PROGRESS AND CHALLENGES PROGRESS AND CHALLENGES © 2019 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. Contents 7 Abbreviations 7 Acknowledgment 8 Overview 9 I. Recent progress in poverty reduction and welfare 9 Poverty was reduced, but progress in shared prosperity slowed in recent years 12 While access to services improved significantly, concerns regarding quality and equity are growing, as dispersed population settlement patterns call into question spatial trade-offs in development 17 II. Drivers of poverty reduction 17 Economic growth was strong, driven by hydropower development, but it was not accompanied by job creation 19 Poverty is highly correlated with being engaged in agriculture 20 Poverty reduction was driven by rural, agricultural households 21 Aggregate data suggest that agricultural earnings gradually increased 24 However, volatile prices of major fruits and crops raise concern over sustainability of these improve- ments 25 Vulnerability is high, partly because farmers are exposed to various uninsured risks, but also because the social protection system is weak overall and nonfarm diversification is low 29 III. Priorities for sustainable poverty reduction 32   Annex Official and comparable poverty measurement methodologies 38 References Boxes 10 BOX 1  Poverty trend estimated using a 28 BOX 2  Recommendation of the 19th ILO consistent methodology International Conference of Labour Statisticians concerning statistics of work, employment, and labor underutilization Figures 9 FIGURE 1  Comparable poverty rates in Bhutan 19 FIGURE 16  Working in agriculture is highly suggest significant progress at the national level correlated with being poor between 2007 and 2017, which was driven by 19 FIGURE 17  Urban households rely mainly improvements in rural areas on wage income, whereas rural households draw 11 FIGURE 2  Bhutan performs relatively well in income from various agricultural products terms of progress in shared prosperity 20 FIGURE 18  Growth incidence curves show that 11 FIGURE 3  Shared prosperity slowed markedly consumption growth was slower among poorer between 2012 and 2017 and was lower than households the consumption growth of the total population 20 FIGURE 19  Datt-Ravallion decomposition 12 FIGURE 4  There were steady improvements confirms that consumption growth was very in asset ownership for poor and nonpoor, unequal 2007-2017 21 FIGURE 20  Huppi-Ravallion decomposition 12 FIGURE 5  Despite significant progress, as shows that poverty reduction was driven by shown by the large decrease in the share of improvements among agricultural households population without education across cohorts, 22 FIGURE 21  Agriculture’s share in the economy education levels remain overall low continued to decline (Value added in Bhutan, by 12 FIGURE 6  Overall educational attainment is sector, 2000–17) lower among the poor, the rural and women 22 FIGURE 23  Output in the agriculture, forestry, 13 FIGURE 7  A significant share of the population and fishing sector increased has accessed nonformal education programs 22 FIGURE 22  However, the output of the in Bhutan agriculture, forestry, and fishing sector grew 14 FIGURE 8  Bhutan has an acute shortage of notably in recent years health care professionals 22 FIGURE 24  Value added per worker in 15 FIGURE 9  Quality of services remains highly agriculture, forestry, and fishing also improved heterogeneous across the country 23 FIGURE 25  Earnings from main fruits and crops 17 FIGURE 10  The shares of industry and services in Bhutan improved in GDP (value added) have increased while the 24 FIGURE 26  The increase in average yields of key share of agriculture has decreased fruits and crops helped improve earnings 17 FIGURE 11  GDP per capita in Bhutan increased 24 FIGURE 27  Prices of key fruits and crops rose by more than ten-fold between 1980 and 2016, but some have been volatile a much faster pace than in South Asia 27 FIGURE 28  80 percent of poor households on average produce mainly or solely for family consumption 18 FIGURE 12  Labor force participation rate in Bhutan, 2017 in Bhutan declined in recent years, though 27 FIGURE 29  There are few opportunities for estimates before and after 2013 are not strictly nonfarm diversification for the rural poor in comparable Bhutan, 2017 18 FIGURE 13  Net job creation has been slow 30 FIGURE 30  Basic services and roads top the 18 FIGURE 14  And has not kept up with the priority areas for government action, selected increase in the labor force by survey respondents 19 FIGURE 15  The share of employment in 28 FIGURE B2.1  Revised statistical standards for agriculture remains high measuring work and employment Map 10 MAP 1  District-level poverty rates in Bhutan vary widely in 2017 Tables 10 TABLE 1  Poverty rates declined significantly 25 TABLE 10  Small changes in the poverty line lead in most districts in Bhutan between 2007 to large changes in the poverty headcount rate and 2017 in Bhutan, 2017 13 TABLE 2  Time taken to reach the nearest 33 TABLE A.1  Comparison of official and hospital or outreach clinic in Bhutan has comparable consumption aggregates declined significantly in the last ten years 34 TABLE A.2  Per capita consumption in Bhutan, 15 TABLE 3  The monetary poor are also more likely official and comparable, 2007–17 to “feel” poor 35 TABLE A.3  Comparison of CPI and survey-based 15 TABLE 4  However, the poor are significantly food price index in Bhutan, 2007–17 less likely to be happy 35 TABLE A.4  Official and comparable poverty lines 16 TABLE 5  Budget share of food groups indicate in Bhutan, 2007–17 small improvements in the dietary composition 36 TABLE A.6  District-level poverty rates in 22 TABLE 6  The crop subsector has been growing Bhutan, 2007–17 steadily 36 TABLE A.5  Official and comparable poverty 23 TABLE 7  Value and amount of livestock trends in Bhutan, 2007–17 products have increased in recent years 10 TABLE B1.1  Official and comparable poverty 24 TABLE 8  A handful of fruits and crops account rates in Bhutan, 2007–17 for most earnings 25 TABLE 9  A few fruits and crops dominate the agricultural export market Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 7 Abbreviations BLSS Bhutan Living Standards Survey CPI Consumer Price Index GDP Gross Domestic Product ICLS International Conference of Labor Statisticians ILO International Labour Organization NCD Noncommunicable Disease PPP Purchasing Power Parity SBPI Survey-Based Food Price Index Acknowledgment This report was prepared by Yeon Soo Kim (economist). Pierre Marion (consultant) helped to collect agricul- tural statistics. The team received overall guidance from Yoichiro Ishihara (resident representative), Christian Eigen-Zucchi (program leader), and Benu Bidani (practice manager). The team is thankful to peer reviewers Trang Van Nguyen (senior economist) and Gabriel Lara Ibarra (senior economist) and to Dean Jolliffe (lead economist) for his very helpful comments and suggestions on the poverty measurement methodology. Maria Eugenia Genoni (senior economist), Nandini Krishnan (senior economist), Hiroki Uematsu (senior economist), and Aziz Atamanov (senior economist) also provided useful comments. The team thanks the National Statistics Bureau for their support in preparing this report. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 8 OVERVIEW This note uses data from the Bhutan Living Standards Survey (BLSS) for 2007, 2012, and 2017 to examine trends in poverty reduction and shared prosperity and to assess the drivers of poverty reduction in the last decade. The note documents the remarkable progress that Bhutan has made in reducing poverty, sharing prosperity, and improving other measures of well-being. To this end, it first establishes a poverty trend that is measured in a consistent manner over time. The resulting poverty trend deviates from official poverty rates for mainly two reasons: (a) the consumption aggregate is being measured in a consistent way over time, resulting in a dif- ferent distribution of consumption in each survey year; (b) the $3.20 World Bank poverty line for lower mid- dle-income countries is used as the cut-off to define poverty. While the resulting measures of poverty deviate somewhat from official poverty rates, the difference is relatively small and the impressive reduction in pov- erty remains unchanged. The analysis in the rest of the note relies on the revised $3.20 poverty rates between 2007 and 2017 for Bhutan. There has been broad-based progress on measures of monetary and nonmonetary welfare. However, spa- tial patterns of poverty reveal that poverty is almost exclusively a rural phenomenon, and some of the poor- est, remote districts made little progress. Following large investments by the government, access to services improved significantly. However, concerns regarding quality and equity are growing, as dispersed popula- tion settlement calls into question spatial trade-offs in development. Economic growth was strong but did not deliver many jobs because it was driven by capital-intensive hydropower development. Poverty reduction was helped by improved earnings in the commercial agriculture sector. Moreover, vulnerability has remained high, partly because farmers are exposed to various uninsured risks, including price shocks, but also because the social protection system is weak overall and nonfarm diversification is low. Going forward, it will be impor- tant to continue increasing agricultural productivity and creating productive jobs outside of the agriculture sector. Efforts to improve service delivery need to continue, but important policy trade-offs may have to be made in deciding where the services are delivered and to whom. Proper prioritization and sequencing of pol- icies may also help in this regard. 9 I. RECENT PROGRESS IN POVERTY REDUCTION AND WELFARE Poverty was reduced, but progress in shared prosperity slowed in recent years Bhutan has made significant progress in poverty reduction in the last decade. Using the World Bank’s international poverty rate of US$3.20 per day (in 2011 purchasing power parity [PPP] terms) for lower-mid- dle-income countries such as Bhutan, the poverty headcount decreased from 36.4 percent in 2007 to 17.8 per- cent in 2012 and then further to 12.1 percent in 2017. Urban poverty did not change substantially between 2007 and 2012, but it subsequently decreased to 1.6 percent FIGURE 1  Comparable poverty rates in Bhutan in 2017. Poverty in rural areas declined dramatically suggest significant progress at the national level between throughout the decade, from 48.1 percent in 2007 to 2007 and 2017, which was driven by improvements 17.4 percent in 2017 (figure 1). As almost all of the poor in rural areas reside in rural areas, the large improvements in rural areas drove progress at the national level. The poverty 50 48.1 estimates referred to throughout this note are based 40 on a consistent methodology, which is summarized in 36.4 box 1 and described in detail in the annex. Poverty rate (%) 30 23.9 Progress was widely achieved at the district lev- 20 el, although welfare in the poorest districts has 17.4 17.8 12.1 deteriorated in recent years. Most regions expe- 10 4.1 rienced steady progress in poverty reduction be- 3.8 0 1.6 tween 2007 and 2017: for example, the poverty head- 2007 2012 2017 count in Chukha, the second-most populous district, Urban Rural Total decreased from 32.2 percent in 2007 to 16.1 percent Source: Calculations using Bhutan Living Standards Survey (BLSS) in 2012 and to 7.4 percent in 2017. The poverty head- for 2007, 2012, and 2017. count in Samtse, which accounts for the largest num- ber of poor people, declined from 53.2 percent in 2007 to 32.3 percent in 2012 and to 17.5 percent in 2017. There was a notable exception to the overall improvements: Dagana, the poorest region in 2017, saw its poverty rate that year revert back almost to 2007 levels. The pover- ty headcount in Zhemgang was cut by more than half between 2007 and 2012 but rose slightly in 2017. These two districts also recorded the highest incidence of poverty in 2017: 42.8 percent in Dagana and 36.5 percent in Zhemgang. Both are remote districts located in the central part of Bhutan. In 2017, just three districts (Dagana, Monggar, and Samtse) accounted for more than a third of the total number of poor (36.6 percent) due in part to their large populations (table 1). As of 2017, poverty varies widely across districts map 1) and remains al- most exclusively a rural phenomenon, with 96 percent of the poor residing in rural areas. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 10 TABLE 1  Poverty rates declined significantly in most BOX 1  Poverty trend estimated using a consistent districts in Bhutan between 2007 and 2017 methodology % of population who are poor The poverty estimates used in this note deviate from the official District 2007 2012 2017 poverty numbers for two reasons: first, the figures in this note are estimated using a consumption aggregate that is constructed Bumthang 29.3 4.8 4.9 based on a consistent methodology across years and, second, the Chhukha 32.2 16.1 7.4 cut-off to define poverty status is based on the US$3.20 poverty line, which measures the minimum living standards for lower-mid- Dagana 43.5 30.6 42.8 dle-income countries. The revisions were undertaken because the Gasa 21.9 — 20.4 methodological choices made for the official estimation of poverty Haa 34.9 14.6 1.8 rates were not always entirely consistent, which can lead to esti- mates that are not strictly comparable across years. The revised Lhuentse 55.2 31.0 8.9 trend presented in this note restores comparability with regard to Monggar 58.2 24.8 22.8 both the consumption aggregate and the intertemporal deflator used to update the poverty line between survey years. As seen Paro 15.3 2.1 0.9 in table B.1.1, the US$3.20 poverty rates using the official and Pemagatshel 50.6 21.9 19.6 revised consumption aggregates are very similar. Full technical details on the methodological differences can be found in the Punakha 31.5 17.4 3.2 annex. Despite differences in the official and comparable trends, Samdrupjongkhar 52.3 26.7 10.9 the overall trend of poverty reduction is the same. Samtse 53.2 32.3 17.5 TABLE B1.1  Official and comparable poverty rates Sarpang 36.5 7.9 18.2 in Bhutan, 2007–17 Thimphu 4.8 2.7 1.1 Indicator 2007 2012 2017 Trashigang 48.1 20.7 16.9 Trongsa 25.5 25.4 15.7 Official poverty rate 23.2 12.0 8.2 Tsirang 47.7 22.2 20.3 US$3.20 poverty rate, 36.4 17.8 12.1 Wangduephodrang 29.0 29.1 8.4 with comparable consumption Yangtse 34.0 14.6 8.1 US$3.20 poverty rate, 30.5 14.5 12.0 with official consumption Zhemgang 72.3 30.3 36.5 Source: Calculations using Bhutan Living Standards Survey (BLSS) for Urban 3.8 4.1 1.6 2007, 2012, and 2017. Rural 48.1 23.9 17.4 Note: The second row shows poverty rates calculated using the World Bank’s international poverty line of US$3.20 per day, which is expressed Total 36.4 17.8 12.1 in 2011 purchasing power parity (PPP) terms and deflated across years using the food and nonfood consumer price index (CPI) series. Full Source: Calculations using Bhutan Living Standards Survey (BLSS) methodological details on the construction of comparable poverty rates are for 2007, 2012, and 2017. described in the annex. Note: Color scale represents districts with high (red) and low (green) poverty rates. — = due to the small sample size, the estimate for Gasa in 2012 is too imprecise to present. MAP 1  District-level poverty rates in Bhutan vary widely in 2017 Gasa (40,45] Lhuentse Trashi (35,40] Yangtse Punakha Bumthang (30,35] (25,30] Paro Thimphu Wangdue (20,25] Phodrang Trongsa (15,20] Haa Trashigang (10,15] Monggar (5,10] Tsirang [0,5] Dagana Sarpang Zhemgang Samtse Chukha Pema Samdrup Gatshel Jongkhar Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 11 Shared prosperity, measured as the consumption growth of the bottom 40 percent, slowed markedly between 2012 and 2017 and was lower than the consumption growth of the total population. Annualized average per capita consumption growth of the bottom 40 percent was 5.2 percent between 2007 and 2012, which was higher than the per capita consumption growth of the total population at 4.2 percent (figure 2). However, the pace of consumption growth of the bottom 40 slowed to 2.6 percent between 2012 and 2017, which was lower than the average consumption growth of the population: 4.8 percent (figure 3). This shows that, in recent years, consumption growth has accelerated in the top of the distribution, but slowed in the bottom of the distribution. The Gini coefficient over this period changed little over the last decade, recording 0.381 in 2007, 0.387 in 2012, and 0.378 in 2017. FIGURE 2  Bhutan performs relatively well in terms of progress in shared prosperity 10 8 6 Consumption growth per capita (%) 4 2 0 -2 -4 -6 -8 -10 China Georgia Burkina Faso Nicaragua Paraguay Vietnam Thailand Sri Lanka The Gambia Maltag Jordan Kazakhstan Uganda Belarus Moldova Macedonia, FYR Dominican Republic El Salvador Indonesia Panama Brazil Kosovo Colombia Uruguay Mauritania Peru Ecuador Pakistan Togo Bhutan Egypt, Arab Rep. Turkey Philippines Tajikistan Armenia Costa Rica Ethiopia Bolivia Russian Federation Mozambique Bangladesh Iran, Islamic Rep. Switzerland Hungary Fiji Côte d'Ivoire France Mongolia Germany Rwanda Honduras Netherlands Sweden Kyrgyz Republic Belgium Denmark Finland Mexico Bulgaria Ireland West Bank and Gaza Argentina Croatia Luxembourg Austria Slovenia Ukraine Albania Iceland United Kingdom Romania Bosnia and Herzegovina Zambia Slovak Republic Portugal South Africa Serbia Spain Italy Madagascar Montenegro Cyprus Benin Greece Bottom 40% Total Population Source: Global Database of Shared Prosperity, circa 2010–15. Calculations for Bhutan using Bhutan Living Standards Survey (BLSS) for 2012 and 2017. FIGURE 3  Shared prosperity slowed markedly between In line with the reduction in monetary poverty, 2012 and 2017 and was lower than the consumption significant improvements were observed in broad- growth of the total population er measures of welfare, such as the ownership of 6 durable assets. Among poor households, the share of Consumption growth per capita (%) those that own a refrigerator rose from 3 percent in 5 2007 to 17 percent in 2017. Over the same period, the 4 ownership of mobile phones increased from 10 percent to 79 percent, while that of television increased from 8 3 percent to 43 percent. Nevertheless, the gap between 2 poor and nonpoor remains wide as the latter experi- 1 enced significant improvements in their welfare as well. Moreover, the ownership of large assets such as 0 cars and computers remains particularly low among B40 T60 Total B40 T60 Total 2007–2012 20012–2017 the poor. Improvements happened mainly among non- poor households, who increased their ownership from Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2007, 2012, and 2017. 14 percent to 25 percent for cars, and from 6 percent Note: B40 = bottom 40% of the population. T60 = top 60% of the population. to 19 percent for computers. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 12 FIGURE 4  There were steady improvements in asset ownership for poor and nonpoor, 2007-2017 a. Poor b. Nonpoor 100 100 % of household % of household 80 80 60 60 40 40 20 20 0 0 Family Compu- Refrige- Mobile Television Rice Stove Family Compu- Refrige- Mobile Television Rice Stove car ter rator phone cooker car ter rator phone cooker 2007 2012 2017 Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2007, 2012, and 2017. Note: No data on mobile phone ownership in 2012. While access to services improved significantly, concerns regarding quality and equity are growing, as dispersed population settlement patterns call into question spatial trade-offs in development Access to services continued to improve, but large disparities persist across districts. Investment in human capital, financed through hydropower revenues, led to significant improvements in education and health outcomes. Infant mortality fell steeply and is now below the South Asia regional average, while, in a single decade, the net primary education enrollment rate increased more than 30 percentage points to 89 per- cent in 2014. Net secondary enrollment also expanded rapidly, surpassing 70 percent in 2017. However, large disparities exist across regions: for example, only 20 percent of individuals ages 15–29 in Zhemgang have obtained upper-secondary education, which is one-third the rate for Thimphu (59 percent). While tertiary enrollment rates are still low, increasingly more emphasis is being placed on higher education. Enrollment rates are higher among women at all levels, which is significant progress considering that literacy rates and educational achievements are lower among older cohorts. Despite tremendous progress in educational attainment, overall education levels are low in the popu- lation, particularly among the poor, the rural, women, and older cohorts. The share of individuals who never attended school declined dramatically in recent decades: 57 percent of adults ages 30–49 never attended school, compared with 84 percent of adults ages 50–64. Yet, among individuals ages 15–29, only 17 percent never attended school as of 2017 (figure 5). Overall educational achievement is lower among the poor, the rural, and women (figure 6). Literacy, defined as the share of the population who can read and write a short FIGURE 5  Despite significant progress, FIGURE 6  Overall educational attainment is lower among as shown by the large decrease in the share the poor, the rural and women of population without education across cohorts, education levels remain overall low 80 70 100 60 % of population % of population 80 60 50 40 40 20 30 0 20 10 No education Primary incomplete Primary complete Secondary complete Some tertiary 0 In boarding Can read/ No Primary Primary Secondary Some school write short education incomplete complete complete tertiary text 15–29 30–49 50–64 Source: Calculations using Bhutan Living Standards Survey Urban Rural Male Female Nonpoor Poor Total (BLSS) for 2017. Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 13 text in any language, is 60.4 percent overall, but is lower in rural areas (53.5 percent), among women (54.4 percent), and among the poor (51.6 percent). Moreover, while primary enrollment is nearly universal and sec- ondary enrollment has expanded rapidly, tertiary enrollment is low. Access to services is challenging given Bhutan’s geography, and equity issues are of increasing concern, as is the quality of education. Access issues are highlighted by the share of children who are attending boarding school, which is 6.7 percent of all chil- dren, 13 percent of poor children, and 8.7 percent of rural children (compared with only 2.5 percent of urban children). The higher share of poor children in boarding schools is likely because poverty is concentrated in remote areas where access is a greater concern. Nonformal education programs have helped to improve literacy and numeracy levels. Nonformal edu- cation programs aim to improve the basic literacy and numeracy skills of individuals who missed the oppor- tunity to participate in formal schooling or training. From 1992 to 2017, 203,471 learners participated in non- formal education programs, of which 71 percent were FIGURE 7  A significant share of the population has female. Survey data suggest that younger age groups accessed nonformal education programs in Bhutan have benefited disproportionately from the program. Low-educated people in rural and urban areas have 25 benefited equally, while the nonpoor are more like- 20 ly to have taken advantage of the program (figure 7). % of population 15 The government has recently adopted a Nonformal Education Equivalency Framework, which provides a 10 formal basis for evaluating alternative educational cre- 5 dentials, while at the same time offering flexible path- 0 ways to achieve education and obtain relevant skills. 15–29 30–49 50–64 Poor Non- Rural Urban Women Men While this framework could facilitate access to formal poor Age Poverty Area Gender jobs for some population groups, the labor market rel- status evance of the program needs to be strengthened as the Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. main focus has been to improve basic competencies Large investments in health care facilities helped to improve access, especially at lower levels of care, however, quality of care and equity of access to services require urgent attention. Bhutan’s health care system does well in terms of coverage, affordability, and availability of care. Coverage is almost universal because public health services are free. As of 2017, 99 percent of the urban and 86 percent of the rural popula- tion live within an hour of a health care facility (outreach clinic, basic health unit, or district hospital). Almost 99 percent of the population live within three hours of a health care facility (table 2). While access to health care improved between 2007 and 2017 in rural areas, it still takes significantly longer to reach a health care facility in rural areas than in urban areas. Out-of-pocket expenditure remains low, at 12 percent in 2014, as access to government health facilities is high, providing financial protection to households. However, a large share of this expenditure is related to transport costs, which may affect access to services. TABLE 2  Time taken to reach the nearest hospital or outreach clinic in Bhutan has declined significantly in the last ten years   Time to a hospital, basic health unit, % of population less than one hour from % of population less than three hours from or satellite clinic (minutes) a hospital or outreach clinic a hospital or outreach clinic  Indicator 2007 2017 2007 2017 2007 2017 Urban 19 18 98.4 98.7 99.2 99.5 Rural 84 62 61.3 86.1 91.2 97.8 Nonpoor 51 44 80.4 91.5 95.7 98.4 Poor 102 70 51.5 81.0 88.2 98.4 Total 65 46 72.5 90.6 93.6 98.4 Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 14 With an ongoing epidemiological transition, communicable as well as noncommunicable diseases (NCDs) are placing an increasing burden on the health system. Health outcomes in Bhutan are among the best in South Asia (after Sri Lanka), but socioeconomic and spatial disparities are significant: one in five children is stunted, with poor outcomes being correlated with poor access to safe drinking water and poor infant feed- ing and caring practices, characteristics more likely to be found among poor households. Infant and under- five mortality is higher in the eastern region than in the western and central regions, and under-five mor- tality in rural areas is twice that in urban areas. Significant changes in lifestyle have led to the emergence of new health challenges. The incidence of NCDs is increasing rapidly, while the burden of communicable diseases remains significant. NCDs, including ischemic heart disease, cerebrovascular disease, hypertension, diabetes, and cancer, are increasing and account for more than 70 percent of the reported disease burden. Such diseases pose a significant risk to people’s health in their productive years. Mental health problems, including alcohol- ism and suicide, are on the rise, owing to sociocultural changes, urbanization, migration, and unemployment. The existing health system is not well equipped to meet the emerging challenges. While the government has continued to invest in health infrastructure, with just 3.7 doctors (300 total) and 15.1 nurses per 10,000 population in 2017, there is an acute shortage of health professionals, especially among specialists (figure 8). The lack of targeted nutrition interventions for low-income and marginalized subgroups of the population and limited geographic access to secondary care services are critical barriers holding back further improve- ments in health outcomes. FIGURE 8  Bhutan has an acute shortage of health care professionals 45 39.5 Number per 10,000 population 40 35 30 26.9 25 21.1 21.2 20 15.1 15 9.8 9.6 10.4 10 6.5 7.8 6.8 3.7 5.3 5.0 5 2.8 3.2 2.1 0.9 3.1 1.9 0.8 0.5 0 0.5 0.5 1.6 0.9 0.9 0.8 0.2 0 Bhutan Afghanistan Nepal Bangladesh India Pakistan Sri Lanka Maldives Doctors Nurses and midwives Dentists Pharmacists Source: World Health Organization Global Health Workforce Statistics database. Similar to education and health services, the quality of basic services is also highly variable across the country. The rapid expansion of electricity access was achieved by implementing off-grid renewable energy projects (solar systems) and donor-assisted grant projects. Almost every Bhutanese now has access to electric- ity, but the quality of access varies substantially: the share of households that experienced a power outage in the last seven days ranges from 11 percent in Tsirang to 95 percent in Dagana, with the overall average at just below 60 percent. It also remains an open question how the renewable energy projects will be sustained going forward. Maintaining these systems requires significant resources and manpower, but the mountainous terrain makes it difficult for maintenance workers to access remote areas for routine maintenance (World Economic Forum 2018). Access to water has improved, but almost 90 percent of the population treats (for example, boils) water before drinking it. Moreover, the share of households with 24-hour water supply varies from 51 percent to 83 percent across regions. Finally, the overall share of households with a flush toilet at home is 81 percent, but in three regions this figure is less than 60 percent (figure 9).1 The lack of safe drinking water and basic sanita- tion could be contributing to the high levels of stunting (21.2 percent) among children under five. Waterborne diseases, including diarrhea and dysentery, are among the diseases with the highest incidence in Bhutan, and 1. All figures are population-weighted estimates. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 15 FIGURE 9  Quality of services remains highly 1,448 children per 10,000 of the population below the heterogeneous across the country age of five were infected with a diarrhea-causing dis- ease in 2017 (Ministry of Health 2018). 100 These trends suggest that the spatial trade-offs in development for a country like Bhutan need to be 80 better understood and reflected in the prioriti- zation and sequencing of policies. Low urbaniza- tion rates (only 40 percent live in urban areas) mean % of population 60 that most people live dispersed across the country- side, often in very small, remote villages and hamlets. Moreover, Bhutan’s unique geographic features—char- 40 acterized by a rugged terrain of deep valleys and steep mountains—make service provision very difficult, costly, and inefficient. Given the need to continue in- 20 vesting in human capital and improving service deliv- ery to help Bhutanese to transition to more productive jobs (including high-value tourism), important trade- 0 offs will have to be made between improving the qual- Access to Electricity 24-hour Treats Flush electricity failure in water water before toilet ity of services and providing access to services for all. last 7 days supply drinking Reliance on technological innovations and the use of Bumthang Chhukha Dagana Gasa Haa alternate modes of delivery (for example, mobile clin- Lhuentse Monggar Paro Pema Gatshel ics) could help to reduce the unit cost of delivering cer- Punkha Samdrup Jongkhar Samtse Sarpang tain services, but important policy trade-offs will still Thimphu Trashigang Trashi Yangtse Trongsa Tsirang Wangdue Phodrang Zhemgang need to be made between which services can be made available to whom and where. Proper prioritization and Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. sequencing of policies could also help in this regard. Nevertheless, overall improvements in monetary as well as nonmonetary dimensions together may have contributed to the large decline in subjective poverty between 2012 and 2017. The share of house- holds that felt “not poor” almost doubled from 10.7 percent in 2012 to 20.2 percent in 2017. The share of house- holds that felt either “poor” or “very poor” fell significantly from 25.7 percent in 2012 to 15.3 percent in 2017. However, poverty was “felt” differently according to monetary poverty status, although the difference was not large: 3 out of 10 households that are considered monetary poor also felt poor or very poor, whereas only about 13 percent of the nonpoor felt the same (table 3). In terms of the perception of happiness, the poor are sig- nificantly less likely to be happy (63.3 compared with 77.7 percent) and more likely to be unhappy (10.8 com- pared with 5.2 percent) than the nonpoor, as shown in table 4. TABLE 3  The monetary poor are also more likely TABLE 4  However, the poor are significantly less likely to “feel” poor to be happy % of respondents % of respondents   Monetary poverty measure Monetary poverty measure Indicator of poverty Not poor Poor Total Indicator of happiness  Not poor Poor Total No 21.3 11.8 20.2 Very unhappy 2.2 3.7 2.4 Neither poor nor unpoor 63.6 56.8 62.8 Moderately unhappy 3.0 7.1 3.5 Poor 11.8 24.7 13.3 Neither happy nor unhappy 17.1 25.9 18.1 Very poor 1.3 5.4 1.8 Moderately happy 39.3 37.8 39.1 Don’t know 2.1 1.3 2.0 Very happy 38.5 25.5 36.9 Total 100.0 100.0 100.0 Total 100.0 100.0 100.0 Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 16 Signs of a shift in the composition of dietary intake are emerging, but malnutrition and food security continue to be a challenge. Analysis of budget shares by food groups suggests a slow change in the compo- sition of dietary intake, as indicated by the slightly smaller expenditure on rice and cereal in the budget and higher expenditure on fruits and vegetables between 2007 and 2017. However, the consumption of meat and fish, an important source of protein, has remained almost unchanged (table 5). Moreover, the consumption of vegetables consists for the most part of two national staples—potatoes and chilies—which do not provide all of the essential vitamins and minerals (Leao and Lhaden 2018). Consequently, malnutrition remains high: 21.2 percent of children under the age of five are stunted, 9 percent are underweight, and 7.8 percent of infants are born with a low birthweight (Ministry of Health 2016). Nearly 27 percent of households consume less than the daily minimum caloric requirement of 2,124 kilocalories (El-Saharty and others 2014). In 2017, 8.6 percent of the poor reported having experienced food insecurity for at least one month in the previous year.2 The budget share does not include food consumed outside the home, a category that is likely to have increased over time, due to some irregularities observed in the data. Moreover, high food price inflation between 2007 and 2012 conflates any other structural trend in the share of food and nonfood consumption. As such, when looking at dietary composition, the budget share of different food groups is mainly considered. TABLE 5  Budget share of food groups indicate small improvements in the dietary composition % of food budget Year Rice, cereal Dairy Fish, meat Fruits Vegetables Tea, oil Spices, seasonings Beverages Share of food consumption at home 2007 15.1 10.8 6.4 2.3 6.3 4.2 4.4 4.1 53.7 2012 14.8 12.1 8.6 2.8 8.4 3.2 4.3 4.7 59.0 2017 13.3 11.1 6.8 3.5 7.9 2.5 5.7 3.7 54.3 Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2007, 2012, and 2017. Note: Does not include the value of food consumed outside the home. 2. . Calculations using BLSS for 2017. 17 II. DRIVERS OF POVERTY REDUCTION Economic growth was strong, driven by hydropower development, but it was not accompanied by job creation Bhutan’s economy continues to grow strongly, driven primarily by hydropower development led by the public sector, but there is limited spillover on the labor market. Over the last few decades, the hydro- power sector drove rapid structural transformation (in value added gross domestic product [GDP]) from agri- culture to industries other than manufacturing—mainly hydropower—and services (figure 10). This trans- formation led to a more than tenfold growth in GDP per capita between 1980 and 2016, a pace that far exceeded the regional average (figure 11). While the revenue from hydropower exports financed investments in human capital in the form of free health and education services, labor market outcomes did not change substantially because the hydropower sector contributes very little to direct job creation. The jobs created during the con- struction phase of a project are typically taken by Indian workers. FIGURE 10  The shares of industry and services in FIGURE 11  GDP per capita in Bhutan increased by more GDP (value added) have increased while the share of than ten-fold between 1980 and 2016, a much faster pace agriculture has decreased than in South Asia on average 100 3,500 2,804 26.1 3,000 GDP per capita (current US$) 80 32.4 36.3 38.6 2,500 60 21.3 % of GDP 2,000 1,640 25.7 6.1 32.2 1,500 40 34.4 8.0 8.3 1,000 20 45.1 31.4 9.1 20.0 500 259 13.0 0 0 1981–90 1991–00 2001–10 2011–17 1980 1982 1984 1986 1988 1990 1992 1994 2000 1996 1998 2002 2004 2006 2008 2010 2012 2014 2016 Agriculture Manufacturing Nonmanufacturing industry Services Bhutan South Asia Source: World Development indicators. Source: World Development indicators. The following briefly describes employment trends, but a change in the timing of the Bhutan Labour Force Survey makes it difficult to interpret longer-term trends. Prior to 2013, the Bhutan Labour Force Survey data were collected in March/April; starting in 2013, the survey was fielded in November/December. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 18 This change may have compromised comparability FIGURE 12  Labor force participation rate in Bhutan of labor market estimates over time because agricul- declined in recent years, though estimates before and tural activities tend to slow down during cold sea- after 2013 are not strictly comparable sons. Although this would not have affected trends 80 in the overall working-age population, the observed Labor force participation rate (%) decline in labor force participation as well as agricul- tural employment is possibly an artifact of the change 70 in the timing of the survey. Labor force participation was consistently above 63 percent prior to 2013, but dropped steeply in 2013 (figure 12). 60 The working-age population expanded rapidly, but labor force participation declined, driven by a large drop among women. Between 2009 and 2016, 50 2009 2010 2011 2012 2013 2014 2015 2016 more than 100,000 individuals ages 15 and above en- tered the labor force. This is a sizable number given Urban Rural Male Female Total Bhutan’s native population of about 680,000 as of 2017. Source: Bhutan Labour Force Survey (Ministry of Labour and Human Resources, various years). The corresponding figure for the period 2013–16 was about 41,000 (figure 13). The labor force participation rate for 2016 was 62.2 percent, which is notably low- FIGURE 13  Net job creation has been slow er than the estimate of 68.5 percent for 2009. Most 250 Number of persons (thousands) of this decline was due to a seemingly precipitous 11 percentage point drop in female labor force participa- 200 tion, driven by trends in both urban and rural areas. Considering only the period 2013–16 when the timing 150 of the survey was consistent, total labor force partic- 100 ipation decreased by a much smaller 3.1 percentage points, and female labor force participation decreased 50 by 5.3 percentage points (figure 12). The drop is still 0 significant, and the reasons for it are not clear. Besides 2009 2010 2011 2012 2013 2014 2015 2016 the lack of nonfarm and flexible part-time jobs, in- Public Private, nonnonagricultural sufficient child care is frequently cited as a barrier to Private, agricultural women’s labor force participation. International ex- Source: Calculations using Bhutan Labour Force Survey data (Ministry of Labour and perience also suggests that, as country’s living stand- Human Resources, various years). ards improve, female labor force participation tends to drop initially, as some women who previously had FIGURE 14  And has not kept up with the increase to work to sustain the household’s livelihood can af- in the labor force ford to become homemakers. 400 800 Number of persons (thousands) Number of persons (thousands) The pace of job creation has not kept up with the 350 700 expanding labor force. Only about 11,260 addition- 300 600 al jobs were created between 2013 and 2016. Most of 250 500 the jobs were added in the private sector, but only a 200 400 relatively small share in the nonagriculture sector 150 300 (figure 14). The increase in inactivity was greater 100 200 among younger cohorts, especially individuals ages 50 100 15–29, and among females. Unemployment rates in 0 0 Bhutan are low, hovering around 2–3 percent in most 2009 2010 2011 2012 2013 2014 2015 2016 years and only slightly higher in urban areas. Most job Employed Unemployed Inactive search is conducted by better-educated youth living Working-age pop (R) in urban areas (World Bank and Ministry of Labour Source: Calculations using Bhutan Labour Force Survey data (Ministry of Labour and and Human Resources 2016). Human Resources, various years). Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 19 Poverty is highly correlated with being engaged in agriculture Nearly 60 percent of the employed are engaged in agricultural activities, which are highly correlated with poverty. At the same time, 57 percent of employed persons were engaged in agricultural activities as of 2016 (figure 15). The share was much higher in rural areas, where it stood at 78.1 percent in 2016 (com- pared with only 5.9 percent in urban areas). The overall figure is lower than the 65.3 percent reported for 2009. However, for the same reasons described above, we conclude that the observed decline in the share of employed in agriculture could be an artifact of the change in the data collection period, especially consider- ing that the numbers change very little between 2006 and 2012 and actually go up between 2013 and 2016. If movements out of agriculture were happening on a systematic basis, one would expect to see a steadier decline over the years; however, this does not appear to have happened. Working in agriculture is highly correlated with being poor: 66 percent of the poor live in a household where the head is engaged in agricul- tural activities (figure 16). FIGURE 15  The share of employment in agriculture FIGURE 16  Working in agriculture is highly correlated remains high with being poor 100 80 70 28.3 30.7 28.8 32.8 32.6 32.6 33.1 33.1 80 60 % of population 6.4 50 9.2 9.0 % of employment 6.8 10.9 10.7 9.5 9.8 60 40 30 40 20 65.3 62.2 60.0 60.1 56.3 56.7 58.0 57.2 10 20 0 working in working in not working agriculture nonagriculture 0 Nonpoor Poor Urban Rural Total 2009 2010 2011 2012 2013 2014 2015 2016 Agriculture Industry Services Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. Note: Figure shows the share of poor, nonpoor, rural, urban and total population Source: Calculations using Bhutan Labour Force Survey (Ministry of Labour and living in a household whose head is working in agriculture, nonagriculture or is not Human Resources, various years). working. Consistent with the high share of workers in ag- FIGURE 17  Urban households rely mainly on wage riculture, wage employment remains low overall income, whereas rural households draw income from and concentrated in urban areas. Only 21 percent various agricultural products of rural households report that their main source of 1.0 income draws from wages (figure 17). More than half Wages Weaving 0.9 Cereal Remittances obtain their main income from farming (51 percent), 0.8 Fruits Pensions followed by business income (18 percent) and sale of Share of income (%) 0.7 Vegetables Real estate assets (6 percent). In contrast, 70 percent of urban 0.6 Meat Inheritance households rely on wage income as their main source 0.5 Dairy products Donations of income, and less than 10 percent rely on farming or 0.4 Eggs received business income. The number of internal as well as ex- 0.3 Forest wood Scholarships ternal Bhutanese migrants has been rising, especially 0.2 products Sale of assets among the younger cohorts who seek education and Forest nonwood Business 0.1 products job opportunities abroad, but the contribution of re- income 0.0 Pottering mittances to households’ living expenses remains lim- Urban Rural ited, with less than 2 percent reporting remittances Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. as their main source of income. Note: Figures show households’ main source of income. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 20 FIGURE 18  Growth incidence curves show that consumption growth was slower among poorer households a. 2007–17 b. 2007–12 c. 2012–17 12 7 8 6 Growth rate (%) Growth rate (%) Growth rate (%) 6 10 5 4 4 8 2 3 6 2 0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 Percentiles Percentiles Percentiles Upper 95% confidence bound/Lower 95% confidence bound Median spline Growth rate in mean Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2007, 2012, and 2017. Consumption growth was lower among poor- FIGURE 19  Datt-Ravallion decomposition confirms that er households, which reduced the pace of pover- consumption growth was very unequal ty reduction. Between 2007 and 2017, consumption 30 growth was unequal across the distribution, with 20 households toward the lower end experiencing slow- 10 er growth in per capita consumption than better-off 0 % change households (figure 18). Results from a Datt-Ravallion –10 –20 decomposition exercise (Datt and Ravallion 1992) con- –30 firm this: results suggest that the 18.6 percentage –40 point reduction in poverty between 2007 and 2017 is –50 mainly explained by consumption growth (−35.1 per- –60 Rural National Urban Rural National Urban Rural National Urban cent) rather than redistribution (16.4 percent) (fig- ure 19). In other words, if consumption growth had been distributed more equally, poverty would have 2007–2012 2012–2017 2007–2017 fallen much faster.3 Consumption growth was signifi- Growth Redistribution cantly more pro-poor between 2007 and 2012 than be- Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2007, 2012, tween 2012 and 2017, when it was still positive along and 2017. Note: Figure shows results from Datt-Ravallion decompositions that break down the distribution, but consumption was growing more changes in poverty into changes due to consumption growth and changes due to slowly for the poor. distribution (Datt and Ravallion 1992). Poverty reduction was driven by rural, agricultural households Bhutan’s mountainous terrain presents large challenges for agriculture, the most importance source of livelihood for the poor. Agricultural land is limited, as Bhutan maintains a constitutionally mandated level of forest cover of at least 60 percent, with current coverage being around 70 percent. Of the available agri- cultural land, more than 30 percent is located on slopes of more than 50 degrees, and an increasing share is being converted to nonagricultural uses (Keturakis and others 2017). The sector’s productivity suffers from low input use, low mechanization, lack of irrigation facilities, wildlife predation, and labor shortages due to continued rural-urban migration. Nevertheless, poverty reduction between 2012 and 2017 was driven by changes in the rural, agri- culture sector. A simple decomposition exercise shows that the reduction in poverty over these years was 3. Poverty measures can be written as a function of the poverty line, mean income, and a vector of parameters that fully describe the Lorenz curve. The Datt-Ravallion decomposition attributes changes in poverty to changes in the mean income of the distribu- tion, while holding the Lorenz curve constant (the “growth” component), and to changes in relative inequality, while holding the mean constant (the “redistribution” component) (Datt and Ravallion 1992). Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 21 driven mainly by changes in poverty among house- FIGURE 20  Huppi-Ravallion decomposition shows that holds whose household head worked in agriculture poverty reduction was driven by improvements among (figure 20). A similar decomposition by urban-rural agricultural households areas shows that almost all (more than 90 percent) of 0.01 poverty reduction took place in rural areas (results not shown here), which is not surprising given that 0 poverty is concentrated in rural areas. Based on the fact that the share of agricultural employment did not –0.01 change much, improvements among rural agricultur- al households are likely to have come from increased % reduction –0.02 agricultural earnings, due either to improved produc- tivity or to higher prices. –0.03 The lack of information on economic activities and –0.04 remuneration in the household survey makes it challenging to establish drivers of poverty reduc- –0.05 tion. Detailed information on household-level agricul- tural activities and household income are key pieces of –0.06 information needed to have a better understanding of Intrasectoral Population shi Interaction effect effect effect progress in poverty reduction and welfare. Only very Agriculture Industry Services limited questions are addressed in the BLSS, which is Unemployed Inactive used for measuring poverty—for example, what are the Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2012 sources of income, and what are the main categories and 2017. of products produced? For a country with a large agri- Note: Figure shows results from Huppi-Ravallion decomposition (Huppi and Ravallion 1991) that breaks down changes in poverty into intrasectoral effects (contribution culture sector (by employment) and widespread infor- of poverty changes within sectors), population shift effects (contribution of changes mal activities, these are critical data gaps. To fill these in poverty due to changes in population shares of sectors), and interaction effects (correlation between sectoral gains and population shifts). Sectors of activity and gaps, we draw on secondary or aggregate data sources employment status (unemployed or inactive) are defined using information on the to describe the likely factors behind poverty reduction. household head. Aggregate data suggest that agricultural earnings gradually increased Despite its declining share of the economy, agricultural productivity has increased gradually. The share of agriculture, forestry, and fishing in the economy has declined steadily (figure 21), mainly because the rest of the economy—in particular, hydropower and related manufacturing sectors—has expanded signifi- cantly (figure 22). However, a longer-term trend also shows that overall output and productivity in the agri- culture sector has made notable improvements, especially since around 2010: between 2007 and 2012, output measured as value added increased 9.2 percent and labor productivity measured as value added per worker increased 9.8 percent (figures 23 and 24). The latter can be attributed to a decline in the number of household members engaged in agriculture (World Bank 2018). Between 2012 and 2017, output and labor productivity growth were even higher, at 18.2 percent and 14.6 percent, respectively. These improvements are noteworthy, and it is important to understand the drivers behind these recent changes, especially since 2010. For example, World Bank (2018) suggests that output growth was driven primarily by greater intensity of input use rather than by land expansion or broad-based productivity growth. We next examine agricultural statistics by sub- sector, followed by price and yield data (which combined determine productivity and earnings in the sector) by type of agricultural product, to identify the reasons for these changes. The crop subsector contributed to steady improvements in agricultural earnings as commercial farm- ers increasingly benefited from cash crop farming. The agriculture sector has three main subsectors: crops, livestock, and forestry. The contribution of livestock and forestry to GDP declined significantly in the last dec- ade. However, the output value in current prices in the crop subsector grew rapidly, at 71 percent, between 2012 and 2016, maintaining its share of GDP. Crops account for an increasingly larger share within the agri- culture sector (table 6). Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 22 FIGURE 21  Agriculture’s share in the economy continued FIGURE 22  However, the output of the agriculture, to decline (Value added in Bhutan, by sector, 2000–17) forestry, and fishing sector grew notably in recent years 40 50 Annual growth of value added (%) 35 40 30 25 30 % of GDP 20 15 20 10 10 5 0 0 2000 2001 2002 2003 2004 2009 2011 2013 2014 2000 2005 2006 2007 2008 2010 2012 2015 2016 2017 2001 2002 2003 2004 2009 2011 2013 2014 2005 2006 2007 2008 2010 2012 2015 2016 2017 Agriculture, forestry, and fishing Manufacturing Agriculture, forestry, and fishing Nonmanufacturing industry Services Industry (including construction) Services, value added Source: World Development indicators. Source: World Development indicators. FIGURE 23  Output in the agriculture, forestry, and FIGURE 24  Value added per worker in agriculture, fishing sector increased forestry, and fishing also improved (value added, constant Nu) (constant 2010 U.S. dollars, 2000–17) 9 50 2.0 Value added (constant US$, thousands) Value added (constant US$, thousands) 8 40 1.6 7 6 30 1.2 Nu (bilions) 5 20 0.8 4 3 10 0.4 2 0 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 1 0 Agriculture, forestry, and fishing (right axis) 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Industry (including construction) Services Source: World Development indicators. Source: World Development indicators. TABLE 6  The crop subsector has been growing steadily Indicator and subsector 2008 2009 2010 2011 2012 2013 2014 2015 2016 GDP at current prices (Nu, millions) Agriculture, livestock, and forestry 10,078 11,159 12,178 13,459 15,558 16,970 20,050 22,008 24,565 Crop 5,061 5,668 6,530 7,488 8,635 9,405 12,029 13,340 14,795 Livestock 2,625 2,895 3,110 3,241 3,997 4,537 4,869 5,237 5,779 Forestry and logging 2,393 2,596 2,538 2,730 2,926 3,028 3,152 3,431 3,991 Sectoral growth rate in constant prices (%) Agriculture, livestock, and forestry 0.71 2.66 0.3 1.54 2.25 2.4 2.37 4.56 3.65 Crop 1.86 2.24 1.17 0.47 2.35 3.38 3.98 5.74 6.54 Livestock 0.54 2.35 1.04 2.22 1.27 2.04 2.31 3.36 5.54 Forestry and logging −1.35 3.88 −2.33 2.93 3.29 0.82 −0.97 3.51 −5.42 Source: Statistical Yearbook for each year (NSB, various years). Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 23 TABLE 7  Value and amount of livestock products have increased in recent years Value of livestock products Livestock products produced Nu, millions (real prices) Growth (%) Metric tons Growth (%) Product  2007 2012 2017 2007–12 2012–17 2007 2012 2017 2007–12 2012–17 Milk 2,553 2,567 5,719 0.6 122.7 20,059 29,625 50,014 47.7 68.8 Eggs 67 481 1,313 612.3 173.2 802 4,759 9,424 493.4 98.0 Cheese 1,422 583 1,123 −59.0 92.5 6,146 2,300 4298 −62.6 86.9 Butter 372 270 527 −27.5 95.4 1,394 1,027 2,053 −26.3 99.9 Chicken 87 165 227 90.7 37.7 515 909 1,349 76.5 48.4 Pork 111 55 189 −50.4 243.1 507 317 919 −37.5 189.9 Beef 109 91 120 −16.4 31.9 638 513 501 −19.6 −2.4 Yak meat 36 33 71 −7.7 114.9 117 106 134 −9.4 26.7 Mutton 60 13 43 −78.9 236.3 239 63 195 −73.6 208.8 Fish 0 9 39 6,635.0 321.2 1 65 200 6,400.0 207.6 Source: Statistical Yearbook and Livestock Statistics (NSB, various years). Note: Table shows the value of livestock production (Nu, millions, real prices). The values are estimated using the quantity produced (from Livestock Statistics) multiplied by the estimated median price of the product (derived from Bhutan Living Standards Survey). Prices were deflated using the food consumer price index. While there is limited information on agricultur- FIGURE 25  Earnings from main fruits and crops al activities at the household level, aggregate ag- in Bhutan improved ricultural statistics point to continued earnings 1,000 increases in cash crop farming. A handful of fruits and crops account for most of the earnings in the sub- sector: among fruits, apple, areca nuts, and mandarin 800 Earnings (Nu, millions) account for more than 90 percent of earnings from fruits nationally, whereas among vegetables, high-val- 600 ue crops such as cardamom, chili, and potato make up 80 percent of earnings from crops nationally (table 8). 400 A large share of harvest land is devoted to maize and paddy (57 percent in 2017), but their contribution to 200 earnings is negligible at 2 percent. This is likely be- cause these crops are staples that are produced mainly for own consumption. Moreover, the incentive to pro- 0 2012 2013 2014 2015 2016 2017 duce cereal for the market is likely weak given high Apple Areca nut Mandarin competition from low-price rice imports from India Cardamom Chilli Maize (Christensen, Fileccia, and Gulliver 2012). The farm- Paddy Potato ing of high-value crops, especially cardamom, con- Source: Agriculture Statistics for each year (NSB, various years). tributed to higher agricultural earnings, but there are Note: Figure shows earnings from main fruits and crops (Nu, millions), deflated large fluctuations in some years (figure 25). using food consumer price index (December 2012 = 100). Growth in the livestock sector was driven mainly by dairy products, which contribute to higher earn- ings, but growth potential is low because this subsector caters to a very small domestic market. Livestock data are difficult to gather, and the household survey only asks about livestock ownership. We rely on national statistics to understand the trend and growth potential in this sector. Butter, milk, and cheese, which are staples of the Bhutanese diet, are top items by production. The production of eggs has seen particularly strong growth recently and is the second most important product by value (table 7). Bhutan’s livestock levels are reported to be nearly twice as high as those of India for cattle and buffalo combined, but a large share of Bhutan’s demand for meat is met by imports: for example, in 2010, 80 percent of beef, 97 percent of fish, and 77 percent of pork that was consumed was sourced from imports (Keturakis and others 2017). Overall, the livestock sector will likely continue to grow and contribute to rising earnings among livestock farmers, but given that the demand Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 24 is driven entirely by the small, primarily urban, do- TABLE 8  A handful of fruits and crops account mestic market and that farmers face strong competi- for most earnings tion from the Indian market, the potential growth of % of national % of national this sector is lower than that of cash crops. Year Fruit bearing trees fruit earnings 2011 Apple 16.05 15.76 However, volatile prices 2011 Areca nut 24.35 11.15 2011 Mandarin 49.47 67.83 of major fruits and crops raise 2017 Apple 11.1 8.4 concern over sustainability of these 2017 Areca nut 38.3 44.7 improvements 2017 Mandarin 37.0 38.3   Subtotal 86.4 91.4 Rural agricultural households likely benefited % of national % of national from overall positive trends that led to improved Year Crop harvested area crop earnings agricultural earnings, but volatile prices increase 2012 Cardamom 2.69 31.37 the variability of agricultural earnings and call sustainability into question. Data on aggregate 2012 Chili 4.13 11.22 yields, earnings, and prices presented here are from 2012 Potato 6.64 32.50 annual Agriculture Statistics (NSB, various years). 2012 Maize 33.61 1.83 Data become more detailed only after 2011, so there 2012 Paddy 27.67 2.17 is less information on sectoral trends in the years pri- 2017 Cardamom 6.8 43.2 or to that. Yields remained largely stagnant for fruit 2017 Chili 3.7 12.1 crops but showed marked improvements for potato and chili (figure 26). However, prices exhibited large 2017 Potato 6.2 25.0 fluctuations in the 2012–17 period, particularly for 2017 Maize 32.2 0.8 cardamom and chili, the most profitable crops (fig- 2017 Paddy 25.0 1.3 ure 27). Higher prices can lead to higher earnings if   Subtotal 16.7 80.3 agricultural households are net producers. Food price inflation was much higher on average between 2007 Source: Agricultural Statistics for each year (NSB, various years). FIGURE 26  The increase in average yields of key fruits FIGURE 27  Prices of key fruits and crops rose but some and crops helped improve earnings have been volatile 5,000 50 1200 4,500 45 1000 Yield (kilograms per bearing tree 800 Mean unit price per kilogram (Nu) 4,000 40 Yield (kilograms per acre) 600 3,500 35 400 3,000 30 80 2,500 25 70 2,000 20 60 1,500 15 50 1,000 10 40 500 5 30 0 0 20 2011 2012 2013 2014 2015 2016 2017 10 2012 2013 2014 2015 2016 2017 Apple Arecanut Mandarin Cardamom Chilli Maize Apple Areca nut Mandarin Paddy Potato Cardamom Chilli Maize Paddy Potato Source: Agriculture Statistics for each year (NSB, various years). Source: Agriculture Statistics for each year (NSB, various years). Note: Yield is measured as kilograms per bearing tree for apple, areca nut, and mandarin (right axis); and as kilograms per acre for cardamom, chili, maize, paddy, Note: Figure shows the mean unit price (Nu per kilogram) of main fruits and crops. and potato (left axis). No yield data for fruits in 2012. Prices are deflated using food consumer price index (December 2012 = 100). Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 25 and 2012 (the food consumer price index [CPI] during this period indicates a 68 percent increase in average food prices), potentially influenced by the global food price crisis, which could have benefited some agricul- tural producers. Subsistence farmers tend to be shielded from price shocks if most of what they consume comes from own-farm production and market access is generally difficult, while any surplus that is market- ed would result in higher earnings. Vulnerability is high, partly because farmers are exposed to various uninsured risks, but also because the social protection system is weak overall and nonfarm diversification is low The exposure to frequent shocks contributes to high levels of vulnerability. While higher prices can increase earnings during good times, subsequent price drops can have the opposite effect if farmers are not sufficiently insured against price shocks. Formal insurance mechanisms to protect farmers are not in place for most subsistence and smallholder farmers. The Himalayan mountains are at substantial risk of climate change, and more extreme weather variations are increasingly expected to affect agricultural production. Commercial farmers face many challenges in marketing their output: access to transport infrastructure has improved, but market access remains an issue, as suggested by the low amount and share of earnings that orig- inate from exports (table 9). The market surplus of these export crops could be much higher if marketing sys- tems were better and postharvest losses were lower (Christensen, Fileccia, and Gulliver 2012). External food price shocks are likely to have wide implications, as Bhutan imports 34 percent of its cereal needs. Wildlife predation and increasing variability in weather further contribute to vulnerability. TABLE 9  A few fruits and crops dominate the agricultural export market Earnings from exports (Nu, millions) Share of earnings from exports (%)   Fruit or crop 2012 2013 2014 2015 2016 2012 2013 2014 2015 2016 Fruit Apple 37 31 31 35 28 34.9 20.3 24 27 13 Areca nut 21 19 31 27 29 28.4 20.2 23 17.9 20 Mandarin 138 138 194 44 117 30.5 29.8 35 26 27 Crop Cardamom 127 332 445 343 113 34 53 45 30 20.2 Chili 3 3 4 3 3 1.9 2 3 2 1 Maize 2 4 3 1 0 7.9 10 17 3 0.3 Paddy 0 1 0 0 0 1 7 1 2 0.8 Potato 113 134 299 138 132 29.1 24 35 26 16.5 Source: Agriculture Statistics for each year (NSB, various years). Note: Data for 2017 are not available. Share in earnings from exports is calculated separately for fruits (apple, areca nut, mandarin) and crops (cardamom, chili, maize, paddy, potato). Although poverty has declined, estimated vulner- TABLE 10  Small changes in the poverty line lead to large ability remains high, with many households re- changes in the poverty headcount rate in Bhutan, 2017 maining just a small shock away from poverty. If poverty rate (%) the poverty line in 2017 had been just 15 percent (or Nu 354 per month) higher, the poverty headcount rate Cutoff Poverty rate in 2017 would have remained unchanged from 2012. 80% of poverty line 5.6 This is equivalent to about US$14 (in 2011 PPP) per Poverty line, 2017 12.1 person per month, or less than half a dollar per day. 120% of poverty line 20.1 To put this number in perspective, selling 3.5 kilo- 150% of poverty line 31.7 grams of chilies, half a kilogram of cardamom, or 10 kilograms of apples could have made the difference Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 26 between being poor and nonpoor in 2017. As another benchmark for these numbers, we perform a back-of-the- envelope calculation for cardamom: in 2017, 1.75 million kilograms of cardamom were produced (Agriculture Statistics for 2017; NSB, various years). Mehta, Rabgyal, and Acharya (2015) report that there are about 17,000 cardamom farmers in Bhutan. This suggests that the average cardamom farmer produced about 100 kilograms of cardamom. Although not all cardamom farmers were poor initially, those who were poor could have been lifted out of poverty thanks to favorable price trends. But small shocks to yields or prices could lead them to fall back into poverty. Put differently, a large concentration of households is tightly clustered around the poverty line, and small movements in the poverty line introduce large changes in the poverty rate. On the one hand, lowering the poverty line 20 percent reduces poverty to just 5.6 percent. On the other hand, raising the pov- erty line a mere 20 percent almost doubles the poverty headcount from 12.1 percent to 20.1 percent. Adding a 50 percent layer to the poverty line almost triples the headcount to 31.7 percent (table 10).4 Despite such vulnerabilities, the social protection system in Bhutan is weak, with limited focus on poverty alleviation or resilience. Coverage of social insurance schemes is limited: pensions are mostly reserved for public sector workers, and there is currently no unemployment insurance scheme. Spending on social assistance schemes accounts for about 0.9 percent of GDP (excluding spending on free health and edu- cation) through a combination of universally provided and targeted programs. With the exception of univer- sal programs (such as free electricity, health care, and education), government-financed programs—Kidu and Support for Agriculture—reach only a small proportion of poor households, but this support is intended as hardship relief as opposed to poverty alleviation. The Kidu Program is an example of such support: it is Bhutan’s main social assistance program and a royal prerogative (as enshrined in the constitution). The pro- gram grants assistance in various forms, mainly through land, but also through citizenship, education schol- arships, and support for people living in destitution, the elderly, and the disabled.5 However, individuals com- ing from these target groups do not necessary belong to the poorest households. Between 2010 and early 2018, many households have been endowed land through the Kidu: 123,254 beneficiaries in 171 gewogs received a total of 133,288 acres of land, implying an average endowment of about 1 acre.6 While the coverage of the pro- gram is significant, little systematic information is available on who receives the land and how it is being used, but it is understood to be used for both residential and agricultural purposes. The majority of endowed land had already been in the families’ use even though they were not the formal land holders.7 The Targeted Household Poverty Programme is the only household-level intervention program; it provides support for hous- ing improvements, supply of agricultural equipment, and income-generating activities such as dairy farm- ing and cash crop cultivation. It relies on easily identifiable indicators to determine beneficiaries, but the tar- geting criteria result in high exclusion and inclusion errors: an assessment conducted in 2017 found that the program’s criteria could include as many as 60 percent of Bhutanese households and fail to reach 50 percent of households that are actually poor.8 Increasing opportunities in nonfarm activities can help households to diversify their income sources and mitigate the impact of shocks, but there are few such opportunities. The employment rate is higher in rural areas, but this reflects the nature of agricultural activities, where the barriers to being classified as “working” are low. Among those who report being engaged in agricultural activities, more than 70 percent report doing so mainly or entirely for their own consumption (figure 28). That figure is a staggering 80 per- cent for the poor. Moreover, almost 80 percent of households indicate that all of their working household mem- bers are engaged in agriculture, indicating few opportunities for diversification outside of the farm sector (figure 29). Despite the larger household size among the poor (6.7 people per household) compared to nonpoor 4. A similar calculation with the official consumption aggregate shows that if the poverty line in 2017 had been just Nu 247 higher, poverty between 2012 and 2007 would not have declined. 5. More detailed information on coverage and benefits is not publicly available and difficult to obtain. 6. See http://www.kuenselonline.com/hm-grants-land-kidu-to-9628-beneficiaries-in-samtse/. 7. This program may have contributed to the slight increase in landownership observed in recent years and potentially to output growth in some local areas, but it is not possible to identify Kidu recipients in the household survey. 8. The assessment was part of a World Bank Technical Assessment to evaluate the targeting performance of the Targeted Household Poverty Programme. About 3,000 households were initially selected under the program. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 27 FIGURE 28  80 percent of poor households produce mainly FIGURE 29  There are few opportunities for nonfarm or solely for family consumption in Bhutan, 2017 diversification for the rural poor in Bhutan, 2017 60 100 % of households or workers 50 80 % of households 40 60 30 20 40 10 20 0 Only for sale Mainly for sale Mainly for Only for family 0 family consumption working households agricultural households with consumption members workers agricultural workers only Total Poor Nonpoor Total Poor Nonpoor Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. Note: Bar graph indicates share of total, poor, and nonpoor households that report Note: Bar graphs indicate (in the following order) share of household members ages engaging in agricultural activities for the purposes of only for sale, mainly for sale, 15 and above who are working; out of those, the share who are agricultural workers; mainly for family consumption, or only for family consumption. and the share of households with all working members engaged in agriculture. households (4.9 people per household), the likelihood of working is similar among the poor and non-poor. The lack of diversification requires further support from the government to enable nonfarm private sector growth (including support for household enterprises) in rural areas. In addition, to achieve its full impact, this support should be combined with others activities to improve productivity of the rural population, such as on-the-job training programs, structured apprenticeship programs, technical assistance to enable farmers to incorpo- rate into value chains, entrepreneurship training in combination with support for access finance, technical assistance in the preparation and execution of business plans, and better support for the creation of favora- ble business ecosystems (for example, logistics, access to finance, regulations). Enhancing the design of social protection programs to promote productive inclusion would help farm- ers to insure against idiosyncratic and covariate shocks, as well mitigate poverty. While a significant number of poor farmers have managed to escape poverty in recent years, many remain near the poverty line and at risk of falling back into poverty due to the high vulnerability of Bhutan. Therefore, tailored produc- tive inclusion programs for the poor and subsistence farmers are necessary to address these risks. Provision of technical support to improve agricultural production, agriculture inputs, self-management of coopera- tives or groups, technical assistance on new technologies as smart agriculture, and microcredit for the poor and subsistence farmers would help to develop more productive and climate-smart farming activities, while strengthening resilience. More and better data would be particularly useful for understanding agricultural production. Better measurement of subsistence agriculture, including identifying good practice standards to separate own-use production work (for example, subsistence agriculture) from employment for pay or for profit, could improve our understanding of the nature of agricultural labor and influence policy decisions. To summarize, poverty reduction was likely helped by improved earnings in the commercial agricul- ture sector, but vulnerability is high and most farming activities are mainly of subsistence nature. A critical data gap makes it difficult to make direct linkages between agricultural activities and poverty reduc- tion, but households that engaged in commercial farming likely would have benefited from positive albeit vol- atile trends in prices and yields. The high share of subsistence farmers is focused on maize and paddy, whereas increases in yields and prices were observed among cash crops such as cardamom, potato, and chili. While still low, these crops also account for an increasingly larger share of agricultural exports. The sector could bene- fit from Bhutan’s proximity to large markets, potential for low-volume, high-value fruits and vegetables, and, most important, government support for commercial farming. Going forward, better measurement of labor statistics and price data could help to inform poverty reduction and other development strategies. There are several areas where better measurement could improve our understanding of the agriculture sector, labor utilization, as well as price trends. First, following Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 28 the 19th International Conference of Labour Statisticians (ICLS) in 2013, the International Labour Organization (ILO) issued new guidelines to narrow the concept of employment and introduce a new concept of work and labor underutilization. An important contribution of the new ICLS standards was to narrow the definition of employment to include only work performed for pay or profit. Unlike under the previous definition, the pro- duction of goods for own consumption is no longer considered employment. In addition, the 2013 resolution introduces a new category of work, which recognizes all productive activities, paid and unpaid, and proposes several measures of labor underutilization. Specific guidelines for how to operationalize these concepts have not been recommended yet, and several pilot studies are being rolled out across countries. See box 2 for more details. Second, analysis revealed an important divergence in food price inflation in different data sources. Between 2007 and 2012, survey data suggest that households faced food prices that were significantly higher than what the official food CPI shows. This divergence is described further in table A.3, which compares the food CPI and a survey-based food price index that uses unit prices calculated from the household survey. For this note, the food poverty line is updated consistently using the CPI. This has the potential to underesti- mate the poverty line and therefore the poverty headcount rate. If this were the case, poverty reduction dur- ing this period would appear less impressive. Despite this potential underestimation, we maintain the CPI as the default deflator throughout this period given data reliability concerns. Better measurement and a better understanding of important labor and price statistics are key to making decisions related not only to poverty reduction but also to economic development. BOX 2  Recommendation of the 19th ILO International Conference of Labour Statisticians concerning statistics of work, employment, and labor underutilization In October 2013, the 19th ICLS of the ILO adopted a ground-breaking resolution concerning the statistics of work, employment, and labor underutilization. The new guidelines narrowed the concept of employment and introduced a new concept of work and several measures of labor underutilization. The implications of the new guidelines are as follows: • Under the revised (narrower) definition of “employment,” farmers who produce only for subsistence purposes are no longer counted as employed and are considered to be outside the labor force (see figure B2.1 for a schematic overview of productive activities). • Given the large number of workers—especially women and children—involved in these activities in low- and middle-income countries, the revised standards are likely to result in significantly lower estimates of employment and labor force participation in the low- and middle-in- come world, especially in countries or regions where subsistence agriculture is common. • Services produced for own final use (such as care for children and the elderly, food preparation, and other household chores), which are often performed by women and children and were not captured by the previous definition of employment, are now recognized as “work” under the category own-use production. The new ICLS standards lay the ground for a better measurement of work activities. • Additional measures of labor underutilization will be useful to complement unemployment statistics and provide a more complete picture of labor underutilization, including time-related underemployment and discouraged workers, among others. The new definitions pose chal- lenges in terms of comparability (for example, series breaks, comparisons between labor statistics and national accounts) and have major implications for the production and communication of key statistics both nationally and internationally. • The new guidelines are being rolled out gradually across countries, with ILO having conducted initial pilot studies in 10 countries in Africa (Cameroon, Côte d’Ivoire, Namibia, and Tunisia), Latin America (Ecuador and Peru), East Asia and Pacific (the Philippines and Vietnam), and Eastern Europe and Central Asia (the Kyrgyz Republic and Moldova). The World Bank conducted initial pilot studies in Ghana and Malawi and is currently conducting a joint World Bank–ILO study in Sri Lanka. FIGURE B2.1  Revised statistical standards for measuring work and employment Productive activities Nonproductive activities Market units Nonmarket units Households Sleeping, learning, (incorporated, unincorporated) (government, nonprofit institutions) (producing for own final use) own-creation, begging, Goods Services Goods Services Goods Services stealing Source: Summary prepared by Isis Gaddis (World Bank). Note: For more details, see https://www.ilo.org/global/statistics-and-databases/meetings-and-events/international-conference-of-labour-statisticians/19/ WCMS_230304/lang--en/index.htm. 29 III. PRIORITIES FOR SUSTAINABLE POVERTY REDUCTION Bhutan has made progress toward reducing poverty in the past decade, but large challenges remain. Poverty fell between 2007 and 2017, and broader measures of welfare, including access to services and asset ownership, improved significantly. However, many of the rural poor are engaged in agricultural activities, which, in the absence of formal insurance mechanisms, exposes them to weather and price shocks. As a result, a large share of the population remains vulnerable to falling back into poverty. Over the next two decades, Bhutan has an opportune window to catalyze private sector–led growth that can help to create more productive jobs. Hydropower revenues are likely to remain an important source of government financing over the medium term. With a dependency ratio that is projected to decline until 2040, the next two decades offer an opportunity for Bhutan to reap its demographic dividend if the expanding work- force can be absorbed by more productive jobs. The labor force is increasingly educated, and urbanization is likely to continue as a result of steady rural to urban migration. This trend, combined with rising aspirations and increased access to information, will generate growing pressure to create jobs outside the agriculture sec- tor—the latter is particularly important to sustain poverty reduction and promote broad-based shared pros- perity in the long run. At the same time, it is important to address the low and declining female labor force par- ticipation. Although female labor force participation tends to fall as living standards rise and then to rebound, it is important to understand and address the drivers behind the relatively large drop in the last decade. A few priorities emerge from our analysis: first, the creation of productive jobs is key to achieving sus- tainable welfare improvements in the long run. Bhutan’s public sector–led growth model has helped to boost economic growth and direct large investments in human capital. However, the private sector remains highly underdeveloped, accounting for only half of the jobs outside of the agriculture sector. The other half consists of public sector jobs that are generously remunerated with comprehensive benefits and contribute to high res- ervation wages and unemployment among the educated urban youth (World Bank and Ministry of Labour and Human Resources 2016). Given that land is an important binding constraint to productivity, it will be difficult for the agriculture sector to generate large income gains in the long run. Continued investments in physical and human capital could facilitate the transition to jobs in higher value-added markets, including high-end tourism. Such a transformation could provide a significant lift to the incomes of the poor, many of whom are less skilled and have few means with which to generate income other than their labor. In this context, the spatial trade-offs in development in a country like Bhutan need to be better under- stood and reflected in the prioritization and sequencing of policies. Despite rapid urbanization in recent years, the share of population living in urban areas remains relatively low at around 40 percent. Most people live dispersed across the countryside, often in very small, remote villages and hamlets. The mountainous ter- rain and poor access isolate a high share of the rural population, making service provision costly and ineffi- cient. While tremendous progress has been achieved in terms of access, quality is the main bottleneck. Clear Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 30 priorities emerge from the household survey: a com- FIGURE 30  Basic services and roads top the priority bined 44.3 percent of respondents requested improve- areas for government action, selected by survey ments in water supply and electrification. The lack of respondents access to infrastructure is also a strong factor: almost 25 a quarter of respondents selected transport services (building roads, providing public transport service) 20 % of respondents as a priority (figure 30). The poor and nonpoor selected 15 the same top three priority areas by a wide margin. 10 First, given the need to continue investing in 5 human capital, important trade-offs will have to 0 be made between improving the quality of services Water supply Electrification Build roads Public transport service Waste management Hospitals/medical facilities Job creation Housing Other Building new schools Improving existing schools Provision of medicines Subsidized agricultural equip Improved sanitation Vocational training Food assistance Local religious services Boarding for students Family planning Credit and providing access to services for all. Progress was achieved in access to education mainly at the primary and secondary levels and in access to health services mainly in primary care. Going forward, the challenge will be to improve quality across-the-board, further raise educational attainment (while at the same time providing training to the large working-age population with no formal education), and increase the availability of specialized health care. Reliance Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2017. on technological innovations and the consideration of alternate modes of service delivery could help to reduce the unit cost of delivering certain services but does not address important policy trade-offs regarding which services can be made available to whom and where. Proper prioritization and sequencing of policies could also help in this regard. At the same time, a shift in pol- icies could be considered, away from a focus on spatially balanced development and toward a focus on improv- ing efficiency and promoting and reaping the benefits of urban agglomeration. Second, poverty reduction will be difficult to achieve in the short to medium term without raising agricultural productivity. Most of the current generation will remain engaged in agriculture, and it will take time for the nonfarm private sector to develop and create jobs. To increase rural incomes, it is critical to address the constraints to growth in the commercial agriculture sector. The recent expansion in the latter is seen in the increased number of agribusinesses, many of which are located in the Thimphu/Paro area, where labor and skills are concentrated, or along the Indian border, where there are trade points (World Bank 2018). Further interventions are needed to improve productivity (with regard to input use, irrigation facilities, agri- cultural research and extension services, and others), combined with efforts to develop marketing systems. Market infrastructure and wholesale systems are largely absent at the domestic level, and marketing activi- ties continue to rely heavily on public agencies such as the Food Corporation of Bhutan (Christensen, Fileccia, and Gulliver 2012). Strengthening the social safety net and formal insurance mechanisms would further pro- tect vulnerable households from shocks. Finally, further efforts are needed to strengthen statistical capacity, improve data quality, and fill data gaps. Much progress has been made in strengthening data standards, but large gaps remain. Agriculture is by far the most important sector of work, but there is very little information on household-level production activities that would allow for a better understanding of the drivers of welfare and poverty of rural house- holds. The next round of the BLSS could consider including a standard Living Standards Measurement Survey– type agricultural module, which collects detailed information on inputs and outputs at the crop level for all households that engage in agricultural activities. The lack of information on household income is another data gap. Income data are largely absent from the BLSS, except for some limited information on income sources. The 2012 BLSS asked about income amounts, but the questions were removed in the 2017 survey round, pos- sibly due to data quality concerns. The annual Labour Force Surveys capture wages, but given that the poor are reliant primarily on agricultural and other informal activities, the data are less suited to capture welfare changes among the poor and bottom 40 percent. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 31 There is also a continued need for strong technical assistance and statistical capacity. The following are several areas of priority: (a) improve the measurement and analysis of poverty to ensure consistency and comparability; (b) improve the quality of data collected and improve sampling to enhance precision of esti- mates derived from household surveys ; and (c) ensure regular collection of essential socioeconomic data—for example, no recent data are available on population health, which could be collected through a Demographic and Health Survey or a Multiple Indicator Cluster Survey. The latter survey was last collected almost 10 years ago. On this point, given the high cost of standard surveys and difficulty of raising funding, alternative ways of data collection could be explored that are more cost-effective for small states such as Bhutan. Finally, given that the official poverty line is relatively low, with official poverty estimates at 8.1 percent as of 2017, and bet- ter data are available (especially on durables consumption) starting with the 2017 BLSS, the official poverty line should be reestimated and updated based on the latest consumption patterns for the next round of the BLSS. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 32 Annex Official and comparable poverty measure- ment methodologies Steps for estimating a comparable poverty trend The poverty trend that is used throughout this note (referred to as the “comparable” trend) deviates from the official poverty estimates, which were published in Poverty Analysis Reports (NSB, various years) following the completion of data collection. The use of a comparable trend aims to ensure that the analysis undertaken to understand the drivers behind changes in poverty rates is based on a trend that is deemed more comparable across years. The rest of this annex explains in more detail how the comparable consumption aggregate is con- structed and how the method of choice for deflating food consumption across years is made comparable. Notably, the mode of data collection switched to computer-assisted personal interviewing (CAPI) for the 2017 survey round, although it is not clear whether this change introduced additional comparability issues across years. The annex first describes how the comparable consumption aggregate is constructed in nominal terms and then how it is spatially deflated to account for price differences across Bhutan. The consumption distribution is compared against a threshold—that is, a poverty line—below which households are classified as poor. There are at least three possible choices for the poverty line: (a) the original poverty line established in 2007; (b) the reestimated poverty line using the consistent consumption aggregate; and (c) the World Bank’s US$3.20 poverty line for lower-middle-income countries. While we compute and compare the resulting poverty trend using all three lines, we ultimately adopt the US$3.20 poverty rate because it is well established and widely accepted and the trends under other alternatives are not significantly different. A final decision regards the choice of the intertemporal deflator. In Bhutan, two different deflators are used: the CPI is used to update the poverty line from 2007 to 2012, and a survey-based food price index (SBPI) is used to update it from 2012 to 2017. The exercise reveals that during 2007–12, households observed large food price increases (as indicated by the SBPI), which are not captured in the CPI. It is not clear whether food price inflation truly diverged during these years or whether other issues related to data quality contaminated household expenditure data and esti- mates of unit values. For these reasons, we use the CPI to update the poverty line throughout all survey rounds. Data sets Poverty estimates in Bhutan are produced using BLSS for 2007, 2012 and 2017. The survey was implemented in 2003 for the first time, but the poverty line was reestimated in 2007, and so a trend can only be established for 2007–17. In 2007, 2012, and 2017, respectively, 9,798, 8,968, and 11,660 households were surveyed. The ques- tionnaires include the following sections, with only relatively minor changes between survey years: demo- graphics; education; health; fertility; employment; housing; asset ownership; access and distance to services; remittances sent; priorities, credit, and opinions; sources of income; food consumption; nonfood expendi- ture; home-produced nonfood items; retrospective and mortality (2012 and 2017); and social capital (2012). Construction of the comparable consumption aggregate The consumption aggregate is typically constructed as the sum of food and nonfood expenditures. The meth- odological inconsistencies across the three years of poverty measurement are mainly due to choices made regarding the inclusion or exclusion of subcomponents of either food or nonfood consumption and the treat- ment of outliers. To the extent possible, we follow best practice in poverty measurement, which means that we are able to follow closely the measurement that was conducted in 2017. There is one notable exception: in 2017, detailed data on durables ownership became available, including information on the timing of acquisition Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 33 and value at purchase. For durables that were acquired within 12 months of the survey, the purchase value is divided by 12 and is included in the consumption aggregate. While it is possible to improve the estimate of the value of services from these goods using the newer data, such an estimation is not possible for 2007 and 2012 due to lack of relevant information. Durables purchases are therefore excluded from 2017, so that a consist- ent aggregate can be constructed using data from all three rounds. The rest of nonfood consumption remains unchanged. Food consumption is estimated as the sum of all purchased, home-produced, and in-kind consump- tion, plus the value of food consumed outside the home, and there are no changes to food consumption in 2017. We revise the consumption aggregates for 2007 and 2012 according to the method adopted for aggregating con- sumption in 2017. This revision also minimizes changes in how the consumption aggregate was built in 2017. One small change remains not fully comparable across time. In the official aggregate, purchased food con- sumption is treated for outliers at the primary sampling unit (PSU) level only in 2017. While we can adjust the unit value for outliers at the PSU level in 2012 in the same way it was done in 2017, this is not possible with the 2007 data because the PSU variable is not available. The outlier treatment is instead performed at the district level, which represents the best attempt at achieving comparability. Table A.1 summarizes the changes that were applied to the three rounds of survey data. It should be noted that the official consumption aggregate for 2007 is reconstructed based on a “best guess” of how it was likely done back then, by comparing raw data and values of consumption components that make up the official aggregate. TABLE A.1  Comparison of official and comparable consumption aggregates Official consumption aggregate Comparable consumption aggregate Consumption good 2007 2012 2017 2007 2012 2017 Food Purchased Yes Yes Yes Yes Yes Yes Home production Yes Yes Yes Yes Yes Yes In-kind Yes No Yes Yes Yes Yes Food outside of home Yes Yes Yes Yes Yes Yes Outlier treatment No No Yes Yesb Yes Yes Nonfood Clothing Yes Yes Yesa Yesa Yesa Yesa Transport Yes Yes Yes a Yes a Yes a Yesa Household operations Yes Yes Yesa Yesa Yesa Yesa Recreation Yes Yes Yes a Yes a Yes a Yesa Furnishing Yes No No No No No Miscellaneous Yes No No No No No Medicine Yes Yes Yes Yes Yes Yes Other health expenditures Yes No No No No No Education Yes Yes Yes Yes Yes Yes Energy Yes Yes Yes Yes Yes Yes Rent Yes Yes Yes Yes Yes Yes Tobacco/doma Yes Yes Yes Yes Yes Yes Durable purchase No No Yes No No No Home production Yes Yes Yes Yes Yes Yes In-kind No  Yes Yes  Yes Yes  Yes  a Winsorized. b Outlier treatment conducted at district (not PSU) level. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 34 Other minor inconsistencies are also taken care of in this process. For example, the questions regarding food consumed outside the home changed in 2017. Previously, the survey collected data on number of meals, cost per meal, and average weekly cost of meals consumed outside the home. The value of free meals did not count toward the households’ food consumption. In 2012, the structure of the questionnaire was maintained, but the value of purchased as well as free meals was accounted for in the household’s food consumption. In 2017, questions regarding food consumption outside the home were rephrased, and the values of breakfast, lunch, dinner, snacks, and beverages consumed outside the home were collected. TABLE A.2  Per capita consumption in Bhutan, Nonfood consumption is estimated as the sum of official and comparable, 2007–17 expenditures on clothing, transport, household (Nu) operations, recreation, medicine, education, energy, rent, tobacco or doma, home-produced nonfood goods, Indicator 2007 2012 2017 and the value of those items received in kind. Some Per capita consumption, official aggregate 2,314 4,603 6,758 expenditure categories are treated for outliers in 2007 Per capita consumption, comparable and 2012, the same way it is done in 2017. Further, some 1,941 3,659 6,236 aggregate categories are included or excluded for comparabil- Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2007, 2012, ity. Housing expenditure (rent) is usually taken from and 2017. the actual rent paid or from a self-reported value of Note: For 2007 and 2012, the official estimates of per capita consumption shown what the household would pay for a house with simi- in this table differ from those reported in the Poverty Analysis Report of the corresponding year (NSB, various years). This is because the estimates in this table lar characteristics. The latter is potentially subject to are all population estimates, whereas the published figures were calculated using large measurement errors in places where the hous- household weights in those two years (only). ing market is small and transactions are scarce; how- ever, for the purpose of this exercise, no further cor- rections are made for either self-reported or subsidized rent. Table A.2 shows the mean values of the official and comparable consumption aggregates for all three years. Spatial deflator Next, the consumption aggregate is deflated across districts to reflect spatial variations in prices. For this, a spatial Paasche index is constructed from the food consumption data according to the following formula: P = [∑ wh P h k pk 0   (    ph ⁄  k )], -1 (A.1) w h k is the share of household h’s budget devoted to good k, pk 0 is the national median price of good k, and phk   is the unit value paid by household h for good k. The household-level   Paasche is constructed using recorded expend- iture on all food items from all households and normalized using the national median. The spatial deflator is drawn from the regional median, separately for urban and rural areas, and then used to deflate per capita consumption. Finally, the resulting spatially adjusted consumption aggregate is normalized so that the mean of the spatially adjusted aggregate matches the mean of the unadjusted aggregate. This is done to ensure that households are not made arbitrarily richer or poorer because of spatial deflation. The next step is to compare this value against the poverty line to identify the poor. Poverty line update The official poverty line was originally estimated in 2007, using a cost of basic needs approach. The method calculates a poverty line that represents the cost of a food basket for a reference population (bottom 2nd to 4th decile of the welfare distribution) that meets a minimum nutritional standard of 2,124 kilocalories per per- son per day, plus an allowance for nonfood purchases. This approach resulted in a poverty line equivalent to Nu 1,097 in 2007, which was subsequently updated to Nu 1,705 (2012) and Nu 2,196 (2017). While the nonfood Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 35 poverty line was adjusted consistently using the nonfood CPI between 2007 and 2012 and then again between 2012 and 2017, the deflator chosen to update the food poverty line is the food CPI in 2007–12 and a survey-based food price index in 2012–17. Adjusting the poverty line in a comparable way across the three survey years involves two decisions: (a) which poverty line to use and (b) which deflator to use (CPI or SBPI). We discuss the decisions made and the rationale behind both decisions. For the poverty line, we ultimately decided to use the World Bank’s US$3.20 poverty line (in 2011 PPP). This is a well-established poverty line that is appropriate for lower-middle-income countries such as Bhutan and widely accepted for analytical purposes. The alternative is to undertake a complete reestimation of the poverty line using 2007 data and the new consumption aggregate and taking the same approach adopted back then, as described above. The nonfood allowance is estimated using a nonparametric method that calculates mean nonfood consumption for households that have food consumption amounts within a bandwidth of x percent of the food poverty line, where x = 1,2,…,10. This approach ensures that the consumption patterns of house- holds closer to the food poverty line are assigned a higher weight. The second choice involves the harmonization of the intertemporal deflator across years. Bhutan’s CPI, as in many other low- and middle-income countries, is based exclusively on data points from urban areas. Price inflation in urban areas could differ significantly from that in rural areas, which can be problematic given the concentration of the poor in rural areas. An alternative method for updating the food poverty line, which was adopted in 2017, is to calculate a survey-based food price index using information on food consumption from rural households. This SBPI can be constructed using a core food basket consisting of around 50 food items that were consumed by rural households in the bottom quintile of the consumption distribution. A democratic weighting method is used (rather than plutocratic weighting, which assigns a greater weight to households with more purchasing power). This cost of living index is approximated by the Laspeyres index as follows: P h zL = z0 p i h ⁄ i=1 i (     p i ) ∑ n w     0  . (A.2) Estimates from the food price deflator derived from the BLSS and comparisons with the official CPI based on price data in urban areas are presented in table A.3. Between 2012 and 2017, this index yields a food price infla- tion rate of 1.27, which is slightly lower than the official food CPI of 1.38. While unit values of food items from TABLE A.3  Comparison of CPI and survey-based food rural households are used, the index does not yield price index in Bhutan, 2007–17 drastically different results than those of the urban- Time period Food CPI Survey-based price index Nonfood CPI based food CPI. However, between 2007 and 2012, the 2007–12 1.68 2.66 1.35 same price index yields a value of 2.66, which is signif- icantly higher than the official food CPI of 1.68. 2012–17 1.38 1.27 1.31 Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2007, 2012, It is not fully clear what is driving this divergence. and 2017. The value of the SBPI is not driven by a particular set Note: CPI = consumer price index. of food items or a particular geographic area. A closer look at the data reveals some peculiar patterns in the TABLE A.4  Official and comparable poverty lines unit values in 2007—that is, the share of reported unit in Bhutan, 2007–17 values that are at or right around the national median (Nu) is twice as high as in 2012 and 2017. It is difficult to Indicator 2007 2012 2017 determine whether what we observe in the data is driven by potential data quality issues (for example, Poverty line, official 1,096.9 1,704.8 2,195.9 measurement errors) or by real trends—that is, for US$3.20 poverty line, deflated with CPI 1,253.5 1,831.3 2,442.9 some reason the CPI did not manage to capture high Reestimated poverty line 1,170.9 1,796.2 2,430.2 food price inflation properly. This time frame is con- Source: Poverty Analysis Report for 2017 (NSB, various year). Calculations using current with the period of the global food price crisis. Bhutan Living Standards Survey (BLSS) for 2007, 2012, and 2017. In the absence of better information, we maintain the Note: CPI = consumer price index. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 36 choice of the CPI as the preferred intertemporal deflator. The resulting poverty lines are shown in table A.4. The reestimated poverty line is slightly lower than the US$3.20 poverty line. Estimating poverty comparably Based on a newly revised consumption aggregate TABLE A.5  Official and comparable poverty trends in and the preferred US$3.20 poverty line, a compara- Bhutan, 2007–17 ble poverty trend is estimated and shown in table A.5. % of population who are poor Estimates for district-level poverty rates are shown in Table A.6. The first row of table A.5 shows official Indicator 2007 2012 2017 poverty rates, estimated with the official consump- Poverty rates, official 23.2 12.0 8.2 tion aggregate and the official poverty line that was US$3.20 Poverty rate, with comparable 36.4 17.8 12.1 estimated in 2007. The second row shows the US$3.20 consumption poverty rate applied to the comparable consumption US$3.20 Poverty rate, with official 30.5 14.5 12.0 aggregate. The third row shows the US$3.20 poverty consumption rate applied to the official consumption aggregate. The Poverty rates with reestimated 32.2 16.7 11.9 fourth row shows poverty headcount rates estimated poverty line using a comparable consumption aggregate and a pov- Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2007, 2012, erty line reestimated using the revised consumption and 2017. distribution and then deflated across years using the food and nonfood CPI. The last set of poverty estimates is not very different from the US$3.20 poverty rates. The impressive progress in poverty reduction remains the same for all calculations. Going forward, we recommend that the official poverty line be updated once the next round of the BLSS is collected, to reflect the rising living standards and changing consumption patterns of the population and to make use of better data that became available starting with the 2017 survey. Specifically, the latter relates to better measurement of the value of durables consumption using information on year of purchase, value at ac- quisition, and current value. Although this component was omitted in this exercise to maintain comparability, durables consumption is an important part of household welfare and should be included when data allow. TABLE A.6  District-level poverty rates in Bhutan, 2007–17 % of the population who are poor Official poverty US$3.20 poverty Reestimated poverty District 2007 2012 2017 2007 2012 2017 2007 2012 2017 Bumthang 10.9 3.4 2.1 29.3 4.8 4.9 26.0 4.8 4.9 Chhukha 20.3 11.2 3.5 32.2 16.1 7.4 28.6 14.4 7.2 Dagana 31.1 25.1 33.3 43.5 30.6 42.8 39.8 30.2 42.8 Gasa 4.1 — 12.6 21.9 — 20.4 17.4 — 20.4 Haa 13.2 6.4 0.9 34.9 14.6 1.8 28.3 13.2 1.8 Lhuentse 43.0 31.9 6.7 55.2 31.0 8.9 46.7 29.2 8.5 Monggar 44.4 10.5 17.1 58.2 24.8 22.8 52.8 24.0 22.5 Paro 3.9 0.0 0.3 15.3 2.1 0.9 10.7 2.1 0.6 Pema Gatshel 26.2 26.9 13.7 50.6 21.9 19.6 45.1 20.4 19.1 Punkha 15.6 10.0 2.6 31.5 17.4 3.2 27.7 16.4 3.2 Samdrup Jongkhar 38.0 21.0 6.2 52.3 26.7 10.9 48.5 25.4 10.9 Samtse 46.8 22.2 12.3 53.2 32.3 17.5 49.2 30.0 17.4 Sarpang 19.4 4.2 12.1 36.5 7.9 18.2 31.5 7.2 17.6 Thimphu 2.4 0.5 0.6 4.8 2.7 1.1 4.5 2.2 1.0 Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 37 Official poverty US$3.20 poverty Reestimated poverty District 2007 2012 2017 2007 2012 2017 2007 2012 2017 Trashigang 29.3 11.5 10.7 48.1 20.7 16.9 42.1 19.7 16.6 Trashi Yangtse 14.3 13.5 11.9 25.5 25.4 15.7 21.5 24.6 15.3 Trongsa 22.2 14.9 14.0 47.7 22.2 20.3 43.3 21.1 20.3 Tsirang 13.9 14.8 4.8 29.0 29.1 8.4 22.5 26.4 8.4 Wangdue Phodrang 15.8 10.9 5.4 34.0 14.6 8.1 26.7 13.5 8.1 Zhemgang 52.9 26.3 25.1 72.3 30.3 36.5 68.3 30.3 35.5 Total 23.2 12.0 8.2 36.4 17.8 12.1 32.2 16.7 11.9 Source: Calculations using Bhutan Living Standards Survey (BLSS) for 2007, 2012, and 2017. Note: — = estimates for Gasa in 2012 are too imprecise to present. Poverty, vulnerability and welfare in Bhutan  PROGRESS AND CHALLENGES 38 References Christensen, Garry, Turi Fileccia, and Aidan Gulliver. 2012. Bhutan Agricultural Sector Assessment. Vol. I: Issues, Institutions, and Policies. Washington, DC: World Bank; Rome: Food and Agriculture Organization, Investment Centre Division. Datt, Gaurav, and Martin Ravallion. 1992. “Growth and Redistribution Components of Changes in Poverty Measures: A Decomposition with Applications to Brazil and India in the 1980s.” Journal of Development Economics 38 (2): 275–95. Deaton, Angus, and Salman Zaidi. 2002. “Guidelines for Construction Consumption Aggregates for Welfare Analysis.” LSMS Working Paper 135, World Bank, Washington, DC. El-Saharty, Sameh, Naoko Ohno, Intissar Sarker, Federica Secci, and Somil Nagpal. 2014. “Bhutan: Maternal and Reproductive Health at a Glance.” Health, Nutrition, and Population Global Practice Knowledge Brief, World Bank, Washington, DC. https://openknowledge.worldbank.org/handle/10986/21293. Huppi, Monika, and Martin Ravallion. 1991. “The Sectoral Structure of Poverty during an Adjustment Period: Evidence for Indonesia in the Mid-1980s.” World Development 19 (12): 1653–78. Keturakis, Ed, Winston Dawes, Maria Ruth Jones, Blair Edward Lapres, Loraine Ronchi, and Massimiliano Santini. 2017. “Increasing Agribusiness Growth in Bhutan.” Policy Note, World Bank, Washington, DC. http://documents.worldbank. org/curated/en/322451506624385502/Increasing-Agribusiness-Growth-in-Bhutan. Leao, Izabela, and Tenzin Lhaden. 2018. “Promoting Better Nutrition in Bhutan.” End Poverty in South Asia (blog), May 14. http:// blogs.worldbank.org/endpovertyinsouthasia/promoting-better-nutrition-bhutan. Mehta, Meeta Punjabi, Jimba Rabgyal, and Sagar Acharya. 2015. “Commodity Chain Analysis of Large Cardamom in Bhutan.” Technical Report for Food and Agriculture Organization, Rome, November. Ministry of Health. 2016. National Nutrition Survey 2015. Thimphu: Nutrition Program, Department of Public Health, Ministry of Health. ———. 2018. “Annual Health Bulletin 2018.” Health Management and Information System Policy and Planning Division, Ministry of Health, Thimphu. http://www.health.gov.bt/wp-content/uploads/ftps/annual-health-bulletins/Annual%20Health%20 Bulletin-2018. Ministry of Labour and Human Resources. Various years. Bhutan Labour Force Survey. Thimphu: Ministry of Labour and Human Relations. NSB (National Statistics Bureau). Various years. Agricultural Statistics. Thimphu: NSB. — ——. Various years. Livestock Statistics. Thimphu: NSB. — ——. Various years. Living Standards Survey Report. Thimphu: NSB. — ——. Various years. Poverty Analysis Report. Thimphu: NSB. — ——. Various years. Statistical Yearbook. Thimphu: NSB. World Bank. 2014. “Bhutan—Poverty Assessment 2014.” World Bank, Washington, DC. — ——. 2018. “Bhutan Agriculture Policy Note.” World Bank, Washington, DC. World Bank and Ministry of Labour and Human Resources. 2016. Bhutan’s Labor Market: Toward Gainful Quality Employment for All. Washington, DC: World Bank. https://openknowledge.worldbank.org/handle/10986/25703. World Economic Forum. 2019. “Bhutan Has Achieved 100% Electricity Access. Here’s How.” Nepali Reporter, February 17. https://nepalireporter.com/local-renewable-energy-technologies/.