December 2017 TECHNICAL POVERTY ESTIMATION REPORT Myanmar Poverty and Living Conditions Survey Ministry of Planning and Finance This technical report is a joint effort of the Government of Myanmar, Ministry of Planning and Finance, and the World Bank. The report is a product of the World Bank's programmatic poverty work in Myanmar. ACKNOWLEDGMENTS The team would like to thank, in particular, the Minister of Planning and Finance for his support to the Myanmar Poverty and Living Conditions Survey (MPLCS) and its analysis. The team warmly thanks the Deputy Minister of Planning and Finance for guidance in the MPLCS Steering Committee.Other special thanks go to members of the MPLCS Steering Committee and the MPLCS Technical Committee for their substantial and substantive guidance and support. The study team would also like to acknowledge the key role played by the Planning Department in supporting survey operations and analysis, and specifically U Maung Maung Tint, Director General, Planning Department, Daw Thway Thway Chit, in her capacity as Deputy Director General, Planning Department, and Dr Wah Wah Maung, Director General, Central Statistics Organization. Additional contributions were made by the National Nutrition Center, Department of Public Health, Department of Labour, Department of Human Resources and Planning and Training, Department of Myanmar Education Research (Department of Education Research, Planning and Training), Department of Labor, Department of Planning, Department of Agricultural Land Management and Statistics, and Department of Population. Special thanks also go to the World Bank for their support to the survey and analysis, and more specifically to Ellen Goldstein, World Bank Country Director for Myanmar, Laos and Cambodia; Abdoulaye Seck, in his capacity as Country Manager for Myanmar, and Salman Zaidi, Program Manager of the Poverty and Equity Global Practice in the East Asia and Pacific Region. The task team leader at the World Bank was Dr Reena Badiani-Magnusson, Senior Economist. TABLE OF CONTENTS Abbreviation iv Executive Summary 1 1 Overview of Content 5 2 Institutional Arrangements 7 3 A Revised and Rebased Measure of Poverty 9 reflecting living conditions in 2015 3.1 Step 1: Definition of New Consumption Aggregates Using the 2015 MPLCS 11 3.1.1 Food expenditures 11 3.1.2 Non-food expenditures 19 3.1.3 Durables 23 3.1.4 Housing 26 3.1.5 Household size and composition adjustment 28 3.1.6 Spatial and temporal price deflation 29 3.1.7 The consumption aggregate and its components 29 3.2 Step 2: Definition of the New Poverty Line 32 3.2.1 Nutritional benchmark 32 3.2.2 Estimating the food poverty line 33 3.2.3 Non-food Allowance 35 3.2.4 Estimation and iteration approach 36 3.3 Step 3: Poverty Estimates 38 3.3.1 The poverty line 38 3.3.2 Basket of food at the poverty line 38 3.3.3 Composition of food and non-food at the poverty line 39 3.3.4 Estimated poverty rate, by strata 40 3.4 Step 4: Sensitivity Analysis and Robustness Checks 42 3.4.1 Inclusion of durables 42 3.4.2 Inclusion of health expenditures 43 3.4.3 Spatial price index 44 3.4.4 Adult equivalence scales 45 3.4.5 Calorie norm 46 3.5 Poverty Trends using the new 2015-rebased estimate 49 References 51 Annex A1 Variants of the Consumption Aggregate 56 Annex A2 Estimation of Poverty Line: Calories 60 Annex A3 Poverty Estimates and Sensitivity Analysis 72 LIST OF TABLES Table 3.1 Conversion to kilogram 12 Table 3.2 Items included in the non-food aggregate 20 Table 3.3 Health and education elasticity of expenditure 21 Table A1.1 Variants of food aggregate 56 Table A1.2 Variants of non-food aggregate 57 Table A1.3 Comparison of depreciation rates by item 59 Table A2.1 Adult equivalence parameters 60 Table A2.2 Overview of calories per kilogram per item 61 Table A2.3 Calories assigned to other categories 70 Table A2.4 Nutritional requirements in Myanmar 71 Table A3.1 Calories per item in poverty line food basket – detailed 72 LIST OF FIGURES Figure 3.1 Methodology of price calculation 13 Figure 3.2 Mean food expenditure per adult equivalent per day 16 Figure 3.3 Food consumption expenditure by item 17 Figure 3.4 Urban and rural food consumption expenditure by item 18 Figure 3.5 Non-food expenditure composition 22 Figure 3.6 Durables composition across durables use value distribution 25 Figure 3.7 Durables ownership across per adult equivalent expenditure distribution 26 Figure 3.8 Actual versus imputed rent from generalized linear model 27 Figure 3.9 Total Consumption per adult equivalent by components (spatially deflated) 30 Figure 3.10 Share of food, non-food and durables expenditures in total consumption 31 Figure 3.11 Calories per adult equivalent 34 Figure 3.12 Share of calories from different food groups 35 Figure 3.13 Example of poverty line estimation process 37 Figure 3.14 Convergence of poverty line and headcount 38 Figure 3.15 Reference food basket of poverty measures in Myanmar 39 Figure 3.16 Composition of basket of goods among those near poverty line 40 Figure 3.17 Poverty headcount, poverty incidence and poverty intensity in Myanmar 41 Figure 3.18 Food poverty headcount, by urban and rural strata and agro-zone 42 Figure 3.19 Sensitivity of poverty estimates to durables aggregate used 43 Figure 3.20 Sensitivity of poverty line and poverty estimate to health expenditures 44 Figure 3.21 Sensitivity of poverty estimate to spatial price adjustment 45 Figure 3.22 Nutrition norms used to anchor poverty lines in low- and middle-income countries 47 Figure 3.23 Poverty estimates using different calorie norms 48 Figure 3.24 Estimated trends in poverty rates using the new rebased estimate 49 Abbreviations ADB Asian Development Bank CBN Cost of Basic Needs EA Enumeration Area GOM Government of Myanmar IGC International Growth Centre IHLCA Integrated Household and Living Conditions Assessment ILO International Labour Organization MI Multiple Imputation MNPED Ministry of National Planning and Economic Development MOH Ministry of Health MOPF Ministry of Planning and Finance MPLCS Myanmar Poverty and Living Conditions Survey NGO Non-governmental Organization SC Steering Committee TWG Technical Working Group UNDP United Nations Development Programme UNFPA United Nations Population Fund UNICEF United Nations Children’s Fund UNOPS United Nations Office for Project Services WFP World Food Programme iv Executive Summary 1 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY A joint analysis of poverty and living standards was conducted by a technical team from the Ministry of Planning and Finance, Government of Myanmar, and the Poverty and Equity Global Practice of the World Bank. Poverty has previously been estimated using data from the Integrated Household Living Conditions Survey conducted in 2004/05 and 2009/10. Using this earlier data, poverty in Myanmar has been estimated using two different approaches. Poverty was initially measured by the Government of Myanmar and its development partners using data from IHLCA-I (“MNPED et al (2007)” methodology); this first measure of poverty based the poverty line and estimate in the living conditions of 2004/05. Poverty was estimated to be 32.1 percent in 2004/05 and was estimated to have dropped to 25.6 percent in 2009/10 (MNPED et al, 2007 and MNPED et al, 2011). A poverty estimate based on 2009/10 standards of living was put forward by the World Bank in 2014 (“World Bank (2014)” methodology), using data from the IHLCA-II. The World Bank estimated poverty to be 37.5 percent in 2009/10 (World Bank, 2014). The objectives of the joint poverty analysis by the Ministry of Planning and Finance and World Bank were to: (i) put forward trends in poverty between 2004/05, 2009/10 and 2015; (ii) present a measure of poverty that reflects the situation of poverty in Myanmar in 2015 and; (iii) conduct analysis about the situation and nature of poverty in Myanmar that informs policy choices and strategies. The first report emerging from the joint assessment, Part One, put forward trends in poverty over time using the two poverty estimation methodologies previously used in Myanmar. The first report also made the recommendation to revise and rebase the poverty estimates to reflect the needs of the poor in 2015 emerged from the initial stages of the joint analysis of poverty. Updates to a country’s welfare aggregate and poverty line are recommended approximately every ten years to reflect changes in living conditions that occur as a country gets richer (such as a shift in the basket of goods from food to non-food goods) and to reflect changes in survey and poverty estimation methodology. A second poverty report, Part Two of the Poverty Assessment, puts forward a revised and rebased poverty estimate and method to reflect the needs of Myanmar’s poor in 2015 (“MOPF and World Bank (2017)” methodology). 2 TECHNICAL POVERTY ESTIMATION REPORT This technical report accompanies Part Two of the Poverty Assessment. The report is intended for a technical reader who would like to explore the details of the poverty measurement exercise. The results and deeper analysis emerging from the poverty measurement exercise are presented in the Part Two report. The revised and rebased poverty estimate is based on a revised measure of welfare that is calibrated to living conditions in 2015, and a re-estimated poverty line. There are three key differences between the new welfare measure and the welfare measure previously used by the Ministry of National Planning and Economic Development. First, durable use value is included to reflect the growing importance of home assets, such as electric fans, solar batteries and mobile phones in households in Myanmar. Durables were not included in the MNPED et al (2007) methodology. Second, the calorie norm and adult equivalent parameters used were revised to reflect updated calorie estimates produced by the Ministry of Health and Sports. The new poverty line is based on a basket of 2238 calories, compared to 2300 calories used in the two previous poverty methodologies. The calorie estimates used in this poverty measurement exercise are more finely cut than those used in the previous exercises. In previous poverty estimations, all children under the age of 15 were treated as having similar needs while in this estimation, for example, a 2-year-old is treated as having different needs to a 10-year-old. Finally, the new consumption aggregate and poverty line are based on the food and non-food consumption patterns of the population in 2015, compared to 2004/05 in the case of MNPED et al (2007) methodology. Based on the recommended revised and rebased poverty estimation method, we find that 32.1 percent of the population were poor in 2015 and 9.8 percent of the population were food poor. Using this new method, we see a decline in poverty over time between 2004/05 and 2015. The decline in poverty is consistent with the patterns of poverty reduction reported in Part One of the Myanmar Poverty Assessment (MOPF and World Bank, 2017). Poverty in Myanmar’s farms and villages (rural areas) is substantially higher than that in its towns and cities: 38.8 percent of the rural population are estimated to be poor compared to 14.5 percent of those in its towns and cities. This amounts to 15.8 million poor in total, of which 13.8 million are found in rural areas and 2 million are found in urban areas. We are unable to estimate state or region level poverty due to the small sample size of the survey. There is substantial geographic variation in poverty. The headcount rate of poverty in the Coastal areas and in the Hills and Mountains is significantly higher than the average seen in Myanmar, and that seen in the Delta and Dry Zone. Despite having lower rates of poverty, the population dense Delta and Dry Zone are home to two thirds of Myanmar’s poor. We find similar patterns in the new poverty method based in 2015 living conditions over time, notably we see a decline in poverty from 48.2 percent in 2004/05 to 32.1 percent in 2015. 3 4 1. Overview of Content 5 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Overview of Content The technical poverty measurement report describes the estimation of poverty in Myanmar by a joint team from the Government of Myanmar, Ministry of Planning and Finance (MOPF) and the World Bank’s Poverty and Equity Global Practice and Living Standards Measurement Survey Team. The joint analysis had three interlinked objectives: 1. To construct comparable poverty estimates over three survey waves using the two methodologies previously used to estimate poverty in Myanmar. 2. To propose a method that could be used to estimate poverty in 2015. 3. To conduct analysis of the correlates and determinants of poverty, to provide an overview of the critical human and economic development needs in Myanmar. The technical collaboration between the World Bank and MOPF has led to four reports: 1. Survey Conduct and Quality Control Report for the Myanmar Poverty and Living Conditions Survey, (MPLCS), 2015; 2. Analysis of Poverty in Myanmar: a. Part I: Trends between 2004/05, 2009/10 and 2015, based on previous measurements b. Part II: Poverty trends and profile based on the new poverty estimates 3. Technical Poverty Estimation Report, accompanying Part II of the Poverty Analysis This report proceeds as follows. Section 2 discusses the institutional arrangements for poverty estimation. Section 3 documents the steps taken to estimate poverty in Myanmar in 2015, and puts forward a revised and rebased method for doing so. Detailed annexes explore the assessment of assumptions used to construct the aggregates, the food basket, and calories, among others. 6 2. Institutional Arrangements 7 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Institutional Arrangements A Steering Committee for the Myanmar Poverty and Living Conditions Survey was established in July 2015 by the President’s Office. The Ministry of Planning and Finance was represented by the Chair and Secretary, and the Ministries of Health, Education, Agriculture and Rural Development, Livestock and Fisheries were represented by members. Representatives from the development partner community participated as members of the Technical Working Group and Steering Committee. The Technical Working Group included representatives from the United Nations Development Programme (UNDP), Asian Development Bank (ADB), United Nations Children’s Fund (UNICEF), World Food Programme (WFP), International Labour Organization (ILO), United Nations Population Fund (UNFPA) and International Growth Centre (IGC). The Steering Committee included representation from the World Bank, UNDP and ADB. 8 A Revised and Rebased Measure of Poverty reflecting living conditions in 2015 9 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY This section presents a revised and rebased estimate, using 2015 standard of livings to base the basket of the poor. The recommended revision is timely from an international perspective, aligned with the end of the Millennium Development Goals and the subsequent start of the more ambitious Sustainable Development Goals. The update is also well timed from a national perspective: elections in late 2015 marked a historic change in administration that defines a new era in Myanmar’s democratic process. The methodology used has been revised over previous methodologies, to reflect updates in poverty measurement techniques. Ensuring the rigor and technical soundness of the poverty measure is a necessary precondition to a meaningful debate around poverty issues and in the development of a coherent strategy for poverty reduction. In this section, we provide a detailed description of the methodological steps and assumptions followed in the construction of a poverty line and measures for Myanmar using the 2015 household survey data.2 The layout of this section follows the four steps involved in constructing a measure of poverty for Myanmar: Step 1: The first and most important step in poverty measurement is the construction of a comprehensive and replicable consumption aggregate. Since this is the measure on which all subsequent steps are reliant, this is a fundamental step. The guiding principle in defining a comprehensive and credible measure of household welfare is that it should create a reliable ranking of individuals. Step 2: Second, we apply an appropriate poverty line to this aggregate that reflects the basic minimum needs of society. This captures an absolute standard of living below which a household is considered poor. The guiding principle in this step is to create a definition of poverty that is applicable to the basket of goods consumed by poor households in Myanmar and that can be consistently applied across all households in the country. Step 3: We produce measures of poverty that aggregate household welfare into a summary statistic. The poverty headcount, the most commonly used measure, captures the share of the population that is considered to be poor. Two other common measures are the poverty gap and the squared poverty gap. These second measures are more sensitive to how far a household is from the poverty line. Step 4: We conduct a series of sensitivity checks to assess how robust poverty measures and the poverty profile are to the assumptions made in the process of constructing the poverty profile. These assumptions range from the use of adult equivalence scales to the inclusion of various components of the sub-aggregates (such as health). 2 This discussion draws heavily on a number of influential texts in poverty measurement, notably Ravallion (1994), Deaton (1997), Deaton and Zaidi (2002) and Haughton and Khandker (2009). 10 TECHNICAL POVERTY ESTIMATION REPORT 3.1 Step 1: Definition of New Consumption Aggregates Using the 2015 MPLCS The most commonly used indicators of welfare are either based on household consumption or on household incomes. The choice of welfare indicator depends on the availability and quality of data from household surveys, as well as on socio-economic conditions in the country. In the context of Myanmar, a welfare aggregate based on consumption expenditures is likely to provide a more accurate indication of a household’s well-being than an aggregate based on income.3 There are two clear reasons for this. First, income typically fluctuates across seasons, while consumption is more likely to remain stable as households shelter themselves from shorter-term fluctuations in incomes. Second, and more practically, households are more likely to accurately recall what they have consumed than what they have earned. Household income is difficult to measure accurately in a context of high self-employment and subsistence agriculture. For such reasons, and in line with previous poverty analysis in Myanmar (MNPED et al., 2011), we use consumption expenditure as our welfare measure. As discussed in Deaton and Zaidi (2002), the component of consumption can be aggregated into four main classes, namely, (i) food items, (ii) non-food items, (iii) consumer durables, and (iv) housing. Consumption includes both goods and services that are purchased and those that are provided from one’s own production (in-kind). In this next section, we describe each component constructed including key assumptions. We then discuss how nominal household expenditure is deflated to take into consideration spatial price differences, and temporal variation over the four months of the survey and household composition. 3.1.1 Food expenditures4 Households typically consume food from various sources, from market purchases, to producing them at home or receiving them as a payment for labor or as a gift. The food expenditure aggregate captures all food consumed by a household, regardless of its source. Estimating food expenditures requires information on both the quantities that were consumed and the prices of these items. The food module of the MPLCS, conducted in 2015, asks households to report consumption for foods in twelve broad categories: (i) rice and cereals; (ii) pulses, beans, nuts and seeds; (iii) roots and tubers; (iv) meat, dairy and eggs; (v) fish and other seafood; (vi) vegetables; (vii) fruits; (viii) oils and fats; (ix) spices and condiments; (x) other food products; (xi) alcoholic beverages; and (xii) food consumed away from home. 3 Ravallion (1994), Deaton (1997), and Deaton and Zaidi (2002) provide a more detailed theoretical and empirical introduction to identifying the welfare aggregate. We note that the narrow focus on tangible components of living standards does not imply that we regard other dimensions that are more difficult to capture in a single monetary measure, such as access to improved water sources or sanitation, as less important. These dimensions need to be carefully scrutinized separately. 4 We tested the sensitivity of our food aggregates, and subsequently the poverty estimates, to differences in underlying assumptions in unit values, specifically the source of prices and the price imputation approach used. The descriptive analysis presented below puts forward the final and preferred approach described in the text; the alternative assumptions are described in Annex A1. 11 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Quantities Two principal questions capture household food consumption and are the foundation of the food aggregate: (i) Was a food consumed by the household over the last 7 days? (ii) How much was consumed during this time frame? Households were allowed to report quantities using the measurement unit with which they felt most comfortable. This varies between foods. For example, mangoes are typically reported by the number of fruit while rice is reported in condensed milk cans. We convert all quantities and expenditures on food items to a uniform reference period. Before applying prices, quantities are converted from non-standard units to a standard unit (kilograms). The approach used depends on whether units have standard imperial or metric magnitudes, or are “non- standardized”. To convert imperial and metric units to kilograms, the conversion factors were derived from Appendix 5 of the 2009/2010 Integrated Household and Living Conditions Assessment (IHLCA-II) Technical Report (MNPED et al., 2011b). Table 3.1 Conversion to kilogram Standard unit Kilogram 1 gram 0.001 kg 1 pound 0.453592 kg 1 Kyattha 0.0163293 kg 1 Viss 1.63293 kg To convert non-standard volume and count units (i.e. pyi, number, bundle) additional information was collected when conducting the survey. Available information includes: (i) weights of non-standard volumes and count units collected in the community questionnaire using scales during market observation; and (ii) the calculated weight of each food item reported in the household survey in non-standard units, as determined by the application of these collected conversions factors after the interview was conducted. Since both sets of data showed variation caused by outliers, we calculated for each item the national median of conversion factors determined during market observation, based on a large number of observations, and apply these conversion factors to both household and market quantities. This allows us to measure all foods in kilograms, regardless of the unit of purchase. The MPLCS asked households to identify how much of the food consumed in the last seven days was self-produced and how much was received as a gift or in-kind transfer (e.g. payment for labor). This distinction of source does not enter into the consumption aggregate and therefore does not affect poverty estimates. It is however useful information for welfare analysis, allowing examination of how many households are reliant on own-production to meet their dietary needs. 12 TECHNICAL POVERTY ESTIMATION REPORT Prices Prices are needed to estimate how much a household consumes in monetary terms. The MPLCS has two sources of information for prices. First, households report the total value of purchases during the past 30 days for each item, along with the quantity purchased. Dividing expenditures by quantities gives “unit values”, i.e. implicit prices. These unit values are influenced by quality choices—someone who buys better quality fish will pay more per unit than someone who buys lower quality fish. Second, a price questionnaire was fielded as part of the community module. The price questionnaire measures prices in the markets used by survey households. While in principle this provides a direct measure of goods, the enumerators do not make purchases, potentially resulting in differences between measured prices and those paid in actual transactions. We use both household and market sources to value the quantities consumed in the last seven days. Substantial sensitivity analysis of prices was undertaken before settling on the approach that was used, since they are a key variable in the estimation of food expenditures and subsequently of poverty. Figure 3.1 Methodology of price calculation Community Questionnaire Market prices (obtained in enumeration area of household) Quantities consumed Value of goods or multiplied (past 7 days) equals consumed Household Questionnaire by Total quantity, quantity (past 7 days) from home production, Unit values = quantity from gifts cost of purchases (past 30 days) quantity purchased (past 30 days) Prices can be reported in both standard units and non-standard units. This occurs where households in the enumeration area consumed an item in multiple different units. For example, one household reports consuming rice in condensed milk cans while the other reports consumption in pyi. Where both information in standard and non-standard units is available (756 cases), we use the market price derived from non-standard units, under the assumption that this more accurately represents household purchasing behavior. Price Imputation To calculate (impute) prices when necessary, we use a mix of household- and community-level data. Household unit values can only be calculated for those households that report consuming the item over the last 7 days and purchasing the item in the previous 30 days. A household may consume an item but not purchase it for a variety of reasons. For example, a household may consume certain items exclusively from home production or a household may consume some non-perishable items such as oil or spices over a long time period. 13 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY In the instances where we do not have household-level prices, we replace missing household unit values for consumed items with unit values from the local market. We do not however have prices for all goods at either the household- or local market-level. Market prices could be missing because, for example, a particular item was not offered on the day the community survey was conducted in a particular market. In a very limited number of cases both market prices and unit values can be missing because merchants or households report non-standard volume or count units that cannot be converted into a common standard unit. If both household and local market prices are missing, we use two strategies to find feasible proxies. First, we use an approach that follows the recommendation of Deaton and Zaidi (2002). This approach is based on imputing the median price or unit value from increasingly larger geographic levels of aggregation. We replace missing unit values of an item with the median unit value of non-missing observations if we have at least three observations of a price for this item in a given geographic level. Community-level prices are treated as one observation. For example, if all enumerated households in a village tract report consuming rice but rice prices are only captured in one household and in the community module, this would count as two observations coming from this village tract. If there are not enough non-missing unit value observations in the same enumeration area, we repeat the procedure at the level of wards/village tracts, townships, districts, states or regions, and finally at the national level. Note that we differentiate between rural and urban areas, for example replacing rural missing values first with the nearest rural observations before moving to higher levels of aggregations. We examine the sensitivity of poverty and inequality measures to our approach to measuring prices by using different combinations of household and community prices. We also examine the use of a nearest neighbor approach which involves matching a community with a missing price to the most geographically proximate community with a reported price, to impute the nearest enumeration areas with prices for an item, as an alternative to the geographic level approach. Other foods consumed For each of the twelve broad food categories covered in the MPLCS, the respondent is asked to report consumption of pre-defined sub-categories. Respondents are also allowed to report the consumption of items that fall outside of the listed sub-items. For example, a list of 24 vegetables was specified in the household survey, but if a household ate a vegetable that fell outside of this list they could report it under “other”. The items consumed under “other” were specified to some level of detail in the processed data, allowing for detailed analysis and post-coding as necessary. Items falling under the “other” categories are more heterogeneous than specified items. The share of “other” categories in the total number of items consumed is 2.12 percent.5 Of these “other” items, 20 percent are “other peas”, about 30 percent “other vegetables”, and about 17 percent “other oil and fat”. We looked at the distribution of unit values per item and found them to be unproblematic. Thus we applied unit values to those items in the same manner as specified items. 5 A detailed fish and seafood section is included in the MPLCS. This section includes commonly found river and sea fish as well as non-specified river and sea fish that are separated by their length, for example “other medium river fishes between 5 and 10 inches”. These fish categories are not included in our assessment of “other” items due to the level of detail and disaggregation they include. Fish in the “other” category are separated into 12 different size, source and processing groups that are important determinants of calorie content and one broad “other” category. The disaggregated groups provide detailed information that can be used for assessing nutrition content. For example, dried fish has a higher calorie content than fresh, while river and sea fish differ quite substantially in oil and fat content. We have 12314 reports of fish products of which 3011 are reports within the “other” fish categories. Of these 3011 “other” reports, only 116 reports are within the broader “other” category. 14 TECHNICAL POVERTY ESTIMATION REPORT Foods away from home A final component of total food consumption is the total value of meals consumed outside the household, for example, meals eaten at a friend’s house, at their employer, in a tea shop or at a restaurant. Respondents are asked to report whether they have taken any meals outside the house in the last seven days. If so, they are asked whether they consumed breakfast, lunch, dinner or snacks away from home. After capturing this information, they are asked how much was spent in total on each of these meal types and also the value of any meals received in kind. Since households report total expenditure on foods eaten away from home, there is no need for imputation of unit values. Outlier correction We correct outliers on an item-by-item basis. To correct outliers in unit values, we replace unit values below (or equal to) the 1st percentile and above (or equal to) the 97th percentile to missing and impute these as described below. We choose a smaller percentile at the lower end of the distribution, as the distribution is right skewed. For outlier correction of quantities, we replace outliers by item (and variant) above the 97th percentile per capita with the value of the 97th percentile per capita to keep information on relatively large amounts consumed. Outliers in rice consumption were corrected aggregated over all rice varieties.6 Descriptive Analysis Figure 3.2 presents median spatially deflated food expenditure per adult equivalent per day broken down by quintile, agro-ecological zone and urban and rural areas. It is notable that per adult equivalent expenditures per day are similar and low in the bottom half of the food expenditure distribution in Myanmar and subsequently these households are likely to have similar spending patterns. Mean expenditures are 476 kyat per adult equivalent per day for the bottom quintile of the food expenditures distribution, 731 kyat per adult equivalent per day in the second quintile and 1094 kyat per adult equivalent per day for the mean food expenditures in Myanmar. Food expenditures are higher in urban areas than in rural areas, and highest in the Delta region, likely reflecting higher spending in the urban areas of Yangon. This is not merely reflection of higher prices in urban areas—although prices for the same food products do tend to be higher in urban areas rather than in rural—but a reflection of higher total expenditures and a higher quality food basket. 6 We tested the sensitivity of the poverty estimates to an alternative median correction approach. 15 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Figure 3.2 Mean food expenditure per adult equivalent per day 2500 Kyat per adult equivalent per day on food 2040 2000 1500 1266 1206 1214 1038 1048 1094 959 976 985 1000 731 476 500 0 1 2 3 4 5 Hills & Mountains Dry Zone Delta Coastal Urban Rural Union Spatially deflated food quintile Agro Zone National Figure 3.3 shows the share of expenditures devoted to these various food products by quintile. The share of expenditures on rice and pulses drops as one climbs the food expenditure distribution, a finding that is expected and consistent with households diversifying their food intake as they become wealthier. The share of expenditures devoted to proteins such as fish, meat, dairy and eggs rises moderately as households become wealthier. Dietary diversity is lower in rural areas than in urban. Households in rural areas spend more on rice and pulses than those in urban areas, both in absolute terms and as a percentage of total expenditures. The average rural household devotes 26.1 percent of food spending on rice and pulses, compared to 19.2 percent for the average urban household. Urban households meanwhile devote 32.2 percent of total expenditures to protein-intensive food, such as meat, dairy, fish and eggs, compared to 26.6 percent of spending in rural areas. 16 TECHNICAL POVERTY ESTIMATION REPORT Figure 3.3 Food consumption expenditure by item 100% 7% 7% 5% 5% 4% 90% 5% 7% 10% 13% 17% 80% 13% 14% 17% 18% 70% 18% 16% 15% 60% 16% 16% 16% 10% 50% 12% 12% 13% 40% 12% 14% 14% 16% 30% 18% 18% 20% 35% 29% 24% 10% 19% 15% 0% Q1 Q2 Q3 Q4 Q5 Spatially deflated food quintiles, national Rice, pulses and nuts Meat, Dairy and Eggs Fish and Seafood Vegetables, roots, fruits Spices and other Food away from home Oils and Fats The share of expenditures on food away from home is 3.7 times higher in the top quintile compared to the bottom quintile of households. There are relatively more urban dwellers in the top quintile than the bottom— 10 percent of those found in the bottom quintile are in urban areas, compared to 80 percent of those in the top. The higher share of food away from home is partly a reflection of differences in the urban-rural composition of food; in urban areas where individuals may work far from their homes, and where tea shops and food stalls are readily found, all households tend to spend more on food away from home. 17 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Figure 3.4 Urban and rural food consumption expenditure by item 100% 5% 5% 3% 6% 6% 90% 7% 10% 13% 14% 24% 80% 13% 14% 14% 18% 70% 16% 14% 16% 60% 17% 17% 50% 12% 13% 15% 12% 40% 14% 16% 15% 20% 30% 19% 20% 20% 19% 28% 22% 19% 10% 16% 11% 0% Q1 Q2 Q3 Q4 Q5 Spatially deflated food quintiles, urban Rice, pulses and nuts Meat, Dairy and Eggs Fish and Seafood Vegetables, roots, fruits Spices and other Food away from home Oils and Fats 100% 4% 8% 7% 6% 5% 90% 4% 7% 8% 12% 15% 80% 13% 14% 15% 17% 17% 70% 16% 15% 16% 60% 15% 9% 16% 50% 11% 13% 10% 13% 40% 14% 13% 14% 30% 17% 17% 20% 37% 31% 26% 10% 21% 16% 0% Q1 Q2 Q3 Q4 Q5 Spatially deflated food quintiles, rural Rice, pulses and nuts Meat, Dairy and Eggs Fish and Seafood Vegetables, roots, fruits Spices and other Food away from home Oils and Fats 18 TECHNICAL POVERTY ESTIMATION REPORT 3.1.2 Non-food expenditures In addition to food expenditures, our expenditure aggregate includes non-food items that households require to achieve a basic minimum standard of living. Non-food items span various goods and services, including expenditure on energy and fuel, education, transportation and clothing. The MPLCS collects information on a wide range of non-food items such as energy, water, items of daily use and cosmetics, apparel, transportation, and telecommunications. In addition, the survey collects data on health and education expenditures for all household members. To account for different purchasing patterns of various items, data is collected for different reference periods. Because non-food items are largely heterogeneous goods, the MPLCS only collects data on the total value of non-food purchases, not on the quantities of items. The difficulty of the non-food component lies in the choice of items to include. This choice depends on data availability and ultimately on the analytical objective of the welfare measure. We follow three guiding criteria to determine whether an item is included or excluded: (1) Does the expenditure related to the particular item represent an investment with the expectation of a future welfare increase (such as schooling), as opposed to consumption that more immediately increases welfare (such as soap or petrol)?; (2) Does the particular item represent a large but infrequent ("lumpy") expenditure (such as marriages, dowries, births and funerals) and, if so, how can we treat it in the consumption aggregate?; and (3) Is the expenditure in any way related to household welfare? Such items include taxes and fees paid, and transfers to other households. Expenditures that might not reflect an increase in household welfare (such as health spending) are treated in more detail below and expenditures on durable consumption goods such as household appliances are treated separately in the next section. Following these inclusion criteria, we exclude the following items from the non-food aggregate: - Weddings and funerals - Expenditures on gold, jewelry, gems and precious stones - House repairs and expenses, including property taxes - Transfers to other households After choosing the appropriate list of items to be included (Table 3.2), we calculate the total value in kyat for each item by adding cash expenditures and the value of the items that the household received in-kind for consumption. Spending is converted to a uniform annual reference period by multiplying an appropriate factor. We then sum spending across items for each household. In composing the non-food aggregate, we use total expenditures on education as reported by households. Households report total expenditures and in some cases disaggregated data by category of expenditure (school fees, donations, uniforms, textbooks, coaching/tutoring, room and board, transportation, pocket money, and other costs). For households that report disaggregated data but where the sum of sub- aggregates is different from total reported expenditure, we use total expenditures on education. An area of some debate in the field of poverty measurement is the inclusion of health expenditures. An argument commonly made for excluding health expenditures is that they often reflect a “regrettable necessity” that does little to increase household welfare (Deaton and Zaidi, 2002). However, not including health expenditures would not capture differences between two individuals, one of whom can afford treatment, whereas the other one cannot. 19 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Table 3.2 Items included in the non-food aggregate Items Inclusion Non-food items Energy for household use √ Water √ Cosmetics and personal apparel √ Medicines and drugs not included in health module √ Local transport √ Other non-food items, incl. telecommunications √ Clothing and apparel √ of which: Gold jewelry, gems, and precious stones x Home equipment √ House repairs and expenses, incl. property taxes x Travel and trips, excl. for medical and health reasons √ Other expenses, e.g. weddings, funerals, transfers x Health and education Health, including medicines x Education, incl. vocational training and travel √ The fundamental problem with health expenditures lies in our inability to measure the cost of welfare losses from sickness. If we could value precisely such welfare losses, we could compare them to the presumed welfare gains of receiving treatment. The financing of health expenditures poses additional problems. Simply including total expenditures when some individuals receive free or subsidized treatment can be misleading. The MPLCS includes information on whether an individual received such subsidized or free treatment, but it does not break down the amounts subsidized. An assessment of demand-side programs to support health care provision conducted by the World Bank in 2015 suggests, however, that the magnitude of such programs is small in Myanmar, thus this concern is likely to be mitigated (World Bank, 2015). Estimating the elasticity of health expenditures with respect to total expenditures can help to determine whether they should be included in the non-food aggregate. A high elasticity would support inclusion, while a low elasticity would suggest exclusion. The elasticity of health expenditures in Myanmar is 0.73, a result that can be contrasted with the higher elasticity of education expenditures with respect to total expenditures (Table 3.3). 20 TECHNICAL POVERTY ESTIMATION REPORT The low elasticity of health expenditure signals a reasonably strong case for excluding it from the consumption aggregate. Table 3.3 Health and education elasticity of expenditure Health Elasticity Education Elasticity Country Year Elasticity t-statistic Elasticity t-statistic Vietnam 1992-93 0.85 33.2 1.35 46.8 Nepal 1996 0.75 20.9 1.65 43.5 South Africa 1993 1.14 58.7 1.32 67.2 Brazil 1996-97 0.85 31.0 1.25 47.9 Myanmar 2015 0.73 10.5 1.11 22.9 Source: Deaton and Zaidi (2002) It should be noted that excluding health care from the aggregate does not imply that health care expenditures are not important and shouldn’t be analyzed – to the contrary, we would recommend that health is extensively analyzed in Myanmar given the high level of out of pocket spending that can be seen in the MPLCS as well as the substantial spending on health care by the poor. Analysis of health care spending in the poverty assessment signals that poor health is linked to difficulties for households, from needing to take often high interest loans to cover health care expenditures to limiting labor force participation. The findings in the poverty assessment uphold the decision not to include health care in the aggregate: these expenditures are associated with shocks that limit household welfare. They also signal that there is a great need to do further study on the implications of poor health for poverty and well-being in Myanmar. Descriptive Analysis Household resources on non-food items are spread across a diverse range of items. Necessities such as energy, personal apparel, and cleaning and sanitary products, dominate the baskets of the households spending the least on non-food expenditures. Among these households, 36 percent of total non-food expenditure is devoted to energy, which includes electricity from the national grid, electricity from a private source, firewood, fuels and candles.7 The share of non-food expenditures devoted to energy is substantially higher in rural areas than in urban, on average as well as among richer and poorer households. The share of non-food resources devoted to energy in rural areas remains above 24 percent in all households, while in urban areas it ranges from 12 percent for the best off households to 31 percent among poorer households (Figure 3.5). Relatively few households in rural areas are connected to the public electricity grid, resulting in a diversity of spending to alternative sources of energy. The share of non-food expenditures devoted to transportation (excluding maintenance costs) rises substantially across the expenditure distribution, from 4 percent to 18 percent. The increase in spending on transportation can be seen in both rural and urban areas. 7 We do not impute the value of gathered firewood, an important source of energy among poor households. 21 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY The rise in education spending across the distribution – from 17 percent to 22 percent of the total non- food budget- reflects higher spending in richer urban households compared to more limited spending in rural areas. The rise in education spending is particularly pronounced in urban areas, where it rises from 16 percent of non-food expenditures for the bottom quintile to 27 percent for the top quintile. Among richer urban households, education expenditures are the largest single category of non-food expenditures. Conversely, among rural households the share of total non-food expenditures spent on education remains fairly stable, between 16 and 20 percent, across the distribution although it does rise in absolute terms. Figure 3.5 Non-food expenditure composition 100% 3% 7% 7% 8% 8% 10% 17% 90% 12% 13% 15% 4% 15% 80% 6% 10% 11% 11% 70% 12% 10% 14% 18% 60% 18% 13% 11% 13% 15% 11% 50% 12% 10% 13% 17% 9% 40% 17% 17% 23% 19% 30% 22% 20% 36% 29% 26% 10% 22% 18% 26% 0% Q1 Q2 Q3 Q4 Q5 All Non-food expenditure quintile, national Energy Education Cosmetics Clothing Local transport Telephone Other 100% 6% 8% 9% 10% 7% 16% 90% 16% 16% 18% 80% 18% 19% 12% 4% 70% 10% 10% 12% 12% 15% 19% 60% 8% 16% 9% 17% 10% 11% 50% 12% 10% 10% 8% 11% 40% 16% 21% 30% 25% 23% 25% 27% 20% 31% 10% 22% 19% 16% 12% 12% 0% Q1 Q2 Q3 Q4 Q5 Urban All Non-food expenditure urban quintile, urban Energy Education Cosmetics Clothing Local transport Telephone Other 22 TECHNICAL POVERTY ESTIMATION REPORT 100% 3% 6% 7% 7% 8% 9% 17% 90% 11% 11% 12% 4% 14% 80% 5% 9% 11% 10% 11% 70% 16% 16% 9% 14% 18% 13% 60% 12% 11% 15% 14% 11% 50% 13% 9% 18% 40% 16% 17% 20% 18% 18% 30% 20% 36% 31% 28% 29% 25% 24% 10% 0% Q1 Q2 Q3 Q4 Q5 Rural All Non-food expenditure rural quintile, rural Energy Education Cosmetics Clothing Local transport Telephone Other 3.1.3 Durables Households purchase a variety of durable goods such as household appliances, televisions, radios, motor bikes, cars, etc. These goods contribute to a household’s welfare, but last longer than the time horizon over which consumption is typically measured. Most households are unlikely to make such purchases within the reference period of the survey because these items are not depleted within a one-year period, but typically offer a flow of services to households over several years. For this reason, the purchase expenditures for durable goods cannot directly be added to the consumption aggregate. Instead, the appropriate measure of consumption for such durable goods is the value of services or “user cost” that households receive from such assets. Their contribution to measured consumption should include the amount of the good that is used or consumed within the year. This is measured as the change in value of the asset during the year plus the cost of allocating money to the asset over the course of the year. For a broad range of household assets, the MPLCS collects information on the age of assets owned, the original purchase price, the current (remaining) value at which the household could sell the asset at the time of the survey, and the hypothetical price for purchasing a new model of the same item at the time of the survey. If a household owns several assets of the same type, the MPLCS records the number of functioning items owned (including ones rented to other households) and collects the aforementioned information for the most valuable item. Based on the information in the MPLCS it is possible to estimate depreciation rates, which are then used to calculate a monthly or annual stream of services. Estimating durable goods user cost is challenging in Myanmar. The 2004/05 IHLCA-I survey recorded information on durables in an approach similar to the MPLCS and found several estimated depreciation rates to be positive, partly attributing this to import restrictions that were in place at the time the survey was fielded. More generally, large and rapid price changes and rapidly changing market conditions over the past years, combined with exceptionally long usage durations and a large heterogeneity of goods, makes it difficult to accurately estimate depreciation rates that can serve as the basis for a user cost calculation. 23 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Despite the challenges in measuring durable use values, the benefit of having a more comprehensive and inclusive basket of items is likely to outweigh the cost of the added measurement error. Empirical evidence from other contexts suggests that excluding any valuation of consumer durables from the consumption aggregate can lead to a bias in poverty estimates, which in turn could lead to misleading policy conclusions. In an environment of economic opening with more widespread access to global markets and high rates of overall economic growth, consumer durable ownership can expand considerably. This expanded ownership can be a key driver of improvements in household welfare. If durable goods are excluded from the consumption aggregate, we may find poverty estimates to appear relatively stagnant while household welfare has in fact increased. There is evidence from the successive household surveys in Myanmar that durables are likely to be a growth component of consumption. The income elasticity of durables is high in Myanmar. As incomes grow, the inclusion of durables is likely to be important for capturing changes in welfare and poverty over time. Comparisons of durable holdings in the IHLCA-II (2009/10), Population and Housing Census (2014) and MPLCS (2015) also suggest that the exclusion of durables would miss an important component of changing welfare over time. Estimating use value We estimate depreciation rates using information collected in the MPLCS data, following the approach presented in Deaton and Zaidi (2002) and most commonly applied in living standards measurement surveys.8 In a first step, we estimate depreciation rates for each type of good. In a second step, we use these depreciation rates, together with a real interest rate, to determine user cost as the proportion of the current value of each good owned by the household. In cases where households report having access to several goods of the same type, we calculate a user cost estimate for the item with the highest value and multiply it by the number of items owned.9 For each good and household, we use the age of the durable good in years T, the current value of the good, Pt, and the value of the good at the time of purchase, Pt-T, and the rate of inflation for the relevant period between t and T, Πt, to calculate the good-specific depreciation rate δ as follows: To minimize the influence of extreme values and measurement error, we take the median across all households that report owning a specific asset to determine a good-specific depreciation rate. Since the ratio of the real current value and real value at purchase, , is likely to include a scarcity value as well as the depletion of the good over time, we use an alternative measure of the real value of the good when purchased, Pt-T. We notably instead use the reported value of the good if purchased new today and compare the price of the new good to that when it is T years old to estimate depreciation rates. 8 We closely follow the notation of Deaton and Zaidi (2002). To simplify notation, we omit household and item subscripts. 9 This assumes that the value of services that households receive from each item of the same type is the same as the value received from the most valuable item. While this assumption may not always hold in reality, we find that in the absence of detailed data on each item this approach of establishing a reliable welfare ranking is preferable to omitting multiple items. 24 TECHNICAL POVERTY ESTIMATION REPORT Descriptive Analysis Durable use values rise sharply across the expenditure distribution, consistent with the high elasticity of these goods with respect to per capita expenditure. The share of durables use value varies substantially across the distribution, with transportation and communications as the greatest area of absolute growth (Figure 3.6). The growth of communications is likely linked to the expansion of the mobile phone network, which was on-going at the time of conducting the survey. In early 2015, urban areas had denser and greater coverage than rural areas, and easier-to-reach and, often, better-off parts of the country were more likely to have been covered earlier. Figure 3.7 shows durables ownership by expenditure quintile. The ownership and value of vehicle expands as you move up the distribution. Forty-two percent of poorer households report owning a vehicle compared to 71 percent of richer households. Poor households own bikes and motorbikes with equal shares, while richer households are more likely to own motorbikes over bikes. Cars are only owned by the richest households, among whom 17 percent own a car. Car ownership is dominated by urban dwellers—32 percent of the top quintile of urban households report owning a car, compared to 4 percent of the top quintile of rural households. In contrast, motorbike ownership is common in both rural and urban richer households. Figure 3.6 Durables composition across durables use value distribution 100% 6% 8% 9% 8% 7% 11% 13% 90% 9% 14% 12% 16% 80% 31% 21% 16% 8% 11% 70% 15% 6% 21% 22% 60% 22% 21% 20% 50% 30% 14% 35% 40% 20% 30% 48% 53% 5% 44% 43% 20% 35% 27% 10% 23% 0% FQ1 FQ2 FQ3 FQ4 FQ5 Urban Rural Durables Use Value Quintile Bike, Motorbike, Car Computer, Phone Home assets Audio Other household 25 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Figure 3.7 Durables ownership across per adult equivalent expenditure distribution Percentage of household owning durable 90% 79% 80% 70% 60% 60% 54% 47% 50% 43% 45% 46% 37% 39% 39% 36% 35% 40% 30% 24%24% 28% 30% 23% 19% 20% 17% 10% 5% 0% Q1 Q2 Q3 Q4 Q5 Expenditure quintiles defined over spatially deflated consumption aggregate Charcoal stove Bicycle Motorbike Mobile Phone Note: Analysis conducted using household weights. Quintiles are estimated using spatially deflated per adult equivalent expenditures in January 2015 prices and using population weights. 3.1.4 Housing As with constructing the sub-aggregate for durable consumer goods, the challenge in constructing a housing sub-aggregate lies in measuring in monetary terms the flow of services that the household receives from occupying its dwelling. In an environment with complete rental markets for housing and where all households rent their dwellings, rent reported by households is the first-best measure of household welfare. In Myanmar, however, only very few households rent their dwelling. In the MPLCS, only 6.9 percent of all households and less than 1 percent of households in rural areas report actual rents. This makes it challenging to reliably use imputation-based approaches and estimate a hedonic pricing model for rents of the non-reporting majority of households. We have addressed these challenges in various ways, described in greater detail in Meyer (2015). Housing use values are typically estimated using a standard ordinary least squares (OLS) approach. Given that the literature finds that the use of the wrong retransformation method or no transformation method at all can lead to appreciable biases in estimation, we explore alternative methods. One such alternative is the use of a generalized linear model (GLM), which is a class of models that allows for more flexible error distributions than classical least squares. An alternative approach that was explored is multiple imputation (MI), a flexible technique that offers an alternative to dealing with missing data. Originally developed by Rubin (1976) to address survey-non-response, it has become more widely used in a variety of applications (Rubin 1987, 1996). In MI approaches, missing values are replaced with a (relatively small) number of simulated alternatives.10 10 See Schafer (1997) for an introduction. 26 TECHNICAL POVERTY ESTIMATION REPORT Our preferred approach is to use a generalized least squares model with log links. This approach was selected for its flexibility in modeling heteroscedasticity, for its transparency and replicability and for its improvement in accuracy over OLS. An added and key advantage of using GLM with log links is that we can skip any retransformations.11 We use actual and hypothetical rent for observations that are not exceeding the largest actual rent in the sample. Assuming a normal (Gaussian) distribution of the dependent variables, we estimate a GLM using maximum likelihood and with the following independent variables: state/region, stratum (rural/urban), log number of rooms, type of dwelling, construction material of floor, type of toilet facility, source of electricity, and main fuel used for cooking. Figure 3.8 plots predicted rents using the GLM versus actual mean rents by stratum and by agro-ecological zone. Predictions from GLM are shown with a 95 percent confidence interval using heteroscedasticity-robust standard errors. As can be seen, the reported rental values lie, with two exceptions, in the confidence intervals of the predicted rents. In rural areas, the model tends to predict lower values than are reported while in urban areas there is no clear pattern. Figure 3.8 Actual versus imputed rent from generalized linear model Actual versus imputed rent Predictive margins from generalized linear model versus actual means Yangon Coastal Delta Dry Zone Hills & Mountains 0 50000 100000 150000 Rent (Kyats per month) urban (mean GLM prediction) rural (mean GLM prediction) urban (actual) rural (actual) 11 We do however note that there are substantial trade-offs in choosing various estimators, and in particular that GLM can yield imprecise estimates if the log-scale error is heavy-tailed (Manning and Mullahy, 2001). 27 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY 3.1.5 Household size and composition adjustment To estimate the share of the population of Myanmar that is poor, we need to capture individual welfare. However, consumption is measured at the household level since it is not possible to accurately measure individual consumption. To move from household welfare to individual welfare, we need to adjust household expenditure by a measure that captures the number of people in the household and to then assign a measure of welfare to each household member as an individual. A straightforward method to convert household consumption to individual consumption is to divide household expenditures by the number of people in the household. While per capita measures are the most common metric used, they must be treated with some caution as they do not take into account two possibilities: 1) First, individuals have different needs. A baby typically needs less food than a teenager and a working-age adult more than a pensioner. A manual laborer requires more food than an identically-aged office worker. 2) Second, there are economies of scale in consumption. Some goods and services consumed by a household have public good characteristics, whereby consumption by one member of the household does not necessarily diminish the amount available for other members. An example of a service with such a characteristic is lighting. Housing and durable goods are also important goods with public good elements—it costs less to house a couple than to house two individuals separately. Additionally, even though food is generally a private good there may be economies of scale due to bulk purchasing in larger households. The use of an equivalency scale that accounts for differences in the composition of household is likely to have important implications on the relative poverty of different groups. It should not however alter the level of poverty (Ravallion, 2015). For example, if we use equivalence scales that attribute low needs to children relative to adults, or that incorporate large economies of scale, we will find that there are relatively few children in poverty, but a relatively large number of the elderly. This is because the elderly often live in households that are relatively small and contain few children, while children, who never live by themselves, live in households with children. We focus on adult equivalent expenditure, using the Government of Myanmar’s nutritional norms as the base for the scale. We assume that caloric needs vary across the population according to age, but that non-food needs are the same for all age groups. In making this choice, we have examined how the poverty profile responds to different adult equivalence scales and made an assessment based on both robustness and knowledge of well-being in Myanmar from other sources (for example, qualitative studies such as LIFT 2015). The adult equivalence parameters were based on the Ministry of Health’s estimates of dietary needs for various age groups, tabulated in Annex A2. We simplify the dietary needs in two ways. First, we assume that the needs of boys and girls, and men and women are the same. Since the Ministry of Health assumes higher calorie needs for men than for women, we take the average of these needs to reach an average calorie need for a particular age group. Second, since we are unable to identify pregnant or breastfeeding women, we are unable to accommodate differences in needs for this group. The adult equivalence parameters are a weighted average of food needs that vary by age, and non-food needs that do not. Since food accounts for an average of 70 percent of total expenditures, we use a food weight of 0.7 and a non-food weight of 0.3. A vital additional issue to note in the application of adult equivalence scales is the need to normalize the poverty line. We apply the two step process recommended in Deaton and Zaidi (2002) and Ravallion (2015) in the application of adult equivalence scales: (i) we choose an equivalence parameter to represent the needs of different demographics in the population; and (ii) we normalize to adjust the poverty line after adjustment. 28 TECHNICAL POVERTY ESTIMATION REPORT Using this two-step approach, differences in adult equivalence scales do not affect the level of poverty but instead impact the profile. In the absence of normalization, however, estimated poverty declines with the application of an equivalence scale. (Ravallion, 2015). This is an important difference between the new methodology and the methodology used in MNPED et al. (2007), with non-trivial implications for measured poverty. 3.1.6 Spatial and temporal price deflation Spatial price differences are likely to be substantial in Myanmar, where transportation costs and times can be substantial. To make comparisons in the welfare of households living in different locations, it is necessary to account for these variations in prices by constructing price indices that account for these differences. Two price indices that are commonly used to capture spatial price variation are the Paasche and Laspeyres price index. A geometric average of the two indices (Fisher index, Törnqvist index) may also be used, which may come closer to the true cost of living. To make these indexes more relevant to the poor, they can be restricted to a poor reference group (Deaton and Zaidi, 2002). Another approach is to derive the spatial price index from regional poverty lines (Gibson, 2007). There are two potential advantages to this approach. In comparison to the indices mentioned above, prices are not required for all products, as poverty lines can be calculated when certain product prices are missing in an area. Second, it means that the products included in the price index reflect those included in the poverty basket. However, this approach may not keep utility constant across areas. We spatially price using Laspeyres welfare-ratio index, where the weights in the index are budget shares at the poverty line. The Layspeyres index calculates the relative price in each region for the base regions basket. As the basket is fixed, the index does not allow for households to substitute expensive products, and therefore overstates the cost of living in high-price areas. Note: k is the base region, i any other region, and j are different items of the basket. Q are quantities, P are prices. Median prices are estimated at the agro-ecological zone level, and then normalized to leave average expenditures unchanged. The spatial price deflator includes food items, housing and non-food expenditure items. The price of non-food items is not captured in the MPLCS. The non-food component of the spatial price deflator therefore relies on prices gathered for the Consumer Price Index. Since the survey was conducted between January and April 2015, we need to normalize expenditures to account for inflation over the course of the survey. We normalize all values to January 2015, and use the Consumer Price Index from the Central Statistical Organization of Myanmar to do this. 3.1.7 The consumption aggregate and its components Figure 3.9 shows the level of total consumption expenditures separated into spending on food, non-food expenditures, education, housing and durables use value. Figure 3.10 shows the share of total consumption devoted to rice, food, durables and non-food expenditures (including education). Median total consumption expenditures in Myanmar are 1644 kyat per adult equivalent per day in January 2015 prices, or approximately US$1.60 (at 1025K=US$1 on January 1st 2015). Median expenditures in urban areas are 60 percent higher those in rural areas, at 2362 kyat per adult equivalent per day compared to 1491 in rural areas. Mean expenditures show differences of similar magnitudes: 3163 kyats per adult equivalent per day in urban areas compared to 1707 kyats in rural areas. 29 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY As real household incomes grow, there is typically a change in the composition of the household budget: a decrease in the share of expenditures going to food items and an increase in the share going to non-food items. Myanmar follows this expected pattern. Food accounts for over half of consumption expenditures for the bottom 80 percent of households. Households in the bottom 20 percent devote 66 percent of the total expenditures to food. In the top quintile, the share of food drops to 46 percent but remains the largest single component of total consumption. The share of spending devoted to non-food expenditures, excluding spending on education, rises moderately across the distribution from 18 percent for the second quintile to 19 percent for the fourth quintile. Figure 3.9 Total Consumption per adult equivalent by components (spatially deflated) 5000 4500 Housing 4000 Durables Use Value Kyat per adult equivalent per day Education 3500 Non-Food Expenditures 3000 Food 2500 2000 1500 1000 500 0 Q1 2 3 4 Q5 Urban Rural National Based on per adult equivalent spatially deflated expenditures 30 TECHNICAL POVERTY ESTIMATION REPORT Figure 3.10 Share of food, non-food and durables expenditures in total consumption Share of food expenditures in total household expenditures Share of non-food expenditures in total household expenditures (Spatially deflated) (Spatially deflated) 6 8 Local polynomial fit (degree zero) 95 percent CI 5 6 4 4 3 2 2 Local polynomial fit (degree zero) 95 percent CI 0 0 11 12 13 14 15 16 11 12 13 14 15 16 Log total per adult equivalent household consumption expenditures Log total per adult equivalent household consumption expenditures Share of durables expenditures in household expenditures (per adult equivalent and spatially deflated) 3 Local polynomial fit (degree zero) 95 percent CI 2 1 0 11 12 13 14 15 16 Log total household per adult equivalent consumption expenditures 31 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY 3.2 Step 2: Definition of the New Poverty Line A poverty line is a threshold for assessing whether an individual achieves a basic minimum level of well- being. The poverty line consists of two components—the food poverty line and a non-food allowance that captures basic non-food needs. The process of defining a poverty line can be summarized by two related questions:12 i. What is the minimum level of well-being at which an individual is not considered poor? ii. How can we identify the minimum amount of money corresponding to that level of well- being? An absolute poverty line is “fixed in terms of the standards indicator used, and fixed over the entire domain of the poverty comparison” (Ravallion 1992, p 25). In other words, the poverty line is set so that it represents the same purchasing power over time. The cost of basic needs (CBN) approach is the standard approach to address these questions. The CBN approach first stipulates a consumption bundle that is deemed adequate for basic consumption needs and then estimates the cost of this specific bundle. The bundle captures two sets of basic capabilities: obtaining sufficient nourishment to maintain an individual’s health and sufficient access to non-food goods that are necessary for full participation in the society in which he or she lives. To estimate the poverty line using the CBN approach, we follow these steps: 1) Identify a nutritional requirement for good health, basing the requirement on calories per person per day. 2) Estimate the cost of meeting the food energy requirement by using a diet that reflects the choices of households whose income and expenditure places them in the vicinity of the poverty line. This is the food component of the poverty line. 3) Add a non-food component that reflects the additional needs of households to be satisfied to allow for full participation in society. 4) Estimate the total poverty line that includes both the food poverty line and the non-food component. We iterate on steps 2 to 4 until we get a final poverty line estimate, narrowing the reference group used in the vicinity of the estimated poverty line until we converge upon a final estimate. Note that the basket of the spatial price indicator is also updated in these steps. This section describes the steps taken in setting the new poverty line. 3.2.1 Nutritional benchmark We anchor the food poverty line to a caloric intake of 2238 calories per person per day. The benchmark is based on calorie recommendations from the Ministry of Health and Sports (n.d.) applied to the age and sex profile of households in the MPLCS (Table A2.4). As there are many people clustered around the poverty line, poverty in Myanmar is quite sensitive to the choice of nutritional benchmark, as discussed in Section 3.4.5. 12 This paragraph is adapted from World Bank (2015). 32 TECHNICAL POVERTY ESTIMATION REPORT 3.2.2 Estimating the food poverty line To estimate the food poverty line, we estimate the calories consumed by households and subsequently estimate the level of per capita expenditure at which the basic nutritional needs are met. Calories We first convert household food consumption into kilocalories. Our approach to setting calories deviates from that of previous poverty measures in Myanmar. Quantified goods We use the same calorie table as previous poverty estimates (found in MNPED et al 2011). This calorie table provides information on edible portions—the component of the food product that is eaten, for example the flesh of the banana without the peel. While this is the correct metric for a range of items, such as rice, pulses and fish sauce, it may overestimate the calories consumed from other sources of food such as meat, fish, vegetables and fruits that typically include a component of bones, refuse, shell or inedible peel. For these items, we apply wastage factors. Where available, we use those provided by the Ministry of Health and Sports (n.d.). Where not available we use the calories of items “as purchased” from the FAO (n.d.), which provides calories for items including any refuse in the weight. Finally, for items also not included in this list, we use information on wastage factors from the FAO 1972 list on food composition in South East Asia (FAO 1972), applied to the calories provided in MNPED (2011). For verifying our estimates, we compared values from all three sources, and, in a few cases, made case- by-case replacements. A list of calories per item, source of information and reason for changes can be found in Table A2.2. “Other” categories For each food item category, the survey instrument allows for an “other” category item—for example, “other cereals or grains”. These account for 5 percent of total consumption expenditures. Table A2.3 shows “other” food item categories, data source for caloric content of the food item, and number of cases in which information is provided for such categories in the MPLCS data. Calorie content for these items were sourced from the calories tables used in previous poverty analysis (MNPED et al, 2007) and from FAO (2015). The estimation of caloric contents for remaining “other categories” is based on country-specific information of the FAO (n.d.). Food away from home and prepared food The MPLCS asks respondents if they have eaten any meals away from home and, if so, the type of meal and the value of the meal both in cash and in kind. We use the caloric intake per kyat for all items consumed at home from that household to estimate a caloric intake for these meals; the assumption here is that the foods consumed outside the home are similar to those consumed inside the home. For prepared food eaten at home we apply the same methodology.13 13 An alternative approach that was tried was to use data on food consumed outside the home in the IHLCA-II, which asked respondents to report consumption of a detailed list of items eaten away from home. We identified meals from the IHLCA-II list that are usually eaten for breakfast, lunch, dinner and as a snack. We then estimated average calories per kyat for each of the foods that fall into these meal categories, weighted by the share of expenditures devoted to different items. This value was inflated to 2015 prices using temporal price deflators per agro-ecological zone, retrieved using population-weighted averages of state/region-level Consumer Price Index. The average calories per kyat from this method were substantially lower than those eaten at home, likely a reflection of the fairly limited items included in the IHLCA (e.g. there were no rice based dishes, even though rice constitutes the majority of calories consumed within the house). 33 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Descriptive Analysis The low food expenditures in the bottom quintile in Myanmar is mirrored in calorie consumption. Within households in the bottom quintile of total expenditures, individuals consume an average of 1959 calories per adult equivalent per day, compared to an average of 2463 nationally (Figure 3.11). The median in the nation, 2388 calories are consumed per adult equivalent per day. The lowest calorie consumption occurs in the Hills and Mountains, where individuals consume an average of 2255 calories a day. Approximately 41 percent of households consume less than 2238 calories per adult equivalent per day. Calorie consumption in urban areas is lower than that in rural areas, but expenditures are higher in urban areas. The difference between calorie consumption in rural and urban areas reflects multiple factors, including higher physical activity levels in rural areas linked to manual labor. Figure 3.11 Calories per adult equivalent 3000 2,831 2,726 2,509 2,604 2,509 2,507 2,512 2,463 Calories per adult equivalent per day 2500 2,292 2,255 2,097 1,959 2000 1500 1000 500 0 1 2 3 4 5 Hills & Mountains Dry Zone Delta Coastal Urban Rural Union Spatially deflated total quintile Agro Zone National Figure 3.12 shows the share of calories from various food groups. Although average per capita calorie consumption in urban areas is lower than in rural areas, the calorie composition of urban areas reflects a richer basket that includes a greater share of calories coming from more nutrient-rich meat, dairy, eggs, fish, vegetables and oils. Rural residents consume 65 percent of their calories from rice, compared to 55 percent among urban residents. The bottom quintile consumes 72 percent of calories from rice and only 5 percent of calories from protein, fat, calcium and amino-acid rich meat, eggs, dairy and fish. There is a stark and clear gradient in the share of calories sourced from rice compared to meat and other proteins. The share of calories from rice drops monotonically across quintiles, while the share of meat and others rises from 5 percent of calorie consumption to 11 percent. The share of calories from oil and fats remains constant across the expenditure distribution, at approximately 11 to 13 percent of total calories. 34 TECHNICAL POVERTY ESTIMATION REPORT In urban areas, higher food expenditure per capita and adult equivalent and lower calorie consumption of reflects a slightly greater share of calories coming from food away from home and from protein- intensive foods. Drawing from the experience of neighboring countries, the share of expenditures and calories coming from food away from home will continue to rise as Myanmar’s population prospers. It is therefore vital in further work on poverty in Myanmar to understand better items consumed from food away from home. Figure 3.12 Share of calories from different food groups Poor 71% Not Poor 58% Rice, pulses and nuts Q5 51% Oils and fats Q4 58% Meat, Dairy, Eggs, Fish Q3 63% Vegetables and fruits Q2 68% Food away Q1 72% from home Spices Urban 55% Other Rural 65% Union 62% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Reference basket and food poverty line We calculate a consumption basket in quantities and calories per item. As consumption patterns and the cost of calories vary with welfare, we use a reference group close to the poverty line. In our first iteration, we use the second quartile of the welfare distribution. We estimate mean quantities consumed by the median household in the reference group. For food away from home, we calculate calories only, as quantities are not available. We iterate on this basket until the poverty estimate converges, defined using a 3 decimal point gap between estimates. The consumption basket is priced using the mean price of the item by strata (agro-ecological zone). Food away from home is priced with median kyat per calorie, by strata. The food basket is scaled up or down to the quantities needed to consume 2238 calories per capita with the same food composition. The food poverty line captures the expenditures needed to meet the minimum caloric needs based on the reference basket of the poor. 3.2.3 Non-food Allowance In contrast to the absolute benchmark of calories used to set the food poverty line, there is no readily available absolute benchmark for setting non-food poverty lines. Following the cost of basic needs method, we estimate the non-food allowance for the subset of households whose per capita expenditure is relatively close to the poverty line. 35 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY We follow the approach of Ravallion (1998) in examining what people close to the poverty line consume in terms of non-food. This approach estimates an upper and lower bound, and an average poverty line: A lower bound of the poverty line is determined by observing households whose total expenditures equal the food poverty lines. These households displace some of their food consumption with necessary non-food consumption. This non-food consumption can be therefore seen as the minimum basic need of non-food consumption. An upper bound of the poverty line can be defined as the total non-food expenditure of households whose food consumption equals the food poverty line. The average poverty line is calculated as the average of both poverty lines. There are two approaches to estimating the non-food component of the poverty line: a parametric approach and non-parametric approach (Ravallion, 1994, Ravallion, 2001, Gibson, 2007). We apply the non-parametric approach for estimating lower and upper poverty lines. We verified the consistency of the estimates to the use of the parametric method, which allows for the inclusion of demographic covariates. The estimates are statistically indistinguishable. 3.2.4 Estimation and iteration approach We use an iterative approach to estimating the poverty line and the food poverty line, revising the reference basket used to benchmark the food poverty line until the national poverty rate remains unchanged. The process is illustrated by Figure 3.13 below. In the first iteration, we start by anchoring food preferences to the basket of foods consumed by households in the second quartile of the food expenditure distribution. We estimate the food poverty line based on the food preferences of this group and then add non-food expenditures using the approach described above. This gives us the poverty line produced from this first iteration. We estimate the fraction of the population who are poor using the first iteration poverty line. We use this estimate of poverty to refine the reference basket of the poor. Instead of using the second quartile to define the reference basket, we take households living in the vicinity of the first iteration poverty line. In the second iteration, we use the food preferences of households living within 10 percentage points of the first iteration poverty line to define the food preferences upon which we define the second iteration food poverty line and use this to add non-food expenditures, giving the second-iteration poverty line This second iteration poverty line is then used to define the reference food basket of the poor, and so on and forth. We continue to refine the food basket until the difference in the poverty rate between successive iterations is less than 0.01 percentage points. 36 TECHNICAL POVERTY ESTIMATION REPORT Figure 3.13 Example of poverty line estimation process ● Reference group: 2nd quartile. Iteration 2 ● Take the reference group's food expenditure basket to capture the food choices of the poor. ● Reference group: people living 10% either ● Estimate the poverty line and the poverty side of poverty rate - iteration 1. rate - iteration 1. Example: reference group 25-45 ● ● Example: Poverty rate = 35% percentiles ● Take the reference group's food expenditure basket to capture the food choices of the poor. ● Estimate the poverty line and the poverty rate - iteration 2. ● Compare. Stop if rate from iteration 1 is close to iteration 2. If not, continue. ● Example: Poverty rate 2 = 33%, difference is 2 percentage points. Continue Iteration 1 until difference is less than 0.01 An additional step in the iteration process is to convert the food and consumption aggregate from nominal to real, through spatial deflation. In the first iteration, we use nominal consumption to estimate poverty. During the iterative process we update the basket of the spatial price deflator using the basket of the poor to benchmark the spatial price index. In the second iteration, we use the food poverty basket from the first iteration to spatially deflate the food aggregate and further include non-food price deflators for housing and non-food expendables to spatially deflate the full consumption aggregate. We estimate the expenditures on housing and non-food items among households whose spatially-deflated per capita food expenditure is close to the poverty line. We use a weighted average of expenditures of individuals on either side of the line, where those individuals whose expenditures lie closer to the line are given a greater weight.14 14 We include individuals whose food expenditures lie within 10 percent of the line. The individuals for whom food expenditures lie within 0 to 2 percent of the line are given a higher weight than those with expenditures between 8 and 10 percent of the line. 37 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY 3.3 Step 3: Poverty Estimates We now turn to the poverty line and estimates. We first examine the poverty line, showing the iterative procedure and discussing the item and calorie content composition of the line. We subsequently turn to the poverty estimates produced. 3.3.1 The poverty line The poverty line and rate was estimated using an iterative procedure, as described in the preceding section. Convergence was attained after 5 iterations. As can be seen in Figure 3.14, the estimated poverty rate declined over these iterations as the reference basket and spatial price index were adjusted to accommodate the preferences of households around the poverty line. Figure 3.14 Convergence of poverty line and headcount Poverty line, Poverty headcount, by iteration by iteration 1400 0.5 0.45 1300 0.4 average poverty line rural 1200 0.35 0.3 Headcount Kyats 1100 0.25 0.2 1000 0.15 900 food poverty line 0.1 urban 0.05 800 0 1 2 3 4 5 1 2 3 4 interation Iteration 3.3.2 Basket of food at the poverty line The food poverty line is estimated to be 850 kyat per adult equivalent per day using January 2015 prices. The food poverty line only includes the food items needed to attain basic minimum food needs; as such it is lower than the total poverty line. Since earlier poverty estimates in Myanmar used adult equivalence scales without adopting a normalization approach, a direct comparison of food poverty lines is difficult to make. Our estimates suggest that 9.8 percent, or roughly one in ten individuals in Myanmar, suffers from food poverty. Food poverty is defined as a household having total (food and non-food) expenditures below the food poverty line. This is a form of severe hardship, which captures the basic minimum expenditure needed to meet minimum calorie requirements. Food poverty is predominantly a rural phenomenon: 12.5 percent of the rural population fall into food poverty compared to 2.7 percent of the urban population. Food poverty incidence is estimated to be highest in Coastal areas and the Hills and Mountains, where respectively an estimated 19.1 and 15.9 percent of individuals are estimated to fall into food poverty. Although food poverty 38 TECHNICAL POVERTY ESTIMATION REPORT is estimated to be lower in the Delta and Dry Zone, at 6.9 and 7.4 percent respectively, these figures still suggest substantial incidence of deprivation. While we cannot make direct comparisons of the poverty line over time, we are able to compare the calorie content of the poverty line over time, subject to the caveat that the food items included in the baskets diverged slightly. Both our estimations and those of MNPED (2011) excluded alcohol from the basket of the poor. MNPED (2011) also excluded non-quantified items, such as “other meat” or “other fish”, that accounted for 15 percent of calories in the new basket using the 2009/10 IHLCA-II data by World Bank (2014). The new poverty line presented here includes a greater share of calories from oil and rice than in previous estimates of poverty in Myanmar and a lower share for meats, vegetables and dairy (Figure 3.15). The decline in calories from meat, fish, vegetables and fruits predominantly reflects the application of wastage factors that take into consideration the inedible portion of meat or fruit, as discussed in Section 3.2.2. The new poverty line applies wastage factors to meat, vegetable, fish, egg and fruit products; benchmarks established by the FAO are applied to the calories consumed (FAO, 1972). Figure 3.15 Reference food basket of poverty measures in Myanmar Calories per category Rice and cereals 2004/05 Poverty Line 100 124 1,522 168 157 128 (MNPED et al, 65 Oil and fats 2007) 2009/10 Meat, dairy, fish, eggs Poverty Line 85 106 1,292 142 133 109 55 (World Bank, Vegetables, fruits, 2014) roots and tubers Pulses, beans, nuts Revised 123 95 71 and seeds Poverty Line 1,505 260 133 81 (2017) Food and beverages taken outside the home 0% 20% 40% 60% 80% 100% 3.3.3 Composition of food and non-food at the poverty line The basket of goods among those living near the poverty line is dominated by food, with only 38 percent of total consumption expenditures devoted to non-food items. This is shown in Figure 3.16 below. Of these non-food items, the largest share of expenditures goes to non-food expendables which include necessities such as energy, cleaning products and toiletries. The share of total expenditure going to food and non-food items is similar for the bottom quintile and the median household, a finding that reflects the large share of households that are clustered at low levels of expenditures. Engel noted that, as household income grows, food spending rises but less quickly than total income—as a consequence the proportion of expenditure devoted to food declines as households become better off (Engel, 1857; Deaton and Muellbauer, 1980). Although the share of expenditures devoted to food declines slightly between the bottom quintile and the median, the small decline is consistent with high rates of poverty in Myanmar where households prioritize spending on the most basic food needs. Households at the median of the distribution have a more diverse diet that includes more protein than poorer and bottom quintile households. 39 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Figure 3.16 Composition of basket of goods among those near poverty line 100% 0.04 0.03 0.10 Durables 0.05 0.05 90% 0.10 0.05 Education 0.12 80% 0.14 Housing 0.17 70% 0.18 Non-Food 60% 0.19 Food 50% 40% 0.65 30% 0.62 0.52 20% 10% 0% Poverty Line Bottom 20 Mean 3.3.4 Estimated poverty rate, by strata We present three measures of poverty in Figure 3.17: 1. Poverty Incidence captures the percentage of the population who are poor. It does not capture the extent to which they are poor. It is commonly known as the headcount rate of poverty. 2. Poverty Intensity is a measure of the depth of poverty. The intensity of poverty is captured using the average shortfall from the poverty line. Known as the poverty gap, it is calculated as the poverty incidence multiplied by the average shortfall among the poor. 3. Poverty Severity multiplies the poverty incidence by the square of the average shortfall from the poverty line. This gives more weight to people who are further away from the poverty line, unlike poverty intensity, which equal weight to each kyat that households are away from the poverty line. Based on the revision of the poverty line and on the preferred consumption aggregate, we estimate that 32.1 percent of the population of Myanmar was poor in 2015. Poverty in Myanmar’s farms and villages is substantially higher than that in its towns and cities: 38.8 percent of the population in rural areas are estimated to be poor compared to 14.5 percent of those in urban areas. This amounts to 15.8 million poor in total, of which 13.8 million are found in rural areas and 2.0 million are found in urban areas. The estimated number of poor is based on the enumerated and estimated non-enumerated population living in conventional households, following the definition of the Housing and Population Census of Myanmar (Ministry of Immigration and Population, 2014). The poverty estimates therefore do not include the 2.35 million individuals in Myanmar living in student dormitories, monasteries, convents, barracks and other such living arrangements. We are unable to currently estimate state or region level poverty due to the small sample size of the survey. Small area estimates will be estimated through a subsequent poverty mapping exercise. 40 TECHNICAL POVERTY ESTIMATION REPORT We are however able to estimate poverty at an agro-ecological zone level and find that there is substantial geographic variation in poverty. The headcount rate of poverty in the Coastal areas is 43.9 percent, nearly double the rate of 26.2 percent in the Delta. The headcount poverty rate in the Hills and Mountains is also substantial, at 40.0 percent. The Coastal and the Hills and Mountains areas have the highest poverty intensity and severity, consistent with the substantial food poverty recorded in these areas. We note that the standard error of poverty estimates in the Coastal areas is considerable, likely reflecting the substantial diversity in these areas. Although poverty in the Coastal areas is estimated to be higher than in the Hills and Mountains areas, due to high standard errors the difference between the two zones is not statistically significant. Figure 3.17 Poverty headcount, poverty incidence and poverty intensity in Myanmar 50% 43.9% 40% 40.0% 38.8% Poverty Headcount 32.1% 30% 32.1% National 26.2% 20% 14.5% 10% 0% Mountains Coastal Hills & Urban Delta Rural Dry Ten percent of the population of Myanmar are food poor, meaning that their total consumption expenditures are not considered sufficient to cover their food needs. This measure of poverty captures a form of extreme deprivation, where even the most basic of food needs are not being met. Rates of food poverty are substantially higher in rural areas than in urban, with 12.5 percent of the rural population suffering from food poverty compared to 2.7 percent of the urban population. Food poverty rates are considerably higher in Hills and Mountains and Coastal areas, consistent also with their higher rankings in the poverty gap and poverty severity measures for both food and total poverty. 41 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Figure 3.18 Food poverty headcount, by urban and rural strata and agro-zone 25% 20% Food Poverty Headcount 19.1% 15.9% 15% 12.5% 10% 9.8% 7.4% 6.9% National 5% 2.7% 0% Mountains Coastal Hills & Urban Delta Rural Dry 3.4 Step 4: Sensitivity Analysis and Robustness Checks We estimate the sensitivity of the headcount rate of poverty to variants of the preferred consumption aggregate, and to alternative price formulations for setting the poverty line and to spatially adjusting prices. We find little sensitivity to the assumptions on how to estimate depreciation of durable goods, the inclusion or exclusion of health care and the formulation of the spatial price index. We find substantial sensitivity to the choice of calorie norm, and that the profile of poverty is sensitive to the adult equivalence scale used. Subsequent to the sensitivity analysis, consultation and analysis was conducted around the choice of calorie norm and the equivalence scale to be used in order to ensure that the profile appropriately represents the nature of poverty in Myanmar. 3.4.1 Inclusion of durables We examine the sensitivity of the poverty estimates to the inclusion of durables, and to the method of producing depreciation rates discussed in Section 3.3. We present three scenarios in Figure 3.19 below. In scenario 1, we remove durables use value from the consumption aggregate. In scenario 2, we estimate depreciation rates using the traditional approach of using the reported purchase value at time of purchase. In scenario 3, we estimate depreciation rates using the preferred depreciation method, using perceived purchase value if purchased at the date of interview. The poverty line is responsive to changes in the approach to estimating durables, but not overly sensitive. This is unsurprising given the small share of durables accounted for by individuals living near the poverty line. Durables account for 3 percent of consumption expenditures among those in the bottom 20 percent. Very little difference is seen in the use of the two depreciation methods on poverty estimates. The headcount rate of poverty remains the same at the national level, and there is a very small change in poverty rates at the urban and rural level. There is, however, a slightly larger impact on poverty estimates from removing durables from the consumption aggregate. In this scenario, poverty estimates rise by approximately 1 42 TECHNICAL POVERTY ESTIMATION REPORT percentage point. Although the limited sensitivity of poverty estimates to durables is intuitive given their relatively small importance in the expenditure aggregate for poorer households, the inclusion of durables is likely to have important implications for poverty trends and future poverty estimates. Durables typically have a high elasticity to expenditures, and are an important area for welfare growth as has been seen in the recent past across neighboring countries including India and Vietnam (Basole and Basu, 2015; World Bank 2012). Figure 3.19 Sensitivity of poverty estimates to durables aggregate used 45.0% 40.0% 35.0% Headcount poverty rate 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% Urban Rural Union Exclude Durables Traditional depreciation Modified depreciation 3.4.2 Inclusion of health expenditures As discussed in Section 3.1.2, it was decided to omit health care expenditures from our measure of the consumption aggregate. We find in our sensitivity analysis that including health expenditures raises the poverty line from 1310 kyat per adult equivalent per day in January 2015 prices to 1325 kyat. This translates into an additional 4745 kyat per year needed to cover health care expenditures at the poverty line. Including health spending additionally raises median per capita expenditures from 1676 kyat per adult equivalent per day to 1727 kyat per adult equivalent per day. The inclusion of health care expenditures reduces the headcount rate of poverty from 32.1 percent to 31.9 percent. The reduction in poverty as a consequence of including health care is potentially a reflection of health shocks that require spending on health care among poor households. An otherwise poor household that suffered a health incident on which they had to spend resources may, after the inclusion of their spending on the treatment, appear non-poor. 43 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Figure 3.20 Sensitivity of poverty line and poverty estimate to health expenditures 45.0% 40.0% 35.0% Headcount poverty rate 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% Urban Rural Union Include Education Include Health Include Education and Health 3.4.3 Spatial price index Prices are likely to vary substantially across Myanmar, where markets are unlikely to be integrated due to the geography and infrastructure. The spatial deflators used in the poverty estimation include food and non-food items. Since food prices are readily available but the prices of non-food items and housing are not, we test the sensitivity of poverty to the approaches that could be used. We control for spatial price differences in two ways. In the first approach, we additionally take into consideration food price variation, using prices estimated from the MPLCS survey. In the second approach, we take into consideration differences in food as well as non-food items and the price of housing. We estimate differences in the price of housing using a regression approach, where we factor out differences in the quality of housing units across locations. Housing accounts for 10 percent of expenditures among individuals in the bottom quintile. We find that the inclusion of housing in the spatial price deflator does not make a big impact on estimated poverty rates. Since spatial price deflators have the potential to affect the headline rate of poverty as well as the spatial distribution of poverty, we examine both level changes and its disaggregated impact. We see that there is a very small impact on national poverty—from 31.4 percent when including only food to 32.1 percent when including also non-food and housing. Urban poverty is slightly higher when housing is included, likely a reflection of higher housing prices in urban areas. The spatial dimension of poverty changes slightly when the spatial price deflator reflects a more comprehensive basket, a reflection of higher non-food and housing prices in Yangon, the Dry Zone and Coastal areas and relatively lower prices in the Hills and Mountains and Delta region. 44 TECHNICAL POVERTY ESTIMATION REPORT Figure 3.21 Sensitivity of poverty estimate to spatial price adjustment 45.0% 40.0% 35.0% Headcount poverty rate 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% Urban Rural Union Food Food and Non-Food Food, Non-Food and Housing 3.4.4 Adult equivalence scales In this section, we examine the robustness of the poverty profile to alternative approaches to converting household welfare into an individual welfare measure. Specifically, we look at the effects of using per capita expenditure and alternative adult equivalence parameters using the equivalence scales and economies of scale that were used in past analyses of poverty in Myanmar. Notably, we deviate from these previous approaches in that we normalize the expenditure aggregate prior to the estimation of the poverty line, as recommended in Deaton and Zaidi (2002). Previous poverty estimation in Myanmar employed adult equivalence scale by applying the following formula, from Deaton and Zaidi (2002) with modified notation: Where refers to the number of adult males, refers to the number of adult females; refers to the number of children. The adult male equivalence parameters are given by , the cost of a female relative to that of an adult male, and , the cost of a child relative to that of an adult male. The parameter θ captures the extent of economies of scale. Since the elasticity of adult equivalence with respect to “effective size” is θ, this implies that the measure of economies of scale is (1-θ). The parameters used to account for adult equivalence varies across the two previous poverty estimates produced in Myanmar. In the previous poverty estimates presented in MNPED et al. (2011), different equivalence scales were used for food and non-food expenditures. For the food consumption equivalence, was set to 0.9, was set to 0.7 and θ was set to 0.9. For the non-food consumption equivalence, was set to 1, was set to 0.3 and θ was set to 0.9. A child was defined as an individual aged between 0 and 14 years of age. Nutritional norms were used to rationalize the food adult equivalence scales, and the non-food equivalence scales were based on those used in countries at similar levels of development, as put forward in Deaton and Zaidi (2002). 45 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY In the alternative methodology used by the World Bank (2014) to also estimate poverty in 2009/10, the same equivalence scale was used for food and non-food expenditure. In this equivalence, no difference was applied to the needs of men and women, therefore was set to 1. For children aged between 0 and 6 years, was set to 0.5. No economies of scale were used (θ was set to 1), reflecting the limited fraction of goods with public goods properties in Myanmar’s consumption aggregate. The adult equivalence scales were based on assessments of equivalence used in countries in the region and at similar levels of development (Kozel and Kim, 2015). 3.4.5 Calorie norm The country norms used by low- and middle-income countries range from 1800 calories per capita (India) to 2800-3000 calories per capita in some African countries (Figure 3.22). The calorie benchmark used in Myanmar falls within the international norms, and within the range of 2200 to 2300 used in most countries. 46 TECHNICAL POVERTY ESTIMATION REPORT Figure 3.22 Nutrition norms used to anchor poverty lines in low- and middle-income countries Uganda Chile Cameroon Sierra Leone Malawi South Sudan Egypt Senegal Jordan Iraq Uganda Panama Colombia Nicaragua Myanmar Vietnam Honduras Mexico Paraguay Guatemala El Salvador Mozambique Ecuador Bangladesh Indonesia Timor Leste Sril Lanka India 0 500 1000 1500 2000 2500 3000 3500 Source: World Bank (2015). These calorie norms reflect the calories used to define national poverty measures based at different points in time. 47 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Poverty estimates are sensitive to the choice of calorie norm used to benchmark daily needs. The current calorie norm is estimated by applying the daily nutritional guidance from the Ministry of Health and Sports to the demographic profile of the population of Myanmar. This generates a per adult equivalent calorie norm of 2238. Since a substantial fraction of individuals live within 200 kyat of the poverty line, changes to the calorie norm used have a substantial impact on poverty rates. Using a 2100 or 2200 calorie benchmark reduces the headcount rate of poverty to 23.8 percent and 29.3 percent respectively. Similar to many low income countries, a large fraction of Myanmar’s population is clustered around the poverty line resulting in sensitivity to the choice of calorie norm. The clustering of households in the vicinity of the poverty line signals substantial vulnerability to poverty where small changes in circumstances—such as climatic shocks, illness or a good yield—can lead to significant movements into and out of poverty. Figure 3.23 Poverty estimates using different calorie norms Varying Calorie Basis of Poverty Line 1 0.8 2300: 33.7% 0.6 2200: 27.3% 0.4 2238: 32.1% 0.2 2100: 18.8% 0 6.5 7 7.5 8 8.5 9 Log total per capita household consumption expenditures Cumulative Distribution Line 2100 Line 2200 Line 2238 Line 2300 48 TECHNICAL POVERTY ESTIMATION REPORT 3.5 Poverty Trends using the new 2015-rebased estimate The imputations methods discussed in Part One of the Poverty Assessment (MOPF and World Bank, 2017a) were used to produce trends of poverty estimates between 2004/05 and 2009/10 that are comparable with those from 2015. We find that poverty declined from 48.2 percent in 2005 to 42.4 percent in 2009/10 and subsequently to 32.1 percent in 2015. Similarly, we see a decline in other measures of poverty that capture the intensity and severity of poverty, as well as is the fraction of the near-poor – those who are living within 20 percent of the poverty line. The analysis of trends using the new aggregate closely mirrors the trends seen in the World Bank (2014) aggregate, notably that poverty declined faster in urban areas than in rural and that growth in mean welfare in urban areas was higher on average than in rural areas. Figure 3.24 Estimated trends in poverty rates using the new rebased estimate 55% 48.2% 50% 42.4% 45% Headcount poverty rate 40% 32.1% 35% 30% 25% 20% 15% 10% 2004/05 2009/10 2015 Imputed Actual Both the trends and poverty estimate are explored in greater detail in the accompanaying Poverty Profile. 49 50 References 51 Basole, Amit and Deepankar Basu. 2015. "Nonfood Expenditure and Consumption Inequality in India." Economic and Political Weekly, Vol. L, No. 36, September 5, 2015, pp. 43-53 Briggs, Andrew., Taane Clark, Jane Wolstenholme, and Philip Clarke. 2003. “Missing.... Presumed at Random: Cost-Analysis of Incomplete Data.” Health Economics 12 (5): 377–92. doi:10.1002/hec.766 Deaton, Angus.1997. The Analysis of Household Surveys. The John Hopkins University Press, Baltimore and London. Deaton, A.; Muellbauer, J. 1980. Economics and Consumer Behavior. Cambridge: Cambridge University Press. Deaton, Angus., and Salman Zaidi. 2002. “Guidelines for Constructing Consumption Aggregates for Welfare Analysis”. Living Standards Measurement Study (LSMS) Working Paper 135. Washington, DC: The World Bank. Douidich, Mohamed., Ezzrari, Abdeljaouad., Van der Weide, Roy., and Verme, Paolo. 2015. "Estimating quarterly poverty rates using labor force surveys: a primer." World Bank Economic Review, pp 1-26, The World Bank. Elbers, Chris, Jean O. Lanjouw, and Peter Lanjouw. 2003. “Micro-Level Estimation of Poverty and Inequality.” Econometrica, 71(1): 355-364. Elbers, Chris., Van der Weide, Roy. 2014. "Estimation of normal mixtures in a nested error model with an application to small area estimation of poverty and inequality." Policy Research Working Paper Series 6962, The World Bank. Engel, Ernst. 1857. "Die Productions- und Consumtionsverhältnisse des Königreichs Sachsen". Zeitschrift des statistischen Bureaus des Königlich Sächsischen Ministerium des Inneren 8–9: 28–29. FAO. 1972. Food composition table for use in East Asia. Food and Agriculture Organization of the United Nations (FAO): Rome. FAO (n.d.). Nutritive Factors for Myanmar. Food and Agriculture Organization of the United Nations (FAO): Rome. Filmer, D., and LH Pritchett. 2001. "Estimating Wealth Effect Without Expenditure Data or Tears: An Application to Educational Enrollments in States of India." Demography 38, no. 115-32. Gibson, J. 2007. “A guide to using prices in poverty analysis.” Working Paper. Department of Economics, University of Waikato. Jonathan Haughton., and Shahidur R. Khandker. 2009. "Handbook on Poverty and Inequality," World Bank Publications, The World Bank, number 11985, January. Jolliffe, Dean. 2001. “Measuring Absolute and Relative Poverty: The Sensitivity of Estimated Household Consumption to Survey Design.” Journal of Economic and Social Measurement Volume 27, no. Numbers 1-2 (2001): 1–23. LIFT. 2015. “Household Survey 2013” Livelihoods and Food Security Trust Fund Manning, Willard G, and John Mullahy. 2001. “Estimating Log Models: To Transform or Not to Transform?” Journal of Health Economics 20 (4): 461–94. doi:10.1016/S0167-6296(01)00086-8. 52 Megill. 2014. “Recommendations for Developing Master Sample for National Household Surveys in Myanmar: Sampling and Estimation Methodology”, Background Note for the MPLCS. Meyer, Christian. 2015. “Imputing housing rents in Myanmar.” Background Note for the technical report. Ministry of Health and Sports (n.d.). “Nutrition information booklet.” Nutrition Unit, Ministry of Health (MoH), The Republic of the Union of Myanmar: Nay Pyi Taw. Ministry of Immigration and Population. 2014. “2014 Myanmar 2014 Housing and Population Census.” Department of Population, The Republic of the Union of Myanmar: Nay Pyi Taw. Ministry of Immigration and Population. 2014. “Provisional Results; Census Report Volume (1): 2014” Myanmar Population and Housing Census: Ministry of Immigration and Population, Union of Myanmar, August 2014 Ministry of Immigration and Population. “The Union Report; Census Report Volume 2: 2014” Myanmar Population and Housing Census: Ministry of Immigration and Population, Union of Myanmar, May 2015 Ministry of National Planning and Economic Development, IDEA, UNDP and UNOPS. 2007. Integrated Household Living Conditions Survey in Myanmar: Poverty Profile. June 2007 Ministry of National Planning and Economic Development, IDEA, UNDP and UNOPS. 2010. “Integrated Household Living Conditions Survey in Myanmar: Technical Report.” June 2010 Ministry of National Planning and Economic Development, SIDA, UNDP and UNICEF. 2011a. Integrated Household Living Conditions Survey in Myanmar (2009-2010): Poverty Profile. June 2011 Ministry of National Planning and Economic Development, SIDA, UNDP and UNICEF. 2011b. Integrated Household Living Conditions Survey in Myanmar (2009-2010): Technical Report. June 2011 Ministry of Planning and Finance and World Bank. 2017a. An Analysis of Poverty in Myanmar, Part One: Trends between 2004/05 and 2015. Ministry of Planning and Finance and World Bank. 2017b. An Analysis of Poverty in Myanmar, Part Two: Poverty Profile. Nguyen, M. C. and R. Van der Weide. 2016. “Updating estimates of poverty using an up-to-date model.” mimeo, The World Bank Ravallion, Martin. 1992. "Poverty Comparisons - A Guide to Concepts and Methods." Papers 88, World Bank - Living Standards Measurement. Ravallion, Martin. 1994. “Poverty Comparisons.” Fundamentals of Pure and Applied Economics 56. Chur, Switzerland: Harwood Academic Publishers. Ravallion, Martin. 1998. “Poverty Lines in Theory and Practice”. Living Standards Measurement Study (LSMS) Working Paper 133. Washington, DC: The World Bank. Ravallion, Martin. 2001. “ Poverty lines: Economic foundations of current practices.” World Bank, Washington D.C. Ravallion, Martin. 2015. “On testing the scale sensitivity of poverty measures.” Economic Letters, 137 (2015) 88–90 53 Ridder, Geert, and Robert Moffitt. 2007. “The Econometrics of Data Combination.” in Handbook of Econometrics, edited by James J. Heckman and Edward E. Leamer, Volume 6, Part B, 5469–5547. Handbooks in Economics. Amsterdam: North-Holland. http://www.sciencedirect.com/science/ article/pii/S1573441207060758. Rubin, Donald B. 1976. “Inference and Missing Data.” Biometrika 63 (3): 581–92. doi:10.1093/ biomet/63.3.581. Rubin, Donald B. 1987. Multiple Imputation for Nonresponse in Surveys. New York, NY: Wiley. Rubin, Donald B. 1996. “Multiple Imputation after 18+ Years.” Journal of the American Statistical Association 91 (434): 473–89. doi:10.1080/01621459.1996.10476908. Rubin, Donald B., and Nathaniel Schenker. 1986. “Multiple Imputation for Interval Estimation from Simple Random Samples with Ignorable Nonresponse.” Journal of the American Statistical Association 81 (394): 366–74. doi:10.1080/01621459.1986.10478280. Schafer, J. L. 1997. “Analysis of Incomplete Multivariate Data.” Monographs in Statistics and Applied Probability 72. Boca Raton, FL: Chapman & Hall/CRC. Shepard, Donald. 1968. “A Two-Dimensional Interpolation Function for Irregularly-Spaced Data.” In Proceedings of the 1968 23rd ACM National Conference, 517–24. New York, NY: ACM. doi:10.1145/800186.810616. United Nations Statistics Division. 1986. Sampling Frames and Sample Designs for Integrated Household Survey Programmes, National Household Survey Capability Programme, United Nations, New York. World Bank. 2012. “Vietnam poverty assessment: Well begun but not yet done” The World Bank. Hanoi Office World Bank. 2014. “Myanmar: Ending poverty and boosting shared prosperity in a time of transition: Systematic Country Diagnostic.” World Bank Group World Bank. 2015. “A Measured Approach to Ending Poverty and Boosting Shared Prosperity: Concepts, Data, and the Twin Goals.” Policy Research Report. Washington, DC: The World Bank. Yoshida, Nobuo. 2014. “Comparison of MI and ELL: Empirical results.” Paper presented at the Poverty and Inequality Measurement and Analysis Practice Group special session on multiple imputation, Washington, DC, June 18.   54 Annexes 55 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Annex A1 Variants of the Consumption Aggregate Before converging upon the preferred method of measuring the welfare aggregate, multiple approaches were examined to assess the implications of assumptions made. This annex addresses all variants, but not all variants are presented in the text assessing the robustness of the poverty estimate to permutations in the consumption aggregate. We ruled out variants by assessing whether they impacted the aggregate component and whether they produced a credible distribution and poverty profile. Food aggregate Six variants of the food aggregate were estimated. Food constitutes the majority of consumption expenditures among poor households and is the base of the poverty line estimation. As such, substantial attention was devoted to estimating permutations in the aggregate. The primary difference between the variants of the food aggregate is the prices used (Table A1.1). All variants use the same outlier correction approach described in Section 3.1.1. After an assessment of the level, composition and relationship between the variants, variant 4 was determined to most accurately capture the food expenditures of the population of Myanmar. There was relatively little difference between variants 2 through 4 in the geographic distribution of expenditures. Mean and median expenditures are highest in Yangon and lowest in the Delta across these variants. Variant 1, by contrast, produces a different ordering across zones and a lower aggregate (by approximately Table A1.1 Variants of food aggregate No Prices Imputation method 1 Unit value prices from household survey; unit values Administrative: Nearest non-missing value based on rural/urban estimated using supervisor-determined item conversion definition and administrative area conditional upon at least 3 for non-standard units. households in the area reporting a price. 2 Unit value prices from household survey; unit values Administrative: Nearest non-missing value based on rural/urban estimated using median supervisor-determined item definition and administrative area conditional upon at least 3 conversion for non-standard units. households in the area reporting a price. 3 Prices from community module Administrative: Nearest non-missing value based on rural/urban definition and administrative area conditional upon at least 3 communities in the area reporting a price. 4 Combined unit value prices from household survey and Administrative: Nearest non-missing value based on rural/urban prices from community module. Household unit values definition and administrative area, conditional upon at least 3 estimated using median supervisor-determined item households or communities in the area reporting a price. conversion for non-standard units. 5 Combined unit value prices from household survey and Distance: Nearest non-missing value based on physical distance prices from community module. Household unit values using distance weights, conditional upon at least 3 households in estimated using median supervisor-determined item the area reporting a price. conversion for non-standard units. 6 Combined unit value prices from household survey and Distance: Nearest non-missing value based on physical distance prices from community module. Household unit values using square-root distance weights, conditional upon at least 3 estimated using median supervisor-determined item households in the area reporting a price. conversion for non-standard units. 56 TECHNICAL POVERTY ESTIMATION REPORT 20 percent). Variants 5 and 6 were computationally intensive with little clear additional benefit, and were therefore excluded. An in-depth assessment of unit values and conversion factors led us to exclude these variants from the estimation. Non-food aggregate The non-food component of expenditures includes items that: (i) contribute to household well-being; (ii) are not “lumpy” in nature; and (iii) do not constitute investment expenditures. We present variants 1 through 6 below for completeness; we however narrow down our focus on variants 7 through 9 in the analysis. Variants 1 through 6 were excluded based on following the recommendations from the poverty measurement literature (Deaton and Zaidi, 2002). Table A1.2 Variants of non-food aggregate Variant Included items 1 2 3 4 5 6 7 8 9 Non-food items Energy for household use          Water          Cosmetics and personal apparel          Medicines and drugs not included in health module          Local transport          Other non-food items, incl. telecommunications          Clothing and apparel          of which: Gold jewelry, gems, and precious stones   Home equipment          House repairs and expenses, incl. property taxes     Travel and trips, excl. for medical and health reasons          Other expenses, e.g. weddings, funerals, transfers    Health and education Health, including medicines   Education, incl. vocational training and travel   Assessed in text   Preferred  57 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Durables As noted in MNPED et al. (2011), the estimation of durable use value in Myanmar was complicated by the finding of positive depreciation rates in 2004/05, attributable in part to the incomplete and imperfect market for some goods. The removal of restrictions on imports and expansion in goods available is likely to result in a decline in scarcity value and in general prices of goods over time. As such, the conventional method to estimate depreciation rates—using the ratio of real current value to real purchased value— would likely capture a decline in scarcity values for multiple goods as well as a decline in value due to depletion and usage. The conventional approach to estimating depreciation is therefore likely to overstate depreciation. Estimating depreciation rates using the cost if purchased at the time of the survey (and subsequently deflated to real purchase value using inflation rates) suffers from one drawback: goods that were originally purchased second hand would be priced today as though they were purchased new. As a consequence, we would expect to find the depreciation rate estimated from current replacement costs to be upward biased. In Section 3.1.3 we describe two options for estimating depreciation rates: using the cost at the time of purchase and the cost if purchased new today. In the table below, we show the depreciation rates under these two scenarios. Although ultimately the use of different depreciation rates did not substantially impact the poverty estimates, some tangible differences between the two sets of depreciation rates are clear. The depreciation rates calculated using the cost at the time of purchase are higher than those estimated using the cost today; this suggests that there was a component of scarcity decline in these rates. Indeed, the comparison suggests that the scarcity premium bias is likely to be greater than any bias caused by using (potentially) new rather than second-hand items. This analysis uses the second, non- conventional depreciation method, notably using cost if purchased at the time of the survey. 58 TECHNICAL POVERTY ESTIMATION REPORT Table A1.3 Comparison of depreciation rates by item Depreciation estimated using: Cost at purchase Cost new today Battery 0.53 0.50 Electric inverter 0.33 0.27 Generator 0.21 0.14 Gas stove 0.33 0.25 Charcoal stove 0.58 0.54 Hot plate 0.40 0.35 Electric stove 0.46 0.42 Rice cooker 0.40 0.35 Electric iron 0.36 0.29 Electric fan/ Air cooler 0.40 0.32 Refrigerator/ Deep freezer 0.29 0.22 Washing machine 0.30 0.23 Air conditioner 0.29 0.21 Water heater 0.26 0.21 Electric heater 0.31 0.23 Radio 0.48 0.42 Stereo/ Hi-Fi cassette (with CD player) 0.33 0.19 Color TV 0.29 0.20 Satellite dish (any type including Paid TV) 0.34 0.25 VCD/DVD player 0.37 0.29 Loudspeaker 0.27 0.20 Computer (any type) 0.29 0.20 Printer 0.24 0.17 Line telephone 0.31 0.06 Mobile phone (including SIM card) 0.45 0.29 Bicycle 0.30 0.24 Motorcycle/moped/tuk tuk 0.23 0.20 Motor vehicle 0.21 0.07 Canoe/Boat (not used for fishing) 0.18 0.22 Trishaw 0.19 0.15 Truck (6 Wheels and above) 0.12 0.06 59 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Annex A2 Estimation of Poverty Line: Calories Adult Equivalence Scales The Ministry of Health’s estimates of dietary needs are used to define an adult equivalence scale. We however deviate from scales based entirely on calorie needs in three ways. First, we treat the needs of girls and boys the same. Second, we apply an additional need factor of 1.3 to the calorie needs of a child under 3 years of age, to account for the higher quality calories needed in the first 1000 days of development. Finally, we assume that the non-food needs of adolescents are equivalent to those of adults, and that the share of expenditures from non-food items is 0.3. Table A2.1 Adult equivalence parameters Calories needs from Age in years Calorie need applied Adult equivalence parameter Ministry of Health Child 1 850 850 0.55 1-3 1260 1260 0.67 4-6 1670 1670 0.78 7-9 1800 1800 0.83 Adolescents 10-12 2400 (male) 2200 (female) 2300 0.97 13-15 2600 (male) 2400 (female) 2500 1.04 16-19 3000 (male) 2500 (female) 2750 1.11 Adult (moderate activities) 2800 (male) 2000 (female) 2400 1 Caloric adjustment of food items Earlier poverty estimates in Myanmar used calories of edible portions; this revision to the poverty estimation methodology applies the calories of food as purchased. Tangibly, food as purchased includes non-edible parts such as banana peel or the carcass of a chicken. The rationale for doing this is that household surveys capture food weights as purchased—a household will report eating a bunch of bananas or a large watermelon—and to convert this into calories consumed. We therefore apply a “wastage factor” to account for the component that is not consumed. For items such as rice and pulses, where there is no wastage factor, the calories applied are the same as those used in previous poverty estimation in Myanmar. For other items such as fish, meat, vegetables and fruits, we apply a wastage factor that tends to reduce the calories consumed. In the case of meat, the calories consumed depend on the age and fat of the animal. We reviewed the calories consumed by meat and noted that the leanest of animals were previously assumed. The calories of meat have to therefore often be revised upwards, to reflect a more average animal. 60 TECHNICAL POVERTY ESTIMATION REPORT Table A 2.2 Overview of calories per kilogram per item Original Updated Reason for revising Item code Item Label MPLCS Calories per calories Source Description calorie value kilo per kilo Rice and cereals 8001 Rice (Ngasein) 3820 3820 MNPED (2011) 8002 Rice (Emata) 3540 3540 MNPED (2011) 8003 Rice (Medone) 3510 3510 MNPED (2011) 8004 Rice (Nga kywe) 3570 3570 MNPED (2011) 8005 Kaukhnyin (Sticky 3590 3590 MNPED (2011) Rice) 8006 Other rice (local 3570 3570 Same as 8004 variety) 8007 Corn 3560 3560 MNPED (2011) 8008 Other rice and 3400 3400 FAO Wheat cereals(specify) 8009 Other rice and 3400 3400 FAO Wheat cereals(specify) Pulses, beans, nuts, and seeds 8010 Pegyi (lablab beans) 3520 3520 MNPED (2011) 8011 Pegya 3610 3610 MNPED (2011) 8012 Pe poke 3550 3550 MNPED (2011) 8013 Sadawpe (green 2470 2470 MNPED (2011) peas) 8014 Gram (Chick pea) 3680 3680 MNPED (2011) 8015 Green gram 3470 3470 MNPED (2011) (Pedesane) 8016 Penilay (Peyaza) 3400 3400 MNPED (2011) 8017 Butter Bean 3350 3350 MNPED (2011) 8018 Other peas 3575 Mean of green peas & chick peas 8019 Sesame 5820 5820 MNPED (2011) 8020 Groundnut without 5480 5480 MNPED (2011) shell 8021 Coconut 3120 1500 MNPED (2011) – Wastage factor: 52 Updated to applied wastage remove non-edible components. 8022 Other (specify): 3400 3400 FAO Pulses not elsewhere specified 8023 Other (specify): 3400 3400 FAO Pulses not elsewhere specified 8024 Other (specify): 3400 3400 FAO Pulses not elsewhere specified 61 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Original Updated Reason for revising Item code Item Label MPLCS Calories per calories Source Description calorie value kilo per kilo Roots and tubers Updated to 8025 Sweet potatoes 1300 950 FAO remove non-edible components. Updated to 8026 Potatoes 840 670 FAO remove non-edible components. Updated to 8027 Radish 260 190 FAO 1972 remove non-edible components. Updated to 8028 Taro 940 860 FAO remove non-edible components. http://caloriecount. Pemyit / Root of about.com/calories- 8029 1480 1480 bean tree winged-bean- tuber-i11599 Applied wastage Updated to FAO 1972, WF from 8030 Palm shoot 1150 577 factor of bamboo remove non-edible ministry shoot: 44 components. Roots and Tubers 8031 Other (specify): 910 910 FAO not elsewhere specified Roots and Tubers 8032 Other (specify): 910 910 FAO not elsewhere specified Roots and Tubers 8033 Other (specify): 910 910 FAO not elsewhere specified Meat, dairy, eggs Increased value up to reflect Young bird, dressed, international 8034 Chicken 990 1220 FAO refuse, head, feet, values. Updated to viscera and bones remove non-edible components. Increased value up to reflect international 8035 Duck 2240 2910 FAO values. Updated to remove non-edible components. Increased quality of MNPED (2011) - Applied wastage 8036 Beef 1500 2180 meat from very thin wastage factor 15 to medium-fat. Changed to 8037 Pork 3380 3359 FAO medium-fat meat. MNPED (2011) - Changed to 8038 Mutton 1790 2060 Wastage factor: 14 wastage medium fat. 62 TECHNICAL POVERTY ESTIMATION REPORT Original Updated Reason for revising Item code Item Label MPLCS Calories per calories Source Description calorie value kilo per kilo 8039 Dried Meat 2800 2800 MNPED (2011) Updated to remove non-edible 8040 Chicken eggs 1700 1390 FAO components. Corrected to reflect international values Updated to MNPED (2011) - 8041 Duck eggs 1890 1660 Wastage factor: 12 remove non-edible wastage components. Updated to MNPED (2011) - 8042 Quail eggs 1610 1430 Wastage factor: 12 remove non-edible wastage components. Correcting to Cow Milk, Whole, 8043 Fresh milk 750 630 FAO reflect international Fresh values Branded condensed 8044 3370 3370 MNPED (2011) milk Meat not elsewhere 8045 Other (specify): 1260 1260 FAO specified Meat not elsewhere 8046 Other (specify): 1260 1260 FAO specified Meat not elsewhere 8047 Other (specify): 1260 1260 FAO specified Fish and other seafood Updated to Ngamyitchin (river, MNPED (2011), WF Wastage factor: 42, 8048 970 682 remove non-edible medium) from FAO 1972 from item 8058 components. Updated to Ngagyin (river, MNPED (2011) - 8049 900 648 Wastage factor: 40 remove non-edible medium) wastage components. Updated to MNPED (2011) - 8050 Ngayant / Catfish 890 622 Wastage factor: 43 remove non-edible wastage components. Updated to Ngakhu (small MNPED (2011) - 8051 980 699 Wastage factor: 41 remove non-edible catfish) wastage components. Updated to Ngagyee (small MNPED (2011), WF Wastage factor: 29, 8052 1040 828 remove non-edible river fish) from FAO 1972 from item 8057 components. Updated to Ngapyayma MNPED (2011), WF Wastage factor: 42, 8053 1370 963 remove non-edible (medium river fish) from FAO 1972 from item 8058 components. Updated to Ngaton/ Ngamyinn MNPED (2011), WF 8054 1150 691 Wastage factor: 57 remove non-edible (Butter fish) from FAO 1972 components. Updated to Ngathalauk (helsa MNPED (2011), WF 8055 2730 1870 Wastage factor: 45 remove non-edible fish) from FAO 1972 components. 63 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Original Updated Reason for revising Item code Item Label MPLCS Calories per calories Source Description calorie value kilo per kilo 8056 Fish meat 890 890 MNPED (2011) MNPED (2011), WF Applied average Updated to Other small river 8057 930 740 from calculations wastage factor from remove non-edible fishes (<= 4") based on FAO 1972 FAO 1972: 29 components. MNPED (2011), WF Applied average Updated to Other medium river 8058 890 626 from calculations wastage factor from remove non-edible fishes (5" - 10") based on FAO 1972 FAO 1972: 42 components. MNPED (2011), WF Applied average Updated to Other large river 8059 750 512 from calculations wastage factor from remove non-edible fishes (11+") based on FAO 1972 FAO 1972: 45 components. Updated to MNPED (2011) - 8060 Kakatit / Seabasse 810 437 Wastage factor: 46 remove non-edible wastage components. Updated to Ngamoke (small MNPED (2011) - 8061 1190 821 wastage factor: 31 remove non-edible sea fish) wastage components. Updated to Ngapokethin (small MNPED (2011) - 8062 1020 500 Wastage factor: 51 remove non-edible sea fish) wastage components. Updated to MNPED (2011), WF 8063 Sardine (All Kinds) 980 529 Wastage factor: 46 remove non-edible from FAO 1972 components. Updated to Pazun Kyawt / MNPED (2011) - 8064 820 377 Wastage factor: 54 remove non-edible Shrimp wastage components. Updated to Pazun Doke / River MNPED (2011) - 8065 940 432 Wastage factor: 54 remove non-edible prawn wastage components. Updated to MNPED (2011); WF 8066 Squid/ octopus 750 690 Wastage factor (7) remove non-edible from FAO 1972 components. MNPED (2011), WF Applied average Updated to Other small sea 8067 1020 723 from calculations wastage factor from remove non-edible water fishes (<=4") based on FAO 1972 FAO 1972: 29 components. Other medium sea MNPED (2011), WF Applied average Updated to 8068 water fishes (5" - 990 542 from calculations wastage factor from remove non-edible 10") based on FAO 1972 FAO 1972: 45 components. MNPED (2011), WF Applied average Updated to Other large sea 8069 1000 552 from calculations wastage factor from remove non-edible water fishes (11+") based on FAO 1972 FAO 1972: 45 components. Nga Yantchauk / 8070 3280 3280 MNPED (2011) Dry Catfish Other dried small 8071 2800 2800 MNPED (2011) river fish (<=4") Other dried 8072 medium river fish 2980 2980 MNPED (2011) (5"-10") 64 TECHNICAL POVERTY ESTIMATION REPORT Original Updated Reason for revising Item code Item Label MPLCS Calories per calories Source Description calorie value kilo per kilo Other dried large Value from medium 8073 2980 2980 river fishes (11+") sized fish Ngakunshutchauk / 8074 2070 2070 MNPED (2011) Dry seafish Other dried small 8075 sea water fish 3350 3350 MNPED (2011) (<=4") Other dried 8076 medium sea water 3350 3350 MNPED (2011) fish (5"-10") Other dried large Value from medium 8077 sea water fishes 3350 sized fish (11+") 8078 Dried Prawns 3380 3380 MNPED (2011) 8079 Shrimp paste 1650 1650 MNPED (2011) 8080 Fish/ shrimp sauce 400 400 MNPED (2011) 8081 Ngapiyae / fishcake 1520 1520 Nagpikaung/ Salted 8082 1280 1280 MNPED (2011) fish Ar Bye Gyauk / dry Average of small 8083 3113 3075 small fish dried fish Dried Prawn Dried shrimp, edible 8084 0 3620 FAO 1972 powder portion Average from Updated to 8085 Other (specify): 940 527 MNPED (2011) medium sized remove non-edible marine and river fish components. Average from Updated to 8086 Other (specify): 940 527 MNPED (2011) medium sized remove non-edible marine and river fish components. Average from Updated to 8087 Other (specify): 940 527 MNPED (2011) medium sized remove non-edible marine and river fish components. Vegetables Updated to MNPED (2011) + 8088 Pumpkin 270 224 Wastage factor: 17 remove non-edible wastage components. Updated to MNPED (2011) + 8089 Ash pumpkin 120 86 Wastage factor: 28 remove non-edible wastage components. Updated to MNPED (2011) + 8090 Brinjal/ Egg plant 260 250 Wastage factor: 4 remove non-edible wastage components. Updated to MNPED (2011) + 8091 Tomato 200 188 Wastage factor: 6 remove non-edible wastage components. 65 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Original Updated Reason for revising Item code Item Label MPLCS Calories per calories Source Description calorie value kilo per kilo Updated to MNPED (2011) + 8092 Cabbage 220 187 Wastage factor: 15 remove non-edible wastage components. Updated to MNPED (2011) + 8093 Cauliflower 290 174 Wastage factor: 40 remove non-edible wastage components. Updated to MNPED (2011) + 8094 Chayote 190 146 Wastage factor: 23 remove non-edible wastage components. Updated to MNPED (2011) + 8095 Water leaf 420 336 Wastage factor: 20 remove non-edible wastage components. Updated to 8096 Roselle leaf 440 420 FAO 1972 remove non-edible components. Updated to MNPED (2011) - WF 8097 Horseradish leaf 720 440 remove non-edible FAO 1972 components. Updated to MNPED (2011) + 8098 Radish leaf 330 264 Wastage factor: 20 remove non-edible wastage components. Updated to 8099 Pumpkin leaf 210 210 MNPED (2011) remove non-edible components. Updated to MNPED (2011) + 8100 Cucumber 120 96 Wastage factor: 20 remove non-edible wastage components. Updated to 8101 Horseradish 420 510 FAO 1972 remove non-edible components. Updated to GMNPED (2011) + 8102 Bean/ Long bean 280 266 Wastage factor: 5 remove non-edible wastage components. Updated to MNPED (2011) + 8103 Bamboo shoots 280 157 Wastage factor: 44 remove non-edible wastage components. Updated to MNPED (2011) + 8104 Bean sprouts 300 279 Wastage factor: 7 remove non-edible wastage components. Updated to MNPED (2011) + 8105 Carrots 440 310 Wastage factor: 17 remove non-edible wastage components. Updated to MNPED (2011) + 8106 Lettuce 200 150 Wastage factor: 26 remove non-edible wastage components. 8107 Fresh chillie 550 1000 MNPED (2011) 66 TECHNICAL POVERTY ESTIMATION REPORT Original Updated Reason for revising Item code Item Label MPLCS Calories per calories Source Description calorie value kilo per kilo Updated to 8108 Mustard leaf 230 FAO 1972 remove non-edible components. Kinmoon / small MOH Nutrition Unit 8109 330 0 wastage sweet gourd book 8110 Gourd leaf 270 270 FAO 1972 Updated to MNPED (2011) + 8111 Vegetable gourd 160 136 Wastage factor: 15 remove non-edible wastage components. 8112 Other (specify): 220 220 FAO Vegetables fresh nes 8113 Other (specify): 220 220 FAO Vegetables fresh nes 8114 Other (specify): 220 220 FAO Vegetables fresh nes Fruits Updated to remove non-edible components. 8115 Bananas 1080 600 FAO Corrected value to reflect international values Updated to MNPED (2011) + 8116 Papaya 450 324 Wastage factor: 28 remove non-edible wastage components. 8117 Grapefruit 423 160 FAO Updated to remove non-edible components. 8118 Watermelon 320 170 FAO Corrected value to reflect international values Updated to MNPED (2011) + 8119 Apple 510 418 Wastage factor: 18 remove non-edible wastage components. 8120 Pomelo 379 160 FAO 8121 Oranges 629 340 FAO 8122 Plums 455 520 FAO Updated to MNPED (2011) + 8123 Mango 650 468 Wastage factor: 28 remove non-edible wastage components. Updated to MNPED (2011) + 8124 Lime 360 277 Wastage factor: 23 remove non-edible wastage components. 8125 Lemon 150 150 FAO Updated to MNPED (2011) + 8126 Pineapple 470 259 Wastage factor: 45 remove non-edible wastage components. 67 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Original Updated Reason for revising Item code Item Label MPLCS Calories per calories Source Description calorie value kilo per kilo 8127 Grapes 674 530 FAO Updated to MNPED (2011) + 8128 Guava 690 676 Wastage factor: 2 remove non-edible wastage components. Updated to MNPED (2011) + 8129 Mangosteen 570 165 Wastage factor:71 remove non-edible wastage components. Tropical fruit Fresh 8130 Other (specify): 410 410 FAO nes Tropical fruit Fresh 8131 Other (specify): 410 410 FAO nes Tropical fruit Fresh 8132 Other (specify): 410 410 FAO nes Oil and fats 8133 Groundnut oil 8840 8840 MNPED (2011) 8134 Sesamum oil 8810 8810 MNPED (2011) 8135 Palm oil 9000 8840 MNPED (2011) 8136 Mustard oil 8810 8810 MNPED (2011) 8137 Other (specify): 8840 8840 FAO Average of oils 8138 Other (specify): 8840 8840 FAO Average of oils 8139 Other (specify): 8840 8840 FAO Average of oils Spices and condiments 8140 Dried chilies 2460 2460 MNPED (2011) 8141 Chilly powder 2460 2460 MNPED (2011) 8142 Lemon grass 210 210 MNPED (2011) 8143 Onions 380 380 MNPED (2011) Updated to 8144 Garlic 1170 1060 FAO 1972 remove non-edible components. 8145 Tumeric powder 3470 3470 MNPED (2011) Updated to remove non-edible components; 8146 Ginger 930 420 FAO 1972 Corrected value to reflect international values 8147 Salt 0 0 MNPED (2011) 8148 Seasoning powder 4267 0 8149 Black pepper 0 3250 MNPED (2011) 8150 Marsala 0 0 MNPED (2011) 8151 Other (specify): 0 0 MNPED (2011) 8152 Other (specify): 0 0 MNPED (2011) 68 TECHNICAL POVERTY ESTIMATION REPORT Original Updated Reason for revising Item code Item Label MPLCS Calories per calories Source Description calorie value kilo per kilo 8153 Other (specify): 0 0 MNPED (2011) Other food products Corrected to 8154 Dried rice noodle 880 3600 FAO 1972 reflect international values. 8155 Dried wheat noodle 3630 3630 MNPED (2011) 8156 Rice vermicelli 3570 3570 MNPED (2011) 8157 Bread 2820 2820 MNPED (2011) 8158 Cake 3930 3930 MNPED (2011) 8159 Biscuits 4070 4070 MNPED (2011) Value from soy bean 8160 Pone Ye Gyi 2000 1480 paste 8161 Bean curd (white) 1913 330 FAO 1972 8162 Tofu 1452 630 FAO 1972 8163 Soy bean paste 2000 1480 FAO 1972 8164 Vermicilli (bean) 3750 3400 FAO 1972 8165 Green tea leaves 0 300 MNPED (2011) 8166 Sugar 3510 3510 MNPED (2011) 8167 Palm jaggery 3590 3590 MNPED (2011) 8168 Cane jaggery 3830 3830 MNPED (2011) Fermented tea 8169 300 300 MNPED (2011) leaves 8170 Fritters with pea 4740 4740 MNPED (2011) 8171 Kun yar 420 420 MNPED (2011) Coffee mix or tea 8172 2150 2150 MNPED (2011) mix 8173 Cereal mix 2090 2090 MNPED (2011) http://caloriecount. Ovaltine, horlick, about.com/calories- 8174 3756 3756 etc. ovaltine-chocolate- malt-i116817 Prepared food bought to eat 8175 at home - rice or noodle based dishes Prepared food 8176 bought to eat at home - snacks Prepared food 8177 bought to eat at home - other 69 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Original Updated Reason for revising Item code Item Label MPLCS Calories per calories Source Description calorie value kilo per kilo Other food 8178 products (specify): Other food 8179 products (specify): Other food 8180 products (specify): Alcoholic beverages Replaced beer / alcohol with zero calories, as the 8181 Beer 0 0 reference basket should not contain beer. Toddy alcohol / 8182 0 0 palm alcohol Local liquors/ 8183 0 0 alcohol Imported liquor/ 8184 0 0 alcohol Note: WF: wastage factor, nes: not elsewhere specified Source: MNPED (2011), FAO (n.d.), FAO (1972), Ministry of Health (n.d.) Table A2.3 Calories assigned to other categories Number of cases in Food Item Source Food Item in Source Calories per kg MPLCS Other rice and cereals 1 FAO Wheat (as in specification) 3400 Other pulses, nuts, beans 59 FAO Pulses nes 3400 Other peas 502 IHLCA Mean of chick peas and green peas 3075 Other tubers 33 FAO Roots and tubers nes 910 Other meat & diary 85 FAO Meat nes 1260 Different specifications of fish (by Different 2101 IHLCA Follow definition of IHLCA size) categories Mean of medium fresh sea& river Other fish 181 IHLCA fish. Alternative: Marine fish nes 527 640 (FAO) Other vegetables 822 FAO Vegetables fresh nes 220 Other fruits 208 FAO Fruit tropical fresh nes 410 Other Oil and fats 388 FAO Oils 8840 30 obs are “tobacco” with 0 Other Food products 38 / calories; the other 8 items did not 0 readily translate into any category. Sources: MNPED et al. (2011b), FAO (n.d.). NES denotes not elsewhere specified. 70 TECHNICAL POVERTY ESTIMATION REPORT Table A2.4 Nutritional requirements in Myanmar Age (Years) Calories Frequency (weighted) in MPLCS Child 1 850 287 1-3 1260 908 4-6 1670 1042 7-9 1800 1137 Male (Adolescent) 10-12 2400 624 13-15 2600 513 16-19 3000 569 Female (Adolescent) 10-12 2200 565 13-15 2400 538 16-19 2500 706 Adult Male (Moderate activities) 2800 4436 Adult Female (Moderate activities) 2000 5326 Add on values: Pregnant (five months before delivery) 350 120 Lactating Female (post-partum six months) 550 431 Total population (MPLCS) 16651 Calories per capita 2247 Calories per adult equivalent 2238 Calories per adult male equivalent 2800 Source: Ministry of Health (n.d.) 71 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY Annex A3 Poverty Estimates and Sensitivity Analysis Table A3.1 Calories per item in poverty line food basket – detailed 2004/05 Poverty 2015 Poverty Line, 2009/10 Poverty Line, Food Category Food Item Line, MNPED et al Updated Measure World Bank (2014) (2007) Total Calories 2238 2300 2340 Pegyi (lablab beans) 8.17 4.61 8.29 Pegya 7.48 4.59 12.96 Pe poke 2.92 2.79 6.61 Sadawpe (green peas) 3.79 3.55 5.82 Gram (Chick pea) 14.49 17.99 9.38 Green gram (Pedesane) 3.86 1.88 9.60 Penilay (Peyaza) 4.87 5.51 8.29 Butter Bean 8.15 6.26 10.65 Other peas 12.84 0.00 0.00 Pulses, beans, nuts and Sesame 0.66 0.76 5.42 seeds Groundnut without shell 17.49 23.13 13.96 Coconut 6.36 3.36 6.84 Pulses, beans, nuts, and seeds 0.66 0.00 0.00 Other (specify): Pulses, beans, nuts, and seeds 0.25 0.00 0.00 Other (specify) Boiled Pea (any kind of peas) 0.00 9.25 6.50 Cashew nuts 0.00 0.11 18.83 Black gram (Matpe) 0.00 0.43 9.43 Pepyin 0.00 0.79 9.58 Chicken 13.95 7.07 3.61 Duck 2.71 1.21 9.33 Beef 3.86 9.37 6.86 Pork 27.90 35.30 17.50 Mutton 0.04 0.72 5.30 Meat, dairy, eggs Dried Meat 0.27 0.36 4.99 Chicken eggs 12.66 8.49 4.15 Duck eggs 5.64 6.48 4.92 Quail eggs 0.16 0.03 1.50 Fresh milk 0.77 0.77 5.34 Other meats 2.68 0.00 0.00 72 TECHNICAL POVERTY ESTIMATION REPORT 2004/05 Poverty 2015 Poverty Line, 2009/10 Poverty Line, Food Category Food Item Line, MNPED et al Updated Measure World Bank (2014) (2007) Ngamyitchin 1.07 2.22 3.75 Ngagyin 1.42 3.90 3.08 Ngayant 1.12 3.10 2.97 Ngakhu 0.61 1.60 2.93 Ngagyee 0.12 0.38 2.88 Ngapyayma 1.20 1.98 4.69 Ngaton/ Ngamyinn 0.76 1.43 3.69 Ngathalauk 0.59 3.39 9.20 Fish meat 0.08 0.62 1.68 Other small river fishes 1.86 3.73 3.41 Other medium river fishes 0.92 1.11 3.39 Other large river fishes 0.04 0.11 2.59 Kakatit 0.12 0.34 2.82 Ngamoke 0.81 0.51 3.94 Ngapokethin 0.18 1.21 2.91 Sardine (All Kinds) 0.42 0.52 3.84 Pazun Kyawt 0.32 1.48 2.34 Pazun Doke 0.02 0.11 1.98 Fish and other seafood Squid and like sea fishes 0.17 0.24 2.73 Other small sea water fishes 2.51 1.46 5.53 Other medium sea water fishes 0.77 0.71 4.67 Other large sea water fishes 0.03 0.13 3.40 Nga Yantchauk 0.83 1.80 5.84 Other dried small river fish 2.04 2.95 5.98 Other dried medium and above 2.47 1.08 5.96 river fish Other dried medium and above 0.18 1.08 5.96 river fish Ngakunshutchauk 0.05 0.38 3.57 Other dried small sea water fish 5.07 2.55 8.26 Other dried medium and above 1.40 1.36 7.62 sea water fish Other dried medium and above 1.00 1.36 7.62 sea water fish Dried Prawns 1.29 1.60 2.78 Shrimp paste 5.71 4.19 2.94 Fish/ shrimp sauce 0.84 0.82 0.84 Ngapiyae 6.88 5.19 4.58 Nagpikaung/ Salted fish 0.84 0.78 2.91 73 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY 2004/05 Poverty 2015 Poverty Line, 2009/10 Poverty Line, Food Category Food Item Line, MNPED et al Updated Measure World Bank (2014) (2007) Ar Bye Gyauk 1.74 4.44 0.00 Dried prawn powder 0.24 0.33 0.00 Other fish products 0.47 0.00 0.00 Fish and other seafood Fish and other seafood Other 0.01 0.00 0.00 (specify) Fish and other seafood Other 0.01 0.00 0.00 (specify) Ngashwe 0.00 0.28 3.76 Sweet potatoes 2.81 3.24 8.23 Potatoes 11.00 8.35 3.75 Radish 0.81 0.81 1.04 Taro 1.52 1.44 5.61 Pemyit 3.68 1.37 0.00 Roots and tubers Palm shoot 0.67 2.16 0.00 Roots and tubers Other (specify) 0.16 0.00 0.00 Palawpenan 0.00 0.44 0.00 Yams 0.00 0.14 2.81 No Ko 0.00 0.03 0.00 Arrow shoot 0.00 0.09 0.00 Pumpkin 0.91 0.74 1.26 Ash pumpkin 0.14 0.20 0.59 Brinjal/ Egg plant 4.41 2.58 1.21 Tomato 5.50 2.80 1.33 Cabbage 3.27 1.38 1.01 Cauliflower 1.34 1.32 0.96 Chayote 0.05 0.47 1.48 Water leaf 3.26 10.93 4.35 Roselle leaf 3.39 6.53 2.78 Vegetables Horseradish leaf 0.25 0.37 1.10 Radish leaf 0.11 0.30 0.61 Pumpkin leaf 0.16 0.16 0.56 Cucumber 0.29 0.49 0.40 Horseradish 0.87 0.43 0.97 Bean/ Long bean 0.94 1.39 0.92 Bamboo shoots 0.30 0.43 0.63 Bean sprouts 0.41 0.73 0.69 Carots 0.07 0.10 0.66 Lettuce 0.08 0.08 0.24 74 TECHNICAL POVERTY ESTIMATION REPORT 2004/05 Poverty 2015 Poverty Line, 2009/10 Poverty Line, Food Category Food Item Line, MNPED et al Updated Measure World Bank (2014) (2007) Chillie/ pepper/ sweet pepper 4.30 1.29 0.72 Mustard leaf 1.13 0.00 0.00 Kinmoon 0.01 0.00 0.00 Gourd leaf 1.40 0.00 0.00 Gourd 2.78 1.45 0.80 Other vegetables 0.36 0.00 0.00 Vegetables Vegetables Other (specify) 0.07 0.00 0.00 Vegetables Other (specify) 0.03 0.00 0.00 Lime 0.19 0.22 0.24 Fresh pepper/Sweet pepper 0.00 0.02 0.00 Kha We 0.00 0.19 0.00 Citrics 0.00 0.04 0.00 Banana 10.26 20.10 8.79 Papaya 1.43 4.33 4.09 Grapefruit 0.00 0.16 0.00 Watermelon 5.80 0.65 2.04 Apple 0.61 3.53 2.71 Pomelo 0.01 0.48 0.00 Oranges 0.69 6.15 0.00 Plums 1.34 0.35 0.00 Mangoes 1.00 18.13 6.66 Lemon 0.07 0.00 0.00 Pineapples 0.00 0.58 5.45 Fruits Grapes 0.17 0.11 0.00 Guava 0.50 0.49 3.42 Mangosteens 0.00 0.19 1.56 Other fruits 0.25 0.00 0.00 Rambutan (Kyetmouk) 0.00 0.07 0.53 Pear 0.00 0.02 2.87 Durian 0.00 0.42 5.67 Strawberry 0.00 0.00 0.00 Custard Apple 0.00 0.01 1.67 Sunkist 0.00 0.06 0.00 Jackfruit 0.00 0.04 0.00 75 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY 2004/05 Poverty 2015 Poverty Line, 2009/10 Poverty Line, Food Category Food Item Line, MNPED et al Updated Measure World Bank (2014) (2007) Dried chilies 9.65 6.56 3.30 Chilly powder 3.68 2.58 1.89 Lemon grass 0.01 0.04 0.14 Onions 8.08 3.92 1.71 Garlic 5.91 3.10 1.35 Spices and condiments Tumeric powder 2.82 2.28 1.14 Ginger 1.14 0.78 0.56 Seasoning powder 0.00 6.07 0.00 Black pepper 0.21 0.00 0.00 Mustard seeds 0.00 0.02 0.00 Dried rice noodle 2.94 0.74 3.45 Dried wheat noodle 5.58 5.89 10.54 Rice vermicelli 3.10 4.65 8.80 Bread 2.96 17.25 17.85 Cake 3.58 6.73 9.58 Biscuits 3.16 9.99 14.50 Pone Ye Gyi 0.54 0.29 0.00 Bean curd (white) 0.42 0.41 0.00 Tofu 1.99 0.30 0.00 Soy bean paste 0.69 15.71 0.00 Other food products Vermicilli (bean) 0.10 18.24 0.00 Green tea leaves 1.01 0.00 0.00 Sugar 7.91 4.75 2.98 Palm jaggery 7.67 5.81 4.23 Cane jaggery 6.96 1.67 4.09 Green tea leaves 0.45 0.00 0.00 Fritters with pea 13.14 4.84 3.51 Areaca nuts/ Betel nuts 1.70 1.08 0.00 Coffee mix or tea mix 4.15 1.41 1.12 Cereal mix 0.51 0.21 0.92 Ovaltine, horlick, etc. 0.68 0.31 0.00 Bean curd (brown) 0.00 0.06 0.00 Beer 0.00 0.31 0.00 Toddy alcohol 0.00 4.37 0.00 Alcoholic beverages Local liquors/alcohol 0.00 1.32 0.00 Imported liquors/alcohol 0.00 0.04 0.00 Rice wine (Khaung Ye) 0.00 0.92 0.00 76 TECHNICAL POVERTY ESTIMATION REPORT 2004/05 Poverty 2015 Poverty Line, 2009/10 Poverty Line, Food Category Food Item Line, MNPED et al Updated Measure World Bank (2014) (2007) Prepared food bought to eat 4.20 0.00 0.00 at home - rice or noodle based dishes Prepared food bought to eat at 12.22 0.00 0.00 home - snacks Prepared food bought to eat at 0.29 0.00 0.00 home - other Breakfast away from home 33.09 0.00 0.00 Lunch away from home 34.30 0.00 0.00 Dinner away from home 29.40 0.00 0.00 Snacks away from home 15.73 0.00 0.00 Butter spread nan/bread 0.00 0.24 8.38 Rice based traditional snacks 0.00 8.49 21.77 Palata (parala - an Indian pancake) 0.00 2.14 6.33 Hot tofu 0.00 1.53 15.16 Rice noodle/vermicelli salad/soup 0.00 0.20 3.91 Shan noodle/ Mie Shay 0.00 3.08 19.28 Food and beverages Alcoholic beverages taken outside 0.00 0.08 0.00 taken outside home home Fried rice 0.00 4.75 19.40 Pauksi (Chinese steamed bun 0.00 0.59 6.77 with stuffing) Myanmar vermicelli (for 0.00 3.85 5.88 breakfast) Ekyarkway 0.00 4.30 6.55 Wheat noodle salad/soup 0.00 0.15 3.53 Brewed tea/coffee taken outside 0.00 2.82 3.02 home Fried noodle/ Cutkyikite 0.00 0.69 10.23 Softdrinks and/or juices taken 0.00 0.34 1.54 outside home Mohingar/ Nhyat noodle 0.00 9.84 7.41 Fried bean curd (tohu) 0.00 0.63 3.96 Samosa (fried stuffed pastry) 0.00 1.89 4.32 Fried snack (gourd/ pulses ... etc.) 0.00 9.49 6.20 Nan with boiled bean 0.00 0.37 3.24 77 MYANMAR POVERTY AND LIVING CONDITIONS SURVEY 2004/05 Poverty 2015 Poverty Line, 2009/10 Poverty Line, Food Category Food Item Line, MNPED et al Updated Measure World Bank (2014) (2007) Rice (Ngasein) 297.76 490.07 374.67 Rice (Emata) 289.05 501.24 331.60 Rice (Medone) 39.96 53.79 295.22 Rice (Nga kywe) 40.84 93.38 232.39 Kaukhnyin (Sticky Rice) 3.19 7.62 41.90 Other rice (local variety) 790.43 143.99 0.00 Rice and cereals Maize seeds (dry) 3.60 0.60 66.80 Millet 0.00 0.37 0.00 Sorghum 0.00 0.04 0.00 Wheat 0.00 0.23 0.00 Flour (Rice) 0.00 0.14 5.31 Flour (Wheat) 0.00 0.07 7.40 Groundnut oil 76.54 55.27 41.41 Sesamum oil 33.20 41.81 56.48 Palm oil 115.77 40.25 40.19 Mustard oil 4.66 1.87 22.69 Oil and fats Other cooking oil and fat 22.70 0.00 0.00 Pork fat 0.00 0.47 0.00 Ghee 0.00 0.00 1.40 Sunflower oil 0.00 2.74 46.44 Branded condensed milk 2.70 2.29 4.25 Domestic condensed milk 0.00 0.02 3.23 Milk products Formula milk for infants 0.00 0.30 7.64 Milk powder 0.00 0.21 4.05 Butter 0.00 0.03 2.40 Potato chips 0.00 2.37 0.00 Pickled tea 0.00 0.16 0.12 Betel leafs 0.00 0.69 0.62 Other food items Coffee (grinded or beans) 0.00 0.00 0.03 Horlick 0.00 0.02 0.00 Non-dairy creamer 0.00 2.07 56.65 78 The World Bank Myanmar No 57, Pyay Road, (Corner of Shwe Hinthar Road) www.worldbank.org/myanmar 61/2 Mile, Hlaing Township, www.facebook.com/ Yangon, Republic of the Union of Myanmar myanmar@worldbank.org