male-youths against their female counterparts in the age group 15-24 year olds is about 48:52 then, in absolute terms, there is even less female youths that are literate. 65 Figure 10: Ratio of literate females to males 15-24 years old Ratio of literate females to males 15-24 years 1.5 1.4 current path 1.3 1.2 Linearly projected value oit 1.1 MDG Target 1 1.0 Ra0.9 0.97 0.8 0.86 0.7 0.82 0.6 0.5 1990-92 1995 1998-00 2004-05 2010 2015 Source: 2000 and 2004 MDHS Indicator 11: Share of women in wage employment in non- agricultural sector. It is expected that given equal opportunities, there should be equal proportions of men and women, in the labor sector especially in the professional jobs. In practice, women often have lower participation in the labor force, especially in professional jobs. One reason for the lower participation in wage employment among women could be that very few women are educated and qualified, and that many forms of wage employment normally require a minimum of skills. The 1998 Integrated Household Survey indicates that approximately 8 percent of individuals aged 10 and older were involved in wage work at that time.16 Women made up 25 percent of all non-agricultural workers, and 20 percent of all wageworkers. The share of women in wage employment in the non-agricultural sector was 13 percent. The 2005 Integrated Household Survey, using a slightly different definition of employment (in line with international standards), indicates that only approximately 6 percent of the individuals 10 years and older were involved in wage work in 2005.17 Women made up 40 percent of all non-agricultural workers, and 22 percent of all wageworkers. The share of women in wage employment in the non-agricultural sector was 12.5 percent overall. Figure11 depicts the current status and trends in the share of women in wage employment in the non-agricultural sector as well as projections to the year 2015. While the two numbers for 1998 and 2005 are not strictly comparable due to the changes in definition, it seems clear that women occupy a very small share of wage employment in the non-agricultural sector, and that there appears to have been little improvement in recent years. Hence, the target of having an equal ratio of men and women in wage employment by the year 2015 appears unlikely. 16Defined as those who report that their activity status was employee or employer. 17Defined as those who report having worked between 1 and 60 hours for a wage or salary in the last 7 days 66 Figure 11: Share of women in wage employment in the non-agricultural sector Share of women in wage employment in non-agricultural sector 100.0 90.0 current path 80.0 70.0 Linearly projected value oit 60.0 MDG Target 50.0% 50.0 Ra 40.0 30.0 13.1% 20.0 12.5% 11.6% 10.0 0.0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1998 and 2005 IHS Indicator 12: Proportion of seats held by women in the national parliament The fourth indicator under Target 4 of the Millennium Development Goals is to reach an equal proportion of seats held by men and women in the national assembly. According to the Malawi Electoral Commission, out of 193 elected members of parliament in the 1999 Malawi general elections, only 18 members of parliament were women, corresponding to about 9 percent. In the 2004 presidential and parliamentary elections, there were 193 MPs and only 27 of them were women, representing around 14 percent (Figure 12). Figure 12: Proportion of seats held by women in parliament Proportion of seats held by women in parliament 100.0 90.0 current path 80.0 70.0 Linearly projected value oit 60.0 MDG Target 50% 50.0 Ra 40.0 30.0 25.3% 20.0 5.6% 9.3% 14.0% 10.0 0.0 1994 1999 2004 2010 2015 Source: Malawi Electoral Commission According to these results, there has been only a modest increase in the number of women in decision-making positions. In fact, Malawi remains well below the Southern African Development Community (SADC) target of 30 percent share of women in 67 parliament by 2005. Further, at current trends, the share of women in parliament is projected to reach about 25 percent by 2015, indicating that Malawi is not on track to reach the target of having a 50 percent share of women in the national assembly (Figure 12). 68 GOAL 4: REDUCING CHILD MORTALITY The fourth Millennium Development Goal seeks to reduce child mortality. Under this goal, the target is to reduce the under-five mortality rate by two-thirds, between 1990 and 2015. Three main indicators will help monitor progress towards achieving this goal. The indicators are under-five mortality rate; infant mortality rate; and finally, the proportion of 1-year-old children immunized against measles. Target 5: Reduce under-five mortality rate by two-thirds, between 1990 and 2015 Indicator 13: Under-five mortality rate Malawi has made good progress towards reducing under-five mortality (Figure 13). The country recorded an under-five mortality rate of 234 deaths per 1000 live births in 1992. This rate declined to 189 in the year 2000 and has further declined to 133 in 2004. At this rate, under-five mortality would decline to around 21 deaths per 1000 live births by the year 2015, implying that the country may be able to reduce under-five mortality by more than two-thirds of its 1992 rate. Figure 13: Under Five Mortality Rate Under-five mortality rate 300 hstrbi 234 250 evil 189 200 00 133 10rep 150 100 current path 78 sh Linearly projected value ateD 50 MDG Target 21 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, and 2004 MDHS Indicator 14: Infant mortality rate According to the 1992, 2000 and 2004 Malawi Demographic and Health Surveys, infant mortality rate has been declining in Malawi from a very high 134 in 1992 to 76 in 2004. Figure 14 shows infant mortality rates for Malawi since early 1990s. At this rate of improvement, Malawi would almost eliminate infant mortality by 2015. This means that the country is on track in reducing the country's infant mortality rate by two-thirds of its 1992 rate by the year 2015. 69 Figure 14: Infant Mortality Rate Infant mortality rate 200 shrtib 175 134 evil 150 125 103 00 10r 100 76 pe 75 current path 47.5 hstaeD 50 Linearly projected value 25 MDG Target 1.8 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, and 2004 MDHS Indicator 15: Proportion of 1-year-old children immunized against measles The proportion of children immunized against measles was 86 percent in 1990, but has since declined reaching 79 percent by 2004. If it continues to deteriorate at the same rate, Malawi may have a 66 percent immunization rate for 1-year-old children by the year 2015. Figure 15 shows the changes in the percentage of 1-year-old children immunized against measles, and the projections to 2015. The dotted line depicts the path to reaching this MDG target of improving the rate of immunization of 1-year-old children against measles by two-thirds by the year 2015. While it is unlikely that progress will be as rapid as indicated by this simple linear projection, nevertheless recent improvements suggest that substantial progress is being made to reduce infant mortality in Malawi. Figure 15: Proportion of 1-year-old children immunized against measles Proportion of 1-year-old children immunized against measles shrtib 100 90 80 95.3% evil 70 85.8% 83.2% 78.7% 60 00 66.3% 10r 50 current path 40 pe 30 Linearly projected value hstaeD 20 MDG Target 10 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, and 2004 MDHS 70 GOAL 5: IMPROVING MATERNAL HEALTH The fifth goal under the Millennium Development Goals is to improve maternal health. Under this goal there is only one target which is to reduce by three-quarters, between 1990 and 2015, the maternal mortality ratio. Two indicators were set to monitor the progress towards achieving this target. These indicators are the maternal mortality ratio and the proportion of births attended by skilled health personnel. The following sections will discuss the current status, trend and the projections of these indicators against the MDG target. Target 6: Reduce by three quarters, between 1990 and 2015, the maternal mortality ratio Indicator 16: Maternal mortality ratio According to the World Health Organization, UNICEF and UNFPA 2000 estimates, Malawi is one of the countries with the highest maternal mortality rate in the world. The Demographic and Health Surveys estimate that the maternal mortality ratio in Malawi increased sharply from 620 deaths per 100,000 live births in 1992, to 1120 deaths per 100,000 live births in 2000. The 2004 MDHS indicates that maternal mortality has declined to 984 deaths per 100,000 live births. If the recent rate of improvement is maintained, Malawi would have a maternal mortality ratio of about 610 deaths per 100,000 live births by the year 2015. This is roughly the same level as in 1992. Under the MDG Target 6, Malawi is expected to have a maternal mortality ratio of about 155 per 100,000 live births by the year 2015. Hence, unless additional measures are put in place, this target is unlikely to be met by 2015. Figure 16 illustrates the recent trends in maternal mortality and the projected trend to 2015, as well as the desired path towards reaching the MDG. Figure 16: Maternal mortality ratio Maternal Mortality Ratio 1400 hstrbi 1120 1200 evil 984 1000 000 800 620 610 0, 10r 600 pe 400 hst 155 200 eaD 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, and 2004 MDHS 71 Indicator 17: Proportion of births attended by skilled health personnel The proportion of births attended by skilled health personnel is a measure of the health system's ability to provide adequate care for pregnant women. It combines information on the presence of skilled health personnel and the accessibility of facilities with skilled health personnel. This indicator aims to reduce the proportion of births not attended by skilled personnel by three-quarters between 1990 and 2015. According to the Demographic and Health Surveys, the proportion of births attended by skilled health personnel in Malawi was 55 percent in 1992, 56 percent in 2000, and 57 percent in 2005. At this rate of change, the proportion of births attended by skilled health personnel in 2015 will be about 63 percent. Hence Malawi is not on track to reach the target of having 100 percent of births attended by skilled health personnel by 2015. Figure 17 shows the recent trends and the projected changes in the share of births attended by skilled health personnel. The dotted line indicates the desired path towards meeting the MDG target. Figure 17: Proportion of births attended by skilled health personnel Proportion of births attended by skilled health personnel 100 90 80 88.7% 70 tnecr 60 50 54.8% 55.6% 57.0% 60.9% Pe40 current path 30 20 Linearly projected value 10 MDG Target 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, and 2004 MDHS 72 GOAL 6: COMBAT HIV/AIDS, MALARIA AND OTHER DISEASES Two targets have been set to achieve this goal: Target 7 aims to halt and begin to reverse the spread of HIV/AIDS by the year 2015, while Target 8 aims to halt and begin to reverse the incidence of malaria and other major diseases by 2015. Target 7: Halt by 2015 and begin to reverse the spread of HIV/AIDS Three indicators have been selected to monitor progress towards Target 7: the HIV prevalence among 15-24-year-old pregnant women; the percentage of current users of contraception who are currently using condoms (currently married women 15-49 years old); and the ratio of school attendance of orphans compared to non-orphans aged 10-14. Indicator 18: HIV prevalence among pregnant women18 Malawi is making progress towards reducing the spread of HIV (the virus that causes AIDS). According to the National AIDS Commission Sentinelle Surveillance report, HIV prevalence amongst pregnant women was at 17 percent in 1994. The prevalence rate increased to 24 percent by 1998, but it has now started to decline reaching 17 percent as of 2004. If the recent trend were to continue, over the next decade the prevalence rate would decrease to around 5 percent by the year 2015 (Figure 18). It should be emphasized that projecting the HIV prevalence rate to come down to 5 percent by 2015 does not seem very realistic, as we are talking about an epidemic and about largely unpredictable issues like behavior change. While it is unlikely that progress will be as rapid as indicated by this simple linear projection, nevertheless recent improvements suggest that some progress is being made to reverse the spread of HIV/AIDS in Malawi. Figure 18: HIV prevalence among pregnant women HIV prevalence among pregnant women at sentinel sites 50 45 current path 40 Linearly projected value 35 tnecr MDG Target 30 24.1% 25 17.4% 16.9% 17.4% Pe20 15 10 4.9% 5 0 1990 1994 1999 2005 2010 2015 Source: NAC Sentinelle Surveillance Report, 1994, 1999, and 2005 Note: Indicator should focus on women age 15-24; however data by age groups is not readily available. 18The official indicator focuses on pregnant women aged 15 to 24. However, data for this age group was not readily available for Malawi. 73 Indicator 19: Percentage of current users of contraception who are using condoms (currently married women 15-49) The use of condoms is not only seen as a family planning mechanism rather it also prevents the spread of HIV. Like many countries, Malawi has been on a campaign to promote the use of condoms to halt the rapid population growth as well the spread of the deadly HIV virus. It should be pointed out that there is no particular MDG target for this indicator, but it is expected that the trends would improve over time. The 1992 MDHS indicates that about 12 percent of users of contraception among currently married women aged 15-49 reported to be using condoms. This number decreased to 5.2 percent by the year 2000, mainly as a result of the increase in the overall number of current users of other methods of contraception. More recently, the 2004 MDHS indicates that the percentage of current users of contraception using condoms has increased slightly to 5.5 percent. At this modest rate of improvement, it is anticipated that by 2015, the proportion of currently married women aged 15-49 that are using condoms will increase to about 6.4, which is a modest improvement (Figure 19). Figure 19: Percentage of current users of contraceptive who are using condoms (currently married women 15-49) Condom use to overall contraceptive use among currently married women aged 15-49 50 45 40 Current path 35 tnecr Linearly projected value 30 25 Pe20 12.3% 15 10 5.2% 5.5% 6.4% 5 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000 and 2004 MDHS Indicator 20: Ratio of school attendance of orphans to school attendance of non- orphans aged 10-14 The indicator is used to assess whether orphans are attending school, particularly in primary education. The 2000 and 2004 MDHS have found that the ratio of school attendance of orphans to school attendance of non-orphans aged 10-14 has remained constant at unity. Like the previous indicator, there is no particular MDG target for this indicator. Figure 20 shows a constant trend in the ratio since the year 2000. 74 Figure 20: Ratio of school attendance of orphans to school attendance of non- orphans aged 10-14 Ratio of school attendance of orphans to school attendance of non-orphans aged 10-14 1.5 1.4 1.3 1.2 1.1 1 1 1 ent rceP1.0 0.9 0.8 Current path 0.7 Linearly projected value 0.6 0.5 1990-92 1995 1998-00 2004-05 2010 2015 Source: 2000 and 2004 MDHS It should be highlighted, however that the analysis of 2005 IHS2 data, suggests that fostering (whether due to orphanhood or to other reasons) is associated with lower probabilities of attending school (Government of Malawi and World Bank, 2006).19 There findings may signal the need for additional analysis of this indicator. Target 8: Have halted by 2015 and begun to reverse the incidence of malaria and other major diseases Seven indicators have been selected to monitor progress towards Target 8. Due to lack of easily available data we only examine the percentage of under five children using bednets (overall and treated). Indicator 22a: Proportion of population in malaria risk areas using effective malaria prevention (percent of under five children using bednets) The proportion of children using a bednet has increased from 7.6 percent in 2000 to 20.2 percent in 2004 (Figure 21). If this rate of improvement continues, about 55 percent of children will be using bednets by 2015. It should be noted that while no specific target 19The findings of the Poverty and Vulnerability Assessment indicate that fostering (whether due to orphanhood or to other reasons) is associated with lower probabilities of attending school (Government of Malawi and World Bank, 2006). Secure single orphans (single orphans living with one parent) are 2.7 percentage points less likely to attend school compared to children living with both parents. Attendance of children who are double orphans or not living with either surviving parent is 8 percentage points lower (from a mean of 82 percent) than their counterparts living with both parents. It is important to note, however, that children that are fostered out, but whose parents are alive, are even less likely to be attending school. 75 level has been set under this indicator, the aim is to increase the proportion of children using bednets. Figure 21: Ratio of school attendance of orphans to school attendance of non- orphans aged 10-14 Percent of under five children using bednets 100 90 80 current path 70 54.9% 60 Linearly projected value entc 50 MDG Target erP40 30 20.2% 20 7.6% 7.6% 10 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 2000 and 2004 MDHS Indicator 22b: Proportion of population in malaria risk areas using effective malaria treatment measures (percent of under five children using treated bednets) The proportion of children using a treated bednet has increased from 2.8 percent in 2000 to 14.8 percent in 2004 (Figure 22). If this rate of improvement continues, about 48 percent of children will be using bednets by 2015. As above, no specific target level has been set under this indicator, but the aim is to maintain or increase the proportion of children using bednets. Figure 22: Ratio of school attendance of orphans to school attendance of non- orphans aged 10-14 76 Percent of under five children using treated bednets 100 90 80 current path 70 60 Linearly projected value 47.8% entc 50 MDG Target erP40 30 14.8% 20 10 2.8% 2.8% 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 2000 and 2004 MDHS 77 GOAL 7: ENSURE ENVIRONMENTAL SUSTAINABILITY The seventh Millennium Development Goal is to ensure environmental sustainability. There are three targets under this goal. Target 9 is to integrate the principles of sustainable development into country policies and programs and reverse the loss of environmental resources. The second target, Target 10, is to halve by 2015 the proportion of people without sustainable access to safe drinking water and sanitation. Finally, Target 11 is to achieve, by 2020, a significant improvement in the lives of at least 100 million slum dwellers. In this report we do not examine progress towards Target 11 due to lack of easily accessible information. Target 9: Integrate the principles of sustainable development into country policies and programs and reverse the loss of environmental resources Five indicators assist to monitor progress towards Target 9: the proportion of land area covered by forest; the ratio of area protected to maintain biological diversity to surface area; the proportion of population using solid fuels; the energy use per $1 GDP; and finally, carbon dioxide emissions per capita and consumption of ozone-depleting CFCc. Due to lack of readily available data, in this report we only examine progress towards the first three indicators. Indicator 25: Proportion of land area covered by forest The proportion of land area covered by forest has been declining in Malawi from around 41 percent in 1990 to around 36 percent by the year 2005 (Figure 23). If this rate of deforestation continues, only about 33 percent of land area will be covered by forest in 2015. It should be noted that no specific target level has been set under this indicator, but the aim is to maintain or increase the proportion of area under forest. Figure 23: Proportion of land area covered by forest Proportion of land area covered by forest 50 45 41.4% 41.4% 40 35 37.9% 30 36.2% ent 32.8% rceP25 20 current path 15 Linearly projected value 10 MDG Target 5 0 1990 1995 2000 2005 2010 2015 Source: FAO Global Forest Resources Assessment, 1990, 2000, 2005 78 Indicator 26: Proportion of area protected to maintain biological diversity. In Malawi, the proportion of protected area has remained constant at 0.16 since 1990 (Figure 24). Again, while no specific target has been set under this indicator, the aim is to see the proportion of protected area increase, or at least stay constant. Figure 24: Ratio of area to surface area protected to maintain biological diversity Ratio of area protected to surface area to maintain biological diversity 1.0 0.9 current path 0.8 0.7 Linearly projected value tnecr 0.6 MDG Target 0.5 Pe0.4 0.3 0.16 0.16 0.16 0.2 0.1 0.0 1990 1995 2000 2005 2010 2015 Source: UNEP, online databank: http://unstats.un.org/unsd/mi/mi_series_results.asp?rowId=616 Indicator 29: Proportion of population using solid fuels The percentage of population using solid fuels has not decreased over the past decade. The 1998 IHS reveals that around 97 percent of the population used solid fuels, which is about the same level as reported under the 2005 IHS2 (Figure 25). Figure 25: Proportion of population using solid fuels Proportion of population using solid fuels 100 90 97.4% 97.9% 98.8% 80 70 tnecr 60 50 current path Pe 40 30 Linearly projected value 20 MDG Target 10 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1998 Census and 2004 MDHS 79 Target 10: Halve, by 2015, the proportion of people without sustainable access to safe drinking water Two indicators have been selected to monitor progress towards achieving this target: the proportion of population with sustainable access to an improved water source; and the proportion of population with access to improved sanitation. Indicator 30: Proportion of population with sustainable access to an improved water source. The data from the MDHS surveys suggests that proportion of population with sustainable access to an improved water source improved significantly during the 1990s, but has not increased since the year 2000. According to the 1992 Malawi Demographic and Health Survey, around 47 percent of the population had sustainable access to improved water source.20 This share reached 65 percent by the year 2000 and 64 percent in 2004. Unless further improvements are introduced, therefore, only around 60 percent of the population is expected to have access to an improved water source by the year 2015. Figure 26 shows the recent trend and projections compared to the path towards the MDG. Figure 26: Proportion of households with access to improved water source Households with sustainable access to improved water source 100 90 80 73.6% 65.2% 70 63.5% tnecr 60 47.1% 50 58.8% Current path Pe 40 30 Linearly projected value 20 MDG Target 10 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, and 2004 MDHS Note: Improved water sources include: piped (into dwelling or yard plot), public standpipe, borehole (protected public well), protected dug well. 20Improved water sources include: piped (into dwelling or yard plot), public standpipe, borehole (protected public well), protected dug well. 80 Indicator 31: Proportion of population with access to improved sanitation Malawi is making good progress in increasing the access to improved sanitation. In 1992, the proportion of households with access to improved sanitation was 72 percent. According to the MDHS, the proportion increased to 84 percent by 2004. At this rate of increase, almost 88 percent of households will have sustainable access to improved sanitation by the year 2015, which is above the MDG target (Figure 27). Figure 27: Proportion of households with access to improved sanitation Households with access to improved sanitation 100 83.9% 88.5% 90 81.4% 80 86.2% 70 tnecr 60 72.3% 50 Pe 40 current path 30 Linearly projected value 20 10 MDG Target 0 1990-92 1995 1998-00 2004-05 2010 2015 Source: 1992, 2000, and 2004 MDHS Note: Improved sanitation includes: flush toilet (own or shared), traditional pit toilet, ventilated improved pit (VIP) latrine. 81 the 2015 015.2 ack,tr ee yam thr on 2015. 2015 by Hence etgr by the ack.tr iss bytem by Ta of et.m 2015. on achieved achieved be two, be by feasibility beot iss get get get to etg get achieved get get get be tar tar tar indicator tar tar to to Off Off Off two On Off beot tar tar wever tar ho elyk achieved Off On Off Target ely the unlikely ely unli indicator be unlikely of is ess; unlik the is one etg are unlik ogr not pr oalG of etg tar tar ileh the oalG None W Good indicators Target 2015 0 27. 08. 0 0 8 20. 14. 11. 100 100 100 MDG Value de hoolingcs ect ojrP ary Value 3 7 2 4 7 m 5 8 5 50. 16. 10. 13. 20. pri 92. 103. 77. 2015 of Linearly urse daya co Rate dollar Change 210.- 110.- 010. 80.0- 170.- fulla te 950. 631. 630. Linear of one ple m than er co to ear) 0052 less hung able Recent (Y to ise 4 8 1 m 2 4 0 9 6 m 52. 17. 10. fror 22. 22. be 82. 85. 70. Most Value 2004 inco suffe illw,e te alik edia 0002 seohw how 9 6 0 4 6 2 4 1 m to girls value 53. 18. 10. 25. 23. 78. 79. 68. people people and Inter 1998 of of boys ear) 9921 (Y 0 0 0 here, 0 4 ATORS to - oportion - noi - hunger 54. 16. pr 28. at 58. 64. Baseline Value 1990 proportioneht eryw the ev NDIC and I educ S 2015, daya in 2015, dren n ary ting G and chil arey national $1 and eightw poverty at energyy in star 24 MD 1990 tioar ]yt quintile mirp 1990 age under opulatiop 2015, al ratio 15- ver dietar upilsp 5 of ON treme eenw po een of of by poorest of ent of tear LE ex bet headcount,yt gapyt ersv of of ofsarey betw that, lmor grade level five en cya ABT Indicator atec n) lve,aH ptionm oportion Pover Pover depthx Share lve,aH oportion Prevalence Pr mu niueve Net Pr reach teriL Ensure adi 1. 2. 3. consu 4. 6. 7. 8. A: 1: 2: Er populatio underne 5. nimim ptionm 3: education of EX 1: Achi:2 ary who1 im Target ndicatorI %( ndicatorI [incidence Indicator national Target ndicatorI childr ndicatorI below consu Target ndicatorI pr ndicatorI grade ndicatorI olds NN al A Goal Go ears k. tor t.em acrt 2015 likelynu 83 2015 2015 on by by indica be to by are Target five achieved the likely etg etg get get get be get etg get achieved feasibility achieved of is tar tar tar tar tar achieved dicatorsin toy tar tar tar track.f of be On On Off On On 2015.ybtem get target tar two rget Off Off be Off beot Off Off are Target ay Ta likel ay three ely be m ess; m to the oalG ogr track. pr on oalG of Target unlik Two oalG indicators Good Both Target 2015 2015 0 7 3 7 anht 1 1 1 50 50 78. 44. 95. 155.0 88. MDG Value er lat de no ect ion ojrP Value educat 031. 161. 970. 6 3 0 11. 25. 21. 81. 3 9 66. 610.0 60. 2015 of Linearly levels Rate allot Change and 010. 040. 010. 090.- 0 940. 14.- 8.6- 131.- -34.0 350. Linear of 2005, by oi e ear) 0052 0 Recent (Y to 950. 750. 860. 5 0 raty raty 0 7 0 12. 14. Most Value 2004 preferably alitrto 133. 76. 78. alitrto 984.0 57. m m te vei rnaleta edia 0002 education, 0 0 m to value 920. 60. 820. 1 13. 39. 2 6 Inter 1998 menow under-feht 189. 103. 83. meht 1120.0 55. secondary 15, 2015,d ear) 9921 er 20 an and (Y to pow ary 870. 50. - - 65. 0 0 and 8 8 234. 134. 85. 1990 620.0 54. Baseline Value 1990 em im 1990 pr eenw andyti in to by bet ins ins wage la between tear held ldo- ualqe disparity boy boy lity tear eary thaleh ers, ratio hst tosl tosl lesamfe innem culturi seats agr of ntem rta irds,ht alitytrom 1- easlesm quart of aln bir of personnel gir gir nder genderet literate wo lia mo o-wt ree-ht alityrtom of of of ge olds ofe non- tion par by ivef- alitytrom tion againstd tera by tion rnale health ina the opor child Indicator ote Ratio Ratios Ratio year Shar in Pr Under fantnI opor opor Pr Mat Pr Elim 9a. 9b. om 4: educationy 10. 24 11. 12. national Reduce 13. 14. 15. 5: unizemmi mevo skilled Reduce pr 16. 17. by 15- ntem 6: Pr:3 ary Reduce Target ndicatorI im pr ndicatorI secondar Indicator lesam ploy innem ndicatorI em sector ndicatorI wo al 4:l Target ndicatorI ndicatorI ndicatorI children mI:5 al Target Indicator ndicatorI attended Go Goa Go the k; 2015 acrt 84 on 2015. by by are etg get etg etg etg feasibility achieved tar tar tar tar tar be dicatorsin achieved Target ay m three beyam On Off On On On oalG the etg of tar Two Target 2015 4 - - - 6 8 - - - - 17. >7. >2. MDG Value de ect ojrP Value 904. 46. 10. - 9 8 - - - - 54. 47. 2015 Linearly Rate Change 2.1- 080. 000. - - - - - Linear of diseasesroja 153. 003. m ear) 0052 her ot Recent (Y 9 to 16. 55. 10. - 2 8 - - - - and 20. 14. Most Value 2004 lariaa te m edia 0002 V/AIDSIH of of 1 m to value 24. 25. 10. - 67. 82. - - - - Inter 1998 diseases spreadeht incidenceeht ear) 9921 (Y 4 3 to other - - - - - - - - 17. 12. Baseline Value 1990 anda reverseot reverseot begun ce n- begun se five on of no rat on and ongma the malari ulosisc ulosisc of of and 100000 under se) se) attendan death DS, 2015 sites tear e 2015 opulatip rat of ectivefef of opulatip of ectivefef (percent eatedtr se erp( bertu ectlyirD( bertu rat associated of S Court of ectlyirD( Court by use by and S V/AI ed evalence school tion tion tion tion pr sentinel ed of encel centerp( ulosisc tesar DOT Shor Shor alence DOT HI halt Pr are bed Pr are Prev Death Pr under ntem opor ntem Indicator aveH HIV atnem mo attendanceloo 41 halt lariaam poro usingsa ets)n opor usingsa asuresem Cond prevalence Ratio sch 10- Preva ntem usingne tuber poro Pr 18. wo 19. to with childr with 7: 20. aveH 21. 22a. skir evention 22b. skir 23a. 23b. ulosisc 24a. eatrT 24b. eatrT aged n) 8: underde Combat phans 6: egnant or phans rialaam prai usingne eattr ivef- tuber detected edv cur edv Target ndicatorI pr ndicatorI contraceptive ndicatorI of or Target Indicator associated ndicatorI in laram rialaam childr ndicatorI in rialaam under bednets) Indicator associated populatio ndicatorI with ndicatorI cases Obser ndicatorI cases Obser Goal ehT 2015 by ack.tr ack.trno 85 is on 2015 achieved feasibility notears 2015.ybtem by be get get get indicator get get beot to tar tar tar Off On Off oney tar tar unlikely onl is Off On Target ely indicator unlikely ee is ess; unlik thr etg ogr etgraT oalG ofowT tar pr Good Target 2015 4 6 2 MDG Value latne 41. 160. - - 0 - 73. 86. nm de viro ect en ojrP of Value 8 loss 32. 160. - - 8 8 5 - 98. 58. 88. 2015 the Linearly erse ellers rev dw m Rate ands Change m 340.- 000. - - 080. teraw 430.- 420. slu - Linear of ingk illion progra inrd m0 10 ear) 0052 and safe Recent (Y 2 to 36. 160. - - 9 5 9 least 97. to 63. 83. at - Most Value 2004 policies of tryn access lives te the edia 0002 cou to 9 4 3 4 m to value int 37. 160. - - 97. sustainable 65. 81. intne - en m Inter 1998 m prove ear) 9921 y mi (Y veloped ithoutw 4 1 3 to Baseline ilitb 41. 160. - - - - people 47. 72. of Value 1990 naiastusl sustainablefo significanta to e- acef n ortionp n d d ds ta eaar ilo sur am issions ozon pro ciples en otected to em of pulatio land po the pulatio provemi ban pr po ur ho mn prine of sity PPP)( de an of of of provedmi achieve usehol roivne thet eaar ilogrk( ptionm tons) 2015, to larur to have of tion to tenure of diver use dioxi GDP nsu tion by tion and tion tion re opor $1 co ODP( opor opor access ban opor access 2020, opor Indicator Pr ste Ratio ergynE onb Car Pr Pr Pr Pr secu reu Integra for per and fuels lve,aH ur to 25. 26. biological 27. 28. CFCs 29. 30. ce, By 9: Ens byde 31. withn 32. 10: 11: capita) solid sustainable sour access 7:laoG Target resources ndicatorI cover ndicatorI intainam area ndicatorI equivalent) ndicatorI erp( depleting ndicatorI using Target ndicatorI with water ndicatorI populatio sanitation Target ndicatorI with REFERENCES Government of Malawi, National Economic Council. 2000. Profile of poverty in Malawi, 1998: Poverty analysis of the Malawi Integrated Household Survey, 1997-98. Lilongwe, Malawi. Government of Malawi. 2006. 2004-2005 Integrated Household Survey Report. Zomba, Malawi. Government of Malawi and UNDP. 2003. Malawi Millennium Development Goals Report 2003. Government of Malawi and World Bank. 2006. Poverty and Vulnerability Assessment 2006: Investing in our future. Lilongwe, Malawi. Grown, Caren, Geeta Rao Gupta, and Aslihan Kes. 2005. Taking action: achieving gender equality and empowering women. Sterling, Va: Earthscan. NEC, 2000 ORC MACRO. 1992. Malawi Demographic Health Survey Report: 1992. (http://www.measuredhs.com) ORC MACRO. 2000. Malawi Demographic Health Survey Report: 2000. (http://www.measuredhs.com) ORC MACRO. 2004. Malawi Demographic Health Survey Report: 2004. (http://www.measuredhs.com) United Nations. 2003. Indicators for Monitoring the Millennium Development Goals: Definitions, Rationale, Concepts and Sources. New York: United Nations. 86 ANNEX 2A: GENDER INEQUITIES IN MALAWI 21 1. This Annex focuses on gender inequities that originate from social and cultural factors. In addition to constituting a form of discrimination, unequal opportunities and access to resources based on sex, may lead to lower productivity and economic growth for society as a whole. At the fourth conference on women in Beijing it was claimed that as many as many as two third of the world's poor are women (Chant 2003). Such gender disparity is largely due to the fact that household members do not enjoy the same level of welfare. Unfortunately the poverty status of individual household members cannot be calculated from the information in the IHS2.22 In this Annex, therefore, we do not assess the links between gender and poverty, but limit the analysis to an assessment of the differences in access to resources and bargaining power as an indirect measure of a person's poverty status. 2. The Annex is structured as follows. The first section discusses inequities in access to resources, notably in education, access to credit, and participation in income generating activities. The latter includes a focus on agricultural production (access to land, fertilizer, and extension services), wage employment, and entrepreneurship. The next section analyses the division of labor between women and men both in the labor market and in the domestic sphere. The third section discusses the empowerment of women both with regards to decision making within the household in general and within agriculture. The fourth section discusses the victimization of women as measured by the prevalence of violence in general and especially domestic violence. The final section provides a summary of the main findings, as well as some policy implications. Additional information is presented in the Table Appendix at the end of the note. GENDER INEQUITIES IN ACCESS TO RESOURCES Education 3. The analysis in Chapter Two has shown that the level of education is one of the most important determinants of the poverty. The higher the level education attained by the household head, the lower the probability for the person to be poor. Hence, gender equity in education is an important predictor for gender equity in poverty. Also studies have shown that increasing female education has a substantial positive impact on economic growth. This relationship is robust across most developing countries (Klasen 2002). 4. Men have better access to education (Table 1 and Table 2). About 76 percent of adult men (15 years and above) are literate, compared to about 52 percent of the women. 21Several of the variables presented in this Annex are calculated using different definitions from those used in the rest of the report; as a result there may be minor discrepancies between some of the statistics presented in this annex as compared to the rest of the report. 22Information on distribution of consumption within the households is generally not collected in household surveys. 87 About 35 percent of adult women have never attended school, compared to about 16 percent of adult men. Similarly, about 16 percent of all adult men have completed at least secondary education, compared to about only 8 percent of adult women. Table 2A.1 Literacy rate and school attendance for persons 15 years and above Proportion literate Proportion literate 15-24 years Malawi 64 76 Male 76 81 Female 52 72 Source IHS2 Table 2A.2 Percentage distributions of men and women by highest level of education completed Never attended Attended school, but no Secondary level school completed level Primary level and above Malawi 26 48 10 11 Male 16 52 12 16 Female 35 48 7 8 Source IHS2 Note: The rows do not sum to 100 due to some persons with missing on education. 5. On a positive note, gender inequities appear to be decreasing over time. For the younger population (15-24 years old) the gender gap in literacy is only 9 percentage points, as compared to 24 percentage points for the whole adult population (Table 1). Enrolment rates in primary education confirm that gender equity in access to education is improving. There is no difference in primary school attendance between boys and girls. However, there remains a large gender gap in attendance to secondary education (Table 3). About 49 percent of girls 15-18 years old attend school, compared to 59 percent among boys. Table 2A.3 School attendance of girls and boys in different age groups by sex of household head. Girls in school Boys in school 5-9 10-14 15-18 10-14 15-18 years year years 5-9 years year years Malawi 72 90 49 69 90 59 Male-headed households 72 90 47 69 91 59 Female-headed households 71 90 54 69 88 61 Source: IHS2 6. These trends are confirmed when looking at the ratio of females to males in primary and secondary school, which is a measure of the gender equity of education. It is calculated by looking at the proportion of females to males in primary and secondary education. The ratio of girls to boys in primary school is almost 1:1 across wealth group and in rural and urban areas (Table 4). In urban areas, the results show 7 percent more girls than boys attending school. 88 Table 2A.4 Ratio of Females to Males in Primary and Secondary Schools (percent) Malawi Poor Non-Poor Urban Rural Primary 95.0 0.93 100 1.07 0.96 Secondary 70.0 43.9 78.4 0.91 0.58 Note: Primary education includes all students attending Standard 1 through 8 in 2004 and 2005. Secondary education includes all students attending Form 1 through Form 6 in 2004 and 2005. Source: IHS2 7. In secondary school, the number of male students is substantially greater than the number of female students: for every 100 male students in secondary education, 70 are female. This was also seen in the enrollment figures presented earlier. The gender gap for secondary education is smaller in urban households compared to rural households, and in non-poor households. 8. The gender of the household head is an important determinant of children's school attendance. A preliminary analysis would suggest that the gender of the household head has no impact on school attendance (Table 3).23 However, once we control for differences between male-headed and female-headed households in levels of consumption, household size, and educational level of head and spouse, we find that female-headed households tend to spend a significantly larger share of their total budget on education compared to male-headed households. In fact the results of a regression explaining the share of household expenditure allocated to education suggest that female household heads are significantly more likely to send children to school (see Annex 3B).24 Access to credit 9. Access to credit is important in order to be able to invest in other income generating activities, i.e. agriculture, as well as for investment in durable goods or for smoothing consumption in difficult times. However, only about of 5 percent of the adult population obtained loans in 2004, and only a third of these were women (Table 5). Among the women it is those that are most likely to live without a man that obtained credit. Table 2A.5 Access to credit by sex and marital status Female Male Access to credit All 3 7 Divorced/separated6 4 Married 3 10 Never married 0,4 0,8 Widowed 7 6 23 The only difference would be that girls from female-headed households are somewhat more likely to attend secondary school (54 percent) than girls from male-headed households (47 percent). 24The share of expenditure also increases with the educational level of the parents. 89 10. There is a clear gender difference when it comes to what credit is used for. Men, in most cases, obtain loans to purchase input for agricultural production, and in particular for tobacco production. To the extent that women obtain loans for agricultural production, it is mainly as input for food crop production. The majority of women obtain a loan to start up a non-agricultural business (Figure 1)Error! Reference source not found.. However, Figure 1 shows that even though starting up a business is the most common reason for obtaining a loan among women, only 17 percent of the female operated enterprises were started up based on loans. In particular credit from formal credit institutions is rare as a source for start-up capital for a business. This may be due to lack of collaterals because of the small size of most enterprises, as well as lack of other forms of collateral. Rather, both women and men seem to rely on own savings and gifts from family and friends to start up a business (Figure 2). This indicates that there is a scope for encouraging female entrepreneurs by making credit more readily available. Figure 2A.1 Reason for obtaining a loan among men and women 60 50 40 30 20 10 0 buy land input, food input, input, oth Buisness non-farm Other crops tobacco cash crops start up input Male Female Source: IHS2 11. Both females and males tend to get loans mostly from informal sourcesError! Reference source not found., confirming the pattern found for start-up capital for enterprises. Relatives and neighbours are important sources for credit for both sexes, while few get loans from formal credit institutions. It can be argued, that stringent loan administration policies by formal lending institutions (collaterals, high interest rates, etc) especially might discourage women from obtaining loans from such sources This might be due to the fact that women in general are more risk adverse than men, and also that more women are illiterate and less able to handle the formalities for to obtaining a loan. This is confirmed when looking at the average education among those that obtained loans and those who didn't (Table 6). Those that obtained loan both among the women and men have considerable higher education than those who did not. Table 2A.6 Average years of education by sex and access to credit Female Male All 3,9 5,8 90 Access to credit 5 6,2 No access to credit3,9 5,8 12. It is interesting to note, however, that Non-governmental organizations are the number one source of credit for women. This might be due to the importance given to support female income generating activities, not the least from international organizations. Figure 2A.2 Source of loan for men and women 35 30 25 20 15 10 5 0 relative neighbour lmerchant eylender ploye RC nk MF CCO Ba NGO Other em SA loca mon religious inst. Male Female Source: IHS2 13. Further, the bigger the loan, the lower percentage of women who obtained a loan. Error! Reference source not found. shows that only for loans of less than MK1000, women more often than men are the borrowers, (38 versus 23 percent). Further, the amounts borrowed, reflect the limited capital available for loans from informal sources (relatives, money lenders, etc.). Figure 2A.3 Amount of credit obtained last 12 months by sex of borrower Amount of loan obtained, by sex 45 40 35 30 25 Male 20 Female 15 10 5 0 <1000 1000-5000 5000- 10000- >500000 10000 50000 Source: IHS2 91 Occupation in income generating activities 14. One of the most important factors used to explain gender inequity is the possibility to participate in income generating activities and thus to contribute to their own welfare as well as the welfare of the household and of their children. The hypothesis is that participation in income generating activities, and especially in formal employment with high remuneration, would be of special importance, both when it comes to the amount of income generated by the woman, to the bargaining power of the woman within the household, as well as the empowerment of the woman, both within and outside the household (Wold 1997). 15. Subsistence farming is by far the most important type of occupation for both men and women (Table 7), and more so for women than for men. Men are more likely (than women) to be engaged in income generating activities that can generate cash. Men are more likely to work in a family business outside the agricultural sector, or to work as ganyu (casual work), or to work for a wage or salary. Men also worked more hours than women in each of these activities. The result is increased gender disparity in access to cash remuneration for work. Poor women spend less time in wage employment and they are at a disadvantage when it comes to access to cash remunerating employment. 16. As indicated in the poverty profile, there is a difference in the proportion of persons engaged in various income generating activities according to poverty status. Further, it is generally assumed that poverty accentuates gender disparities.25 The gender gap does not appear to worsen with poverty status, however. In fact, IHS-2 data highlights that average hours worked does not vary much according to poverty status. The only exception is the gender gap in the number of hours worked in wage employment that is larger for the poor than for the non-poor. 17. In what follows we focus specifically on gender disparities in the main categories of occupation, namely agriculture, wage employment, and household owned non- agricultural activity. Agriculture 18. Agriculture is the largest income-generating sector in Malawi. It supports about 85 percent of the population, accounts for almost 35 percent of the Gross Domestic Product, and contributes most of the income for the rural population, whether poor or non-poor. Agriculture is also the most important sector for female employment, (see above). Hence, the outputs generated from those activities may play an important role when it comes to gender inequities, not only in employment, but also in other areas of life. This is so whether we talk about crop growing, access to land or factors that can influence productivity, such as receiving starter packs and access to extension services. 25It is often observed that the gender gap for poor women (compared to poor men) is worse than the gap for non-poor women (compared to non poor men). 92 19. Since little information on allocation of resources and responsibility in agriculture within the households is available, sex of head of household is used as a proxy for sex when further analyzing gender issues related to agriculture. 20. In IHS2, agricultural/farming households were defined as those households with at least one member engaged in farming in the last cropping season (2003/04). According to this definition about 90 percent of all Malawian households can be labeled agricultural households. The proportion of agricultural households is even higher among female-headed households than among male headed, 95 percent as compared to 88 percent (Table 7). This difference is mainly due to due to the larger proportion of urban female-headed households engaged in farming, and can be an indication to the larger need of those household to increase their welfare through consumption of own produce, and/or through sale of agricultural produce. It can also be noted that among the poor households, there are no sex differences as to engagement in farming households, but it seems as if among the non-poor female headed households, farming activities may to some extent be a strategy to avoid or alleviate poverty than among male-headed households, 89 percent among the non poor female-headed households were engaged in farming activities as compared to 83 percent of male-headed non-poor households. Table 2A.7 Proportion of male-headed and female-headed households engaged in farming by place of residence and poverty status. Male-headed Female-headed Malawi 88 95 Urban 40 51 Rural 95 96 Poor 95 96 Non-poor 83 89 Source: IHS2 Crops cultivated 21. Males are often said to be growing cash crops in order to provide cash income to the households. The females, on the other hand, are supposed to be more prone to grow crops for own consumption, in order to provide food for the household members. 22. From the available data it is not possible to analyze whether, or to what extent, different types of crops are grown among men and women. We can, however, get an indication of what are typically female crops by looking at crops grown by female-headed versus male-headed households. Preliminary analysis of the IHS2 data suggests that there is indeed a difference between female headed and male-headed households when it comes to type of crop grown. Male-headed households have a greater tendency to grow cash crops, while the female-headed households to a larger extent grow crops for own consumption. 23. The most important 'female' crop is local maize, which is grown by more than 60 percent of the female-headed households, even though it is also important for the male- headed households, as half of all of these households grow local maize. Local maize is often used for own consumption. The same is the case with hybrid maize, but, hybrid 93 maize can only be used for consumption and not for seeds, and also tend to yield higher output if fertilized, such that the possibility of selling some of the produce is larger than for local maize. About the half the male-headed households grow hybrid as local maize, as compared to 40 percent of female-headed households. The relative larger proportion of female-headed households growing local maize, therefore, not only indicates that they grow crops for own consumption, but also that they grow crops where the need for commercial input is less. 24. It can also be readily seen that the most important cash crop in Malawi, tobacco, is more often grown in male headed than in female headed households, 19 percent as compared to 7 percent. If those findings can be generalized to the cultivation of male crops and female crops also within the household, it is not surprising to find this pattern in crop production. Traditionally, women have had the responsibility of feeding the family, and thus make sure that they grow crops that will ensure that food is available in their household. They are less willing than men to take risks that can jeopardize the food security of the household. 25. Another reason could be that the colonial administrators and technical advisors introduced cash crops and modern agricultural technologies to men (Eicher et al. 1990). Men therefore, developed a productive export-oriented farming sector while women were left behind in the traditional low-yielding subsistence sector. ACCESS TO LAND AND INCOME GENERATED FROM LAND 26. Access to land and the size of the land holding, are also important resource indicators, especially in an economy dominated by agriculture. Also, access to land and holding size has a bearing on food security and even on poverty, since the lesser the holding size, the lesser the productivity and the lesser the possibility to grow crops that can both provide food security and the necessary cash to provide for other household needs, and eventually, to avoid poverty. Table 8 shows that there are no significant difference between male-headed and female headed households in access to land and holding size per capita. Most holdings are small, more than half the holdings are less than 0.25 ha per capita, and only a small fraction has more than 1 ha of land per capita. Table 2A.8 Percentage distribution of male headed and female-headed households by landholding size per capita Male-headed Female-headed All households households households 0 ha 12 13 9 0-0.25 ha 54 53 55 0.25-0.5 ha 24 24 23 0.5-1 ha 8 8 9 above 1 ha 3 3 3 Total 100 100 100 Source: IHS2 94 27. There is a link between poverty and access to land and size of landholding. Female-headed households without land have a larger proportion of poor than landless male-headed households, 35 percent as compared to 27 percent. This may indicate the greater possibility for land less male-headed household to find employment outside agriculture, as compared to their female counterparts. The table also shows that the smaller the holding size, the larger proportion of poor households, but this relationship is more pronounced among female headed than among male headed households The larger the holding size, the smaller the proportion of poor households, both for male headed and female headed households, and the gender inequity disappears. Table 2A.9 Proportion of poor male headed and female-headed households by landholding size per capita All Households Male-headed Female-headed households households Total proportion poor 52 51 58 0 ha 28 27 35 0-0.25 ha 63 62 70 0.25-0.5 ha 42 41 45 0.5-1 ha 21 21 22 above 1 ha 29 29 31 Source: IHS2 28. The relationship between sex and whether crops were grown for sale or own consumption were discussed above, concluding that male headed households had a greater propensity to grow cash crops than female headed households. Table 10 seems to confirm this finding. Regardless of holding size, female-headed household to a greater extent than male-headed households grow produce only for own consumption, that is, they do not sell any of their produce. Table 2A.10 Proportion of male headed and female-headed households that sold something of their production, by size of holding per capita Male-headed Female-headed All households households households Did not sell any produce 64 68 52 0-0.25 ha 51 55 38 0.25-0.5 ha 72 76 58 0.5-1 ha 75 80 61 above 1 ha 77 79 71 Source: IHS2 29. Using sex of household head as a proxy for sex of a person, this section has shown that there exists gender inequities both in access to land, holding size, cash crop production as well as the probability of being poor, and that females are the ones at a disadvantage. This will again have consequences both for the bargaining power of women related to decision making both within and outside the household, as well as the empowerment of women both in the private and public sphere. 95 Starter packs 30. The Government of Malawi did for some time distribute starter packs to agricultural households (up to 2005) in order to enable households to either grow, or grow more effectively, various types of crops. The starter packs included for instance, seeds, fertilizer and pesticides. The program was intended to reach out to households having difficulties acquiring the necessary agricultural inputs. Over the period covered in IHS2, 2001-2004, the share of agricultural households that received starter pack varied between 35 and 35 percent. Female-headed households benefited somewhat more from this program than male-headed household, the sex difference being about 7 percent for each year covered, in the favor of women. Hence, the program most probably contributed to reducing gender disparities in the agricultural sector, and would most probably also do so in the future, if continued. Table 2A.11 Proportion of agricultural households who received starter pack 2001- 2004 by sex of household head Proportion who received starter pack 2001 2002 2003 2004 Malawi 35 42 46 42 Sex of household head Male 34 40 45 40 Female 41 47 51 47 Source: IHS2 Use of extension services 31. Malawi has a system of agricultural extension services, with extension officers giving advice on various aspects of agricultural activities. Since most female-headed households are involved in agricultural activities, they would have an equal need as their male counterparts to access advice on how to select the right produce to grow and how to boost productivity through the use of the most effective seeds, use of fertilizer and pesticides etc. From IHS2, information is available on which farming households who received any kind of advice from agricultural extension workers26. Only seven percent of female-headed farming households received such advices as opposed to thirteen percent of the male-headed farming households. Table 2A.12 Proportion of farmers that received advice from Agricultural Extension Workers, by sex of head, place of residence and poverty status All households Male-headed Female-headed households households Malawi 11 13 7 Place of residence Urban 2 2 3 Rural 13 14 8 26This advice capture one or more of the following; general crop production, new seed varieties, fertilizer use, pest control, irrigation, general animal care, animal diseases or animal vaccinations, marketing, access to credit, and growing and selling of tobacco. 96 Poverty status Poor 11 13 7 Non-poor 11 13 7 Source: IHS2 32. This indicates first of all that access to, or use of, the agricultural extension service is limited; both for male headed and female-headed households, but also that there are gender disparities in the access to, or use of those services. There are no data available to shed light on this gender inequity, but one hypothesis could be that in most cases, the agricultural extension service is dominated by male extension workers who might have problems relating to female farmers and hence neglect their need for advice. Also, since normally the extension service is normally provided through meetings, and not on a one-to-one basis, it might pose problems for females to attend those meetings. Access to extension services may require travelling for long distances to reach the Extension Planning Area where agricultural meetings and demonstrations are carried out. Concerns about safety may make women hesitant to travel, and their responsibility for cooking and childcare may make it difficult for them to be absent from home for the required length of time. 33. One can suspect that this gender disparity in access to advice from the extension service also prevails within the household. Extension workers would normally consider a household as a homogeneous unit of production and consumption, thus assuming that when men are offered training and extension services, they will pass on the information to their wives. However, this is not necessarily the case, and the women's access to the necessary knowledge required for efficient farming will be restricted. Wage Employment27 34. Even though agricultural activities are the most important income generating activity for Malawians, wage employment may be more important when it comes to securing cash income for a person or household. 35. IHS2 provides data on participation in some kind of wage employment for all persons 15 years and above for some time during the 12 month period prior to the survey. As discussed in Chapter Two, wage employment is not a very widespread phenomenon in the Malawian economy, and even less so among women than among men, only 6 percent as compared to 22 percent were in wage employment some times last twelve months. The table also shows that the remuneration received from wage labor is low, and again even less for women than for men. The median daily wage for women was MK78, as compared to MK124 for men. 27The statistics on wage differentials presented in this section need to be cautiously qualified as they are just averages which are based on very few observations, and also because, as highlighted in the text, these averages do not control for other aspects that lead to such differences. For example, there is need to distinguish between ganyu and salaried employment, and also the educational levels, skills, and level of experience have to be taken into account. 97 36. What the data does not show, however, is whether the lesser participation and the lesser hours worked in wage employment among women is chosen by the woman or imposed by the woman, either because lesser opportunities for wage employment or greater responsibilities for chores outside the labor market. If participation in, and amount of, wage employment, is not determined by the woman herself, a gender inequity exists wage employment, with women at a disadvantage. 37. One reason for the lower participation in wage employment among women could be that very few women are educated and qualified, and that many forms of wage employment normally require a minimum of skills. Table 13 confirms that education level is higher for those employed than for those not employed for a wage. Table 2A.13 Average years of education by sex and wage employment Female Male All 3,9 5,8 Wage-employed 6,5 6,9 Not wage-employed3,7 5,5 These figures hide large discrepancies between the different occupations, and Figure 4 shows that administration and management, professionals and clerical occupations not surprisingly require much higher education than the other occupational groups. Among these occupations there is no gender disparity in terms of proportion of respectively women and men employed. Figure 5 shows that the proportion of female workers in professional and technical, jobs as well as in clerical jobs are higher than the proportion of men, although men still outnumber women in absolute terms. In administrative and managerial jobs the proportions are equal. Further, Table 14 shows that for these three occupations, which are also the best-paid ones, women are paid about the same as men. This indicates that with a higher level of education, women may compete with men for employment opportunities in the white-collar sector. 38. The lower average wage for employed women than men is mainly due to the lower wage paid for women employed in production activities, and as laborers and in service compared to men. 39. Both for women and men, laborer, not elsewhere classified, is the largest occupational group, but relatively more common among women than among men. One reason for this may be the character of the Malawian labor market. It is dominated by informal sector employment, where few skills are required and it can be very difficult to classify to which occupational category a person belongs. Laborers are characterized with slightly below average education, and lower education for women than men, corresponding to the lower wage paid for women. Also being occupied with some kind of service work is fairly common, both among men and women, but with a majority of men. In service work, however, the women are paid on average less than men, even though the average education for the employees is about the same. Workers in production activities are predominantly men. The average education among men in this occupation is considerably higher than for women. 98 Figure 2A.4 Average years of education by sex and type of occupation 18 16 14 12 10 8 6 4 2 0 employed d manage gric. Fish Clerical rers tion s les sional Sa rvice Labo Se not in. an A Produc Profes Adm Male Female Figure 2A.5 Percentage distribution of wage employment by sex and type of occupation Source: IHS2 30 25 20 15 10 5 0 nage d ma ric. Fish Clerical Laborers Sales Service in. an Ag Production Professionals Adm Male Female Table 2A.14 median wages per day by sex and type of occupation Administration and Agriculture Production management and Fishing Clerical Laborers activities Professionals Sales Service Male 1200 40 156 70 120 184 100 88 Female 1680 35 168 48 45 166 100 60 Malawi 1580 40 160 60 112 183 100 80 40. Finally, Table 15 shows that a relatively large group of the divorced or separated women are employed compared to the married, never married, and widowed Table 2A.15 Proportion of women and men in wage employment, by marital status Male Female Malawi 22 6 Married 28 5 Never married 11 6 Separated/divorced30 11 Widowed 22 6 Entrepreneurship 99 41. An important source of income for Malawian households is a non-agricultural household enterprise. As many as 31 percent of all Malawian households owned and operated an enterprise producing goods or services, a shop or a trading business. It can be expected that running an enterprise would be an especially important factor for empowering women for at least two reasons: It can be combined with domestic obligations because it allows for flexible working hours and it is one way to surpass possible discrimination in the formal labor market. Table 16 shows that 10 percent of women owned and managed their own enterprises as compared to 16 percent of men28, even though not all of them put in work in their enterprise the week preceding the survey. Women most probably combine running an enterprise with other activities or have restricted time available for running the enterprise due to other obligations. While a man on average put in 29 hours a week in his enterprise, a woman on average put in 20 hours per week in hers (Figure 6). Table 2A.16 Proportion of women and men running an enterprise, by marital status Female Male All 10 16 Divorced/separated 17 21 Married 9 22 Never married 2 4 Widowed 16 12 42. Operating an enterprise seems to have no correlation with educational status for women, while men operating an enterprise tend to have slightly lower education than those men who don't (Table 17). Table 2A.17 Average years of education by sex and whether running an enterprise Female Male All 3,9 5,8 Running an enterprise 3,9 5,4 Not running an enterprise3,9 5,9 43. Women were on the average obtaining a substantially lower profit from their enterprises than men, MK160 per day compared to MK280. These numbers are corrected for the amount of time worked in the enterprise. The difference may be a result of the lower input of time in the female enterprises and thus less possibility to accumulate skill and further investments, it might be related to the lower education for women, also among those operating enterprises and it might be related to the type of enterprises operated. For both men and women, however, enterprises categorized, as unspecified retail is the most important category. In addition many women are engaged in traditional female activities like baking, either beer distilling or beer brewing and providing street food, while men on the other hand, are engaged in manufacturing handicrafts and fishing, that is, traditionally male activities. 28Here, managed refers to the enterprises owned and managed by only one individual. Such enterprises come to 88 percent of all enterprises. The remaining were owned and managed by more than one person. 100 44. It is interesting to see that the divorced and separated and the widowed women, i.e. those that are most likely to live without a man show a stronger tendency to entrepreneurship than the married and never married ones. DIVISION OF LABOUR Division of labor between women and men 45. There is a clear gender disparity in the use of time. Women spend more time on domestic work than men do, and hence have less time to engage in income generating activities than men have, as presented in Figure 6. The figure shows that men spend considerably more time on income generating activities than women, on average about 27 hours a week, as compared to about 17 hours a week for women. Women spend considerably more time on domestic tasks, about three and a half hour daily, as compared to just about half an hour daily among men. If time used on a daily basis for housework is converted to hours per week, the figure also shows that women has a considerably heavier workload than men, averaging 40 hours per week, as compared to about 30 hours among men. This gender inequity in workload is likely to be even larger, since tending to children, and even caring for the sick, traditionally female tasks, are not included. Especially caring for the sick may bring an extra burden for women and girls, because of the AIDS pandemic (see Chapter Five). Figure 2A.6 Average hours worked "last week", 15 years and older, by sex 45 40 35 30 sr 25 housew ork hou20 inc.gen w ork 15 10 5 0 man w oman Source: IHS2 46. A considerable share of the time spent on domestic activities is used for heavy tasks such as collecting water and firewood. Three out of four Malawian women reported that they collected water last week and that they spent on average one hour and 15 minutes every day doing this. Only one out of ten men collected water and spent considerably less time doing so, 3 quarters of an hour. 101 47. One fifth of all women collected firewood, and they spent daily one and a half hours doing this work29. Only one of twenty men did this type of work and they spent on average slightly less than an hour doing this work. In addition to collecting firewood being time consuming and heavy work, it may have detrimental health effects, and even other dangers associated with it, varying from snakebites to rape. Nankuni (2004) also shows that stunting in children is associated with absence of safe water in household and long time spent by the woman collecting water, even when controlling for important variables as education and poverty. 48. It should be noted that in a predominantly agricultural economy, time spent on income generating activities is likely to vary over the year, with a peak in the main cropping season. Wodon and Beegle (2005) found that both women and men spend more time on income generating activities in December-January. They also found that the gender disparity in hours worked between the gender were lower during those months. This indicates that the possibilities of finding work for cash are highest during the main cropping season. Poor farmers in need for cash income might thus reduced input of labor on their own fields, affecting the productivity of land. This might in particular affect women who in general seem to lack the same job opportunities as men. Division of labor between girls and boys 49. An important question from a gender perspective is whether the gender inequities in division of labor is likely to be repeated for future generations, or whether there are signs that changes might occur in the future. One possible way of coming to grips with what the future may bring, is to look at what the situation is for the new generations growing up, that is, the division of labor between girls and boys. 50. The gender inequities documented seem likely to be reproduced for future generation. Figure 7 shows that a girl between 10 and 15 years of age, works on average about 16 hours a week while a boy in the same age group work about 10 hours. The same pattern is even repeated for younger children. The differences observed are mainly due to less time spent on housework for boys. Figure 2A.7 Average hours worked per week for children between 5 and 15 years 29Collecting firewood may not necessarily be daily tasks, depending on availability and season. Hence, while the numbers show that one fifth of all women collected firewood last week, one should expect that a higher share is involved in such work since in many cases it is preferable to collect wood once in a while and rather pile it up. 102 18 16 14 12 10 8 6 4 2 0 boys 5-10 years boys 10-15 years girls 5-10 years girls 10-15 years housew ork inc.gen w ork Source: IHS2 51. In analyzing whether gender inequities are likely to be reproduced in the future, it will also be of interest to look at the relationship between child labor and school attendance. A plausible hypothesis would be that even if girls do more work than boys, but that this is not reflected in school attendance. If this is the case, it will affect the girl's possibility for doing their homework and thus the performance at school. 52. The Malawi Demographic Education Survey (2002) tried to capture some of the factors that led to not attending school among children aged 6 to 15 years. Table 2A.18 Reason for never attending school among boys and girls Needed for Travel to work at Student not school is Distance to home Cost to great interested unsafe school Girls 6 15 31 14 35 Boys 3 20 35 12 25 Source: DHS edData Survey(2002) Table 2A.19 Reason for dropping out of primary school among boys and girls Illness for Needed for Unsafe Student no longer than 3 work at Distance to travel to Failed longer months home Cost to great school school examination wanted Girls 11 37 25 12 4 10 41 Boys 12 23 23 7 3 17 48 Source: DHS edData Survey (2002) 53. One of the reasons listed was not attending school because work was needed at home. According to the survey, a greater proportion of girls than of boys cited 'needed for work at home' as reason for not attending school, but the for neither sex, the proportion is large, 6 percent as compared to 3 percent. The reasons most often cited both for girls and boys were 'not interested' and 'distance to school' (. However, looking at dropouts, more girls than boys drop out of primary school because they are needed for work at home, 37 percent as compared to 23 percent (Table 18). This may be an indication that the gender inequity in division of labor is likely to continue in the future. Older children (10-15 years) and both girls and boys, do more work, whether domestic or income generating, 103 than younger children (Table 19). Children in this age group not attending school, engage more often in income generating activities and also spend more time doing them, than those attending school. 54. It is, however, a puzzle that among the youngest, a lower share of those not attending school participate in housework and income generating activities compared to those in school, see Table 19. This is however, not in contrast with the findings from MDHS, indicating that labor needed at home is not a main reason for not sending the children to school. Table 2A.19 Children's engagement in domestic work and income generating work, by sex, age and school attendance Income generating work Housework Proportion that did Average hours Proportion that did Average hours this work last week worked last week* this work last week worked last week* Boys (10-15) not in school 58 19 31 11 in school 49 12 42 10 Boys (5-10) not in school 9 12 12 8 in school 19 9 25 9 Girls (10-15) not in school 52 19 80 18 in school 42 11 83 14 Girls (5-10) not in school 8 8 24 9 in school 16 8 47 10 Source: IHS2 *For those doing this type of work EMPOWERMENT AND DECISION MAKING 55. Typically expenditure surveys, as the IHS-2, only report aggregated household consumption and one is not able to detect actual allocation, for example allocation of food or resources to education, within the household. There is a large gender literature concerned about bargaining power and distribution within the household, because the households "are arenas of competing claims, rights, power, interests and resources" Chant (2003), which will be reflected in the consumption level and poverty status for the individual household members. This section focuses on intra household circumstances that shape individual opportunities and entitlements, and on the bargaining power among household members. This covers women's participation in decision making, and distribution of responsibilities and benefits within the household, i.e.; women's contribution to, and control over, the household earnings: and gender based violence. Decision making within the households 56. Participation in decisions about one's own needs and household needs is an indicator of women's empowerment. In MDHS (2004), women were asked whether they were involved in specified decisions in their household. 104 57. Among currently married women, only one out of five say they are having a say when it comes to larger purchases (Figure 8). One out of three say they have a say when it comes to making daily purchases, and 60 percent say they have a say when it comes to visits to family or relatives. This makes it plausible to conclude that the women's influence in the household depends upon the importance given to the task and the money involved. The greater the importance given to the task and the more money involved, the less the influence of the woman. Figure 2A.8 Proportion of women who say that they alone or jointly have the final say in specific decisions 70 60 50 40 30 20 10 0 Making large purchases Making daily purchases Visits to family or relatives Source: MDHS 2004 58. It is worth highlighting that there are no clear relation between wealth30, and decision making for women. The shares reported that they are involved in the three decisions as listed are highest for those in the lowest wealth quintile (not shown). Neither do education seems to empower the women in this regard. Those with no education are more often involved in the specific decisions than those with education. The table show, however, that employment for cash seem to empower women. Those employed for cash are to a much higher extent involved in the decision. For example 40 percent of those women employed for cash said that they alone or jointly had the final say in making large purchases compared to the average 24 percent. The same table shows that participation in decision-making increases with age. This verifies the hypothesis that a woman's empowerment increases with age. Decision making in agricultural tasks 59. From the analysis in the section on participation in agricultural households, it was concluded that crops mainly used for own consumption could be classified as 'female' crops and cash crops as 'male' crops, even though no information was available on intra- household responsibilities for growing various crops. 30The Wealth index in the DHS is based on the respondent's household assets, amenities and service. 105 60. The IHS2, however, looked at who made decisions as to what inputs to use and the timing of planting crops on a plot. In female-headed household the woman of course normally makes those decisions. Within male-headed households, however, the picture is different. Figure 9 shows that only in a minority of cases do women make the important decisions on when to plant and which inputs to use for the crops. It is also shown that to the extent women are involved in such decisions, it is mainly on inputs and planting for leguminous crops such as ground beans, peas, beans and ground nuts and other cereals like sorghum, pearl millet and finger millet. Even for these crops the woman makes less than half of the decisions. These are crops that do not require fertilizer application and where most seeds are recycled. However, for local maize, where also the seeds are recycled, but where fertilizer is applied, women make about 10 percent of decisions. For cash crops such as burley tobacco, cotton and vegetables, men almost solely made the main decisions. These crops also normally require buying the seeds as well as the application of fertilizer and pesticides. This indicates that for crops that required spending money on inputs, men made the decisions on when to plant and which inputs to use. It would have been interesting to analyze the relationship between the labor input of women, their role in decision-making and finally, their control over the outputs. This type of information is unfortunately not available. Figure 2A.9 Percentage of women who make decisions on input use and planting 25 20 15 10 5 0 ze ze cco Local MaiHybridMai Cassavaroundnutset Potato Rice Cotton bles Toba Soybean legumeserCereals G Vegeta Swe Other Oth Source: IHS2 Women's contribution to household earnings and their control over own earnings 61. As noted by Sen (1985) and Wold (ed.) (1997), women's relative bargaining power depends on their contribution to the total household income. Thus, it is expected that employment and earnings would empower women if they perceive their earnings as important for meeting the needs of their household. However, women may not necessarily be empowered by their earnings if they have no control of their use. 62. The pattern of time use shown earlier, is also reflected in the earnings of women and women's contribution to the households' cash earnings. Their larger responsibility for domestic tasks reduces their possibilities to engage in gainful employment, and hence 106 may reduce their contribution to the household income. The gender disparities in time use therefore have repercussions for the bargaining power of women within the household. 63. Only 38 percent of all women had some cash earnings during the last 12 months31. Table 20 illustrates that the women's contribution to the households' cash expenditures. About one in five of the married women who earned any cash contribute nothing or very little to the household's cash expenditure, and about half contributed half or more of the household cash earnings. It is worth noticing that among the previously married women, only slightly more than one third contributed provided all cash income needed for the household's consumption. This group most likely consists of many female- headed households, indicating that they, to a high degree, rely on cash support from other household members or from other persons outside the household. There is no clear relation between share contributed and education or wealth index. Table 2A.20 Proportion of cash expenditure that the woman's cash earnings pay for, by marital status, education and wealth index Almost none/none Less than half Half or more All Marital status Never married 20 30 31 19 Married or living together 9 36 39 16 Divorced/separated/widowed 9 25 29 36 Education No education 10 33 31 25 Primary 1-4 11 33 35 20 Primary 5-8 12 33 37 18 Secondary+ 8 32 41 19 Wealth index Lowest 9 33 31 27 Second 6 38 31 23 Middle 11 32 36 21 Fourth 11 33 42 14 Highest 12 31 39 18 Source: MDHS 2004 64. To assess women's control over their earnings, the women who earned any cash income during the 12 months preceding the survey were asked who decided how their earnings should be used. Further, they were asked how much of their households' expenditure their earnings pay for. Table 21 shows that among the married women one out of three said that somebody else (mainly husband) decided about the use of their earnings. Note, however, that we do not have the corresponding information for men. Some men may perceive that the wife has the final say. As expected, never married and previously married women more often 31Unfortunately, at the time writing we do not have information on how these women are distributed by marital status. 107 made the decision on how their cash earnings should be spent. It should be noted, however, that about one fourth of the never married women say that someone else made the decision. This might be a reflection of the situation of unmarried daughters living in their parents' household. Table 2A.21 Women's control over own cash earnings, by marital status and type of location Percentage deciding how earnings are used Self only Jointly Someone else only Marital status Never married 70 4 26 Married or living together 38 28 34 Divorced/separated/widow 95 1 3 Wealth index Lowest 55 14 31 Second 46 18 35 Middle 42 19 39 Fourth 47 22 30 Highest 63 23 13 Education No education 48 13 38 Primary 1-4 46 21 33 Primary 5-8 52 21 27 Secondary+ 64 24 12 Total 52 20 27 Source: MDHS 2004 65. The control over own earnings is highest for women in the highest wealth quintile. The relationship between control over the earnings and wealth is, however, not clear. The relationship between education and control over own earnings are, on the other side clear, and education seems to empower women in this respect. 66. The data showed in this section, thus indicates that women, and especially married women, have very little cash income, and also contribute little to cover household expenditures, and as a result have little bargaining power within the household. 108 GENDER BASED VIOLENCE 67. Gender based violence is defined as any act that results in, or is likely to result in, physical, sexual or psychological harm, including threats of such acts and deprivations of liberty, whether occurring in public or private life, United Nations (1993). 68. Violence is a fundamental violation of human rights and freedom. It lowers mental and physical health for the victims, affects the opportunities to participate in the economy and has a detrimental impact on poverty. In addition to the direct cost of expenditure to health care, judicial and social services, there are costs related to lowering productivity of paid and unpaid work, Bott et al. (2005). A culture of silence has been surrounding gender based violence in Malawi, and it was not until the 1990s that the government, NGO's and international organizations started implementing various initiatives aimed at creating awareness on the prevalence and negative consequences of such violence. Recently, the Government of Malawi endorsed A National Strategy to combat gender-based violence (2002-2006). The National Strategy provides guidance to government institutions, civil society and donors who address issues of gender based violence within their programs. 69. Two recent surveys give data that can shed some light on the amount, and content, of gender based violence. IHS2 provides data on violence against women in general, while the Malawian Health and Demographic Survey provides data on domestic violence. 70. A point of departure for the analysis is that the lack of access to various types of resources, the lesser influence on decision making among women and the lesser bargaining power and empowerment of women demonstrated in earlier sections of this chapter, will make the Malawian women vulnerable to violence, both in general, and especially in the domestic sphere. Gender and security against crime 71. Security against crime is a basic human right. Fear of being exposed to violent acts will restrict a person's activities, whether the fear is realistic or not. Hence, it is more often the fear, than the experience of actual criminal acts that is debilitating to a person. Also, to which the police or other authorities can be relied upon to take action in case of a crime being committed is also an important aspect of safety. It is often assumed that women, both because of their lesser physical strength and of their lower status in general in many societies, will be more exposed to, or feel more vulnerable in relation to, criminal acts, whether inside or outside their homes, and whether the perpetrator is known or not, than men. 72. According to results from IHS2, both women and men feel most safe in the neighborhood during the day. They are feeling more unsafe at home, but most unsafe outside the home during the night, regardless of marital status. This indicates that married women do not fear for violence in their own home, compared both to married men and other women. However, women more often feel unsafe venturing outside the home 109 during the night than men do, 37 percent as compared to 29 percent (Table 22) indicating more fear from violence from outside the household than from inside the household, regardless of marital status. Table 2A.22 Proportion who feel unsafe in various locations by sex and marital status Unsafe in own home Unsafe in the Unsafe in the neighborhood during the neighborhood during the day night Malawi 14 4 33 Sex Male 13 4 29 Female 15 5 37 Marital status males Married 15 4 30 Separated/divorced 11 1 27 Never married 11 4 31 Widowed 14 5 27 Marital status females Married 16 5 39 Separated/divorced 15 4 38 Never married 12 5 38 Widowed 15 5 35 Source: IHS2 73. The fear of crime is greater than actually being a victim of crime. A very small proportion of the Malawian population had been attacked during the last year (Table 22) only about 4 percent. However, men are somewhat more exposed to this type of crime than women, 5 percent as compared to 3 percent. However, women are to a larger extent than men attacked either by somebody in the household or by other relatives, or by a neighbor, more than 50 percent, while men are more likely to be attacked by a stranger, 64 percent. Married and divorced women are more exposed to violence from somebody within the household or other relatives than other women and also than men. Widows are more exposed to violence from other relatives than other women. Women, regardless of marital status are most likely to have experienced attacks from a neighbor. 74. Hence, women seem to be more prone to attacks caused by known persons, and also to some extent by household members and relatives. This may reflect the low status that women have in many localities, and also that domestic violence against women is quite common. (See next section). Domestic violence 75. The traditional attitude of considering gender-based violence as an internal family matter makes collection of such data particularly challenging. Women who want to convey their experiences of domestic violence may find it difficult to do so, because of feelings of shame or fear. Complete privacy is essential for ensuring the security of the respondent and the interviewer. Asking about, or reporting, violence, especially in households where the perpetrator may be present at the time of interview, carries the risk 110 of future violence. Thus, one should bear in mind that these aspects may lead to an under- reporting of the incidence of domestic violence. Attitudes towards wife beating 76. Wife beating is a cultural phenomenon penetrating large parts of Sub-Saharan Africa. Thus, before discussing various aspects of domestic violence, attitudes towards wife beating will be analyzed. It is disturbing that nearly 30 percent of all women agreed to at least one of the following five reasons that would justify that a husband beat his wife; if she burns the food; if she argues with him; if she goes out without telling him; if she neglects the children; and if she refuses to have sex with him. The most accepted reason for beating is that the wife neglects the children. Married women have a somewhat greater tolerance toward domestic violence than either the never married and previously married (Table 23). Table 2A.23 Women's attitude toward wife beating. Proportion of women who agree that a husband is justified in hitting or beating his wife for specific reasons, by background characteristics Husband is justified in hitting or beating his wife if she: Goes out Neglects Refuses to Agrees with at Burns the Argues without the have sex least one food with him telling him children with him specified reason Marital status Never married 11 12 13 18 11 27 Married or living together 12 12 15 18 15 29 Divorced/separated/widowed 10 9 11 14 11 25 Wealth index Lowest 13 13 14 18 15 30 Second 13 13 15 18 17 32 Middle 13 14 15 20 16 32 Fourth 12 12 16 19 14 30 Highest 7 8 11 12 8 19 Education No education 12 11 13 16 15 28 Primary 1-4 14 13 14 17 15 30 Primary 5-8 12 13 16 19 14 30 Secondary+ 7 9 10 14 8 20 Age 15-19 14 15 16 21 14 32 20-24 12 13 14 18 14 30 25-29 11 11 15 17 14 28 30-34 9 11 12 15 12 24 35-39 11 10 12 14 13 26 40-44 10 11 14 16 14 27 45-49 10 8 12 14 13 25 Residence 111 Urban 6 7 10 11 9 18 Rural 13 13 15 19 15 30 Total 11 12 14 17 14 28 Source: MDHS 2004 77. However, on the positive side it should be noted that the acceptance of violence in Malawi, is at a (much) lower level than corresponding figures from some neighboring countries32. 78. There is a lower acceptance of violence among the wealthiest quintile. For women, there are, however not much difference among the remaining wealth. The data further shows that rural women are more likely to accept wife beating, and that only a quite high level of education, secondary or more, reduces the likelihood of accepting wife beating. Age reduces the acceptance of wife beating, but it is still widely accepted among women, regardless of age. Those results seem to confirm that wife beating is a widely accepted cultural phenomenon among Malawian women, and that access to resources and empowerment therefore will not alone reduce the tendency to accept this kind of behaviour within a marriage. 79. It is a puzzle that in general, women's acceptance of wife beating is higher than that of men. Only 16 percent of the men accepted that one of the mentioned reasons justified wife beating (Table 24). Quite interestingly, it is the young and the never married men who most readily justifies wife beating. If this can be taken as an indication of future attitudes towards wife beating, the acceptance of wife beating is on the increase among men. This is not a promising scenario for the future. 32Compare to; Zimbabwe 1999: 51 percent, Benin 2001: 60.4 percent, Uganda 2000/01: 76.5 percent, Ethiopia 2000: 84.5 percent, Zambia 2001: 85.4 percent, Mali 2001: 88.8 percent. 112 Table 2A.24 Men's attitude toward wife beating. Percentage of men who agree that a husband is justified in hitting or beating his wife for specific reasons, by background characteristics Husband is justified in hitting or beating his wife if she: Goes out Neglects Refuses to Percentage who Burns Argues without the have sex agrees with at least the food with him telling him children with him one specified reason Marital status Never married 7 12 10 13 10 24 Married or living together 3 5 5 6 5 12 Divorced/separated/widowed 5 8 6 6 11 15 Wealth index Lowest 7 12 8 11 10 19 Second 5 9 8 11 7 18 Middle 4 8 8 9 7 16 Fourth 3 7 5 7 5 15 Highest 3 5 7 6 6 14 Education No education 3 6 4 5 6 12 Primary 1-4 5 10 7 11 10 20 Primary 5-8 5 10 9 10 7 18 Secondary+ 3 4 6 6 4 11 Residence Urban 5 6 6 6 7 14 Rural 4 8 7 9 7 17 Age 15-19 8 14 11 14 11 28 20-24 5 12 12 13 9 22 25-29 4 5 5 7 4 12 30-34 3 5 5 6 7 13 35-39 2 4 5 5 5 10 40-44 1 2 3 3 2 7 45-49 2 5 3 2 2 5 Total 4 8 7 8 7 16 Source: MDHS 2004 113 The Extent of Physical Violence 80. Whether action follows from acceptance or acceptance just reflect action might vary, but not only is violence accepted, physical violence against women is quite prevalent in Malawi. One third of all women between 15 and 49 years33 reported that they have experienced physical violence since the age of 15, while one out of every five women reported that they had experienced physical violence in the 12 months prior to the survey (Table 25). 81. Given that physical violence can be a cause for divorce, it can be expected that women who are currently divorced/separated report the highest incidence of physical violence. Almost half of the divorced or separated women reported having experienced violence since the age of 15, as compared to one third of married women. However, no such difference was found when looking at the 12 months period preceding the survey. One in five women, whether presently married or divorced, had been exposed to violence. Widowed women seem to have been the least exposed to violence, both since the age of 15 and during the past 12 months. This may possibly be related to age. Widows tend to be older than married and divorced women, and the oldest women in the survey (40-49 years old), report less violence than those in the 'prime' marital ages (20-39 years of age) (Table 25). Table 2A.25 Proportion of women who has experience of physical violence, by background characteristics Since Age In last 12 15 months Martial Status Currently married 34 21 Divorced/separated 48 20 Widowed 15 2 Never married 23 11 Wealth index Lowest 35 20 Second 33 20 Middle 32 19 Fourth 33 19 Highest 31 14 Age 15-19 28 17 20-29 36 20 30-39 34 18 40-49 29 13 Education No education 29 16 Primary 1-4 36 22 Primary 5-8 34 19 33All the tables from MDHS 2004 are based on information from respondents in this age group only. 114 Secondary+ 30 13 Employment status Employed for cash 36 21 Employed, but not for cash 34 18 Not employed 30 17 Total 33 18 Source: 2004 MDHS 82. The incidence of violence seems to be highest among the poorest. Education, which could be an indicator for access to resources and empowerment, has no clear relationship with the experience of violence, neither in the present, nor in the past. One should not, however, necessarily expect a consistent negative relationship between violence and education of a woman, since high education implies higher status, which again might constitute a threat to men and thus higher incidence of violence, see also Bott et al (2005). That economic empowerment can be a threat is indicated by the higher proportion of victims of physical violence among the women that were employed for cash than the others. Perpetrators of Physical Violence 83. Husbands are the main cause of physical violence (Table 26). Two out of three currently married women who had experienced physical violence, reported that their husband were the perpetrator. Not surprisingly, the husband is reported to be the perpetrator to an even higher degree in those marriages that has broken up. Three out of four divorced and separated women who have experienced violence, report that the perpetrator was the former husband. Table 2A.26 Percentage distribution of women reporting physical violence since age of 15, by perpetrator and current marital status Perpetrator Current marital Current Last/previous Any husband and Persons other status husband* husband only other persons than husband Currently married 66. 2. 13 19 Divorced/separated na 8 12 13 Widowed na 19 7 74 Never married na na na 100 Source: MDHS 2004 *Includes women that were also beaten by a former husband na=not applicable 84. In the cases where the husband is the perpetrator, the wife is at greater risk of being exposed to violence if he often gets drunk, or he is older than the wife, or he has an education below secondary education,. Among the married women, the incidence of marital violence for a women living with a man that gets drunk very often is almost 50 percent (Table 27). However, the incidence of violence is high (25 percent) also in relationships where the man does not drink. If the wife is older than the husband, she is less likely to experience violence from the husband, and the same is the case where the 115 husband has an education above primary level. Whether the wife is more educated than the husband, does not seem to have an impact on her risk for being exposed to violence from the husband. Those results might imply that cultural factors, and not so much the relative bargaining power or access to resources, are important when analyzing domestic violence, especially when keeping in mind the relatively widespread acceptance of women of such violence. Table 2A.27 Proportion that never experienced physical or sexual violence by characteristics of husband Husband's education No education 69 Primary 1-4 68 Primary 5-8 69 Secondary+ 73 Missing 74 Difference in age between husband and wife Wife older than husband 3+years 82 Same age or 1,2 years different 68 3-4 years 71 5-9 years 70 10+ years 68 DK/Missing 61 Differences in education Husband has more education 70 Wife has more education 68 Both have equal education 71 Neither educated 72 DK/missing 74 Alcoholic consumption of husband Does not drink 75 Drinks, never gets drunk 65 Drinks, gets drunk sometimes 65 Drink, gets drunk very often 48 Total 70 85. Marital violence may be a result of a wife's "undesired" behavior, which the husband tries to control. The husband can try to control his wife in different ways, by restricting her access to money, restrict her access to family and friends and to control how she spends her time. Of course, jealousy may also be an important factor in marital violence. Figure 10 shows the extent to which married women perceive to be, or are, controlled by their husbands. 86. The most important forms of restrictions of social interaction are control over time-use and jealousy. About one in six married women said that the husband insists on knowing where she is at all times (control her use of time), and about half the women 116 said that the husband gets jealous or angry if she talked to another man. Only one in five of the married women did not perceive being controlled by their husband in the areas mentioned, and as many as one out of three women perceived that they were controlled in at least three of the areas mentioned. 87. The incidence of control decreases with education of the husband, from 32 percent among those with no education to 25 percent for those with secondary education and above; it also decreases with wealth from 35 to 24 percent. Figure 2A.10 Proportion of married women reporting that their husband restrict their social interactions, by type of restriction Doesn't trust her with money Insists on knowing where she is at all times Tries to limit contact with family Does not permit meetings with girl friends Accuses her of being unfaithful Is jealous / angry if she talks to other men Does none of these acts Does at least 3 of these acts 0 10 20 30 40 50 60 70 Source: MDHS 2004 Forms of Marital Violence 88. Marital rape appears to the most common violence against women as ten percent of all married women experienced forced sexual intercourse in the 12 months preceding the survey (Figure 11). Considering the danger of HIV, and the predominant pattern of unsafe sex and extramarital affairs among husbands (Chapter Five on HIV/AIDS) these numbers are even more alarming. Figure 2A.11 Proportion of ever married women who have experienced various forms of violence 117 Pushed, shaken, or thrown Slapped, twist arm Punched Kicked or dragged Strangled or burned Threatened with weapon Attacked with weapon Forced intercourse Forced sexual acts 0 2 4 6 8 10 12 14 16 18 percent Ever Last 12 months Source: MDHS 2004 89. Domestic violence is not a one-time experience for most women, indicated by the fact that there is a relative low difference between those that experienced violence during last 12 months and "ever". Of the married women who reported that they "ever" experienced violence about 30 percent said that they did not experience violence in the last year. 41 percent said that it happened one or two times last year, 20 percent said it happened three or four times and 9 percent said it happened more than five times. Help seeking for women who experience violence 90. Domestic violence has for a long time been treated as a private issue in Malawi, hence it was stated earlier that domestic violence to a great extent is a 'silent ' form of violence, where no help is sought and no course orders are issued. This is confirmed by the survey results. 91. Less than half the married women who experienced violence from their present husband sought some help, 41 percent. It seems that women felt freer to report incidences that occurred with their previous husband as compared with their current husband; 52 percent of the women who suffered violence by their earlier husbands sought help (Table 28). "Seeking help" is loosely defined to also include talking about the abuse to someone, and to the extent that abused women seek help, they mainly consult family or friends. This may be a result of the culture of silencing violence, and also of accepting violence. The victim may also think that help will be of no use and may fear further violence from the perpetrator if the abuse gets to be known. 118 Table 2A.28 Proportion of women who have ever experienced violence who sought help, by type of help sought 34 Persons perpetrating Percent who have Own Other violence sought help family In-laws relatives/friends Other Current husband only 41. 24 15 6 13 Earlier husband(s) only 52. 20 18 4 16 Husband and others 47 19. 10 8 19 Others only 37 58 1 4. 14 Source: MDHS 2004 92. The results might also imply that either the women have no knowledge about the legal institutions that deal with women's rights that these institutions are not readily available for abused women, or that they are not seen as of any help. In particular, rural areas might lack institution securing the abused women. The traditional village authorities rarely seem to settle cases of violence in favor of the woman, National Gender Policy (2004). FEMALE-HEADED HOUSEHOLDS 93. As discussed, the available data, does not allow for assessing poverty at the individual level. Thus, one is not able to analyze the effects on individual poverty related to different opportunities and allocation of resources within the household. 94. However, it is possible to analyze male headed and female-headed households, and to use sex of the household head as a proxy for sex differences on the individual level.35. In doing so, important to bear in mind that the group of female-headed households is a heterogeneous group consisting of many types of households, as well as that there are different reasons for them to be female-headed. This section seeks to identify vulnerable female-headed households, since this is one group that could be relatively easy to target for direct policy actions when it comes to poverty reduction and poverty alleviation. 95. One can expect to find particularly vulnerable households among the female- headed households, since in many of these households, the woman will bear the entire responsibility for generating income and taking care of the family. Another reason for studying female-headed households separately, is that their consumption pattern can reveal differences in priorities/necessities as compared to male-headed households. We have already seen that there is a stronger focus on education in female-headed household compared to male-headed households. 34Note multiple responses were possible in this question 35According to the enumeration manual for the IHS2 survey; "The Head of household is the person commonly regarded by the household members as their Head. The Head would usually be the main income earner and decision maker for the household, but you should accept the decision of the household members as to who is their Head." In general in Malawian households the male will be the Head, and the female will be the Head if the male is absent. 119 Poverty status by types of female-headed households 96. Almost one fourth of the Malawian households are female. The incidence of poverty is higher among female headed than among male-headed households. The proportion of poor and ultra-poor is 58 and 27 percent respectively in female-headed households and 51 and 21 percent when the head is male (Table 29). Table 2A.29 Poverty Headcount ratios by sex of head, and place of residence Poor Ultra- poor Malawi 52 22 Male-headed 51 21 Female-headed 58 27 Rural: Male-headed 55 23 Female-headed 61 28 Urban: Male-headed 24 7 Female-headed 32 11 Source: IHS2 97. As mentioned above, female-headed households are not a homogeneous group, and they may be female headed for different reasons. One way of defining various groups of female headed households, is to group them according marital status of the head into; the widowed; separated and divorced; those never married; and those that are married, but where the women for some reason is considered the head, e.g. women married to a polygamous or a monogamous man. In the latter case the woman will (most likely) be considered the head if the husband is absent most of the time. 98. The largest group of female-headed households are headed by widows (Table 30). The other main group is the divorced and separated ones, constituting about 40 percent of all female-headed household. Few female-headed households are headed by married women, whether monogamous or polygamous, and never married. Table 2A.30 Poverty status by material status of sex of household head Ultra- percent of Poor poor households Female-headed Divorced/seperated63 29 41 Married 58 29 10 Never married 17 5 3 Widow 56 25 47 Male-headed Divorced/seperated15 7 3 Married 52 22 92 120 Never married 9 4 3 Widow 30 6 2 All households Divorced/seperated59 27 12 Married 52 22 73 Never married 11 4 3 Widow 54 23 12 Source: IHS2 99. Female-headed households are more prevalent in rural than in urban areas. 24 percent as compared to 15 percent. This is due to the higher proportion of divorced and separated female heads in rural areas, and may be explained by increasing urbanization with men moving to town and leaving the wife behind. 100. Figure 12 below shows the incidence of poverty by marital status of the head of household. The incidence of poverty is higher in all types of female-headed households as compared to male-headed households. Figure 2A.12 Poverty headcount by type of household by sex and marital status of Head 70 60 50 40 30 20 10 0 Divorced/seperated Married Never married Widow Female-headed Male-headed All households 101. It is not surprising that the never married female heads have a lower poverty level since this group is expected to consist of younger well-off women studying or working away from their family. This is reflected in the relatively high education level among these groups, Table 31 below. They constitute however a very small group of all households (about 5 percent). Table 2A.31 Characteristics of female and male-headed household by marital status 121 Number of years of Dependency Average age of education Head Household size ratio36 Head Female-headed All 2,6 3,8 0,52 48 Divorced/separated 2,7 3,9 0,52 42 Married 3,4 4,6 0,48 38 Never married 8,1 2,3 0,24 26 Widowed 2,1 3,7 0,55 58 Male-headed All 5,4 4,7 0,43 41 Divorced/separated 3,9 1,6 0,25 45 Married 5,5 4,9 0,45 40 Never married 8,2 1,7 0,08 26 Widowed 4,4 2,5 0,45 57 All households All 4,7 4,5 0,45 42 Divorced/separated 2,9 3,3 0,46 42 Married 5,2 5 0,45 41 Never married 8,2 1,8 0,12 26 Widowed 2,4 3,6 0,54 57 Source: IHS2 102. The incidence of poverty does not vary much according to marital status among the remaining female-headed households groups (Figure 12) Among the male-headed household there is a much larger difference in poverty levels according to marital status. 103. The group of male-headed households headed by a widower has a much lower poverty incidence compared to households headed by widows. This could be attributed to property grabbing which is widespread in Malawi, though not well documented. When the husband dies, the relatives from the husband side take the property of the deceased without leaving the widow anything. If, however, the wife dies, the property is normally left with the husband. (See also the discussion on property grabbing in Chapter Five on HIV/AIDS). 104. Even though male headed households headed by a widower is better off than their female counterparts, and in spite of the tradition of property grabbing, households headed by widowed persons have a lower poverty level than other households, regardless of sex of household head. This could be explained by several factors, for instance by age. Widowed persons are in general relatively old (the average age being 58 and 57 years respectively for female and male widowers) and, thus one could expect them to have accumulated wealth or be supported by their own children. This is, however, not necessarily the case, because those households have a relatively high family responsibility reflected in the high dependency ratio. 36Dependency ratio is here defined as the number of dependants, i.e. those younger than 15 or older than 65 relative to all members in household. 122 105. The high dependency ratio can also be due to several factors, some of which are related to the widespread HIV/IDS prevalence. One such factor could be that the Head her/himself is a dependant below 15 years, another factor could be that many women/men, become widowed due to AIDS early in life when they still have young children. One would expect that households where one spouse died from AIDS to be particularly vulnerable due to possible additional burdens due to other AIDS sick member in the household. To throw some light on this issue, young widowed, defined as those below 40, were analyzed separately. Contrary to what we expected we found that the young widowed seems to have a lower poverty ratio than the older ones. Since information about the cause of death is not available, it was not possible to assess the role of HIV/AIDS related issues in determining poverty status in households headed by widowed persons 106. It is striking to see the difference in poverty levels between divorced/separated female and males. While the female-headed divorced and separated are worse off then the married ones, the male divorced and separated seem to be rather well off37. The large difference in poverty level between the female and male divorcees indicates that the woman tends to be the looser in terms of economical welfare after a divorce. It might be because the main breadwinner "grabs it all" and the woman is left without support while resuming the responsibility for the children. The considerably higher dependency ratio for divorced and separated women than separated and divorced men confirms that she tends to resume the responsibility for the children after a break up between the spouses. We do not have any explicit information on whether the wife receives transfers from the children's father. 107. Many of the divorced/separated from both female-headed and male-headed households consider themselves to be severely negatively affected by break up of the household over the past five years (46 and 44 percent of respectively separated and divorced female Heads and respectively 36 and 45 percent of separated and divorced male heads). Female-headed households with and without support 108. Even though a person is female headed it does not necessarily mean that the female head bear all the economic burdens of the households. For instance, widowed are among the better off female-headed households. One could expect that they, to a larger extent, get support from adult children. This leads to another classification of female- headed households, those who get economic support from outside the household and those who do not. In male-headed households there are normally at least two adult persons supporting the household, while in female-headed household there may often be only one person to fulfill the same tasks. In addition, we have seen in previous 37While 588 female Heads in the survey reported that they are divorced, only 110 male Heads said they were divorced. For separated the corresponding figures are 423 and 127 for respectively female and male Heads. These numbers indicate that most of the men that break up from a relationship do not live alone (for a long time); they may already be in a polygamous relationship or they quickly start to live with somebody else. 123 discussions, that women often do not have the same opportunities in the labor market as men and also have more domestic obligations than men. Thus one can expect that women without a husband or without any other adult support are more vulnerable than those having such support. 109. We define the group of female-headed household with support to include those where the head is married monogamously or polygamous (assuming that the husband in such situations in some way support their wife and the household); those receiving net transfer to the households, either from safety nets or from individuals outside the household, of more than 10 percent of their expenditure; and those having at least an adult male (between 18 and 60) in the household. The remaining female-headed households will be referred to as the no-support group. More than 60 percent belongs to the latter group. Hence, according to this definition most female heads have sole economic responsibility for their households as well as sole responsibility for most domestic tasks. Table 32 however, shows that there is no significant difference in poverty levels between these female-headed households with or without support. This indicates that dichotomy with or without support is not a good indicator of vulnerability, probably due to heterogeneity even within those groups. Table 2A.32 Poverty status of female-headed households with and without support Poor Ultra poor Without support 59 26 With support 58 27 Source: IHS2 Female-headed households and reliance on child labor 110. In female-headed households one could expect a larger extent of child labor, as a substitute for the lack of an additional adult to support the household. 111. There is a slight tendency that both girls and boys work more both in income generating activities and housework in female-headed household, the difference is however, not strong. Among those working, they work about the same number of hours. Figure 2A.13 Proportion of children between 5 and 14 that worked in income generating activities and housework, by sex of household Head 124 0,70 0,60 0,50 0,40 0,30 0,20 0,10 0,00 boy, male head hh girl, male head hh boy, female head hh girl, female head hh housework inc. gen. work 112. Neither do teenagers tend to work more in income generating activities in female- headed compared to male-headed household. The number of hours spent for girls and for boys in income generating activities is about the same in the two types of households (about 20 hours). Further, both girls and boys tend to help out more in the housework in female-headed than in male-headed households. This is in accordance with results showing that women in female-headed households spend on a daily basis on average half an hour less on housework than the woman in a male headed household (see Chapter Two). Thus they rely more on support from the children in order to be able to spend more time on income-generating activities They spend on a weekly average 2 hours more in income generating activities than women in male-headed households. Figure 2A.14 Proportion of children aged between 15 and 18 that worked in income generating activities and in housework, by sex of head 0,80 0,70 0,60 0,50 0,40 0,30 0,20 0,10 0,00 Girls 15-18, male Boys 15-18, male Girls 15-18, female Boys 15-18, head hh head hh head hh female head hh housework inc.gen act SUMMARY AND RECOMMENDATIONS 113. Gender disparities are prevalent in many areas of Malawian society. As documented in this report the literacy levels of women are considerable lower than those 125 of men. Only about half of the adult female population, could read and write in their mother tongue or English, compared to three fourth of the men. Currently there is a one to one ratio in the school attendance in primary school, which may tend to equalize the gap in literacy ratio over time. There is, however, a significant gender gap in the education attendance rate in secondary school. 114. Education is a key factor for empowering women could bring desired changes to many of the gender disparities described above. We have seen that education level having an even strong positive impact on welfare in the households. Thus, greater attention needs to be paid in addressing existing disparities in levels of literacy and education. Education increases knowledge and understanding of development issues. Infant mortality, maternal mortality and family size decrease with increase in education levels of women. Also women's education has a positive impact for children's health and upbringing. Educated parents, and in particular educated women, tend to give a stronger priority to educating their children. Despite of the heavy burden on single women in taking care of their family, they choose to educate their children to a higher extent. In particular when women are making the major decisions in a household they tend also to give a higher priority to girl's education. Thus this may portrait the positive implication of empowering women to bring upon desired changes in terms of education for the next generation. 115. Even though women work longer hours than men, they spend considerable less time in income generating activities. The gender disparity extends to girls and boys, with girls working a lot more at home. This division of labor has a negative impact on girls' school attendance and learning. More than one third of all girls who dropped out of school indicated that the main reason is that they are needed for work at home. 116. Education is one of the main factors influencing women's participation in, and choice of, wage employment. Also, the higher the education, the better the remuneration. A much lower proportion of women than men are employed for a wage. On the other side, a higher share of the wage-employed women works in professional jobs. This highlights the importance of closing the gender gap in educational attainment in order to expand employment opportunities for women. Also, there continues to be a considerable wage gap in across gender in lower paid occupations. 117. Agriculture is the one sector where women are involved to the same extent as men. However, women continue to face lower access to, and control over, production factors such as land, agricultural inputs, and technology. More support should be given to women in agriculture for instance by increasing the number of female agricultural extension workers, to increase the likelihood that female farmers are assisted on equal footing as male farmers. While men are involved in cash crop production, women are mainly involved in growing the food crops. Outside household agriculture women still continue to be primarily engaged as self- employed in small, informal sector. Small-scale enterprises are an important source of income for many households. However, women need to be encouraged to engage in more productive enterprising activities and credit should be made available for these purposes. 126 118. There is a strongly gendered division of labor within the household, with the women doing most of the domestic tasks including collecting water and firewood. Thus, providing water facilities and access to other energy sources than firewood would be a large improvement in everyday life for most women. 119. Protection from violence is a fundamental human right. Domestic violence lowers women's self-esteem and erodes their mental health. It is alarming that one out of five married women experienced physical violence in the year preceding the survey, and in most cases it is the husband that is the perpetrator. Most of these women did not seek help, which may be because violence is considered acceptable to many, both men and women. This may be due to the culture of silence surrounding domestic violence and lack of availability and knowledge of legal institutions and support systems for the victims. The expansion of the ongoing awareness programs to empower the women and strengthen activity to reduce such violence is clearly of uttermost importance. The analysis carried out in this chapter indicates that when analyzing domestic violence cultural factors are much more important than relative bargaining power or access to resources. 120. There are reasons to expect that women's low bargaining power affect their share of benefits and obligations within the household. We have seen that many women cannot influence decisions highly relevant to themselves and to their family. Women's low share of contribution of cash to the household gives them a low bargaining power. This may also be a rational consensual decision based on market and shadow prices the household is facing. However, in one third of the cases when she contributed, someone else (mainly the husband) decided how the earnings are spent, which are not likely to be in accordance with her preferences. 121. Women have triple burdens with the responsibility for housework, childcare and as contributors to income and food. These responsibilities are particularly heavy for women managing a household alone and these households tend on average to be poorer. A vulnerable group among the female-headed households is the divorced and separated, in particular when compared to the same group of male-headed. Hence, special focus should be given to these households, which are likely to be particularly needy for support. 122. There is a highly seasonal market for agricultural ganyu labor in Malawi. As a result one can expect female-headed household depending on seasonal ganyu labor to be particularly exposed for food-shortage and poverty, because of the lack of alternative employment opportunities during the rest of the year. 123. Further, the seasonal labor pattern likely results in lower agricultural productivity, as the poorest are forced to engage in ganyu labor (in order to earn some cash) rather than spending sufficient time on own fields. The seasonality in agricultural labor market is likely to have a particular negative implication for single farming women. In fact, the seasonality in the labor market, together with the fact that women have fewer 127 employment opportunities than men over the rest of the year,38 implies that they are to a higher extent 'forced` to prioritize cash earnings at the time when their labor are most valuable at their own fields. In addition they are already being constrained with time due to the additional burdens in the household, and are therefore less able to utilize their land for productive production. 124. These observations are in accordance with the findings that female-headed households engage in cash cropping to a much lesser extent than male-headed households and that they rarely borrow to invest in agriculture, or other activities. The impact is a disproportionably higher rate of poor among the female- compared to the male-headed households with small landholding sizes. 38In fact, the gender gap in time spent in income generating activities is lower during the main cropping season (October-December). 128 TABLE APPENDIX Table 2A.33 Regression explaining share of expenditure used on education No. Observations 11218 R2 adjusted 0.183 Variable Estimate St.error Intercept -0,056 (0.005)*** If female-headed 0,008 (0.0011)*** Children below 15 years 0,002 (0.00021)*** Boys (15-18 years) 0,009 (0.00056)*** Girls (15-18 years) 0,009 (0.00056)*** Men (19-65 years) 0,008 (0.00042)*** Woman (19-65 years) 0,006 (0.00047)*** Age of Head 0,0002 (0.000016)*** Log expenditure per capita 0,004 (0.00043)*** Years education woman39 0,0005 (0.0001)*** Years education man 0,001 (0.000083)*** is female spouse in household -0,006 (0.000099)*** if rural -0,004 (0.00080)*** Dummys for agro-ecological zones Source: IHS-2 * Dependent variable: Share of household expenditure on education. Table 2A.34 Proportion of women and men that worked in the various income generating activities and hours worked for these, by sex, poverty status and place of residence Agriculture activities Own business Ganyu Worked for wage % who % who % who % who did this Average did this Average did this Average did this Average work last hours work last hours work last hours work last hours week worked week worked week worked week worked Sex Male (above 14) 57 22 14 29 15 17 15 44 Female (above 14) 63 20 10 20 9 14 4 36 Non-poor Male (above 14) 50 22 17 31 12 18 19 44 Female (above 14) 55 20 12 22 6 14 5 38 Poor Male (above 14) 66 22 11 24 17 17 11 44 Female (above 14) 71 20 8 18 12 15 2 29 Rural Male (above 14) 65 22 14 25 16 16 12 41 39 If female headed this will correspond to the head, if male headed it will correspond to the Head's spouse. 129 Female (above 14) 69 20 10 18 9 14 3 30 Urban Male (above 14) 11 19 16 47 8 30 37 49 Female (above 14) 14 19 12 36 4 20 13 45 Source: IHS2 Table 2A.35 Time-use for household agricultural activities last week Male (above 14) Female (above 14) Proportion who Average weekly Proportion who Average weekly worked in hours among those worked in hours among those agricultural that did agricultural agricultural activities that did agricultural activities last week work last week work Sex 57 22 63 20 Poverty status Non-poor 50 22 55 20 Poor 66 22 71 20 Location Rural 65 22 69 20 Urban 11 19 14 19 Marital Status Divorced/separated 54 23 67 20 Married 63 24 68 21 Never married 50 17 53 17 Widowed 47 19 62 20 Source: IHS2 Table 2A.36 Proportion of farming households who cultivate different crops by sex of household head Female-headed Male-headed Type of crops household household All households Local Maize 62 50 53 Hybrid Maize 40 52 49 Ground nuts 19 25 24 Sweet Potato 8 13 12 Rice 6 7 7 Cassava 7 11 10 Tobacco 7 19 16 Soybean 4 5 4 Cotton 1 3 3 Vegetables 2 4 4 Other Cereals 3 3 3 Source: IHS2 Note: Includes only rainfed cultivation.Vegetables include Cabbage, onions, nkhwani, okra and tanaposi and sugarcane. Other cereals include pearl millet sorghum and finger millet. Other legumes include beans and peas. 130 Table 2A.37 Economic activity and payoff40 by sex and place of residence Percent Employed Percent doing Median Percent Operating some times last 12 Median Ganyu some times Ganyu non-agricultural Median months wage/day last 12 months wage/day enterprise last month profit/day Malawi Men 22 124 35 52 16 279 Women 6 78 26 36 10 159 Urban Men 39 416 17 294 16 439 Women 15 514 10 199 14 102 Rural Men 19 129 38 112 15 255 Women 5 117 28 72 9 125 Source: IHS2 Table 2A.38 Percentage distribution of enterprise types by sex of owner41 Male Female Fishing 7 2 Baking 0 7 Distilling 1 12 Beer brewing 1 10 Handicraft manufacture 11 3 Furniture 3 0 Food and beverage retail 5 6 Cloth retail 3 2 Other retail 36 34 Street food 6 10 Other 25 16 Total 100 100 Source: IHS2 Table 2A.39 Main Source of Start up Capital Male Female Family/friends 5 8 Gift, family/friends 9 21 Sale of assets 4 3 Proceeds from other business 9 5 Own savings from Ganyu 16 12 Own savings from agriculture 25 23 Own savings, non-agriculture, non-Ganyu 10 7 Non-agricultural credit, bank or other institutions 2 3 Money lender 2 3 Inheritance 5 2 Other 0 1 None 15 14 Total 100 100 Source: IHS2 40All monetary values have been deflated 41Only those enterprise types that sum up to two percent or more are listed, the others are included in the "others" group 131 ilyade t en oseht agervA spemti estunim, ong am collecting oodw 77 88 85 19 77 88 85 91 60 71 80 86 81 91 83 87 75 86 87 132 atht oodw tionr week opo llected Pr co last 3 8 4 12 3 9 4 23 0 1 1 6 3 7 4 19 4 9 3 ilyade es t en oseht nutim agervA spemti ong r,e am collecting wat 48 57 52 27 48 58 53 73 45 48 46 55 48 57 50 70 48 57 55 tionr last collected opo Pr that water week 18 47 11 37 19 50 11 77 13 26 13 47 19 46 13 70 18 48 9 day ilyade t en oseht k,r per agervA spemti wo es ong ng use am doi ho nutim 75 78 96 241 0 4 8 2 3 7 60 73 91 14 13 10 11 16 79 81 97 14 71 76 93 timeuse kr tionr wo and opo ng week use Pr doi ho last 17 38 18 58 15 36 16 86 35 52 30 80 19 42 22 85 15 35 14 e ci activities, agervA ilyad onemit est es , dom task nutim 25 66 28 991 3 6 8 22 65 25 20 56 74 43 16 28 72 33 19 23 62 22 mestic s do tionr ci ng re) last udi ca opo ng est Pr doi dom tasks week (Excl child 30 60 25 19 28 60 23 92 42 60 33 83 32 63 28 91 28 58 20 various did ) that )41 14 4)1 4) 5-( -15( na na na na Proportion oyB rl oveb(ana above(namo Gi M W oyB rl na Gi M Wom oyB rl na y rl na Gi M Wom Bo Gi M Wom oyB rl na Gi M Wom 2A.40 or n po- dsolh dsolh IHS2 l Table Al uralR ba Ur onN use or use ho Po ho Source: Table 2A.41 Time use on main activities by sex of head Proportion of Average Proportion of Average time Average hours persons who time spent persons who spent weekly worked per did domestic weekly did income among those week, income tasks among those generating doing inc. Gen generating and (Excluding doing activities activities domestic child care) housework Female-headed Boy (5-14) 0.32 10 0.32 12 7 Girl (5-14) 0.64 13 0.30 10 11 Man 0.35 14 0.64 25 20 Woman 0.87 23 0.77 26 40 Male-headed Boy (5-14) 0.29 10 0.29 11 6 Girl (5-14) 0.59 123 0.25 11 10 Man 0.23 13 0.82 34 31 Woman 0.92 27 0.72 24 41 Source: IHS2 Table 2A.42 Proportion of persons being attacked last 12 months and type of perpetrator Proportion Type of attacker being Household Other Neighbor Stranger attacked member relative Malawi 4 4 14 25 58 Sex Male 5 1 12 24 64 Female 3 9 18 27 46 Marital status males Married 7 - 13 22 66 Separated/divorced 7 - 10 29 67 Never married 4 2 10 33 55 Widowed* 1 - - - - Marital status females Married 3 11 20 24 46 Separated/divorced 4 10 16 26 48 Never married 3 6 11 40 44 Widowed 3 - 28 27 45 *Only 3 observations Source IHS2 Table 2A.43 Women's participation in decision making by background characteristics. Proportion of women who say that they alone or jointly with other household member(s) have the final say in specific decisions. Alone or jointly have final say in: Making large Making daily Visits to family or purchases purchases relatives Marital status Never married 9 10 21 Married or living together 18 33 60 Divorced/separated/widowed 77 80 86 133 Wealth index Lowest 34 43 63 Second 21 30 59 Middle 20 30 55 Fourth 20 31 56 Highest 26 39 52 Education No education 28 38 63 Primary 1-4 21 32 58 Primary 5-8 22 34 55 Secondary+ 26 34 49 Employment Not employed 17 28 46 Employed for cash 40 53 68 Employed not for cash 24 34 64 Residence Urban 28 43 53 Rural 23 33 57 Total 24 35 57 Source: MDHS 2004 134 fo 135 types fferentid of erb ne 65 30 35 79 37 89 60 50 27 71 61 59 5 45 68 54 14 11 14 16 17 16 15 22 22 26 92 80 Num wom none demonstrates(d) 22 22 19 11 15 19 22 21 23 20 19 21 23 20 these Does of acts husband at of3 acts 30 29 33 32 35 33 30 28 24 32 31 30 25 30 Does eastl these tn' 19 18 21 16 21 20 19 17 14 19 19 18 14 18 current/last 2004i rehts Does tru with oneym the on ing Malaw ehser all that Insists nowk at whe is esmti 59 58 63 45 61 59 55 58 54 57 59 56 57 57 lyi to fam 20 20 21 20 21 23 20 17 18 21 20 19 19 20 reported :dn Tries mitli contact with who characteristics, ba hus ton nd it ngsi rlig ds 19 19 21 19 21 21 20 17 16 20 20 19 16 19 omenw osehwne Does ermp eetm with frien l fu backgrou wom fo gn 17 16 18 21 22 20 18 15 12 19 19 16 11 17 married to of her bei faithnu ever geatn /Accuses ify tos ofn jealous gr talke nemr 51 50 55 45 52 52 47 49 50 50 50 51 49 50 according Perce Is an sh heot ceno Proportio behaviors, anht 2A.44 sutatslatira ceno reom deirramy ed ed ed rria rria rria uslovi xednihtlae + nd e h west ddli urt hestg Table controlling M M M M Pre W Lo Seco M Fo Hi noitacudE noi 1-4 5-8 ry ucat ed arym arym nda la No Pri Pri Seco Tot Table 2A.45 Percentage distribution of female-headed households by marital status by Head Rural Urban All 24 15 Marital status Divorced/separated 42 34 Married 10 8 Never married 5 4 Widow 17 12 100 100 Source: IHS2 Table 2A.46 Activities carried out by 15-18 years olds, and average time utilized for the activity among those who did it, by sex of household head. Income generating activities Housework Proportion that did Average hours Proportion that did Average hours this work last week worked last week this work last week worked last week Male-headed Girls 15-18 not in school 77 27 65 25 in school 44 11 64 16 Boys 15-18 not in school 80 29 53 22 in school 51 13 48 13 Female-headed Girls 15-18 not in school 73 26 75 22 in school 52 13 70 17 Boys 15-18 not in school 75 26 59 20 in school 55 15 57 14 All Girls 15-18 not in school 76 26 67 24 in school 46 12 65 16 Boys 15-18 not in school 78 28 54 21 in school 52 13 51 13 Source: IHS2 Table 2A.47 Educational qualification by sex of head Sex of Head Female Male Educational qualification Never attended school 52 21 None42 38 52 PSLC 4 11 JCE 3 9 MSCE and above 2 7 Source: IHS2 42Attended school but failed to pass any exam. 136 ANNEX 2B: FORESTS, BIOMASS USE AND POVERTY IN MALAWI INTRODUCTION 1. Household well-being and livelihoods critically depend on the location of the household, and the environmental and resource constraints it faces as a result. Poverty and forest degradation appear to go hand in hand in Malawi. Forest degradation by all accounts is rampant. A third of Malawi is considered forested, but these patchy forests have been declining at a rapid rate. Between 1990 and 2000, the area under forest cover in Malawi decreased at a rate ranging from 1% to 2.6% per annum (FAO 2005, WRI 2003).43 In a country where 90% of the poor live in rural areas and share space with and use forests and shrub lands, it is not surprising that many express concern about the effect of poverty on the environment.44 2. Fuelwood scarcity in Malawi can affect household welfare in different ways. Poor households in particular may not have access to alternate energy resources and may increase the time spent on fuelwood collection, reducing time on productive activities, leisure or household care. An interesting recent report by Nankuni (2004), for example, shows that stunting in children in Malawi is associated with the long periods of time spent by women in collecting water. Households may also reduce their use of energy and switch to lower quality wood (Brouwer et al. 1997). This may limit their ability to cook or light their homes. More generally, forest cover losses can also reduce the availability of other forest products such as fruit, mushroom, poles, bush meat etc, and can have long- run effects on local hydrological services and agricultural output. 3. This chapter investigates the implications of biomass scarcity for livelihoods, and household activities. Specifically, the chapter discusses the relationship between fuelwood availability and poverty in Malawi. Biomass availability is used a proxy for fuelwood availability in this study. The analysis examines the hypothesis that biomass scarcity in Malawi has a `net' negative effect on household welfare. We test this hypothesis for different regions in the country and for rich and poor households. In a recent paper, Monica Fisher (2004) studies three villages in southern Malawi and shows that forests can prevent poverty by supplementing income and may even help raise standards of living. This chapter focuses on biomass availability and asks similar questions for the whole of Malawi. 4. Households of course engage in efforts to `manage' biomass scarcity. From a practical and policy perspective, it is important to understand how households respond to fuelwood depletion.45 In this chapter, our focus is to investigate the increase in time 43Estimates of deforestation in Malawi done by different global organizations differ. 44Note a recent New York Times articles (Nov, 1, 2005) by Michael Wines entitled "Malawi is Burning and Deforestation Erodes Economy." 45Households cope with scarcity in many ways: some plant trees, others increase the time allocated to collection, and still others decrease consumption or change cooking practices. Strategies may include improvements in fire management practices, sharing fires for cooking, preparation of fewer meals or changes in diet favoring fast-cooking (Dewees 1989). 137 spent on collecting fuelwood, which is a frequent and important response to biomass scarcity (Cooke 1998a, 1998b, Kumar and Hotchkiss 1988, Dewees 1989). Fisher et al. (2005) show that households in Malawi increase labor allocated to forest extraction as returns to these activities increase. They argue that with increasing wood scarcity, the returns to forest labor and time allocated to forest product extraction are likely to increase. As elsewhere, Malawian households use other coping strategies as well: they switch to lower quality wood, economize on wood use and increase the number of collectors (Brouwer et al. 1997). The analysis presented in this chapter uses this broad understanding of household responses to fuel wood scarcity to carefully examine how forest-dependent households allocate labor. We focus on time spent on fuel wood collection and ask whether scarcity has an effect on collection time and on labor allocated to agricultural tasks. 5. The chapter is structured as follows. Next section attempts to assess the extent and distribution of biomass availability for meeting the energy needs of the poor in Malawi, and provides a descriptive analysis of fuelwood use in Malawi. Section 3 explores the extent to which poverty and forest degradation are inter-linked and what are the factors that affect this relationship. In other words the section presents the analysis of the determinants of household poverty and the role played by fuelwood scarcity. Section 4 examines how households cope with scarcity. It investigates the impact of fuelwood scarcity on household coping strategies. In particular, the analysis explores whether women and men spend more time in fuelwood collection activities and less time in agriculture in response to fuelwood scarcity. Section 5 presents conclusions. BOX XX: ARE THE POOR VICTIMS OR AGENTS OF ENVIRONMENTAL DEGRADATION? Poor households often depend on environmental resources for their survival. Fertile soils help maintain crop yields, forests offer food, medicines and energy, and pollution affects the health and productivity of the poor (Jodha 1986, Cavendish 2000, Ezzati and Kammen 2001, WRI 2005). Poverty can, arguably, also be a cause of environmental losses (World Commission on Environment and Development 1987, World Bank 1992). Unsustainable use of renewable resources ­ deforestation, soil erosion or over-fishing ­ can also result from the livelihood practices of poor people. Quite often rational decisions based on limited information can contribute to depletion of natural assets. This dual relationship between poverty and natural resource changes has sparked a discussion about whether the poor are victims or agents of environmental degradation (Duraiappah 1998, Ekbom and Bojo, 1999). Another strand within this debate advances the idea of a circular mutually reinforcing relationship between poverty and degradation (Nerlove 1991, Dasgupta 1995). Given the complexity of the linkages between poverty and the environment, the debate about whether the poor are victims or agents of environmental degradation is perhaps less important than what factors might jointly or independently contribute to environmental pressure and poverty. The relationship between the poor and natural resources is mediated by a number of micro and macro factors such as labor and credit markets, property rights, information about best practices and so on (Duraiappah 1998, Bluffstone, Adhikari 2005, Fisher et al. 2005b). Thus, the nexus is endogenous to a given geo-political system and dependent on prevailing institutions, infrastructure and policies. While under varying circumstances, it may be optimal for poor people to mine natural resources, if this degradation leads to greater poverty, then environmental management becomes an important instrument for poverty reduction. 138 EXTENT AND DISTRIBUTION OF BIOMASS AVAILABILITY, AND FUELWOOD USE IN MALAWI 6. Loss of forest cover in Malawi is attributed to agricultural expansion, biomass use for fuelwood, charcoal production and a number of other factors including tobacco curing, brick burning and so on (Malawi SDNP 1998, Minde et al. 2001). While land under agriculture in Malawi IS FUELWOOD HARVESTING BAD FOR THE ENVIRONMENT? increased by some 31 percent As with much of the debate on poverty and environment, the between 1972 and 1990 impact of wood fuel harvesting on wood resources and forest (Eschweiler 1993), this type of health is not straightforward (Dewees 1989, 1995). Forest expansion may have slowed degradation can occur as a result of un-sustainable use to meet because of the limited supply of livelihood needs. In a 46 country based study, Barnes et al. cultivable land. In 1998, (2001) show that widespread tree removal and forest depletion around urban areas can occur as urban energy demand grows. tobacco and tea curing was But this demand slows as cities get wealthier. Further, as estimated to account for 30 Arnold et al. (2003) point out, that the extent to which fuel percent of total wood demand, wood harvesting contributes to deforestation or degradation and 70 percent was estimated to depends on factors such as source of demand and supply, be for household energy use nature of fuelwood and charcoal markets, household practices and so on. Wood fuel scarcity need not contribute to (Malawi SDNP 1998). Over deforestation as long as entire trees and root stock are not ninety percent of total energy destroyed. In fact, as Dewees (1989) argues, when harvesting demand is met by biomass. involves hacking of branches and collection of dead wood, this With population growth in can lead to increased productivity of woodland areas. Malawi being one of the highest Nonetheless, in the case of Malawi, there is both evidence of deforestation and concern about the implications for fuelwood in Southern Africa, this demand availability (Walker 2004, Malawi SDNP 1998). is considered a serious threat to forests (Malawi SDNP 1998). 7. The next question is what feeds this concern. Is it simply a reflection of long- standing global environmental worries related to the gap between fuelwood supply and demand, which is echoed by Malawian officials (Dewees 1995, Walker 2004)? Or, is fuelwood scarcity sufficiently high that it affects household wellbeing? Physical scarcity of wood fuel does not always mean that households perceive it to be scarce. Whether wood fuel is economically scarce will depend on household constraints and whether and how they are able to respond to physical scarcity (Dewees 1989, 1995). Farmers in Malawi do seem to consider fuelwood scarcity a problem. In a recent paper by Walker (2004), 92 percent of the farmers he interviewed in Southern Malawi were aware of decreasing tree availability and 94 to 100 percent of them expected this to cause problems for meeting future needs. 8. In this chapter we combine biophysical data from satellite spatial information with household data from the 2005 IHS2, to assess biomass availability in Malawi (Box XX). In order to understand the relationship between poverty and fuelwood use, we estimate biomass availability in Malawi based on satellite data (Box XY). In order to estimate biomass at the year 2004, we need to identify the relationship between remotely sensed data and biomass. For this purpose, we use a study undertaken by the Ministry of Forests and Natural Resources, Government of Malawi in 1993: Forest Resources 139 Mapping and Biomass Assessment for Malawi. The report provides detailed land cover maps for Malawi derived from visual interpretation of Landsat remotely sensed data. The land cover maps describe land cover status in 1990 and 1970 and provide a comparison of changes over this period. Based on field surveys, the report estimates total biomass and average biomass volume per hectare for different land cover classes at the district and regional level.46 We used the data available in this report to help us model the relationship between biomass and remotely-sensed data. BOX XX REMOTE SENSING AND BIOMASS DATA Several sets of remotely sensed data were used to estimate biomass distribution in Malawi. The first is 8 days composite data recorded by the Moderate Resolution Imaging Spectroradiometer (MODIS) that are atmospherically corrected and cloud cleared.1 These data, which provide average surface reflectance over 8 days with a spatial resolution of 463.3 meters, are used to map the amount and spatial distribution of standing biomass at the year 2004. The second is a set of Orthorectified Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper (ETM) data for the year circa 1990 and circa 2000 respectively.2 These data provide surface reflectance at the resolution of 30 meters and are used to accurately identify training samples for the estimated biomass model. The MODIS Continuous Field of View (MOD44B) product provides information on vegetation density and is used to separate forested land from areas of non-forest. Elevation information for Malawi is extracted from the Global Digital Elevation Model (GTOPO30) with a horizontal grid spacing of 30 arc seconds (about 1 km).3 Notes: 1. MODIS data are available at http://delenn.gsfc.nasa.gov/~imswww/pub/imswelcome/ 2. Landsat images are available at http://www.landsat.org/ 3. GTOPO30 was developed at the U.S. Geological Survey's Center for Earth Resources Observation and Science (EROS). 9. Optical remote sensing measurements reflect the interaction between radiation and forest canopies and are influenced by the canopy's structural properties. For this reason remote sensing data have been extensively used to map land cover and forest structural variables such as forest density and timber volume (Franklin 1986, Cohen 1992, Puhr 2000). Remote sensing also has the potential for providing spatial and temporal information on biomass availability because forest characteristics such as crown size and forest density are correlated with biomass (Shugart 2000, Franklin 1986). New techniques and the availability of remotely sensed data with higher radiometric and spatial characteristics such as MODIS, now make it possible to assess biomass availability at the regional and local scale (Baccini et al. 2004). 10. Our estimates of biomass or forest area estimates are not directly comparable with the estimates provided in the Government of Malawi 1993 report and cannot be used to assess land cover change between 1990 and 2004. This is because we used different definitions for forested land and the remote sensing data have different spatial characteristics. 46Because one of the goals of the project was also to characterize the land cover classes in terms of woody biomass, defined as "the volume of trees including bark from ground to top including branches with a diameter above 2 centimeters over bark" the project carried out a biomass assessment in the evergreen forest, Brachystegia forest in hilly and flat areas, in intensive agriculture land and, in extensive agriculture in forested land. 140 HOW DO WE ESTIMATE BIOMASS (FUELWOOD) AVAILABILITY FROM SATELLITE IMAGERY? A variety of techniques are used to classify vegetation and forest structure from satellite data. Unsupervised clustering algorithms and parametric supervised algorithms such as the maximum likelihood algorithms are the most commonly used methods. Vegetation structures such as forest canopy density and above ground forest biomass are commonly analyzed using linear statistical models based on ordinary least squares. These methods present limitations when the dependent and predictor variables are not linearly related, or when the predictors include high-dimensional input data. Further, multi-collinearity among the regressor variables can result in misleading inferences. Alternative methods such as neural networks and tree based-models overcome these limitations, and have been shown to be efficient in dealing with complex non-linear relationships (Friedl 1997, Gopal 1998). In this study, we derive biomass estimates for the year 2004 by applying a regression tree algorithm to identify the relationship between the output variable, biomass, and two predictors: remotely sensed data and elevation. Tree-based models have been previously used to predict both categorical (Friedl 1997) and continuous variables (Michaelsen 1994). The basic theory behind such models is reported in Breiman (1984).1 For the work reported here, the specific methodology is referred to as a regression tree, because the model is predicting a continuous-valued response variable. The regression tree is a supervised algorithm that uses a set of examples to identify the relationship between the regressor (biomass) and the predictors. In order to identify examples to train the algorithm, 1990 Landsat data similar to the ones used in the Department of Forestry, Government of Malawi, report (GOM 1993) were first analyzed. Regions with vegetation formations were selected from the Landsat 1990 scenes and then compared with the scenes for the year 2000 to verify that no changes occurred within the area selected. If changes occurred the sample was discarded. A last check was then performed using MODIS 500 meters data for the year 2004. If there was no discrepancy between the Landsat data for the year 2000 and the MODIS data for the year 2004, the area was selected as training area.2 For each training area, the spectral reflectance derived from MODIS 500 meters and the elevation data were extracted and used for the training of the algorithm. During this phase the algorithm identifies the relationships between above ground forest biomass and the predictors (remotely sensed and elevation data). Once the relationships are defined, they were applied to the entire set of 2004 data to predict forest biomass for the area where remotely sensed and elevation data were available. An important step in assessing biomass is to define and identify areas as forests and non-forests. To separate forested from non-forested areas, we used a MODIS product called MOD44B (Vegetation Continuous Field) that provides information on the density of the vegetation defined as ground cover from 0 to 100 percent. The combined visual analysis of the vegetation density product with the high resolution Landsat data indicates that a threshold of 20 percent best separates forested from non- forested land.3 We estimate the amount of forest area in Malawi as areas with vegetation density (according to the MOD44B) above the threshold of 20 percent. We estimate biomass density for the same areas. The regression tree algorithm allows us to estimate biomass per unit area for Malawi in all 463.3 m x 463.3 m pixels. Notes: 1. Tree-based models (i) make no assumptions regarding the distributional properties of the input data, (ii) are able to capture nonlinear relationships between the response and predictor variables, (iii) and provide easily understandable outputs. 2. We extracted MODIS observations taken at the beginning and end of the drying season. At the beginning of the drying season most vegetation still have leaves on, while at the end of the dry season only evergreen trees have leaves on them. This information is useful in separating evergreen vegetation from deciduous vegetation. 3. Setting the tree density threshold to 20 percent, may result in a value of zero biomass for regions characterized by a mixture of agriculture and trees. 141 11. Figure 1 and Table 2 present a picture of biomass and forest cover in Malawi.47 The total amount of biomass in the country is 388 million cubic meters. The average biomass available is about 40 cubic meters per hectare; however, this number does not present an accurate picture because of the skewed distribution of biomass. As Figure 1 indicates, Malawi's forests are mainly in the Northern areas. 12. The northern region with 159 million cubic meters of biomass has 41 percent of the country's biomass. The central region has 38 cubic meters per hectare and 38 percent of the biomass. The southern region is the area with the least amount of biomass (24 biomass). 13. It should be noted that these estimates are different than reported in the Biomass Assessment for 1990 (Government of Malawi 1993). A preliminary comparison of these two studies would suggest that biomass in Malawi has decreased by about 16 percent between 1990 and 2004. Specifically, the northern region went from 218 million cubic meters to 159 million cubic meters (27 percent decline), the central region from 142 to 137 million cubic meters, a decrease of 14 percent and, the southern region from 102 to 92 million cubic meters, a 10 percent decrease. In fact it should be clarified that a precise comparison between the two studies is not feasible because of definitional differences between our study and the 1993 report: the two studies are based on different data with different spatial resolution and different methodologies.48 Nonetheless, it seems safe to conclude that much of the degradation has occurred in the biomass rich north, with less occurring in the central and southern regions which had less forests in them. 14. We combine these estimations with household level data from the IHS2. Specifically the analysis uses the data from the 492 rural enumeration areas in the IHS2, covering 9,840 households, and the collected community level information from each of the enumeration areas (see Annex 1A). The data on the economic activities at the community level contain information about three important sources of income. The various sources of income include agriculture and fishing as well as use of fuelwood and charcoal. The summary statistics for the community indicators are in Table 1. 15. The data on the fuelwood collection, its use and source are in the household part of the survey. The household level of welfare is measured by the real annual per capita consumption expenditure of the household. The socio-demographic characteristics of the households are captured by: an indicator variable that takes the value 1 if the head of the household is female, a set of indicator variables that takes the value 1 if the age of the head of household are in range 26-35, 36-45, 46-55, 56-65 and 66 and above, an indicator variable that takes the value 1 if the head of the household is a widow or a widower, 47Figure 1 shows a map of biomass at the spatial resolution of 463.3 meters. 48The main reason for the difference between the 1993 report and our results is because of different definitions of forested land. In the 1993 report a large amount of land is classified as agriculture mixed with trees/forest; under our assumptions (namely the 20% threshold) some of this land area is classified as forest resulting in an increase of the forested area with respect to the 1993 estimates. 142 household size, a set of indicator variables that takes the value 1 if the highest level of education in the household is some primary, completed primary, and post primary. 16. The agricultural characteristics of the household are represented by: an indicator variable that takes the value 1 if the households reported agriculture as the main economic activity, an indicator variable that takes the value 1 if the household grew tobacco in the last growing season, area under rain-fed and dry (dimba) agriculture cultivated by the household, area owned but not cultivated by the household, area with trees owned by the household. Availability of alternative to fuelwood collection is measured by an indicator variable that takes the value 1 if the household uses other cooking fuels or purchases fuelwood. 17. A third type of data collected by the survey relates to individual members of a household. Along with demographic characteristics of the members such as age, sex, marital status, the survey collected data on the amount of time spent by each household member above five years in various domestic and productive activities. These activities include fuelwood collection and household agriculture. 18. We define active men and women as those ten years and above who reported time spent on at least one of the activities for which time use data was collected. We also looked at children between age five and ten. However, preliminary analysis showed that very few children below age ten participated in fuelwood collection and agricultural production. Thus, by considering men and women above age ten we take into account most of the active labor force in the household. 19. We separate the households living in areas with less than 20 cubic meters per hectare of biomass (Less) and in areas with more than 20 cubic meters per hectare of biomass (More). The areas with less than 20 cubic meters per hectare of biomass are where biomass is scarce and areas with more biomass are where the scarcity is not as severe. Sixty five percent of the rural households surveyed reside in biomass scarce areas. 20. Along with the labor allocation data for the active men and women, we control for the following demographic characteristics: age, an indicator variable that takes the value 1 if the active person is married, number of living children less than five years old, an indicator variable that takes the value 1 if the active person suffered from any illness or injury in the past two weeks, an indicator variable that takes the value 1 if the active person suffered from any chronic illnesses, and the number of other members of the households suffering from chronic illnesses. 21. Table 3 presents evidence of the dependence of households on biomass for energy based on the household data. As indicated in Table 3, as many as 99 percent of poor households use fuelwood as cooking fuel. Though the non-poor consume significantly more fuelwood as compared with the poor, proportionately more poor households collect fuelwood. A large majority of the households (55 percent) collect fuelwood from unfarmed community areas. Less than 10% collect from their own lands. On average, a Malawian household spends the equivalent of MK 2,558 per year per, per capita on 143 fuelwood.49 That is, about 12 percent of total per capita annual consumption expenditure is spent on fuelwood. 22. The household heads were asked the time it takes for a one way walk to the location of fuelwood collection from their respective houses. We found no significant differences between the poor and non-poor households in terms of time taken for a one way walk to the location of fuelwood collection. 23. On an average an active women spent 1 hour and 30 minutes to collect fuelwood. The time spent in collection is slightly lower in the northern region but otherwise there is little difference in collection time between regions and rich and poor households. Eighty four percent of all the active individuals who collected fuelwood were women. 24. Figure 3 combines household data with biomass estimates by overlaying district level biomass availability on district level poverty estimates. There is considerable variation in biomass at the district level. Furthermore, at the district level there is no one- to-one connection between poverty and biomass. While the south faces considerable biomass scarcity and poverty, even in the south there are districts that are much better off than others. 49These numbers are self-reported. Since most Malawians collect and do not buy fuelwood, we expect that households used local prices to estimate the value of the fuelwood they use. 144 HOUSEHOLD WELFARE AND BIOMASS USE 25. Biomass scarcity may be less problematic in slightly richer countries or in countries where there are substitutes available, but in Malawi, the almost singular dependence on forests for energy implies that scarcity can have an impact on household welfare, particularly for poor households. Households faced with scarcity may reduce their consumption of energy, walk longer distances to collect fuelwood or use lower quality fuelwood (Brouwer et al. 1997). Lower biomass also implies that there are fewer wild foods and products to be had and there may be indirect effects from changes in eco- system functions. We hypothesize that all of these factors can affect household well- being either by affecting household consumption directly, affecting income (from any limited sales of fuelwood and labor re-allocation) or through its impact on household health or leisure. 26. To test the link between household welfare and biomass, we estimate a reduced form equation with log of real annual per capita consumption expenditure as the measure of household welfare on the left hand side. Consumption expenditure covers a variety of food and non-food expenditures, including estimated expenditures on some products obtained from forests such as charcoal, poles, grasses, wild foods and meat. It does not include estimated expenditures on collected fuelwood. Thus, any effect of biomass on consumption is an overall `net' effect that captures behavioral changes undertaken as a result of scarcity in fuelwood and other forest products. 27. We build our model of consumption expenditure based on an understanding of economic theory. Thus, our exogenous variables are those that are identified by economic theory and we test to see if there is a causal relationship between these variables and poverty. Our model relies strongly on a previous effort to estimate the determinants of poverty in Malawi using 1997-98 household data (National Economic Council, National Statistical Office, Government of Malawi and International Food Policy Research Institute 2001). This model is also documented in Mukherjee and Benson (2003). This model is also similar to the model used in Chapter 2 of this volume to assess determinants of poverty. Our main extension is in the use of biomass as a determinant of welfare. 28. In our reduced form model, the variable of interest on the right hand side is amount of biomass per hectare in the community. We postulate that household welfare and biomass stock have a nonlinear relationship. As biomass increases, welfare does not linearly increase but increases at a decreasing rate. We include biomass and biomass squared in the right hand side of the household welfare equation: lnCons = 1 + 2Bio + 3Bio2 + 4 AgroZones + 5Community + 6Household (1) 29. The community level explanatory variables are biomass, biomass squared, regular bus service in the community takes the value 1 if such service exists and 0 145 otherwise, health clinic in community takes the value 1 if there is a clinic in the community and 0 otherwise. If the EA is a Boma or a trading center, the indicator variable takes the value 1 and 0 otherwise. Four indicator variables capture the travel time to the nearest Boma. The respective variables take the value 1 if travel time is respectively 20 to 30 minutes, 30 to 45 minutes, 45 to 60 minutes and more than 60 minutes. The ADMARK market in the community, bank in the community, daily market in the community, and tarmac/asphalt road in the community are respective indicator variables that take the value 1 if the corresponding facilities exist in the community and 0 otherwise. 30. The household variables include an indicator variable that takes the value 1 for female-headed households and 0 otherwise and an indicator variable that takes the value 1 if the head of the household is a widow or a widower and 0 otherwise. Age of the head of the household is captured by five indicator variables that take the value 1 if the age of the head of household is in the respective ranges of 26 to 35, 36 to 45, 46 to 55, 56 to 65, and 66 and above. These five age indicators of the household head capture the nonlinear relationship between age and consumption expenditure. Household size and size squared scaled by hundred, capture the non-linear relationship between household size and consumption expenditure. The number of children in the age groups of 0-4, 5-10, and 11- 14 are used as corresponding three variables. The highest level of education in the household is measured by three indicator variables that take the value 1 respectively if the highest level of education is, some primary, completed primary or post primary and 0 otherwise. The chronic illness in the household is an indicator variable that takes the value 1 if such condition prevails and 0 otherwise. 31. Four religion indicator variables take the value 1 if the household head reported Islam, Catholic, CCAP, or Other Christian religion and 0 otherwise. 32. The economic characteristics of the household are captured by the following indicator variables. Household has wage or salary income takes the value 1 if this condition is satisfied and 0 otherwise. Household has non-farm enterprise takes the value 1 if this condition is satisfied and 0 otherwise. 33. Eight Agricultural Development Districts (ADD) are used to control for agro ecological zonal variations in various parts of the country. The ADD Ngabu left out as the default ADD in the analysis.50 34. The land holdings of a household were classified into four categories: rain-fed, dimba, uncultivated and trees. Since rain-fed and dimba are the land used for cultivation in the previous wet and dry season respectively, we expect areas under these to positively affect household welfare. We found no relationship between the amount of uncultivated and tree land and consumption expenditure and dropped these variables from the analysis. Thus, area of land used by the household for rain-fed cultivation in the last rainy season, area of land used by the household for dimba cultivation in the last dry 50The eight Agricultural Development Districts are: Karonga, Mzuzu, Kasungu, Salima, Lilongwe, Machinga, Blanyre, and Ngabu. 146 season, measures land assets of the household. DTobacco is an indicator variable that takes the value 1 if the household cultivated tobacco in the last cropping season. It measures the effect of the cash crop on welfare. We expect tobacco cultivating households to have higher welfare. 35. We estimate equation 1 for all households in the sample. We then estimate the same equation for households in the three rural regions: North, Central and South. We anticipate that the impact of biomass may be different for rich versus poor households. Thus, our final estimation of this equation uses the poverty line to divide the data into rich and poor categories. Results of the analysis of biomass scarcity on consumption per capita 36. Household welfare is expected to be influenced by biomass availability. We estimate the welfare-biomass relationship in equation 1 for the overall sample as well as for the sub samples of poor and non-poor and the North, Center and South regions. The summary relationship between consumption expenditure and biomass is presented in Table 4. The complete estimation results are in Appendix Table 2 and 3. 37. Table 4 indicates that the average effect of biomass on rural per capita consumption expenditure is significant but small. For example, a 10 percent increase in biomass per hectare is associated with approximately 0.05 percent higher annual per capita consumption expenditure. These numbers are somewhat higher if only poor households are considered ­ a 10 percent increase in biomass per hectare is associated with a 0.1 percent higher annual per capital consumption expenditure. In the rural South, a 10 percent increase in biomass per hectare is associated with a 0.2 percent increase in welfare. Because these effects are small we conclude that households may be adjusting to biomass scarcity without suffering much welfare loss. The small loss of welfare is due to fuelwood as well as other uses of and benefits from biomass. 38. To put the impact of biomass on consumption in perspective, we compare it with the impact of rain-fed agricultural land. For example, a 10 percent increase the size of the average rain-fed land holding by a poor household would result in a 0.4 percent increase in the per capita consumption expenditure. Thus, even though the effect of biomass scarcity on household welfare is small, it is not insignificant as compared with increases in agricultural land. 39. The quadratic relationship between biomass and consumption expenditure implies benefits from having access to more biomass increases at a decreasing rate. We find that the impact of biomass and its squared term are statistically significant for poor households, but they are not significant for non-poor households. This supports the hypothesis that the welfare of the poor is dependent on the stock of biomass as compared with that of the non-poor. 40. Figure 5 suggests that for an average rural household in Malawi consumption expenditure declines after biomass reaches 26 cubic meters per hectare in the 147 community.51 Seventy two percent of the rural households are in areas with biomass less than 26 cubic meters per hectare. However, as indicated in Figure 6, a rural poor household continues to benefit from biomass stock till 39 cubic meters per hectare. Eighty percent of the rural poor households are in areas with biomass less than 39 cubic meters per hectare. Thus, most of the rural poor would benefit if average biomass per hectare almost doubles from 20 to 39 cubic meters. 41. Figure 6 shows the total welfare loss for the rural poor in various districts resulting from a decline of biomass per hectare.52 The vertical axis shows the loss in total annual welfare incurred by all poor rural households in the district resulting from a 1 cubic meter decline in biomass per hectare. The horizontal axis shows the average biomass concentration in each district. This figure identifies the combination of districts from the South and Center regions where the poor are most likely to gain from investments in biomass. These are the districts with large concentration of rural poor and low concentration of biomass. All the districts in the figure 6 would suffer an annual welfare loss as a result of increase in scarcity of biomass. The five districts that would incur a welfare loss of 4 million Malawi Kwacha or more per year are Lilongwe, Zomba, Dedza, Mangochi, and Kasungu. The overall welfare loss by all the poor for rural Malawi is 44.8 million Malawi Kwacha per year. 42. Some caveats apply to the above findings. Our measure of biomass may not adequately capture mixed and low density biomass such as sparse bush or bush mixed with agricultural land. It is likely that households shift to lower quality biomass for their fuelwood needs when they face extreme scarcity. A more refined measure of biomass may help explain the biomass-welfare relationship better. We also note that our measure of household welfare does not account for other benefits such as biodiversity conservation that may not translate directly to higher annual household consumption. FUELWOOD SCARCITY AND HOUSEHOLD COPING STRATEGIES: LABOR ALLOCATION 43. The second question that interests us is whether biomass scarcity affects fuelwood collection and whether it reduces production-oriented activity. Thus, in our study we model two personal labor allocation decisions made by active men and women of age 10 and above: fuelwood collection and time spent on household agricultural production. 44. There is a small set of literature that looks at the effect of scarcity on time allocation. An early study in this area by Kumar and Hotchkiss (1988) examined the impact of deforestation on women's labor allocation. Their work on Nepal suggested that women increase their time spent on fuelwood collection when wood becomes scarce and this affects agricultural output. More recent papers by Cooke (1998a, 1998b) on 51 At the community level, the average biomass per hectare is 20 cubic meters per hectare. This number differs from the country average of 40 cubic meters per hectare because of differences in averaging weights. 52Three districts in the North region, Chitipa, Nikhata Bay, and Rumphi were excluded from Figure 6 because biomass concentrations in these districts are greater than the threshold level for the poor to benefit from an increase in biomass. 148 Nepal show that scarcity motivates increased time on collection but has no significant effect on labor in agriculture. Adding to this thin literature is a paper by Fisher et al. (2005) using data from southern Malawi, which supports some of Cooke's findings. 45. Cooke (1998a,b) and Fisher et al.(2005), while their empirical models differ, derive their labor allocation models from a neo-classical household production function framework. Similar models are used in Illahi and Grimard (2000) and Khandker (1988) who study labor allocation in different contexts in Pakistan and Bangladesh. We do not develop a formal model of household behavior for this paper; however, our partial equilibrium personal labor allocation model is rooted in the same household production and utility function framework. 46. The idea here is that of a rural representative household that functions in a subsistence setting and produces the goods that it consumes. Households choose the optimal levels of consumption goods and leisure. Some home produced goods (for example maize) may be generated with the use of market inputs, labor and land. Others, such as fuelwood are generated by the use of biomass and labor. Household behavior and utility is conditioned on several fixed household characteristics such as household size, health stock and education and so on. Households maximize utility by allocating optimal amounts of labor to different home production tasks and by purchasing market inputs subject to a full income constraint. At the margin, households equate the marginal value product of leisure to the net returns to labor in different tasks. This yields a set of labor supply equations that can be empirically estimated. Men and women's labor supply is a function of output prices, wages, and household and community characteristics that may affect productivity. Scarcity of biomass is expected to impact labor supply decisions. 47. Following Illahi and Grimard (2000) we model labor allocation as an individual decision influenced by household characteristics. This level of disaggregation allows us to control for individual characteristics and their relationship with the rest of the household in the labor allocation decision. For example, we can control for age, sex, health, and marital status of the individual.53 48. The model described below is estimated separately for men and women above 10. However, in Malawi most of the fuelwood collection is done by women. Thus, we describe of the model in terms of a representative woman. 49. In estimating labor supply equations, an empirical problem that needs to be accounted for is the presence of active men and women who do not participate in agriculture or fuelwood collection tasks. The determinants of the decision to collect 53Alternate models (Cooke 1998a,b, for example) estimate labor demand and supply based on average household behavior versus individual allocative decisions. In these models, the wage rate is for an `average' male or female member of the household. However, wages differ based on individual age, sex, education, and experience rather than because of household characteristics. This is the reason why we choose to estimate individual male and female labor supply functions versus a household labor supply function. 149 fuelwood may be different from those of the decision to allocate hours to fuelwood collection. Whether a woman collects fuelwood may be a function of availability of substitutes and whether or not the household purchases fuelwood, which in turn may be influenced by household wealth. To address this, we model the personal labor allocation decisions in two stages. In the first stage an active woman decides whether or not to participate in the specific activity such as fuelwood collection or agricultural production. Those who decide to participate then decide how much time to spend on the respective activities. 50. The participation decision for activity j = fuelwood collection, agricultural production may be represented by: Pr(Participat ion = 1) = (1 + 1Bio + 3Region j (2) + 4Community + 5 Household ) 51. The time spent on activity j may be represented by: Hours (3) j = d 1 + d 2 Bio + d 3 Region + d 4 Community + d 5 Household 52. The variable HAssets measures real value of household assets and is a proxy for wealth of the household.54 DWet is an indicator variable that takes the value 1 if the household was interviewed during the wet season. Since the active woman is likely to work in the rain-fed plots in the wet season and in the dimba plots in the dry season, the season dummies are interacted with the respective size of rain-fed and dimba plots. Bigger the size of uncultivated land in the household, it is likely the active woman will have more time and thus is more likely to participate in fuelwood collection. Similarly we expect larger size of the area with trees will make the active women in the household more likely to collect fuelwood. 53. It is important to provide some additional discussion about the biomass variable. Many of the papers (Kumar and Hotchkiss, 1988, Cooke 1998a,b) that have thus far studied the effects of natural resource scarcity on labor allocation do not have an exogenous measure of scarcity. Cooke estimates the shadow wage rate and uses this in her model as a measure of scarcity. Our paper because of its unique data set is able to directly assess the impact of biomass scarcity on labor allocation. Our hypothesis is that households will increase the time spent on fuelwood collection when faced with increased scarcity. They may consequently reduce time on agricultural productive activities. However, in modeling the effect of biomass on labor decisions we are influenced by information on some specific ways in which Malawian households react to 54HAssets is the total current value of household assets measured in constant MK. The assets considered are mortar/pestle, bed, table, chair, fan, air conditioner, radio, tape or CD player, television and VCR, sewing machine, kerosene stove, electric or gas stove, refrigerator, washing machine, bicycle, motorcycle, car, mini-bus, and lorry. These assets represent the wealth stock of the household. Most of these assets are for household use rather than used for productive activities (exception: mini-bus, lorry). We do not use principal component based index of assets for these as the monitory value is available. 150 scarcity. Brouwer et al. (1997) suggest that households when faced with great scarcity increase their time in collection up to a point, but after that they switch to alternate and lower quality sources of fuel wood. Thus, the labor supply function switches at a certain level of biomass scarcity. We model this by estimating two separate labor supply functions ­ one for biomass quantities lower than 20 cubic meters per hectare and one for higher quantities. We hypothesize that the active women who live in biomass scarce areas cope with the biomass scarcity in many different ways as compared with women who live where biomass is not scarce. 54. An important determinant of collection behavior is the opportunity cost of time. We expect that a higher opportunity cost would make woman less likely to collect fuelwood. We assume that the wage rate is a reasonable measure of the opportunity cost of time. However, not all women enter the formal labor market and do not report the wage rate. Therefore, in a fashion similar to Ilahi and Grimard (2000), we follow Heckman's model to instrument the real average wage for women based on the wages and salaries reported by women and their socio-demographic characteristics.55 Our measure of opportunity cost of time is individually predicted wages, but does not account for the possible presence of different shadow wages for different activities (Cooke 1998a, Fisher et al. 2005). 55. We use Heckman's sample selection model (Heckman, 1978) to model the participation and labor allocation decisions of the active women in fuelwood collection. The choice of Heckman's model is influenced by the fact that some of the factors that affect both the participation decision as well as the hours spent on collecting fuelwood may have opposite effects on these two decisions. For example, easy access to fuelwood as measured by the size of household tree plot is expected to have positive effect on participation decision and negative effect on time allocation. The probability of participation is expected to be higher for women residing in households with larger woodlots. However, we expect women in these households to take less time in collecting fuelwood as compared with women who do not have access to personal wood lots. As a result, a Tobit model is not appropriate for fuelwood collection. 56. The hours spent on household agriculture is modeled as a Tobit equation. The use of Tobit model for the estimation of hours spent on household agriculture is based on the reasoning that the same variables that affect decision to take part in agricultural activities also affect the hours spent on agriculture. Unlike fuelwood collection, where some variables may have opposite effects on the participation and hours decision, for agricultural activities all the variables have similar effects. That is, the variables that are likely to increase the probability of participation in agriculture are also the variables that are likely to increase the hours spent on agriculture and vice versa. Thus, Tobit 55The active woman's participation in the labor force is modeled as a function of whether or not she is married, number of children she has, the wage earned by the spouse, number of other active women in the household, the household size, and wet season dummies. The real wage of the active woman is modeled as a function of her age and education and respective squared terms, and wet season dummy. Both the models included regional intercept dummies to control for between region variation in the labor market. 151 estimation method is appropriate for the time allocation to agricultural activities and not for fuelwood collection. 57. All but one right hand side variable are the same as that of hours spent on fuelwood collection. PAssets is an index of productive assets owned by the household.56 The sign of the coefficient for the productive asset index is ambiguous. More productive assets may imply higher level of capital intensity and labor saving technology and thus fewer hours spent on agriculture. On the other hand, more assets may also point to more intense cultivation and thus more hours spent on agriculture. We estimate the agricultural activity labor allocation models separately for less and more biomass areas. Results of the analysis of coping strategies and labor allocation 58. The estimates of fuelwood collection equations 2 and 3 for active women in overall rural Malawi, less biomass, and more biomass areas are shown in Table 5. Columns 1 and 2 show estimates for the overall rural Malawi. Columns 3 and 4 show estimates for the biomass scarce areas where biomass stock is less than 20 cubic meters per hectare and columns 5 and 6 present results for areas with more biomass. 59. For overall rural Malawi, higher levels of biomass appear to make an active woman more likely to participate in fuelwood collection but does not affect hours spent on fuelwood collection. Some of the factors such as Dry Season Dimba Area Interaction that are significant for the overall rural Malawi estimation of the active women's fuelwood collection estimation are only significant for more biomass areas (Columns 1 and 5). Other factors such as Number of Children below five appear to affect overall rural Malawi, and less biomass areas estimations (Columns 1 and 3). Thus, we concentrate on the separate estimates for less biomass and more biomass areas rather than overall rural Malawi. 60. For both the biomass scarce and sufficient areas, higher stock of biomass increases the likelihood of collection and decreases the hours spent on fuelwood collection. In both cases, factors such as age of the active woman, uncultivated land held by the household and the marriage dummy appear to make an active woman more likely to participate in fuelwood collection. Age squared, wage rate, and substitute for collected fuelwood appear to make an active woman less likely to participate in fuelwood collection in both cases. Our interest is mainly in the number of hours spent in collection of fuelwood. We find that biomass scarcity has a significant but small effect on hours spent in collecting fuelwood. A one cubic meter decrease in biomass per hectare results, on average, in a one minute increase in the time spent by women on collection. 61. An important question is why there is such a small change in time spent in the face of increasing scarcity. One explanation is that women undertake multiple tasks 56PAssets is an index of productive assets computed from the principal component analysis. The assets considered are: beer brewing drum, boat or canoe, fishing net, ox cart, wheel barrow, hand sprayer, panga, hoe, axe, and sickle. These assets are mainly used for agricultural and other productive activities and is thus assumed to affect agricultural labor hours but not hours spent on fuelwood collection. 152 when they collect fuelwood, which may also be why hours spent in fuelwood do not differ much between the biomass rich north and the biomass poor south. Households also use multiple strategies to cope with biomass scarcity and do not just change their labor allocation behaviors. These conclusions are supported by micro-studies such as that by Brouwer et al. (1997), who indicate that women in Ntechu District in central Malawi combine fuelwood collection with field work or change the frequency of collection in response to scarcity. Another explanation for the low variance in labor time in our study may lie in the nature of the data collected, since the labor allocation responses were rounded to half hour segments. 62. Table 6 shows the results from our estimation of female agriculture labor supply in overall rural Malawi, less and more biomass areas. Though for overall rural Malawi higher biomass appear to decrease hours spent by active women in agricultural productions, variation in biomass stock within less and more biomass areas does not appear to affect hours spent by an active woman in household agriculture. Other significant factors have the expected signs. For example, larger household size implies larger supply of labor in the household and thus lesser number of hours for the active woman in the agricultural fields. Larger size of cultivated land in dry and in wet season implies more hours whereas larger size of uncultivated fallow implies shorter hours. Higher wages means higher opportunity costs for the active woman and thus shorter hours in the fields. 63. We find strong evidence of negative impact of chronic illnesses in the family on the number of hours spent in agriculture by an active woman. However, we do not find such a strong effect on either the participation or the number of hours an active woman spends on fuelwood collection. Similarly recent personal illness appears to have a negative effect on the number of hours in agriculture but has no effect on the hours spent on fuelwood collection. This is likely to be because agricultural labor is more strenuous than labor required for fuelwood collection. 64. Relatively small numbers of men collect fuelwood. Men do not report collecting any fuelwood in 94 percent of the households surveyed. We find no evidence of biomass stock affecting fuelwood collection decision and hours spent collecting fuelwood by active men. We report the estimation results for men's labor allocation in fuelwood collection and agricultural activities in Appendix Table 2 and 3. Interestingly while married women are more likely to collect fuelwood, married men are less likely. Larger household size also makes men less likely to collect fuelwood. This effect is not always significant for women. 65. Biomass shows perverse sign for hours spent on agricultural production by men. That is, higher biomass stock results in less number of hours by men in agricultural production in the more biomass areas and in overall estimates. We also find similar results for women in the overall estimates but they are not significant for more and less biomass areas. We speculate that men (and women) may spend time collecting other forest products instead of agricultural production where biomass is not scarce. 153 SUMMARY OF MAJOR FINDINGS AND CONCLUSIONS 66. Malawi is a country in biomass distress. It is also one of the poorest countries in Africa. This chapter has attempted to understand how biomass access varies across the country, how this varied access affects the poor and how the poor cope with scarcity. 67. Remote sensing data shows that a large proportion of the biomass available in Malawi is in the northern region, which is the least populated region in the country. The South with much less biomass and much higher population density, faces considerable scarcity. Overlaying biomass and poverty maps on each other shows that at the district level there is no one-to-one relationship between poverty and forest degradation. 68. As many as 97 percent of rural Malawians depend on fuelwood as their primary source of energy for cooking. Almost 80 of households collect fuelwood directly from areas close by. Also, 90 percent of the poor collect their own fuelwood. On an average day when women go out to collect fuelwood, they spend one and half hours on this task. Thus, biomass scarcity matters for the welfare of rural Malawians. 69. Our analyses suggest that biomass scarcity is associated with small but significantly lower household welfare, particularly for the rural poor. At current high levels of scarcity, 80 percent of the rural poor households are likely to benefit moderately () from an increase in biomass in the community. The districts that are likely to gain the most from investments in biomass are Lilongwe, Zomba, Dedza, Kasungu, and Mangochi. 70. There have been attempts in the past to invest in fuelwood in Malawi and to induce rural households to plant more trees. Unfortunately, these efforts have largely been un-successful (French 1986, Dewees 1995, Walker 2004). Our results may provide an economic explanation for the households' decision not to plant more trees despite various policy interventions. The net returns to households from participating in biomass related projects may not be adequate. This does not mean that Malawians do not perceive biomass to be scarce. Rather, it indicates that given various constraints, households may choose to not use scarce land and labor to plant fuelwood oriented tree crops. This suggests that understanding household responses to scarcity has to be integral to any strategy to resolve Malawi's deforestation problems. 71. In this chapter, we examine whether households increase their time in fuelwood collection in response to fuelwood scarcity and whether this has an impact on agricultural labor. Our analysis confirms that rural women cope with biomass scarcity by increasing the time they spend on fuelwood collection. However, this increase is small suggesting that households use multiple strategies to cope with scarcity. On a positive note, biomass scarcity does not have a negative effect on the amount of agricultural activity undertaken by women or men. Of more concern in this regard is the effect of chronic illness within the family. 154 72. While poverty and environment are inter-linked in Malawi, physical scarcity of biomass may not always translate to economic scarcity of fuelwood. When faced with biomass scarcity, the rural poor may cope by collecting lower quality biomass and undertaking simultaneous tasks with fuelwood collection. A better understanding of these strategies may help policy-makers target poverty eradication and biomass conservation more effectively in rural Malawi. 155 REFERENCES Adhikari, B. 2003. "Property Rights and Natural Resources: Socio-Economic Heterogeneity and Distributional Implications of Common Property Resource Management." Working Paper 1-03. South Asian Network for Development and Environment Economics (SANDEE): Katmandu, Nepal. Baccini, A, M.A. Friedl, C.E. Woodcock and R. Warbington (2004). Forest biomass estimation over regional scales using multisource data. Geophysical Research Letters, 31(L10501). Bluffstone, R.A. 1995. The effect of labor market performance on deforestation in developing countries under open access: An example from rural Nepal. Journal of Environment and Development Economics, 4; (2): 161-76. Brouwer, I.D., J.C.Hoorweg and M.J. Van Liere. 1997. When households run out of fuel: Responses of rural households to decreasing fuelwood availability, Ntcheu District, Malawi. World Development, Vol 25, No. 2: 255-266. Breiman, L., Friedman, J. H., Olshen, R. A., and Stone, C. J. 1984. Classification and Regression Trees. Wadsworth, Belmont, CA. Cavendish, W. 2000. Empirical regularities in the poverty-environment relationship of rural households: evidence from Zimbabwe. World Development (U.K.); 28, No. 11:1979-2003, November 2000. Cooke, P. A. 1998a. Intra-household labor allocation responses to environmental good scarcity: a case study from the hills of Nepal. Economic Development and Cultural Change vol 46, no. 4: 807-830. Cooke, P. A. 1998b. The effect of environmental good scarcity on own-farm labor allocation: the case of agricultural households in rural Nepal. Environment and Development Economics, No. 3: 443-469. Cohen, W. B. and Spies, T. A. 1992. Estimating structural attributes of Douglas- Fir/Western Hemlock forest stands from Landsat and Spot imagery. Remote Sensing of Environment, Volume 41, 1-17. Dasgupta, P. 1995. Population, poverty and the local environment. Scientific American, vol 272, no. 2: 26-31. Dasgupta, S. U. Deichmann, C. Meisner, and D. Wheeler. 2005. Where is the poverty- environment nexus? Evidence from Cambodia, Lao, PDR and Vietnam. World Development, vol 33, No. 4: 617-638. Dewees, P. 1989. The woodfuel crisis re-considered: Observations on the dynamics of abundance and scarcity. World Development, vol 17, no. 8: 1159-1172. 156 Dewees, P. 1995. Trees on farms in Malawi: Private investment, public policy and farmer choice. World Development, vol 23, no, 7: 1085-1102. Ekbom, A. & J. Bojö 1999. Poverty and Environment: Evidence of Links and Integration in the Country Assistance Strategy Process. World Bank Africa Region Discussion Paper No. 4. Washington, DC: World Bank. Duraiappah, A. 1998. Poverty and environmental degradation: A review and analyses of the nexus. World Development, vol 26, No. 12: 2169-2179. Ezzati, M and D. M. Kammen. 2001. Indoor air pollution from biomass combustion and acute respiratory infections in Kenya: an exposure-response study. The Lancet, vol 358, 619-24. Fisher, M. 2004. Household welfare and forest dependence in southern Malawi. Environment and Development Economics 9: 135-154. French, D. 1986. Confronting an Unsolvable Problem: Deforestation in Malawi. World Development, vol 14, no. 4: 531-540. Fisher, M and G. Shively. 2005. Can income programs reduce tropical forest pressure? Income shocks and forest use in Malawi. World Development, vol 33, No. 7: 1115-1128/ Fisher, M, G. Shively, and S. Buccola. 2005. Activity Choice, Labor Allocation and Forest Use in Malawi. Land Economics, vol 81, no. 4: 503-517. Food and Agriculture Organization (FAO 2005). Forestry 2005 Country Tables, Malawi. www.fao.org/forestry/site/32085/en Franklin, J. 1986. Thematic mapper analysis of coniferous forest structure and composition. International Journal of Remote Sensing, Volume 7, 1287-1301 Friedl, M. A. and Brodley, C. E. 1997. Decision Tree classification of land cover from remotely sensed data. Remote Sensing of Environment, Volume 61, 399-409 Government of Malawi, Ministry of Forestry and Natural Resources, Department of Forestry, 1993, Forest Resources Mapping and Biomass Assessment for Malawi. Lilongwe, Malawi. Heckman, J.J. 1978. Sample selection bias as a specification error. Econometrica, Volume 47, 153-161. Ilahi, N and F. Grimard (2000). Public infrastructure and private costs: Water supply and time allocation of Women in rural Pakistan. Economic Development and Cultural Change, 49 (1): 45-75. 157 IFPRI, 2000. Poverty Monitoring Service Policy Brief No. 2. Malawi State of the Environment Report 1998 Jodha, N. 1986. Common property resources and rural poor in dry regions of India. Economic Political Weekly, vol 21, No. 27. Khandker, S. R. 1988. Determinants of women's time allocation in rural Bangladesh. Economic Development and Cultural Change, 37 (1): 111-126. Malawi Sustainable Development Network Programme (SDNP), 1998. State of the Environment Report. http://www.sdnp.org.mw Michaelsen, J., David, S. S., Friedl, M. A, Davis, F. W. and Dubayah, R. C. 1994. Regression Tree analysis of satellite and terrain data to guide vegetation sampling and survey. Journal of Vegetation Science, Volume 5, 673-686. Minde, I.J., G. Kowero, D. Ngugi and J. Luhanga. 2001.Agricultural Land Expansion and Deforestation in Malawi. Forest, Tress and Livelihoods, vol 11: 167-182. Mukherjee S. and T. Benson. 2003. The determinants of Poverty in Malawi, 1998. world Development vol 31, no. 2: 339-358 R. B. Myneni, J. Dong, C. J. Tucker, R. K. Kaufmann, P. E. Kauppi, J. Liski, L. Zhou, V. Alexeyev, and M. K. Hughes. 2001. A large carbon sink in the woody biomass of Northern forest. Proceedings of the National Academy of Sciences, Volume 98, 14784- 14789. Nankuni. F. 2004. Environmental degradation, resource scarcity and children's welfare in Malawi : School attendance, school progress and children's health. A Ph.D. thesis in agricultural economics and demography. Pennsylvania State University, the Graduate School, College of Agricultural Sciences. National Economic Council, National Statistical Office, and IFPRI, 2001, The determinants of Poverty in Malawi: An analysis of the Malawi Integrated Household Survey 1997-97 Nerlove, M. 1991. Population and the environment: A parable of firewood and other tales. American Journal of Agricultural Economcs. Vol 73: 1334-47. Puhr, C. B. and Donoghue, D. N. M. 2000. Remote sensing of upland conifer plantations using Landsat TM data: A case study from Galloway, South-West Scotland. International Journal of Remote Sensing, Volume 21, 633-646. Shugart, H. H., Chavez, L. B. and Kasischke, E. S., 2000. Determination of stand properties in boreal and temperate forests using high-resolution imagery. Forest Science, Volume 46, 478-486. 158 Walker, P. A. 2004. Roots of crisis: Historical narratives of tree planting in Malawi. Historical Geography, vol 32: 89-109. World Resources Institute (WRI) 2003. Earth Trends Table, Forests, Grasslands and Drylands. http://earthtrends.wri.org/pdf_library/data_tables/for1_2003.pdf 159 Figure 28: Biomass distribution at the year 2004 derived from remotely sensed data recorded by MODIS sensor 160 Figure 29: Average district biomass at the year 2004 intersected with poverty rates at the district level 161 Figure 30: Overall quadratic relationship between biomass and household consumption Change in Welfare for All Rural 1.2 26 1 0.8 0.6 0.4 0.2 0 0 20 40 60 80 100 120 140 160 180 Biomass m3 / Ha 162 Figure 31: Quadratic relationship between biomass and household consumption for the poor Change in Welfare for Poor Rural 1.2 39 1 0.8 0.6 0.4 0.2 0 0 20 40 60 80 100 120 140 160 180 Biomass m3 / Ha 163 Figure 32: Annual welfare loss for the rural poor from 1 percent decline in biomass in the districts 6,000,000 Lilongwe Zomba 5,000,000 Dedza Mangochi Kasungu 4,000,000 K Machinga M s Chiradzulu Los 3,000,000 Ntcheu lfare Mchinji We Phalombe Dowa Thyolo 2,000,000 Mzimba Balaka Nsanje Salima Chikwawa 1,000,000 Nkhotakota Blantyre Mwanza Karonga Mulanje Ntchisi 0 0 5 10 15 20 25 30 35 40 Biomass m3 / Ha Note: The names of the districts from South are in red, those from Center are in blue and those from North are in green. Three districts in the North region, Chitipa, Nikhata Bay, and Rumphi were excluded. Biomass concentrations in these districts are greater than the threshold level for the poor to benefit from an increase in biomass. 164 Table 1: Community Level Indicators in the regions Rural Rural Rural North Centre South Total Population Density 0.18 0.14 0.25 0.20 Number of Agricultural Households in Community 302.3 295.3 267.2 283.3 Distance to Asphalt Road KM 38.9 16.8 16.5 19.8 Daily Market Distance KM 12.5 9.5 5.3 8.0 Weekly Market Distance KM 7.5 5.3 5.2 5.5 ADMARC Market Distance KM 12.9 12.9 9.0 11.1 Distance to Nearest Market KM 3.5 3.2 2.4 2.9 Post Office Distance KM 12.5 10.1 9.3 10.1 Primary School Distance KM 26.9 2.3 3.1 5.5 District Boma Distance KM 55.9 39.2 38.5 41.4 Urban Center Distance KM 170.5 106.9 94.2 110.4 Bank Distance KM 62.3 29.2 23.0 31.3 Average Distances to Various Facilities KM 46.3 27.2 23.4 28.3 165 Table 2: Biomass Distribution Biomass Biomass Biomass Volume 2004 Volume Volume 1990 Area distr. Fuel coll. Poverty (1000 cubic (cubic (1000 cubic DistrictID (1000 ha) District name Hours/day rates meters) meters/ha) meters)@ North 101 425 Chitipa 1.22 0.67 25146 59.1 33126 102 342 Karonga 1.12 0.55 17250 50.5 34788 103 418 Nkhata Bay 1.25 0.63 33664 80.6 32658 104 466 Rumphi 1.76 0.62 35747 76.7 35069 Mzimba/Mzuzu 105 1064 City 1.12 0.47 47196 44.3 81855 Region subtotal 2715 159004 58.6 217496 Central 201 807 Kasungu 1.79 0.45 33264 41.2 34541 202 433 Nkhotakota 2.26 0.48 23190 53.5 33260 203 172 Ntchisi 0.78 0.47 6014 34.9 5999 204 307 Dowa 1.66 0.37 6438 21.0 7867 205 215 Salima 2.00 0.57 5289 24.6 7909 Lilongwe/Lilongwe 206 623 City 1.15 0.33 7186 11.5 20545 207 315 Mchinji 1.41 0.60 12525 39.8 7495 208 376 Dedza 1.25 0.55 32674 86.8 13417 209 326 Ntcheu 1.78 0.52 10122 31.1 10716 Region subtotal 3575 136702 38.2 141749 South 301 674 Mangochi 1.94 0.61 7277 10.8 32831 302 394 Machinga 1.56 0.74 13465 34.2 13046 303 312 Zomba/Zomba City 1.41 0.64 2936 9.4 3144 304 76 Chiradzulu 1.13 0.64 9456 124.1 812 Blantyre/Blantyre 305 203 City 0.88 0.32 9408 46.3 6822 306 233 Mwanza 1.44 0.56 4160 17.9 12595 307 167 Thyolo 1.91 0.65 2689 16.1 3214 308 201 Mulanje 1.55 0.69 523 2.6 8092 309 143 Phalombe 1.35 0.62 8479 59.5 na 310 490 Chikwawa 1.71 0.66 18663 38.1 14650 311 195 Nsanje 1.41 0.76 6530 33.5 6569 312 214 Balaka 1.09 0.67 8405 39.3 na Region subtotal 3300 91992 27.9 101775 Total 9590 387698 40.4 461020 @From Government of Malawi (1993) 166 Table 3 Fuelwood use, source, expenditure and collection time by households Less More Non- biomass Biomass variable Poor Poor North Center South Areas Areas Overall Percent of households using fuelwood as cooking fuel: Fuelwood is primary cooking fuel 95.9% 98.6% 98.2% 97.5% 96.6% 96.9% 97.6% 97.2% Fuelwood is secondary cooking fuel 2.6% 1.1% 0.6% 1.7% 2.3% 2.1% 1.3% 1.9% Fuelwood is primarily collected 78.5% 90.5% 88.5% 82.8% 83.9% 82.2% 88.0% 84.2% Fuelwood is sometimes collected 7.8% 5.4% 2.6% 7.2% 7.4% 7.9% 4.2% 6.6% Percent of households with different sources of fuelwood: Own woodlot 9.8% 8.7% 4.0% 9.7% 10.6% 11.6% 4.9% 9.3% Community woodlot 4.2% 3.9% 1.5% 7.6% 2.0% 4.4% 3.5% 4.1% Forest reserve 12.5% 17.9% 13.4% 10.1% 19.8% 14.3% 16.5% 15.1% Unfarmed areas of community 50.8% 58.9% 71.4% 47.7% 55.2% 51.8% 59.9% 54.6% Other 9.3% 6.7% 2.4% 15.1% 4.0% 8.4% 7.5% 8.1% Distance, hours, and expenditure One way walk to the source (Hours) 0.6 0.6 0.5 0.7 0.6 0.6 0.6 0.6 Hours spent (women) yesterday collecting firewood 1.5 1.5 1.2 1.5 1.5 1.5 1.5 1.5 Hours spent on agricultural activities per day 2.9 2.8 2.7 2.8 3.0 2.9 2.8 2.9 Expenditure on Fuelwood (MK 2004)@ 3484 1536 1557 3205 2327 2731 2233 2558 Total Consumption Expenditure (MK 2004)@ 31139 10946 21742 23809 19683 21850 21115 21595 Percent of Expenditure on Fuelwood 11% 14% 7% 13% 12% 12% 11% 12% Of those who collected fuelwood: Percent women age 10 and above 81.6% 86.8% 94.9% 81.6% 83.4% 83.5% 86.3% 84.4% @ Per capita per year. 167 Table 4: Biomass and Per Capita Consumption Expenditure After Controlling for other Variables Coeff Coeff Biomass Mean Elasticity Biomass Squared Biomass @ (cum per ha) Overall 0.0010209 ** -0.0000194 *** 20.22 0.005% Non-Poor 0.0000051 -0.0000040 19.35 Poor 0.0015476 *** -0.0000200 *** 21.20 0.01% Rural North 0.0016728 -0.0000176 * 42.97 Rural Center -0.0001699 -0.0000079 13.70 Rural South 0.0025929 *** -0.0000347 *** 18.61 0.02% @These numbers differ from Table 3 because of averaging differences. 168 Table 5: Labor allocation for fuelwood collection by women 10 years and older (1) Collect (2) Hours (3) Collect (4) Hours (5) Collect (6) Hours Overall Less Biomass More Biomass Age in years 0.011*** 0.009 0.008* 0.003 0.015** 0.017 (0.004) (0.007) (0.005) (0.009) (0.006) (0.013) Age Square -0.0001*** -0.0001 -0.0001** -0.0001 -0.0002** -0.0002 (0.0000) (0.0001) (0.0001) (0.0001) (0.0001) (0.0002) Head Education: Snr. Prim or more -0.053** 0.088* -0.073** 0.110* -0.012 0.009 (0.026) (0.049) (0.033) (0.062) (0.045) (0.085) HH Size -0.014*** 0.015 -0.021*** -0.003 0.004 0.038** (0.005) (0.010) (0.006) (0.012) (0.009) (0.016) Real Value of Assets '000 -0.002*** 0.000 -0.003*** 0.002 -0.001* -0.001 (0.001) (0.001) (0.001) (0.002) (0.001) (0.002) Wet Season & Wet Area Interaction 0.009 0.042 0.003 0.056 0.026 -0.014 (0.021) (0.044) (0.024) (0.051) (0.041) (0.088) Dry Season and Dimba Area Interaction 0.153** -0.150 0.068 -0.105 0.312*** -0.251 (0.065) (0.108) (0.081) (0.148) (0.121) (0.174) Owned & uncultivated last season 0.117*** 0.021 0.129*** 0.084 0.090*** -0.052 (0.021) (0.040) (0.031) (0.055) (0.030) (0.061) Total hectares of trees 0.259 -0.378 1.392** -0.776 -0.124 0.229 (0.259) (0.449) (0.549) (0.663) (0.344) (0.666) Predicted Real Avg Wage for Women -0.001*** 0.001 -0.001*** 0.000 -0.001** 0.002* (0.000) (0.001) (0.000) (0.001) (0.000) (0.001) Number of <5yr children 0.076*** -0.075** 0.095*** -0.053 0.038 -0.085 (0.020) (0.036) (0.024) (0.046) (0.034) (0.061) Married for age 12+ 0.175*** -0.074 0.201*** -0.049 0.147*** -0.176* (0.031) (0.063) (0.038) (0.080) (0.053) (0.106) Rural Centre -0.033 0.359*** 0.020 0.278** 0.019 0.388*** (0.045) (0.085) (0.063) (0.119) (0.072) (0.139) Rural South -0.073 0.410*** 0.088 0.322*** -0.206*** 0.489*** (0.045) (0.083) (0.065) (0.125) (0.066) (0.137) Wet Season Dummy -0.105*** -0.001 -0.107*** -0.046 -0.120** 0.179 (0.032) (0.064) (0.040) (0.079) (0.058) (0.122) ADMARC Market Distance KM 0.0002 0.0011** 0.0003 0.0010** -0.0020 0.0091*** (0.0003) (0.0005) (0.0003) (0.0005) (0.0015) (0.0029) Distance to Asphalt Road KM -0.0011*** 0.0004 0.0005 -0.0017 -0.0018*** 0.0014 (0.0004) (0.0009) (0.0005) (0.0013) (0.0007) (0.0015) Elevation - -0.00010** -0.00004 0.00000 -0.00007 0.00029*** 0.00011 (0.00004) (0.00008) (0.00005) (0.00011) (0.00008) (0.00016) Biomass (m3/ha) 0.001** -0.001 0.009*** -0.018*** 0.005*** -0.006*** (0.001) (0.001) (0.003) (0.005) (0.001) (0.002) Price of Maize Grain MK/Kg 0.002 -0.018 -0.011 -0.015 0.020* -0.026 (0.006) (0.013) (0.008) (0.016) (0.011) (0.023) Recent illness/injury 0.023 -0.084* 0.035 -0.061 0.013 -0.135 (0.027) (0.050) (0.034) (0.062) (0.047) (0.089) Do you suffer from a chronic illness -0.035 -0.149** -0.026 -0.188** -0.047 -0.122 (0.040) (0.075) (0.048) (0.088) (0.072) (0.144) Other Members Chronically Ill in HH -0.031* -0.086*** -0.027 -0.085** -0.031 -0.081 (0.018) (0.034) (0.022) (0.040) (0.031) (0.062) 169 Substitute for collected fuelwood -0.536*** -0.472*** -0.732*** (0.041) (0.047) (0.088) Constant -0.596*** 1.674*** -0.562*** 1.848*** -1.031*** 2.145*** (0.145) (0.383) (0.185) (0.485) (0.247) (0.736) Observations 13281 13281 8690 8690 4591 4591 -0.244 -0.135 -0.688* Inverse Mills Ratio (0.214) (0.270) (0.359) Standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 170 Table 6: Labor allocation hours for household agriculture by women 10 years and older (1) (2) (3) Overall Less Biomass Areas More Biomass Areas Age in years 0.133*** 0.138*** 0.126*** (0.007) (0.008) (0.011) Age Square -0.001*** -0.001*** -0.001*** (0.000) (0.000) (0.000) Head Education: Snr. Prim or more -0.142*** -0.029 -0.360*** (0.048) (0.059) (0.082) HH Size -0.031*** -0.038*** -0.042** (0.010) (0.012) (0.017) Real Value of Assets '000 -0.005*** -0.005*** -0.004*** (0.001) (0.001) (0.001) Wet Season & Wet Area Interaction 0.181*** 0.176*** 0.229*** (0.036) (0.041) (0.074) Dry Season and Dimba Area Interaction 0.623*** 0.582*** 0.596*** (0.121) (0.144) (0.220) Owned & uncultivated last season -0.150*** -0.090 -0.167*** (0.043) (0.061) (0.061) Total hectares of trees 0.224 1.041 0.184 (0.479) (0.934) (0.559) Predicted Real Avg Wage for Women -0.003*** -0.003*** -0.003*** (0.000) (0.001) (0.001) Number of <5yr children 0.256*** 0.245*** 0.289*** (0.036) (0.043) (0.061) Married for age 12+ 0.563*** 0.585*** 0.462*** (0.055) (0.067) (0.094) Rural Centre -0.036 -0.385*** 0.514*** (0.081) (0.109) (0.129) Rural South -0.032 -0.455*** 0.340*** (0.082) (0.113) (0.123) Wet Season Dummy 0.740*** 0.642*** 0.891*** (0.058) (0.071) (0.104) ADMARC Market Distance KM 0.003*** 0.002*** 0.016*** (0.000) (0.000) (0.003) Distance to Asphalt Road KM 0.006*** 0.001 0.010*** (0.001) (0.001) (0.001) Elevation -0.000*** -0.001*** 0.000 (0.000) (0.000) (0.000) Biomass (m3/ha) -0.005*** 0.003 0.000 (0.001) (0.005) (0.002) Price of Maize Grain MK/Kg 0.045*** 0.085*** 0.000 (0.012) (0.015) (0.019) Recent illness/injury -0.215*** -0.213*** -0.209** (0.049) (0.060) (0.083) Do you suffer from a chronic illness -0.176** -0.219*** -0.172 (0.070) (0.085) (0.125) Other Members Chronically Ill in HH -0.153*** -0.113*** -0.282*** (0.032) (0.038) (0.057) Productive Assets Index B 8.078*** 6.945*** 11.463*** (0.831) (0.952) (1.709) Constant -2.028*** -2.156*** -2.388*** (0.262) (0.330) (0.437) Observations 12740 8399 4341 Standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 171 Appendix Table 1: Household consumption and biomass relationships for overall, non- poor, poor, north, center and south. (1) (2) (3) (4) (5) (6) Rural Non-Poor Poor North Central South Female household head -0.148 -0.104 -0.061 -0.117 -0.149 -0.158 [0.014]*** [0.017]*** [0.013]*** [0.041]*** [0.022]*** [0.020]*** Age of hh head: 26-35 years 0.058 0.049 0.026 0.032 0.053 0.059 [0.017]*** [0.017]*** [0.015]* [0.045] [0.025]** [0.025]** Age of hh head: 36-45 years 0.072 0.064 0.024 0.029 0.028 0.118 [0.019]*** [0.020]*** [0.017] [0.049] [0.028] [0.028]*** Age of hh head: 46-55 years 0.010 0.016 0.014 -0.029 -0.018 0.041 [0.019] [0.021] [0.018] [0.053] [0.030] [0.028] Age of hh head: 56-65 years -0.053 -0.039 -0.029 -0.026 -0.102 -0.024 [0.020]*** [0.022]* [0.019] [0.055] [0.032]*** [0.029] Age of hh head: 66+ years -0.099 -0.069 -0.023 -0.032 -0.134 -0.091 [0.022]*** [0.023]*** [0.020] [0.056] [0.032]*** [0.033]*** Widowed household head 0.058 0.053 0.028 0.079 0.039 0.052 [0.019]*** [0.023]** [0.017]* [0.056] [0.028] [0.028]* Household size -0.290 -0.216 -0.090 -0.255 -0.299 -0.301 [0.025]*** [0.013]*** [0.013]*** [0.028]*** [0.014]*** [0.042]*** Household size squared (/100) 1.313 1.182 0.288 1.079 1.461 1.286 [0.208]*** [0.103]*** [0.092]*** [0.209]*** [0.097]*** [0.361]*** Number of children 0-4 -0.059 -0.032 -0.027 -0.061 -0.047 -0.060 [0.009]*** [0.010]*** [0.008]*** [0.024]** [0.013]*** [0.015]*** Number of children 5-10 -0.028 0.012 -0.017 -0.002 -0.029 -0.028 [0.009]*** [0.010] [0.007]** [0.021] [0.012]** [0.014]** Number of children 11-14 -0.001 0.015 -0.003 -0.005 -0.013 0.012 [0.010] [0.012] [0.008] [0.025] [0.014] [0.016] Highest education: some 0.064 0.030 0.081 -0.057 0.072 0.064 primary [0.022]*** [0.021] [0.019]*** [0.109] [0.030]** [0.030]** Highest education: completed 0.161 0.101 0.122 -0.072 0.196 0.176 primary [0.026]*** [0.025]*** [0.022]*** [0.112] [0.036]*** [0.036]*** Highest education: post 0.364 0.219 0.194 0.074 0.357 0.443 primary [0.026]*** [0.025]*** [0.022]*** [0.112] [0.035]*** [0.037]*** Chronic illness in household 0.059 0.021 0.034 -0.033 0.046 0.082 [0.011]*** [0.012]* [0.010]*** [0.036] [0.015]*** [0.016]*** Religion: Islam -0.001 -0.008 -0.014 0.113 0.024 0.010 [0.025] [0.028] [0.021] [0.157] [0.038] [0.043] Religion: Catholic -0.001 -0.018 -0.039 -0.103 0.017 -0.008 [0.021] [0.023] [0.018]** [0.095] [0.025] [0.040] Religion: CCAP 0.061 0.019 -0.022 -0.082 0.083 0.083 [0.022]*** [0.024] [0.020] [0.092] [0.028]*** [0.045]* Religion: Other Christian -0.000 -0.021 -0.025 -0.117 0.008 -0.004 [0.019] [0.021] [0.017] [0.090] [0.023] [0.038] HH has wage/salary income 0.114 0.060 0.052 0.117 0.107 0.113 [0.011]*** [0.013]*** [0.010]*** [0.036]*** [0.017]*** [0.016]*** Household has a non-farm 0.142 0.053 0.066 0.181 0.136 0.134 enterprise [0.011]*** [0.012]*** [0.010]*** [0.030]*** [0.017]*** [0.015]*** HH grew tobacco in last 0.072 0.032 0.007 0.107 0.079 0.029 172 cropping season [0.015]*** [0.015]** [0.013] [0.033]*** [0.020]*** [0.032] Total hectares of dimba plots 0.073 0.044 0.049 0.037 0.066 0.101 [0.023]*** [0.023]* [0.018]*** [0.029] [0.027]** [0.059]* Total hectares of rain-fed plots 0.111 0.066 0.051 0.112 0.126 0.086 [0.007]*** [0.007]*** [0.008]*** [0.017]*** [0.010]*** [0.011]*** Regular bus service in 0.006 0.008 -0.020 -0.006 -0.027 0.011 community [0.012] [0.013] [0.010]* [0.032] [0.018] [0.018] Health clinic in community 0.061 0.030 0.051 0.069 0.099 0.032 [0.011]*** [0.012]** [0.010]*** [0.029]** [0.018]*** [0.016]** EA is a Boma or Trading 0.203 0.169 0.046 0.331 0.242 0.137 center [0.025]*** [0.026]*** [0.023]** [0.068]*** [0.043]*** [0.036]*** Travel to nearest boma: >20- -0.009 -0.030 0.021 0.007 0.073 -0.057 30mins [0.016] [0.018]* [0.014] [0.055] [0.026]*** [0.023]** Travel to nearest boma: >30- -0.102 -0.058 -0.027 -0.267 -0.008 -0.103 45mins [0.016]*** [0.018]*** [0.013]** [0.050]*** [0.022] [0.024]*** Travel to nearest boma: >45- -0.039 -0.020 -0.025 -0.278 0.074 -0.044 60mins [0.017]** [0.019] [0.015]* [0.052]*** [0.024]*** [0.027]* Travel to nearest boma: -0.018 -0.007 0.000 -0.128 0.103 -0.068 >60mins [0.017] [0.019] [0.014] [0.052]** [0.028]*** [0.024]*** ADMARC market in the -0.050 -0.035 -0.032 0.032 -0.077 -0.049 community [0.014]*** [0.016]** [0.013]** [0.044] [0.024]*** [0.020]** Bank in community -0.014 -0.032 0.004 0.423 -0.042 -0.017 [0.022] [0.025] [0.020] [0.096]*** [0.035] [0.032] Daily market in community 0.004 0.039 -0.004 -0.077 0.047 0.014 [0.013] [0.015]*** [0.011] [0.037]** [0.024]** [0.018] Tarmac/asphalt road in 0.029 0.016 0.045 -0.194 0.194 -0.003 community [0.017]* [0.020] [0.015]*** [0.044]*** [0.032]*** [0.025] Biomass (m3/ha) 0.0010209 0.0000051 0.0015476 0.0016728 -0.00017 0.0025929 [0.001]** [0.001] [0.0004]*** [0.001] [0.001] [0.001]*** Biomass / Ha Squared -0.0000194 -0.000004 -0.00002 -0.000018 -0.000008 -0.000035 [0.000005]* [0.000007] [0.000005]* [0.00001]* [0.000009] [0.000008]* ** ** ** North region 0.007 [0.029] Central region 0.270 [0.022]*** ADD: Karonga 0.016 0.012 -0.097 [0.034] [0.024] [0.032]*** ADD: Mzuzu 0.113 0.077 0.042 [0.028]*** [0.029]*** [0.022]* ADD: Kasungu 0.080 0.215 0.133 [0.025]*** [0.021]*** [0.027]*** ADD: Salima -0.088 0.054 0.119 [0.024]*** [0.030]* [0.024]*** ADD: Lilongwe -0.043 0.128 0.080 0.092 173 [0.017]*** [0.024]*** [0.019]*** [0.026]*** ADD: Machinga -0.047 0.005 -0.055 -0.053 [0.023]** [0.027] [0.020]*** [0.025]** ADD: Blantyre 0.004 0.023 0.003 -0.012 [0.021] [0.025] [0.018] [0.022] Observations 9752 5140 4612 1415 3819 4518 R-squared 0.44 0.26 0.21 0.43 0.45 0.44 Robust standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 174 Appendix Table 2: Labor allocation for fuelwood collection by men 10 years and older Male (1) Collect (2) Hours (3) Collect (4) Hours (5) Collect (6) Hours Overall Less Biomass More Biomass Age in years 0.002 0.053* -0.003 0.079** 0.013 0.005 (0.009) (0.029) (0.010) (0.039) (0.015) (0.037) Age Square 0.0000 -0.0006* 0.0001 -0.0009** -0.0001 0.0000 (0.0001) (0.0003) (0.0001) (0.0004) (0.0002) (0.0004) Head Education: Snr. Prim or more -0.092** -0.151 -0.054 -0.028 -0.169** -0.419* (0.045) (0.161) (0.054) (0.205) (0.083) (0.232) HH Size -0.040*** -0.051 -0.036*** -0.048 -0.052*** -0.055 (0.009) (0.041) (0.011) (0.055) (0.017) (0.052) Real Value of Assets '000 -0.002 -0.003 -0.002 0.000 -0.006 -0.015 (0.001) (0.006) (0.001) (0.007) (0.005) (0.016) Wet Season & Wet Area Interaction 0.009 0.123 -0.053 -0.032 0.148** 0.336** (0.033) (0.099) (0.043) (0.177) (0.058) (0.149) Dry Season and Dimba Area Interaction 0.144 -0.121 0.127 -0.163 0.207 -0.334 (0.093) (0.269) (0.105) (0.311) (0.205) (0.616) Total hectares of land uncultivated 0.038 0.207 0.050 0.190 0.011 0.361* (0.038) (0.131) (0.051) (0.166) (0.067) (0.194) Total hectares of trees 0.081 0.364 0.439 1.040 -0.965 -3.707 (0.408) (1.471) (0.468) (1.816) (0.951) (2.563) Predicted Real Wage for Women -0.001 -0.002 0.000 -0.002 0.000 -0.001 (0.001) (0.002) (0.001) (0.003) (0.001) (0.002) Number of <5yr children 0.077 -0.009 0.131 0.048 -0.122 -0.634 (0.118) (0.420) (0.131) (0.517) (0.276) (0.707) Married -0.468*** -0.368 -0.467*** -0.243 -0.482*** -0.613 (0.061) (0.365) (0.073) (0.496) (0.110) (0.420) Rural Centre 0.475*** 0.557 0.441*** 0.552 0.538*** 0.821 (0.098) (0.494) (0.138) (0.752) (0.152) (0.520) Rural South 0.536*** 0.832 0.535*** 0.611 0.524*** 1.450*** (0.098) (0.524) (0.142) (0.810) (0.145) (0.513) Wet Season Dummy 0.027 -0.096 0.053 0.051 -0.043 -0.144 (0.055) (0.177) (0.068) (0.262) (0.099) (0.225) ADMARC Market Distance KM 0.0008*** 0.0046*** 0.0009*** 0.0042*** -0.0021 0.0047 (0.0003) (0.0010) (0.0003) (0.0012) (0.0030) (0.0077) Distance to Asphalt Road KM -0.0005 -0.0052 0.0006 -0.0115* -0.0020 -0.0026 (0.0009) (0.0039) (0.0011) (0.0065) (0.0014) (0.0040) Elevation 0.00010 0.00016 0.00015 -0.00002 0.00005 0.00050 (0.00008) (0.00030) (0.00009) (0.00041) (0.00014) (0.00037) Biomass20 (m3/ha) -0.001 -0.003 0.001 -0.015 0.001 0.000 (0.001) (0.003) (0.004) (0.017) (0.002) (0.004) Price of Maize Grain MK/Kg -0.012 -0.007 -0.016 -0.001 -0.015 -0.030 (0.010) (0.036) (0.013) (0.051) (0.020) (0.047) Recent illness/injury 0.113** 0.050 0.078 -0.116 0.180** 0.401* (0.047) (0.179) (0.058) (0.233) (0.083) (0.234) Chronic illness 0.030 -0.078 0.122 -0.013 -0.199 -0.264 (0.064) (0.208) (0.075) (0.291) (0.127) (0.317) Other Members Chronically Ill 0.048* -0.061 0.055* -0.136 0.041 0.041 175 (0.026) (0.092) (0.032) (0.132) (0.047) (0.110) Substitute for collected fuelwood -0.395*** -0.354*** -0.614*** (0.072) (0.080) (0.176) Constant -1.454*** -1.408 -1.400*** -0.766 -1.487*** -3.020 (0.250) (1.837) (0.314) (2.577) (0.452) (1.966) Observations 11169 11169 7336 7336 3833 3833 Inverse Mills Ratio 1.041 0.693 1.972** (0.801) (1.110) (0.813) Standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 176 Appendix Table 3: Labor allocation hours for household agriculture by men 10 years and older (1) (2) (3) Overall Less Biomass Areas More Biomass Areas Age in years 0.113*** 0.115*** 0.116*** (0.010) (0.013) (0.017) Age Square -0.001*** -0.001*** -0.001*** (0.000) (0.000) (0.000) Head Education: Snr. Prim or more -0.164*** -0.166** -0.137 (0.056) (0.069) (0.093) HH Size -0.021** -0.032** -0.031* (0.011) (0.013) (0.019) Real Value of Assets '000 -0.003*** -0.003*** -0.004*** (0.001) (0.001) (0.001) Wet Season & Wet Area Interaction 0.252*** 0.259*** 0.274*** (0.039) (0.045) (0.077) Dry Season and Dimba Area Interaction 1.063*** 0.982*** 1.040*** (0.132) (0.159) (0.235) Total hectares of land uncultivated 0.030 0.139* -0.121* (0.047) (0.072) (0.062) Total hectares of trees -0.233 -1.166* 1.047* (0.463) (0.692) (0.607) Predicted Real Wage for Women -0.005*** -0.005*** -0.004*** (0.001) (0.001) (0.001) Number of <5yr children -0.114 -0.149 -0.004 (0.141) (0.168) (0.259) Married 0.453*** 0.434*** 0.448*** (0.079) (0.098) (0.131) Rural Centre -0.530*** -0.767*** -0.038 (0.097) (0.132) (0.151) Rural South -0.487*** -0.684*** -0.380*** (0.095) (0.135) (0.138) Wet Season Dummy 0.740*** 0.616*** 0.950*** (0.066) (0.082) (0.114) ADMARC Market Distance KM 0.002*** 0.001*** 0.018*** (0.001) (0.001) (0.003) Distance to Asphalt Road KM 0.009*** 0.005*** 0.011*** (0.001) (0.001) (0.001) Elevation 0.000 -0.000 0.000 (0.000) (0.000) (0.000) Biomass20 (m3/ha) -0.006*** 0.004 -0.004** (0.001) (0.006) (0.002) Price of Maize Grain MK/Kg -0.001 0.056*** -0.088*** (0.013) (0.017) (0.023) Recent illness/injury -0.244*** -0.282*** -0.189* (0.060) (0.075) (0.098) Chronic illness 0.107 0.205** -0.232* (0.083) (0.102) (0.141) Other Members Chronically Ill -0.075** -0.068 -0.161*** (0.035) (0.043) (0.059) Productive Assets Index B 11.334*** 9.565*** 19.574*** (0.854) (0.974) (1.828) Constant -0.193 -0.644* 0.080 (0.302) (0.381) (0.507) Observations 10935 7186 3749 Standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 177 ANNEX 2C: DISTRIBUTION AND BREAKDOWN OF TIME USE BY LOCATION AND GENDER Figure 1: Distribution of individual working time by sex and area (individuals aged 15+) Men Women 4 2 .0 .0 3 .0 51 .0 ytisn 2 .0 ytisn 1 .0 De De 1 .0 50 .0 0 0 0 50 100 150 0 50 100 150 Total hours (water,firewood,ag,non-ag,salary/wage) Total hours (water,firewood,ag,non-ag,salary/wage) Urban areas Rural areas 4 .0 2 .0 3 .0 51 .0 yti 2 yti ensD .0 ensD 1 .0 1 .0 50 .0 0 0 0 50 100 150 0 50 100 150 Total hours (water,firewood,ag,non-ag,salary/wage) Total hours (water,firewood,ag,non-ag,salary/wage) Source: Authors' estimation using IHS2. 178 Table 1: Total time spent working by area and consumption quintile, national sample Poorest Richest quintile 2nd quintile 3rd quintile 4th quintile quintile Total National, adults (age 15 and over) March 2004 34.0 32.6 36.0 35.6 38.4 35.7 April 2004 34.1 34.7 35.6 35.7 37.8 35.8 May 2004 28.3 31.7 33.7 34.5 38.3 33.8 June 2004 34.0 33.2 35.6 36.0 35.4 35.1 July 2004 28.4 33.1 32.2 34.4 35.2 33.5 August 2004 33.3 34.9 33.2 32.9 36.3 34.3 September 2004 34.5 35.5 35.4 38.3 34.5 35.7 October 2004 34.9 39.2 38.5 38.6 37.5 37.6 November 2004 39.8 40.2 40.3 38.1 40.3 39.8 December 2004 44.5 42.1 42.1 37.9 41.3 41.7 January 2005 38.9 41.1 43.1 41.1 42.1 41.2 February 2005 35.1 36.1 37.6 38.8 35.3 36.5 March 2005 33.8 38.5 36.7 36.2 41.2 37.0 Annual average 35.0 36.2 36.5 36.3 37.5 36.4 National, children (below 15 years old) March 2004 8.0 7.7 6.9 8.5 8.5 7.9 April 2004 9.5 10.7 9.9 10.4 10.1 10.2 May 2004 5.6 7.8 7.3 9.8 6.4 7.3 June 2004 5.3 6.8 8.3 7.9 7.4 7.2 July 2004 5.4 7.2 7.5 8.6 7.8 7.6 August 2004 8.3 9.7 9.8 9.3 10.8 9.6 September 2004 6.8 8.2 7.3 8.2 7.2 7.5 October 2004 7.4 8.4 8.6 6.9 7.6 7.8 November 2004 8.3 9.6 8.6 10.9 8.1 9.0 December 2004 12.8 12.3 13.2 15.4 11.3 12.9 January 2005 8.8 7.7 7.9 7.8 7.6 8.1 February 2005 7.4 8.2 9.0 9.4 7.3 8.1 March 2005 6.5 10.1 7.6 10.4 12.4 8.5 Annual average 7.7 8.9 8.6 9.3 8.6 8.5 Source: Authors' estimation using IHS2. 179 Table 2: Total time spent working by area and consumption quintile, rural areas Poorest Richest quintile 2nd quintile 3rd quintile 4th quintile quintile Total Rural areas, adults (age 15 and over) March 2004 34.3 32.9 36.4 37.3 39.5 36.3 April 2004 34.1 34.8 36.0 36.2 38.4 36.1 May 2004 29.0 31.8 35.0 34.8 38.2 33.8 June 2004 34.0 33.3 36.1 36.1 35.7 35.2 July 2004 27.4 33.1 31.9 34.8 35.6 33.6 August 2004 33.7 34.6 32.9 32.7 35.4 33.9 September 2004 33.7 35.0 35.0 37.9 35.7 35.6 October 2004 35.1 39.3 38.7 37.8 38.8 37.8 November 2004 40.1 40.7 41.2 39.4 41.5 40.6 December 2004 45.0 42.9 43.1 38.5 42.0 42.5 January 2005 39.1 40.2 41.5 41.4 44.2 41.1 February 2005 35.5 36.6 38.2 38.8 35.1 36.8 March 2005 34.3 38.2 37.5 35.9 40.2 36.9 Annual average 35.2 36.2 36.8 36.5 37.8 36.5 Rural areas, children (below 15 years old) March 2004 8.1 7.9 7.2 9.4 9.8 8.4 April 2004 9.6 11.0 10.0 10.8 10.9 10.4 May 2004 5.7 7.9 7.4 9.8 6.9 7.4 June 2004 5.3 6.4 8.2 7.7 7.9 7.1 July 2004 5.2 7.0 7.5 9.0 8.0 7.7 August 2004 8.5 9.6 9.8 8.8 11.2 9.5 September 2004 6.8 7.7 6.4 8.1 7.5 7.2 October 2004 7.5 8.5 8.7 6.5 9.2 8.0 November 2004 8.3 9.5 8.9 12.3 6.9 9.2 December 2004 12.9 12.3 15.0 15.5 8.7 13.2 January 2005 7.9 7.7 7.8 7.5 9.5 7.9 February 2005 7.6 8.3 8.4 6.2 6.5 7.8 March 2005 6.8 10.0 7.3 7.4 7.5 8.0 Annual average 7.8 8.8 8.5 9.1 8.9 8.5 Source: Authors' estimation using IHS2. 180 ginkr er sr Mo anth .6 181 4 Wo hou07 08. 37. 45. 74. 35. 45. 77. 56. 67. 56. 29. 54. 17. .4 .9 .9 .4 .9 .8 .1 .6 .2 .1 .3 .1 17 12 10 14 10 11 12 12 14 12 13 11 10 200 ginkr anth sr .4 .0 .3 .0 .7 .8 .6 .7 .3 .1 .3 .8 .8 .9 .7 .5 Wo less hou01 27 18 24 22 19 21 19 17 12 86. 88. 16 16 11 11 10 10 11 18. 69. 77. 17. 36. 47. 58. 96. national,­i tal rk .5 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .5 .0 .0 .5 .0 .5 .5 To wo edian) 24 30 26 30 27 25 28 30 34 36 35 30 30 38 37 38 38 35 37 38 43 45 45 46 41 40 (m rk Malaw wo ean) .1 .5 .3 .6 .0 .6 .0 .1 .4 .7 .9 .6 .2 .9 .8 .0 .3 .7 .6 .3 .7 .2 .9 .9 .9 .5 tal (m 29 31 28 29 29 28 31 31 34 36 35 30 32 41 39 39 40 37 39 40 43 45 45 45 41 41 To survey, the in rk Salaried wo 06. 36. 66. 58. 75. 06. 09. 46. 98. 75. 67. 76. 57. dedr naloi naloi 41. 21. 31. 51. 21. 11. 21. 61. 32. 01. 81. 21. 61. co nat nat re art-p u al, emti ny rk Ga wo time & ),ervod er), 52. 52. 81. 52. 23. 52. 03. 43. 72. 23. 52. 22. 62. of Casu an 15 ovdna51 71. 21. 01. 71. 61. 01. 90. 01. 31. 71. 21. 61. 31. ing r ag. fo essn Help on-n 70. 70. 80. 70. 50. 40. 40. 20. 20. 30. 10. 20. 50. busi (agesela 11. 60. 60. 50. 30. 20. 40. 30. 30. 20. 50. 20. 30. categories the ing ag. to nn essn (ageselamtul femtul Ad Ru on-n 43. 45. 54. 83. 45. 75. 93. 14. 03. 24. 43. 13. 93. Ad 22. 52. 61. 32. 62. 03. 52. 02. 61. 61. 01. 31. 51. busi ral according rk .9 .5 .8 .9 .5 .7 .8 .8 .8 .6 .5 .6 .2 .5 .0 .2 .0 .0 .6 .4 .2 .4 .1 .2 age ricultu wo 13 13 11 10 10 10 10 13 15 20 18 15 14 12 13 13 11 68. 69. 11 14 16 20 17 15 13 and gAg llectin fire- oodw 50. 50. 30. 30. 60. 30. 40. 30. 40. 30. 30. 30. 30. 03. 12. 32. 32. 42. 32. 12. 12. 02. 61. 02. 71. 12. . month, Cog llectin water 50. 60. 60. 90. 90. 70. 80. 60. 70. 60. 01. 70. 70. 35. 15. 45. 16. 46. 07. 96. 96. 76. 95. 87. 46. 86. IHS2g gender, Co by , ing usinnioat dry .8 .0 .6 .0 .5 .3 .2 .3 .6 .6 .2 .1 .7 time ok Co unal and 61. 91. 91. 91. 22. 42. 72. 22. 62. 81. 42. 81. 42. 14 14 13 15 14 15 15 15 14 13 14 14 14 cleaning estim s'r Work tho 04 4 05 04 4 05 3: 20hc Au 200 0042 0042 ne 0402y 2004 04 04 st 20.t 20 20hc 20hc 2004 04 200 0042 0042 st 20.t 0402 04 20 0502 0502 20hc ce:r Table Mar April May Ju Jul uguA Sep 0042t.cO 0402 0502 0502 v. No ec.D ne 0402y Jan. Feb. Mar Mar April May Ju Jul uguA Sep 0042t.cO v. No ec.D Jan. Feb. Mar Sou ginkr anhter sr 92. 62. 02. 22. 01. 22. 02. 11. 32. 11. 60. 31. 31. 03. 72. 31. 41. 90. 62. 02. 41. 12. 90. 60. 11. 23. 182 Wo 4 Mo hou07 200 ginkr anth sr .0 .3 .2 .1 .3 .9 .5 .6 .3 Wo less hou01 .9 .0 .8 .9 .5 .7 .6 .6 .5 .2 .4 .5 .6 .7 .9 .1 .1 77 71 77 76 79 68 77 79 72 60 74 77 78 61 58 61 67 64 56 62 62 59 49 58 67 67 national,­i tal rk To wo edian) 00. 03. 50. 50. 00. 00. 00. 00. 01. 53. 00. 00. 00. 54. 07. 07. 53. 54. 07. 07. 07. 07. .0 10 07. 53. 53. (m Malaw tal rk ean) To wo 35. 87. 65. 06. 35. 27. 85. 55. 57. .9 .7 .5 .8 .0 .6 .2 .6 .7 (m 10 55. 66. 46. 10 12 29. 48. 79. 11 19. 10 10 15 10 79. 10 survey, the rk in Salaried wo 20. 10. 10. 10. 20. 20. 10. 00. 10. 10. 00. 10. 20. 30. 00. 00. 20. 00. 10. 10. 00. 50. 00. 00. 00. 30. dedr e u co al, ny rk naloi naloi re Ga Casu wo nat 20. 30. 30. 20. 40. 20. 20. 30. 30. 80. 20. 30. 20. nat 30. 20. 20. 30. 40. 10. 00. 10. 00. 60. 20. 10. 10. part-tim & ), ), time of ing r ag. fo essn 14ot5 14ot5 Help on-n 20. 20. 10. 10. 30. 20. 00. 10. 00. 00. 10. 00. 00. 60. 40. 10. 20. 40. 30. 20. 10. 10. 00. 00. 00. 00. busi (agesrl categories ing ag. nn essn ge(asyoB Gi the Ru on-n 20. 10. 20. 10. 10. 00. 00. 00. 00. 10. 00. 00. 10. 30. 20. 10. 10. 10. 00. 00. 10. 00. 10. 10. 00. 00. busi to ral rk ricultu wo 42. 54. 42. 32. 81. 03. 91. 82. 14. 86. 72. 33. 52. 12. 33. 91. 31. 01. 91. 01. 12. 92. 95. 33. 82. 62. according Ag g age and llectin fire- oodw 10. 30. 50. 40. 20. 30. 20. 20. 30. 40. 10. 20. 30. 80. 11. 70. 70. 01. 80. 80. 01. 60. 01. 40. 50. 70. Co g . month, llectin water 80. 01. 90. 21. 31. 41. 21. 01. 90. 11. 11. 11. 01. 03. 43. 03. 62. 53. 14. 43. 33. 13. 33. 63. 33. 82. Co IHS2g gender, , by ing dry ok Co unal and 01. 31. 21. 61. 01. 91. 12. 11. 81. 61. 31. 51. 02. 43. 83. 23. 13. 43. 44. 53. 43. 53. 34. 92. 92. 24. usinnioat time cleaning estim s'r Work tho 04 4 05 04 4 05 4: 20hc Au 200 0042 0042 ne 0402y 2004 04 04 st 20.t 20 20hc 20hc 2004 04 200 0042 0042 st 20.t 0402 04 20 0502 0502 20hc ce:r Table Mar April May Ju Jul uguA Sep 0042t.cO 0402 0502 0502 v. No ec.D ne 0402y Jan. Feb. Mar Mar April May Ju Jul uguA Sep 0042t.cO v. No ec.D Jan. Feb. Mar Sou ginkr anhter sr .1 .2 .4 .6 .1 .3 .0 .5 .0 .1 .9 .6 .3 Wo Mo hou07 18. 96. 84. 74. 94. 64. 07. 85. 86. 55. 47. 34. 25. 18 13 10 14 11 11 13 12 15 12 12 11 10 183 ginkr anth sr .4 .2 .2 .0 .1 .2 .4 .5 .9 .7 .6 .5 .8 .1 2004 Wo less hou01 26 17 24 22 19 22 18 16 10 04. 47. 15 15 10 10 68. 59. 11 67. 98. 96. 25. 54. 55. 47. 86. rural,­i tal rk .0 .0 .0 .0 .0 .0 .0 .0 .0 .5 .5 .0 .0 .5 .5 .5 .5 .0 .0 .0 .0 .0 .0 .0 .0 .0 To wo edian) 25 29 25 28 26 24 26 30 32 34 34 28 30 38 38 38 39 35 38 40 44 46 47 47 43 41 (m Malaw tal rk ean) .0 .9 .0 .3 .3 .9 .4 .3 .9 .2 .2 .9 .8 .0 .7 .9 .7 .5 .2 .4 .6 .2 .7 .4 .1 .4 To wo (m 29 30 27 28 28 26 29 30 33 36 35 29 30 43 40 39 41 38 40 41 44 47 47 46 43 42 survey, rk the 94. 63. 74. 16. 15. 44. 76. 54. 36. 03. 17. 25. 25. 80. 90. 40. 40. 21. 60. 50. 70. 11. 40. 71. 21. 01. in Salaried wo ral dedr e u ru ralur al, ny rk er), co Ga re Casu wo part-tim & ),ervod 62. 52. 91. 62. 92. 42. 72. 43. 13. 13. 42. 32. 82. time r an fo 15 ovdna51 81. 11. 11. 91. 61. 01. 80. 11. 51. 02. 01. 71. 51. of ag. ing essn on-n (age 70. 70. 50. 70. 60. 30. 20. 20. 20. 30. 10. 20. 20. (age 21. 60. 30. 50. 40. 20. 40. 30. 30. 20. 50. 30. 20. busi Help lesam lesa fem categories ing ag. ult ult the nn essn Ad on-n 62. 95. 63. 33. 94. 44. 63. 04. 22. 82. 43. 03. 33. Ad 91. 42. 31. 02. 42. 52. 12. 71. 21. 01. 90. 31. 41. to Ru busi ral rk .6 .2 .9 .6 .2 .2 .8 .5 .8 .7 .4 .7 .3 .6 .1 .8 .4 .6 .4 .8 .0 .3 .2 .9 .8 according ricultu wo 15 15 13 12 11 12 12 15 18 24 19 16 16 13 14 14 12 39. 10 12 15 19 23 18 15 14 Ag age g and llectin fire- oodw 60. 50. 40. 40. 60. 30. 40. 30. 40. 30. 30. 30. 30. 33. 22. 62. 62. 52. 52. 42. 32. 32. 81. 91. 81. 22. Co . month, g llectin water 50. 60. 60. 90. 80. 70. 80. 60. 70. 60. 90. 70. 80. 65. 45. 85. 76. 76. 37. 47. 47. 47. 46. 28. 86. 37. IHS2g gender, Co by , ing dry .7 .0 .6 .1 .5 .4 .4 .3 .3 .6 .8 .0 .0 usinnioat ok time Co unal and 51. 71. 51. 61. 12. 22. 42. 91. 12. 31. 51. 61. 91. 14 14 13 15 14 15 15 15 14 12 13 14 14 estim cleaning s'r Work tho 5: 04 4 05 04 4 05 20hc Au 200 0042 0042 ne 0402y 2004 04 04 st 20.t 20 20hc 20hc 2004 04 200 0042 0042 st 20.t 0402 04 20 0502 0502 20hc ce:r Table Mar April May Ju Jul uguA Sep 0042t.cO 0402 0502 0502 v. No ec.D ne 0402y Jan. Feb. Mar Mar April May Ju Jul uguA Sep 0042t.cO v. No ec.D Jan. Feb. Mar Sou ginkr anhter sr Wo Mo hou07 23. 62. 32. 12. 70. 12. 32. 70. 42. 11. 70. 41. 21. 13. 42. 41. 31. 01. 72. 91. 41. 71. 70. 70. 21. 13. ginkr anth sr .5 .9 .5 .0 .2 .4 .0 .9 .7 Wo less hou01 .6 .5 .0 .9 .8 .0 .3 .8 .4 .8 .0 .9 .7 .5 .1 .9 .1 184 76 70 78 77 78 70 79 79 72 58 75 78 78 58 57 60 67 65 56 62 61 57 49 58 67 66 rk 2004 wo tal edian) 00. 03. 00. 00. 00. 00. 00. 00. 02. 04. 00. 00. 00. 06. 07. 07. 53. 54. 07. 07. 07. 07. .5 10 07. 53. 53. To (m rural,­i tal rk ean) To wo 65. 08. 65. 75. 65. 96. 35. 65. 67. .3 .3 .0 .1 .3 .8 .4 .3 .1 (m 11 05. 16. 06. 11 13 49. 48. 79. 12 09. 10 10 15 10 49. 10 Malaw rk wo 30. 10. 10. 10. 20. 20. 20. 00. 10. 10. 00. 00. 20. 30. 00. 00. 10. 00. 10. 10. 00. 10. 00. 00. 00. 10. survey, Salaried the e al, nyu in rk Ga Casu wo dedr part-tim & ralur 30. 30. 30. 20. 50. 20. 20. 30. 30. 90. 10. 30. 20. ralur 30. 20. 20. 40. 50. 10. 00. 10. 00. 70. 10. 10. 10. r ), ), co fo re ag. ing essn on-n 15ot5 15 30. 30. 10. 10. 20. 20. 10. 10. 00. 00. 20. 10. 00. to5 60. 50. 10. 10. 20. 30. 20. 10. 10. 00. 00. 00. 00. busi time Help (age of ing ag. nn essn ge(asyoB Girls Ru on-n 20. 10. 30. 10. 10. 00. 10. 00. 00. 10. 00. 00. 10. 30. 20. 10. 10. 10. 00. 00. 20. 00. 10. 10. 00. 00. busi categories the to ltural rk wo 72. 74. 62. 52. 91. 23. 12. 03. 54. 77. 72. 53. 72. 22. 63. 12. 41. 01. 12. 21. 22. 33. 96. 23. 92. 82. Agricu according g age llectin fire- oodw 10. 30. 50. 40. 20. 30. 20. 20. 30. 40. 10. 20. 40. 90. 21. 80. 80. 01. 90. 90. 11. 70. 11. 40. 60. 90. and Co S. g month, HI4 llectin water 90. 01. 90. 31. 41. 41. 21. 01. 90. 21. 21. 21. 01. 23. 53. 23. 92. 53. 44. 73. 43. 43. 63. 83. 53. 03. 200g Co gender, by , ing dry usinnioat ok time Co unal and 90. 21. 90. 11. 01. 31. 31. 90. 51. 90. 80. 80. 21. 63. 83. 03. 72. 33. 04. 92. 33. 23. 03. 52. 32. 23. estim cleaning s'r Work tho 04 4 05 04 4 05 6: 20hc Au 200 0042 0042 ne 0402y 2004 04 04 st 20.t 20 20hc 20hc 2004 04 200 0042 0042 st 20.t 0402 04 20 0502 0502 20hc ce:r Table Mar April May Ju Jul uguA Sep 0042t.cO 0402 0502 0502 v. No ec.D ne 0402y Jan. Feb. Mar Mar April May Ju Jul uguA Sep 0042t.cO v. No ec.D Jan. Feb. Mar Sou ANNEX 2D: SUBJECTIVE WELLBEING This section is based on the analysis of answers on subjective wellbeing (SWB) questions. Quality of life, happiness and well-being are broad, multi- dimensional concepts that include not only material achievements but also other aspects of life, such as health, respect of others, employment, and having children. This implies that objective economic indicators (income or expenditure) fall short on assessing fully the satisfaction with life. SWB questions are important to poverty practitioners for various reasons. First, they can be used to examine the structure of household's welfare and wellbeing and thus assist in understanding households' preferences and in predicting behavior. Second, SWB questions allow the evaluation of many socio-economic policies. Third, measuring welfare and well-being contributes to the assessment of distributional problems, as well as to the understanding of who is, or is not, relatively well-off, and why. Fourth, understanding the structure of welfare and wellbeing sheds light on the potential trade-off between such factors as income, health and children. Box A3.1 A special section of the IHS2 includes questions about the subjective wellbeing of each household. The current analysis is based on three sets questions. A block of these questions deals with the adequacy of consumption expenditure for major consumption groups. The questions are formulated as follows: Concerning your household's food consumption over the past one month, which of the following is true? 1. It was less than adequate for household needs 2. It was just adequate for household needs 3. It was more than adequate for household needs Similar questions are asked about expenses related to clothing, housing, and health. We call these the Consumption Adequacy Questions (CAQ). The minimum income question (MIQ) was asked in the following form: What income level do you personally consider to be absolutely minimal ­ below which you could not make ends meet? The economic ranking questions (ERQ) were formulated as following: Imagine six steps, where on the bottom, the first step, stand the poorest people, and on the highest step, the sixth, stand the rich. 1. On which step are you today? 2. On which step are most of your neighbours today? 3. On which step are most of your friends today? Consumption Adequacy 185 The distribution of answers to consumption adequacy questions (CAQ) for food, housing, clothing and health care are shown in Table A3.1. More than half of households in Malawi indicated that their expenditures on food, clothing, and housing are less than adequate to meet the households' needs. Almost 73 percent of households are dissatisfied with their expenditure on clothing. For all consumption categories less than 6 per cent of the households perceived their expenditures as more than adequate. Table A3.1: Perceived adequacy of consumption in Malawi, proportion of population. Less then adequate Just Adequate More than adequate Food 57.42 36.50 6.08 Housing 55.41 39.99 4.58 Clothing 72.58 25.61 1.78 Health 60.94 35.93 3.11 Figure A3.1 shows how the subjective perceptions about the consumption adequacy depend on the level of household income. Approximately 75 percent of households from the lowest deciles of expenditure distribution consider their expenditure on food as inadequate. At the same time, that proportion is twice as low for the wealthiest households. A similar tendency could be observed for expenditures on clothing. Poor households are much more likely to categorize their expenditures on clothing as inadequate compared to better-off households. The proportion of households with perceived inadequate expenditures on clothing declines from 80 percent for the lowest expenditure deciles to less than 40 percent for the highest deciles. Trends in perceived adequacy of housing and health expenditures reveal lower income elasticities for these consumption groups. In particular, the share of households who thought that their health expenditures were inadequate stays almost constant up to the 60th percentile of expenditure distribution. The perceptions of consumption adequacy vary geographically. Table A3.2 presents the average proportion of households with less than adequate expenditures across different regions of Malawi. The highest proportion of households dissatisfied with food consumption is found in Rural South (64%) and Rural Central (59%) regions. On the other hand, only 35 percent of families living in Rural North perceived their food expenditures as inadequate. The rankings of expenditure adequacy for housing and health show similar patterns. About 60 percent of households in Rural Centre and Rural South thought that they did not spend enough on housing. Households living in Urban and Rural North areas seemed to be more satisfied with their housing expenditure levels. Satisfaction with clothing expenditures is very low in Rural Center and Rural South. Even households residing in the relatively well-off, according to other indicators, Urban and Central North regions report low levels of satisfaction 186 with expenses on clothing. Overall, households from the poorest regions of Malawi were less satisfied with expenditure levels than households from the richer regions. Table A3.2: Regional differences in the proportion of households who consider their level of expenditures inadequate by consumption categories. Consumption category Food Housing Clothing Health Poor Urban 48.27 44.05 54.52 51.78 15.41 Rural North 35.11 36.13 52.07 42.96 48.47 Rural Centre 59.20 60.32 84.24 65.55 31.67 Rural South 63.87 58.78 71.78 63.66 51.31 Total 57.40 55.39 72.56 60.93 39.48 Table A3.3 shows the proportion of households with inadequate consumption by type of household and consumption category. Again, general trends are clear ­ poorer (and larger) households were least satisfied with their level of consumption. In comparison with other household types, poor households consisting solely of elderly members had the second lowest objective poverty rate (14%). Nonetheless, 63 percent of these households considered their food consumption inadequate. The poverty rate of extended households (46%) is almost three times higher than that of elderly households, and 59 percent of such households indicated that their expenditures on food were inadequate. Interestingly, even though the poverty rates among the single adults were the lowest among all household types, still at least 40 percent of such households indicated consumption inadequacy in all categories. Single- parent households were among the least satisfied for almost all consumption categories. Table A3.3: Proportion of households who consider their level of expenditure inadequate by type of household* and consumption category. Consumption category Food Housing Clothing Health Poor Households with elderly only 63.09 50.56 80.73 70.51 14.35 Single parent with less than 3 children 65.54 52.17 77.63 58.08 35.68 Nuclear family with less than 3 children 55.71 53.23 71.36 59.44 31.19 Extended family with less than 3 children 59.07 54.79 74.97 67.53 45.52 Households with 3 and more children 59.64 61.04 75.57 61.70 54.62 Single adult households 45.56 40.41 62.27 52.11 0.90 Total 57.40 55.39 72.56 60.93 39.48 187 *) Household typology is not exhaustive. Economic Ranking Questions (ERQ) Figure A3.2 presents the non-parametric estimation of the relationship between the answers of ERQ (own, friends and neighbors) by per capita expenditure. The graph indicates a strong correlation between the subjective economic rankings and objective economic welfare measure. The poorer people are the lower is their own perceived economic ranking and the economic rankings of their friends and neighbors. Table A3.4 shows the regional distribution of average own, friends and neighbors subjective economic ranks. Households living in Rural Centre and Rural South areas report lowest ranking of their economic wellbeing. Households residing in Urban Malawi perceive their economic wellbeing as the highest among all other regions. Interestingly, the gap between the own economic ranking and those of friends and neighbors is the largest for household from Rural North. Household from the poor Rural South give low ranking to living standards of their friends and neighbors. Table A3.4: Average own economic rank and economic ranks of friends and neighbors by regions of the country Average economic rank Own Friends Neighbors Urban 2.29 2.69 2.49 Rural North 1.90 2.36 2.09 Rural Centre 1.73 2.14 1.87 Rural South 1.63 2.00 1.67 Total 1.77 2.17 1.88 The multivariate analysis of factors affecting the subjective economic wellbeing shows (second column of Table A3.6, below) that level of per capita consumption expenditure is a strong determinant of the subjective economic wellbeing ­ objectively better-off households perceived themselves richer. However, economic perception is not fully proportional to household's income. Controlling for the level of household per capita expenditures large households, households with larger share of adult females, households owing an enterprise report higher levels of economic wellbeing. Relative to the households with the head holding university degree, households with less educated heads have a lower perception of their economic rank. Overall, rural households rank their economic wellbeing lower than similar households from the urban areas. Coefficients on the regional dummies indicate that households residing in Lilongwe are less satisfied with their economic wellbeing, compared to the households from Karonga, Mzuzu, Kasungu, and Machinga. Objectively poor 188 regions of Ngabu, Salima, and Blantyre report lower rankings of economic wellbeing than the households from the capital. Subjective Poverty Lines The subjective poverty line (SPL) is calculated by two methods (see for the review of methods Lokshin, Umapathi, and Paternostro 2005). The first method is based on the minimum income question and the second utilize the CAQ. Table A3.6 at the end of this section gives the estimates for the ordered probit and linear regression models for CAQ and minimum income questions MIQ respectively. Using these estimates it is possible to derive the so-called subjective poverty lines. These poverty lines could vary by household and regional characteristics reflecting different needs of the households with different composition residing in various regions of the country. On average, Log subjective poverty line derived based on MIQ is 11.04 and the Log SPL based on CAQ is equal to 10.21. Table A3.5 shows the estimated subjective poverty rates for the selected socio-demographic groups. On the national level, almost 80 percent of household have their expenditures lower than the MIQ based poverty line an 69 percent of household are poor according to CAQ poverty line. For all groups of household subjective poverty rates are higher than the poverty rates calculated based on the absolute poverty line. However, looking at the ranking of the groups by poverty rates, the MIQ-based poverty ranking is close to rankings based on the absolute poverty line. For example, households consisting of elderly only and households with a single adult with no children are two least objectively poor households. These two groups of households also have lowest subjective poverty rates. The poorest according to both definitions are the households with more than three children, followed by extended households with children. At the same time, CAQ-based poverty rates demonstrate different trends. In the context of Malawi that most likely indicates that CAQ are not precise enough to derive meaningful poverty inferences. Table A3.5: Subjective poverty rates and poverty ranking calculated based on MIQ and CAQ, objective poverty rates and rankings. MIQ CAQ Objective poverty Rate Rank Rate Rank Rate Rank Households with elderly only 0.21 1.00 0.86 6.00 0.14 2.00 Single parent with less than 3 children 0.73 3.00 0.86 5.00 0.36 4.00 Nuclear family with less than 3 children 0.86 4.00 0.69 2.00 0.31 3.00 Extended family with less than 3 children 0.89 5.00 0.79 4.00 0.46 5.00 Households with 3 and more children 0.98 6.00 0.75 3.00 0.55 6.00 Single adult households 0.22 2.00 0.53 1.00 0.01 1.00 189 Figure A3.3 demonstrate similar patterns of the changes in the objective and subjective poverty rates by household size. Both MIQ and absolute poverty rates increase monotonically for larger households. CAQ-based poverty rate however, is almost constant, or even inverted U-shaped, increasing slightly for the households with up to seven members and then declining for larger households. Box A3.2 Ravallion and Lokshin 2005 looked at the effect of relative deprivation and positive externality among households in Malawi. The relative deprivation effect arises from social comparisons, whereby the utility of certain consumption goods depends on one's own consumption relative to some salient reference group; by this social effect, having well-off friends and neighbors is a source of disutility. The positive external effects arise instead from institutions for risk-sharing, the provision of local-public goods or from local employment opportunities or productivity-enhancing spillovers. These arguments predict that it is an asset to have well-off friends and neighbors. If full risk-sharing is attainable then the positive effect of an increase in the community's mean income will always dominate relative deprivation. With partial risk sharing the outcome is less clear. The empirical results in the paper suggest that perceived economic disparities relative to friends and neighbors are not a welfare-relevant concern for the poor in Malawi. Economically well-off people essentially care about relative position, side-by-side with a large number of poorer people (by far the majority) who appears to care far more about their absolute levels of living. The results suggest that for the bulk of the households in the Malawi, the utility-consistent poverty line would fall as the comparison-group income increases. However, on splitting the sample between urban and rural areas, our results are consistent with the idea that relative deprivation will be more important as the economy becomes more urbanized. The results could also give some insights for understanding the causes of rural under- development in Malawi. The dominance of the positive externality for the rural poor implies that there will be too little incentive to take actions that increase one's own income, given the spillover effect to others. The fact that signs of relative deprivation become the most important social effect amongst the upper economic stratum suggests the possibility that the incentive effect reverses amongst the relatively well-off. The distributional implications will depend on the extent to which economic gains to the rich spillover to benefit the poor. If the positive external effect from better-off friends and neighbors is concentrated amongst poor, middle-income and predominantly rural groups, while the relative deprivation effect is largely confined amongst urban elites, then there will also be too much inequality from the point of view of aggregate efficiency. 190 Perceived food adequacy Perceived housing adequacy 1 1 .8 .8 ds hole .6 .6 hous of oni .4 .4 oportrP .2 .2 0 0 0 20 40 60 80 100 0 20 40 60 80 100 Perceived clothing adequacy Perceived health adequacy 1 1 .8 .8 ds holes .6 .6 hou of onit .4 .4 or oprP .2 .2 0 0 0 20 40 60 80 100 0 20 40 60 80 100 Real per capita expenditure percentiles Real per capita expenditure percentiles Figure A3.1: Proportion of households who consider their expenditures inadequate by consumption categories. 191 15 3 (%) ution 10 distrib Own ranking ranking Ranking of friends 2 Ranking of neighbors expenditure Welfare of 5 Histogram 0 1 0 25000 50000 75000 100000 120000 Real per capita expenditure Figure A3.2: Economic rankings superimposed over a histogram of real per capita expenditure distribution 192 1 .8 poor .6 the ofn .4 Proportio MIQ poverty rate .2 CAQ poverty rate Absolute poverty rate 0 1 2 3 4 5 6 7 8 9 10 11 12 Household size Figure A3.3: Poverty rates derived based on subjective and absolute poverty lines by household size. 193 Table A3.6: Ordered probit of CAQ and OLS of MIQ Ordered probit of CAQ Ordered probit of SEW OLS of MIQ coef sd coef sd coef sd Log real PCE 0.384*** 0.023 0.557*** 0.022 0.406*** 0.022 Household demographics Household size 0.061** 0.021 0.193*** 0.018 0.222*** 0.018 Household size 2 -0.004** 0.002 -0.007*** 0.001 -0.008*** 0.001 Share of children -0.105 0.088 0.008 0.083 -0.229** 0.083 Share of elderly -0.118 0.091 -0.176* 0.092 -0.494*** 0.086 Share of female adults -0.026 0.079 0.135* 0.077 -0.080 0.076 Share of employed -0.051 0.061 -0.026 0.058 -0.039 0.058 Female household head -0.078* 0.033 -0.141*** 0.032 -0.159*** 0.031 Age of Household Head 0.002* 0.001 0.001 0.001 0.004*** 0.001 Agricultural household -0.027 0.045 0.004 0.042 -0.052 0.043 Owns an enterprise 0.145*** 0.025 0.108*** 0.024 0.019 0.024 Unemployed in the family -0.227* 0.100 -0.096 0.091 -0.042 0.096 Education of the head Reference category Pre-school -0.627*** 0.094 -1.195*** 0.090 -1.061*** 0.089 Junior Primary -0.536*** 0.093 -1.028*** 0.089 -1.022*** 0.088 Senior Primary -0.443*** 0.091 -0.817*** 0.086 -0.898*** 0.085 Junior Secondary -0.510*** 0.122 -0.717*** 0.115 -0.868*** 0.117 Senior Secondary -0.200* 0.090 -0.419*** 0.086 -0.680*** 0.085 University degree Reference category Lilongwe Reference category Karonga 0.971*** 0.060 0.152* 0.059 -0.666*** 0.064 Mzuzu 0.954*** 0.043 0.414*** 0.040 -0.551*** 0.043 Kasungu 0.170*** 0.043 0.205*** 0.041 -0.797*** 0.041 Salima 0.094 0.058 -0.408*** 0.059 0.206*** 0.049 Machinga 0.261*** 0.036 0.276*** 0.034 -0.428*** 0.037 Blantyre -0.059 0.036 -0.080* 0.034 0.040 0.032 Ngabu -0.508*** 0.060 -0.550*** 0.055 0.027 0.045 Major urban areas Reference category Boma/Large Urban -0.091 0.094 -0.183* 0.087 -0.045 0.087 Small Urban 0.202** 0.066 -0.037 0.062 -0.267*** 0.068 Rural 0.244*** 0.045 -0.141*** 0.041 -0.491*** 0.042 cut1 3.968*** 0.304 5.049*** 0.287 cut2 5.473*** 0.306 6.384*** 0.289 cut3 7.508*** 0.292 cut4 8.611*** 0.300 _cons 7.243*** 0.287 N 11,217 11,219 7,286 Adjusted/Pseudo R2 0.093 0.129 0.315 Log-Likelihood -8,987.28 -10,895.13 -9,760.23 194 195 196 *** ) ) ) ) ) ** ) ) ) ) ) *** ) ) *** ) *** ) ) South coef/sd -0.132 (0.020 0.087*** (0.024 0.140*** (0.028 0.050* (0.028 -0.010 (0.029 -0.069 (0.032 0.060** (0.030 0.165*** (0.035 0.512*** (0.036 0.040 (0.027 -0.272 (0.042 1.293*** (0.364 -0.097 (0.016 -0.053 (0.014 -0.022 (0.016 E *** ) ) ) ) ** ) *** ) ) ) ) ) *** ) ) *** ) *** ) ) TUR Central coef/sd -0.154 (0.022 0.077*** (0.025 0.057** (0.028 0.010 (0.030 -0.082 (0.033 -0.129 (0.032 0.057* (0.031 0.167*** (0.036 0.421*** (0.036 0.035 (0.029 -0.316 (0.015 1.676*** (0.107 -0.066 (0.013 -0.033 (0.012 -0.022 (0.015 ENDIPX *** ) ) ) ) ) ) ) ) ) ) *** ) ) *** ) ) ) E North coef/sd -0.149 (0.040 0.084** (0.043 0.061 (0.049 -0.006 (0.050 -0.006 (0.056 -0.005 (0.054 -0.072 (0.106 -0.108 (0.109 0.098 (0.110 0.110** (0.053 -0.224 (0.028 0.997*** (0.197 -0.088 (0.024 -0.019 (0.022 -0.019 (0.025 APITA C ) ) ) ) ) * ) ) ) ) ) *** ) ) *** ) * ) * ) ER an f/sd P 054 046 059 070 075 094 089 092 092 069 031 240 026 025 032 Urb coe -0.074 (0. 0.124*** (0. 0.142** (0. 0.019 (0. 0.010 (0. -0.177 (0. 0.053 (0. 0.134 (0. 0.667*** (0. -0.068 (0. -0.319 (0. 2.016*** (0. -0.143 (0. -0.049 (0. -0.063 (0. OGL *** ) ) ) ) ) ** ) ) ) ) ) *** ) ) *** ) *** ) ) LD 274 Rural coef/sd -0.147 (0.014 0.074*** (0.017 0.097*** (0.019 0.045** (0.020 -0.012 (0.021 -0.046 (0.022 0.077*** (0.023 0.183*** (0.026 0.390*** (0.026 0.053*** (0.019 -0. (0.027 1.308*** (0.220 -0.071 (0.010 -0.036 (0.009 -0.009 (0.010 OUSEHO *** ) ) ) ) * ) *** ) ) ) ) ) *** ) ) *** ) *** ) ) H OF 063*** .034 .007 S Rural coef/sd -0.140 (0.014 0. (0.017 0.081*** (0.019 0.022 (0.020 -0 (0.021 -0.066 (0.022 0.068*** (0.023 0.165*** (0.026 0.371*** (0.026 0.062*** (0.019 -0.285 (0.027 1.333*** (0.222 -0.064 (0.010 -0.032 (0.009 -0 (0.010 *** ) ) ) ) * ) *** ) NANT ** ) ) *** ) ) *** ) ) *** ) *** ) ** ) 128*** Malawi coef/sd -0.138 (0.014 0.073*** (0.016 0.092*** (0.019 0.027 (0.019 -0.036 (0.020 -0.075 (0.021 0.052 (0.022 0. (0.025 0.403 (0.025 0.055*** (0.019 -0.277 (0.026 1.369*** (0.215 -0.084 (0.010 -0.041 (0.009 -0.022 (0.010 ERMI ET D yr ON N prima sr sr sr sr sr imaryrp ary tede m yea yea yea yea pri ad (/100) EGRESSIO R eadh 5 5 5 5 yea he red some compl post 4-0 01-5 -1411 26-3 36-4 46-5 56-6 66+ old squa OLS: eholds ad: ad: ad: ad: ad: seh ildren ildren ildren size size ch ch ch 2E ouh he he he he he ucation: ucation: ucation: hou hh hh hh hh hh ed ed ed d of of of EX of of of of of est est est we holdes holdes er er er NN A Female Age Age Age Age Age High High High Wido Hou Hou Numb Numb Numb 197 ) ) ) ) *** ) ) ) ) ) ) *** ) ) *** ) ) ) ) South coef/sd 0.120*** (0.016 0.136*** (0.015 0.028 (0.033 0.082*** (0.017 -0.125 (0.026 0.046*** (0.007 -0.008 (0.017 0.011 (0.016 0.057 (0.035 -0.025 (0.022 -0.088 (0.024 -0.011 (0.026 -0.066 (0.023 -0.017 (0.021 -0.016 (0.031 0.068*** (0.018 ) ) ) ) ** ) ) ) ) ) ) ) ) ) *** ) ) ) Central coef/sd 0.142*** (0.017 0.125*** (0.017 0.076*** (0.021 0.059*** (0.015 -0.080 (0.032 0.130*** (0.013 -0.019 (0.017 0.080*** (0.019 0.088** (0.043 0.154*** (0.027 0.015 (0.022 0.091*** (0.024 0.066** (0.028 -0.086 (0.024 -0.040 (0.038 0.133*** (0.023 ) ) ) ) ** ) ) ) ) ) ) *** ) *** ) *** ) ) ) ) North coef/sd 0.180*** (0.034 0.224*** (0.029 0.085** (0.034 0.049* (0.027 -0.108 (0.047 0.081*** (0.023 -0.017 (0.031 0.089*** (0.029 0.243*** (0.069 -0.028 (0.057 -0.129 (0.047 -0.267 (0.051 -0.140 (0.052 -0.037 (0.042 0.330*** (0.100 0.026 (0.035 ) ) *** ) ) ) ) ) ) ) ) an f/sd 036 036 052 040 039 039 066 074 047 Urb coe 0.100*** (0. 0.106*** (0. -0.191 (0. 0.060 (0. 0.074*** (0.028 0.077** (0. -0.031 (0. 0.102 (0. -0.096 (0. 0.097** (0. ) ) ) ) *** ) ) ) ) ) *** ) ) ) ** ) ) ) Rural coef/sd 0.095*** (0.011 0.143*** (0.011 0.135*** (0.015 0.094*** (0.011 -0.081 (0.023 0.007 (0.012 0.056*** (0.011 0.178*** (0.025 -0.009 (0.017 -0.102 (0.016 -0.026 (0.017 -0.014 (0.017 -0.033 (0.014 -0.023 (0.022 -0.007 (0.013 ) ) ) *** ) ) ) ) ) ) ) *** ) ** ) ) *** ) ) ) 108*** .004 Rural coef/sd 0. (0.011 0.148*** (0.011 0.100*** (0.015 0.085 (0.011 -0.028 (0.024 0.080*** (0.006 0.007 (0.012 0.057*** (0.011 0.194*** (0.025 -0.018 (0.017 -0.108 (0.016 -0.039 (0.017 -0.027 (0.017 -0.039 (0.014 -0.022 (0.022 -0 (0.013 ) *** ) ) ) ) ) ) ) ) ) *** ) ** ) ** ) *** ) ) ) 014 077*** 012 154*** 095 037 Malawi coef/sd 0.122*** (0.011 0.140 (0.011 0.089*** (0.015 0.070*** (0.010 -0. (0.020 0. (0.006 0. (0.011 0.066*** (0.011 0. (0.025 -0.001 (0.017 -0. (0.015 -0. (0.016 -0.038 (0.017 -0.044 (0.014 -0.025 (0.022 0.010 (0.013 nos iser sea (0/1) s s s plots min0 min5 min0 s e enterp pping ed plots mmunity comin rm cro mmunity center >20-3 >30-4 >45-6 >60min co n-fa last plot no in rainfyna co rainfed in unity ingd ma: ma: ma: ma: the unity of ec a Tra bo bo bo bo in comm salary/e ash cco dimba comm rms res or unity fa servi inc rest rest rest rest rket m ni wag toba hecta nea nea nea nea ma com ash holdes werg anysnwo usb holdes clini Boma a to to to to in market total ular is HH Hou HH HH Hou Ln Reg Health EA Travel Travel Travel Travel ADMARC Bank Daily 198 lude inc ) Regressions 5,176 South coef/sd 0.086*** (0.023 0.4657 1%. at ) shown). 4,219 Central coef/sd 0.423*** (0.035 0.4982 (not *** ) significant rict 1,637 *** dist North coef/sd -0.195 (0.049 0.4056 5%; ent ) at an f/sd 044 1,431 Urb coe 0.441*** (0. 0.5078 developm ) ) ) significant 9,601 4212 ** Rural coef/sd 0.030* (0.018 0.002 (0.030 0.284*** (0.023 0. ricultural ag 10%; ) ) ) at and 9,601 Rural coef/sd 0.030* (0.018 -0.037 (0.030 0.249*** (0.023 0.4329 ) ) ** ) ) shown), 032 (not Malawi coef/sd 0.132*** (0.017 0.310*** (0.022 -0.061 (0.030 0.140*** (0.026 11, 0.4697 significant*. brackets ligionre in age, mmunity errors co head's in ns roadlta standard n servatio household sph on gio ob /ac re of Robust for regi er ared an Tarma Urb North Central Numb R-squ Notes: controls ANNEX 3A: SHOCKS REPORTED BY HOUSEHOLDS BETWEEN 1999-2004 IN IHS1 COMPLEMENTARY PANEL SURVEYS Type of shock All households Drought 66.2 Too much rain 31.5 Erosion 6.0 Frosts, hail 2.6 Pests, disease affecting crops 19.6 Pests, diseases affecting stored crops 10.8 Pests, disease affecting livestock 25.4 Lack of access to inputs 10.4 Large increase input prices 65.7 Large decrease in output prices 6.3 Lack of demand for ag. Products 4.7 Lack of demand for nonag. Products 2.2 Theft of crops 19.4 Theft of livestock 14.7 Death of husband 5.4 Death of wife 2.8 Death of another person 24.4 Illness of husband 11.0 Illness of wife 11.6 Illness of another person 17.7 Source: Sharma and Yohannes (2005) using 529 households in the Complementary Panel Survey (CPS) 5th round. 199 200 ) ** ) *** ) ) ) ) ) ) *** ) ) ) *** ) *** ) *** ) *** ) LDS South coef/sd 0.329*** (0.065 -0.172 (0.084 -0.273 (0.094 -0.152 (0.096 -0.026 (0.099 0.067 (0.105 -0.119 (0.086 0.562*** (0.079 -2.167 (0.575 0.090** (0.036 0.048 (0.036 -0.327 (0.090 -0.607 (0.111 -1.148 (0.117 -0.281 (0.055 ) ) ) ) ) ) ) ) *** ) ) ) ** ) *** ) *** ) *** ) OUSEHO H Central coef/sd 0.316*** (0.087 -0.063 (0.130 0.095 (0.136 0.090 (0.142 0.329** (0.144 0.303** (0.152 -0.076 (0.116 0.590*** (0.064 -2.805 (0.436 0.094** (0.040 0.111*** (0.042 -0.294 (0.138 -0.614 (0.161 -0.977 (0.169 -0.244 (0.074 OF SU ) ** ) * ) ) ) ) ) ) *** ) ) ) ) ) ** ) *** ) TATS North coef/sd 0.167 (0.133 -0.343 (0.171 -0.295 (0.176 -0.152 (0.186 -0.215 (0.194 -0.302 (0.219 -0.287 (0.191 0.564*** (0.087 -2.260 (0.538 0.095 (0.058 -0.015 (0.061 -0.592 (0.431 -0.523 (0.437 -1.045 (0.443 -0.428 (0.128 OOR ) ** ) ) ) ) ) ) ) *** ) ) ) ) ) ) ) -P an RA Urb coef/sd 0.335 (0.251 -0.741 (0.361 -0.359 (0.370 0.095 (0.392 -0.374 (0.443 0.796* (0.445 0.162 (0.298 1.186*** (0.216 -6.718 (1.527 0.101 (0.094 0.035 (0.096 4.345*** (0.727 4.201*** (0.727 3.223*** (0.757 -0.216 (0.168 LT U ) ** ) ** ) ) ) ) ** ) ) *** ) ) ) *** ) *** ) *** ) *** ) OF TS Rural coef/sd 0.283*** (0.050 -0.134 (0.067 -0.149 (0.072 -0.089 (0.074 0.054 (0.076 0.045 (0.081 -0.162 (0.067 0.539*** (0.060 -2.208 (0.424 0.098*** (0.026 0.065** (0.027 -0.335 (0.075 -0.534 (0.088 -0.958 (0.091 -0.265 (0.043 NAN ) ** ) ** ) ) ) ) ** ) ) *** ) ) ) *** ) *** ) *** ) *** ) MI ER Malawi coef/sd 0.290*** (0.048 -0.160 (0.065 -0.150 (0.070 -0.081 (0.072 0.043 (0.074 0.070 (0.079 -0.141 (0.065 0.539*** (0.062 -2.245 (0.436 0.097*** (0.024 0.066** (0.026 -0.314 (0.074 -0.507 (0.086 -0.979 (0.090 -0.259 (0.041 ET D ON yr NO SSI prima sr sr sr sr sr (/100) imaryrp ary tede m e EGRE yea yea yea yea ad R eadh 5 5 5 5 yea he red 7-0 41-8 pri comin some compl post 26-3 36-4 46-5 56-6 66+ old squa ROBITP: eholds ad: ad: ad: ad: ad: seh ildren ildren he he he he he size size hou ch ch salary/e 3B ouh ucation: ucation: ucation: hh hh hh hh hh d of of ed ed ed EX of of of of of we holdes holdes er er est est est wagsah NN A Female Age Age Age Age Age Wido Hou Hou Numb Numb High High High HH 201 *** ) ) *** ) ) *** ) ) *** ) ) ) ) ) ) ) ) *** ) ) South coef/sd -0.378 (0.050 0.013 (0.099 -0.291 (0.056 0.269*** (0.093 -0.123 (0.022 0.013 (0.056 -0.161 (0.055 -0.039 (0.112 0.003 (0.075 0.091 (0.076 0.007 (0.082 0.099 (0.073 0.167** (0.066 -0.084 (0.109 -0.158 (0.057 -0.043 (0.074 *** ) ) *** ) * ) *** ) ) *** ) *** ) * ) ) ** ) ) ) ) ) *** ) Central coef/sd -0.325 (0.072 -0.083 (0.085 -0.173 (0.064 -0.217 (0.128 -0.325 (0.046 0.159** (0.072 -0.358 (0.081 -0.536 (0.206 -0.175 (0.105 0.048 (0.088 -0.209 (0.102 -0.128 (0.107 0.242** (0.098 0.049 (0.159 -0.012 (0.089 -0.697 (0.141 *** ) ) ) ) ** ) ) ) *** ) ) ) ) ) ) *** ) ) ) North coef/sd -0.354 (0.099 -0.120 (0.117 -0.129 (0.093 0.377* (0.193 -0.160 (0.075 0.049 (0.108 -0.073 (0.094 -1.162 (0.403 -0.170 (0.180 0.098 (0.153 0.323** (0.159 -0.093 (0.170 -0.162 (0.146 -1.440 (0.527 0.098 (0.123 0.264 (0.177 *** ) ) ) ) ) ) ) ) *** ) *** ) an Urb coef/sd -0.655 (0.167 -0.028 (0.236 -0.284 (0.176 0.004 (0.060 0.277 (0.201 0.030 (0.202 0.213 (0.264 -0.039 (0.398 -0.725 (0.210 -0.632 (0.212 *** ) * ) *** ) ) *** ) ) *** ) *** ) ) ) ) ) ) ) ) ) Rural coef/sd -0.324 (0.039 -0.094 (0.056 -0.238 (0.039 -0.004 (0.089 -0.189 (0.018 0.032 (0.041 -0.197 (0.041 -0.327 (0.092 -0.003 (0.060 0.155*** (0.056 0.078 (0.059 0.024 (0.057 0.118** (0.050 -0.049 (0.089 0.026 (0.045 -0.093 (0.062 *** ) * ) *** ) ) *** ) ) *** ) *** ) ) ) ) ) ) ) ) ** ) Malawi coef/sd -0.338 (0.038 -0.097 (0.055 -0.231 (0.038 0.009 (0.075 -0.173 (0.018 0.029 (0.040 -0.201 (0.039 -0.291 (0.090 -0.008 (0.058 0.143*** (0.053 0.046 (0.057 0.039 (0.056 0.122** (0.049 -0.051 (0.085 -0.029 (0.044 -0.127 (0.059 nos iser sea (0/1) s s s plots min0 min5 min0 s enterp pping ed plots rm cro mmunity center mmunity mmunity >20-3 >30-4 >45-6 >60min co co n-fa last plot no in rainfyna co rainfed in unity ingd ma: ma: ma: ma: the unity in of ec a Tra bo bo bo bo in comm ash cco dimba comm rms res or fa servi inc rest rest rest rest rket munity ni roadlta toba hecta nea nea nea nea ma sph holdes werg anysnwo usb holdes clini Boma a to to to to com in market /ac total ular is Hou HH HH Hou Ln Reg Health EA Travel Travel Travel Travel ADMARC Bank Daily Tarma tn ** ** 202 ) 9 ]* ]* ]* ] 4 ] 8 ] ] (6 hut me So 070. 0.016[ .0290- .0170[ .0470- .0170[ 52.0 -0 .0190[ 000. .0210[ 020. .0250[ .0230- 0.017[ *** develop ) percent;5 (5 ralnteC ** * 7 ]* ] at 030. 0.012[ .0060- .0130[ 010.0 ] 0 ] 3 ]* 4 ]* ] .0150[ 010. .0160[ 040. .0220[ 040. .0240[ .0060- 0.011[ agricultural ficant *** ) and ) hrt * (4 8 ] ]* ]* ] ] ] ] signi No 5,176 20.9 n), 42.0 33.0 14.0 South coef/sd -2.606 (0.245 0.2589 020. ** 0.025[ .0520- .0230[ .0450- .0230[ -0 .0260[ -0 .0250[ -0 .0260[ .0380- 0.023[ show *** ) (not n ] ] ] ] ] ] ] 4,219 11.5 percent; ) ba 2 1 6 Central coef/sd -3.318 (0.253 0.2430 (3 10 ion Ur 000. 0.002[ .0040- .0030[ .0010- .0010[ 000. .0030[ 10.0 -0 .0010[ 010. .0190[ 100.0 *** ) att .0020[ relig 1,637 17.4 North coef/sd -3.063 (0.596 0.2585 age, ** * * ) ]* ]* ]* ] (2 6 ] 8 ] ]* an significan*. raluR head's 060. 0.011[ .0220- .0120[ .0250- .0120[ 31.0 -0 .0130[ 010. .0160[ 010. .0170[ .0270- .0110[ 1,431 5.0 Urb coef/sd -8.432 0.4085 hold ) *** ) brackets house (1 wiala ** * * * 3 ]* ]* ]* ] 2 ] 0 ] ]* in M 9,601 18.3 050. for 0.010[ 0.023- .0100[ .0210- .0100[ 1.0 -0 .0110[ 010. .0130[ 020. .0150[ .0200- .0090[ Rural coef/sd -2.426 (0.187 0.2394 errors *** ) *** ) 7 controls Malawi coef/sd -0.511 (0.083 -2.384 (0.182 11,032 16. 0.2541 andardst include Robust ssions s s effects. Regre d year5 year5 earsy ns d 6-32 -463 earsy55- earsy56- eadhdlo servatio Poor uare arginalm 46: 56: 66+: percent.1 shown). ob at of Ultra R-sq (not ablesiraVd headolh use ead:h ead:h useh hh hh hodew er o Probit hoela hol an sn use Fem ofegA ofegA headhhfo headhhfo headhhfo doi Age Age Age W Urb _co Numb Percent Pseud Notes: significant district Ho ** ** * ** ** ** ** ** ** ** ** ** 203 ) (6 hut 3 ]* ]* 8 ]* 9 ] ]* ]* ]* ]* ]* ]* So 120. 0.016[ .4750- 0.119[ 010. .0080[ 000. .0080[ .0680- .0190[ .0960- 0.013[ .1780- .0130[ -0.044 0.010[ .0560- 0.011[ .0700- 0.010[ 100.0 ] ]* 4 ]* ]* 2 ] .0210[ .0560- 0.010[ 050. .0160[ .0280- .0050[ 000. .0120[ .0360- ) (5 ralnteC ** ** * * * ** ** ** ** ** ** ** * 0 ]* ]* 9 ]* 1 ]* ]* ]* ]* ]* ]* ]* ] ]* ] ]* 6 ]* 060. 0.006[ .2780- 0.041[ 000. .0040[ 010. .0040[ .0300- .0140[ .0440- 0.008[ .0700- .0090[ -0.017 0.006[ .0210- 0.006[ .0280- 0.006[ .0070- .0080[ .0190- 0.006[ 32.0 -0 .0170[ .0330- .0050[ 010. .0070[ .0330- ) hrt ** ** * ** ** * * (4 3 ]* ]* 4 ] No 090. 0.013[ .3660- 0.083[ 010. .0100[ 20.0 ] ] ] ]* ] ]* ]* ] ]* 2 ]* ]* 8 ] -0 .0100[ .0850- .0530[ .0750- 0.053[ .1560- .0610[ -0.017 0.018[ .060- .0150[ .0540- 0.014[ .0190- .0170[ .0240- 0.014[ 050. .0210[ .0260- .0120[ 000. .0170[ .0100- ) n ** * ** ** (3 ba 6 ]* ]* 1 ] 0 ] Ur 000. 0.002[ .0340- .0130[ 000. .0010[ 000. .0000[ 996.0 ]* 0.227[ 937.0 ]* 0.225[ 450.0 ] ] ] ]* ] ] ] 1 ] 0.045[ -0.001 0.001[ .0010- 0.001[ .0030- 0.002[ .0010- 0.001[ 10.0 -0 .0010[ 000.0 0.000[ 000. .0010[ 000.0 ** ** ** * ** ** ** ** ** ** ** ** ) (2 raluR 6 ]* ]* 7 ]* 1 ]* ]* ]* ]* ]* ]* ]* ]* ]* 2 ] ]* 6 ] 100. 0.010[ .429 -0 .0760[ 010. .0050[ 010. .0050[ .0640- 0.015[ .0810- 0.010[ .1340- 0.009[ .037 -0 0.007[ .0460- 0.007[ .0540- 0.007[ .0170- .0100[ 046. -0 0.007[ 000. .0170[ .0380- 0.003[ 000. .0080[ .0370- ) (1 wiala ** ** ** * ** ** ** ** ** ** ** ** 1 ]* M 090. 0.009[ 37.3 ]* 5 ]* 0 ]* ]* ]* ]* ]* ]* ]* ]* ]* 2 ] ]* 5 ] -0 .0660[ 010. .0040[ 010. .0040[ 05.0 065 -0 0.012[ -0. 0.008[ .1230- 0.008[ -0.030 0.006[ .0390- 0.006[ 94.0 -0 0.005[ .0150- .0080[ 83.0 -0 0.006[ 000. .0120[ .0290- 0.003[ 000. .0070[ -0.032 ary seri yt )001(/d arym prim yra rpe seasong pin im ld ent unim 7 41 pr em crop sotpldfen last rai itynu uare 0-n 8-n priemos letedpm ehosuoh coiny stolpdefinar omcni in mm sq of ce zesid zesid drel drel on:i co:nio stop:nio in armf-nona rvi coni chi chi ess ccoa se ic of of ducate catu catu ed ed illn hasd tob hol hol otplabmdiynad anysm fard ectaresh variables usb ar clin ber ber use use esthg est est icn ro /salaregawsah hol ew hol gr ha talot Ho Ho Num Num Hi High ighH Ch HH use Ho HH use HH Ho Ln guleR Health Community ** * ** 204 ) (6 hut ]* ] ] 1 ] 5 ] ] 9 ]* So 0.011[ .0020- 0.023[ .0010- .0160[ 020. .0170[ 000. .0180[ 020. .0160[ 030. .0160[ 02.0 ] ]* ] .9 5 -0 .0210[ 43.0 -0 .0120[ .0110- .0150[ 20 215, ) (5 ralnteC ** ** * * ** ]* ]* ] 1 5 ] 0.007[ .0380- 0.009[ .0140- .0090[ 000. .0090[ 91.0 ]* ] 8 ]* 7 ] ] ]* .5 11 -0 .0080[ 21.0 -0 .0090[ 020. .0130[ 000. .0180[ 10.0 -0 .0090[ .0460- .0060[ 284, ) hrt ** ** ] ]* ] 0 (4 8 ] 0 ]* No 0.015[ .0950- .0120[ .0260- .0240[ 010. .0260[ 060. .0330[ 41.0 ] ] ]* 6 ] ] .4 17 -0 .0260[ 32.0 -0 .0200[ 09.0 -0 .0110[ 010. .0210[ 050.0 .0380[ 651, ) n 3 (3 ba ] 1 ] 0 ] Ur 0.001[ 000. .0020[ 000. .0020[ 70.0 ]* ]* 05. -0 .0040[ .0030- .0010[ 431, ** ** ** * ) (2 raluR ]* ]* ] 3 2 ]* 7 ] 5 ] 4 ]* 0.007[ .0510- 0.012[ .0010- .0120[ 030. .0120[ 010. .0120[ 000. .0110[ 020. .0110[ 90.0 ] 5 ] ]* .3 -0 .0160[ 000. 18 .0090[ .0190- .0110[ 719, ) (1 wiala ** ** ** * * ** ]* ]* ] 5 ]* 9 ] 6 ] 1 ]* 64,1 M 0.006[ .0390- 0.010[ .0010- .0090[ 020. .0100[ 000. .0100[ 000. .0090[ 020. .0090[ 80.0 ] ] ]* ]* .7 16 -0 .0130[ 50.0 -0 .0070[ -0.022 .0080[ 66.0 11 -0 0.008[ snim yt ityn center 30-0 nsim5 nsim0 sni unim mmu ing ad >2:am Trroam bo est >30-4:amobts >45-6:amobts >60m:amobts com yt hetni yt rketam unim unimmocni coni ad Boa earnot roltah neareot neareot neareot C rooP AR rat s is el omcni ac/asp n oni av EA Tr Travel Travel Travel ADM nkaB rketamyl ba Dai Tarm Ur Ultn vatr Perce Obse Table 3A.3: Probit Regression on Determinants of Ultra-Poor Status of Children All Age 0-5 Age 6-11 Age 12-17 coef/sd coef/sd coef/sd coef/sd Child Variables Age 0.037*** (0.006) Age squared -0.002*** (0.000) Female 0.003 -0.014 -0.005 0.045 (0.017) (0.028) (0.029) (0.034) Orphan living with one parent -0.042 0.177* -0.031 -0.148** (0.043) (0.099) (0.069) (0.069) Double or 'virtual' double orphan -0.060* 0.150 0.034 -0.214*** (0.036) (0.112) (0.056) (0.055) Non-orphan living with one parent 0.092*** 0.171*** -0.001 0.053 (0.031) (0.050) (0.052) (0.062) Non-orphan living away from parents -0.236*** -0.092 -0.215*** -0.405*** (0.034) (0.070) (0.052) (0.062) Household Variables PA death past two years in household 0.051 -0.047 0.030 0.105 (0.041) (0.078) (0.068) (0.071) Female head -0.041 0.032 -0.041 -0.128** (0.029) (0.051) (0.047) (0.052) Head has some primary education -0.348*** -0.309*** -0.375*** -0.356*** (0.020) (0.033) (0.033) (0.039) Head has some post primary education -1.025*** -0.942*** -1.037*** -1.118*** (0.034) (0.053) (0.060) (0.071) Urban -0.527*** -0.610*** -0.535*** -0.406*** (0.035) (0.059) (0.060) (0.063) Central -0.535*** -0.555*** -0.507*** -0.547*** (0.019) (0.031) (0.032) (0.038) North -0.031 -0.017 -0.066 0.003 (0.025) (0.040) (0.042) (0.049) _cons -0.165*** -0.176*** -0.020 -0.002 (0.029) (0.045) (0.047) (0.054) Number of observations 26,610 10,497 9,239 6,874 Pseudo R-squared 0.082 0.085 0.076 0.094 Notes: Probit marginal effects. Robust standard errors in brackets. * significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent. Virtual double orphans are those with one parent deceased and not living with surviving parent. 205 ANNEX 3C: CHARACTERISTICS OF SHOCKS AND ASSOCIATION BETWEEN SPECIFIC HOUSEHOLD CHARACTERISTICS AND LIKELIHOOD OF EXPERIENCING A SHOCK Table 1: Shocks in the past 5 years as reported by households Expenditure Quintile Poorest Richest Type of shock 20 20 All percent 2 3 4 percent households Lower crop yields due to drought or floods 71.4 68.1 67.1 62.5 50.8 62.5 Crop disease or crop pests 20.9 24.7 26.2 26.4 20.9 23.7 Livestock died or were stolen 29.9 35.2 35.5 36.1 30.1 33.3 Household business failure, non-agricultural 13.5 19.0 21.1 24.8 26.7 21.9 Loss of salaried employment or non-payment of salary 4.7 8.0 8.2 9.4 10.6 8.5 End of regular assistance, aid, or remittances from outside household 5.8 6.9 6.8 7.7 7.8 7.2 Large fall in sale prices for crops 33.7 37.8 40.5 41.3 36.1 38.0 Large rise in price of food 74.3 75.5 79.7 78.6 76.1 77.0 Illness or accident of household member 41.8 46.3 47.1 49.0 43.7 45.7 Birth in the household 11.1 12.9 13.0 11.8 7.7 11.0 Death of HH head 4.8 5.0 4.7 4.6 4.8 4.8 Death of working member of household 7.9 9.9 9.6 8.4 8.0 8.7 Death of other family member 35.5 39.3 43.4 41.8 41.1 40.6 Break-up of the household 10.1 10.2 9.9 8.9 10.3 9.9 Theft 13.2 17.6 17.8 21.7 22.9 19.3 Source: IHS2 206 Table 2: Covariance of shocks by type of shock (percent of households reporting) Some Most Type of shock Own hh other other All other only HHs HHs HHS Total Lower crop yields due to drought or floods 2.8 18.2 46.7 32.3 100 Crop disease or crop pests 7.7 35.2 37.3 19.9 100 Livestock died or were stolen 54.6 30.5 13.4 1.5 100 Household business failure, non-agricultural 80.0 10.2 4.7 5.1 100 Loss of salaried employment or non-payment of salary 75.2 18.9 3.8 2.2 100 Large fall in sale prices for crops 3.9 16.4 42.2 37.5 100 Large rise in price of food 3.2 9.1 41.4 46.3 100 Illness or accident of household member 78.0 19.4 1.3 1.3 100 Death of working member of household 37.6 60.6 1.4 0.4 100 Death of other family member 31.4 65.2 2.9 0.5 100 Break-up of the household 89.7 8.9 1.2 0.1 100 Theft 77.7 17.8 3.4 1.1 100 Note: Among the sample of top three shocks reported by households (excluding those shocks less than 2 percent of the top three shocks). Source: IHS2 207 Table 3: Sources of income earnings (percent of households reporting) Urban households Poorest Richest Source 20 20 percent 2 3 4 percent All Crop sales (non-tobacco) 9.5 21.5 21.6 14.0 9.9 13.2 Tobacco sales 8.4 9.4 4.9 4.4 1.1 3.3 Livestock sales 7.7 9.0 5.6 6.2 3.5 4.9 Tree crop sales 7.9 5.6 3.6 2.5 0.6 2.1 Enterprise income 11.6 39.4 42.8 41.4 40.0 39.3 Retail/wholesale 9.5 27.5 28.8 32.1 24.8 26.4 Manufacturing 2.1 9.7 5.9 3.4 5.2 5.1 Other enterprise 0.0 3.0 5.4 3.0 7.8 5.8 Wage/salary income 41.7 40.2 50.9 51.6 67.4 58.7 Ganyu income 61.5 50.5 35.0 29.1 17.3 26.7 Rural households Poorest Richest Source 20 20 percent 2 3 4 percent All Crop sales (non-tobacco) 41.7 51.3 55.6 58.8 55.8 53.3 Tobacco sales 11.6 12.6 14.4 18.4 19.5 15.7 Livestock sales 25.1 29.8 33.2 33.9 28.5 30.3 Tree crop sales 17.3 18.1 18.5 19.4 18.2 18.4 Enterprise income 27.3 30.4 32.2 37.5 37.8 33.5 Retail/wholesale 10.2 13.0 13.0 18.5 18.8 15.1 Manufacturing 9.6 9.6 10.9 9.8 8.7 9.7 Other enterprise 5.2 5.2 7.3 7.7 8.1 6.8 Wage/salary income 21.0 25.1 23.2 25.4 30.4 25.3 Ganyu income 67.5 59.3 55.3 47.0 37.5 52.0 Note: Income source for crops and livestock refers to some sales of output in the last season. Sample includes non-agricultural households (landless). Agricultural households are not restricted by definition to rural areas. Source: IHS2 208 Table 4: Sources of income earnings (percent of households reporting) Rural Rural Urban landed landless households households households Source Crop sales (non-tobacco) 13.2 56.0 0.0 Tobacco sales 3.3 16.4 0.0 Livestock sales 4.9 31.4 8.5 Tree crop sales 2.1 19.1 3.1 Enterprise income 39.3 33.3 38.7 Retail/wholesale 26.4 14.8 20.3 Manufacturing 5.1 9.8 7.2 Other enterprise 5.8 6.7 10.2 Wage/salary income 58.7 24.1 50.2 Ganyu income 26.7 53.2 28.6 Number of Sources 0 4.4 5.1 9.3 1 58.0 22.0 61.1 2 28.1 31.5 22.4 3 7.0 24.9 6.4 4-7 2.6 16.6 0.7 Note: Income source for crops and livestock refers to some sales of output in the last season. Source: IHS2 209 210 Total 100 100 100 100 100 100 therO 1.9 0.6 0.9 1.7 1.1 1.3 Nothing 14.6 52.5 27.8 23.8 54.6 29.9 Pray 0.7 0.1 0.2 11.9 13.8 4.0 less Consume 15.3 3.4 28.5 0.7 1.0 14.2 households by ance Response Receive 4.4 0.2 0.9 assist 12.2 0.5 3.4 portedre reporting) S2 1.8 1.3 1.4 8.0 4.1 2.8 Borrow IH: shocks joram households Source of Labor 35.6 18.2 21.8 7.5 7.2 21.1 ocks. toe s sh (percent Sold asset 13.2 17.5 8.0 11.4 6.3 10.7 major respons evalent Spend cash 6.3 pr First savings 12.6 10.5 22.8 11.4 12.6 5: most5 sdo the Table flo or erbmem ong am oughtrd crops to for food household mberem householdsyb due prices of of ily fam reported k yields sale price dent in in nse acci Shoc other crop of fall rise or of respo First Type Lower Large Large Illness Death Total Note: ) 9 *** *** * *** ** ** *** *** ** *** (9 Theft 060. 8].3 8 2 2 9 3 [4 190. 7].2 [9 040. 7].8 [1 030. 2].5 [1 030. 0].3 [1 040. 7].2 [4 .1280- 3].3 42.0 6].4 1.0 6].1 [2 -0 [2 -0 [1 400.0 7].3 [0 990.0 3].4 [8 040. 2].7 [3 020. 1].3 [2 030. 9].7 [2 211 ) *** *** *** *** *** *** *** * (8 Death 1.0 9].7 7 1 2 5 3 5 8 7 -0 [0 010. 2].0 [1 060. 1].3 [3 070. 8].0 [3 060. 1].8 [2 030. 6].7 [3 .1450- 5].8 [2 000. 3].7 [0 000. 4].4 [0 .0060- 6].6 [0 540.0 6].7 [5 060. 5].3 [7 000. 5].8 [0 010. 9].7 [1 * ) (7 /injury 3 *** 6 ** 9 1 9 *** *** ** ** ** ess 050. 6].3 1.0 6].4 5 5 3 8 8 [3 -0 [0 040. 2].1 [2 030. 1].5 [1 010. 3].4 [0 020. 3].1 [3 .1680- 2].0 [3 010. 0].5 [1 000. 3].5 [0 .0050- 2].4 [0 641.0 4]73. 020. 2].0 [2 020. 5].1 [2 000. 0].7 Illn [1 [0 ) ** *** *** ** * *** * (6 easer icerp 8 8 3 6 4 6 6 Inc 010. 5].5 91.0 6].2 4].5 5].1 8].4 5].9 8].5 9].5 20.0 8].2 7].8 0].3 4].3 1].7 8].1 food [1 -0 [1 030. [2 050. [3 020. [1 020. [2 .1020- [2 000. [0 -0 [0 .0150- [1 030.0 [0 030. [4 010. [1 010. [1 in ) ** *** *** * * *** *** (5 price 15.0 *** 2].1 8 4 3 2 -0 [3 000. 5].3 80.0 5].3 41.0 5].5 95.0 7].2 [0 -0 [0 -0 [0 -0 [2 040. 0].1 5].2 [4 [3 010. 4].3 1.0 3].1 0].8 [1 -0 [1 [1 220.0 2].8 [1 070. 9].0 6.0 7].5 1.0 9].7 Decrease crop .2230- .020- [6 -0 [4 -0 [0 ) 3 4 *** 3 *** 4 *** 7 ** 9 *** 3 *** 3 9 (4 Business failure 010. 1].0 22.0 1].3 [1 -0 [1 100. 2].0 [5 130. 1].1 [5 140. 7].6 [5 010. 1].2 [2 .0660- 7].4 [1 000. 4].1 50.0 6].7 [1 -0 [0 .0040- 4].4 [0 70.0 6].1 [7 040. 7].7 [4 010. 4].3 [1 170. ***]118. [1 k ) oc 3 *** *** *** *** *** * *** *** ** (3 vest loss 000. 1].2 60.0 0].3 7 100. 9].8 9 110. 4].5 3 120. 9].7 040. 2].5 7 7 010. 000. 9].0 4 050. 1].9 1 010. 1].9 9 2].5 Li [0 -0 [0 [4 [4 [4 [4 .2410- 4].5 8].7 2].8 10.0 9].0 [4 [1 [0 -0 [0 470.0 [4 [4 [0 020. [2 households Rural ) (2 seaseid 3 ** *** * *** * *** *** ** op 000. 0].2 3 4 1 9 5 5 [0 040. 5].3 [2 050. 3].8 [2 040. 8].7 90.0 0].4 [1 -0 [0 020. 8].6 [3 .0680- 5].4 41.0 2].7 40.0 9].4 21.0 3].3 [1 -0 [1 -0 [0 -0 [1 310.0 5].0 [3 040. 7].6 71.0 2].5 [4 -0 [1 020. 6].4 [2 years,5 Cr t past ) ** 3 1 7 7 *** *** ** *** *** (1 the ughorD oodlfro 1 030. 9].1 [2 000. 4].1 [0 030. 3].6 [1 010. 8].7 33.0 5].4 [0 -0 [1 020. 2].1 [3 .1340- 3].6 50.0 3].5 30.0 5].3 2.0 5].0 [2 -0 [0 -0 [0 -0 [2 290.0 2].7 70.0 9].6 13.0 1].6 30.0 7].2 [2 -0 [0 -0 [2 -0 [0 in seri shocks arym ) rpe of e ent d prie ld headd oms letedpm arymrip st 4 01 ehos po ouh comni farm Correlates headolh ehols on:i co:nio /100(dera 0-n 5-n 14-11n uretdin on:i squ in aryals n-ona 6: ablesiraVd use ouhd ducate catu chi chi chi ess ge/ ed ducate zesid drel drel drel expeatpi hasd hoela of of of illn wa est hol sizedlo car hol Table hol owedi hestg arym hestg use use Fem W Hi High pri Hi Ho ehsuoH ber erb erb icn ro Peg has use Num Num Num Ch Lo HH Ho Ho ) eft 8 *** *** *** *** * ** *** *** *** *** *** (9 Th 040. 6].9 7 5 8 5 8 [2 080. 2].5 [7 060. 5].8 3.0 9].6 [2 -0 [4 420.0 5].9 92.0 2].3 [1 -0 [2 030. 8].4 [1 .0940- 6].1 [5 .0730- 2].3 68.0 9].7 [4 -0 [4 .0720- 6].0 [4 000. 4].3 [0 020. 8].7 [0 040. 9].3 [3 .0810- 212 ) 7 *** *** ** *** *** * *** *** (8 Death 010. 8].3 1 7 2 6 3 2 [1 050. 2].5 [5 000. 9].3 10.0 2].2 [0 -0 [0 .0530- 5].5 [5 020. 5].2 [2 000. 9].2 [0 .0480- 7].5 [3 .0380- 3].0 62.0 2].9 [3 -0 [1 .0790- 2].1 90.0 7].7 [6 -0 [0 000. 8].1 [0 060. 5].2 [5 800.0 ryu ) 2 *** *** *** ** *** *** *** *** (7 ess/inj 070. 9].2 4 1 4 4 6 [4 040. 8].6 [3 070. 5].0 40.0 0].6 [3 -0 [0 .030- 2].4 2.0 9].5 [2 -0 [1 020. 4].9 [0 .0850- 2].6 [4 .0650- 6].7 74.0 9].5 [3 -0 [2 .1010- 1].6 [5 000. 8].2 [0 040. 8].4 [1 010. 3].1 [1 .0110- Illn ) *** *** * *** *** *** *** *** *** (6 easer icerp 7 7 8 7 9 8 Inc oodf 050. 9].7 [4 020. 9].0 [3 020. 6].6 70.0 4].3 [1 -0 [1 .0060- 4].6 [0 030. 2].0 77.0 5].7 [4 -0 [3 .0470- 8].1 [3 .0870- 6].2 8.0 5].3 [6 -0 [5 .0160- 2].1 31.0 8].0 [1 -0 [1 020. 4].5 [1 030. 1].8 [3 .0670- in ) *** *** *** *** (5 price 320. ***]158. * ** 150. 7]62. 5 190. 6].2 5 2 3 2 [8 060. 7].1 6].8 [9 [2 020. 4].5 73.0 8].3 [1 -0 [1 620.0 5].3 5].9 [1 [0 020. 9].1 2].4 [1 [1 000. 6].1 [0 000. 8].0 65.0 1].8 Decrease crop [1 [1 .0360- .0170- .0260- [0 -0 [3 .0970- ) 1 3 7 ** 3 *** * * *** ** *** 5 2 *** (4 Business failure 000. 5].0 [0 000. 8].2 [0 020. 8].5 [1 010. 1].0 [2 .0150- 7].5 20.0 7].1 [1 -0 [0 070. 8].4 [3 .0270- 4].9 [1 .0230- 6].7 14.0 9].9 [1 -0 [2 .0350- 2].5 13.0 1].6 [2 -0 [2 010. 5].7 [0 070. 5].9 [5 900.0 k ) oc 8 * * *** *** ** ** ** ** *** ** ** *** (3 vest loss 020. 7].7 9 010. 5].6 7 180. 2].7 8 030. 2].2 .027 5].2 20.0 6].1 45.0 2].2 .039 1].2 2 8 6 6].6 .064 Li [1 [1 [8 [6 -0 [2 -0 [0 -0 [2 -0 [2 .0380- 7].3 010. 8].6 [2 [0 045.0- 6].6 030. 2].5 050. 1].0 [2 [2 [2 060. [4 -0 ) *** *** *** * *** *** *** *** *** (2 disease 85.0 *** 1].4 5 5 7 9 2 9 -0 [4 060. 5].5 [6 130. 3].8 [6 010. 7].7 [2 010. 7].5 [1 010. 8].7 69.0 0].6 [1 -0 [4 000. 6].1 [0 .0830- 6].0 91.0 9].2 [6 -0 [1 .0950- 9].6 11.0 1].8 [6 -0 [0 080. 8].7 53.0 6].8 72.0 [3 -0 [2 -0 Crop t ) 4 1 *** * 9 ** 3 *** *** *** *** *** *** *** *** (1 ughorD oodlfro 010. 9].9 6 8 6 6 [0 040. 2].9 [3 330. 8]15. 020. 7].7 41.0 9].2 9].9 41.1 8].6 5.0 4].9 4].3 5.0 8].9 7].5 4].1 8].5 4].4 18.0 [1 [3 -0 [1 050. [4 -0 [4 -0 [2 .0690- [4 -0 [2 .0610- [3 050. [4 030. [1 000. [0 -0 ngippocr sotpld yt nfe stalnioc otpl rai sotpldfen yt snim unimmocni ertn nsim0 45-0 nsim0 unim yt cegnid sinm0 >3:am >6:am com yt anysm hetni unim munitymocnid bacot bamdi raifo ce Tra bo bo yt fard esar rvi >20-3:amobts est >45-6:amobts est rketam unim omcni roa ewrg nyad hol ha use hectlatot variables sesub ar unimmocniciniclht neareot earnot neareot earnot C el el AR omcni ac/asphalt HH season HH Ho Ln guleR oramoBasi av av Heal EA Travel Tr Travel Tr ADM nkaB rketamyl Dai Tarm Community ) eft *** and (9 Th 8].2 7129 [4 213 religion, e, ag ) (8 Death 5].5 7129 [0 head'sdl eho ryu ) housr (7 ess/inj 7].5 7109 fo [0 rolst Illn con ) *** (6 easer icerp 2].5 7129 include Inc oodf [4 essions in ) *** (5 price Regrt. 8].8 7129 en Decrease crop [4 perc1ta ) cant (4 Business failure 1].6 7129 [0 gnifiis k ***; ) oc *** (3 vest loss 7].4 cent 7129 Li [3 per5 att can ) * (2 disease 6].6 7129 gnifiis [1 ** Crop t; t ) *** enrcep (1 ughorD oodlfro 7].5 10 7129 at s.t [4 cant effecl gnifiis*.st rginaam ackerb in Probit.cti ticsisattszfo distr ent lopm ns lue va veedl atio servbO luteo ltura Abs agricu 214 ) 9 2 *** 6 *** *** 7 *** 1 (6 Theft 020. 6].6 [0 290. 9].9 [4 .0510- 0].6 24.0 8].4 [0 -0 [0 .0780- 8].8 [0 080. 0].5 1.6 8].1 71.0 8].7 81.0 9].8 [3 -0 [3 -0 [0 -0 [0 .0150- 8].5 [0 930.0 7].2 [1 060. 6].0 51.0 1].5 [3 -0 [0 010. 6].3 [0 ) 8 5 7 6 * ** 5 *** 4 6 (5 Death 000. 4].3 [0 030. 4].1 [1 789.0 1].0 [0 980. 1].0 [0 17.0 1].0 [0 010. 5].4 96.0 3].8 40.0 1].4 20.0 4].2 [1 -0 [0 -0 [0 -0 [0 .0230- 9].8 [1 930.0 5].4 [2 030. 3].3 [3 010. 5].9 [0 000. 6].4 [0 ) (4 31.0 *** ** * * *** ** 1].3 4 170. 3].0 3 9 040. 000. 000. 7].4 5].4 4].4 90.0 2].4 31.0 7].4 95.0 9].1 households -0 [0 [3 .1580- 7].1 32.1 2].6 [2 -0 [1 .1660- 4].9 2].8 8.1 2].1 3].1 [1 [1 -0 [1 [0 [0 .0110- [0 761.0 [5 -0 [0 -0 [0 -0 [2 Illness/injury Urban food ) 1 3 6 6 *** ** 9 (3 years,5 easer price 040. 6].7 24.0 1].6 [0 -0 [0 .1840- 0].5 61.1 2].9 [1 -0 [0 .1620- 3].3 21.0 4].4 [1 -0 [0 070. 4].3 [0 030. 4].3 [1 000. 3].2 [0 .0120- 9].3 [0 301.0 7].6 56.0 9].4 30.0 8].0 [2 -0 [2 -0 [0 010. 5].5 [0 Inc past the ) in (2 71.0 * * *** *** 8].4 7 010. 8].3 2 1 7 000. 030. 000. 9].3 4].7 9].1 8].0 010. 0].6 10.1 3].3 7 6].4 Business failure -0 [0 [0 .0140- 0].2 2].0 [0 [0 .0520- 9].6 0].7 95.1 4].1 72.0 [0 [1 -0 [1 [0 -0 [1 400.0 [0 -0.002 [0 [0 -0 [4 120. [5 shocks of ) rot 8 *** (1 ughorD odofl 030. 8].3 51.0 3].5 74.0 0].2 94.0 ** 4].2 1 6 1 8 [1 -0 [0 -0 [1 -0 [1 .120- 2].2 [2 020. 4].5 93.1 3].4 [1 -0 [1 000. 1].5 11.0 4].9 [0 -0 [0 .0090- 6].6 [0 580.0 3].3 11.0 8].8 [4 -0 [0 010. 7].6 [0 000. 9].4 [0 Correlates 7: arym rpe Table im ld e ent pr 4 d -1 headolh headd priemos arymripdetelpmoc seri yra ehols on:i on:i stop:nio )001(/d 4 01 uare 0-n 5-n 11 ehosuoh reutidn comni sq in farm-nona ablesiravd use ouhd ducate ducate catu ed zesid zesid drel drel enrdil chi chi ess expeatpi aryals ge/ hasd hol hoela of of chfo illn car wa hol owedi esthg esthg est hol hol ber erb has use Fem W Hi Hi ighH use use Ho Ho Num Num erbmuN icn ro Peg use Ch Lo HH Ho Ho 215 ) eft (6 74.0 * * *** ** 5].0 2 7 5 6 3 3 Th -0 [1 060. 5].9 [1 010. 4].8 [0 010. 5].4 20.0 6].0 87.0 2].7 [0 -0 [0 -0 [1 260. 4].6 64.0 7].2 [4 -0 [1 080. 1].2 [2 431, for rolst con ) 1 8 1 2 ** ** 3 (5 Death 010. 0].5 [0 020. 1].3 40.0 5].4 [1 -0 [0 000. 2].5 [0 000. 6].0 [0 050. 4].0 40.0 7].1 63.0 8].0 22.0 7].2 [2 -0 [0 -0 [2 -0 [1 431, include ressions ryu ) (4 12.0 * ** 2].5 70.0 4].2 21.0 4].7 7 8 1 5 6 3 ess/inj -0 [0 -0 [0 -0 [0 030. 7].1 [1 010. 5].6 [0 080. 5].8 [1 110. 6].1 81.0 6].5 [2 -0 [0 030. 4].0 [1 431, Reg.1ta Illn cant gnifiis food ) 8 * * ** * (3 easer price 040. 4].8 2 0 7 5 4 3 [0 070. 2].9 [1 000. 2].0 11.0 7].2 85.0 0].7 [0 -0 [0 -0 [1 100. 8].0 [2 110. 4].8 65.0 7].3 [1 -0 [1 060. 3].5 [1 431, ***; cent Inc per5 effects. att ) (2 72.0 2 * ** 9 ** 2].8 33.0 8].3 12.0 4].5 030. 0].2 73.0 7].6 5 000. 3].1 4 110. 1].5 1.0 6].3 060. 7].3 3 can marginal Business failure -0 [0 -0 [1 -0 [1 [1 -0 [1 [0 [2 -0 [0 [2 431, gnifiis ** Probit t; ct.ir ) rot 8 *** dist (1 ughorD odofl 000. 9].3 9 0 1 3 6 3 enrcep [0 190. 6].4 [9 000. 3].0 22.0 9].0 32.0 7].3 63.0 7].3 [0 -0 [1 -0 [1 -0 [1 000. 3].0 [0 010. 8].6 [0 010. 9].6 [0 431, 10 at cant velopmentedla sotpldfen yt yt otpl rai sotpldfen yt unim gnifiis*.st unimmocni agricultur yt com ackerb hetni in and, bamdi anysm raifo ce yt unitymmoc unimmocnid esar rvi rketam unim in roat nyad fard ticsisattszfo religion age,s hol ha use hectlatot Variables sesub ar unimmocniciniclht C AR omcni rketam asphal s ac/ oni lue va head' HH Ho Ln guleR Heal ADM nkaB vatr Daily Tarm luteo Community Obse Abs household ANNEX 3D: TEMPORARY WITHDRAW FROM SCHOOL, STUDENTS 10-15 (1) (2) (3) With rainfall variation With rainfall variation and wealth interaction Rainfall in 2004 as percent of long-run average -0.226 -0.225 [4.26]*** [4.23]*** Rainfall in 2004 as percent of long-run average * good flooring -0.008 [0.50] Household Variables Female household head -0.006 -0.007 -0.008 [0.51] [0.61] [0.65] Widowed household head 0.042 0.041 0.041 [2.42]** [2.38]** [2.40]** Highest education: completed primary -0.013 -0.012 -0.011 [1.14] [1.05] [1.01] Highest education: post primary -0.042 -0.041 -0.039 [3.87]*** [3.74]*** [3.50]*** Household size 0.004 0.003 0.003 [0.52] [0.41] [0.40] Household size squared (/100) -0.041 -0.034 -0.033 [0.85] [0.72] [0.71] Number of children 0-4 0.015 0.015 0.015 [2.23]** [2.32]** [2.32]** Number of children 5-10 -0.007 -0.006 -0.006 [1.14] [1.12] [1.11] Number of children 11-14 0.015 0.014 0.014 [2.17]** [2.04]** [2.06]** Chronic illness in household 0.017 0.015 0.015 [1.95]* [1.73]* [1.73]* HH has wage/salary income 0.018 0.016 0.017 [1.83]* [1.68]* [1.72]* Household has a non-farm enterprise -0.001 -0.002 -0.002 [0.14] [0.19] [0.18] HH grew tobacco in last cropping season -0.016 -0.014 -0.014 [1.34] [1.11] [1.11] HH had any dimba plot 0.01 0.01 0.01 [1.12] [1.08] [1.05] Household farms any rainfed plots 0.014 0.012 0.012 [0.77] [0.69] [0.66] Ln total hectares of rainfed plots -0.003 -0.002 -0.002 [0.51] [0.40] [0.38] Community Variables Regular bus service in community -0.047 -0.044 -0.044 [4.82]*** [4.49]*** [4.48]*** Health clinic in community 0.018 0.017 0.017 [1.82]* [1.69]* [1.71]* EA is a Boma or Trading center 0.032 0.032 0.034 [1.49] [1.51] [1.55] 216 (1) (2) (3) With rainfall variation With rainfall variation and wealth interaction Travel to nearest boma: >20-30mins 0.003 0.007 0.007 [0.22] [0.46] [0.47] Travel to nearest boma: >30-45mins -0.013 -0.011 -0.011 [0.97] [0.85] [0.85] Travel to nearest boma: >45-60mins -0.003 0.000 0.000 [0.22] [0.00] [0.01] Travel to nearest boma: >60mins 0.005 0.006 0.006 [0.35] [0.43] [0.42] ADMARC market in the community 0.032 0.032 0.031 [2.51]** [2.50]** [2.49]** Bank in community -0.042 -0.044 -0.044 [2.60]*** [2.75]*** [2.75]*** Daily market in community -0.003 -0.006 -0.006 [0.24] [0.59] [0.56] Tarmac/asphalt road in community 0.041 0.042 0.042 [2.58]*** [2.65]*** [2.66]*** Urban -0.048 -0.041 -0.04 [3.04]*** [2.59]*** [2.46]** Observations 4,408 4,408 4,408 Absolute value of z statistics in brackets. * significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent. Regressions include controls for household head's age and religion. Probit marginal effects. 217 ANNEX 4A: DETAILS OF THE ESTIMATION OF MALNUTRITION INDICATORS AND COMPARISON WITH 1998 IHS1 This Annex details the steps used in estimating the malnutrition figures for the IHS-257. One consideration was comparability with IHS-1 and DHS. The other was international comparability based on methodological recommendations from CDC/WHO. Estimations were calculated using the 1978 NCHS/CDC/WHO reference curve. This is the same curve used for the IHS1 and the DHS. In terms of outliers for the calculation of z-scores, CDC/WHO recommends the use of (-6, +6) SD for HAZ and WAZ and (-4; +6) for WHZ. This is in contrast with the default thresholds of (-5; +5) used by Stata. In addition, two additional criteria for "flagging" a record are recommended (for an definition of the flags, see the end of this Annex). Specifically, the following cross-check is proposed: (HAZ > 3.09 and WHZ < -3.09) or (HAZ < -3.09 and WHZ > 3.09). Based on these flagging criteria, a flagging system is created in EpiInfo: Flag code HAZ WHZ WAZ Notes 0 No indices flagged 1 Y Only HAZ flagged 2 Y Only WHZ flagged 3 Y Y Both HAZ and WHZ flagged 4 Y Only WAZ flagged 5 Y Y Both HAZ and WAZ flagged 6 Y Y Both WHZ and WAZ flagged 7 Y Y Y All three indices flagged Source: EPI Info manual 57Part of this Annex draws from the Epi Info manual. 218 Based on these flagging system, and unable to further check possible outliers and inconsistencies in the data, each index is estimated using the flowing flags: Flags used HAZ (0,2,4,6) WAZ (0,1,2,3) WHZ (0,1,4,5) Source: EPI Info manual Thus, we calculated the figures using the recommended thresholds and cross z-score checks, and only using the "un-flagged" observations. The corresponding estimates are the following: TABLE A3.1: IHS-2: (-6; +6) FOR STUNTING AND UNDERWEIGHT; (-4; +6) FOR WASTING; CROSS FLAGGING Stunting Underweight Wasting Moderate Severe Moderate Severe Moderate Severe (-2SD) (-3SD) (-2SD) (-3SD) (-2SD) (-3SD) National 43.7 18.4 18.3 3.5 2.2 0.4 Urban 40.0 14.8 17.2 3.5 1.8 0.7 Rural 44.1 18.8 18.4 3.5 2.2 0.4 North 39.6 18.0 16.1 2.6 2.8 0.2 Centre 47.9 20.9 20.0 4.7 1.8 0.3 South 40.8 16.2 17.2 2.7 2.3 0.6 As noted above, the IHS-1 is using the same reference chart. However, we found two different sources reporting substantially different figures. For moderate stunting, for example, the Statistical Abstract of the IHS-1 (2000) reports an estimate of 49.7 percent, while another publication, "A relative profile of poverty in Malawi, 1998" (2001) by the GoM/PMS, also using the IHS-1, reports a figure of 59.1 percent. Consequently, the next step was to understand which is the right number and what type of assumptions were made in estimating the IHS-1 figures. After different attempts, using the IHS-1 data available to us, we were able to reproduce almost exactly the figures reported in the 2001 publication by the GoM/PMS. The assumptions made are: 1. reference curve 1978 NCHS/CDC/WHO 2. outlier thresholds: (-6, +6) for HAZ, WAZ and WHZ 219 3. no additional flagging criteria used (i.e. no cross-zscore checks) The resulting estimates are the following: TABLE A3.2: IHS-1 (-6; +6) for HAZ, WAZ, WHZ Stunting Underweight Wasting Moderate Severe Moderate Severe Moderate Severe (-2SD) (-3SD) (-2SD) (-3SD) (-2SD) (-3SD) National 59.0 35.8 29.4 10.6 9.2 3.5 Urban 53.9 34.5 18.9 7.9 11.3 7.6 Rural 59.3 35.9 30.3 10.8 9.1 3.3 North 58.5 36.9 24.8 9.0 10.2 4.5 Centre 60.7 36.0 31.1 10.9 8.0 3.1 South 56.8 35.2 28.9 10.6 10.5 3.8 We then re-estimated the IHS-1 malnutrition figures using the same assumptions we proposed for the IHS-2 in Table A3.1. The corresponding table is below: TABLE A3.3: IHS-1 (-6; +6) for stunting and underweight; (-4; +6) for wasting; cross flagging Stunting Underweight Wasting Moderate Severe Moderate Severe Moderate Severe (-2SD) (-3SD) (-2SD) (-3SD) (-2SD) (-3SD) National 57.5 33.0 29.4 10.6 7.9 1.9 Urban 45.2 22.2 18.9 7.9 9.7 5.4 Rural 58.1 33.6 30.3 10.8 7.9 1.7 North 56.4 32.9 24.8 9.0 8.2 2.0 Centre 59.7 34.1 31.1 10.9 7.0 1.9 South 54.7 31.5 28.9 10.6 9.2 1.9 The estimates are somewhat different from the IHS-1 published estimates (see Table A3.2) due to two factors: a threshold of (-4,+6) was used for WHZ, and cross z-score checks were applied to flag observations. All other assumptions remain the same for the two surveys. The advantage of using the more limited flagging for the IHS-2 as well (as in Table A3.2) is to improve comparability with IHS-1. However, comparability with other surveys and internationally recommended standards is somewhat reduced. In theory a decision would have to be made. 220 In practice, however, if we re-estimate IHS-2 using the same reduced flagging adopted in IHS-1 (Table A3.2), the results do not change. As can be seen below, Table A3.4 virtually reproduces Table A3.1. TABLE A3.4: IHS-2 (-6; +6) for HAZ, WAZ, WHZ. No cross-checks. Stunting Underweight Wasting Moderate Severe Moderate Severe Moderate Severe (-2SD) (-3SD) (-2SD) (-3SD) (-2SD) (-3SD) National 43.7 18.4 18.3 3.5 2.2 0.4 Urban 40.0 14.8 17.2 3.5 1.8 0.7 Rural 44.1 18.8 18.4 3.5 2.2 0.4 North 39.7 18.0 16.1 2.6 2.8 0.2 Centre 47.9 20.9 20.0 4.7 1.8 0.3 South 40.7 16.2 17.2 2.7 2.3 0.6 Table A3.4 or Table A3.1, with their different underlying assumptions, can be used interchangeably for the IHS-2 without losing on comparability with either the IHS-1 or the DHS as apparently the additional flags have no effect on the estimates. Thus, for the IHS-2 it would be preferable to use of the "stricter" flagging system, as recommended by CDC/WHO, which is what is done in this report. The following is an interpretation of the flags, as reported in EpiInfo documentation: Flag 0: This means that none of the indices were flagged. However, this does not necessarily mean the information is correct. Either sex, age, weight, or height could be incorrect but not extreme enough to be flagged. Flag 1: HA is flagged but not WH or WA. This could be an extremely short or tall individual. Assure that the height information entered onto the computer file is correct. If height is incorrect, then WHZ would generally be close to -3.09 or 3.09 (a WHZ value beyond these would produce a flag error number 5). The other alternative is that the age information is incorrect, which would make the WAZ extreme (near -6 or 6). Flag 2: WH is flagged but HA and WA are not. First, check the age and height of the child and make sure they are within the limits described above. If the child is within the age and height limitations, then either height or weight may be incorrect. If height is incorrect, then HAZ would be expected to be near an extreme value (but not extreme enough to be flagged), and if weight is incorrect, then WAZ would be close to an extreme value (but not extreme enough to be flagged). Finally, this could truly be an extremely thin or obese child. 221 Flag 3: HA and WH are both flagged but WA is not. This is an indicator that height may be incorrect or missing. Flag 4: WA is flagged but not HA or WH. If the weight is incorrect, then WHZ would be near an extreme value (but not extreme enough to be flagged), and if age is incorrect, then HAZ is likely to be near an extreme value (but not extreme enough to be flagged). Flag 5: HA and WA are flagged but not WH. This is an indication that the age information is incorrect, missing, or out of range. Flag 6: WH and WA are flagged but not HA. This is an indication that weight is likely to be incorrect or missing. Flag 7: All three indices are flagged. This can occur if sex is unknown or incorrectly coded; or at least two of the following are missing, incorrectly coded, or beyond the limitation of the growth curve: age, weight, or height. 222 ANNEX 4B: THE FOSTER, GREER, AND THORBECKE 1984 POVERTY MEASURES The FGT (Foster, Greer, and Thorbecke - 1984) class of poverty indicators are the poverty headcount, the poverty gap, and the squared poverty gap. These are defined as follows: P = 1 p ( z - yi) with 0 N i=1 z Where z is the poverty line, y is the population consumption (or income), and measures the sensitivity of each index to poverty. All three measures are defined for 0 . More specifically: the headcount index is defined for = 0, the poverty gap index is defined for =1, and the squared poverty gap index is defined for = 2. Poverty Headcount: This is defined as the share of the population which is poor, i.e. the proportion of the population for whom consumption (or income) is below the poverty line z. Poverty Gap: The poverty gap, which is often considered as representing the depth of poverty, is the average distance of the population from the poverty line, with the distance of the non-poor equal to zero. The poverty gap measures the poverty deficit of the entire population, with the idea that the poverty deficit should capture the resources that would be needed to lift out of poverty all the poor through perfectly targeted cash transfers. Squared Poverty Gap: This measure picks up the severity of poverty. Whereas the poverty gap is weighted by itself, so as to give more weight to the very poor, using the squared poverty gap allows accounting for the degree of inequality among the poor. The analysis based on the incidence of hunger is similar to using the headcount index of poverty, as it equally weights a household just below the minimum threshold vs. a household falling far below the threshold. However, their level of deprivation and their chances reaching food secure status are quite different. To account for the distance of households from caloric adequacy, we use other two indices proposed by Foster, Greer and Thorbecke (1984). The first is the depth of hunger, or the hunger gap, in which the index is weighted by distance to the minimum caloric requirement. The second is the severity of hunger in which more weight is given to individuals further away from the threshold. What we are interested is seeing is whether the ranking of food insecurity among groups varies according to the indicator used. This is likely to occur if the actual caloric shortfalls are significantly different across regions, and may have implication for the targeting of nutritional programs. We find no re-ranking of geographic areas when using these alternative measures as compared to the incidence of caloric inadequacy (see Table 1 below). 223 Table 1: Depth and Severity of Caloric availability (by region, urban-rural and poverty status) NON-POOR POOR TOTAL Region Region Region North Centre South Total North Centre South Total North Centre South Total Depth Urban 0.03 0.04 0.06 0.05 0.12 0.15 0.18 0.16 0.06 0.07 0.09 0.08 Rural 0.04 0.03 0.03 0.03 0.18 0.17 0.19 0.18 0.12 0.10 0.13 0.12 Total 0.04 0.04 0.04 0.04 0.18 0.17 0.19 0.18 0.12 0.09 0.13 0.11 Severity Urban 0.01 0.01 0.02 0.01 0.04 0.05 0.07 0.06 0.02 0.02 0.03 0.03 Rural 0.01 0.01 0.01 0.01 0.07 0.07 0.07 0.07 0.05 0.04 0.05 0.04 Total 0.01 0.01 0.01 0.01 0.07 0.07 0.07 0.07 0.04 0.04 0.05 0.04 224 ANNEX 4C: DIFFERENT INDICES OF DIETARY DIVERSITY: DEFINITIONS AND A FULL BREAK DOWN OF RESULTS Definition of different food diversity measures No standard definition of measurement or optimal level of food diversity exists in the literature, and food diversity can be quantified in a number of ways. One possibility is based on the number of different kinds of food consumed (simple count), and another upon the number of commodities consumed within a broad commodity group (simple count of food groups). An alternative approach to measuring variety is through diversity indices, which take into account not only whether or not each food is consumed, but also the relative magnitudes of the amount of each food consumed. Simpson index = 1- 2 i Shannon index = - ) ilog( i i Revealed Optimal Diversity index = 1- ( i - i)2 i where i is the calorie share of food i (i=1, 2, ....X), i is the average calorie share of good i consumed by the top four deciles of the distribution of per capita total consumption. If only one food was consumed, the first two indices would be zero; variety thus increases with the index value. The Revealed Optimal Diversity index equals one in the case of optimal diversity, which is assumed to be the average consumption basket of the households of the top four expenditure deciles (Ruiz, 2002). 225 DIET DIVERSITY, BY REGION, URBAN-RURAL AND POVERTY STATUS NON-POOR POOR TOTAL Region Region Region North Centre South Total North Centre South Total North Centre South Total Urban Simple count 20.23 25.39 19.09 21.91 13.68 13.84 14.70 14.19 18.54 22.98 18.30 20.37 Group count 9.07 9.53 8.76 9.12 7.56 7.16 8.16 7.64 8.68 9.04 8.65 8.82 Simpson 0.81 0.90 0.88 0.89 0.54 0.74 0.71 0.70 0.74 0.87 0.85 0.85 Shannon 1.19 1.44 1.24 1.32 1.06 1.09 1.24 1.15 1.16 1.37 1.24 1.29 Revealed Optimal Diversity 0.90 0.95 0.95 0.95 0.73 0.87 0.85 0.85 0.86 0.94 0.93 0.93 Share of food consumption 0.54 0.54 0.53 0.54 0.59 0.62 0.60 0.61 0.55 0.55 0.55 0.55 Rural Simple count 15.67 17.34 16.61 16.85 12.34 12.17 11.94 12.06 14.13 15.34 14.10 14.61 Group count 8.31 8.12 8.19 8.17 7.40 6.60 6.93 6.87 7.89 7.53 7.51 7.56 Simpson 0.87 0.86 0.90 0.88 0.68 0.66 0.71 0.69 0.78 0.78 0.80 0.79 Shannon 1.17 1.20 1.06 1.14 1.20 1.08 1.03 1.07 1.18 1.15 1.04 1.10 Revealed Optimal Diversity 0.94 0.94 0.96 0.95 0.83 0.82 0.84 0.83 0.89 0.89 0.90 0.89 Share of food consumption 0.68 0.59 0.60 0.61 0.66 0.62 0.61 0.62 0.67 0.60 0.61 0.61 Total Simple count 16.26 18.61 17.09 17.71 12.42 12.29 12.06 12.18 14.56 16.31 14.59 15.30 Group count 8.41 8.34 8.30 8.33 7.41 6.64 6.98 6.91 7.97 7.72 7.64 7.71 Simpson 0.86 0.87 0.90 0.88 0.67 0.67 0.71 0.69 0.78 0.79 0.80 0.80 Shannon 1.17 1.24 1.09 1.17 1.20 1.08 1.04 1.07 1.18 1.18 1.07 1.13 Revealed Optimal Diversity 0.94 0.94 0.96 0.95 0.82 0.82 0.84 0.83 0.89 0.90 0.90 0.90 Share of food consumption 0.66 0.58 0.59 0.59 0.65 0.62 0.61 0.62 0.66 0.60 0.60 0.61 226 ANNEX 4D: THE DETERMINANTS OF CHILD MALNUTRITION AND HOUSEHOLD CALORIC AVAILABILITY Table 1: OLS Regression of Child Malnutrition (1) (2) (3) Pooled Under two Over two Child characteristics Boys -0.155*** -0.262*** -0.091** (0.035) (0.056) (0.044) Age in months -0.050*** -0.180*** -0.104*** (0.006) (0.032) (0.022) (Age in months)2 0.001*** 0.005*** 0.001*** (0.000) (0.001) (0.000) Birth spacing§ (less than 24 months younger than older sibling) -0.178*** -0.082 -0.194*** (0.054) (0.114) (0.062) Dummy = 1 if child over two and born in rainy season -0.031 -0.082* (0.045) (0.046) Dummy = 1 if child under two and born in rainy season -0.106* -0.022 (0.057) (0.060) Parental characteristics Woman head hh de jure 0.004 -0.070 0.074 (0.062) (0.103) (0.074) Woman head hh de facto 0.028 -0.107 0.075 (0.113) (0.153) (0.154) Mother completed up to 8 yrs of education 0.238*** 0.316** 0.190** (0.074) (0.124) (0.086) Mother completed more than 8 yrs of education 0.150** 0.099 0.166* (0.071) (0.105) (0.089) Age of the mother 0.074*** 0.073*** 0.095*** (0.014) (0.028) (0.018) (Age of the mother)2 -0.001*** -0.001** -0.001*** (0.000) (0.000) (0.000) Dummy = 1 if mother received prenatal visits 0.085* -0.055 0.216*** (0.046) (0.110) (0.056) HH characteristics No children under 5§ -0.005 -0.069 -0.010 (0.035) (0.051) (0.045) No women over 14 0.017 0.060 -0.014 (0.029) (0.044) (0.035) ln(owned land+1) 0.035 -0.005 0.055 (0.055) (0.088) (0.066) dummy = 1 if hh does not own any land 0.084 -0.063 0.163 Income Ln (real pc exp)§ 0.115*** 0.072 0.153*** (0.040) (0.062) (0.048) 227 (1) (2) (3) Cont'd Pooled Under 2 Over 2 Non self reported community characteristics % of HHs with toilet 0.000 0.002 -0.001 (0.001) (0.002) (0.001) % of HHs with piped or protected water source 0.000 -0.001 0.001 (0.001) (0.001) (0.001) % of HHs without grass roof 0.001 0.001 0.002 (0.001) (0.002) (0.001) % of HHs with child in under 5 programme§ 0.003** 0.003 0.003* (0.001) (0.002) (0.002) % of HHs with tv 0.010*** 0.010* 0.009** (0.003) (0.005) (0.004) Spatial characteristics Centre -0.267*** -0.079 -0.417*** (0.081) (0.124) (0.099) South -0.081 0.185 -0.297*** (0.093) (0.143) (0.114) Rural 0.197** 0.194 0.190* (0.085) (0.141) (0.100) ln (rainfall corresponding to previous crop year) 0.213** 0.225 0.200* (0.091) (0.147) (0.107) Dummy = 1 if hh does not engage in ag. Activity 1.419** 1.698 1.193 (0.641) (1.038) (0.759) coefficient of variation for 2000-2004 rainfall -0.363 -1.165* 0.103 (0.387) (0.616) (0.472) Other controls distance to nearest daily mkt -0.009*** -0.009*** -0.009*** (0.002) (0.003) (0.002) real district maize price -1.465* 0.857 -3.141*** (0.825) (1.336) (0.915) number of severe shocks in last 5 yrs. -0.080 -0.107 -0.044 (0.084) (0.149) (0.092) Constant -4.830*** -4.055*** -3.753*** (0.872) (1.406) (1.089) Observations 6220 2471 3749 R-squared 0.09 0.09 0.08 Notes: Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. The omitted categories are: no religion, male headed households, North region, Urban households. § indicates endogenous variables. Dummies for month of interview and household head religion are not reported. 228 Table 2: 2SLS Regression of Child Malnutrition (1) (2) (3) Dep. variable: z-score on Height-for-age Pooled Under 2 Over 2 Child characteristics Boys -0.156*** -0.261*** -0.094* (0.039) (0.065) (0.049) Age in months -0.052*** -0.201*** -0.074*** (0.006) (0.038) (0.026) (Age in months)2 0.001*** 0.005*** 0.001** (0.000) (0.001) (0.000) Birth spacing -0.173*** -0.087 -0.196*** (0.059) (0.125) (0.066) Child over two and born in rainy season -0.013 -0.056 (0.050) (0.051) Child under two and born in rainy season -0.095 -0.028 (0.063) (0.068) Parental characteristics Woman head hh de jure 0.155* 0.092 0.212** (0.081) (0.133) (0.096) Woman head hh de facto 0.046 -0.084 0.063 (0.131) (0.187) (0.169) Mother completed up to 8 yrs of education 0.131 0.173 0.098 (0.083) (0.141) (0.097) Mother completed more than 8 yrs of education -0.084 -0.099 -0.101 (0.094) (0.144) (0.116) Age of the mother 0.099*** 0.122*** 0.117*** (0.015) (0.035) (0.017) (Age of the mother)2 -0.001*** -0.002*** -0.001*** (0.000) (0.001) (0.000) Dummy = 1 if mother received prenatal visits -0.002 -0.419** 0.148** (0.056) (0.175) (0.066) HH characteristics No children under 5 0.088* -0.011 0.093 (0.045) (0.064) (0.058) No women over 14 0.009 0.084 -0.032 (0.034) (0.053) (0.039) ln(owned land+1) -0.069 -0.107 -0.052 (0.065) (0.104) (0.077) dummy = 1 if hh does not own any land 0.031 -0.099 0.091 (0.103) (0.165) (0.124) Income Ln (real pc exp)§ 0.608*** 0.605*** 0.637*** (0.106) (0.176) (0.126) Non self reported community characteristics % of HHs with toilet 0.003* 0.006** 0.001 (0.001) (0.002) (0.002) % of HHs with piped or protected water source -0.001 -0.002 -0.000 (0.001) (0.001) (0.001) % of HHs without grass roof 0.002 0.002 0.002 (0.001) (0.002) (0.002) 229 (1) (2) (3) Cont'd Pooled Under 2 Over 2 % of HHs with child in under 5 programme§ 0.047*** 0.058*** 0.042*** (0.011) (0.020) (0.013) % of HHs with tv 0.005 0.003 0.005 (0.003) (0.006) (0.004) Spatial characteristics Centre 0.184 0.494* -0.032 (0.157) (0.257) (0.186) South 0.431*** 0.886*** 0.122 (0.159) (0.281) (0.181) Rural -0.033 -0.014 -0.044 (0.117) (0.175) (0.147) ln (rainfall corresponding to previous crop year) 0.398*** 0.432** 0.386*** (0.114) (0.180) (0.140) Dummy = 1 if hh does not engage in ag. Activity 2.709*** 3.144** 2.481** (0.799) (1.261) (0.976) coefficient of variation for 2000-2004 rainfall 0.390 -0.353 0.927 (0.481) (0.754) (0.597) Other controls distance to nearest daily mkt (km) -0.007*** -0.009** -0.007** (0.002) (0.004) (0.003) real district maize price 1.685 5.269*** -0.649 (1.152) (2.002) (1.269) number of severe shocks in last 5 yrs. 0.108 0.128 0.129 (0.105) (0.181) (0.117) Constant -14.483*** -15.292*** -13.507*** (2.098) (3.658) (2.681) Observations 6106 2423 3683 Centered R2 -0.12 -0.23 -0.10 0.85 0.78 1.65 Hansen J statisticT [H0: joint validity of instruments] P-value 0.65 0.68 0.44 Hausman t-test (H0: OLS consistent, H1: IV consistent and efficient) ln(pc exp.) 5.02 3.24 4.16 % of HHs with child in under 5 programme 4.02 2.76 3.04 F test on excluded instruments ln(pc exp.) 202.69 90.65 160.88 % of HHs with child in under 5 programme 38.00 16.49 29.85 Notes: Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. The omitted categories are: no religion, male headed households, North region, Urban households. Dummies for month of interview and for the household's head religion are not shown. Excluded instruments are: wealth index on durable assets derived with PCA, the highest education level in the HH other than the mother's, and the average level of per capita expenditure in the community, and a dummy for whether the household head is polygamous. The definition of North region, Centre region, and South region includes urban areas. § These are variables are treated as endogenous. 230 Table3: OLS and 2SLS Estimates of calorie demand elasticity OLS estimates (Income spline) OLS estimates IV estimates Dependent variable is ln(daily pc calories) Income Ln (real pc exp.) 0.598*** 0.523*** 0.394*** (0.016) (0.012) (0.017) Ln (real pc exp.)*Q2 Dummy -0.002* (0.001) Ln (real pc exp.)*Q3 Dummy -0.007*** (0.002) Ln (real pc exp.)*Q4 Dummy -0.016*** (0.002) Share non food. exp -0.798*** -0.798*** -0.721*** (0.042) (0.042) (0.040) Household head characteristics Household Head Employed (ILO definition) 0.037** 0.038** 0.055*** (0.015) (0.015) (0.015) Female household head -0.034** -0.043*** -0.068*** (0.014) (0.015) (0.015) Age of Household Head 0.003** 0.003** 0.003*** (0.001) (0.001) (0.001) (Age of Household Head)2 -0.000** -0.000** -0.000*** (0.000) (0.000) (0.000) Polygamous 0.005 0.008 0.017* (0.010) (0.010) (0.010) Separated 0.048** 0.065*** 0.081*** (0.019) (0.019) (0.020) Divorced 0.092*** 0.106*** 0.124*** (0.019) (0.019) (0.019) Widowed or widower 0.044*** 0.056*** 0.072*** (0.017) (0.017) (0.017) Never married 0.097*** 0.119*** 0.154*** (0.020) (0.020) (0.020) Household composition and characteristics Highest education level in household 0.011*** 0.011*** 0.007** (0.003) (0.003) (0.003) (Highest education level in household)2 -0.001*** -0.001*** -0.001*** (0.000) (0.000) (0.000) Household size -0.028*** -0.042*** -0.055*** (0.003) (0.002) (0.003) Proportion of HH members 0-14 yrs -0.856* -0.890* -1.137** (0.445) (0.459) (0.455) Proportion of HH members 15-64 yrs -0.671 -0.670 -0.851* (0.446) (0.460) (0.455) 231 Cont'd OLS estimates OLS estimates IV estimates (Income spline) Non- self reported Community characteristics % of HHs with toilet in EA -0.001* -0.001* -0.001** (0.000) (0.000) (0.000) % of HHs with improved water source 0.000 0.000 0.000 (0.000) (0.000) (0.000) % of HHs with improved roof -0.001** -0.001** -0.000 (0.000) (0.000) (0.000) % of HHs with child in under 5 programme -0.000 -0.000 -0.000 (0.000) (0.000) (0.000) % of HHs with radio 0.001** 0.001** 0.001** (0.000) (0.000) (0.000) Distance to nearest daily mkt -0.002*** -0.002** -0.002** (0.001) (0.001) (0.001) Real price of maize -0.244 -0.289 -0.241 (0.400) (0.416) (0.384) Spatial characteristics Centre -0.000 -0.006 0.016 (0.034) (0.035) (0.033) South -0.039 -0.045 -0.057 (0.042) (0.043) (0.040) Rural 0.021 0.022 0.007 (0.029) (0.029) (0.028) Rainfall -0.000 -0.000 -0.000 (0.000) (0.000) (0.000) (Rainfall)2 0.000 0.000 0.000 (0.000) (0.000) (0.000) CV for 2000-2004 rainfall 0.134 0.140 0.124 (0.127) (0.129) (0.133) Constant 3.259*** 4.020*** 5.543*** (0.509) (0.509) (0.540) Observations 11272 11272 11269 R-squared 0.69 0.69 0.68 F-test on income segments 16.55 (H0: slope dummy variables for all quartiles = 0) P-value (0.00) Hansen J statistic [H0: joint validity of instruments] 1.48 P-value (0.22) -11.29 Hausman t-test (H0: OLS consistent, H1: IV consistent and efficient) F test on excluded instruments 676.36 P-value (0.00) Notes: Robust standard errors in parentheses. Household clusters employed. * significant at 10%; ** significant at 5%; *** significant at 1%. EA cluster effects included Omitted categories are: HH head monogamous, HH head has no religious belief, proportion of HH members older than 64, Urban, North. Dummies for month of interview and household's head religion not reported. Excluded instruments are: wealth index on durable assets derived with PCA, and the proportion of land under tobacco cultivation. Note: The definition of North region, Centre region, and South region includes urban areas. 232 ANNEX 5A: PREVALENCE OF HOUSEHOLDS IN WHICH A CHILD RESIDES WITHOUT EITHER PARENT (EITHER DUE TO BEING ORPHANED OR FOSTERED); LIST OF ASSETS AND THE RELATIONSHIP BETWEEN THE ASSET INDEX AND PER CAPITA EXPENDITURE Table 1: Households with children aged 0-17 years that are orphaned in Malawi Households Population of children Without orphans With orphans Non-orphans Orphans Region North 82.6 16.4 87.9 12.1 Central 84.3 15.7 88.1 11.9 South 82.7 17.3 85.6 14.4 Urban/Rural Urban 83.3 16.7 85.7 14.3 Rural 83.4 16.6 87.0 13.0 All 83.4 16.6 86.9 13.1 Source: IHS2 Table 2: Proportion of orphans among children (aged 0-17) by gender Orphan by type: Non-orphans Paternal Maternal Double All 86.9 7.5 2.9 2.7 Gender Girls 87.1 7.4 2.9 2.7 Boys 86.7 7.6 3.0 2.7 Age 0-5 95.9 2.8 1.0 0.3 6-12 84.7 8.8 3.6 2.8 13-17 73.8 13.9 5.4 6.9 Source: IHS2 233 Figure 1: Percentage of children that are orphan by age 35 30 enrdli 25 chlalfot Paternal orphans 20 Maternal orphans 15 Double orphans All orphans cen 10 erP 5 0 1 3 5 7 9 11 13 15 17 Age Source: IHS2 Table 3: Living arrangements for orphans Living arrangement Both parents One parent Neither All children 64.9 19.1 16.0 Non-orphans 74.7 15.0 10.3 Orphans Paternal orphans n.a 72.1 27.9 Maternal orphans n.a 22.4 77.7 Single orphans n.a 58.1 41.9 Double orphans n.a n.a. 100.0 Source: IHS2 234 Appendix Figure 1: Relationship between household asset index and per capita expenditure )K 50000 M( 45000 erutid 40000 35000 30000 en 25000 exp 20000 ati 15000 10000 cap 5000 erP 0 1 2 3 4 5 Asset Index 235 Appendix Table 1: Types of Households in IHS2 Mean number of Percent of children (0-17) in these types of Household Characteristic households households Any children (0-17 years) 82.4 2.87 Any children (0-17 years) with 1 or both parents in 72.1 3.01 the household Any grandchildren (0-17 years) of the household 12.0 2.78 head with no parent in the household Any children (0-17 years) who are not grandchildren 11.9 3.33 of the household head and have no parent in the household Household head is child (14-17 years) 0.2 1.56 Source: IHS2. 236 ANNEX 5B: RESULTS OF THE MODEL PROBIT ON LIKELIHOOD OF ORPHANS ATTENDING SCHOOL Appendix Table 2: Probit model for children's school attendance (Age 6-17) Marginal effects (1) (2) (3) (4) All All Boys Girls Orphan status Single orphan -0.023** [0.009] Double orphan -0.042** [0.016] Orphan living with one parent -0.027* -0.018 -0.035 [0.014] [0.018] [0.020] Double or "virtual" double orphan -0.079** -0.076** -0.078** [0.012] [0.017] [0.017] Non-orphan living with one parent -0.040** -0.045** -0.032 [0.012] [0.017] [0.016] Non-orphan living away from parents -0.083** -0.043** -0.113** [0.012] [0.016] [0.017] Child characteristics Age 0.197** 0.196** 0.183** 0.207** [0.006] [0.006] [0.008] [0.008] Age squared -0.009** -0.009** -0.008** -0.010** [0.000] [0.000] [0.000] [0.000] Female -0.010 -0.009 [0.006] [0.006] Household characteristics PA death past two years in household 0.003 0.003 -0.014 0.019 [0.013] [0.013] [0.018] [0.017] PA chronically ill with AIDS in household -0.016** -0.019** -0.022** -0.018* [0.006] [0.006] [0.009] [0.009] Poor household -0.049** -0.053** -0.056** -0.050** [0.006] [0.006] [0.008] [0.008] Female head 0.036** 0.053** 0.049** 0.054** [0.007] [0.008] [0.011] [0.011] Head has some primary education 0.086** 0.083** 0.083** 0.085** [0.007] [0.007] [0.009] [0.010] Head has some post primary education 0.126** 0.127** 0.135** 0.120** [0.006] [0.006] [0.007] [0.008] Urban 0.057** 0.057** 0.054** 0.061 [0.009] [0.009] [0.013] [0.012] Central -0.004 -0.009 -0.020* 0.009 [0.006] [0.006] [0.009] [0.009] North 0.096** 0.096** 0.107** 0.089** [0.007] [0.007] [0.009] [0.009] Dependent variable: % attending school 0.821 0.821 0.822 0.821 Number of observations 16,030 16,030 7,976 8,054 237 Pseudo R2 0.127 0.132 0.126 0.154 Source: IHS2. Notes: Standard errors in brackets. ** indicates statistical significance at 1% and * at 5%. indicates binary indicator variables (=1 if true, else 0). 238 Appendix Table 3: Probit model for children's school attendance (Age 15-17) Marginal effects (1) (2) (3) (4) All All Boys Girls Orphan status Single orphan -0.028 [0.024] Double orphan -0.027 [0.034] Orphan living with one parent -0.081* -0.049 -0.107* [0.037] [0.046] [0.056] Double or "virtual" double orphan -0.186** -0.147** -0.224** [0.029] [0.040] [0.042] Non-orphan living with one parent -0.153** -0.161** -0.142** [0.036] [0.048] [0.053] Non-orphan living away from parents -0.313** -0.170** -0.423** [0.030] [0.043] [0.038] Child characteristics Age -0.104** -0.101** -0.062** -0.141** [0.010] [0.010] [0.013] [0.016] Female -0.123** -0.113** [0.017] [0.017] Household characteristics PA death past two years in household -0.033 -0.039 -0.087 -0.009 [0.037] [0.037] [0.051] [0.055] PA chronically ill with AIDS in household -0.021 -0.030 -0.040 -0.027 [0.018] [0.018] [0.023] [0.027] Poor household -0.014 -0.049** -0.046* -0.058* [0.018] [0.018] [0.023] [0.029] Female head 0.066** 0.120** 0.105** 0.127** [0.021] [0.022] [0.027] [0.034] Head has some primary education 0.092** 0.094** 0.074** 0.114** [0.020] [0.020] [0.025] [0.032] Head has some post primary education 0.191** 0.207** 0.186** 0.226** [0.021] [0.020] [0.024] [0.033] Urban 0.060* 0.071** 0.034 0.103** [0.025] [0.025] [0.035] [0.037] Central -0.003 -0.013 -0.034 0.012 [0.019] [0.019] [0.024] [0.029] North 0.098** 0.110** 0.173** 0.058 [0.023] [0.023] [0.025] [0.037] Dependent variable: % attending school 0.697 0.697 0.749 0.642 Number of observations 3,028 3,028 1,560 1,468 Pseudo R2 0.068 0.101 0.084 0.128 Source: IHS2. Notes: Standard errors in brackets. ** indicates statistical significance at 1% and * at 5%. indicates binary indicator variables (=1 if true, else 0). 239 Appendix Table 4: Probit models for children being absent from school in past two weeks among students attending (Ages 6-17) Marginal effects (1) (2) (3) (4) All All Boys Girls Orphan status Single orphan -0.008 [0.012] Double orphan -0.043* [0.019] Orphan living with one parent 0.011 0.001 0.023 [0.017] [0.024] [0.025] Double or "virtual" double orphan -0.030* -0.031 -0.029 [0.014] [0.020] [0.020] Non-orphan living with one parent 0.008 -0.013 0.029 [0.015] [0.021] [0.021] Non-orphan living away from parents 0.003 -0.007 0.012 [0.014] [0.020] [0.020] Child characteristics Age -0.007** -0.007** -0.007** -0.008** [0.001] [0.001] [0.002] [0.002] Female -0.010 -0.010 [0.008] [0.008] Household characteristics PA death past two years in household -0.030 -0.031 -0.047 -0.016 [0.017] [0.017] [0.024] [0.025] PA chronically ill with AIDS in household 0.058** 0.057** 0.062** 0.052** [0.009] [0.009] [0.012] [0.012] Someone ill in household and others must 0.009 0.009 0.003 0.021 stop working [0.009] [0.009] [0.013] [0.013] Poor household 0.021* 0.021* 0.036** 0.006 [0.008] [0.008] [0.012] [0.012] Female head 0.023* 0.017 0.028 0.006 [0.011] [0.013] [0.018] [0.018] Head has some primary education -0.013 -0.013 -0.016 -0.010 [0.010] [0.010] [0.014] [0.014] Head has some post primary education -0.099** -0.099** -0.091** -0.104** [0.012] [0.012] [0.017] [0.017] Urban -0.113** -0.113** -0.118** -0.109** [0.011] [0.011] [0.016] [0.016] Central 0.010 0.010 0.004 0.017 [0.009] [0.009] [0.013] [0.013] North -0.020 -0.020 -0.041** 0.000 [0.011] [0.011] [0.016] [0.016] Dependent variable: % attending school 0.282 0.282 0.287 0.277 Number of observations 13,005 13,005 6,483 6,522 Pseudo R2 0.024 0.025 0.020 0.021 Source: IHS2. Notes: Standard errors in brackets. ** indicates statistical significance at 1% and * at 5%. indicates binary indicator variables (=1 if true, else 0). 240 Appendix Table 5: Probit models for children being absent from school in past two weeks among students attending (Ages 15-17) Marginal effects (1) (2) (3) (4) All All Boys Girls Orphan status Single orphan 0.011 [0.026] Double orphan -0.031 [0.036] Orphan living with one parent 0.025 0.031 0.014 [0.037] [0.050] [0.054] Double or "virtual" double orphan -0.031 -0.019 -0.043 [0.029] [0.042] [0.040] Non-orphan living with one parent -0.035 -0.020 -0.054 [0.034] [0.048] [0.045] Non-orphan living away from parents 0.011 0.018 -0.001 [0.034] [0.044] [0.052] Child characteristics Age -0.007 -0.007 -0.009 -0.004 [0.012] [0.012] [0.016] [0.018] Female -0.026 -0.025 [0.019] [0.019] Household characteristics PA death past two years in household -0.017 -0.019 -0.077 0.043 [0.039] [0.039] [0.050] [0.060] PA chronically ill with AIDS in household 0.042* 0.041* 0.047 0.031 [0.020] [0.020] [0.027] [0.029] Someone ill in household and others must 0.006 0.007 -0.020 0.041 stop working [0.022] [0.022] [0.030] [0.034] Poor household -0.037 -0.036 -0.024 -0.053 [0.021] [0.021] [0.028] [0.031] Female head 0.024 0.029 0.017 0.049 [0.025] [0.029] [0.039] [0.043] Head has some primary education -0.046* -0.048* -0.060 -0.029 [0.023] [0.023] [0.032] [0.035] Head has some post primary education -0.126** -0.126** -0.120** -0.130** [0.026] [0.026] [0.036] [0.038] Urban -0.080** -0.080** -0.087* -0.068 [0.026] [0.026] [0.036] [0.036] Central 0.031 0.032 0.025 0.041 [0.022] [0.022] [0.029] [0.032] North 0.061* 0.061* 0.058 0.061 [0.030] [0.030] [0.041] [0.042] Dependent variable: % attending school 0.238 0.238 0.253 0.221 Number of observations 2,064 2,064 1,144 920 Pseudo R2 0.025 0.026 0.024 0.034 Source: IHS2. Notes: Standard errors in brackets. ** indicates statistical significance at 1% and * at 5%. indicates binary indicator variables (=1 if true, else 0). 241 Appendix Table 6: Probit models for children's chores and paid work Marginal effects (1) (2) Did economic Did chores yesterday activity last week (Ages 6-17) (Ages 12-17) Orphan status Orphan living with one parent 0.001 0.008 [0.013] [0.019] Double or "virtual" double orphan 0.038** 0.006 [0.012] [0.015] Non-orphan living with one parent 0.027* 0.003 [0.014] [0.016] Non-orphan living away from parents 0.037** 0.040** [0.013] [0.015] Child characteristics Age -0.012** 0.197** [0.002] [0.010] Age squared -0.007** [0.000] Female -0.031** 0.386** [0.006] [0.008] Household characteristics PA death past two years in household 0.016 -0.012 [0.014] [0.019] PA chronically ill with AIDS in household 0.018** -0.002 [0.007] [0.009] Someone ill in household and others must 0.020** -0.005 stop working [0.008] [0.010] Poor household 0.005 -0.059** [0.007] [0.009] Female head 0.037** 0.035** [0.011] [0.014] Head has some primary education -0.003 0.058** [0.007] [0.010] Head has some post primary education -0.017 0.040** [0.010] [0.015] Urban 0.047** -0.057** [0.008] [0.014] Central 0.038** 0.017 [0.007] [0.010] North -0.027** -0.150** [0.009] [0.013] Dependent variable: % engaged in activity 0.089 0.514 Number of observations 6,781 15,901 Pseudo R2 0.060 0.178 Source: IHS2. Notes: Standard errors in brackets. ** indicates statistical significance at 1% and * at 5%. indicates binary indicator variables (=1 if true, else 0). 242 ANNEX 6A: TABLES ON NATIONAL ACCOUNTS SECTORAL COMPOSITION OF GDP, AND EMPLOYMENT Table 6A.1 National Accounts: Gross Domestic Product (GDP) by Sector of Origin at 1994 Factor Cost Prices (annual percentage growth rates) 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005* GDP at Factor Cost 13.8 10.0 6.6 1.1 3.5 0.8 -4.1 2.1 3.9 4.6 2.1 Agriculture, Forestry and Fishing 39.6 25.5 0.1 10.3 10.1 5.3 -6.0 2.7 5.9 2.7 -6.7 Smallscale 43.6 31.7 -3.5 18.8 13.4 1.6 -4.8 -0.4 12.4 -1.4 -7.4 Largescale 30.3 9.6 11.2 -12.3 -1.9 21.0 -10.3 14.2 -15.4 20.5 -4.2 Industry Mining and Quarying 9.2 336.6 -23.5 4.2 3.4 10.8 7.5 -38.7 18.6 -11.6 40.1 Manufacturing 5.5 -0.6 1.0 1.5 1.8 -3.0 -14.2 -0.1 3.2 6.9 3.6 Electricity and Water 2.0 -0.2 5.8 7.2 -0.4 10.2 -7.0 5.8 2.4 7.5 5.8 Construction 3.0 11.6 6.8 1.9 15.5 -2.2 -4.7 14.1 13.3 10.9 17.1 Services 5.6 1.2 14.8 -5.7 -1.5 -2.6 0.2 2.9 1.5 5.8 7.7 Distribution 2.2 -0.6 16.8 -6.6 -1.8 -0.3 1.1 1.6 -0.8 6.9 10.0 Transport and Communication 17.4 -7.3 8.3 -0.3 4.8 -4.2 -0.6 17.5 8.3 7.2 7.7 Financial and Professional Services 10.2 20.8 35.1 -8.3 -0.3 2.0 -3.0 6.7 6.1 7.3 9.7 Ownership of Dwellings 2.0 2.1 2.1 2.1 2.1 2.6 2.8 2.8 2.8 2.8 2.8 Private Social and Community Services 1.9 9.8 10.0 0.7 0.7 2.7 2.9 2.9 2.9 2.8 2.8 Producers of Government Services 7.1 -2.6 1.7 -4.9 -1.8 -9.9 0.8 -0.5 1.7 2.3 2.4 Unallocable Finance Charges 9.9 3.7 13.8 -5.1 10.5 2.4 -0.3 13.5 9.6 7.5 10.0 Per capita real GDP growth 11.7 7.9 4.6 -0.8 1.6 -1.1 -5.9 0.2 2.0 2.7 0.2 Source: National Statistical Office *Estimate Table 6A.2 Sectoral Composition of GDP: GDP by sector as a share of GDP, at 1994 constant market prices (percentage contribution to GDP) 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005* GDP at Factor Cost 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Agriculture 30.4 34.7 32.6 35.6 37.8 39.5 38.8 39.0 39.8 39.1 35.7 Smallscale 21.9 26.2 23.7 27.9 30.6 30.8 30.6 29.9 32.3 30.5 27.6 Largescale 8.5 8.5 8.8 7.7 7.3 8.7 8.2 9.2 7.5 8.6 8.1 Industry 21.3 19.6 19.3 18.1 18.3 18.2 17.9 16.7 16.1 16.4 16.6 Mining and Quarying 0.4 1.8 1.3 1.3 1.3 1.4 1.6 1.0 1.1 0.9 1.3 Manufacturing 15.8 14.3 13.5 13.6 13.4 12.9 11.5 11.3 11.2 11.4 11.6 Electricity and Water 1.4 1.3 1.3 1.4 1.3 1.4 1.4 1.4 1.4 1.5 1.5 Construction 2.0 2.0 2.0 2.0 2.2 2.2 2.2 2.4 2.6 2.8 3.2 Services 48.4 44.5 47.9 44.7 42.5 41.1 43.0 43.4 42.4 42.8 45.2 Distribution 24.3 22.0 24.1 22.3 21.1 20.9 22.0 21.9 21.0 21.4 23.1 Transport and Communication 5.1 4.3 4.4 4.3 4.4 4.2 4.3 5.0 5.2 5.3 5.6 Financial and Professional Services 6.5 7.1 9.0 8.2 7.9 8.0 8.1 8.5 8.6 8.9 9.5 Ownership of Dwellings 1.6 1.4 1.4 1.4 1.4 1.4 1.5 1.5 1.5 1.5 1.5 Private Social and Community Services 2.0 2.0 2.1 2.1 2.0 2.1 2.2 2.2 2.2 2.2 2.2 Producers of Government Services 13.3 11.8 11.2 10.6 10.0 9.0 9.4 9.2 9.0 8.8 8.8 Unallocable Finance Charges -2.9 -2.7 -2.9 -2.7 -2.9 -2.9 -3.1 -3.4 -3.6 -3.7 -4.0 Source: National Statistical Office *Estimate 243 244 Table 6A.3. Sectoral contribution to GDP growth (percentage) 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005* Agriculture 12.0 8.8 0.0 3.7 3.8 2.1 -2.3 1.0 2.4 1.1 -2.4 Small scale 9.5 8.3 -0.8 5.2 4.1 0.5 -1.5 -0.1 4.0 -0.4 -2.0 Large scale 2.6 0.8 1.0 -0.9 -0.1 1.8 -0.8 1.3 -1.1 1.8 -0.3 Industry Mining & quarrying 0.0 5.9 -0.3 0.1 0.0 0.2 0.1 -0.4 0.2 -0.1 0.5 Manufacturing 0.9 -0.1 0.1 0.2 0.2 -0.4 -1.6 0.0 0.4 0.8 0.4 Electricity & water 0.0 0.0 0.1 0.1 0.0 0.1 -0.1 0.1 0.0 0.1 0.1 Construction 0.1 0.2 0.1 0.0 0.3 0.0 -0.1 0.3 0.4 0.3 0.5 Services Distribution 0.5 -0.1 4.0 -1.5 -0.4 -0.1 0.2 0.3 -0.2 1.5 2.3 Transport & communication 0.9 -0.3 0.4 0.0 0.2 -0.2 0.0 0.9 0.4 0.4 0.4 Financial & professional services 0.7 1.5 3.2 -0.7 0.0 0.2 -0.2 0.6 0.5 0.6 0.9 Ownership of dwellings 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Private social and community services 0.0 0.2 0.2 0.0 0.0 0.1 0.1 0.1 0.1 0.1 0.1 Producers of government services 0.9 -0.3 0.2 -0.5 -0.2 -0.9 0.1 0.0 0.2 0.2 0.2 Unallocable finance charges -0.3 -0.1 -0.4 0.1 -0.3 -0.1 0.0 -0.5 -0.3 -0.3 -0.4 Source: National Statistical Office *Estimate Table 6A.4. Employment: Distribution of the Working Population aged 5 years or over by Economic Activity in 2004 (percent) Malawi Rural Urban Agriculture 78.8 84.1 10.5 Mining and quarrying 0.4 0.4 0.5 Manufacturing 1.9 1.6 6.5 Construction 2.3 1.9 7.9 Transport 0.6 0.3 4.5 Trade or selling 9.6 7.5 38.1 Other services 3.4 1.9 23 Education and Health 1.8 1.5 6.3 Administration 0.3 0.2 1.9 Other 0.7 0.6 0.9 Source: National Statistical Office 245 ANNEX 7A: LAND HOLDINGS BY HOUSEHOLD AND PER CAPITA Table 7A.1 Land Holdings: Average Hectares of Land (per household and per capita) per capita Total household Total Total Total Total Total Total Cultivated Rainfed Dimba Tree Rent Uncultivated (rainfed+dimba+tree) All 0.318 1.208 1.024 0.947 0.075 0.002 0.03 0.154 North 0.411 1.681 1.186 1.09 0.092 0.004 0.008 0.487 Central 0.33 1.322 1.100 1.012 0.086 0.002 0.05 0.172 South 0.287 1.001 0.922 0.859 0.061 0.002 0.017 0.064 Poor 0.227 1.130 0.970 0.898 0.07 0.002 0.022 0.139 Non-poor 0.399 1.276 1.072 0.991 0.079 0.002 0.037 0.167 1st decile 0.166 1.008 0.865 0.812 0.053 0.000 0.018 0.126 2 0.190 1.078 0.955 0.888 0.064 0.003 0.014 0.108 3 0.215 1.106 0.949 0.883 0.063 0.003 0.017 0.141 4 0.246 1.198 1.023 0.929 0.093 0.001 0.023 0.152 5 0.256 1.149 0.979 0.906 0.072 0.001 0.031 0.139 6 0.287 1.225 1.032 0.956 0.075 0.001 0.026 0.168 7 0.316 1.246 1.005 0.920 0.081 0.004 0.040 0.201 8 0.334 1.231 1.042 0.961 0.078 0.003 0.037 0.152 9 0.415 1.302 1.086 1.006 0.078 0.002 0.048 0.168 10 0.534 1.342 1.156 1.075 0.079 0.002 0.029 0.157 Notes: Includes urban and rural households. Excludes landless. Excludes households for which we cannot calculate total land. Includes rented in land. Deciles are defined for sample n=9759. Source: National Statistical Office, IHS2 Table 7A.2: Comparison of land holdings from IHS1 and IHS2: Average Hectares of Land (per household and per capita) IHS1 Cultivated land IHS2 Cultivated land IHS2 Total land (hectares per household) (hectares per household) (hectares per household) Malawi 0.992 1.024 1.208 North Region 1.134 1.186 1.681 Center Region 1.195 1.100 1.322 South Region 0.759 0.922 1.001 IHS1 Cultivated land IHS2 Cultivated land IHS2 Total land (hectares per capita) (hectares per capita) (hectares per capita) Malawi 0.222 0.237 0.282 North Region 0.256 0.249 0.372 Center Region 0.257 0.245 0.298 South Region 0.176 0.228 0.249 Source: National Statistical Office, IHS1 and IHS2. The data on land from IHS1 is taken from: NEC (2000), Profile of Poverty in Malawi 1998, Page 61, Table 26. Notes: Excludes landless. Cultivated land excludes land rented out and uncultivated land. Total land includes cultivated land, uncultivated land, and land rented out. 246 ANNEX 7B: PROPORTION OF FARMERS THAT RECEIVED ADVICE FROM FIELD ASSISTANT BY TYPE OF ADVICE AND BY QUINTILE Any Crop New Fert Pest Irriga- Animal Animal Crop Credit Tobacco advice Prodn Seed Use Control tion Care Disease Sales Access Overall 11 11 9 10 7 7 6 6 6 6 4 Poorest 20% 18 18 19 19 17 17 16 16 16 16 15 2nd Quintile 21 21 20 21 21 22 21 21 20 22 17 3rd Quintile 23 23 23 22 22 23 23 23 24 22 20 4th Quintile 22 22 22 22 23 22 22 23 22 23 28 Richest 20% 16 16 17 16 17 16 18 17 18 17 20 Male Headed 13 12 11 11 8 8 7 7 7 7 5 Hhs Poorest 20% 18 18 19 18 17 17 16 16 17 15 15 2nd Quintile 21 21 20 21 21 21 21 20 19 22 18 3rd Quintile 23 23 23 23 22 24 22 23 24 23 20 4th Quintile 22 22 22 22 24 22 23 24 22 23 28 Richest 20% 16 16 16 16 17 15 18 17 18 18 20 Female 7 7 6 6 4 4 3 3 3 3 2 Headed Hhs Poorest 20% 19 18 20 19 18 17 14 14 14 23 13 2nd Quintile 22 23 19 22 22 23 25 25 25 22 14 3rd Quintile 23 23 23 21 22 22 25 28 22 16 25 4th Quintile 21 20 18 22 23 21 18 17 24 26 28 Richest 20% 16 16 19 16 16 17 18 17 16 13 20 Urban 2 2 2 2 1 1 1 1 1 1 0.4 Poorest 20% 0 0 0 0 0 0 0 0 0 0 0 2nd Quintile 12 12 6 7 0 11 10 11 12 16 0 3rd Quintile 43 43 40 44 45 65 63 64 41 43 72 4th Quintile 19 19 29 21 22 12 20 21 26 25 9 Richest 20% 26 26 26 28 33 12 8 5 21 17 19 Rural 13 12 10 11 8 8 7 7 7 7 5 Poorest 20% 18 19 19 19 18 18 16 16 17 16 15 2nd Quintile 21 21 20 21 21 22 21 21 20 22 18 3rd Quintile 23 23 22 22 21 23 22 23 24 22 20 4th Quintile 22 22 22 22 23 22 22 23 22 23 28 Richest 20% 16 16 17 16 17 16 18 17 18 17 20 North 22 21 20 21 17 17 16 15 15 15 10 Poorest 20% 21 20 20 21 17 20 18 18 17 18 19 2nd Quintile 20 19 18 19 17 18 17 16 16 18 16 3rd Quintile 21 21 21 21 23 23 23 23 22 20 18 4th Quintile 22 23 23 23 26 24 26 25 25 26 30 Richest 20% 16 17 17 16 18 15 17 17 19 18 17 Centre 11 10 8 9 6 7 6 5 6 5 5 Poorest 20% 9 9 9 9 10 9 9 8 9 9 9 2nd Quintile 18 18 16 18 16 18 19 19 18 18 17 3rd Quintile 21 24 24 24 22 25 25 25 26 23 21 4th Quintile 23 29 29 29 30 27 25 28 27 27 30 Richest 20% 24 20 22 21 22 21 22 21 22 23 23 South 10 9 8 8 6 5 5 5 4 4 2 Poorest 20% 25 26 27 26 24 25 23 22 24 21 24 2nd Quintile 24 25 24 25 27 28 27 25 25 29 21 3rd Quintile 23 24 23 22 21 22 21 22 23 22 19 4th Quintile 16 15 15 16 16 15 17 17 14 17 19 Richest 20% 12 11 12 11 11 10 14 14 13 10 16 247 ANNEX 7C: SMALLHOLDER FARMERS' EFFICIENCY IN HYBRID MAIZE AND BURLEY TOBACCO PRODUCTION Methodology Theoretical aspects of technical, allocative and economic efficiency are dealt with within the traditional theory of production economics, where economic productive efficiency derives from technical as well allocative or factor price efficiency. According to Forsund et al. (1980), technical efficiency implies a combination of inputs that for a given monetary outlay maximizes the level of production. Whereas technical efficiency reflects the ability of the farmer to maximize output for a given set of resource inputs, allocative efficiency reflects the ability of the farmer to utilize the inputs at his/her disposal in optimal proportions given their respective prices and the available production technology. Empirically, many studies on farm productivity and efficiency use a production function approach in contrast to a profit function approach because although the later is less restrictive, its estimation requires substantial variation in output and input prices, which does not normally obtain in the case of cross-section data (Khandker et. al. 1986). In our analysis we use the value of crop production as a representation of a production technology, with the assumption of constant returns to scale (CRS). To obtain the parametric measure of efficiency, a functional form for the stochastic production frontier is chosen. Ideally, the functional form should be flexible, computationally straightforward and more importantly it has to be self dual in order to allow the estimation of allocative efficiency. To satisfy these properties, most empirical studies widely use the translog function. Following Battesse and Coelli (1995), the translog specification is mathematically expressed as: n ln(qj) = 0 + 1 n n 1 iln(xij) + ij ln(xi )ln(x j ) +v j - u j i=1 2i=1 j=i+1 where qj is the value of crop output, xij are the inputs, 0....ij are the parameters to be estimated, vj is a two-sided random error and is assumed to be identically and independently distributed with zero mean and constant variance and is independent of the one-sided error, uj. Economic efficiency is estimated from a cost frontier given the duality assumption of the production function specified in equation 1. Following the Shephard's lemma, the system of minimum cost input demand equations can be obtained by differentiating the cost frontier with respect to the input prices (Bravo-Ureta and Pinheiro 1997). From the input demand equations, we can recover the economically efficient input quantities, given respective input prices and the quantity of output. Thus economic efficiency estimates are obtained from estimating a dual cost frontier to the production technology. And as indicated by Farrell (1957), economic efficiency is the product of technical and allocative efficiency, thus allocative efficiency is obtained by 248 dividing the economic by the technical efficiency measures for each farmer. We also control for the value of crop output in the estimation of allocative efficiency and the average labour wage. Description of the data The data that have been used in the estimation of efficiency are presented in Table 1 which shows aggregate productivity and factor endowment. In the estimation of efficiency, the farmers' crop technology represented by the value of hybrid maize and burley tobacco is expressed as a function of the key factors: land, labour, fertilizer and seed. For the sake of ensuring consistency and controlling bias in the estimation of yields and efficiency, we analyse data from monocrop plots separate from intercrop plots. This is important because research evidence indicates marked differences in terms of crop yields as a result of the cropping pattern. The extent of yield differences between monocropped and intercropped plots largely depends on different levels and intensity of management as well as whether the relationship between crops grown in an intercrop is complementary, competitive or non-competitive. In general, monocropping is associated with higher intensity of management and the crop does not face competition. Complementary relationship occurs in the case of maize/leguminous intercropping where leguminous crops enhance soil fertility through biological nitrogen fixation. In cases where non-leguminous or non-symbiotic crops are grown together in an intercrop, there is likely to be competition for nutrients with the result that the yield potential for one or both crops is compromised (see Kanyama-Phiri et al. 2000). We therefore present the description of the production factors by cropping pattern. Land Land is measured in hectares and we consider the actual cultivated area per crop as opposed to total land holding size, because the later is likely to underestimate the levels of efficiency. Land endowment is much higher among farmers that grow hybrid maize and burley tobacco compared to those that grow local/composite maize and other tobacco types. Labour In the case of labour, we use the amount of family labour, measured in person days devoted to agricultural work per year. There are no marked differences in terms of total family labour supply among households that grow local and hybrid maize. However, average total family labour for households that grow burley is much higher. Labour intensity is likely to be higher in burley tobacco in particular, because of the many activities that have to be undertaken from nursery management, transplanting of seedlings, weeding, de-suckering, harvesting, drying, tying, sorting/grading, baling and marketing. Due to the low and highly skewed levels of hired labour, we have lumped it together with family labour in the estimation of efficiency. The average returns to labour are as expected higher in hybrid maize and burley tobacco due to the relatively higher yields from these crops compared to local/composite maize. 249 Improved seed and fertilizer We also control for the value of seed, and as expected seed cost for hybrid maize is on average much higher than that of local maize, because local or composite maize seed is often recycled, and therefore costless, but also the cost of composite seed is cheaper than that of hybrid maize seed. As expected, fertilizer intensity is higher in hybrid maize and tobacco than the local varieties. This is likely attributed to the higher productivity (and profitability) of the improved crop technologies which enables farmers to afford higher levels of fertilizer. Also due to risk-aversion, most farmers tend to apply sub-optimal levels of fertilizer and other inputs to local maize to avoid high losses in the event of adverse weather conditions. Table 1: Aggregate maize productivity measures and factor endowment by crop technology Productivity and factor Hybrid maize Local maize Burley tobacco endowment measures Monocrop Intercrop Monocrop Intercrop Monocrop Intercrop (n=1992) (n=3680) (n=1703) (n=4498) (n=932) (n=380) Crop yield (kg/ha)58 925.3 692.1 639.0 539.4 872.0 853.2 (799.2) (884.0) (571.9) (504.5) (594.5) (517.2) crop value (Kwacha/ha) 11,538 9,143 3,752 3,082 32,743 23,957 (43,564) (35,933) (4,078) (3,772) (113,171) (40,487) Family labour intensity59 475.1 455.9 454.1 468.6 501.3 434.8 (person days/ha/annum) (487.2) (429.7) (428.8) (421.0) (385.3) (324.9) Hired labour intensity in 5.2 4.5 3.0 2.8 15.7 10.8 (mandays/ season) (14.5) (13.8) (12.0) (7.7) (136.2) (44.3) Fertilizer intensity (kg/ha) 65.4 51.7 35.1 23.5 108.5 94.6 (118.8) (69.2) (25.5) (26.1) (174.0) (115.2) Family labour productivity 59 51 14 10 301 147 (K/ person day/ha) (171) (148) (40) (27) (1,759) (389) Maize seed cost 530 307 106 153 (Kwacha/ha) (469) (383) (269) (287) Rain-fed owned land (ha) 0.93 0.88 0.96 0.89 1.40 1.43 (0.85) (0.80) (0.83) (0.69) (1.14) (1.04) Total cultivated land (ha) 1.12 1.04 1.11 1.01 1.64 1.68 (0.96) (0.90) (0.98) (0.74) (1.28) (1.15) Note: Figures in parentheses are standard deviations. Kwacha is the local currency (1US$ = K122.00 as of January 2006) These descriptive figures indicate that some households grew both local and hybrid maize on separate plots. 58In the calculation of crop yields, all yields that are greater than the 95% percentile are replaced by the median yields. 59Includes total family labor supply for agricultural activities per year 250 The efficiency results are derived from the estimated frontier models as explained in Table 2. Table 2: Mean technical, allocative and economic efficiency of Malawian smallholder farmers Average efficiency (%) Min Max Monocrop Intercrop Monocrop Intercrop Monocrop Intercrop Hybrid maize TE 48.7 47.7 34.5 34.6 88.4 83.5 (7.0) (6.5) AE 41.0 40.0 27.1 27.2 80.8 75.4 (7.0) (6.5) EE 20.4 19.5 9.4 9.4 71.5 63.0 (6.9) (6.2) Local/composite maize TE 47.1 46.8 40.6 40.5 60.4 57.2 (2.6) (2.3) AE 30.9 30.6 26.1 25.9 40.4 38.0 (1.9) (1.6) EE 14.6 14.4 10.6 10.5 24.4 21.7 (1.8) (1.5) Burley tobacco TE 63.4 61.7 47.8 45.8 95.9 93.9 (7.2) (7.5) AE 55.2 51.0 35.5 31.0 87.8 83.3 (7.1) (7.4) EE 35.5 29.9 16.9 10.9 84.2 78.2 (9.3) (9.7) Figures in parentheses are standard deviations 251 Figure 1: Distribution of the efficiency indicators for hybrid maize and burley Tobacco (based on the monocrop samples) 300 200 y enc equfr 100 0 20 40 60 80 100 technical efficiency for hybrid maize (%) 0 25 200 ycn 150 ue eqrf 100 50 0 30 40 50 60 70 80 allocative efficiency for hybrid maize(%) 400 300 y 200 equencrf 100 0 0 20 40 60 80 economic efficiency for hybrid maize(%) 252 200 0 15 yc quenerf 0 10 50 0 50 60 70 80 90 100 technical efficiency for burley tobacco (%) 250 0 20 0 y 15 enc equfr 0 10 50 0 40 50 60 70 80 90 allocative efficiency for burley tobacco(%) 0 25 200 0 y 15 nc ue eqrf 0 10 50 0 20 40 60 80 economic efficiency for burley tobacco(%) 253 0 40 0 30 y nc 0 queerf 20 100 0 40 45 50 55 60 technical efficiency for local maize (%) 400 300 yc 0 20 equenrf 100 0 10 15 20 25 economic efficiency for local maize(%) 500 400 0 yc 30 equenrf 200 100 0 25 30 35 40 allocative efficiency for local maize(%) 254 The Determinants of Technical Efficiency in Hybrid Maize and Burley Tobacco Production: Factors that determine the levels of smallholder production efficiency From a policy perspective the most interesting aspect is to know the factors that affect efficiency. We then specify the one-sided technical efficiency effect as being related to the exogenous factors that influence crop production: uj = f (z) 2 where z is a vector of determinants of efficiency, such as household socio-economic characteristics, policy and institutional variables, and asset endowment as highlighted in the description of the explanatory variables in the crop technology adoption analysis conducted as part of this study. In the estimation of the translog function, we impose some key regularity conditions such as monotonicity, diminishing marginal productivity for all inputs, and quasi-concavity in order to ensure consistency of the results to economic regularity.60 The estimation for the efficiency model is conducted in STATA's frontier model. The estimation of equation 2 is important because this study not only aims at providing empirical measures of different efficiency indices but also aims at identifying key correlates to these efficiency indices that have to be addressed through policy interventions. The analysis of the main factors that influence efficiency is obtained from the estimation of the stochastic frontier (equation 2 above). The results are presented in Table 3. Table 3: Determinants of technical efficiency among Malawian smallholder farmers (OLS estimates based on the monocrop sample for both hybrid maize and burley tobacco) Dep variable: technical efficiency (%) Hybrid maize Burley tobacco Coefficient Robust Coefficient Robust Variables and measurement standard standard error error Socio-economic factors Household head sex (1=M; 0=F) 0.037 0.023 0.351 0.510 Age of household head (years) 0.003 0.005 -0.004 0.009 Highest level of education in the household 0.107 0.030*** 0.018 0.047 (years) Factor endowment and assets Dimba plot size (ha) 0.0560 0.008*** 0.016 0.009 Dimba plot size squared (ha) -0.060 0.321 -0.064 0.084 60Monotonicity requires positive marginal products with respect to all inputs and thus non-negative elasticities. In the translog form, this is obtained by multiplying the logarithmic marginal product with the average product of the input. Diminishing marginal productivity implies negative second order partial derivatives with respect to all input. Quasi-concavity condition implies a convex input requirement set and this is satisfied if the Hessian matrix is negative semi-definite. The third condition has only been imposed locally rather than globally. 255 Total land holding (ha) -0.072 0.041*** 0.079 0.022*** Total landholding squared (ha) -0.047 0.008*** -0.045 0.025* Tropical livestock units (TLU) 0.342 0.106*** 0.075 0.014*** Household has a radio (dummy) 0.225 0.173*** 0.217 0.336 Household has a bicycle (dummy) 0.203 0.196* 0.314 0.323 Household has an oxcart (dummy) 0.188 0.064* 0.205 0.078** Household has a w/barrow 0.304 0.570 0.438 1.041 Household has a sprayer 0.123 0.102 0.718 1.486 Quality index of dwelling unit 0.243 0.081*** 0.385 0.139** Total non-farm income (`000 K) 0.087 0.002*** 0.041 0.001** Access to institutional amenities Distance to banking facility (km) -0.006 0.005* -0.001 0.008 Availability of banking facility (dummy) 0.096 0.046* 0.070 0.069 Agricultural credit (dummy) 0.118 0.064* 0.123 0.062* Credit sources (private money lenders dummy) -0.023 0.011** -0.153 0.092* Credit source (Finance inst. dummy) 0.092 0.055*** 0.175 0.646* Farmer /credit club membership 0.051 0.005*** 0.118 0.052* Extension worker resident in community 0.075 0.021** 0.038 0.376 Extension message useful 0.147 0.027*** 0.139 0.047*** Cumulative hybrid maize adopters in 0.010 0.003** -0.017 0.005*** community (%) Cumulative burley tobacco adopters in 0.006 0.003** 0.005 0.006 community (%) Labour and dependency Household size 0.112 0.054** 0.200 0.125* Dependency ratio -0.176 0.125 -0.152 0.215 Infrastructure and access to markets Distance to the Boma (km) -0.019 0.004* -0.019 0.008*** Availability of daily market 0.042 0.022** 0.290 0.503 Availability of weekly market 0.044 0.221 0.071 0.398 Distance to daily market (km) -0.043 0.015*** -0.037 0.022* Distance to weekly market (km) -0.021 0.018 -0.071 0.035* Distance to ADMARC market (km) -0.040 0.017* -0.033 0.034 Access indicator (time) -0.003 0.003 -0.006 0.007 Road type (1=tarmac; 0=others) 0.103 0.027*** 0.050 0.690 Distance to main road (km) -0.010 0.004* -0.002 0.008 Land tenure, soil quality and husbandry practices Did household use purchased seed 0.093 0.017*** Did household apply manure -0.213 0.284 -0.054 0.046 Land tenure: customary 0.453 0.347 -0.287 0.074*** Land tenure: leasehold 0.267 0.738 0.170 0.120 Low cation exchange capacity (cec) dummy -0.042 0.013** -0.135 0.131 Low nitrogen (N) dummy -0.079 0.032*** -0.158 0.068** Low phosphorus (P) dummy 0.022 0.020 0.105 0.063 Availability of irrigation scheme 0.046 0.024* 0.126 0.483 Availability of farmers' cooperative 0.076 0.023** 0.162 0.319 Returns and profitability/ shocks Maize/tobacco: urea price ratio 0.264 0.098** 0.312 0.102** Did the household receive free fertilizer and -0.089 0.019*** -0.266 0.033*** seed Water requirement index 2004 (WRSI2004)61 0.033 0.005** 0.055 0.108 61WRSI is the ratio of the seasonal evapotranspiration (ET) to the crop water requirement. It provides an indication of a specific crop's potential performance based on the availability of water during its growing season. It is derived from the data on rainfall, evapotranspiration, amount of water that the soil is able to hold and other crop specific information (FEWS NET 2003). 256 Water requirement index_standard deviation -0.278 0.198 -0.596 0.382* Did the household experience an agricultural -0.226 0.183 -0.604 0.341* shock in the last 5 years No. of observations: 1992 932 Adjusted R2 value: 0.701 0.634 F-statistic 89.94*** 29.41*** Note: * P<0.10; ** P<0.005; *** P<0.001 For the dummy variables, the marginal effect shows the discrete change from 0 to 1. 257 ANNEX 7D: WEATHER-BASED INSURANCE This annex provides a description of weather based insurance instruments. For details the reader is referred to Hess and Syroka (2005), available on line at http://siteresources.worldbank.org/INTARD/Resources/Malawi_text_final.pdf. The aim of weather-indexed insurance is to define a weather index that most accurately represents the impact of weather on a farmer's physical crop production. The objective is then to design an insurance contract that compensates the farmer when these adverse weather conditions occur. Weather based insurance instruments thus provide financial protection based on the performance of a specified rainfall-index in relation to an agreed trigger, and offers protection against uncertain revenues that result from volume volatility. Buyers are compensated against unfavorable weather fluctuations that impact physical volumes produced, thus a farmer or a group of farmers could buy such a product. Because weather-indexed insurance products can be reinsured in the global weather-risk market, the risk is transferred from Malawi to the international reinsurance and capital markets. New risks and locations provide diversification and hence enhanced risk/return characteristics for their portfolios, and weather market players, from both the reinsurance and financial communities, are interested in these new developing country transactions. This ultimately leads to more aggressive pricing, market liquidity attracting new market players and thus to business growth and expansion through broadening product offerings and increasing global networks. The critical periods, when maize is most vulnerable to low rainfall and therefore water stress, are the emergence period immediately after sowing and the tasseling period ( Figure ). On the basis of farmer interviews, agro-meteorological studies and models such as the FAO water satisfaction index, a specific maize rainfall index has been developed. The maize rainfall index is defined as a weighted sum of cumulative rainfall during the 130 growing period of maize, with individual weights assigned to specific phases of the crop's evolution. That is, the index gives more weight to the more critical periods when maize is most vulnerable to rainfall variability. The individual weights are determined by maize water requirements, as advised by the FAO, for each of the following stages of growth and development: Table 1: Weights for Maize Growing Phases Phase Length of Phase Relative Weight Sowing/Establishment 20 days 1.75% Vegetative Growth 30 days 1.75%-13% Flowering (Tasseling & Silk) 20 days 13% Yield Formation 40 days 13% Ripening 20 days 1% Sowing is determined to take place during the first decade, 10-day period, after 1st October when rainfall is greater than 30mm. This level of rainfall indicates the commencement of 258 rains for the farmers in the region and hence, given the long growing cycle of maize, the optimal time of sowing. For illustrative purposes we identify the Lilongwe area as an important maize-producing region of Malawi. Rainfall is a very accurate proxy for measuring the maize production variability for farmers in the Lilongwe and Salima regions. Error! Reference source not found. demonstrates that the correlation between this rainfall index and actual yields in the Lilongwe area and the nearby Salima region is very high. The correlation between the inter-annual variations in the index and the inter-annual variations in both the hybrid and local maize yields in Lilongwe and Salima is 60%, significant at the 99% confidence level. It is a commonly held belief in Malawi that February is the most critical rainfall month for the country's maize crop. However, Error! Reference source not found. shows that simple cumulative February rainfall measured at the Lilongwe weather stations does not correspond well with the yields in the region. The correlation between the inter-annual variations in the index and the inter-annual variations in the hybrid and local maize yields in Lilongwe and Salima are 24% and 50% respectively, not significant at the 99% confidence level. This indicates that a more sophisticated underlying weather indicator, such as weighted rainfall index outlined above, capturing sowing dates and critical maize periods, is more appropriate for insuring farmers' maize crops. The next step in constructing an insurance policy is to establish the financial impact of adverse weather events for the farmer, in other words to find the farmer's weather exposure in terms of MKW per unit of the defined maize rainfall index. The first step is to find the yield fluctuation of maize per 1mm of the defined maize rainfall index. This can be derived through a regression analysis using the historical yield and weather data; the overall objective is to minimize the mismatch between payouts triggered by the rainfall index and the actual yield based on past data. A simple linear regression using data from 1990-2002 indicates 1mm of the defined rainfall index corresponds to a 20 kg/ha fluctuation in hybrid maize yield and a 10 kg/ha fluctuation in local maize yield (not shown). The yield fluctuation per mm of the index can then be converted into MKW per mm by either assessing the input costs or the expected sales margin of the farmer. For example for local maize, if 1 mm of the rainfall index corresponds to 10 kg/ha of yield (using data from the regression analysis above) and is worth 15 MKW per kg to adequately cover production costs, then the unit exposure of Lilongwe rainfall index is 150 MKW per mm for the farmer if he is seeking to cover input and production costs. In this simple example a farmer could buy a stand-alone insurance product with a straight up-front premium to be paid before the protection period. Perhaps lenders such as the Malawi Rural Finance Company (MRFC) could finance this premium as the farmers may not have the cash to pay before harvest. The average maize rainfall index for Lilongwe weather station is 70mm (Error! Reference source not found.). A farmer could purchase a weather-indexed insurance contract with a trigger level of 50mm, i.e. a contract that will compensate the farmer if the maize rainfall index for the growing season is recorded to be less than 50mm. From Error! Reference source not found. we can see that 50mm corresponds to approximately a 1500 kg/hct yield for hybrid maize and a 650 kg/hct yield 259 for local maize, and thus farmers would in effect have protection against situations where yields would drop below these levels. An example of the payout structure of such an insurance contract is shown in Figure . The maximum compensation a farmer could receive in a worst-case drought scenario is 4000 MKW to cover his production costs (Figure ). The historical payouts of such a contract are shown in Figure . It shows that a farmer would have received payouts to compensate for drought-related drops in maize production in 1992, 1994 and 2001 ­ low yield years in the Lilongwe and Salima regions. Clearly the retention or trigger level, the deductible in insurance parlance, is key to pricing. In the example above the farmer retains the first 30% of risk himself before the protection begins. Weather-indexed insurance contracts are priced using an actuarially fair assessment of the risk an insurance company selling the contract takes. Hence a long and gap-free historical record of rainfall data is essential in order for the insurer to be able to price the risk and hence charge an appropriate premium to the farmer. The expected loss for the insurer is determined by the average payout of the contract using historical rainfall data. In the case of this Lilongwe based structure the 30-yr average payout is 340 MKW (Figure ). However a commercially minded insurer will obviously charge more than this and will want to be compensated for the possibility of having to pay 4000 MKW to the farmer in a severe drought year. The standard deviation, a measure of volatility of payouts and hence the risk for the seller of such a product, is 760 MKW. Therefore a very preliminary indication of the premium an insurance company could charge is 340 + 25%*760 = 530 MKW62. This corresponds to an insurance premium rate of 13%, i.e. a premium of 530 MKW for 4000 MKW of compensation (Figure 4). The final step in designing a weather-indexed insurance program is to find a national insurance company to intermediate the insurance contract. This might be required by national insurance regulations. Otherwise one needs to weigh costs and benefits of local intermediation versus a direct international contract. If a local intermediary is used, the risk needs to be reinsured ­ this is key for the national insurer who cannot retain much systemic risk. The risk transfer structure illustrated below assumes that the farmer buys the product through a national insurance company, which is not necessarily the case. In practice farmer aggregators such as NASFAM or MRFC will act as agents for the distribution of the product. Reinsurers in the global weather risk market, where the actors are both insurers and banks, are interested in this type of risk. It provides diversification to their books through new locations and risks, this leads to enhanced risk/return portfolio characteristics which ultimately lead to more competitive pricing. Hence these new developing country transactions develop weather market liquidity thus attracting new market players. In addition, for reinsurers, these transactions lead to business growth and expansion through 62This formula as well as the number of 25% is a rule of thumb that varies between risk takers. In essence the premium calculation involves an actuarial calculation asking for the expected payout of the policy based on past data, and a measure of the magnitude of these expected payouts. If past payouts are very volatile the premium increases as sellers will want to be compensated for taking on more uncertain risk. The 25% measures the risk aversion of the seller. 260 broadening product offerings and increasing global networks. The reinsurability of weather-indexed insurance products is another benefit of these types of instruments. Example of an innovative instrument to provide credit to smallholders: the Lilongwe weather indexed maize production loan A financial/ insurance company could package a loan-plus-insurance based on the Lilongwe maize farmer index described above into one product, the weather indexed maize production loan.63 The farmer would enter into a loan agreement with a higher interest rate that accounts for the weather insurance premium that MRFC pays to the insurer.64 For a loan of MKW 6000, normal interest rate is 2.9% per month. Assuming a tenor of six months, MRFC would charge another 1.5% per month for the weather risk protection, totaling 4.4% per month. In return the farmer does not repay all its dues in case of a drought. In case of a severe drought with the rainfall index at 20 mm or below, instead of paying MK 7000 (principal and interest), the borrower pays only MK 3500, or half the usual dues. In case of a severe drought impacting maize yields, therefore, the farmer would pay only a fraction of the initial loan borrowed, and would therefore be less likely to default, strengthening the banks portfolio and risk profile. Historical simulations of such a product in Malawi (Figure ) demonstrate that the years of reduced loan-dues payments coincide with the drought years where farmers suffered from much lower yields, mainly the years 1992 and 1994. The assumption is that this type of risk transfer makes defaults more unlikely for three reasons. Firstly, the farmer simply pays according to his repayment ability, which is severely reduced in times of drought crises. Secondly, a strategic, that is willful default, becomes more unlikely as the weather-indexed loan structure distinguishes between systemic weather risk and idiosyncratic farmer risk. In other words, farmers cannot use the weather hazard as an excuse for not repaying their loans. Strategic default of joint liability groups of smallholder farmers becomes more unlikely. Thirdly, the insurance company is able to continue to lend to farmers throughout crises periods, without painful rescheduling or even moratoriums that are inevitably associated with smallholder loans in times of crises. Instead of defaulting on the whole loan, farmers would pay half their loan dues in a severe crisis, and the insurance company collects on the insurance policy and the borrower can maintain their credit-worthy record and continue to borrow in the following season. Basis Risk ­ or: how good is this insurance? A major concern with insurance based on weather or other indexes is basis risk-- the potential mismatch between insurance payouts and farmers' losses. Jerry Skees writes that "[t]he effectiveness of index insurance as a risk management tool depends on how positively correlated farm-yield losses are with the underlying area yield or weather 63For detailed elaboration of this product, see Ulrich Hess, Innovative Financial Services for India, Monsoon Indexed lending and Insurance for smallholders, ARD Working Paper 9, 2004. 64The insurance/financial institution could pass on only part of the premium, since it would reduce the risk premium it used to charge the farmer through relatively high interest rates. 261 index."65 This concern relates to the question of whether insurance based on a weather index can substitute for traditional crop insurance and indemnify the farmer for his losses. The usual answer is that basis risk can be managed if: 1. The correlation between index and yields is high and the index is measured well; and 2. Efficiency gains with index insurance allow for lower deductibles, which partially compensate for the basis risk. An example of basis risk is seen in Error! Reference source not found. - it is clear the maize rainfall index indicates a lower expected maize yield than was actually experienced in the Lilongwe and Salima regions. The basis risk is to the farmer's advantage in this particular example. The experience of the Commodity Risk Management Group at the World Bank shows that the relevant question is whether the payout from insurance based on a weather index effectively reduces the insured's value-at-risk (VAR) rather than compensating for a single crop loss only. Value-at-risk is a measure of potential dollar loss from an adverse change in prices occurring in a normal market environment. The farmer's value-at-risk is an effective measure of his overall vulnerability, his exposure to income shocks--such as a wedding, a disease, or a big drought. The farmer is interested in maximizing his overall income while minimizing his value-at-risk. Income comes from multiple sources--such as, off-farm labor, livestock, as well as field and perennial crops. As stated earlier, diversifying income sources is clearly a way of managing risks and minimizing VAR by sacrificing some of the benefits that could come with specialization and economies of scale. Basis risk, or in other words the effectiveness of the insurance, is always an issue to be considered when dealing with index-based risk management solutions. It is important to note that not all food security issues are caused by weather-risk: civil strife, poor farm- management and inadequate seed and fertilizer supplies may as important as weather in triggering food emergency situations. Although unlikely, given the extreme nature of the risk being considered for SADC in the proceeding sections of this paper, there is also the possibility that simple weather indices may not fully capture the financial impact of a weather event. Hence the need for World Bank and donor assistance still remains within the SADC region. The 2005 smallholder groundnut production pilot In 2005, 900 groundnut farmers in Malawi bought weather insurance to increase their ability to manage drought risk and in turn access credit for better inputs. National Smallholder Farmer Association of Malawi, in conjunction with the Insurance Association of Malawi and with technical assistance from the World Bank, designed an index-based weather insurance contract that would payout if the rainfall needed for groundnut production in four pilot areas was insufficient for groundnut production. Because these weather contracts could mitigate the weather risk associated with lending to farmers, Opportunity International Bank of Malawi and Malawi Rural Finance Corporation agreed to lend farmers the money necessary to purchase higher-yielding certified groundnut seed if the farmers 65 "Risk Management Challenges in Rural Financial Markets: Blending Risk Management Innovations with Rural Insurance" by Jerry Skees, Prepared for presentation at: Paving the Way Forward for Rural Finance: An International Conference on Best Practices June 2 ­ 4, 2003 Washington DC. 262 bought weather insurance as part of the loan package. Given the success of last year's pilot, the aim is to target a maximum of 10,000 farmers through weather insurance-linked cash crop production loans in 2006. Figure 1: Maize crop calendar *Maize yields are particularly sensitive to rainfall during the tasseling stage and the yield formation stage ­ rainfall during the latter phase determines the size of the maize grain 2% 2% 2% 2% 13% 13% 13% 13% 13% 13% 13% 1% 1% x Cumulative Rainfall in each decade = Maize Rainfall Index Weights and diagram taken from the FAO's maize water requirement report* 263 Figure 2: Maize yield Lilongwe Salima vs. maize rainfall index 264 Figure 3: Payout structure of maize rainfall index (net of premium) Payout per Hectare for Maize Drought Protection, Lilongwe Region 4000 3500 3000 2500 hct) 2000 per 1500 (MKW 1000 Payout 500 0 0 10 20 30 40 50 60 70 -500 -1000 Maize Rainfall Index 265 Figure 4: Historical payouts of drought protection cover Historical Payouts of Drought Protection Cover 4000 3500 3500 3000 3000 t) t) 2500 hcr g/hc 2500 (k pe 2000 dl KW 2000 Yie (M 1500 izea outy 1500 M Pa brid 1000 1000 Hy Payouts per Hectare Insured 500 500 Hybrid Yield (kg/hct) 0 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Year Indicative Pricing: 30-yr Average Payout = 340 MKW , Stdev = 760 MKW Premium = 340 + 25%*760 = 530 MKW (13% ROL) Figure 5: Historical payouts of rainfall indexed loan Historical loan repayments for drought indexed seasonal loanover 8000 3500 7000 3000 6000 2500 ) 5000 (kg/hct) 2000 (MKW 4000 Yield dues 1500 Maize loan 3000 Interest Principal 1000 Hybrid 2000 Hybrid Yield (kg/hct) 500 1000 0 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Year 266 ANNEX 8A: DETERMINANTS OF COMMERCIALIZATION (AND ADOPTION OF TOBACCO PRODUCTION) Regression Analysis In this section, we undertake the empirical analysis of the determinants of agriculture commercialization in Malawi. Broadly, we want to set up a regression model that includes two sets of regressors: household controls and district controls. The model takes the following form, (1) shc = z'hcs +sIc + hc s where shc is the share of land allocated to tobacco by household h in community c. This is our measure of commercialization or export participation. This formulation is in line with our aim of trying to identify the role of individual determinants vis-a-vis village determinants in agriculture commercialization. We consider 6 different models: Model (1) includes only essential household demographic variables such as household size, demographic composition, and educational attainments of the head. Model (2) includes also district level characteristics. Model (3) is equal to Model (2) but adds district dummies to control for district fixed effects. Model (4) includes a dummy that equals 1if the farmer belongs to a tobacco club. In Model (5) we include an alternative definition of access to transport along with variables to capture assets' possessions. Lastly, Model (6) replaces the tobacco club dummy with dummies of different benefits the farmer obtains from the tobacco club if he belongs to one. Since the dependent variable, the land shares, are censored at zero, the model has to be estimated with a Tobit model. For the sake of exposition, we begin with OLS models. Results for the 6 models are presented in Table 5. In general, there are only minor differences between the models and the results are robust across specifications. We first discuss the household level determinants of export crops farming. From the results we find that household characteristics do matter. For instance, if the head of the household is male, that has a positive effect on the share of land allocated to tobacco. Also the marital status of the head is important: the tobacco share of land is greater when the head is married. Household size has a positive effect even though it is weak. This result goes in line with the hypothesis that larger households have a larger labor supply and thus face lower constrains to commercialization. On the other hand, it could be argued that larger households will try and secure food needs before engaging into commercialization. But, do note that we need to take into account the household structure as well.66 The way to interpret the household age composition coefficients is the following. Each variable (prop. age 0-7, prop. age 8-12, etc) represents the proportion of total members in that age bracket. In the regression, the omitted category is the proportion of 66For example, in terms of labor supply and food security a household with 8 members of which 6 are 8 years old or less is completely different than a household with 8 members of which all are over 18 years old. 267 members between 18 and 45 years old. Therefore if, for example, the coefficient of the variable proportion age 0-7 is negative (positive) this should be interpreted as that raising the share of members between 0 and 7 years old, and at the same time decreasing the share of members in the 18-45 category by the same amount, remaining the rest of the shares constant, will decrease (increase) the share of land devoted to tobacco. The results are as expected. Substituting adults between 18 and 45 years old with either younger or older members has a negative effect on the share of land. Two forces are in play in the same direction. Having more children raises the food security concern and at the same time restricts the available labor force, therefore, limiting commercialization activities. Another component of the household structure is the proportion of males. In this case, there seems to be no effect on export activities to a change in the proportion of males after controlling by household size and age composition. This may indicate that women are also engaged in the production of tobacco. There is no clear association between commercialization and the education or literacy status of the head. Only in Model (1) those heads with primary or junior secondary studies completed seem to have a larger area of tobacco planted than those with no education. One possible explanation is that education is extremely low across the population, so that there is no sufficient variability to capture the effects of education. It is surprising to find that the health status of the household (measured as the proportion of sick members) does not seem to affect the decision to participate in export activities. This is also true for the health status of the household head. This evidence does not support the notion that sick members, including the head, reduce effective labor, or that sick children compromise the effort of the women in agriculture. The reason may be that these variables can not distinguish between chronic and short-term sickness. Possession of household assets like beds, table and chairs, fan, radio, etc. has a positive effect. Also the possession of agricultural tools, like a hoe or sickle, has a positive effect. 268 Table 5 Determinants of Participation in Tobacco: land share (OLS results) (1) (2) (3) (4) (5) (6) constant -2.37** -3.18** 6.63** 5.57** -3.74** -4.09** rural 3.23*** 1.25 3.39*** 3.04*** 2.29** 2.04** household size 0.33*** 0.27*** 0.11 0.02 -0.14* -0.14* head male 2.03*** 1.82*** 1.91*** 1.89*** 1.50*** 1.48** head married 1.24** 1.34** 1.09** 0.93* 0.74 0.76 prop. males 0.19 -0.26 -0.32 -0.36 -0.61 -0.42 prop. age 0-7 -0.11 -0.09 0.40 0.38 1.10 1.41 prop. age 8-12 -2.47** -2.36** -1.45 -1.49 -1.07 -0.89 prop. age 12-18 -1.46 -1.47 -1.20 -0.82 -0.68 -0.66 prop. age 46+ -1.56** -1.85*** -1.99*** -2.02*** -2.50*** -2.42*** prop. sick 1.39* 0.97 0.84 0.89 0.89 0.78 head sick -0.83** -0.65* -0.55 -0.56 -0.60* -0.56 head literacy (chichewa) 0.48 0.53 0.34 0.07 -0.09 -0.10 head literacy (english) -0.88 -0.48 0.01 -0.50 -0.60 -0.51 head educ.: attends school -1.66 -1.92 -1.19 -0.77 -0.46 -0.36 head educ.: primary completed 1.30* 0.99 -0.04 0.01 0.02 -0.05 head educ.: secondary (jr.) completed 1.58** 1.03 -0.06 0.39 0.33 0.32 head educ.: secondary (sr.) completed -0.88 -0.72 -1.30 -0.82 -0.89 -1.17 head educ.: superior completed 1.85 2.99 2.19 2.54 1.36 1.32 land area 0.00** 0.00 large weekly market 1.28* 0.89 1.02** 0.99* 0.94* permanent ADMARC market -1.12 -0.39 -0.13 0.01 0.00 cooperative 5.64*** 2.62*** 2.15* 2.16*** 2.21** credit club 2.63*** 1.51** 1.42** 1.39** 1.23* post office -0.85 -1.72** -1.68** -1.90** -1.92** public telephone -1.40** -0.56 -0.25 -0.25 -0.32 health clinic -0.48 0.37 0.28 0.41 0.43 irrigation scheme -0.84 -0.71 -1.06 -0.86 -0.76 graded graveled road 2.83* 0.69 0.73 0.45 0.46 dirt road (maintained) 2.05* 1.26* 1.00 0.64 0.69 dirt track 1.18 0.91 0.78 0.43 0.44 dist. to urban center -0.00 -0.00 dist. to asphalt road 0.01 0.01 remoteness 0.82 0.90 household assets 0.43 0.43 bicycle 1.24*** 1.30*** other mean of transp. 1.78 1.91 oxcart 6.17*** 7.05*** wheelbarrow 1.31 1.24 hoe 0.60 0.43 sickle 1.13*** 1.11*** tobacco club member 17.46*** 16.23*** club benefit: credit 8.93*** club benefit: inputs at lower price -1.45 club benefit: better tobacco prices -0.20 club benefit: extension service 0.96 club benefit: quota access 6.23*** club benefit: transport to market 4.37** An interesting result is that the existence of food markets in the village is associated with higher export agriculture. This evidence supports the hypothesis of de Janvry et. al (1991). If food markets are thin, households will be reluctant to engage in commercialization and be at the mercy of potentially high prices in the presence of low food supply. In this case, farmers will choose to focus on food production for home consumption. Instead, if food markets are available, households may face, in the end, a lower risk of engaging into export crops. Also note that the presence of ADMARC (Agricultural Development and Marketing Corporation, originally a government marketing board) markets has no effect. In a study by Nthara (2002) one of the findings is that ADMARC does not play anymore a role as a 269 buyer of agricultural produce. The reasons are that the producers may get higher prices when selling to private traders and at local markets, also ADMARC starts buying from farmers very late in the year. Therefore, producers are forced to sell to other buyers when cash is in need. One last reason is that in some areas ADMARC tends to run out of cash. We turn now to the aggregate (district level) determinants. An interesting finding is that credit plays a significant role. The presence of a credit club at the village has a strong positive effect on commercialization activities. If we look further into Model (6) if the farmer is a member of a tobacco club and if he receives credit from the club the effect on the share of land allocated to tobacco is positive and quite large. This gives us a sense that not only having credit available, but having met the requirements to obtain a credit plays an important role. This finding supports the start-up investment hypothesis. The presence of a cooperative at the village also has a positive effect in export crops farming, but more attractive is the effect of tobacco clubs. If the farmer is a member of such a club the share of land devoted to tobacco increases significantly as shown in Model (5). More interesting is to disentangle the aggregated effect of tobacco clubs into the benefits it has to its members. This is reported in Model (6) where the tobacco club dummy is replaced by dummies of the different benefits the members receive from the club. As we mentioned earlier, credit plays a very important role in the decision of engaging into export oriented crops. There are also another two important factors. A tobacco grower produces tobacco to sell it in order to obtain cash to later be able to buy food and other goods. In this sense, the availability of traders or produce markets to where the farmer can sell its produce is a crucial condition. This is why the quota access to auction floors has a big positive effect in the decision of growing tobacco. The district road variables included in the regressions seem to have little impact on our measure of commercialization when controlling by the rest of the covariates. This is true for two different measures of road infrastructure: dummies for different types of roads (Models (2) to (4)), and distance to asphalt road in Km (Models(5) and (6)). This does not mean that roads are not important. To see why, notice first that some of our measures of road access are dummies indicating the main type of road in the district. The omitted category in our regressions is asphalt roads; thus, the right interpretation of the sign and significance of the road dummies is that the different qualities of roads have no further effect on participation. We experiment with a set of different road access specifications later (see below). Second, even though we found no effect of different road access and distance on the share of land allocated to tobacco, we do find that access to transport of bales to the market is linked to higher participation in export agriculture. This is an important result because it suggests that the presence of intermediaries linking farmers with markets facilitates participation. In other words, these services create markets for commercializable crops and act as impulses towards participation. This conclusion is consistent with our findings regarding market costs in export agriculture reported in Balat and Porto (2005). Third, notice also that if the household has a bicycle or an oxcart this too has a positive effect in participation. These two means of transport are used by tobacco farmers to carry their produce to the selling point. This indicates that even though we don't find a positive effect for different types of access to roads, we do find positive and significant effects of services related to transport. 270 In conclusion, the type of road access that the farmer have does not seem to be a great constraint to agriculture, provided there are roads (asphalt, graveled, or maintained) and provided farmers are endowed with tools to reach the selling points, have access to intermediaries or other marketing channels, and have transport facilities through, for instance, tobacco clubs. To acknowledge the fact that our independent variable, i.e., the share of land allocated to tobacco, is censored at zero (and also atone), thus turning OLS estimation inconsistent, we should estimate a Tobit model instead. We reestimated the six models using Tobit. Marginal effects results are presented in Table 6. For the sake of making the table simpler, we only report the unconditional marginal effects, i.e., dE[s|x]/dxi, except for models (5) an (6) where we also report the marginal effects conditional on being uncensored (i.e., dE[s|x,s>0]/dxi) in columns (5b) and (6b), respectively. All marginal effects are evaluated at the average values of z, and I. As can be seen from the results, the conclusions regarding participation derived from the OLS estimation remain unchanged. As expected, the sign of the effects are as before, as well as the statistical significance of the variables. However, the magnitudes are different. Since the share variable is censored, it is perhaps convenient to focus on the Tobit coefficients when interpreting the results and assessing the quantitative effect of selected variables. In this case, the distinction between the two marginal effects becomes more important. In what follows, we briefly elaborate on this. The marginal effects in Table 6 refer to i) the change in the average share of land allocated to tobacco for all Malawi farmers and ii) the change in the average share of land allocated to tobacco only for those already producing tobacco. One way to think about the difference in these marginal effects is as follows. The latter coefficient would indicate the response of tobacco production if we did not allow farmers to adopt tobacco; in contrast, the unconditional marginal effect would allow for supply responses from all farmers, including those that were not participating to begin with. Notice that, in general, the unconditional marginal effects dE[s|x]/dxi are much smaller than the conditional marginal effect dE[s|x,s>0]/dxi. This is because the unconditional effects average across a larger number of producers with zero shares. However, the total effect on the whole population, or the total supply response, will be much larger. This is because, in our sample, there are approximately only 10 percent of farmers that produce some tobacco. For example, becoming a tobacco club member would increase total land shares by 1.52 (if supply responses are allowed) but only by 0.53 if supply responses of non-producers are not allowed). Similar, having an oxcart would increase land shares by 0.196 (with supply responses) and by 0.154 (without supply responses).67 In tables 7 through 9 we reestimate the Tobit models under alternative specifications for the road access variables. In Table 7 we replace the type of road dummies and the distance to asphalt variable with an access indicator (in logs), which measures the travel time (in minutes) from each location to the closest market or trading center. Again, we find no effect of our measure of distance on the tobacco participation decision, but we still find that access to transport services and possession of assets to haul the produce (such as a bicycle or an oxcart), do have a positive effect. The rest of the results remain unchanged. 67As a rule of thumb, one can compare the unconditional marginal effects in columns 5a and 6a to 0.1 times the conditional marginal effects in columns 5b and 6b. 271 It is possible that the relationship between the share of land allocated to tobacco and distance is non-linear. Thus, to take this into account, we brake down the access index into categories of closeness to the trading center. Instead of the (log) of the access index, we include dummies for log access index from 0 to 2.5, 2.5 to 4.5, and greater than 4.5, and 0 to 1.5, 1.5 to 2.5, 2.5 to 3.5, 3.5 to 4.5, and greater than 4.5 in tables 8 and 9, respectively. In both tables the excluded category is the greater than 4.5 one. Once more, no effect of distance is found in neither of the tables under these specifications. To end, we try a different model for equation 3. Instead of land share devoted to tobacco we use as dependent variable the total plot area (in acres) allocated to tobacco. This variable is also censored at zero, but it is no more right censored. We replicate tables 6 to 9, and results are shown in tables 10 to 13, respectively. In general, the main conclusions persist, the magnitudes are different, of course, but the sign of the effects remain unchanged, as expected. Some variables are less statistically significant, in particular, the household characteristics. One reason may be that this new model answers a different question: we are no longer interested in how a household allocates its available land into cash or food crops, but we are now interested in the scale of the tobacco plot. Thus, assets or services available to the household may have greater effect than household composition and characteristics. 272 Table 6 Determinants of Participation in Tobacco: land share (Tobit marginal effects) (1) (2) (3) (4) (5a) (5b) (6a) (6b) constant -17.261***-14.776*** -1.306*** -1.203*** -1.158*** -13.220*** -1.045*** -12.467*** rural 2.827*** 1.133 0.131*** 0.120*** 0.091** 1.608** 0.082** 1.451** household size 0.404*** 0.285*** 0.017*** 0.010** 0.000 0.004 0.000 0.001 head male 1.586** 1.170*** 0.103*** 0.096*** 0.068** 0.880** 0.068** 0.936** head married 1.943*** 1.642*** 0.124*** 0.108*** 0.077*** 1.000*** 0.069*** 0.925*** prop. males 0.586 0.160 0.028 0.020 0.005 0.060 0.017 0.199 prop. age 0-7 -0.652 -0.534 -0.033 -0.040 -0.005 -0.056 0.018 0.218 prop. age 8-12 -2.270** -1.828** -0.123* -0.133** -0.087 -0.990 -0.066 -0.784 prop. age 12-18 -0.852 -0.763 -0.069 -0.045 -0.036 -0.414 -0.027 -0.326 prop. age 46+ -1.160* -1.193** -0.109** -0.110*** -0.123*** -1.400*** -0.111*** -1.327*** prop. sick 1.110 0.550 0.026 0.028 0.021 0.244 0.018 0.212 head sick -0.583* -0.288 -0.020 -0.018 -0.017 -0.194 -0.013 -0.164 head literacy (chichewa) 0.656** 0.579** 0.044** 0.024 0.011 0.124 0.009 0.114 head literacy (english) -0.476 -0.033 0.029 -0.006 -0.010 -0.114 -0.006 -0.078 head educ.: attends school -1.018 -0.692 -0.030 0.000 0.024 0.256 0.026 0.282 head educ.: primary completed 1.398*** 0.716** -0.004 0.003 -0.001 -0.016 -0.005 -0.059 head educ.: secondary (jr.) completed 1.422** 0.658 -0.022 0.004 0.003 0.030 0.004 0.048 head educ.: secondary (sr.) completed -1.214 -0.786 -0.079* -0.055 -0.050 -0.690 -0.053 -0.788 head educ.: superior completed 0.745 2.316 0.078 0.131 0.040 0.411 0.036 0.383 land area 0.000*** 0.001*** 0.000 0.000 large weekly market 1.199** 0.096** 0.106*** 0.077** 0.809** 0.070** 0.765** permanent ADMARC market -1.089* -0.067 -0.045 -0.033 -0.408 -0.031 -0.403 cooperative 5.025*** 0.193*** 0.154*** 0.124*** 1.173*** 0.120*** 1.178*** credit club 1.837*** 0.103** 0.093** 0.078** 0.792** 0.066** 0.708** post office -0.305 -0.084 -0.078 -0.067 -0.929 -0.064 -0.928 public telephone -1.598** -0.080 -0.058 -0.048 -0.613 -0.050 -0.684 health clinic -0.415 0.025 0.014 0.014 0.159 0.015 0.175 irrigation scheme -0.660 -0.032 -0.050 -0.040 -0.501 -0.034 -0.452 graded graveled road 4.327*** 0.141* 0.121 0.084 0.816 0.083 0.842 dirt road (maintained) 2.433*** 0.119* 0.090 0.062 0.694 0.063 0.731 dirt track 2.089* 0.132 0.106 0.079 0.787 0.081 0.837 dist. to urban center -0.000 -0.001 -0.000 -0.002 dist. to asphalt road 0.000 0.003 0.000 0.002 remoteness 0.047 0.486 0.053 0.555 household assets 0.041** 0.528** 0.040** 0.541** bicycle 0.082*** 0.871*** 0.083*** 0.917*** other mean of transp. 0.028 0.290 0.038 0.408 oxcart 0.156*** 1.265*** 0.196*** 1.541*** wheelbarrow 0.066 0.635 0.059 0.594 hoe 0.071 1.108 0.061 0.954 sickle 0.082*** 0.987*** 0.076*** 0.961*** tobacco club member 2.139*** 1.519*** 5.372*** club benefit: credit 0.269*** 1.938*** club benefit: inputs at lower price -0.030 -0.400 club benefit: better tobacco prices 0.018 0.198 club benefit: extension service 0.071 0.702 club benefit: quota access 0.123** 1.097** club benefit: transport to market 0.245*** 1.816*** 273 Table 7 Determinants of Participation in Tobacco: land share (Tobit marginal effects) (2) (3) (4) (5a) (5b) (6a) (6b) constant -14.032*** -0.814*** -0.820*** -1.160*** -13.007*** -1.053*** -12.224*** rural 1.086 0.141*** 0.128*** 0.095*** 1.642*** 0.085** 1.458** household size 0.283*** 0.018*** 0.011** 0.000 0.004 0.000 0.001 head male 1.269*** 0.114*** 0.105*** 0.072*** 0.915*** 0.073*** 0.972*** head married 1.629*** 0.131*** 0.113*** 0.077*** 0.984*** 0.069*** 0.908*** prop. males 0.148 0.026 0.017 0.004 0.040 0.016 0.182 prop. age 0-7 -0.620 -0.053 -0.055 -0.009 -0.101 0.014 0.165 prop. age 8-12 -1.807** -0.127* -0.133** -0.086 -0.964 -0.065 -0.752 prop. age 12-18 -0.728 -0.083 -0.052 -0.040 -0.450 -0.031 -0.357 prop. age 46+ -1.180** -0.126*** -0.120*** -0.125*** -1.406*** -0.115*** -1.340*** prop. sick 0.515 0.031 0.033 0.025 0.285 0.022 0.258 head sick -0.318 -0.022 -0.021 -0.019 -0.212 -0.016 -0.185 head literacy (chichewa) 0.566** 0.047** 0.024 0.009 0.106 0.008 0.089 head literacy (english) -0.012 0.031 -0.005 -0.012 -0.131 -0.008 -0.088 head educ.: attends school -0.974 -0.052 -0.023 0.009 0.093 0.013 0.145 head educ.: primary completed 0.668* -0.008 -0.001 -0.002 -0.021 -0.005 -0.061 head educ.: secondary (jr.) completed 0.672 -0.017 0.012 0.008 0.089 0.007 0.078 head educ.: secondary (sr.) completed -0.875 -0.090* -0.061 -0.052 -0.705 -0.055 -0.804 head educ.: superior completed 1.552 0.059 0.093 0.019 0.198 0.014 0.153 land area 0.000*** 0.001*** 0.000 0.000 large weekly market 1.162** 0.085* 0.096** 0.082** 0.844** 0.076** 0.806** permanent ADMARC market -0.949 -0.071 -0.045 -0.030 -0.357 -0.029 -0.364 cooperative 5.362*** 0.218*** 0.176*** 0.144*** 1.310*** 0.140*** 1.312*** credit club 2.030*** 0.099** 0.092** 0.079** 0.788** 0.068** 0.707** post office -0.327 -0.090 -0.080 -0.067 -0.903 -0.066 -0.919 public telephone -1.764*** -0.098 -0.065 -0.052 -0.654 -0.056 -0.752 health clinic -0.437 0.022 0.010 0.014 0.156 0.016 0.183 irrigation scheme -0.696 -0.031 -0.054 -0.036 -0.448 -0.033 -0.419 lnaccess 0.347 -0.002 0.010 0.013 0.151 0.012 0.135 household assets 0.044** 0.550** 0.043** 0.563** bicycle 0.087*** 0.910*** 0.088*** 0.947*** other mean of transp. 0.022 0.236 0.034 0.359 oxcart 0.164*** 1.302*** 0.208*** 1.582*** wheelbarrow 0.064 0.609 0.058 0.571 hoe 0.075 1.180 0.065 1.018 sickle 0.083*** 0.979*** 0.078*** 0.961*** tobacco club member 2.393*** 1.559*** 5.438*** club benefit: credit 0.327*** 2.181*** club benefit: inputs at lower price -0.035 -0.465 club benefit: better tobacco prices 0.009 0.100 club benefit: extension service 0.084* 0.792* club benefit: quota access 0.110** 0.986** club benefit: transport to market 0.261*** 1.869*** 274 Table 8 Determinants of Participation in Tobacco: land share (Tobit marginal effects) (2) (3) (4) (5a) (5b) (6a) (6b) constant -13.365*** -0.873*** -0.816*** -1.358*** -15.103*** -1.206*** -14.189*** rural 1.169 0.138*** 0.128*** 0.098*** 1.727*** 0.086** 1.536** household size 0.276*** 0.017*** 0.010** 0.000 0.003 -0.000 -0.000 head male 1.248*** 0.108*** 0.101*** 0.070** 0.887** 0.070** 0.942** head married 1.631*** 0.126*** 0.112*** 0.079*** 1.002*** 0.069*** 0.922*** prop. males 0.181 0.027 0.018 0.006 0.064 0.017 0.203 prop. age 0-7 -0.611 -0.052 -0.056 -0.010 -0.110 0.013 0.153 prop. age 8-12 -1.788** -0.124* -0.132** -0.089 -0.995 -0.066 -0.776 prop. age 12-18 -0.694 -0.079 -0.050 -0.039 -0.434 -0.029 -0.345 prop. age 46+ -1.187** -0.121*** -0.119*** -0.126*** -1.405*** -0.113*** -1.334*** prop. sick 0.487 0.025 0.030 0.024 0.263 0.020 0.234 head sick -0.323 -0.022 -0.021 -0.019 -0.220 -0.016 -0.191 head literacy (chichewa) 0.559** 0.045** 0.024 0.009 0.105 0.007 0.087 head literacy (english) 0.026 0.031 -0.004 -0.010 -0.115 -0.006 -0.070 head educ.: attends school -0.898 -0.055 -0.029 -0.000 -0.001 0.004 0.041 head educ.: primary completed 0.605* -0.011 -0.004 -0.005 -0.052 -0.008 -0.090 head educ.: secondary (jr.) completed 0.653 -0.013 0.015 0.009 0.102 0.007 0.086 head educ.: secondary (sr.) completed -0.861 -0.083* -0.057 -0.049 -0.652 -0.052 -0.749 head educ.: superior completed 1.501 0.060 0.093 0.022 0.232 0.016 0.183 land area 0.000*** 0.001*** 0.000 0.000 large weekly market 1.115** 0.085* 0.097** 0.087** 0.884** 0.077** 0.830** permanent ADMARC market -0.910 -0.064 -0.038 -0.021 -0.247 -0.021 -0.256 cooperative 5.274*** 0.209*** 0.172*** 0.145*** 1.308*** 0.138*** 1.306*** credit club 1.981*** 0.100** 0.093** 0.084** 0.824** 0.070** 0.740** post office -0.128 -0.074 -0.069 -0.058 -0.749 -0.055 -0.755 public telephone -1.672*** -0.080 -0.052 -0.038 -0.460 -0.042 -0.552 health clinic -0.420 0.018 0.007 0.011 0.119 0.013 0.149 irrigation scheme -0.703 -0.033 -0.054 -0.039 -0.472 -0.034 -0.440 lnaccess (0 - 2.5) -1.026 -0.026 -0.064 -0.076 -1.080 -0.074 -1.117 lnaccess (2.5 - 4.5) 0.672 0.064 0.032 0.025 0.291 0.018 0.217 household assets 0.045** 0.567** 0.043** 0.577** bicycle 0.087*** 0.899*** 0.085*** 0.933*** other mean of transp. 0.014 0.151 0.025 0.270 oxcart 0.159*** 1.264*** 0.198*** 1.538*** wheelbarrow 0.071 0.657 0.062 0.619 hoe 0.075 1.160 0.064 0.997 sickle 0.083*** 0.982*** 0.077*** 0.958*** tobacco club member 2.364*** 1.567*** 5.433*** club benefit: credit 0.319*** 2.158*** club benefit: inputs at lower price -0.036 -0.487 club benefit: better tobacco prices 0.007 0.085 club benefit: extension service 0.080 0.770 club benefit: quota access 0.107** 0.968** club benefit: transport to market 0.264*** 1.898*** 275 Table 9 Determinants of Participation in Tobacco: land share (Tobit marginal effects) (2) (3) (4) (5a) (5b) (6a) (6b) constant -13.376*** -1.252*** -1.154*** -1.080*** -12.682*** -0.983*** -11.891*** rural 1.189 0.138*** 0.121*** 0.090** 1.654** 0.081** 1.470** household size 0.276*** 0.017*** 0.011** 0.000 0.003 0.000 0.000 head male 1.247*** 0.108*** 0.093*** 0.065** 0.864** 0.067** 0.924** head married 1.627*** 0.129*** 0.109*** 0.075*** 1.009*** 0.068*** 0.937*** prop. males 0.173 0.031 0.025 0.007 0.088 0.019 0.228 prop. age 0-7 -0.618 -0.051 -0.046 -0.009 -0.101 0.014 0.164 prop. age 8-12 -1.782** -0.128* -0.137** -0.087* -1.022* -0.067 -0.808 prop. age 12-18 -0.701 -0.080 -0.048 -0.038 -0.441 -0.029 -0.353 prop. age 46+ -1.191** -0.119** -0.106*** -0.116*** -1.367*** -0.108*** -1.302*** prop. sick 0.476 0.029 0.034 0.024 0.284 0.021 0.260 head sick -0.326 -0.021 -0.020 -0.017 -0.202 -0.014 -0.175 head literacy (chichewa) 0.558** 0.046** 0.023 0.009 0.111 0.008 0.093 head literacy (english) 0.025 0.031 -0.004 -0.009 -0.113 -0.006 -0.070 head educ.: attends school -0.906 -0.056 -0.026 0.001 0.017 0.005 0.060 head educ.: primary completed 0.605* -0.011 -0.003 -0.004 -0.052 -0.007 -0.090 head educ.: secondary (jr.) completed 0.662 -0.014 0.006 0.007 0.085 0.006 0.069 head educ.: secondary (sr.) completed -0.849 -0.086* -0.054 -0.048 -0.685 -0.052 -0.788 head educ.: superior completed 1.545 0.042 0.075 0.010 0.112 0.004 0.053 land area 0.000*** 0.001*** 0.000 0.000 large weekly market 1.102* 0.096** 0.114*** 0.089*** 0.943*** 0.082** 0.898** permanent ADMARC market -0.924 -0.057 -0.021 -0.014 -0.168 -0.014 -0.174 cooperative 5.278*** 0.200*** 0.164*** 0.129*** 1.231*** 0.125*** 1.229*** credit club 2.003*** 0.093** 0.088** 0.072** 0.759** 0.062* 0.676* post office -0.117 -0.081 -0.072 -0.058 -0.800 -0.057 -0.818 public telephone -1.652** -0.084 -0.045 -0.037 -0.480 -0.043 -0.585 health clinic -0.420 0.014 0.004 0.007 0.081 0.009 0.111 irrigation scheme -0.714 -0.030 -0.045 -0.034 -0.439 -0.031 -0.405 lnaccess (0 - 1.5) -0.982 -0.031 -0.084 -0.073 -1.223 -0.068 -1.166 lnaccess (1.5 - 2.5) -1.031 -0.014 -0.073 -0.062 -0.908 -0.063 -0.970 lnaccess (2.5 - 3.5) 0.642 0.136* 0.082 0.061 0.666 0.055 0.617 lnaccess (3.5 - 4.5) 0.783 0.056 0.024 0.012 0.137 0.004 0.050 household assets 0.043** 0.574** 0.043** 0.587** bicycle 0.081*** 0.884*** 0.082*** 0.922*** other mean of transp. 0.010 0.116 0.020 0.231 oxcart 0.156*** 1.289*** 0.200*** 1.574*** wheelbarrow 0.069 0.677 0.063 0.636 hoe 0.071 1.173 0.063 1.020 sickle 0.078*** 0.966*** 0.074*** 0.948*** tobacco club member 2.137*** 1.488*** 5.358*** club benefit: credit 0.317*** 2.179*** club benefit: inputs at lower price -0.032 -0.449 club benefit: better tobacco prices 0.011 0.133 club benefit: extension service 0.081* 0.790* club benefit: quota access 0.094* 0.890* club benefit: transport to market 0.262*** 1.917*** 276 Table 10 Determinants of Participation in Tobacco: plot area (Tobit marginal effects) (1) (2) (3) (4) (5a) (5b) (6a) (6b) constant -35.240** -31.517** -2.511** -2.450** -2.167** -38.246** -2.007** -36.597** rural 5.049* 2.418 0.235* 0.229** 0.159** 4.377** 0.146* 4.044* household size 0.890* 0.687* 0.047 0.038 0.018 0.319 0.017 0.303 head male 2.804* 2.122* 0.160* 0.153* 0.089 1.738 0.096* 1.962* head married 2.974* 2.548* 0.171* 0.158* 0.125 2.495 0.109 2.23 prop. males 1.090 0.386 0.053 0.043 -0.001 -0.024 0.011 0.208 prop. age 0-7 -1.003 -0.791 -0.066 -0.081 -0.082 -1.441 -0.042 -0.762 prop. age 8-12 -4.064 -3.322 -0.238 -0.264 -0.185* -3.269* -0.148 -2.706 prop. age 12-18 -0.568 -0.421 -0.048 -0.016 -0.024 -0.421 -0.002 -0.043 prop. age 46+ -1.315 -1.426 -0.102 -0.110 -0.114** -2.017** -0.096* -1.742* prop. sick 2.678 1.655 0.099 0.100 0.080 1.411 0.078 1.416 head sick -0.879 -0.346 -0.015 -0.013 -0.013 -0.225 -0.012 -0.218 head literacy (chichewa) 1.257 1.183 0.094 0.067 0.040 0.714 0.038 0.700 head literacy (english) -1.536 -0.672 -0.020 -0.066 -0.044 -0.818 -0.046 -0.881 head educ.: attends school -1.457 -1.101 -0.028 0.025 0.114 1.668 0.113 1.702 head educ.: primary completed 4.499 3.026 0.161 0.179 0.115 1.726 0.104 1.625 head educ.: secondary (jr.) completed 3.265 1.833 0.034 0.082 0.058 0.935 0.071 1.159 head educ.: secondary (sr.) completed -1.654 -0.949 -0.083 -0.043 -0.089 -1.916 -0.087 -1.958 head educ.: superior completed 1.306 4.146 0.101 0.177 0.071 1.116 0.085 1.337 land area 0.001*** 0.014*** 0.001*** 0.013*** large weekly market 1.850 0.143 0.164* 0.113* 1.849* 0.100* 1.703* permanent ADMARC market -2.129 -0.133 -0.106 -0.082 -1.628 -0.079 -1.622 cooperative 9.071** 0.348** 0.303** 0.217** 3.141** 0.208** 3.111** credit club 2.785* 0.154 0.147 0.086 1.400 0.069 1.170 post office -0.517 -0.132 -0.128 -0.097 -1.999 -0.094 -2.013 public telephone -2.887* -0.144 -0.117 -0.067 -1.293 -0.069 -1.392 health clinic -0.930 0.011 -0.005 -0.002 -0.043 0.002 0.045 irrigation scheme -1.185 -0.056 -0.086 -0.047 -0.891 -0.038 -0.739 graded graveled road 7.922** 0.277 0.254 0.197 2.820 0.189 2.787 dirt road (maintained) 4.026* 0.185 0.147 0.100 1.721 0.093 1.657 dirt track 4.097 0.258 0.223 0.139 2.136 0.136 2.155 dist. to urban center -0.000 -0.002 -0.000 -0.003 dist. to asphalt road 0.001 0.009 0.000 0.009 remoteness 0.038 0.638 0.043 0.740 household assets 0.068 1.330 0.066 1.344 bicycle 0.133** 2.203** 0.131** 2.239** other mean of transp. -0.025 -0.456 -0.024 -0.454 oxcart 0.108** 1.611** 0.143** 2.086** wheelbarrow 0.064 1.012 0.059 0.974 hoe 0.126 3.116 0.112 2.763 sickle 0.136** 2.544** 0.126** 2.433** tobacco club member 2.164** 1.404*** 9.869*** club benefit: credit 0.209** 2.839** club benefit: inputs at lower price 0.098 1.515 club benefit: better tobacco prices 0.182 2.516 club benefit: extension service 0.085 1.346 club benefit: quota access 0.129 1.929 club benefit: transport to market 0.332** 4.009** 277 Table 11 Determinants of Participation in Tobacco: plot area (Tobit marginal effects) (2) (3) (4) (5a) (5b) (6a) (6b) constant -30.890** -2.561** -2.545** -2.243** -37.510** -1.972** -35.490** rural 2.392 0.249* 0.240** 0.171** 4.473** 0.148* 4.051* household size 0.684* 0.050 0.040 0.019 0.320 0.017 0.301 head male 2.321* 0.173* 0.164* 0.100* 1.861* 0.103* 2.074* head married 2.481* 0.178* 0.162* 0.130 2.448 0.107 2.163 prop. males 0.375 0.057 0.045 -0.003 -0.047 0.011 0.193 prop. age 0-7 -0.930 -0.077 -0.093 -0.092 -1.542 -0.049 -0.874 prop. age 8-12 -3.243 -0.248 -0.271 -0.190 -3.172 -0.143 -2.578 prop. age 12-18 -0.331 -0.057 -0.025 -0.032 -0.527 -0.008 -0.150 prop. age 46+ -1.364 -0.106 -0.112 -0.119** -1.996** -0.095* -1.718* prop. sick 1.614 0.114 0.112 0.089 1.488 0.083 1.502 head sick -0.409 -0.023 -0.020 -0.017 -0.286 -0.016 -0.284 head literacy (chichewa) 1.147 0.096 0.066 0.039 0.660 0.034 0.630 head literacy (english) -0.663 -0.021 -0.069 -0.050 -0.869 -0.048 -0.913 head educ.: attends school -1.562 -0.056 -0.007 0.095 1.365 0.096 1.468 head educ.: primary completed 2.979 0.167 0.185 0.123 1.755 0.106 1.643 head educ.: secondary (jr.) completed 1.921 0.040 0.093 0.073 1.093 0.078 1.247 head educ.: secondary (sr.) completed -1.074 -0.090 -0.042 -0.096 -1.976 -0.089 -1.994 head educ.: superior completed 2.845 0.037 0.117 0.029 0.464 0.040 0.665 land area 0.001*** 0.014*** 0.001*** 0.013*** large weekly market 1.767 0.145 0.164* 0.126* 1.951* 0.109* 1.815* permanent ADMARC market -1.803 -0.125 -0.094 -0.075 -1.383 -0.069 -1.379 cooperative 9.711** 0.406** 0.353** 0.253** 3.422** 0.231** 3.353** credit club 3.092* 0.159 0.153 0.094 1.449 0.072 1.210 post office -0.519 -0.147 -0.133 -0.101 -1.965 -0.094 -1.986 public telephone -3.096* -0.160 -0.120 -0.077 -1.419 -0.077 -1.560 health clinic -0.992 0.008 -0.010 -0.003 -0.047 0.003 0.046 irrigation scheme -1.308 -0.048 -0.084 -0.047 -0.843 -0.039 -0.744 lnaccess 0.785 0.028 0.042 0.024 0.399 0.020 0.360 household assets 0.075 1.390 0.069 1.389 bicycle 0.147** 2.302** 0.137** 2.300** other mean of transp. -0.034 -0.616 -0.031 -0.588 oxcart 0.122** 1.704** 0.152*** 2.165*** wheelbarrow 0.063 0.961 0.056 0.916 hoe 0.139 3.304 0.117 2.901 sickle 0.143** 2.527** 0.127** 2.417** tobacco club member 2.274** 1.487*** 9.995*** club benefit: credit 0.265** 3.368** club benefit: inputs at lower price 0.080 1.256 club benefit: better tobacco prices 0.165 2.309 club benefit: extension service 0.095 1.461 club benefit: quota access 0.110 1.667 club benefit: transport to market 0.342** 4.060** 278 Table 12 Determinants of Participation in Tobacco: plot area (Tobit marginal effects) (2) (3) (4) (5a) (5b) (6a) (6b) constant -28.884** -3.192** -2.978** -2.774** -47.048** -2.384** -43.700** rural 2.457 0.254* 0.239** 0.171** 4.590** 0.147* 4.156* household size 0.674* 0.049 0.039 0.019 0.317 0.016 0.299 head male 2.293* 0.168* 0.157* 0.096 1.792 0.098* 2.007* head married 2.506* 0.179* 0.161* 0.130 2.489 0.107 2.203 prop. males 0.433 0.063 0.050 0.000 0.001 0.013 0.230 prop. age 0-7 -0.896 -0.077 -0.091 -0.092 -1.555 -0.049 -0.899 prop. age 8-12 -3.216 -0.253 -0.271 -0.191 -3.244 -0.144 -2.639 prop. age 12-18 -0.271 -0.052 -0.021 -0.029 -0.485 -0.007 -0.121 prop. age 46+ -1.381 -0.105 -0.109 -0.117** -1.982** -0.093* -1.705* prop. sick 1.603 0.109 0.107 0.086 1.455 0.080 1.468 head sick -0.425 -0.024 -0.021 -0.018 -0.301 -0.016 -0.296 head literacy (chichewa) 1.133 0.096 0.065 0.038 0.658 0.034 0.625 head literacy (english) -0.601 -0.020 -0.066 -0.047 -0.832 -0.045 -0.870 head educ.: attends school -1.319 -0.056 -0.008 0.083 1.235 0.082 1.292 head educ.: primary completed 2.848 0.153 0.170 0.115 1.683 0.100 1.581 head educ.: secondary (jr.) completed 1.859 0.043 0.092 0.072 1.099 0.077 1.241 head educ.: secondary (sr.) completed -1.103 -0.082 -0.036 -0.091 -1.884 -0.084 -1.902 head educ.: superior completed 2.686 0.041 0.115 0.031 0.500 0.041 0.688 land area 0.001*** 0.014*** 0.001*** 0.013*** large weekly market 1.723 0.163* 0.176* 0.130* 2.029* 0.109* 1.849* permanent ADMARC market -1.723 -0.110 -0.075 -0.062 -1.132 -0.057 -1.137 cooperative 9.539** 0.402** 0.343** 0.247** 3.398** 0.226** 3.333** credit club 3.012* 0.167 0.155 0.096 1.505 0.074 1.261 post office -0.265 -0.128 -0.115 -0.087 -1.673 -0.080 -1.682 public telephone -3.026* -0.135 -0.097 -0.057 -1.040 -0.059 -1.173 health clinic -0.956 -0.000 -0.017 -0.007 -0.121 -0.001 -0.018 irrigation scheme -1.314 -0.054 -0.086 -0.049 -0.890 -0.040 -0.784 lnaccess (0 - 2.5) -1.885 -0.094 -0.134 -0.118 -2.457 -0.114 -2.639 lnaccess (2.5 - 4.5) 1.009 0.126 0.079 0.024 0.425 0.006 0.118 household assets 0.075 1.427 0.069 1.426 bicycle 0.143** 2.272** 0.133** 2.272** other mean of transp. -0.043 -0.786 -0.038 -0.759 oxcart 0.115** 1.640** 0.144** 2.103** wheelbarrow 0.069 1.047 0.062 1.012 hoe 0.136 3.280 0.114 2.886 sickle 0.141** 2.528** 0.125** 2.415** tobacco club member 2.205** 1.462*** 9.952*** club benefit: credit 0.259** 3.355** club benefit: inputs at lower price 0.076 1.210 club benefit: better tobacco prices 0.160 2.288 club benefit: extension service 0.091 1.436 club benefit: quota access 0.104 1.613 club benefit: transport to market 0.344** 4.121** 279 Table 13 Determinants of Participation in Tobacco: plot area (Tobit marginal effects) (2) (3) (4) (5a) (5b) (6a) (6b) constant -28.890** -2.524** -2.411** -2.100** -36.397** -1.892** -34.353** rural 2.520 0.242* 0.233* 0.164** 4.442** 0.146* 4.016* household size 0.672* 0.047 0.039 0.018 0.317 0.017 0.300 head male 2.291* 0.162* 0.154* 0.092 1.766 0.098* 1.995* head married 2.491* 0.174* 0.160* 0.128 2.512 0.110 2.246 prop. males 0.401 0.064 0.052 0.003 0.054 0.016 0.293 prop. age 0-7 -0.924 -0.073 -0.088 -0.088 -1.526 -0.047 -0.862 prop. age 8-12 -3.186 -0.248 -0.271 -0.191* -3.315* -0.150 -2.727 prop. age 12-18 -0.290 -0.051 -0.021 -0.029 -0.500 -0.007 -0.129 prop. age 46+ -1.402 -0.098 -0.104 -0.111** -1.926** -0.091* -1.651* prop. sick 1.569 0.108 0.108 0.087 1.502 0.084 1.534 head sick -0.440 -0.022 -0.019 -0.016 -0.273 -0.015 -0.270 head literacy (chichewa) 1.127 0.093 0.065 0.038 0.672 0.035 0.641 head literacy (english) -0.605 -0.020 -0.065 -0.046 -0.829 -0.046 -0.877 head educ.: attends school -1.331 -0.053 -0.007 0.083 1.252 0.085 1.323 head educ.: primary completed 2.844 0.148 0.168 0.113 1.679 0.101 1.588 head educ.: secondary (jr.) completed 1.887 0.040 0.089 0.068 1.060 0.075 1.204 head educ.: secondary (sr.) completed -1.062 -0.083 -0.039 -0.092 -1.956 -0.089 -2.004 head educ.: superior completed 2.834 0.024 0.097 0.015 0.245 0.023 0.400 land area 0.001*** 0.014*** 0.001*** 0.013*** large weekly market 1.668 0.166* 0.182* 0.136* 2.163* 0.120* 2.007* permanent ADMARC market -1.776 -0.099 -0.067 -0.052 -0.969 -0.049 -0.959 cooperative 9.544** 0.377** 0.328** 0.230** 3.255** 0.215** 3.182** credit club 3.090* 0.154 0.146 0.086 1.382 0.067 1.134 post office -0.234 -0.129 -0.118 -0.090 -1.802 -0.088 -1.852 public telephone -2.966* -0.134 -0.099 -0.059 -1.114 -0.065 -1.293 health clinic -0.952 -0.004 -0.020 -0.011 -0.187 -0.005 -0.086 irrigation scheme -1.347 -0.048 -0.081 -0.044 -0.812 -0.036 -0.691 lnaccess (0 - 1.5) -1.557 -0.086 -0.131 -0.111 -2.542 -0.101 -2.389 lnaccess (1.5 - 2.5) -1.964 -0.081 -0.121 -0.104 -2.184 -0.108 -2.447 lnaccess (2.5 - 3.5) 0.711 0.206 0.134 0.071 1.176 0.055 0.955 lnaccess (3.5 - 4.5) 1.287 0.124 0.070 0.006 0.112 -0.013 -0.245 household assets 0.075 1.457 0.072 1.468 bicycle 0.139** 2.252** 0.134** 2.263** other mean of transp. -0.045 -0.856 -0.043 -0.853 oxcart 0.119** 1.720** 0.154*** 2.210*** wheelbarrow 0.070 1.077 0.063 1.031 hoe 0.134 3.338 0.118 2.970 sickle 0.137** 2.508** 0.126** 2.409** tobacco club member 2.175** 1.429*** 9.901*** club benefit: credit 0.270** 3.436** club benefit: inputs at lower price 0.082 1.294 club benefit: better tobacco prices 0.173 2.408 club benefit: extension service 0.096 1.491 club benefit: quota access 0.094 1.466 club benefit: transport to market 0.356** 4.197** 280 ANNEX 9A: WEATHER-BASED INSURANCE: NATIONAL-LEVEL DROUGHT INSURANCE Preliminary details on the concepts and mechanism of a weather-based insurance contract have already been provided in Annex 7D. In the case of the nation-wide insurance, the first step is to construct a rainfall index for each of the (twenty-two) Malawi Meteorological Office weather stations (Figure 1). The rainfall-index is defined as the weighted average of indices measured at each of the weather stations located throughout the country, with each station's contribution weighted by the corresponding average or expected (maize) production in that location. This Malawi Maize Production Index (MMPI) gives a measure of the countrywide exposure of maize production to drought and hence serves as a nation- wide food security indicator (Figures 2 and 3). On this note, including more stations in the basket not only gives better national coverage and hence representation of the index but also increases the placement potential of the structure in the international reinsurance markets. Such insurance would not cover smaller localized droughts which affect only a few of the stations. The government may be able to cope with small, localized droughts by transporting food-supplies from other districts of the country and by sourcing government budget reserves. Retaining such risks will most probably be a more cost-effective solution than seeking insurance and Malawi should be able to take advantage of the natural diversification of the country to reduce its insurance costs. However, in situations where drought affects several districts, or when there is a severe regional drought, this reallocation of resources may not be manageable for the government and it would be appropriate to utilize the nation-wide insurance product to measure when such events occur. A market-based instrument such as a weather-based insurance would provide a supplemental source of emergency financing to support existing resources at the country level. Distinct advantages that can be achieved through index-based ex ante financing, include: immediate cash payment, structured rules for payment, improved correlation between need and provision, flexibility of cash payments, risk assessment, risk mitigation. Further, by transferring the risk to the international financial markets, this instrument allows the government to smooth out expenditure outlays related to drought relief, and thereby improving public financial management (i.e., facilitating better planning by minimizing shocks to the budget). Given the objective nature of the rainfall-index, and the good quality of weather data from the Malawi Meteorological Office, such a structure could be placed in weather-risk reinsurance market. Hess and Syroka (2005) show that Malawi could need up to $70 million to financially cover the crop losses in case of an extreme food emergency. Given this amount, such a transaction would be treated on a stand-alone basis, with an estimated premium of approximately three times the expected loss for the reinsurer. Based on 40 years of historical rainfall data and assuming that the Malawi government retains the cost associated with deviations in maize production (as measured by rainfall) up to 25 percent away from normal, the expected loss is $1.75 million implying a annual premium of about $5.25 million (that is equivalent to an insurance rate around 9 percent for such a product). 281 Figure 1: Location of Malawi's Meteorological Office weather stations Key : 1. Chitipa, 2. Karonga, 3. Mzimba, 4. Mzuzu, 5. Nkhata Bay, 6. Lilongwe Inernational Airport, 7. Chitedze, 8. Nkota Kota, 9. Dedza, 10. Salima, 11. Chileka, 12. Mangochi, 13. Thyolo 282 Figure 2: MMPI vs. Maize Production Figure 3: Historical MMPI Insurance Payouts 283 A drought risk index-based insurance contract would transfer the financial risk of severe and catastrophic national drought that adversely impacts the Government's budget to the international risk markets. The aim of such a contract would be to secure timely and reliable funds for the Government if a contractually specified severe and catastrophic shortfall in precipitation occurs during the agricultural season, as measured by weather stations throughout the country. Access to such contingency funds in a time of crisis would generate a supplemental source of emergency financing in May to complement existing budget resources, giving the Government more flexibility in its drought response and enhancing the Government's ability to launch an efficient and cost-effective drought response. Such a new approach to financing a Government response to drought promises several additional advantages. These include greater autonomy for the Government as Malawi shifts away from an appeals-based model that relies on unpredictable and often untimely donor emergency relief funds, to an ex-ante risk management and financing model. Establishing event-specific, contractually guaranteed contingency funding also creates fiscal stability for the Government and has the added advantage of providing risk price information to assist Government in its investment decisions with respect to managing and mitigating drought risk. Weather market players, from both the reinsurance and financial communities ­ a growing market, valued at US$ 8.6 billion of outstanding risk in the most recent industry survey ­ are extremely interested in new transactions of this type of instruments. In addition to core weather market participants who offer risk management products to customers, professional investors -- such as alternative risk hedge funds -- are also becoming interested in weather risk and are beginning to source excess risk from the primary weather market. Weather is an uncorrelated risk that can enhance their portfolio positions and differentiate them from other funds that deal in traditional financial markets. The new risks, introduced by the new country locations, allow for more diversification and hence enhance the risk/return characteristics of commercial risk portfolios. Ultimately this should lead to more aggressive pricing of weather insurance products in the global market, which in turn should lead to more firms entering the sector attracted by greater market liquidity. In due course this should result in greater business growth and expansion through broadening product offerings and increasing global networks that will benefit the end user customers seeking risk management products, such as the Government of Malawi. The success of a weather risk management program for the Government of Malawi depends on the design of the underlying rainfall index on which the insurance contract is based. Index-based insurance is not insurance in the traditional sense where the insured party is compensated for every dollar loss that can be proved to have occurred. Rather the insurance is based on the performance of a specified index during the insurance period where the index is designed to correlate as closely as possible with the underlying risk. Risk mitigating payouts are made from the insurer if the index crosses a specified trigger level at the end of the contract period, indicating situations where a loss is most likely to have occurred, i.e. payouts are made based on the index and not on the actual loss itself. 284 The benefits of such an approach is that by using a proxy measure of risk, such as rainfall that is available on a real-time basis, insurance payouts can be made as soon as the rainfall data is collected and the index calculated, rather than waiting for the time-consuming and often subjective loss assessments to determine the likelihood and magnitude of a payout. Moreover, if based on an independent, objective, verifiable and replicable dataset, an index also creates an opportunity to transfer the risk to the international markets, thus removing this risk from Southern Africa. The Malawi Meteorological Office's rainfall data is of excellent quality. The Meteorological Office also has the required capacity to communicate rainfall data reliably in real-time, which ensures that the Government will be able to access the international markets and benefit from such risk management products. The disadvantage of such an approach is that proxy indicators are not perfect predictors of actual events on the ground and there will always be some element of mismatch between the index and actual experience. In addition an index based on rainfall does not reflect losses from other events, such as flood or pest infestation that can also negatively effect agricultural production. However, for the purposes of risk management and budgetary support the index does not need to be accurate to the nearest dollar: it must provide protection and reduce the Government's overall fiscal exposure to drought in the critical extreme years when national maize production is severely impacted. As discussed above, the index proposed for Malawi is constructed using rainfall data from twenty-one weather stations throughout the country and is based on the Malawi Meteorological Office's national maize production forecasting model. The model is based on the FAO's Water Requirement Satisfaction Index adapted for Malawian conditions and uses daily rainfall as an input to predict maize yields and therefore production throughout the country. The model captures not only the total amount of rainfall received at each station, but also its distribution during the agricultural season and how that impacts maize yields. Using such an underlying rainfall index an insurance contract can be structured to reflect conditions which would impact national maize production and therefore food security, resulting in the Government need to import maize. An insurance contract based on such an index would have triggered cash payouts to the Government most recently in 2005 and in 1992 and 1994 to finance such purchases. It is important to note that not all production deficits are caused by drought: excessive rainfall and flooding, civil strife, poor farm-management, pest infestations and inadequate seed and fertilizer supplies may be as important as deficit rainfall in triggering food emergency situations. These risks are not captured by the model and cannot be objectively indexed for risk transfer. However this approach provides the Government with a tool which can manage the greatest systematic risk to its budget: drought. The first pilot of this type of instrument has started in early 2006. A the United Nations World Food Programme, with technical assistance from the World Bank, entered into the first-ever humanitarian aid weather insurance contract with a leading European insurer. The contract provides contingency funding in case of an extreme drought during Ethiopia's 2006 agricultural season. The policy is based upon a calibrated index of rainfall data gathered from twenty-six weather stations across Ethiopia. Payment will be triggered when data gathered over a period from March to October indicates that rainfall is significantly below historic averages, pointing to the likelihood of widespread crop failure. 285 While the experimental pilot transaction only provides a small amount of contingency funding ­ a maximum payout of $7.1 million for a premium of $930,000 paid by USAID ­ the funds will be available to WFP at harvest-time which will allow for an intervention four to six months earlier than the traditional appeals-based system. If a catastrophic drought occurs in 2006, WFP will use these funds to assist 65,000 households in November 2006. 286 ANNEX 10A: DETAILS OF THE METHODOLOGY FOR BENEFIT INCIDENCE ANALYSIS The analysis employs the conventional benefit incidence methodology in examining the distributional impact of public spending. The methodology is founded on the principle that the distributional impact of public spending depends on the `behavior' of government in allocating resources among competing uses and also on the behavior of households in accessing publicly provided services. Steps in benefit incidence analysis There are usually four main steps followed in conducting a benefit incidence analysis (BIA). The first step involves calculating the unit cost (also known as subsidy) of providing a particular service.68 The unit cost shows how much the government actually spent in providing a particular unit of service, net of cost recovery payments made by beneficiaries of the public service.69 For example, the unit cost of providing primary education would show how much the government spent on one primary school student, net of school fees paid by that student. Algebraically, the formula for calculating unit costs of providing a service can be presented as follows: Unit cost = Si Ei Where Si = Government spending (net of cost recovery payments) on service i (e.g. primary education) Ei =Total number of users that benefited from service i (e.g. total enrolment in primary school) The second step is to rank the population of users from poorest to richest using a welfare measure and then aggregate them into groups with equal number of users (e.g. quintiles, deciles). The purpose of this step is to categorize users into different income groups before looking at how the benefits of public spending on a particular service are distributed across various income groups. A decision always needs to be made on the choice of analysis (i.e. whether to use individual or household as unit of analysis) and on the welfare measure (i.e. whether to use per capita household expenditure or per adult equivalent household expenditure). The third step entails estimating the benefit incidence of public spending on a particular service using information from the preceding steps. At this stage, it is possible to derive the 68According to Demery (2000), the practice is to confine the analysis to recurrent spending in order to avoid the difficulties encountered in estimating the flow of services/benefits from capital expenditures 69Demery et al (1996) argue that it is only valid to net out revenue from cost-recovery measures if such revenues are returned to treasury rather than retained within the within the collecting institution 287 level of subsidy for a particular public service that accrues to an income group as well as the share of total subsidy (share of benefits) in a particular sector that accrues to an income group. The level of subsidy for a particular service that accrues to an income group is derived by simply multiplying the unit cost (average benefit) by the number of users of the service in an income group. Algebraically, this can be denoted as follows: Total benefits to income group j = X = Si j * Eij Ei Where X = total benefits (e.g. in Malawi kwacha) accruing to income group j j Si = unit cost of providing service i Ei Eij = represents number of users of service i in income group j On the other hand, the share of total subsidy in a particular sector that accrues to an income group is obtained by multiplying the participation rate of a particular income group in a service by the service share of government spending in total sector spending. Algebraically, this can be denoted as follows: Share of total subsidy = xj = Eij SS Ei * = eijsj i X Where xj = j= share of total subsidy in a particular sector that accrues to income S group j eij = Eij = participation rate of income group j in service i Ei si = Si = share of government spending on service i in total sector spending S The final step involves plotting concentration curves for benefits from government services. Concentration curves plot the cumulative proportions of households (individuals), ranked from the poorest to the richest, on the horizontal axis, against the cumulative proportions of benefits received by households (or individuals) on the vertical axis. These curves become more instructive when compared with two benchmarks: (a) the 45 degree line (line of equality) and (b) the Lorenz curve of income or consumption, which depicts income distribution in a population. This will help arrive at some conclusions regarding targeting and progressivity of government spending. With regard to targeting, benefits from government spending on a service are said to be pro-poor if the concentration curve for these benefits is above the 45-degree line. For 288 example, if we have quintiles as income groups, such a curve would show that the bottom 20% (poorest quintile) gains more than 20% of the total subsidy (and the richest quintile, less than 20%). On the other hand, benefits from government spending on a service are said to be progressive if the concentration curve for these benefits is above the Lorenz curve for income or consumption, but below the 45-degree line. This would mean that lower-income groups get a larger share of benefits from government spending than they do of either income or consumption. In other words, a government subsidy would be deemed to be more equally distributed than income. Disaggregation In most benefit incidence analyses, it is usually instructive to disaggregate calculations of benefit incidence by geographical location and gender. Therefore, to the extent possible, a similar approach will be adopted in the chapter. 289 s in 71 oi est A rat or of dna SED unto Indicator included MA pofoear in onipt ght ) Poverty headc Sh tileniuq 290 onali ums alencev rwei de oderate nat con Pre un (m severe of te ta ng bility: Da /Med/h w Lo (usih S) Estima Relia Hig Hig DH ataD n/o g al rtino val hnto h 1/sh OALS m Inter earsy ont hnto )?( G Nor Collecti Rep years5 M1 years5 M1 T years5 5 M1 rsaey5 rsaey4 raey1 ontm3 1. 2. 1. 2. 1. 2. 1. 2. 3. 4. year EN M Office ELOP ncy onsible with of (MoH) EV Age lan D Institution/ )O O O Resp O O O NS NS rystnii O Ho NS M Natio Statistical (NS NS NS NS 1. 2. M Health 3. 4. NNIUME a nte ring ILL atD 70 M )(s stem O) al ) ) ceirP ) ) ) )S Household er ) Monito M gemana Sy THE onti HS(I PI (W Source Mht WH F- ation Na grated eyv nkaBd S ING IH nsumoC (Cx nkaBd S de IH IPC nkaBd S IH IPC nkaBd S* ey rm S)I Heal NICE Inte Sur Worl( 1. 2. In Worl( 1. 2. Worl( 1. 2. Worl( IHS1. DH2. Welfare.3 rv fo Su 4. In (HM (U tn *) ) ONITOR M $1 cou of of line age of low a age low tyr ratiop rendl of of FOR be be depthx lan poorest chini SE Indicators years ortion ation daya eadhy gay ce of tioan ercent(p ovep in ation noipt ght five MDG Share ums Prevalence rwei OURC Prop Povert Povert den de S 1. popul (PPP) 1a. ratio popul national 2. (inci poverty) 3. q tileniu con 4. un (under ATA D of st e than from NAL between rgeaT eht less 5, peoplfo is ,e and09 ATIO Halve, N MDG 1:t 201dna me proportion suffero onti Halv 19 or incoes y 2: the wh r 11A: Targe 9901 prop who daa1$ Target between 2015, people hunge EX NN A G 1: MD Goals icated ger Goal Era extreme poverty and hun t en yar le)a of 5 lm prim otsrlgi starting ry 291 enro in n,oi tionr grade noi ale/fem opo Net ratio educat (m Pr ilspup ohw1ed gra reach ofoitaR nis boy seconda educat 04:0 ni -2 asbi t en rd n no4 pre rd llm woL w wa ro tes;am nwaw ni latio tes;am ighH- 200 Lo up en esti do asbi opup esti Med omfr )(? 1 1 1 1 hs/ )? hs/ )? hs/ )? years5 ontm3 year( years5 ontm3 year( years5 ontm3 )? years year( 10 sraey5 hs/ years5 ontm3 year( 1. 2. 1. 2. 1. 2. 1. 2. 1. 2. fo fo fo fo O rystnii n ) O catio rystnii n ) O catio rystnii n ) O catio rystnii n ) NS M oE NS M oE NS M oE O NS M catio oE 1. 2. Edu (M 1. 2. Edu (M 1. 2. Edu (M NS 1. 2. Edu (M d naloi .atad base foe ates aluv le) nat des O) O) O) rectdi ovirp onipt estim ailab us ums av S IS S IS S IS S S IS No source. IHS con (FAO: calorificno IH EM NESC IH EM NESC IH EM NESC nseC IH IH EM oodf 1. 2. (U 1. 2. (U 1. 2. (U 1. 2. 1. 2. s ratiot n ilspup n o fon wo of wh1 tiolepm to51 y,ar n of nd catio llmen ucatio co tio bel folevel gyr rate boyotslrgi ro ed tionr b5 sd seco edu or op Pr onitalup m ene noipt en ary ary opo radeg -ol y,ar 5. po muinim ums Net Pr ngi grade Prim prim * Literacy ear -y ofoitaR im dietary con 6. in 7a. artst pr tertiaryd reach 7b. rate 8. 24 9. in an by to of y by ar of an that, and able et reu be course boys, niyti lse th econds erablyf Ens will fulla gniolo ar levlla sp pre teralon 3:t children, like,a schy Elimina:4 di andy ar et ar indna Targe 0152 everywhere rlsig m rg m complete pri Ta gender pri education 0052 education al y 2: ar 3: ote m Goal Achieve univers pri education Goal Prom gender equality and empower of lds -o rate rate year4 fiver- rtalityomt rate 292 Literacy -251 de Un rtalityom ternala Infan rate M rtalityom h 72 gh h Hig Hi h Hig High 1. 2. Hig years 10 rsaey5 s earsy01 1. 2. 1. year5.2 earsy4 years4 year1 years4 1. 2. earsy4 )(?tne Ho H Ho Ho of O/M Ho O O NS NS erklC liam O/M NS M O/M Par NS NSO/Mo 1. 2. NS n Electio ntarye O) O) O) O) O) ary/t WH WH WH us us en F- F- 73 F- NESC nseC S HI NESC nseC S )UP-I HI O) on S NICE S NICE SHD IS HS NICE S (U 1. 2. (U 1. 2. (IL Parliam records (Inter-Parliam Uni DH (U DH (U 1. 2. (U DH literatefo -51segan ninemow ehtni or rate seats ent sect of nine rtalityom -eno of ne nsti rtalityom Ratio meotn ofer oympl ralutulc tionr ntemairl fiver- alitytrom aga wom pa tionr de Sha em opo fant opo ildrhcd ternala gri Un In M 10. meow Pr 24 na byd onali Pr unized to 11. wage no 12. hel nat 13. rate 14. 15. ear-oly imm measles 16. ratio two- and by by tween 1990 be duce er-five s, Re tween und tear duce Re 5:t be 0152 the,51 6:t Targe thirds, 20 mortality Targe three-quarter 4: 5: women Goal Reduce child mortality Goal Improve h e ne of tionr eddnetat 'sr ealth theo rate /Rural; M alencv 24 omw tearesu illed m pre 293 opo Pr hsrtbi an skyb nnelo by ofle V pers (Urb and lev education) HI 15-gnoma d nantg ndo raceptive alencev ear-oly eht pre Co of cont pre 74 hg umdie 75 hg hg umdie Hi M Hi 1. 2. 1. ghiH Hi M High 1. 2. years4 year1 rsaey2 s 1. 2. 1. year4 years4 rsaey4 raey1 1. 2. t t t Ho O/M Ho AIDSlan nois PDE O/M Uninoiat DPE M O/ Uninoiat DPE M Uninoiat NS M mmis 1. 2. Natio Co NS1. pul Po )(? ACN NS pul 2. 1. Po )(? ACN O/ pul 2. NS Po )(? te- O) O) an O- at WH WH ce WH ) noiat O) O) WH WH F- F- Site ics )n F- F- NICE SHD IS el eillan clin DS-I ICEF S S S HS NICE tin puloP rv (U 1. 2. (U Sen Su talan NA UN DH NAC N Divisio DH NAC NICE DH NAC NICE (U 1. 2. (U 1. 2. (U 1. 2. (U sthirb ne ofe last -51 DS of dell omw rat *c at of nel nt use use correct tionr ski encelav * by opo onsrep ptive pre gnaerp 24ot rate tage sex ithwsd cen nsive /AIVIHfoe -ol Pr Condom Per ledg 17. ndedetat HIV 15- modnoC ong contrace alencev isk-r ear rehep ealthh 18. am ages 19. the pre 9a.1 ighh -y 19b. 24 com nowk *d eht otiaryti by edt of hal to 5, tal gun spread 201dna mor Have be aln 7:t the and AIDS/ 9901 mater Targe 2015 reverse HIV nal 6: tab AIDS/ laria,a herto mater health Goal Com HIV m, and diseases evipt rate of d s s yl ort sh osi osi race nte ontC alencev alencev 294 rcul fonoirtop detecte rcul rectdir de veder pre Pre ubet Pro ubet cases un obs eatmrt 77 de 76 w/M umdie umdie High Lo M Medium M rsaey4 raey1 H) 1. 2. years4 eary1 rsaey4 rsaey5 yb(raey1 1. 2. 3. Mo year1 eary1 t of DPE M Uninoiat nistryi O/ red Ho pul S NS Po )(? O/M Ho O/M Ho NS Gen M NS NO MoH M - DS noiat AI re. ly) O) UN WH S puloP )n F- onn F- tio en DH NAC N S O) NICE IS O)H SHD (dataS )?(S )?(S IH 1. 2. (U Divisio DH (U WH HM (W 1. 2. evrp NICE O)H (U HSI (W HSI ol veder urse ptive ho sc otsnahpro 4 -1 on 10 dna n dna es cen associated of ffectiveegn )(i cen of cured associated ses co rate osisl ort Contrace alencev ofoitaR of ce cenadnetat ags tion usi cas obsyl an or malaria-nin entioverp e Prevale (i) Prevale rcue tionr i)(idna sh ph ratesh eatmentrt ratesh opo osi nte 19c. pre 20. ndanetat ool or (i) deat alariam Prop latio easra (i) tub Pr rcul rectdir sch on-n 21. (ii) with 22. opup risk lariaam (ii) deat de and measures 23. (ii) with 24. ubet detected un eatmrt by of edt jora m hal to gun incidence other Have be 8:t the and anda Targe 2015 reverse malari diseases h TS; of s yl ort DO­ d sh per of tionr osi witn se cure 295 rcul rectdir nte DOTS- to ur opo de veder deixoidn ons tionr latio edv co Pr ubet cases un obs eatmrt ssii opo course Carbo em capita Pr opup pro access im s year 10.1 rsaey5 rsaey4 raey1 2. 3. 4. A A EARN NREo NREo NREA EARN with O O O NS NS NS H Mo M M Mo Mo 1. 2. 3. Mo ) )C MC nkaBd UNSD) eno )OH W- Worl Oz- NEP A, NEP )?( us S S (W AO)F( O)H nseC IH SHD WM (U (IE (UNFCCC, (U Secretariat) IHS (W 1. 2. 3. 4. P of na land rest de di tain ita) ns to of tionr foybdre otyt rsi cap ption of of ess urce,osr areafo h mainot vedi use gy ilofos GD1$r oxiid arboc solgn er eting on ) opo (psn ro ) nsum tionr usi tionr witn acc wate ble Pr ovec Ratio Ener rbaC Co tons opo opo OTS (D 25. area 26. ectedtorp caligoolbi eara amr pe)nte val ) rofluo Pr * Pr latio ilog surface 27. (k equi (PPP 28a. issiome ne-depl DP onitalup 28b. ozo chlo (O 29. po fuels 30. opup proved sustaina im the the of feas by on to grate stainableus dna ntoit reverse orti t cess Inte of licies Halve, op ac 9:t po and pr pmen environmental 10:t of untry the,51 withou Targe principles develo co program loss resources Targe 20 people sustainable Ensure environme ntal sustainabil ity al)r n /Run 296 itatio rba san (U 1/sh s 1/sh ontm3 )?( year )?( years5 year1 5. year 10.1 rsaey5 rsaey4 raey1 ontm3 2. 3. 4. 5. year 1. 2. with O Ho O O O O NS M NS NS NS H Ho NS M O 4. 5. 1. 2. 3. Mo 4. 5. NS O) O) T)A WH WH C)A C)A SI F- us S F- S HM NICE nseC S IH SHD SI BITAH D/D D/D WM HM NICE IHS WM N- 5. (U 1. 2. 3. 4. 5. (U 1. 2. (U (OEC (OEC to d el an e cab ,n of accessh on,itat of access talto dep a s'ronod AC comni of social sic arym tritio uralr tionr wit sani uralr tionr with reu veloed nu ten as fo tionr ba pri ndan ODA naloi sector-allo CD/DEO opo Pr edv ndan (basic opo Pr oldsh re Net ba pro ba secu leasteth opo for care, tries, geatn DAC nat of Pr un A oss ur 31. population im ur 32. house to 33. to co perce OECD/ gr 34. ilateral,b OD donors services education, ealthh sicab 0 - oth the d a, 10t la tariff- anr rule y --b ve 20 asel po 20 at dwellers vel ase good thefo d ed cludes wate Ha: by improvement of De: financi open, edictable, ludc totn and onti , t, )yll duc Address op on 11t ed, esvil slumn 12t an pr, and (in ance needs (ins m pmeno rey anylal ionat 13: devel drinking sanitati rgeaT on hievca the significant ni millio rgeaT vern vert further based nondiscriminator trading syste commitme go devel po nati interna Target special least countrie 8: apo al nershit Goal Devel obgl for par p developme nt 297 C)A C)A C)A WTO, D, D/D D/D D/D NCTA (OEC (OEC (OEC (U WB) ancest AC in d free of d ssia tiednu ofn ved eirht ved tio onali taltofo dnaeu opele an ittedm )n tionr is val dnaseritnuoc official ent CD/DEO atth recei asseritnuocd ssorg in ngi ofn semocni recei lopevedd opor nat pr rtion rytnuocd from)smra dev ad ng opo water, opm of ke tio rs oc as oss (bys itatio Pr opi eastl A ODA or onali ODA sem ope tries, safe san 35. ilateralb devel OD onod 36. ndlal oprp islanlla nat 37. sm states grreiht coin Propo. port 38 devel im excluding devel omfr un yutd co of ss ali to of more the accee relief hanced offic for en debt of and DAO mmitted uction) Address fos untries nd co isla atesst Program rts of and debt, cos red need theh quota-fre expo ram HIPC ellation tyr 14: ping small and for prog for ncac ialc bilateral generous countrie pove Target spe landlocked and develo (throug 298 ) WTO, )O nkaBd D, WT D) D, NCTA Worl-F (U WB) (OEC (OEC (IM ral ng and pedo ltu opi trop e tag tiles sup d riffstaegar vel ve de agricu tex devel rcenepa esticm DAO ltural doss ildubpleh fo erbm evah HIPC ansnt ha HIPC tsni by on and fo as to nu atht pon tal Ave. tries tscu gro ofnoirtop pacity atht tries tries edid ca eirhtd tries poinosi eirhtd ber e)viatul posed un 39 im co odrp omfrgnhiotcl OECDr un Agricu. team un co 40 esti co eirthfo ducto Pro. pr 41 ovrp .To un letiopm trade 42 co reache deci num reache co (cum ser Smallfo with of and debt long ping the velo 2nd2 mbly nal easum the )s Deal velysi ems countries make in for De and Asse ions 15: probl rehen ping natio to able tion debt order Ac Sustainable Development Island States General provi Target comp the develo through international in sustain es; r line pe onitaul hon ersb lar opp0 299 Telep Cellu subscri 001, earsy5 O NS ) ) nkaBd nkaBd oni cat onali unim Worl-F Worl-F O) O)H (IM (IM IHS (IL (W naternt(I ) ecom on U) Tel Uni (IT of e to , a ts rat HIPC as por ex nderu rvices ntem dsolr-ae d e and rep an tial lea ofnoi accessh lines ersb wit essen cris n putersmoc relief e service ofe tag sedna oypl femd ortp le, nablaistusa sub latio an pu Debt. f ab rd s lar mmitted Personal. 43 co itiativin Debt. 44 encrep Unem. -y42ot-51 goods 45 of male talto Pro. onsgu Telephone. 46 population affo dr basi 47 cellu po001 48a tn for of e on work the ons make In with implement dece countries, In with alc rovidp insg affordable, countries thiw benefits cati dru or,t gies,o informati uni 16: eration ingp andp fors uctive 17: ceuti to In:81 the sec m prod eration nies, ping et technol com rg term Target coop develo develo strategie and youth Target coop pharma compa access essential develo Ta cooperation private available new especially and s, sm user n etn latio siti ndo ing ing egar co 300 Inter 1,000rep 15 opup n,oinuninemowg oklo llaborationoc ohw42­ ontraceptive(c sing hy- in oklo-y (u V 19cr ato HIfon ealtha ealthha atth agesnem UNICEF, ow kn on Indic issio amd smn knowo above, how ndeunredlihcfoegatn rce 19b). tra wh ned andnemow pe of ­24 defi 15 byd easurem only is rate catoridn(iegdelwo xual as es seehtgni and,no getan issi rce catori pe sm entv tran indeht the are agnemdnane easuremeb e.blalaivaer use ndom knSDIA totne pre ofs HIVt late They on.i calcu womfo treatm; zatni co V/ way to ga .d whe,d the HI onro joram abousniot geatn cepn ableeb ndicator.i rce Orr e-treated doptea usebed Since .sl woteth to iscom actual eys the peehtd bo La been fyti rv an onaliat yetton oulhs su ," smission.n goay ofs secticidinred rne of un U) esnilyrt catdininadna tra a)91 HIV of by (IT have ovep e." st use sleepe ent but ratn naloi tingnev oratcidn(i ertvopdna,erdn enid callono mm erb onentp domn co ing Inteht rrectly co num coo mo comowt ent ntets five pmo per nat letiopm pre in ers 0 us n on ear-o ect ndicators,i d co tive 10rep n ary rej latio ffece 24-y athtsr der pu unsis latio MDG basesro bynoi ndeunredl rim who "p chi use cat ares herotni to-51 currentl r),e catoi fectni of eary in opup Internetb. 48 po001 ist onsiatutisksir-hghini ge,ht whsdl twoeth fficientusa nsi agr vel de ealh noty esrper "co itional ndii, enm ressgo of indy V use partn HI tage geatn addsa dsenrtyrt ndomoc areeerht rce only rcenepeth ncesi proxowt morff reutufrof infected r, un edcu peeht by posedrop ovepyrtnuocgnriot elopvedred prgnki acrtni unr thods,em modnocno or ul od herselt etgrateht treated. rea catoi catdini asde weveoH fined faithful,e pr,OH ecotrpn of HIV. easuredm dicatorsines indeivat an by usefosalsi)e isr on to HIV.ti Wd can rso propriatelypa contraceptive onimr altern ong can The* Fo An Am edntemelp rat catoi sexg pea transmit easuremdev ind smnartn are encealv is Th itinmlid cano anSDIANU atht beotnoitnev Pre who (a) (b) (c) sup pre (d) an pers ithw wo kn rsonep (e) five prominA)(f 301