Development Reversed: Poverty and Labor Markets in Myanmar May 2024 Document of the World Bank Produced by the Poverty and Equity Global Practice Equitable Growth, Finance and Institutions Vice Presidency East Asia and Pacific Region. © 2024 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. 1 The World Bank does not guarantee the accuracy, completeness, or currency of the data included in this work and does not assume responsibility for any errors, omissions, or discrepancies in the information, or liability with respect to the use of or failure to use the information, methods, processes, or conclusions set forth. 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. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. 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. Attribution—Please cite the work as follows: Sinha Roy, Sutirtha and van der Weide, Roy. 2024. Development Reversed: Poverty and Labor Markets in Myanmar. World Bank. 2 Acknowledgments The report was prepared under the guidance of Mariam J. Sherman (Country Director for Cambodia, Myanmar, and LAO PDR); Lalita M. Moorty (Regional Director for East Asia and the Pacific); Rinku Murgai (Practice Manager, Poverty and Equity Global Practice); Ulrich Schmitt (Manager, Operations) and Kim Alan Edwards (Program Leader, Equitable Growth, Finance and Institutions Practice Group). The author is most grateful to Valens Mwumvaneza, Kemoh Mansaray, Thi Da Myint, Minn Thu, Mar Mar Thwin, Kyaw Soe Lynn, Han Win Htat, Moh Moh Lwin and other seminar participants for their guidance and thoughtful feedback during a presentation of the main findings of this report. The author also extends gratitude to Sebastian Eckardt (Practice Manager, Macroeconomics, Trade and Investment Global Practice) for providing valuable guidance and insights into drivers of poverty at project inception. Saurav Katwal (Consultant Economist) made central contributions in the analysis of labor market indicators presented in this report.. The author is thankful to Mildred Gonsalvez, Thida Aung and May Oo Mon for offering strong administrative support to the project. 3 Executive Summary The combined impacts of multiple negative shocks, including the Covid-19 pandemic and the February 2021 military coup, have brought poverty headcount, depth, and severity in Myanmar back to levels last seen in 2015. In 2023-24, headcount poverty in Myanmar is estimated to be 32.1 percent at the national poverty line 1. There are approximately 7 million more poor people in 2023 than at the start of the COVID-19 pandemic. Most of Myanmar’s poorest households are concentrated in rural areas, with internally displaced persons (IDPs) and people with disabilities experiencing amongst the highest poverty rates of around 50 percent in 2023. Poverty estimation exercises based on unofficial household survey data involve statistical uncertainties. However, trends in household well-being indicators studied in this report robustly point to a rise in poverty between 2017 and 2023. The last official consumption survey in Myanmar was in 2017. The World Bank’s Myanmar Subnational Phone Surveys (MSPS) have filled the gap in official survey statistics in more recent years by collecting sub-nationally representative data from over 90 percent of Myanmar’s townships. However, unlike official consumption surveys, MSPS cannot collect detailed itemized consumption data of households. Estimating household consumption, and by extension poverty, using such surveys, involves certain statistical uncertainties. Despite these statistical uncertainties, consumption trends from MSPS studied in this report robustly point to a rise in poverty. Poverty is estimated at 32.1 percent in 2023, up around 7 percentage points from 2017. The report shows that had Myanmar sustained its pre-pandemic growth trends until 2023, there would have been 8 million fewer poor people in 2023 and the poverty rate would be just 11 percent. The share of the population that is economically secure (that is, those with average consumption more than 1.5 times the minimum needs basket) has fallen from 45.3 to 35.7 percent between 2017 and 2023. Simultaneously, poverty is becoming increasingly entrenched, with the depth and severity of poverty rising significantly during the two periods. These trends underscore significant reversals from Myanmar's experience of sustained poverty reduction until 2020, which was driven by fast economic growth and an expansion of its manufacturing and services sectors. Headcount poverty in Myanmar had decreased from 42 percent in 2005 to 32.1 percent in 2015 and further to 25 percent in 2017. With an annual per capita growth rate of 7.3 percent between 2011 and 2017, Myanmar's growth used to rank among the highest in East Asia and the Pacific region. Increased manufacturing and services sector activity, particularly in the urban centers of Yangon and Mandalay, created new employment opportunities in many neighboring states and regions, leading to declines in extreme poverty. But by 2023, many of these development gains had reversed. While poverty has risen compared to the last six years and remains highest in rural areas, it has surged particularly rapidly in urban areas and among individuals with higher human capital endowments. Between 2017 and 2023-24, urban poverty increased by about 12 percentage points, compared to a 7.3 percentage point rise in villages. Additionally, poverty is rising among workers with a college education and extensive work experience. This trend poses significant 1 Estimates are based on survey data collected between December 2023 to March 2024. 4 threats to Myanmar’s medium-to-long-term growth prospects, as its most productive human capital struggles to meet their basic needs. The decline in living standards among urban households – especially those that belong to lower- and middle-income families – has likely contributed to a reduction in inequality. Between 2017 and 2023, average consumption levels among the richest urban households in Myanmar decreased by about 4 percent annually, while the average annual decrease for rural households was about 2 percent. Consequently, the Gini coefficient of inequality dropped from 30.0 in 2017 to an estimated 27.0 in 2023. This trend aligns with other global studies suggesting that wars, conflicts, and disasters are often associated with reduced economic disparities (Schneidel, 2018). There are notable differences in 2023 poverty levels across states and regions of Myanmar. As in 2017, Chin continues to be the poorest among all other states and regions. The rise in poverty has been particularly pronounced in conflict-affected areas including Sagaing, Kayah and Kayin. Many other states and regions have seen a significant increase in poverty, including Yangon, where poverty has risen by 4.7 percentage points between 2017 and 2023, and Mandalay, where poverty rose by over 20 percentage points. On the other hand, Rakhine, Magway, and Nay Pyi Taw exhibited a minor reduction in poverty between 2017 and 2023, while Ayeyarwady is the only region to experience substantial reductions in poverty during this period. The rise in poverty during 2023 is associated with unfavorable trends in labor market indicators. In contrast to the rapid economic expansion witnessed in the pre-Covid years, in 2023, Myanmar's labor market conditions—in terms of participation, employment, and formality—were considerably weaker than in 2017 and recovered only marginally from 2022. The employment rate has declined by 7.4 percentage points since 2017, with only a 2.3 percentage points rise between 2022 and 2023. In rural areas, household wellbeing levels declined as employment levels have fallen (from 65 percent in 2017 to 55 percent in 2022 and followed by a recovery of only 2.7 percentage points over the past year) while agricultural input price inflation has surged. Yet, rural poverty would have been higher had agriculture employment share in rural areas declined further. The sector has helped cushion against negative shocks: households with at least one member engaged in agriculture were much less likely to report income losses than households with no exposure to agricultural activities. In urban areas, the rise in poverty between 2017 and 2023 correlates with a decline in the quality of available jobs for Myanmar's most educated workforce. Wage employment has become more restricted among urban populations, falling by 8.7 percentage points between 2017-2022 and recovering less than 1 percentage point in 2022-23. Simultaneously, wage workers with formal contracts and pension provisions have fallen by 3.6 percentage points in the past year, indicating rising informality of work. These changes are associated with elevated urban poverty rates. Poverty estimates in this report are lower than other recently available estimates for Myanmar, largely due to methodological differences and the statistical uncertainties associated with such exercises. The methodology adopted in this report directly estimates household consumption (and poverty by extension) in 2023 by modeling expenditures against household-level indicators that are known correlates of consumption or ideally suited to capture changes in well-being conditions during recent years. This is a different approach compared to studies that simulate poverty based on macro-level indicators (e.g., sectoral growth, wages, employment, etc., many of which are 5 themselves estimates due to the unavailability or unreliability of official statistics). Such simulations implicitly rely on assumptions about the evolution of inequality across households (as inequality is not observed in macro-level data), with the resulting poverty estimates sensitive to these underlying assumptions. The present study makes no assumptions about the evolution of household inequality between 2017 and 2023 and relies on the observed survey data to construct the final estimates. Other studies include income-based measures of poverty for recent years. Since households tend to smooth consumption over time, income-based poverty estimates tend to detect sharper changes in poverty over time. For instance, incomes are generally higher during the harvest season – meaning that surveys conducted at that time would return lower poverty rates. Consumption, on the other hand, is usually prone to fewer changes over time as households tend to smooth consumption across periods. Moreover, poverty estimates based on official household surveys, such as the Myanmar Living Conditions Survey of 2017, adopted consumption-based poverty measurement. This means that income-based poverty measures cannot be compared directly with poverty from pre-Covid years. Despite differences in headline estimates, recent studies (e.g., UNDP 2024, IFPRI 2024, and this study) all highlight several themes regarding poverty in Myanmar. Firstly, all analyses show significant reversal in the progress made against poverty in Myanmar. Poverty is becoming more entrenched, with increasing depth and severity. Secondly, although poverty remains most prevalent in rural areas, urban poverty is experiencing a rapid escalation, swiftly approaching levels seen in rural regions. Third, economic stability in Myanmar is deteriorating rapidly. The proportion of households with consumption levels several times higher than the minimum needs basket in 2017 is now diminishing—a clear indication of a widespread economic insecurity. 6 Introduction A series of negative shocks since the Covid-19 pandemic and the February 2021 military coup have forced vulnerable households in Myanmar back into poverty. However, with no official household survey data since 2017, it has been challenging to determine poverty levels in recent years. Households have been impacted by a severe third wave of COVID-19 (mid-2021), substantial macroeconomic volatility (rapid exchange rate depreciation and price inflation, persistent foreign currency shortages), shortages of electricity, food, fuel, and other essential items such as medicines, and a significant rise in conflict in parts of the country which has led to widespread internal displacement. The compounding effects of these shocks would indicate a rise in poverty. However, due to the absence of official household survey data containing detailed household consumption or income data, measuring the effect of these shocks on poverty has been challenging. The World Bank’s Myanmar Subnational Phone Surveys (MSPS) attempt to fill the gap in survey data but poverty estimation using this data can carry certain statistical uncertainties. State and regional-level indicators from MSPS are based on data collected from over 300, out of 330 townships in Myanmar. With a sample size of over 8,500 households selected proportionately based on their representation in the population, MSPS provides comprehensive spatial and socioeconomic coverage across the country 2. MSPS also ensures that demographic groups are proportionately reflected in the survey. However, unlike official surveys, MSPS cannot capture detailed itemized household consumption data over the phone. In such situations, statistical methods are adopted to measure poverty which carry certain uncertainties. However, despite these uncertainties, trends in household well-being indicators from MSPS robustly point to a rise in poverty. The findings in this report robustly point to a significant escalation in poverty: Myanmar's poverty rate is estimated to have reached 32.1 percent in 2023-24. The report shows that had Myanmar sustained its pre-Covid growth trend until 2023, there would have been 8 million fewer poor people and 1/3rd of the headcount rate than what its 2023 levels. Simultaneously, poverty is getting increasingly entrenched with a significant rise in the depth and severity of poverty, meaning that poorer households are drifting further away from the basic minimum needs basket. About one- third of Myanmar's population has average consumption that is quite close to their minimum needs. In the event of an adverse economic shock, the risk of backsliding into poverty for these households is high. Fourth, while poverty is still higher in villages, urban poverty has risen more sharply. Fifth, internally displaced populations (IDPs) and individuals with disabilities exhibit the highest poverty rates in Myanmar, with poverty rates about 50 percent in 2023-24. Finally, there is a notable disparity in poverty trends across states and regions: poverty has markedly increased in conflict-affected areas such as Kachin, Kayah, Kayin, and Sagaing between 2017 and 2023-24. In Yangon, poverty has risen by 4.7 percentage points between 2017 and 2023. Changes in poverty in Nay Pyi Daw and Magway are relatively subdued in comparison, and, defying national trends, poverty has decreased in Ayeyarwady. 2 Due to the ongoing conflict events and to minimize risk to survey enumerators, MSPS continues to be a phone-based survey. However, the MSPS takes several steps to minimize biases arising from phone survey methodologies. A detailed examination of these methods and a review of MSPS’ sub-national properties can be accessed in Sinha Roy (2024). 7 These findings signify a considerable reversal in development gains made since the liberalization of Myanmar’s economy. Myanmar's GDP growth at an average rate of 7.3 percent annually between 2011 to 2017 had positioned it among the top five fastest-growing economies globally at the time. This economic growth resulted in rapid improvements in living standards across the nation. Poverty dropped from 42 percent in 2005 to 32.1 percent in 2015 and to 25 percent in 2017. Moreover, the poorest households began to narrow the gap between consumption and their minimum needs during this time. In 2005 and 2010, the average consumption of poor households was 14.2 and 12.2 percent less than the poverty line. From 2015 to 2017, this figure progressively dropped from 8.4 to 5.2 percent. According to the World Bank (2022a), the increase in consumption among Myanmar’s poorest households between 2014 and 2019 ranked it among the top three countries worldwide. The latest poverty estimates based on MSPS show that poverty headcount, depth, and severity in 2023-24 have reversed to levels last seen in 2015. As a result, Myanmar has an estimated 7 million additional poor people in 2023 than at the start of the pandemic. The rise in poverty in 2023 reflects challenging labor market conditions. The labor force participation rate (LFPR) has risen by 3.5 percentage points between 2022 and 23 but has yet to recover levels observed in 2015 or 2017 (Table 1). Female labor participation rates (FLFPR) have reversed to levels last observed in 2015 – but the share of women in the labor force who have managed to find work has marginally fallen. The employment rate has dropped by 7.4 percentage points since 2015 and recovered only by 2.3 percentage points between 2022 and 2023. In rural areas, household wellbeing levels declined as employment levels have fallen from 65 percent in 2017 to 55 percent in 2022 and followed by a recovery of only 2.7 percentage points in the past year. Yet, rural poverty would have been higher had agriculture employment share in rural areas declined further. Households with at least one member in the agriculture sector are less likely to report income losses – highlighting the shock-absorbing role of the sector in Myanmar. However, the report also finds that the share of the poorest rural households working in agriculture has fallen between 2017 and 2023-24. Surging agricultural input prices could have contributed to these patterns (in addition to other factors such as conflict, churning in the labor market, transportation and logistics challenges, etc.). In urban areas, poverty has risen between 2017 and 2023, with a substantial decline in consumption among urban lower and middle-income households. Wage employment has become more restricted among urban populations, falling by 8.7 between 2017-2022 and recovering less than 1 percentage point in 2022-23. Simultaneously, wage workers with formal contracts and pension provisions have fallen by 3.6 percentage points in the past year. These trends correlate with significant backsliding in the quality of jobs available to Myanmar’s more educated workers. The deterioration in job quality, especially among college-educated workers, is likely contributing to the rise in urban poverty. The remaining report proceeds as follows: Chapter 1 provides the headline estimates of Myanmar's poverty by location and household characteristics. Chapter 2 uses a labor market lens to understand the emerging trends in poverty. Technical details related to the updated methodology can be found in the Appendix 1. Appendix 2 provides key estimates of labor market indicators from MSPS at the subnational level. 8 Chapter 1: Poverty and Vulnerability in Myanmar during 2023-24 POVERTY TERMINOLOGIES Poverty measures • Poverty headcount: The share of Myanmar population in 2023-24 that is poor, with per adult equivalent consumption below the minimum needs basket, also known as the poverty line. In 2017, the poverty line for the whole of Myanmar was set to 1590 kyats per day in 2017 quarter 1 kyats (approximately 2762 kyats in 2023 nominal terms). Reflecting significant differences in relative prices across states and regions of Myanmar in 2023-24, the poverty line varies at the subnational level. • Poverty gap (depth): The average amount that per adult equivalent consumption falls below the poverty line, expressed as a percentage of the poverty line. The poverty gap captures the depth of poverty by estimating the average distance that the poor live below the poverty line, expressed as a percent of the poverty line. • Squared poverty gap (severity): the squared value of the poverty gap, which gives greater weight to individuals who fall further below the poverty line. Poverty in Myanmar increased from 24.8 percent in 2017 to an estimated 32.1 percent in 2023- 24, nearing levels seen in 2015. Figure 1a shows poverty headcount from official surveys conducted in 2015 and 2017, the 2023 poverty rate from MSPS, and macroeconomic projections of poverty for intervening years (see Box 1 for a technical description of macroeconomic poverty projections). The figure highlights a substantial setback in progress against estimated poverty between 2015 and 2019, the year preceding the pandemic. During this period, poverty declined by approximately 12 percentage points, driven by a 24 percent increase in real GDP per capita, with manufacturing and service sector growth boosting employment. About 6 million individuals emerged from poverty during these years. Projections based on estimated GDP per capita growth indicate that despite the onset of the pandemic's first wave in 2020, poverty further decreased by 2.4 percentage points between 2019 and 2020, lifting an additional 1.2 million people out of poverty. However, these gains were swiftly reversed in the subsequent three years after 2020. The cumulative impact of multiple shocks during this period is estimated to drive real GDP per capita in 2023 to 19 percent below 2020 levels – causing poverty to nearly double in the three years between 2020 and 2023 3. Consequently, Myanmar is now estimated to have 7 million more people living in poverty compared to pre-pandemic levels, with 1 million more individuals experiencing poverty compared to 2015. Alternatively, if Myanmar had continued to embark on its pre-covid growth trends, the estimated poverty rate in 2023 would have been 11 percent. This 3 The doubling of poverty between 2020 and 2023 is based on poverty projections for 2020 using GDP per capita growth estimates and for 2023 based on MSPS. We rely on GDP per capita based poverty projections in 2020 because there is no MSPS survey available for the year. Figure 1a shows a 5.6 percentage point headcount difference in 2023 between MSPS and GDP per capita growth based projection. This gap is likely due to the differences in inequality assumptions underlying the two approaches. See Box 1 for additional details. 9 means that, compared expected poverty rates based on its historical growth trend, Myanmar has 3 times higher poverty rate in 2023 and about 8 million additional poorer people. Figure 1a: Poverty Rate (%) Figure 1b: Number of poor people (in millions) Notes: "Official surveys" refer to poverty rates from official household surveys conducted in 2015 and 2017. “Using Macroeconomic Data” show poverty rates for years between 2017 and 2023 – see Box 1 for additional details. The “Using MSPS” shows poverty rates for 2023 relying on MSPS – se Appendix 1 for additional details. Real GDP per capita growth and population data are available from the Macro Poverty Outlook datasheets, accessible here: https://www.worldbank.org/en/publication/macro-poverty-outlook/mpo_eap Box 1: Projecting poverty during years between MLCS-2017 and MSPS-2023 using GDP per capita growth estimates. Poverty rates for years between 2017 and 2023, as shown in the "Using Macroeconomic Data" series in Figures 1 and 2, are estimated using the World Bank's global poverty imputation method. The method is used to estimate poverty during years that are in between official rounds. The method involves multiplying the real GDP per capita growth recorded in national account statistics by a pass- through rate, which represents the portion of aggregate GDP per capita growth likely to be recorded in household consumption surveys. The methodology assumes a constant distribution of consumption (i.e., inequality) since the last official survey and applies the product of GDP per capita growth and the pass-through rate to update poverty estimates in years when no official survey data is available. For more detailed information, please refer to Lakner, et al (2022). In Figures 1 and 2, the pass-through rate is set at 0.7, the standard value used by the World Bank for its annual poverty rate calculations as part of its global poverty monitoring efforts. The GDP growth projections and pass-through rates are applied to household consumption, as observed in MLCS, 2017, the last official household survey in Myanmar. Unlike estimates derived using the World Bank’s pass-through method, poverty estimates based for 2023 are based on MSPS. They do not rely on the assumption of inequality remaining unchanged since 2017. Instead, they use spatially deflated poverty lines in 2023, reflecting changes in relative price differences across states and regions due to ongoing conflict conditions. This approach relies on actual consumption data and not on estimates of GDP per capita growth and how it is transferred to households. Additional details related to the MSPS poverty estimation exercise can be found in the appendix. Note that estimates of poverty estimates based on macroeconomic data are lower than the MSPS-based estimate for 2023-24. A likely reason for this underestimation is that MSPS, despite being a telephonic survey, can capture changes in welfare at the bottom of the distribution due to the careful sample design and reweighting algorithm used to preserve its sub-nationally representative properties. By assuming that inequality has been unchanged since 2017, simulations based on macroeconomic statistics are not able to capture rising disparities among the poorest households in Myanmar. 10 Estimates of poverty depth and severity reinforce the finding that the hard-won gains of 2015 and 2020 were eliminated within three years since 2020. Poverty depth and severity are estimated to be 8 and 3 percent, respectively, in 2023-24. The levels are almost the same as in 2015 (Figures 2a and 2b). Compared to 2020, poverty depth and severity in 2023-24 are almost 2.5x to 3x levels higher. These indicators suggest that poverty has become more entrenched in Myanmar, with the average consumption of the poorest households now about as far away from their minimum needs basket today as they were in 2015. Figure 2a: Poverty Depth Figure 2b: Poverty Severity Notes:. "Official surveys" refer to poverty rates from official household surveys conducted in 2015 and 2017. “Using Macroeconomic Data” show poverty rates for years between 2017 and 2023 – see Box 1 for additional details. The “Using MSPS” shows poverty rates for 2023 relying on MSPS – se Appendix 1 for additional details. Real GDP per capita growth and population data are available from the Macro Poverty Outlook datasheets, accessible here: https://www.worldbank.org/en/publication/macro-poverty-outlook/mpo_eap Poverty remains concentrated in rural areas, yet urban poverty has risen much faster than rural poverty since 2017. Rural poverty has increased by 5.5 percentage points in 2023-24 from 30.2 percent in 2017 (Figure 3a). Urban poverty in 2023-24 was 12.6 percentage points lower than rural. However, poverty among urban households has grown at a faster rate than those living in villages. Figure 3a shows that urban poverty rose by 11.8 percentage points and 7.3 percentage points in rural areas between 2017 and 2023. At the same time, urban poverty is becoming increasingly entrenched. Whereas rural poverty severity has risen 1.6 times between 2017 and 2023-24, in urban areas, severity has risen three-fold (Figure 3c). Estimates of poverty depth (Figure 3b) confirm that consumption levels of the urban poor have drifted further away from their minimum needs basket during the past six years. Figure 3a: Poverty Headcount by Location Figure 3b: Depth of Poverty by Location Figure 3c: Severity of Poverty by Location 11 Technical notes: For each household in the MSPS survey, we impute 100 consumption values using Stata’s Multiple Imputation (MI) command. Each repetition reflects 100 draws of model parameters from their posterior distribution and a draw from a random normal error term. Each consumption value, corresponding to each repetition, yields 100 estimates of poverty, poverty gap, and squared poverty gap in the MSPS. Individual sampling weights are used in the calculations. The figures reported in the above figure are based on a simple average across 100 repetitions. Spatially deflated poverty lines for 2023, reflecting changes in relative price differences across states and regions due to ongoing conflict conditions, have been used in the above calculations. See the appendix for additional details. Households that are currently not poor are at a high risk of falling into poverty in the future. Approximately 32.3 percent of individuals are currently not poor but have consumption levels only 1.5 times the poverty line in 2023-24 and are therefore vulnerable to falling into poverty. Furthermore, the share of the population secured against the risk of sliding into poverty (with consumption more than 1.5 times the poverty line) has risen by 6.6 percentage points compared to 2017. Putting all of these indicators together confirms the finding that welfare levels in Myanmar in 2023-24 have reversed to patterns observed earlier in 2015. Figure 4: The threat of sliding into poverty looms large among a third of the population Technical notes: The group “poor” indicates the share of population below the poverty line; the series “not poor, insecure” signifies households with consumption above the poverty line but below 1.5 times the value of poverty line. Finally, the series “not poor, secure” denotes household consumption above this upper threshold. Spatially deflated poverty lines for 2023, reflecting changes in relative price differences across states and regions due to ongoing conflict conditions, have been used in the above calculations. See appendix for additional details. 12 A PROFILE OF POVERTY IN MYANMAR The increased entrenchment of poverty and the looming threat of backsliding among non-poor households reflects household consumption declining across all income groups. Consumption declines are more pronounced across urban households. Real consumption in 2023-24 is lower than in 2017 at types of households (Figure 4). But urban households on average have experienced larger declines in consumption than their rural counterparts. Within urban, middle- income households experienced less than 3 percent loss in per capita consumption annually; in comparison, per capita consumption declines for richest and poorest urban households exceeds 4 percent per year since 2017. Consequently, consumption inequality in Myanmar between 2017 and 2023 has likely moderated, not due to an improvement in the welfare conditions of the poor but rather due to substantial setbacks among wealthier segments of the population. Indeed, the Gini coefficient of inequality is found to reduce from 30.0 in 2017 (26.34 in rural and 31.82 in urban) to 27.0 in 2023 (24. 62 in rural and 29.48 in urban). This observation aligns with findings from global studies suggesting that events such as wars, conflicts, and disasters often lead to narrowing economic disparities (Schneidel, 2018). Figure 4: Annualized changes in consumption over the distribution Notes: Consumption growth is in annualized terms. Ventiles are half of a decile or 5th percentile of the consumption distribution. Individual-level sampling weights are used in the above figure. Rising poverty trends among those with college education, those employed in tertiary sectors, and households with more assets underscore the finding that living conditions are backsliding even among relatively wealthier households. In 2017, households with greater endowments—such as better education, more assets, smaller household sizes, and service sector jobs—exhibited lower poverty levels. However, figure 5 shows that consumption has substantially declined among these groups, leading to a rise in poverty. For example, individuals with college or higher levels of 13 education had a poverty rate of 4 percent. By 2023-24, the poverty rates among this group have more than quadrupled to about 18 percent (panel a). Similarly, workers in the service sector had lower poverty rates than those in manufacturing and agriculture in 2017. Yet, poverty is observed to have escalated by 11 percentage points among this group (not shown in the figure). Poverty headcount is significantly higher among households that have moved to a new village or tract in less than 1 year prior to the survey (panel b). This sample of households includes households who have migrated for economic or other reasons, as well as those that have been forcefully displaced. Poverty rates are also high among households who moved during 2020-21 – the starting year of the military coup. In comparison, households who have remained at their existing ward or village for more than 10 years have lower poverty headcount rates. Trends among households owning assets such as refrigerators, TVs, electrical grid connectivity, and automobiles confirm the finding that well-being contractions are substantial among households with more endowments (panel d). While households owning these assets had lower poverty rates in 2023, poverty is escalating faster among these groups. Lastly, poorer households tend to have larger household sizes as assets and incomes must be shared among more family members. This trend has persisted in Myanmar, with households with more members generally exhibiting the highest poverty levels in 2017 and 2023. In particular, since 2017, poverty rates among households with 3 to 5 members (the country's average household size is 4.6, according to the 2019 intercensal survey) have doubled from 14 percent to 29 percent in 2023. Figure 5: Patterns of consumption and poverty by individual and household characteristics Panel a: By Education Panel b: By year since moving to village/ward 50% 5000 45% 4500 42% 40% 4000 37% 35% 3500 Poverty Headcount 32% 30% 3000 Poverty in 2023-24 30% 25% 2500 24% 20% 2000 15% 1500 10% 1000 5% 500 0% 0 Never Less than Upto Upto middle Upto high College and attended primary primary school school above school During 2023-24 During 2022-23 During 2021-22 During 2020-21 More than 10 Consumption 2017 Consumption 2023 years in the same village/tract Poverty 2017 Poverty 2023 Panel c: Household Size Panel d: Asset Ownership 45% 4500 40% Adult equivalent consumption per capita 40% 4000 33% 32% 35% 3500 31% Poverty Headcount 28% 29% 28% 27% 26% Poverty Headcount 30% 3000 24% 25% 2500 20% 20% 20% 2000 14% 11% 9% 15% 1500 3% 10% 1000 5% 500 No Yes No Yes No Yes No Yes 0% 0 1 to 2 members 3 to 5 members 5 to 9 members over 9 members Electrical Grid Refrigerator TV Own Vehicles Connectivity Consumption 2017 Consumption 2023 Poverty 2017 Poverty 2023 Poverty 2017 Poverty 2023 14 Notes: Consumption values are in real terms. Poverty in 2017 is calculated using the 1590 Kyats per capita poverty line; 2023 estimates of poverty use state, rural, and urban-specific poverty lines based on relative price adjustments – more details in the appendix. Education and employment indicators are reported at the individual level, while household size, asset, and identification indicators are calculated at the household level. Appropriate sampling weights were used to produce these estimates. Internally displaced populations (IDPs) and individuals with disabilities rank among the poorest groups in Myanmar. IDP and disability statuses are strongly correlated with elevated poverty levels. In 2023, IDP households' mean per capita consumption value stood at 2066 kyats (in constant 2017 terms), approximately 11 percent lower than non-IDP households. Consequently, poverty rates were 48 percent for IDP households. Note that IDPs predominantly reside in rural areas (Sinha Roy, 2024). Yet, their average poverty rate exceeds the mean rural poverty rate of 35.7 percent in 2023. Similarly, the average consumption of disabled populations has declined by 21 percent between 2017 and 2023, while the poverty headcount rate is 52 percent compared to 32 percent among non-disabled groups. POVERTY AT THE STATE AND REGION LEVEL Although poverty in 2023-24 is estimated to have increased in nearly all areas, the geographic concentrations of poverty have undergone substantial changes. In 2017, Chin had the highest proportion of its population living below the poverty line. In 2023, the poverty headcount rate in Chin has not changed much (Figure 6a), solidifying its status as the poorest state/region in the country. In contrast, poverty has risen significantly across most other states and regions. Kayah, Kayin, and Sagaing have witnessed a notable rise in poverty between 2017 and 2023-24 in rural and urban locations. These areas are now ranked among the second to fourth highest poverty rates. Back in 2017, Rakhine, Magway, and Ayeyawady used to hold these rankings. Moreover, in Tanintharyi, Mandalay, Mon, Bago, and Yangon, poverty headcount rates used to be below 20 percent in 2017. All of these areas have experienced rising poverty in 2023. For example, poverty in urban areas of Mandalay increased by 5½ times in 2023 compared to 2017, while rural areas of Tanintharyi saw poverty levels rise by 3¼ times in 2023 (Figure 6c). Finally, Rakhine, Magway, and Nay Pyi Taw have experienced a slight reduction in the poverty rate, primarily in rural areas. In terms of total state/region’s population that is poor, Sagaing, Mandalay, Shan and Yangon rank among the highest (Figure 6b). In particular, Mandalay has 1.5 million more poorer people in 2023 than in 2017 – the largest rise in total poor population compared to all other states and regions. Poverty trends in Ayeyawady indicate a unique pattern across all states and regions of Myanmar: poverty has reduced from 31.7 percent in 2017 to 17.4 percent in 2023-24. Further, disaggregation by rural and urban areas reveals that a substantial reduction in rural poverty has driven the decline in aggregate poverty at the regional level. Figure 6a: Changes in Poverty Headcount Figure 6b: Changes in poor population (in millions) 15 Figure 6c: Poverty rate by rural and urban of states and regions Notes: Consumption was in real terms and its growth has been annualized. Individual level sampling weights used in the above figure. Chapter 2: Applying a Labor Market Lens to Observed Poverty Trends The rise in poverty between 2017 and 2023 is associated with unfavorable trends in labor market indicators. In contrast to the rapid economic expansion witnessed in the pre-Covid years, in 2023, Myanmar's labor market conditions—in terms of participation, employment, and formality—were considerably weaker than in 2017 and recovered only marginally from 2022. The labor force participation rate has yet to recover to levels observed in 2015 or 2017 (Table 1). Female labor participation rates (FLFPR) have reversed to levels last observed in 2015. The employment rate has dropped by 7.4 percentage points since 2015. 16 Between 2022 and 2023, the LFPR has risen by 3.6 percentage points, but the employment rate has grown by 2.3 percentage points. Therefore, much of the recent rise in LFPR is due to adults trying to seek employment opportunities but facing difficulties in obtaining one. After declining for several years, women's participation in the labor force experienced a modest rebound between 2022 and 2023. However, finding employment opportunities in a weak and sluggish economic environment poses a challenge. Between 2017 and 2023, approximately 4.5 million working-age adults joined Myanmar's labor market (Figure 7). Of these, around 1.9 million are currently active in the workforce. The share of females in the labor force has risen by 3 percentage points between 2022 and 2023 – but the share of women in the labor force who have managed to find work has marginally fallen. Female labor force participation levels in 2023 are back to levels seen in 2015 (Table 1). Over 11 percent of women in the labor force in 2023 were unemployed (totaling approximately 1.3 million adults) – rising from 3 percent in 2017 and 10 percent in 2022. Similarly, male unemployment share in the labor is also rising – from 2 percent in 2017 to 5 percent in 2022 and about 7 percent in 2023. With more women looking for jobs, there are 1.3 million fewer women categorized as those not in employment, education, or training (NEET) in 2023 compared to 2022. Nonetheless, women contribute a disproportionate 70 percent of the NEET category of adults in Myanmar. Moderate improvements in labor force participation and employment rates over the past year mask substantial deterioration in job quality. Table 1 also shows that the share of agricultural employment, after falling 6.6 percentage points between 2017 and 2022, has recovered by 3.3 percentage points over the past year. Moreover, shares of wage earners in overall employment (a proxy for better quality of employment) have progressively fallen by 9.3 percentage points between 2017 and 2022 and further by 2.4 percentage points within just the last year (Table 1). In 2022, wage workers that possessed written employment contracts were found to be better protected from job losses (Sinha Roy et al., 2023) – leading to an 11-percentage point rise in share of wage workers with written contracts between 2017 and 2022. However, by 2023, there has been a 5-percentage points reduction in this share, meaning that that past protections offered to formal contract workers may be weakening. Table 1: Summary of key‐labor market indicators (as a share of 15+ population) Share of Share of wage Share of formal LFP rate Employment employment to agricultural Year LFP rate employment (Females) rate total employment among wage employment (15+) employees 2015 64.7 51.6 64.2 35.3 2016 36.0 2017 64.2 53.4 62.6 38.2 13.0 49.5 2018 62.4 49.3 61.9 34.4 2019 60.5 47.1 60.3 35.1 2020 60.2 45.6 59.3 2021 2022 58.6 45.9 54.5 28.9 24.2 42.9 2023 62.2 51.9 56.8 26.3 20.6 46.1 WDI, MLCS, WDI, MLCS, WDI, MLCS, WDI, MLCS, Sources modelled ILO estimates. MLCS, MSPS MLCS, MSPS MSPS MSPS MSPS MSPS 17 Figure 7: Snapshots of Myanmar’s labor market in 2017, 2022 and 2023 Notes: Population figures are in million. Estimates for 2017 and 2022 are from MLCS and MSPS, respectively. The sample includes individuals aged 15 or over. Workers with more education are better protected from job losses but are experiencing deteriorations in job quality. Reflecting the inverted U-shaped pattern of consumption declines observed in Figure 4, employment shares have declined less (or risen even) for individuals with middle/high school education. In comparison, those with the least or highest educational attainment have experienced a fall in employment as a share of their population. Despite experiencing declines since 2017, 70 percent of college educated workers are in the labor force – higher than all other groups. This means that more educated workers have been protected from job losses. A higher rate of formal employment (proxied by written contracts or pension provisions) served as guardrails against job losses for college educated workers. However, such employment opportunities for college educated workers have declined by 7.7 percentage points between 2017 and 2023 – indications that they may now be exposed to job losses. 18 In comparison to other groups, high school educated workers are more likely to experience a rise in employment (the employment share has risen 7.3 percentage points between 2017 and 2023) with most of this increase coming from rise casual work (a 4.9 percentage rise over the same period) and only a small rise in formal jobs. For all other types of workers, job losses between 2017-2023 have been pervasive combined with rising informality of employment. Table 2: Employment and job quality by education Employment-population ratio Casual work Formal job Education 2017 2023 Diff 2017 2023 Diff 2017 2023 Diff Upto primary 63.4 53.1 -10.3 14.7 19.8 5.1 0.5 0.8 0.3 Upto middle 66.6 61.7 -4.9 16.4 21.9 5.5 1.5 1.4 -0.1 Upto high 50.5 57.8 7.3 11.5 16.4 4.9 3.6 3.8 0.2 College and above 73.7 69.8 -3.9 10.0 11.7 1.7 26.6 18.9 -7.7 Notes: A formal job indicates a position involving a written contract or with pension provisions. Includes populations over 15 years of age. The Consumption Decline Among Richer Urban Households Is Likely Due to Rising Unemployment among Undereducated Groups and Growing Informality of Employment Among Educated Workers The premium associated with higher education and more years of work experience are starting to diminish in Myanmar. Older and more educated workers enjoyed higher consumption in 2017 than those with fewer experience and without a college degree (Table 8a). However, in 2023, the consumption premium experienced by these workers has diminished, and the gap in household well-being levels has converged. The economic incentives for pursuing higher education and more years of work experience are therefore diminishing in Myanmar. This observation is consistent with those not in employment, education or training, rising from 11.2 million in 2017 to 14 million in 2023 (Figure 7). The decline in consumption among college-educated workers is also correlated with a significant drop-in wage work and formal employment, as shown in Figure 8a. The formal employment rate has declined most acutely for those with a college degree and more years of experience (as proxied by age). This rising informality among Myanmar’s most educated, experienced and therefore, most productive workforce – also likely amongst its richer populations—could have contributed to consumption losses among well-off urban households, as seen in Figure 4. The left panel of Figure 8a underscores significant adverse impact on employment rate of undereducated workers. These workers are less likely to hold formal employment to begin with (see Table 2), and therefore, most of the reduction in consumption across ages within the undereducated is associated with loss of employment. Figure 8b confirms the finding that college educated workers are earning less than they used to and compared to middle and high school educated, have experienced bigger losses between 2017 19 and 2023 4. The drop in median rural wages with college education exceeds that of urban areas. Overall, median wages are observed to have fallen by about 8 percent between the two years across the sample. Figure 8a: Formal jobs work and consumption among workers with college education Notes: “Ch. Wage employment” indicates changes in the share of adults with wage employment in the population between 2017 and 2023. “Ch formal jobs” indicate changes over the same period in share of adults in the population that had either a written job contract or had pension provisions in their current employment. “Ch Employment Rate” indicates changes in employment-population ratio at each age group and education level. Consumption values in 2017 and 2023 are on the right vertical axis and are in real Kyats and adult equivalent terms. Figure 8b: Changes in median wages by education and location Notes: Wage information is available only for salaried workers 4 wage data is available only for salaried workers. Figure 8b suppresses wages of up to primary school educated workers because of the very low shares of such workers engaged in salaried jobs. 20 While Subsistence Agriculture Has Absorbed Negative Shocks, Myanmar’s Poorest Households Appear to be Leaving Agricultural Employment, Even as More Educated People Are Entering The Sector Agricultural activities have moderated in rural areas and increased marginally in urban locations. Agricultural employment in rural areas has fallen by 10 percentage points 5 and risen 3 percentage points in cities (table 3a). Consequently, services and manufacturing employment have contracted by 6 and 2 percentage points in urban areas and changed minimally in villages. Signifying structural shifts, Myanmar’s most educated workers are now turning to farming activities, while those with less education are observed to be exiting agriculture. College-educated adults were 10 percentage points more likely to be in agriculture in 2023. Simultaneously, agricultural share of employment among undereducated (up to primary) has fallen 7 percentage points from 65 percent in 2017. Thes workers have instead turned to informal manufacturing and small-scale retail services – evident from their 13 and 5 percentage points rise in sectoral employment shares. Rising agriculture among the most educated underscores shows that structural transformation processes, which were key to rapid poverty reduction during Myanmar’s economy following liberalization, are now reversing. Table 3a: Change in sectoral employment by location Table 3b: Change in sectoral employment by education Rural Urban Agriculture Manufacturing Services 2017 2023 Diff 2017 2023 Diff Education 2017 2023 Diff 2017 2023 Diff 2017 2023 Diff Up to Agriculture 43 34 -10 6 9 3 65 58 -7 14 18 4 21 25 3 primary Up to Manufacturing 10 10 0 15 13 -2 43 48 5 23 23 0 34 30 -4 middle Up to Services 14 14 0 39 34 -6 24 32 8 22 22 0 54 46 -8 high sc. College and 8 18 10 14 11 -3 78 70 -8 above Notes: Shares are as a percent of the population. Notes: Shares are a percentage of the employed population. Changes are in percentage points. Changes are in percentage points. Rapid agricultural input price inflation could be one of the reasons for agricultural moderating among undereducated people. The 7-percentage point drop in agriculture among undereducated workers (Table 3b), is associated with 16 to 19 percent of low-educated workers and 10 to 14 percent of high-education workers reporting reductions in the use of agricultural inputs during the survey (Table 4). In addition to reducing agricultural inputs, undereducated workers were 15 percentage points more likely to borrow money and 5 percentage points less likely to spend savings than people with college degrees. With more debt and a lower ability to service this debt (given limited savings resources), persistent input price inflation could be one of the reasons driving undereducated people to reduce their use of input materials. Fewer inputs could lower outputs and marketable surpluses, making it economically infeasible for undereducated workers to remain in the sector. On the other hand, with potentially more resources at their disposal, college educated workers are able to weather the pressures of higher inflation. Therefore, the agriculture input price inflation channel could be one of the reasons for agriculture moderating among undereducated workers. Other factors not directly related to inflation, such as insecurity, conflict, 5 The remaining adult rural population is either seeking opportunities or has exited the labor force. 21 displacement pressures, transport and logistics constraints are also likely contributors that further reinforce this trend. The Moderation in Agriculture Among Under-Educated and Poorer Workers Between 2017 And 2023 Could Explain why Poverty Depth and Severity is Rising Moderation of agriculture among undereducated workers is correlated with cutbacks in food, education, and healthcare consumption among such families. Undereducated workers were 5 percentage points more likely to reduce food expenditures, 8 percentage points more likely to curtail health expenses, and 1 percentage point more likely to reduce education expenses than college-educated workers (Table 4). As noted, these consumption cutbacks are likely the result of lower marketable agricultural surpluses caused by input price inflation and other factors. The consumption cutbacks reported by undereducated families in Table 4 is also consistent with the declining consumption trend among poorer households in rural areas, as observed in Figure 4 6. Further confirmation can be found in figure 10, which shows food, education and health expenditure contractions are more commonly adopted coping strategies among poorest households (that those, with quintile rank of 1) than others. Table 4: Coping strategies adopted by workers at varying levels of education. Reduction in Reduction in Reduction in Reduction in Increased Asset Sales agriculture Spend Savings education food health expenses borrowing inputs expenses expenditures up to 34 16 44 26 11 46 38 primary up to 37 19 44 25 13 46 35 middle up to high 32 14 48 23 11 45 33 college and 28 10 49 18 10 41 23 above Notes: Shares are reported as a fraction of adult workers. Figure 10: Coping Strategies by Quintiles of Consumption Notes: The horizontal axis shows quintiles of consumption. Households are allowed to select more than 1 coping strategy. As a result, summations may not add up to 1. 6 Poorer households in villages have high shares of members with under primary education. 22 Poorer and undereducated families already had low consumption levels in 2017; as education, food, and health expenditures are cutback further, the depth and severity of poverty starts rising. Recall that in Chapter 1, poverty severity and depth was found to be sharply rising between 2017 and 2023. The reported reductions in food, education, and health expenditures among undereducated/poorer households above, could be one of the potential causes driving the rise in poverty depth and severity. To further investigate this, Figure 11 compares key coping strategies of poorest households – that are differentiated by whether they have at least one member working in agriculture. Households with at least one member in agriculture were less likely to resort to consumption cutbacks than families with no member working in the sector. For instance, 55 percent of households exposed to agriculture reported cuts to food expenses, 36 percent to health expenses and 14 percent on education related expenses. In comparison, households with no member working in agriculture, consumption cutbacks across all three categories were 8, 15 and 9 percentage points higher. Therefore, poorest households with no exposure to agricultural employment, face a greater risk of food, health, and education consumption contractions. The reduction in agriculture employment could therefore raises the risk of depth and severity of poverty. Figure 11: Poverty severity and Consumption cutbacks Notes: Household weights are used in the figure. With Limited Avenues for Agricultural Employment in Cities, Urban Poverty Could Rise Further If Current Labor Market Trends Continue If current labor market trends persist, urban poverty is likely to rise further. Figure 12 shows the share of households reporting a fall in income three months before the survey and the share of agricultural employment by decile rank of rural and urban households. The figure highlights four main stylized facts. First, although there has been a rise in agriculture activities in urban areas and a fall in rural areas between 2017-2023 (likely due to the various factors described above), agriculture continues to be the main sector of employment in villages. Second, the poorest rural households had the highest employment share in agriculture in 2017. Agriculture’s share among this group has moderated by 2023 (even though they are still the most likely to work in the sector than another household group). This is consistent with the drop in 23 agriculture employment among undereducated groups reported in Table 3b. Third, many richer urban households have shifted to agriculture between 2017 and 2023. However, poorer urban households have not made this sectoral switch (or they are unable to). Fourth, households with less agricultural exposure are most likely to report income losses. For instance, nearly sixty percent of urban households in the first decile reported income contractions over the past three months. At the same, these households were amongst the least likely to switch to agriculture in 2023. In contrast, agriculture employment share in 2023 is observed to be high across almost all rural households. Simultaneously, rural households are less likely to report income contractions. These patterns reveal a negative relationship between exposure to agriculture employment and household income losses. This complements the findings patterns presented in Figure 11, which showed that poorest households with exposure to agriculture reported fewer consumption cuts than those that do not have any member working in agriculture. These findings reinforce the role of the agriculture sector for poverty due to its shock-absorbing properties. Opportunities for agricultural employment in urban areas is going to remain limited. This means that with current levels of job losses and rising informality, the chances of urban poverty rising in the future is likely high. In rural areas, with no further reduction in agricultural employment shares, the sectoral guardrails could potentially hold back poverty from rising further. Figure 12: Self-reported income shocks, agricultural employment by consumption deciles Notes: Self-reported fall in household income are over the 3 months preceding the survey. Agriculture shares are relative to total population in decile. Absence Of Conflict-Induced Shocks and Relative Stability in The Household Level Economy Could Explain Ayeyarwady’s Positive Experience of Poverty Reduction Between 2017 And 2023 Figure 13 disaggregates changes in average consumption between 2017 and 2023 at the state and regional level by seven factors that can impact consumption, namely, household demographic structures, patterns of consumption for items typically consumed by poor or richest households, household asset ownership, conflict exposure, service provision, labor market indicators and adoption of various coping strategies. 24 The figure shows that Ayeyarwady has experienced fewer shocks across factors than other states and regions in Myanmar. Ayeyarwady is the only region where demographic changes, coping strategies, and consumption patterns have moved in a positive direction. The figure separates the direct effect of conflict on consumption (indicated by the yellow legend in the graph) from its indirect effects. The indirect effect captures the influence of conflict on consumption through adverse changes to assets, service provisions, adoption of coping strategies, and other factors. For instance, conflict can directly affect household consumption (for instance, by preventing people from growing their own food due to violence in their vicinity) or indirectly through other factors (e.g., conflict can adversely affect livelihoods by reducing aggregate demand for business activities). The labor market factors predict a fall in consumption in every state and region except Nay Pyi Taw. Moreover, changes in asset ownership predict a drop in consumption in almost every state and region. Demographic factors, such as household size and education, also predict a reduction in consumption in almost every state and region except Ayeyarwady. Finally, changes in service provision and coping strategies also point to a reduction in consumption in almost all states but are least likely to influence consumption in Ayeyarwady. The muted influence of these factors, that have adverse impacts in other areas, could explain its exceptional poverty trend between 2017 and 2023. Figure 13: Contribution of factors to changes in average consumption by states and regions. Technical Notes: The vertical axis shows changes in predicted consumption based on different household factors that are indicated in the legend. The contribution of a variable towards consumption can be estimated by multiplying the regression coefficient in Table T1 in the appendix by the average values of the indicator at the state and regional levels from MLCS-2017 and MSPS-2023. Changes in this value over the two periods can be interpreted as the impact of the indicator on changes in average consumption at the subnational levels. The exercise is repeated for all “demographic characteristics”, such as household size, female-headed households, whether there are any married in the household, if there is any member in the household that has no school education, whether the head of the household had below primary education, one-member households, three-member households, whether the head of the household is widowed, if the head of the household has over high school education, whether there are disabled members in the household and finally, if the household is Buddhist. These factors' contribution to average state and regional level consumption, whether net positive or negative, is then summed up to observe the aggregate effect of “demography” on average consumption. 25 Similarly, “consumption” factors include the effect of items such as betelnut, lottery expenses, cheroot, ready-made thanaka, and apple. “Assets” factors include generator ownership, firewood use, and color TV. Conflict data is from ACLED. “Service Provision” aggregates the impacts of access to formal health centers, access to piped water supply, and tube or boreholes. “Labor Markets” captures the impact of variables where the household head works in casual sector jobs, if there is any adult unemployed member in the household, is any member working in manufacturing, whether the head of the household is employed, and if the person is working in either manufacturing or services sector jobs. Finally, “Cope” aggregates the impact of family borrowing from friends or family (but not pawn shops), support from government or non-profit institutions, and whether the household was negatively impacted by illness/accident/death in the past 12 months. 26 Appendix 1: Technical details for estimating poverty in Myanmar using MSPS. Household surveys containing detailed consumption or income data are required to accurately measure poverty across countries. In Myanmar, the most recent official survey that provided such information was the Myanmar Living Conditions Survey (MLCS), conducted in 2017. The absence of subsequent survey data since then means that little is known about poverty trends in Myanmar during a turbulent period featuring the COVID pandemic and 2021 military coup. In an effort to fill this gap, the World Bank's Myanmar Subnational Phone Surveys (MSPS), initiated in 2022, have been collecting household data through telephone interviews. The MSPS may offer improved accuracy in poverty measurement for several reasons. Firstly, its larger sample size represents nearly all demographic groups in proportion to their population share, potentially reducing biases compared to earlier efforts, such as those of Karamba et al. (2022). These attempts had to rely on smaller telephonic surveys with approximately 2000 samples. Secondly, estimates from MSPS are representative at the sub-national levels, a feature lacking in the World Bank's previous high-frequency telephonic surveys. Earlier survey attempts, initiated in the immediate aftermath of the pandemic, were not designed to capture variations at the state and regional levels. Given significant disparities in conflict and cost-of-living shocks across subnational locations, the ability to diagnose shocks at the subnational levels can lead to more precise poverty estimates. Lastly, an internal review conducted by the World Bank identified certain assumptions in past studies that could be relaxed to improve the reliability of poverty prediction models. These findings have been incorporated in this report. Although these surveys cover nearly all subnational units in the country, and collect a range of employment, asset, coping strategies and other household indicators, detailed consumption data underpinning official estimates of poverty remains elusive because it is challenging to collect consumption information over telephonic interviews. We propose to impute consumption values into the MSPS survey to obtain estimates of poverty for 2023. The MSPS serves as an ideal candidate for this purpose. Its representative nature 7 assures that all socioeconomic groups within the population, ranging from the affluent to the middle- income and less privileged, are represented in the survey. Consequently, estimates of poverty, depth, severity and vulnerability, using MSPS are likely to be more accurate than other surveys that could are potentially skewed in favor of certain population groups. The first step in the imputation exercise is to estimate a model that governs the relationship between household characteristics and household (log) consumption. The model is trained using data from the last official survey, MLCS-2017, and uses variable present in both the MLCS-2017 and MSPS-2023 surveys. In anticipation of this imputation task, we designed the MSPS-2023 instrument to include specific questions aimed at capturing indicators that are strongly associated with consumption patterns. These include, for instance, data on household size, gender of the household head, marital status, and consumption of key items typically associated with lower- income households, such as cheroot, thanaka, betelnut, and others. The survey instrument also ensure that the data remains fully consistent with MLCS-2017 in terms of phrasing, contextual 7 see Sinha Roy, 2023 and Sinha Roy, 2024 for a detailed examination of MSPS’ representative properties. 27 framing, and recall mechanisms. This helped minimize the risk of consumption estimates being contaminated by the lack of comparability of household indicators across the two time periods. Having addressed representativeness and compatibility issues, the subsequent challenge was to identify variables that could serve as predictors of consumption in 2023. The imputation model assumes that the relationship between household characteristics and consumption has remained constant over time. This implies that changes in consumption and poverty levels between MLCS- 2017 and MSPS-2023 are driven primarily by changes in household attributes such as employment status, demographics, asset ownership, and discretionary spending habits. With that in mind, we do not exclusively focus on covariates that rank as the strongest predictors of household consumption back in 2017, as these need not necessarily be the variables that are well equipped to pick up changes over time. Specifically, we compiled two candidate sets of potential predictors: (a) those identified as strong predictors using LASSO model selection, ranked by their power to predict consumption in 2017, and (b) covariates that best capture significant recent changes, particularly deteriorations in household conditions, even if they ranked lower in predictive capabilities than variables from set (a). Our base model comprises variables selected from the second list – those that better capture changes in household conditions over time and exhibit significant correlation with (log) household consumption. Subsequently, we enhance this model by incorporating variables from the first set – representing variables that are highly correlated with consumption in 2017. Note that as more variables from the set (a) are integrated into the model, it becomes increasingly tailored to conditions observed in 2017. This enhances the model's goodness-of-fit in 2017 but simultaneously heightens the risk of ignoring changes in household well-being conditions since 2017. Accordingly, our final model aims to strike a balance between the goodness-of-fit observed in 2017 and sensitivity to evolving conditions in later years. We present the resulting model in Table T1 below, fitted separately for rural and urban areas of Myanmar. Relying on this model, we impute 100 consumption values for each household using Stata's Multiple Imputation (MI) command 8. The predicted consumption values are then used to estimate poverty in MSPS 2023. This entails evaluating for each household the likelihood that their consumption is below the poverty line. These probabilities are then aggregated to derive an estimate of the proportion of the population living below the poverty line. In 2017, the poverty line was estimated at 1590 Kyats in adult equivalent per capita terms, representing the average cost of a minimum needs basket across Myanmar at that time. However, even at the time, prices across Myanmar varied significantly by state and regions. In more remote areas, the cost of the minimum needs basket exceeded the average of 1590 Kyats applicable to Myanmar as a whole. Proportionately, the price of items in Yangon were lower given stronger connectivity and improved market access. Consequently, to assess poverty against the 1590 Kyats poverty line, the 2017 adult equivalent consumption in MLCS was adjusted at the state and regional level, in order to account for variations in the cost of living across different locations. By 2023, Myanmar has experienced a substantial increase in conflict and cost-of-living crises, arguably affecting prices in states and regions disproportionately. For instance, Sagaing witnessed 8 Each of these 100 repetitions represent draws from the posterior distribution of the model parameters and an error term drawn from a random normal distribution. 28 a notably higher incidence of conflict compared to Nay Pyi Taw. These differential shocks have likely exacerbated the differences in relative prices across the country, causing the cost of the minimum needs basket to rise significantly beyond the real value of 1590 Kyats that is based on average price of items across Myanmar as whole. Since the model was trained on 2017 data, the potential dispersion in relative prices due to conflict and cost-of-living crises are not captured in model parameters and therefore, not reflected in the imputed consumption values. To incorporate spatial price differences in our poverty estimates for 2023, we modify the 1590 Kyat poverty line according to relative prices of commodities observed in each state and region of Myanmar, further disaggregated into rural and urban sectors. Specifically, we multiply the 2017 poverty line by the relative prices observed for each state and region (separately for rural and urban areas) compared to the average price observed across Myanmar. Unfortunately, detailed sub- nationally disaggregated price or inflation data from recent periods is unavailable to us. Instead, we use price data from IFPRI's Myanmar Household Welfare Survey (MHWS). The survey contains self-reported prices information for approximately 50 items, 17 of which are shared with MLCS. We restrict our calculations of relative prices to commodities shared by the two surveys to facilitate compatibility 9. We utilize price data from 6 rounds of MHWS surveys conducted between 2022 and 2023. Figure F1 shows the distribution of relative prices across 15 states and regions of Myanmar: a distribution centered around 1 indicates no spatial dispersion of prices. In contrast, a flatter spread out of distribution would indicate greater price dispersion across the country. The figure supports our intuition of greater price dispersion in 2023 than in 2017 across Myanmar. In 2017, state and region prices are observed to be closer to the average price for Myanmar – as indicated by the significant density of relative prices clustered around 1 (when subnational prices are exactly equal to the average price across Myanmar, relative price is equal to 1). In 2023, however, a visible increase in price dispersion can be observed across provinces. In rural areas of selected provinces, prices are 1.4 times the average price across Myanmar, while in selected urban areas, prices are seen to be below 0.8 times the Myanmar average. Adjusting the poverty line based on these relative prices is assumed to help account for the divergence in the price of basic needs. Figure F1: Relative prices in 2023 across state and regions of Myanmar 9 Note that differences in quality of items and package sizes consumed in MLCS-2017 and MHWS surveys could still contribute to incompatibilities between the two surveys. We abstract from these differences given limited choices of other price information available in Myanmar. 29 Table T1: Dependent variable: Log (adult equivalent consumption per capita) Rural Urban Household has 1 member 0.666*** 0.723*** -21.45 -14.27 Household has 2 members 0.402*** 0.400*** -21.54 -14.08 Household has 3 members 0.253*** 0.215*** -18.85 -9.57 Household head is female -0.042* -0.041 (-2.09) (-1.18) Household head is casual worker -0.041 -0.023 (-1.15) (-0.35) Household head is working 0.009 -0.041 -0.69 (-1.51) Household head has primary or below levels of education -0.019 -0.025 (-1.56) (-1.27) Household head works in the manufacturing sector 0.081** 0.104** -3.15 -2.88 Household head is widowed 0.032 -0.037 -1.53 (-1.09) Household head works in the services sector 0.104*** 0.100*** -5.49 -3.86 Household head has high school or more levels of education -0.025 0.211*** (-1.45) -8.82 Consumption of Betel Nuts -0.024* -0.093*** (-2.11) (-3.88) Consumption of Firewood in the past 30 days 0.012 -0.138*** -1.09 (-5.86) Consumption of Cheroot -0.085*** -0.128*** (-7.59) (-6.26) Consumption of Readymade Thanaka 0.007 -0.001 -0.66 (-0.06) Consumption of Apples 0.170*** 0.109*** -12.58 -5.95 Expenses on Lottery 0.164*** 0.116*** -12.64 -6.3 Owns a Generator 0.214*** 0.352*** -8.27 -10.34 Owns a Color TV 0.239*** 0.291*** -21.31 -14.4 Access to a private piped water connection 0.082* -0.03 -2.44 (-0.72) Access to water from a tube or a borewell -0.007 -0.079*** (-0.55) (-3.88) Any adult (18+) member of the household that is unemployed -0.012 -0.057 (-0.44) (-1.23) Any married member in the family 0.021 -0.008 -0.95 (-0.23) Any member of the household works in manufacturing sector -0.061** -0.158*** (-3.23) (-6.03) Any adult member (18+) in the family without school education -0.094* -0.073 (-2.45) (-0.93) Any disabled members in the family -0.082*** -0.108*** (-4.80) (-3.66) Received income support from authorities 0.03 0.039 -1.61 -0.89 Received income support from non-profit organizations -0.024 -0.037 (-0.84) (-0.47) Borrowed money from outside the family in the past 12 months 0.024* -0.036 -2.07 (-1.77) Access to Formal Health Facilities 0.053*** 0.001 -4.71 -0.08 Serious illness or death in the family over the past 12 months 0 -0.022 -0.02 (-0.74) Household's primary religion is Buddhism 0.091*** 0.049 -4.31 -1.72 Acled Conflict Data -0.009 -0.035 (-0.75) (-1.92) 30 Constant 7.247*** 7.646*** -200.77 -148.08 Observations 8387 5342 R-squared 0.304 0.35 31 Appendix 2: Detailed Statistical Tables Table A1: Headline Labor Market Indicators 2017 2022 2023 State UE LFPR UE LFPR UE LFPR (%) WPR (%) WPR (%) WPR (%) (%) (%) (%) (%) (%) Ayeyarwady 64% 62% 1% 59% 55% 4% 68% 63% 5% Bago 61% 59% 2% 61% 59% 3% 66% 61% 5% Chin 64% 62% 2% 68% 67% 1% 68% 62% 6% Kachin 60% 56% 4% 60% 55% 5% 64% 54% 10% Kayah 67% 64% 3% 38% 34% 5% 55% 50% 5% Kayin 55% 53% 2% 40% 31% 8% 49% 45% 4% Magway 68% 67% 0% 61% 57% 4% 65% 61% 5% Mandalay 68% 67% 1% 59% 56% 3% 64% 59% 5% Mon 51% 50% 1% 56% 53% 3% 61% 57% 4% Nay Pyi Taw 66% 65% 2% 54% 52% 2% 66% 61% 5% Rakhine 61% 59% 2% 60% 50% 10% 54% 48% 7% Sagaing 70% 69% 0% 57% 54% 3% 60% 57% 4% Shan 72% 72% 1% 65% 62% 3% 55% 49% 7% Tanintharyi 64% 63% 1% 52% 48% 4% 54% 51% 3% Yangon 59% 56% 3% 57% 53% 5% 66% 59% 6% Myanmar 64% 63% 2% 59% 54% 4% 62% 57% 5% Notes: LFPR = labor force participation rate, WPR = Worker population ration, UE = Proportion of the population employed. Sample restricted to 15+. 32 Table A2: Headline Labor Market Indicators (by gender) 2017 2023-24 State Male Female Male Female UE LFPR WPR UE LFPR WPR UE LFPR WPR UE LFPR (%) WPR (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) Ayeyarwady 79% 78% 2% 50% 48% 1% 82% 77% 5% 55% 49% 6% Bago 75% 74% 2% 48% 46% 2% 81% 75% 5% 52% 48% 4% Chin 74% 72% 2% 55% 53% 2% 84% 83% 1% 55% 45% 10% Kachin 77% 73% 4% 44% 40% 4% 72% 64% 8% 55% 43% 12% Kayah 77% 73% 4% 58% 55% 3% 67% 58% 8% 42% 40% 2% Kayin 70% 67% 3% 42% 41% 1% 60% 59% 1% 40% 34% 6% Magway 76% 76% 1% 61% 61% 0% 76% 72% 4% 57% 52% 5% Mandalay 80% 78% 1% 59% 58% 1% 74% 70% 4% 55% 50% 5% Mon 67% 65% 1% 39% 38% 1% 78% 74% 4% 47% 43% 5% Nay Pyi Taw 81% 79% 2% 54% 53% 2% 75% 71% 4% 58% 52% 6% Rakhine 78% 75% 3% 46% 45% 2% 63% 57% 6% 47% 40% 7% Sagaing 78% 77% 0% 63% 62% 0% 76% 72% 4% 48% 44% 4% Shan 82% 81% 1% 63% 63% 0% 63% 56% 7% 48% 42% 6% Tanintharyi 79% 78% 1% 51% 50% 1% 69% 68% 2% 37% 33% 4% Yangon 72% 69% 3% 48% 45% 3% 79% 73% 5% 55% 48% 7% Myanmar 77% 75% 2% 53% 52% 1% 74% 69% 5% 52% 46% 6% Notes: LFPR = labor force participation rate, WPR = Worker population ration, UE = Proportion of the population employed. Sample restricted to 15+. Population in millions are produced using individual-level sampling weights and may not match with projected population estimates of states and regions. 33 Table A3: Headline Labor Market Indicators (by location) 2017 2023-2024 State Rural Urban Rural Urban LFPR WPR UE LFPR WPR UE LFPR WPR UE LFPR WPR UE (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) Ayeyarwady 64% 63% 1% 61% 60% 2% 69% 63% 6% 63% 60% 3% Bago 61% 59% 2% 60% 58% 3% 65% 60% 5% 67% 65% 2% Chin 66% 64% 2% 57% 54% 3% 75% 69% 7% 33% 33% 0% Kachin 62% 57% 4% 57% 54% 3% 61% 54% 7% 69% 54% 16% Kayah 69% 65% 4% 62% 59% 3% 55% 50% 5% 57% 51% 7% Kayin 53% 51% 2% 59% 58% 1% 48% 46% 2% 52% 41% 11% Magway 69% 69% 0% 58% 57% 1% 67% 62% 4% 58% 52% 7% Mandalay 71% 70% 1% 62% 61% 1% 64% 60% 4% 62% 56% 6% Mon 50% 49% 1% 54% 53% 1% 61% 57% 4% 60% 56% 4% Nay Pyi Taw 69% 67% 1% 61% 58% 3% 68% 63% 5% 60% 54% 5% Rakhine 60% 58% 2% 63% 61% 2% 59% 52% 7% 32% 26% 6% Sagaing 71% 70% 0% 64% 63% 1% 62% 58% 4% 53% 47% 5% Shan 75% 75% 1% 63% 62% 1% 53% 47% 7% 60% 53% 6% Tanintharyi 66% 65% 1% 60% 58% 1% 47% 46% 1% 73% 66% 7% Yangon 63% 60% 3% 57% 54% 3% 71% 65% 6% 63% 57% 6% Myanmar 66% 65% 1% 60% 57% 2% 63% 58% 5% 61% 55% 6% Notes: LFPR = labor force participation rate, WPR = Worker population ration, UE = Proportion of the population employed. Sample restricted to 15+. Population in millions are produced using individual-level sampling weights and may not match with projected population estimates of states and regions. 34 Table A4: Headline Labor Market Indicators (by education) 2017 2023-2024 State Primary education Secondary education Tertiary education Primary education Secondary education Tertiary education LFPR WPR UE LFPR WPR UE LFPR WPR UE LFPR WPR UE LFPR WPR UE LFPR WPR UE (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) Ayeyarwady 64% 63% 1% 60% 59% 2% 79% 78% 1% 66% 60% 6% 71% 67% 4% 79% 74% 4% Bago 59% 58% 2% 61% 58% 3% 78% 77% 1% 63% 58% 5% 70% 65% 5% 70% 68% 2% Chin 66% 65% 1% 60% 57% 3% 76% 76% 0% 67% 58% 10% 72% 70% 2% 49% 49% 0% Kachin 60% 56% 4% 57% 53% 4% 79% 74% 5% 56% 46% 10% 70% 60% 11% 64% 62% 2% Kayah 70% 66% 3% 62% 59% 4% 84% 82% 3% 59% 55% 4% 49% 42% 7% 76% 73% 2% Kayin 54% 52% 2% 52% 51% 2% 77% 76% 1% 39% 35% 4% 60% 56% 4% 87% 87% 0% Magway 69% 69% 0% 63% 62% 1% 78% 76% 1% 64% 60% 5% 66% 61% 5% 78% 75% 3% Mandalay 68% 68% 0% 65% 64% 1% 83% 80% 3% 60% 56% 4% 65% 60% 5% 75% 69% 5% Mon 48% 47% 1% 51% 49% 2% 77% 74% 3% 58% 55% 3% 62% 56% 6% 77% 74% 3% Nay Pyi Taw 65% 64% 1% 64% 62% 3% 84% 82% 2% 61% 59% 2% 71% 62% 9% 84% 74% 11% Rakhine 59% 57% 2% 60% 58% 2% 89% 87% 2% 54% 49% 5% 55% 46% 10% 79% 76% 3% Sagaing 68% 68% 0% 69% 68% 1% 82% 81% 1% 55% 53% 3% 67% 62% 5% 70% 64% 6% Shan 74% 74% 1% 66% 65% 1% 81% 80% 2% 52% 47% 5% 60% 50% 10% 77% 68% 8% Tanintharyi 66% 66% 1% 59% 58% 1% 84% 81% 3% 46% 44% 2% 59% 54% 5% 98% 96% 3% Yangon 56% 54% 2% 58% 55% 3% 68% 64% 5% 60% 52% 9% 66% 61% 5% 74% 68% 6% Myanmar 64% 63% 1% 61% 60% 2% 77% 74% 3% 58% 53% 5% 66% 60% 6% 75% 70% 5% Notes: LFPR = labor force participation rate, WPR = Worker population ration, UE = Proportion of the population employed. Sample restricted to 15+. Population in millions are produced using individual-level sampling weights and may not match with projected population estimates of states and regions. 35 Table B1: Formal and Informal Employment 2017 2022-2023 2023-2024 State Self- Self- Self- Salaried Casual Salaried Casual Salaried Casual employed employed employed Ayeyarwady 40% 45% 15% 24% 50% 27% 17% 47% 36% Bago 40% 41% 20% 22% 44% 34% 18% 40% 41% Chin 27% 33% 40% 5% 30% 64% 7% 38% 55% Kachin 40% 24% 36% 24% 52% 25% 26% 42% 32% Kayah 35% 28% 37% 34% 50% 16% 23% 47% 30% Kayin 26% 48% 26% 19% 41% 40% 28% 47% 24% Magway 38% 34% 28% 23% 47% 29% 15% 41% 44% Mandalay 41% 43% 16% 30% 43% 27% 26% 39% 35% Mon 43% 44% 14% 32% 44% 24% 27% 41% 32% Nay Pyi Taw 50% 37% 13% 42% 41% 17% 35% 36% 29% Rakhine 34% 48% 18% 28% 49% 23% 15% 40% 45% Sagaing 29% 37% 34% 22% 46% 32% 16% 41% 43% Shan 22% 37% 41% 30% 35% 35% 30% 42% 27% Tanintharyi 40% 43% 17% 33% 46% 21% 42% 44% 14% Yangon 55% 34% 11% 43% 41% 17% 47% 35% 18% Myanmar 38% 39% 23% 29% 44% 28% 26% 41% 33% Notes: Shares are a proportion of the employed workforce 36 Table B2: Formal and Informal Employment (by gender) 2017 2023-2024 State Male Female Male Female Self- Self- Self- Self- Salaried Casual Salaried Casual Salaried Casual Salaried Casual employed employed employed employed Ayeyarwady 43% 46% 11% 35% 44% 21% 18% 51% 32% 17% 41% 42% Bago 42% 45% 13% 36% 35% 28% 20% 45% 35% 16% 34% 50% Chin 32% 39% 29% 22% 26% 52% 11% 31% 57% 1% 48% 52% Kachin 41% 24% 35% 38% 24% 38% 27% 43% 31% 25% 42% 34% Kayah 40% 34% 27% 29% 20% 51% 32% 42% 26% 8% 56% 36% Kayin 29% 53% 18% 22% 41% 37% 36% 34% 30% 17% 67% 16% Magway 37% 47% 16% 38% 22% 40% 18% 49% 33% 12% 32% 55% Mandalay 43% 46% 11% 39% 40% 21% 32% 41% 27% 20% 37% 44% Mon 49% 42% 9% 34% 46% 19% 29% 44% 27% 25% 36% 39% Nay Pyi Taw 54% 37% 9% 44% 36% 19% 34% 39% 27% 37% 33% 31% Rakhine 37% 52% 11% 29% 43% 28% 18% 42% 39% 11% 38% 51% Sagaing 34% 45% 22% 25% 28% 46% 19% 42% 39% 12% 40% 48% Shan 26% 48% 26% 18% 24% 58% 25% 54% 20% 36% 27% 36% Tanintharyi 46% 41% 14% 32% 47% 21% 44% 40% 16% 38% 52% 10% Yangon 57% 36% 7% 53% 32% 16% 46% 39% 15% 49% 31% 21% Myanmar 41% 44% 15% 35% 34% 32% 28% 44% 28% 25% 36% 39% Notes: Shares are a proportion of the employed workforce 37 Table B3: Formal and Informal Employment (by location) 2017 2023-2024 State Rural Urban Rural Urban Self- Self- Self- Self- Salaried Casual Salaried Casual Salaried Casual Salaried Casual employed employed employed employed Ayeyarwady 39% 45% 15% 40% 46% 15% 15% 47% 38% 29% 47% 24% Bago 37% 41% 22% 49% 39% 12% 16% 40% 44% 29% 43% 27% Chin 20% 33% 46% 57% 31% 11% 3% 37% 60% 44% 45% 11% Kachin 38% 22% 40% 43% 27% 29% 26% 39% 35% 25% 49% 26% Kayah 28% 28% 44% 58% 27% 15% 23% 47% 30% 23% 48% 29% Kayin 25% 49% 27% 30% 45% 25% 29% 48% 23% 28% 43% 29% Magway 36% 34% 30% 48% 37% 15% 13% 40% 47% 30% 46% 24% Mandalay 38% 45% 17% 49% 40% 11% 21% 40% 39% 40% 36% 24% Mon 43% 43% 13% 41% 44% 15% 24% 43% 33% 39% 34% 27% Nay Pyi Taw 31% 50% 19% 49% 41% 11% 13% 40% 47% 38% 41% 20% Rakhine 27% 37% 36% 44% 36% 20% 13% 41% 46% 35% 44% 21% Sagaing 18% 37% 45% 38% 38% 24% 31% 44% 25% 28% 38% 34% Shan 38% 44% 18% 46% 40% 13% 41% 40% 18% 43% 51% 6% Tanintharyi 51% 36% 13% 57% 33% 10% 36% 39% 25% 53% 34% 14% Yangon 34% 40% 26% 49% 37% 14% 20% 42% 38% 41% 38% 21% Myanmar 44% 41% 15% 65% 27% 8% 30% 37% 33% 49% 34% 17% Notes: Shares are a proportion of the employed workforce 38 Table B4: Formal and Informal Employment (by education) 2017 2023-24 State Primary education Secondary education Tertiary education Primary education Secondary education Tertiary education Self- Self- Self- Self- Self- Self- Salaried Casual Salaried Casual Salaried Casual Salaried Casual Salaried Casual Salaried Casual employed employed employed employed employed employed Ayeyarwady 41% 48% 11% 35% 42% 23% 52% 29% 19% 13% 48% 39% 20% 45% 35% 38% 48% 14% Bago 41% 41% 18% 34% 42% 24% 60% 31% 9% 9% 45% 45% 24% 36% 40% 48% 34% 18% Chin 14% 41% 45% 34% 28% 39% 88% 8% 4% 7% 11% 82% 7% 55% 38% 9% 90% 1% Kachin 33% 27% 40% 41% 22% 37% 64% 21% 15% 12% 57% 30% 32% 32% 35% 44% 43% 13% Kayah 25% 33% 42% 40% 23% 36% 72% 17% 11% 35% 44% 20% 11% 49% 41% 19% 52% 29% Kayin 23% 52% 25% 28% 44% 28% 48% 27% 24% 27% 59% 14% 32% 33% 35% 7% 86% 7% Magway 36% 36% 28% 36% 33% 31% 58% 26% 17% 8% 44% 48% 19% 38% 43% 38% 31% 31% Mandalay 38% 47% 15% 40% 43% 17% 61% 27% 12% 20% 38% 41% 31% 37% 32% 35% 48% 17% Mon 42% 47% 11% 38% 44% 18% 59% 30% 11% 20% 49% 31% 32% 31% 36% 41% 38% 20% Nay Pyi Taw 43% 43% 14% 51% 35% 14% 79% 15% 6% 35% 39% 26% 29% 36% 35% 61% 22% 17% Rakhine 32% 51% 17% 29% 48% 23% 79% 16% 5% 10% 42% 48% 15% 42% 44% 79% 16% 5% Sagaing 26% 41% 33% 32% 31% 36% 41% 30% 29% 11% 46% 43% 21% 35% 44% 30% 46% 24% Shan 17% 41% 42% 31% 28% 41% 53% 28% 19% 27% 44% 29% 33% 40% 26% 45% 35% 20% Tanintharyi 36% 49% 16% 45% 37% 17% 51% 23% 25% 51% 36% 13% 26% 55% 20% 53% 43% 4% Yangon 49% 39% 11% 54% 34% 12% 67% 26% 7% 34% 42% 23% 49% 35% 16% 56% 30% 14% Myanmar 34% 43% 23% 40% 37% 24% 60% 27% 13% 18% 44% 37% 30% 38% 32% 46% 37% 17% Notes: Shares are a proportion of the employed workforce 39 Table C1: Total workers by industry Industry 2017 2022-23 2023-24 Agriculture 16296131 12678162 14370407 Mining and quarrying 306448 334415 412123 Manufacturing 3207353 2399982 3315308 Electricity, gas, water supply, waste management 45356 232260 137830 Construction 2120801 2480129 2021248 Wholesale and retail trade 5068531 6297886 5912350 Transportation 1353837 1250464 1517876 Hospitality, communication, finance, real estate 1173405 945449 962212 Professional 97195 169666 149336 Administrative and public admin 445599 408744 508023 Education, health, social work 1101431 1321474 1031838 Art, entertainment 75525 79853 116946 Other 1657387 669515 666503 Notes: Calculated using share of industry’s employment in the total employed workforce * worker- population ratio * population projection for the year. Sample includes population over 15. Table C2: Share of wage workers in total employment (%) Industry 2017 2022 2023 Agriculture 25.4 11.8 5.8 Mining and quarrying 72.5 29.8 50.0 Manufacturing 55.6 49.5 53.6 Electricity, gas, water supply, waste management 84.4 74.0 77.7 Construction 89.3 54.3 76.0 Wholesale and retail trade 20.6 20.8 18.5 Transportation 42.2 29.4 36.4 Hospitality, communication, finance, real estate 36.2 41.2 40.6 Professional 58.8 83.6 59.8 Administrative and public admin 98.9 80.2 88.8 Education, health, social work 86.3 85.4 81.9 Art, entertainment 66.3 59.9 50.8 Other 53.2 52.5 38.0 Notes: Calculated using share of industry’s employment in the total employed workforce * worker- population ratio * population projection for the year. Sample includes population over 15. Table C3: Median wages by industry 2017 2022 2023 40 Agriculture 6081 7644 8000 Mining and quarrying 10424 11678 10000 Manufacturing 8687 8493 7000 Electricity, gas, water supply, waste management 9266 8918 6000 Construction 10424 10192 10000 Wholesale and retail trade 8687 8493 7333 Transportation 10424 12740 10000 Hospitality, communication, finance, real estate 8687 8748 8333 Professional 9671 12740 9333 Administrative and public admin 9555 8918 6667 Education, health, social work 9266 8493 6667 Art, entertainment 7876 10617 10000 Other 6949 8493 6667 Notes: Wages are available for salaried workers only. In per day, real 2023 Kyat terms. References Karamba, R. W., & Salcher, I. (2022). Progress, Setbacks, and Uncertainty: Effects of COVID-19 and Coup on Poverty in Myanmar. Lakner, C., Mahler, D. G., Negre, M., & Prydz, E. B. (2022). How much does reducing inequality matter for global poverty? The Journal of Economic Inequality, 20(3), 559-585. Scheidel, W. (2017). The great leveler: Violence and the history of inequality from the stone age to the twenty-first century. Princeton University Press. Sinha Roy, S. (2023). Myanmar Subnational Phone Surveys (MSPS) of the World Bank: Coverage, Reliability, and Representativeness. World Bank. Sinha Roy, S. (2024). Populations in Peril: Decoding Patterns of Forced Displacement in Myanmar. World Bank. World Bank. (2022). Poverty and shared prosperity 2022: correcting course. The World Bank. 41