Navigating Two Decades of High Poverty and Charting a Course for Change in Madagascar Poverty and Equity Assessment February 2024 Madagascar Poverty and Equity Assessment Acknowledgments © 2024 International Bank for This report was prepared by a World Bank team led by Francis Mulangu Reconstruction and Development / (Senior Economist and Task Team Leader, EAEPV) and Ana Maria Oviedo The World Bank (Senior Economist and co-Task Team Leader, EAEPV) and consisted of a core team including Alexandra (Sasha) Jarotschkin (Economist, EAEPV), 1818 H Street NW, Nobuo Yoshida (Lead Economist, EPVGE) and Javier Baez (Lead Econ- Washington DC 20433 omist, EAEPV). Giovanni Vecchi, Giulia Mancini (Consultants, University Telephone: 202-473-1000 of Rome Tor Vergata), Fatoumata Dieng (Consultant, EAEPV), Nyasha Internet: www.worldbank.org Kurasha (Consultant, EAEPV), Maria Louisette Ranorovololona (Consult- ant), Haingo Randrianarivony (Consultant), Michael Keller (Consultant, EAEPV), Amit Nga (Consultant), Serge Radert (Consultant, EAEPV), Jean Rights and Permissions Razafindravonona (Professor, University of Antananarivo), Jean-Pascal The material in this work is subject Nguessa Nganou (Senior Economist, EA1M2), Takaaki Masaki (Senior to copyright. Because The World Economist, EPVGE), Koichi Ito (Consultant, EAEPV), David Newhouse Bank encourages dissemination of (Senior Economist, DECCS), Dunstan Matekenye (Data Scientist, DECSC), its knowledge, this work may be Antsa Raoelijaona (ET consultant, EA1M2), Faniry Razafimanantsoa reproduced, in whole or in part, for Harivelo (Senior Economist, EA1M2), Maud Juquois (Senior Health Econ- noncommercial purposes as long as omist, HHNGF), Atsushi Iimi (Senior Economist, IAFT2), Gael Fetrani- full attribution to this work is given. aina Raserijaona (Urban Specialist, SAEU2), Ibrahim El ghandour (Public Sector Specialist, EAWG1), Stephen D'Alessandro (Senior Agricultural Any queries on rights and licenses, Economist, SAEA2), Daniel Gerszon Mahler (Senior Economist, DECID) including subsidiary rights, should be and Samuel Kofi Tetteh Baah (Economist, DECID) provided key inputs addressed to World Bank Publications, and feedback to the report. The team is grateful to peer reviewers The World Bank Group, 1818 H Street German Caruso (Senior Economist, HHCDR), Leonardo Lucchetti (Senior NW, Washington, DC 20433, USA; Economist, EECPV), Trang Van Nguyen (Senior Economist, EMNPV), and fax: 202-522-2625; Andres Castaneda Aguilar (Economist, DECID) for their insightful sug- e-mail: pubrights@worldbank.org. gestions during various stages of the report. The report was prepared under the overall guidance and supervi- This work is a product of the staff sion of Zviripayi Idah Pswarayi Riddihough (Country Director, AFCS2), of the World Bank with external Marie-Chantal Uwanyiligira (Country Director, AWCF2), Atou Seck contributions. The findings, (Country Manager, AEMMG), and Pierella Paci (Practice Manager, interpretations, and conclusions EAEPV). Luis-Felipe Lopez-Calva (Global Director, EPVDR) provided expressed in this work do not strategic guidance and participated in high-level discussions with the necessarily reflect the views of The Government of Madagascar in 2023. World Bank, its Board of Executive Directors, or the governments that The team would like to express gratitude to the Government of Mada- they represent. gascar for helping to define the scope of work, sharing data, providing The World Bank does not guarantee inputs, and reviewing initial rounds of recommendations. Key counter- the accuracy of the data included in part institutions include INSTAT (National Statistics Institute) and the this work. Ministry of Economy and Finance. The team is grateful to all counter- parts from the government, civil society, academia and development partners who shared their inputs with the World Bank team during con- sultations in 2023. The team is also grateful to the people of Madagas- car who generously contributed their time to share their personal stories of poverty and resilience during focus group discussions, providing the rich qualitative information presented in this report. ii Madagascar Poverty and Equity Assessment This Poverty Assessment is supported by the Climate Support facility Whole-of-Economy Program, administered by the World Bank. The team would also like to express gratitude to Litera for translation services, Elizaveta Tarasova for design work, Diana Styvanley (Communi- cations Officer, AFREC) and Ainasoa Felana Ramasy (Consultant, ECRAE) for media and communications support. iii Madagascar Poverty and Equity Assessment List of acronyms ACTP Argent Contre Travail Productif BDI Burundi BGD Bangladesh CAF Central African Republic CE Cours Élémentaire CM Cours Moyen COD Democratic Republic of Congo COVID Coronavirus Disease CP Cours Préparatoire DHS Demographic and Health Survey Enquête Nationale sur les Objectifs Millenaires du Développement ENSOMD (National Survey on the Millennium Development Goals) EPM Enquete Permanent auprès des Ménages EU European Union FID Fonds d'Intervention pour le Développement Foibem-pirenena momba ny Fikarohana ampiharina amin’ny FOFIFA Fampandrosoana ny eny Ambanivohitra (National Centre of Applied Research and Rural Development) Fikambanan'ny Ray Aman-drenin'ny Mpianatra (Association des FRAM parents d'élèves) GDP Gross Domestic Product GEL Groupement des Exportateurs de Litchi GFF Global Financing Facility for Women, Children and Adolescents GIS Geographic Information System GRADE Global Rapid Damage Estimation HCI Human Capital Index HFPS High Frequency Pulse Surveys HH Household HTI Haiti ICT Information and Communication Technology IFPRI International Food Policy Research Institute ILO International Labour Organisation INSTAT Institut National de la Statistique IOE International Organisation of Employers KHM Cambodia LIC Low Income Country LPG Liquefied Petroleum Gas MDG Madagascar MGA Malagasy ariary MICS Multiple Indicator Cluster Survey MINAE Ministère de l'Agriculture et de l'Élevage MPO Macro Poverty Outlook iv Madagascar Poverty and Equity Assessment MSME Micro, Small and Medium Enterprise MTDLS Ministry of Territorial Development and Land Services NDVI Normalized Difference Vegetation Index NER Niger NGO Non Governmental Organization NPK Nitrogen, Phosphorus, and Potassium ODA Official Development Aid OECD Organization for Economic Cooperation and Development OLS Ordinary Least Squares OPHI Oxford Poverty and Human Development Initiative PASEC Programme on the Analysis of Education Systems PCER Per capita real consumption PEM Plan Emergence Madagascar PMT Proxy Means Test PPP Purchasing Power Parity RAI Rural Access Index RFMS Rapid and Frequent Monitoring System RWA Rwanda Centre d'Achats de Médicaments Essentiels et de Matériel SALAMA Médical de Madagascar (Medical Procurement Agency) SCD Systematic Country Diagnostic SDI Service Delivery Index SFA Stochastic Frontier Analysis SGBV Sexual and Gender Based Violence SOE State Owned Enterprise SSA Sub-Saharan Africa SWIFT Survey of Well-Being through Instant and Frequent Tracking TFP Total Factor Productivity TMDH Transfert Monétaire pour le Développement Humain TZA Tanzania UGA Uganda UN United Nations UNDP United Nations Development Programme UNICEF United Nations Children's Fund US United States WDI World Development Indicators WFP World Food Programme WHO World Health Organisation v Madagascar Poverty and Equity Assessment Table of Contents Acknowledgments ii List of acronyms iv Executive summary 1 Overview 4 Key findings 4 1. Long-term and recent developments that stifled poverty reduction 4 2. Low agricultural productivity and lack of basic services still trap 8 in 10 rural 7 people in poverty 3. Failing labor markets and recent shocks explain rising and deeper urban poverty 8 4. Progress in long and short-term poverty drivers is mixed 10 5. Repeated shocks wipe out income gains and asset accumulation 11 6. Policies to break the curse of low growth and high poverty 13 1. Two decades of poverty stagnation against a modest growth 14 performance Key findings 14 1. Poverty against broader political and economic trends 15 2. While the number of poor increased by 50 percent in 10 years, inequality 15 decreased for the wrong reason 3. Multidimensional poverty: Urban areas are slipping on nutrition 21 4. Characteristics of the poor, poverty vulnerability, and subjective poverty 26 5. Conclusion 35 Annex 1 38 2. Low agricultural productivity and market access 53 trap rural households in a poverty cycle Key findings 53 1. The link between agricultural productivity and poverty 53 2. Agricultural inputs and low agricultural productivity 56 3. Agricultural potential and efficiency: the productivity gap 60 4. Agricultural productivity: the case of rice 61 5. Marketing and post-harvest management 66 3. As urban markets fail, urban poverty rises 68 Key findings 68 1. Defining poverty in an urban context 68 2. Key drivers of increasing urban poverty 75 4. Low human capital limits options for escaping poverty 85 Key findings 85 1. Human capital development and poverty 86 2. Health and sanitation 90 3. Food security 95 4. Child marriage is a key determinant of intergenerational poverty 96 vi Madagascar Poverty and Equity Assessment 5. High vulnerability to shocks drives short term poverty but 100 has long term consequences Key findings 100 1. An overview of recent shocks and challenges to recovery 100 2. Types of shocks and coping mechanisms 101 3. Systemic shocks and their effects on welfare in urban and rural settings 104 Annex 2 112 6. Charting a course for change 114 Key findings 114 1. Improving agricultural productivity, market connectivity and resilience 114 2. Raising the productivity and quality of services of urban areas 120 3. Increasing resilience to shocks and strengthening basic safety nets 122 4. Establishing a Real-Time Poverty Monitoring System Adapted to Monitor 124 and Measure Impact of Climatic Disasters References 125 List of Figures: Figure 1: Agriculture represents close to one-fifth of GDP 4 Figure 19: Nutrition deprivation increased between 2018 23 but contributes negatively to growth and 2021 Figure 2: Between 2012 and 2022 urban consumption fell 5 Figure 20: Indicators of deprivation in Madagascar over 23 along the distribution, while rural consumption time in rural (left) and urban (right) areas improved more among the poorest Figure 21: Multidimensional poverty is mostly rural 24 Figure 3: National and rural poverty have stagnated while 5 urban poverty increased Figure 22: A quarter of Madagascar's population is illiterate 25 Figure 4: Poverty severity increased in urban areas and 6 Figure 23: Multidimensional poverty is concentrated in the 25 fell in rural areas primary sector Figure 5: Urban population growth between 2012-2022 8 Figure 24: The Human Opportunity Index for children aged 26 is explained by growth in urban poor population 6-16 is below the coverage rate Figure 6: Living conditions have deteriorated in urban 9 Figure 25: Location and assets explain most disparities in 26 areas while they improved nationally access to opportunities for children aged 6-16 Figure 7: Self-employment and family work are the 9 Figure 26: Larger households with children and agricultural 27 largest employment categories work are associated with higher poverty Figure 8: Education attainment has increased but quality 11 Figure 27: Rural vulnerability is significantly higher 31 remains low Figure 28: Rural, large households with no education are 31 Figure 9: Poor households were disproportionately hit by 11 the most vulnerable cyclones in 2022 Figure 29: Women are relatively more likely to work as 32 Figure 10: Coverage of safety nets is extremely low in rural 12 family workers and in subsistence farming areas Figure 30: Services tend to employ relatively more women 32 Figure 11: Urban poor population grew significantly in 16 2012-2022, but the poor remain predominantly Figure 31: Most workers earn poverty wages, regardless of 33 rural type of employment Figure 12: Madagascar remains among the poorest 17 Figure 32: Most people report to be struggling financially 34 countries globally Figure 33: Perceptions of financial difficulty are higher in 34 Figure 13: Poverty Gap and Poverty Gap square have 18 rural areas declined Figure 34: Lack of jobs is the main perceived cause of 35 Figure 14: Consumption growth was higher among the 19 poverty poorest households between 2001-2022 Figure 35: An example for the iterative process if the 41 reference group is too low Figure 15: Madagascar's inequality level remains below 20 peer countries' Figure 36: An example for the iterative process if the 41 reference group is too high Figure 16: Urban-rural consumption gap widens for richer 20 deciles Figure 37: Agricultural productivity is below peer countries 55 and declined in the last 30 years Figure 17: Poverty and Inequality varies widely across 21 regions Figure 38: While TFP and outputs grew util 2010, they fell 56 thereafter Figure 18: Multidimensional poverty fell significantly since 22 2008, despite the recent shocks Figure 39: The gender gap in agricultural land ownership 57 diminishes with age vii Madagascar Poverty and Equity Assessment Figure 40: Rice occupies over half of arable land 58 Figure 70: Primary school completion improved 89 considerably between 2000-2010 and slightly Figure 41: Self-employed household heads tend to 58 declined thereafter cultivate larger land areas, relative to wage employed heads, except for the richest 20 Figure 71: Life expectancy at birth has risen steadily but 91 percent slower than among peers Figure 42: Rice utilizes more labor than any other crop 59 Figure 72: Life expectancy level is above the LIC average 91 Figure 43: A conceptual framework to understand macro 60 Figure 73: Child mortality has declined at a slower pace 91 and micro drivers of agricultural efficiency than among SSA peers Figure 44: Domestic rice production systematically 62 Figure 74: Infant mortality is higher than among peer 91 exceeds consumption, but production has countries, but below LIC average fallen since 2009 Figure 75: Madagascar has less than half the hospital beds 92 Figure 45: Rice yields, poverty and connectivity are highly 63 than its peer countries and LIC average correlated Figure 76: Modern contraceptive use has increased, but 92 Figure 46: Improved rice seed use remains low 63 16 percent of women still lack access to any contraception Figure 47: Rice production in highland regions is 64 significantly larger than elsewhere (average Figure 77: Richer regions have slightly more health facilities 92 2012 – 2015) Figure 78: Madagascar has lower access to drinking water 93 Figure 48: Rice prices bottom up at mid-year 65 than the average LIC Figure 49: Price variation is substantial across regions, 65 Figure 79: Access to improved water and sanitation are 94 reflecting lack of arbitrage highly correlated in rural areas Figure 50: Regional price variation has widened in the last 66 Figure 80: Child marriage is prevalent in Madagascar ... 96 5-10 years Figure 81: ... it is more common rural areas 96 Figure 51: Asset ownership fell in urban and rural areas 70 Figure 82: Child marriage is higher in districts with high 97 Figure 52: Secondary cities drive the expansion of urban 71 multidimensional poverty poverty, 2001–22 Figure 83: Child labor and child marriage are correlated 98 Figure 53: Urban consumption fell throughout the 72 across districts distribution Figure 84: Child marriage is higher in districts with low 99 Figure 54: Demographic characteristics, infrastructure and 73 literacy economic opportunity are the main predictors of urban poverty Figure 85: Child marriage is higher in districts with fewer 99 secondary schools Figure 55: Over half the population 6+ years old needs to 74 work Figure 86: Shocks have repeatedly set back GDP growth 101 in Madagascar Figure 56: Micro firms were most affected by the pandemic 77 Figure 87: Illness and death are the most frequent 101 Figure 57: Access to finance is hampered by exceedingly 78 idiosyncratic shocks (Self-reported) high interest rates Figure 88: Food price and weather shocks are the most 103 Figure 58: Productivity among formal firms suffered a 78 common aggregate shocks (Self-reported) dramatic decline since 2009 Figure 89: Several COVID-19 waves hit Madagascar in 105 Figure 59: School enrollment increased over time 79 2020/2021 but the country's borders fully opened only in late 2022 Figure 60: Primary and secondary education have 80 increased in most regions Figure 90: Perception of food insecurity worsened during 106 the pandemic early stages Figure 61: Malagasy literacy increases with educational 80 access; less so for French and English Figure 91: Poor households were disproportionately 108 affected by cyclones in 2022 Figure 62: Earnings barely increase with education levels 81 Figure 92: Controlling for household characteristics, richer 108 Figure 63: Poverty reinforces low aspirations 83 households suffered larger losses Figure 64: Madagascar has one of the lowest rates of 84 Figure 93: Food prices have increased sharply since 2022 109 financial inclusion Figure 94: Price increase is the main reason for not being 110 Figure 65: A child born in Madagascar can only reach 39 86 able to buy basic staples percent if their potential productivity as an adult Figure 95: Reducing food consumption is the most 110 Figure 66: Expected years of school in Madagascar are 86 common coping mechanism against food price well below Sub-Saharan Africa's and South increases Asia's average Figure 96: Reducing non-food consumption is the most 111 Figure 67: Fertility among the poor is much higher, 87 frequent way of dealing with non-food price including among adolescents shocks Figure 68: Stunting remains high compared to peer 88 Figure 97: Poverty is expected to increase further due to 111 countries inflation Figure 69: Teenage pregnancy is strongly related to 88 Figure 98: Coverage of safety nets is still very low 123 education viii Madagascar Poverty and Equity Assessment List of Tables: Table 1: Inequality has declined as the urban/rural 6 Table 28: Over one-third of children under 11 are engaged 91 wealth gap narrowed in economic activity Table 2: Poverty incidence steadily increased in the last 16 Table 29: Urban average calorie consumption is higher 95 20 years than rural Table 3: Evolution of extreme urban poverty is most 18 Table 30: Poverty is higher for households with child 97 significant in secondary cities marriage, 2021-22 Table 4: National inequality slightly declined over the 19 Table 31: Geography, literacy and education of head of 98 period 2012-2022 (Gini index and shares of household are associated with not attending total consumption at the extremes of the school consumption distribution in %, 2005–22) Table 32: Households have few mechanisms to cope with 102 Table 5: Inequality is largest within broader geographic 19 idiosyncratic shocks (Self-reported) areas and between regions Table 33: Households have few mechanisms to cope with 104 Table 6: Demographic and agricultural variables drive 28 aggregate shocks (Self-reported) poverty Table 34: Urban job losses due to the pandemic far 105 Table 7: Women tend to work fewer hours than men 33 outnumber rural job losses Table 8: Multidimensional poverty is highly correlated 54 Table 35: Access to health services remained high and 106 with agricultural activity increased during the pandemic Table 9: Most agricultural households own their land 58 Table 36: Access to education services was interrupted 106 only briefly in 2020 Table 10: Agricultural households diversify production 59 across crops Table 37: A small percentage of households was not able 109 to buy cooking oil or rice even before the price Table 11: Fertilizer and other input use is minimal across 59 the welfare distribution (% households) Table 38: The current social assistance system has 123 marginal impact on poverty Table 12: Most households produce their own agricultural 60 inputs Table 13: Nitrogen-based fertilizer use has doubled since 63 Table A1.1: Countries which experienced a decline in income 38 2015 Table A1.2: Average daily calorie intake per capita and 42 Table 14: Most sales, except for cash crops, take place 67 cost per calorie for each decile of annual food through markets expenditures per capita Table 15: Farmers lack basic storage solutions for most of 67 Table A1.3: Variables of the MPI 45 their crops Table A1.4: MPI dimensions and indicators 46 Table 16: Asset ownership is heavily urban and male 70 Table A1.5: Poverty by characteristic of the household head 47 Table 17: There are wide gaps in access to services 71 between large and small cities Table A1.6: Stunting across national, rural, and urban areas 49 in 2018 and 2021 Table 18: Poverty gap and poverty gap squared increased 72 in urban areas Table A1.7: Full regression analysis—Determinants of 49 household welfare Table 19: Labor market outcomes show most people work 74 for very low wages Table 20: Close to half of urban low earners work full-time 75 Table A2.1: Only 6 countries have suffered a secular income 112 or more decline since independence Table 21: Poverty among the working-age population is 75 Table A2.2: Median variation in prices for food and cooking 112 higher in rural areas items Table 22: Urban households were disproportionately 76 Table A2.3: Food inflation coping mechanisms adopted by 113 affected by pandemic-related job losses households, by province (%) Table 23: Two-thirds of households lost income in 2020 77 Table A2.4: General inflation coping mechanisms adopted 113 by households, by province (%) Table 24: Returns to education are higher among older 82 cohorts Table 25: The incidence of health shocks more than 87 doubled in 2012-2022 Table 26: Attendance is much lower in rural areas, but 89 the gender gap is small (attendance rates of population aged 3 to 15, by location) Table 27: Fewer than 6 in 10 children finish primary 90 education with basic reading skills ix Madagascar Poverty and Equity Assessment List of Maps: Map 1: Madagascar by geographical areas, regions, and 3 provinces Map 2: Electricity coverage has improved along the 7 highlands but remains limited Map 3: There is a large North-South poverty divide at 22 the commune level Map 4: Multidimensional poverty is higher in the South 24 Map 5: Vulnerability is greatest in the South 31 Map 6: Agriculture land use is widespread and is 57 increasingly unsustainable Map 7: Northern and Eastern regions have the highest 61 agricultural revenue potential Map 8: Rural access and transport costs are correlated 66 Map 9: The South has the lowest number of health 93 facilities per children aged 0-5 Map 10: The South and the Capital have the lowest 93 number of health facilities per women aged 15-49 Map 11: 4.4 million households lack access to improved 94 sources of water Map 12: 4.7 million households lack access to improved 94 sanitation Map 13: Hidden hunger is largely located in the high 95 plateau with a few pockets in the south Map 14: Child marriage is higher in the West and South 96 Map 15: Recent cyclones have affected the East of the 107 island more significantly Map 16: Some regions with high poverty also have 116 untapped agricultural potential Map 17: Large gains in agricultural productivity can be 117 obtained through improved irrigation Map 18: Regions with high poverty and high 118 development potential should be prioritized for feeder roads investment List of Boxes: Box 1: Urban and Rural definitions for poverty analysis 69 Box 2: Predicting poverty through Random Forest 72 Analysis Box 3: Policy priorities from the Systematic Country 115 Diagnostic Update x TOC Madagascar Executive summary Poverty and Equity Assessment Executive summary Breaking the curse of high poverty and low growth For too long, Madagascar has struggled with low unable to serve as engines for growth and poverty and fluctuating GDP per capita, which fell by almost reduction for the country as standard development half between 1970 and 1990 and has remained theories would suggest. below US$500 since. Consistently, brief periods of positive growth have been interrupted by polit- Monetary poverty rates are alarmingly high but, ical crises and climate shocks which have time and they fail to capture the scope of the deprivation again set back any socio-economic progress. In the faced by poor households. About three-quarters past decade, a modest economic recovery between of the population suffers from food insecurity, and 2013 and 2019 was followed by the pandemic-in- this share has remained broadly unchanged for a duced economic crisis, causing GDP per capita to decade or more. Most households, especially in rural fall by 2.3 percent over the 2012-2022 period. areas, lack access to reliable electricity, safe water, or adequate sanitation. Access to healthcare is inade- This report provides an account of the evolution of quate while high fertility, teenage pregnancy (about poverty and living conditions in the decade 2012- one-third of girls 15-19 is a mother already) and low 2022. It finds that at the national level monetary education completion (only about half of all chil- poverty essentially stagnated while urban poverty— dren complete primary school) erode future human admittedly a much smaller in absolute and relative capital (Chapter 4). Weak road connectivity leaves terms—dramatically increased. In 2022, monetary rural communities isolated from markets and public poverty affected about 75 percent of the popula- services. As an illustration, regional differences in tion, a share slightly above the 73 percent in 2012.1 rice prices have widened substantially over the past Rural poverty remained roughly unchanged at 15 years, suggesting that domestic markets are about 80 percent of the rural population, but urban becoming less integrated over time (Chapter 2). poverty increased from 42 to 56 percent over the decade.2 The increase in poverty was especially dra- Looking forward, the country needs a radical matic in secondary cities, where poverty increased change in its growth trajectory to make a real from 46 to 61 percent (Chapter 1). dent on poverty. With an underdeveloped private sector mostly in subsistence agriculture and highly A closer look at the drivers of poverty reveals that exposed to the effects of climate change; low and the trends of the last decade are explained by falling agricultural productivity; and an inadequate market and governance failures, climatic shocks supply of infrastructure and essential services and the COVID pandemic. Structurally, stubbornly including education; a rapidly growing population high rural poverty in particular is the legacy of long- will find itself always more trapped in poverty and term infrastructure underinvestment, isolation, and exert increasing pressure on public goods and nat- low internal demand (World Bank Group, 2022). But ural resources, potentially threatening social and since 2013, this structural failure to launch has also political stability. Despite the increasingly difficult affected urban employment and living conditions external conditions, Madagascar has an opportunity as private investment has persistently declined and to break free from the cycle of low growth and weak competition was suffocated by special interests.3 governance. To avert a new crisis, the government Moreover, the COVID pandemic—which caused an must implement a comprehensive reform agenda exceedingly long border closure and wiped out tour- that establishes the necessary conditions for rapid ism revenues until mid-2022—and a repeated string employment growth in urban areas and renewed of cyclones wreaked havoc on the service economy, productivity gains in the agricultural sector. destroying as many as a quarter of jobs and slash- ing urban incomes (Chapter 3). At the national level, Paid employment growth is an urgent priority. A a decline in inequality occurred driven by a reduc- weak enabling environment for the private sector tion of welfare in urban rather than an improvement discourages job creation, trapping a large share of in the historically much poorer rural areas. This is the country’s workforce in low-productivity agri- a major concern as it implies that urban areas are culture. Almost 90 percent of employment is infor- 1 However, the national poverty headcount is not statistically significantly different between the two years. 1 2 This trend appears to precede 2012, as the last Poverty Assessment (Osborne et al. 2016) already noted that urban poverty increased between 2010 and 2012 by close to 6 percentage points. 3 Private investment as a share of GDP is well below the average for peer countries and has been declining since 2009. TOC Madagascar Executive summary Poverty and Equity Assessment mal, and most workers are either self-employed ing a pro-growth coalition focused on developing or engaged in household enterprises (Chapter the institutional framework necessary for a robust 3). Informal employment is often precarious and and competitive private sector. Firms and investors marked by low pay and poor job quality, while require a stable, predictable, and impartial public household enterprise labor is mostly unpaid. Mean- administration that provides essential infrastructure while, a small elite group controls a large share of and services while imposing limited and reasonable the economy and fiercely resists competition, which regulatory requirements. Unless the government has reduced private investment to an all-time low can establish these conditions, employment growth of 17 percent of GDP as of 2021. Such “elite capture” will remain insufficient to reduce poverty, and Mad- has created a hostile environment for employment agascar will face another decade of missed oppor- creation, further narrowing the path to prosperity tunities. for urban residents. By building consensus around an ambitious agenda for private-sector growth and With different data sources, we revisit poverty job creation, the government could rapidly improve across its multiple definitions. The new Census employment quality in the urban economy while (2018), Demographic Health Survey (DHS, 2021), encouraging competition and expanding the tax and Multiple Indicator Cluster Survey (MICS, 2018), base. The development of the formal labor market Afrobarometer (2015 & 2018), the Enquête Nation- will become increasingly critical as the urban popu- ale sur les Objectifs Millenaires du Développement lation continues to grow. (ENSOMD 2012) and the 2021/2 Enquete Perma- nent aupres des Menages (EPM) are used in this Rural development will be vital to reduce poverty report to look at trends and deliver a detailed snap- rates and ease pressure on urban labor markets. shot of different dimensions of welfare, exploiting Agriculture provides livelihoods for two-thirds of the geographic granularity of the 2018 Census. Madagascar’s workforce, including a large majority For the analysis, administrative delimitations at the of poor households. The limited use of improved province, regional, district, and communal level will seeds and fertilizers, slow technological uptake be used (as delineated in Map 1). among smallholder farmers, and weak commerciali- zation capacity undermine agricultural productivity, while deteriorating transportation networks inhibit access to input and output markets. As a result, about 80 percent of rice production is not mar- keted but consumed directly by producers (Chap- ter 2). Investment in rural infrastructure—especially the dilapidated road system, including new feeder roads—will be a crucial first step in addressing agri- cultural market failures and reducing rural poverty. Climate resilience is a cross-cutting challenge. Madagascar is highly vulnerable to extreme weather events such as tropical cyclones, heavy rains, droughts, and heatwaves (Chapter 5). On average, natural disasters cost the economy an estimated 1 percent of GDP each year and inflict devastating losses on poor communities, especially in rural areas along the Eastern Coast. Extreme weather events have repeatedly prevented gains in poverty reduc- tion, and the six tropical cyclones that hit Madagas- car in early 2022 reduced per capita consumption among affected households by an estimated 30 percent while destroying valuable productive assets. Good governance is essential to climate resilience, as Madagascar’s public sector will need to manage an expanding range of climate-related shocks in the coming years. Breaking the vicious cycle of underinvestment, slow growth, and weak governance will require build- 2 TOC Madagascar Executive summary Poverty and Equity Assessment Map 1: Madagascar by geographical areas, regions, and provinces Geographical areas Provinces Regions North Antsiranana Center Antananarivo Diana East Toamasina West Mahajanga Sava Southeast Toliara Southwest Fianarantsoa Sofia Boeny Analanjirofo Alaotra- Betsiboka Mangoro Melaky Analamanga Bongolava Atsinanana Itasy Vakinankaratra Amoron’i Mania Menabe Vatovavy- Matsiatra Fitovinany Ambony Ihorombe Atsimo- Andrefana Vatovavy Anosy Androy Source: mapchart.net 3 TOC Madagascar Overview Poverty and Equity Assessment Overview 1. Long-term and recent sustainable development. However, job creation in developments that stifled these sectors is still too slow to trigger large shifts in employment, as private investment is restricted by poverty reduction the poor business environment. Therefore, policies Madagascar is among a few countries that have and strategies prioritizing structural transformation experienced long-term decline in real GDP per are necessary to drive economic growth and devel- capita, without experiencing civil wars (Razafind- opment. rakoto et al., 2020). Short periods of relative sta- bility allowed some gains, but these were unfortu- Weather-related shocks have been recurrent over nately reversed by crises such as political instability the past two decades, including eight floods, five in 2001-02, 2009-12, and the COVID-19 pandemic severe droughts and almost 40 cyclones, includ- in 2020. While real GDP per capita saw modest ing three major ones in the past four years alone.5 growth between 2013-2019, it later fell dramati- These events have had a devastating impact on cally, resulting in an average decline of 2.3 percent rural communities and agriculture-based activ- over the 2012-2022 period. ities, particularly along the Eastern Coast, which are now among the poorest regions of the coun- The slow pace of structural transformation in the try. On average, natural disasters are estimated to economy is evident as agriculture and livestock cost the economy about 1 percent of GDP each year, still represented almost 25 percent of GDP in 2021, and up to 8 percent for once-in-a-century events with much of it being subsistence farming (Figure 1). (World Bank Group, 2022). The poorest households Despite its negative contribution to GDP growth, are especially vulnerable to such shocks, with many agriculture employs about two-thirds of the work- located in areas affected by recent cyclones. force and absorbs more new workers than indus- try and services combined. However, construction, In addition to weather shocks, the COVID-19 pan- extractives, and services (excluding tourism) are demic also hit the economy hard (World Bank growing more rapidly and creating more job oppor- Group, 2022). Export revenues and private invest- tunities, indicating a potential for structural trans- ment have collapsed, leading to a contraction of formation.4 The growth in these sectors suggests income per capita by 9.8 percent. The situation was a potential to shift employment from agriculture compounded by historic droughts in the South, to industry and services, which could help to diver- which left 1.3 million people in a state of acute food sify the economy, boost GDP growth, and promote insecurity. Figure 1: Agriculture represents close to one-fifth of GDP but contributes negatively to growth 30 % GDP in 2013 % GDP in 2021 % GDP growth 2013-2021 Percent (cumulative) 25 20 15 10 5 0 gr ion s s r n ds ry y ce th n an ries es s s rt en nce ile l re t ry n ia pe ise ie en nt nk tr io io po tio st ltu st al iti er xt er l tr od eho s t at ct ra pa dm dri m e a pr du e du ra re til Te at m s an ans ca sh u r ru H ic au u pe su te ist ic Fo U m m d in in us d d un u W d fi t st r in ui in an ns Ed co in T n al e ho m re A Eq iv tio d an Co er et d to d om an ct to oo d th M uc k A an ra e Fo oc c O s es ic tr e le nk t rv el Ex ad st ns ic Te Ba ot Se rv ve Co Tr H Se Li Source: Author’s calculations with data from Ministry of Finance. 4 The service sector contributed 55 percent to GDP growth during the 2013-2019 period, while construction and public works also made a significant contribution as 4 public investments increased. Commerce, the largest service subsector, grew slowly as it remained dominated by small informal companies and constrained by subdued consumer demand and limited incentives for formalization. Industry, including manufacturing and the extractive sector, grew at 8.6 percent and contributed about half of GDP growth during 2013-2019. This expansion was driven mainly by two large mining projects. An assessment of the impact of those mining projects on welfare revealed that in poorer and less educated regions, the opening of a mine increased the average wealth score, while in more developed regions, it decreased the local wealth score and average education level due to the influx of lower skilled migrants (Keller, 2022). 5 Madagascar ranks 12th on the 2000-2019 Global Climate Risk index (2021), better than Bangladesh but slightly worse than Cambodia. The Climate Risk Index (CRI) indicates a level of exposure and vulnerability to extreme events. In the CRI 2021, data from 180 countries were analyzed. See Eckstein, Künzel and Schäfer (2021). TOC Madagascar Overview Poverty and Equity Assessment Figure 2: Between 2012 and 2022 urban consumption fell along the distribution, while rural consumption improved more among the poorest 3 Annualized consumption growth 2 Growth Rate (%) 1 0 -1 -2 -3 1 2 3 4 5 6 7 8 9 10 Decile National Annualized consumption growth rate (percent) Rural Annualized consumption growth rate (percent) Urban Annualized consumption growth rate (percent) Source: World Bank estimates based on 2012 ENSMOD and 2022 EPM data. During the past decade (2012-2022), urban con- with differences in trends across urban and rural sumption declined while rural consumption barely areas (Figure 3). In 2022, 75.2 percent of the national increased. Almost all the urban population suffered population was poor (79.9 percent for rural and 55.5 income losses, in part due to COVID-19, which led 25 percent for urban areas). This is a slight (non-sta- percent of households to experience job loss. Over- tistically significant) increase from the 72.9 percent all consumption growth was positive but low for the estimated for 2012. Rural poverty decreased from bottom 60 percent and negative or close to zero for 80.6 percent in 2012 to 79.9 percent in 2022 (a sta- the top 40 percent at the national level. However, tistically insignificant change). At the same time, the overall consumption of poor Malagasy increased poverty significantly increased in urban areas, from slightly more, as the growth incidence curve (GIC) 42.2 percent in 2012 to 55.5 percent in 2022. The slopes downwards between decile 1-9 (Figure 2). largest increases in urban poverty took place in sec- This pro-poor trend is driven by rural areas, where ondary cities (from 46 to 61.1 percent), while in the the GIC is sloping downwards between decile 2 and capital poverty increased slightly from 33.3 to 34.8 9. On the other hand, in urban areas consumption percent (not a statistically significant change). fell in the past decade (it remained unchanged for the wealthiest decile). The strongest consumption Over the past decade, there has been a slight losses affected deciles 3 to 6, followed by deciles 1 - reduction in national inequality due to the narrow- 6. Altogether, the evolution of consumption is con- ing of the urban/rural wealth gap (Table 1). In con- sistent with a simultaneous reduction in extreme trast to the 2005-2012 period, inequality fell during rural poverty and an overall deterioration of urban the last decade, despite the little variation in pov- living standards. erty nationally. This decline in inequality is the result of a deterioration in welfare along the urban income Accordingly, poverty at the national level increased distribution, which has brought urban incomes slightly (the change is not statistically significant), closer to rural ones. At the same time, urban ine- Figure 3: National and rural poverty have stagnated while urban poverty increased 80.6 79.9 75.2 72.9 55.5 *** 56.7 ** 60.2 51.8 * 52.6 42.2 31 *** 22.1 2012 2022 2012 2022 Extreme poverty (% population) Moderate poverty (% population) Urban Rural Madagascar Source: World Bank estimates based on 2012 ENSMOD and 2022 EPM data. Note: *, **, *** denote significantly different from 2012 numbers at the 10%, 5% and 1% levels, respectively. 5 TOC Madagascar Overview Poverty and Equity Assessment Table 1: Inequality has declined as the urban/rural hunger” reflected in higher-than-average stunting wealth gap narrowed rates despite apparent high food security. 2012 (a) 2022 Multidimensional poverty stood at 69 percent in Gini index 2021, among the highest globally. Madagascar Madagascar 38.2 36.8 is the 8th poorest country in the world in multidi- Urban 36.0 39.4 mensional headcount. Deprivations are highest for Rural 34.6 34.1 clean cooking fuel (69 percent), improved sanita- tion (68 percent), safe housing (62 percent), elec- Notes: (a) Revised from the numbers reported in “Shifting Fortunes” Poverty Assessment, 2016, to ensure data tricity (55 percent), clean drinking water (53 percent) comparability over time. Urban estimate is the weighted and education (50 percent). The multidimensional average of capital city & other urban areas. Inequality poverty headcount ratio dropped from 76 percent measures are the average across 100 imputations. Source: in 2008 to 67 percent in 2018 but then increased EPM 2012, 2022. slightly between 2018 and 2021, when the share quality increased slightly and rural inequality fell of electricity and adequate nutrition depriva- significantly, thanks to a significant reduction in tions increased after the country suffered multiple rural extreme poverty. The narrowing of the gap shocks. The deterioration of nutrition occurred in between rural and urban areas reflects a reduced rural and urban areas, returning close to its 2008 economic advantage of urban over rural areas. level, despite a steady improvement in stunting among children under 5. The latest data suggests that poverty has become less deep and severe at the national level, while the Multiple deprivations explain persistently high opposite occurs in urban areas between 2012 and monetary and multidimensional poverty. First, 2021 (Figure 4).6 The national improvement is driven most people are employed in subsistence low-pro- by a reduction in poverty depth and severity in rural ductivity agriculture, which is uneconomic due to areas. However, urban poverty has become more lack of inputs, infrastructure and favorable insti- severe and deeper, with lower asset ownership at tutions. In this sector, 90 percent of households the household level. This trend in urban poverty are poor. Second, a slow accumulation of human pre-dates the pandemic but was likely worsened by capital has prevented people from escaping pov- it, as well as from rural migration into urban areas. erty through more productive and higher paying Still, even if urban poverty has worsened, rural pov- employment. Child vulnerability is extremely high, erty remains deeper. with high malnutrition among children (39.8 per- cent stunting), child labor and high rates of early Figure 4: Poverty severity increased in urban areas marriages and teenage pregnancies, all of which and fell in rural areas reinforce the intergenerational transmission pov- Poverty severity – poverty gap squared, 2012-2022 erty. Other factors that predict household poverty include household size, illiteracy of the household 25 head, lack of ownership of land or livestock, and 20 absence of electricity, water, sanitation, paved 15 roads, transport, internet, and cellphone networks. 10 Finally, repeated weather shocks, including floods 5 and droughts, destroy infrastructure, crops and 0 livestock, and the recent external shocks (COVID-19 Nati onal Urban Rural and the Ukraine invasion) have affected prices and 2012 2022 urban labor markets, diminished employment and Source: World Bank estimates based on 2012 ENSMOD and earnings opportunities. 2022 EPM data. Significant spatial variation in poverty and ine- Malagasy household heads express a sense of quality persists, with Southern provinces showing constant struggle to make ends meet and lack of vastly more acute poverty and deprivation levels. aspirations for the future, citing multiple causes of Southern regions continue to experience higher poverty in their society. The primary factors identi- levels of poverty and extreme poverty. They also fied by respondents of a focus group study include face the highest rates of multidimensional poverty. the lack of jobs (43.5 percent), inflated cost of living Conversely, inequality tends to be higher in certain (13.1 percent), limited access to land (9.1 percent), low northern regions. Finally, small-area estimations salaries (7.8 percent), and insufficient education (6.2 reveal that some of the fertile areas in the high-pla- percent). The scarcity of employment opportunities teaus of the center suffer the most from “hidden severely hampers the ability of poor households to 6 The poverty gap is a measure of poverty that quantifies the depth of poverty in a population by calculating the average shortfall of the total 6 population's income or consumption from the poverty line. The poverty line is typically set as the minimum amount of income or consumption needed to afford a basic standard of living. The poverty gap indicates how much additional income or consumption would be required for those living below the poverty line to reach it, on average. The poverty gap square is a related measure that adds further weight to those who are further below the poverty line. It is calculated by squaring the difference between the income or consumption of each poor individual and the poverty line, and then adding up these squared differences across all poor individuals in the population. The poverty gap square provides a more nuanced measure of poverty than the poverty gap alone, as it accounts for the severity of poverty experienced by each individual. TOC Madagascar Overview Poverty and Equity Assessment generate sustainable income, while the soaring cost especially during the rainy season. This makes it dif- of living makes it increasingly difficult to afford basic ficult to transport goods and access essential ser- necessities. Limited access to land restricts engage- vices. Unfortunately, the infrastructure quality has ment in productive activities, low salaries leave them been declining over the past decade, resulting in unable to cover essential expenses, and insufficient a ranking of 106 out of 131 countries in the 2020 education undermines their prospects for better Global Innovation Survey (Cornell University et al., employment. These firsthand insights underline the 2020). Only 11.4 percent of the rural population has complex nature of poverty and highlight the prior- access to good quality road networks (World Bank, ities of poor households to address the challenges 2021a). The lack of connectivity is isolating rural they face. populations and limiting their access to economic opportunities and essential services. Large rice price 2. Low agricultural productivity and differences across regions show weak market inte- lack of basic services still trap 8 in gration, mostly due to prohibitively high transport 10 rural people in poverty costs. For most of the 80 percent of the population (rural Access to basic services, such as electricity, water, and some urban) whose main activity is agricul- and sanitation, remains a major challenge. While ture, generating enough income to exit poverty is there has been some progress in improving access beyond their reach. Subsistence farming is prev- to electricity (Map 2), only about 15 percent of alent and explains why 90 percent of agriculture the population is connected to the electricity grid workers are poor. Rice, the main crop in the country, (World Bank, 2023), significantly below the Sub-Sa- accounts for 70 percent of total agricultural pro- haran Africa average of 47 percent, and fewer than duction, 40 percent of cultivated land, and all of 10 percent rural households have electricity. Limited the irrigated land. The average national annual rice access to electricity also affects the use of water yield is approximately 2.5 t/ha, similar to other East pumps, storage facilities, milling, and cold chain African countries, but very low compared with the facilities, which could help decrease post-harvest major rice-producing regions in Asia and the avail- losses. In addition, access to water and sanitation able irrigation. remains low, lagging behind most of the country's peers and other low-income countries. These chal- Agricultural productivity severely lags compara- lenges hinder the country's ability to improve its tors, with both land and labor productivity remain- human capital and economic growth. ing low over the past decade. Production growth in recent years was not driven by increased land Map 2: Electricity coverage has improved along productivity but by farmland expansion through the highlands but remains limited slash and burn. Agricultural labor productivity is 2010 2018 significantly lower in Madagascar than the Sub-Sa- haran African average and has dropped by 31 per- cent since 1991. The low quantity and quality of inputs broadly explain low productivity, including low mechanization (5-10 percent) and low use of agrochemicals and improved seeds (fewer than 10 percent use fertilizers/pesticides and 11.9 percent use improved seed types). Inequitable distribution of land and weak ownership rights exacerbate the problem. At the same time, few agricultural workers have formal education or technical training. Farmers lack access to finance (to purchase better inputs) and insurance, which further block the adoption of more productive technologies. Finally, farmers use inefficient storage methods, as half of unsold pro- duce is either stored in suboptimal conditions (on house roofs) or not stored at all, leading to financial Source: Madagascar Poverty Targeting Indicator Available at losses and food insecurity. https://datanalytics.worldbank.org/mdgPT/ Limited market access has also compromised prof- itability in the agricultural sector. Many rural areas face a major challenge with poor road connectivity, 7 TOC Madagascar Overview Poverty and Equity Assessment 3. Failing labor markets and recent creation never fully recovered. Secondly, the limited shocks explain rising and deeper growth which was achieved was driven by mining urban poverty and public works, which had limited impacts outside of mostly rural mining towns. Market concentra- Urban poverty increased from 42.2 percent in 2012 tion further exacerbated the situation as dominant to 55.5 percent in 2022. This represents a 31.5 per- firms maximized rents and avoided competition. cent increase in a decade. In the capital city of Anta- Market capture in key sectors such as telecommu- nanarivo, poverty increased minimally, from 33.3 to nications, petroleum and agricultural exports ended 34.8 percent, but in secondary cities it rose from up increasing prices and worsening the quality of 46 percent in 2012 to 61.1 percent in 2022. Con- services for consumers. Thirdly, the pandemic in sumption dropped for virtually all households along 2020/21 had a larger impact on urban populations the urban income distribution, particularly those in through economic losses in the services sector. middle deciles. Going further back, urban poverty Finally, slow but continuous migration from rural to increased in 2001-2005, 2010-2012, and 2012- urban areas due to high poverty and high rural fer- 2022, which follow closely the 2002, 2009, and tility also contributed to the problem. 2020 crises. This confirms that urban areas are vul- nerable to political and external crises, whereas rural Between 2008 and 2018, multidimensional living areas are more sensitive to weather-related shocks. standards improved overall, but in urban areas, these standards worsened due to the factors While the country remains rural, urban population mentioned above (Figure 6). The increase in mul- growth in the last decade has been driven by growth tidimensional poverty in urban areas was primarily in the number of poor people.7 In urban areas the caused by declines in access to education and living number of impoverished individuals surged by more conditions indicators, with access to water being than 70 percent over the last decade (Figure 5). the most severely affected. The disproportionate This rise in urban poverty stems from several fac- impact of the pandemic on urban households fur- tors, including higher fertility rates among the poor, ther increased urban poverty. Despite a relatively recent economic downturns and income losses, as low number of COVID-19 cases, the country expe- well as in some cases, the migration of impover- rienced a deep recession, causing a significant con- ished individuals from rural to urban areas. Although traction in GDP and income per capita. The govern- rural areas have higher fertility rates, the number of ment's border closures and restrictions on public poor rural individuals increased less quickly than the gatherings helped control the spread of the virus number of non-poor, in contrast to the urban case. but negatively impacted urban households, particu- larly those dependent on trade, transport, hospital- Growth in urban poverty can be attributed to a ity, as well as informal labor. The measures affected combination of factors. Firstly, the impacts of the livelihoods and the gradual recovery of incomes was 2009-2013 crisis, which mostly affected urban observed only after the reopening of the country's areas, still lingered as private investment and job borders in early 2022. Figure 5: Urban population growth between 2012-2022 is explained by growth in urban poor population Population growth Population growth (%) 30 population poor non-poor 71.0% 25 millions 20 15 43.7% 40.9% 10 36.9% 38% 37% 5 30% 24.9% - population poor non-poor population poor non-poor 0.0% 2012 2022 Urban Rural URBAN RURAL TOTAL Source: World Bank estimates based on 2012 ENSMOD and 2022 EPM data. 7 According to the definition of urban/rural poverty used in the Population and Housing Census 1993 and 2018. 8 TOC Madagascar Overview Poverty and Equity Assessment Figure 6: Living conditions have deteriorated in urban areas while they improved nationally Share of deprivation in urban areas Share of deprivation at the national level 2008 2018 2008 2018 100% 100% 72% 75% 76% 75% 75% 69% 67% 67% 60% 59% 56% 56% 52% 54% 49% 49% 50% 50% 41% 41% 33% 36% 36% 33% 27% 28% 27% 25% 25% 25% 26% 25% 22% 20% 20% 22% 25% 16% 13% 13% 8% 6% 5% 3% 3% 0% 0% lity trition dance ooling ssets ter l ty n ce rtali Nutritio endan chooli ng ssets ter sing ctricity itation g fue l Wa sing ctricity itation g fue mo A Wa Hou orta Nu atten of sch A Hou San Cookin d att rs of s Ele San Cookin dm Ele Chil o o l Chil ool Years Sch Yea Sch Source: World Bank estimates based on 2008 DHS and 2018 MICS data. Low quality employment is pervasive in urban (Psacharopoulos & Patrinos, 2018). Each year of areas. At the national level, over one-third of total schooling is associated with a 4.9 percent increase employment in 2022 was still composed of unpaid in earnings at the national level and 5.6 percent workers either within family enterprises or in sub- in urban areas, against 4.2 percent in rural areas. sistence agriculture. Among women, subsistence Further, regression analysis also shows that, con- farming was the largest employment category trolling for experience and other observable char- (Figure 7). In urban areas, labor markets look slightly acteristics, returns to education are higher for older better but the quality is overall still low. According cohorts, indicating that there is likely a fall in the to Census data, employment rates in urban areas education premium (possibly also combined with a are below those of rural areas (56.6 percent vs. 76.6 compounded effect of education and experience). percent of the population 15-59 years-old), which This effect is particularly strong among women. The indicates lower pressure to participate in economic decline in returns to education can be the result of activity but also fewer opportunities. Moreover, only a decline in quality (whereby more years of school- 32 percent of urban jobs are relatively stable wage ing are needed to achieve similar productivity), or jobs (Figure 7), which is still low but much higher an oversupply of skills, which would be the case if than in rural areas (6 percent) and only 12 percent employment creation is slow, such as has been the of employed workers have social security coverage. case in Madagascar. Returns to education have been declining in recent Lack of opportunities and low aspirations are years. Like most countries, returns to education in intertwined factors contributing to urban pov- Madagascar are generally higher for individuals with erty. Insufficient investment in education, health- more education and those who live in urban areas care and urban infrastructure limits human capital Figure 7: Self-employment and family work are the largest employment categories Employment distribution Employer Public wage Private wage Independent Piece worker Trainee Family worker Other 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Male Female Male Female Urban Rural Source: RGPH (INSTAT, 2021a). 9 TOC Madagascar Overview Poverty and Equity Assessment development and economic activity and increases ment, reduced opportunities for income generation, poverty traps. Low real income reduces access to and thus perpetuate a cycle of poverty. Data from finance and investment opportunities. The domi- the Multiple Indicator Cluster Survey (MICS 2018) nance of the elite in political and economic spheres reveals that women in the poorest quintile have creates a sense of hopelessness among the urban three times higher fertility rates (6.6 children per population, leading to diminished aspirations. Low woman) than those in the richest quintile (2.7 chil- education and weak social cohesion further hinder dren per woman) and 48 percent of teenagers aged individuals' ability to demand government services 15-19 in the poorest quintile have already been and participate in their communities, exacerbating pregnant, against 12 percent of those in the richest the poverty trap. According to focus group discus- quintile. Although Madagascar has made some pro- sions, there is a widespread belief that the system is gress in enhancing access to healthcare, education, designed to maintain the status quo and preserve and family planning, further investment is necessary resource distribution for the benefit of the elite. This to overcome challenges such as limited resources, has led to a feeling of hopelessness and a sense that cultural and social barriers, and lack of awareness. individuals are powerless to change their situation (Mulangu, 2023). Consequently, people have low Madagascar stands out with one of the highest expectations of government services and limited child marriage rates, a pressing issue that demands aspirations for productive work and high income. attention. Shockingly, 40 percent of women aged This implies that people's ambitions are being lim- 20-24 were married before the age of 18. The prev- ited by the perception of limited opportunities and alence of child marriages is most pronounced in inequality in society. the poorest regions, where vulnerable communities struggle to escape the cycle of poverty. Although 4. Progress in long and short-term higher rates of child marriage are found in rural poverty drivers is mixed areas alarmingly high rates are found in urban areas as well. There is a strong correlation between the According to the World Bank's Human Capital prevalence of child marriage and other detrimental Index (HCI) and other data sources, Madagascar practices like child labor and high numbers of out- faces significant challenges in terms of human cap- of-school children. The districts with the highest ital development. A child born in Madagascar just occurrence of child marriage are also the ones with before the pandemic is projected to be 39 percent high illiteracy rates, highlighting the interconnec- as productive as they could be with complete edu- tion between these issues and the urgent need for cation and good health. This places Madagascar effective strategies to combat them. If left unad- slightly below the average for Sub-Saharan Africa dressed, intergenerational decision-making perpet- and just above the average for Low Income Coun- uates the vicious cycle of poverty, ensnaring future tries (LICs). However, when considering recent data generations. on school enrollment and child survival, the adjusted HCI drops to 34 percent. Although child survival Child mortality and infant mortality have been rates and school enrollment are relatively favora- decreasing but remain above those of peer coun- ble, educational outcomes are low and declining. tries and the overall LIC average. While child mor- Test scores indicate that students in Madagascar tality levels have improved compared to the 2000s, perform below average compared to other regions. the gains have been smaller than in countries like Additionally, over 90 percent of 10-year-old children Cambodia, Rwanda, and Uganda. These coun- suffer from "learning poverty," defined as being able tries started with similar or higher levels but have to read and understand a simple text. There is also a achieved substantially lower child mortality rates significant disparity in HCI between the richest and than Madagascar. Infant mortality, specifically poorest children, with the gap in future productivity deaths occurring before the first year of life, remains being larger than the global average. These findings a significant issue across peer countries, although it highlight the pressing need for investment in edu- is still lower in Madagascar compared to the overall cation and health to improve Madagascar's human LIC average. capital. The prevalence of stunting remains extremely high High fertility rates in low-income households con- at 39 percent of children under 5, although it has tribute to chronic poverty by overburdening house- decreased over the past 10 years. Regions where holds, hindering educational attainment, and limit- health and nutrition projects were implemented ing income generation opportunities. High fertility have experienced a faster decline in stunting, indi- rates in low-income households can lead to over- cating the effectiveness of donor-financed pro- burdened households, lower educational attain- grams. Regions where health and nutrition projects 10 TOC Madagascar Overview Poverty and Equity Assessment Figure 8: Education attainment has increased but quality remains low Education of the population age 10+ Basic reading skills (% class) 1993 2018 Male Female 47.0 44.7 90 80 34.0 70 31.0 60 50 20.7 18.0 40 30 20 3.6 2.0 10 0 No education Primary Secondary Tertiary CP2 (G1) CP2-CE CE (G3) CM1 (G4) CM2 (G5) Junior (lower/upper) (G2) High School Source: Source: Madagascar, INSTAT (RGPH, 2018) and MICS (2018). were implemented experienced a faster decline in comes such as basic reading skills tend to increase stunting than the country as a whole, indicating that as children advance in their grade, by grade 5 fewer donor-financed health and nutrition programs were than 60 percent of children enrolled in school have well-targeted and effective. Nonetheless, Malagasy acquired basic skills in reading (Figure 8) and fewer households often struggle to provide the nutrition, than 20 percent in math. Child labor remains a sig- healthcare and hygiene conditions to allow children nificant challenge, with one-third of children aged to grow, further emphasizing the importance of 5-11 engaged in labor and over 60 percent of chil- continued investment in health and nutrition. dren aged 12-14 employed. Progress in education attainment has been limited, 5. Repeated shocks wipe out with incremental advancements in secondary edu- income gains and asset cation attainment from 18 percent to 31 percent accumulation from 1993 to 2018. Conversely, the proportion of individuals with no schooling has reduced from 34 The country is particularly vulnerable to weather percent to 21 percent. Nevertheless, the comple- shocks, and the number of people affected tion of primary education remains low, as only 47 increased significantly in recent years. For cen- percent of children complete primary schooling on turies, Madagascar has been ravaged by cyclones time.8 Moreover, teacher absenteeism and natural and droughts. On average, weather shocks are disasters disrupt attendance for over 80 percent of estimated to cost the economy about 1 percent of children in a given year, and although learning out- GDP each year. Poor households are especially vul- Figure 9: Poor households were disproportionately hit by cyclones in 2022 Number of cyclones in 2022 a ecting households by welfare status 100% 90% 80% 70% 60% 2 or more 50% 1 40% 0 30% 20% 10% 0% 1 2 3 4 5 6 7 8 9 10 All Consumption decile Source: World Bank estimates based on 2022 EPM data. 8 The Population Census reports that in 2018, the completion rate at the primary level was 46.9 percent overall, 44.1 percent for boys and 49.8 11 percent for girls. This relatively low rate (half of the children did not complete the cycle) could be due to late entry, high dropout rate, late completion, or a high repetition rate (INSTAT, 2021b). TOC Madagascar Overview Poverty and Equity Assessment nerable and particularly ill-equipped to cope with relatives and friends, or using their savings. Out of these shocks. Drought, irregular rains, and tropical the 68 percent of respondents that had experi- storms are the most frequent and severe systemic enced an increase in the price of major food items shocks faced by households. Each had an impact on usually consumed by their household, 46 percent household wealth (income and assets), agricultural reduced their consumption to cope with the shock. output (crops and livestock) and food (purchases and stock). Weather shocks tend to affect poorer Public transfers and subsidies have remained at households more frequently than richer ones. For less than 3.5 percent of GDP, which is similar to instance, in 2022, over 60 percent of households peer countries but extremely low compared to the in the richest quintile were not hit by any cyclones, extent of poverty in the country. Between 2020 against 9.7 percent of households in the poorest and 2021, the main safety nets (Figure 10) covered quintile (Figure 9). At the same time, about 30 per- close to 300,000 households, equivalent to 5.3 cent of households in the poorest decile suffered percent of the national population (9.3 percent of more than three cyclones, versus only 3 percent of urban and 4 percent of rural population). Despite household across the rest of the distribution. The its low coverage, safety nets appear to be relatively observed pattern indicates that wealthier house- well targeted, as they cover a larger percentage of holds tend to live in better protected areas from extreme poor and moderate poor than non-poor cyclones and its secondary consequences. In con- populations (5.4 percent, 4.4 percent and 2.6 per- trast, poor households are more likely to live in areas cent, respectively). Targeting efficiency seems to especially vulnerable to cyclones—usually remote be more accurate in rural areas, which is consistent areas with minimal access to infrastructure and both with the higher difficulty of targeting the urban poor drainage. poor and with the reference period (during the pan- demic) when urban coverage was a priority. Bene- Apart from these systemic shocks which affect fits received by extreme poor households represent whole communities and sometimes the entire about one-quarter of their overall consumption, nation, households also suffer idiosyncratic shocks which reflects the importance of these transfers for which also affect poverty. The loss of jobs and sal- the extreme poor and suggests that more effective aries had the most adverse effect on households. targeting could raise consumption among more In the absence of employment, people sell assets, extreme poor households. Finally, despite the edu- crops, and livestock in order to meet their needs. cational conditionality on the cash transfers, there is Many households often lack an immediate coping little evidence that among cash transfer beneficiar- strategy, and when they have one it frequently con- ies, children are improving their school attendance sists of buying cheaper food, soliciting help from rates and health outcomes. Figure 10: Coverage of safety nets is extremely low in rural areas Safety net coverage, 2020-21 Safety net coverage (% population), 2020-21 1,335,994 12.0 10.0 8.0 Rural 6.0 Urban 4.0 291,930 2.0 0.0 Urban Rural National Total Extrem e poor Households Individuals Moderate poor Non-poor Note: Safety nets include the main cash transfer program, cash for work and pregnant women/infant care program. Program coverage is the percentage of population in each group that receives the transfer. Extreme poor are households below the food poverty line, moderate poor are households between the food and the moderate poverty line. Coverage data might differ from administrative data due to survey design. Source: EPM 2022. 12 TOC Madagascar Overview Poverty and Equity Assessment 6. Policies to break the curse of low Invest in and protect the quality of human cap- growth and high poverty ital. Human capital accumulation has been slow due to malnutrition, lack of access to education, The evidence on the drivers of poverty, both urban vulnerability to climatic shocks and high food inse- and rural, point to a systemic lack of conditions to curity. To address this issue, the country needs to improve labor and investment returns. This obeys improve the education system by enforcing mer- to a low level of infrastructure and human capital it-based recruitment for teachers as well as aligning and to a lack of basic services that can sustain pri- the academic calendar with the agricultural season. vate sector growth and employment creation. As a It is also important to improve teacher pay in rural result, poverty will not fall sustainably unless there areas (they are paid by parents), improve health and are conditions in place for broad based and sus- nutrition services, increase safety net coverage (in tained growth. An overview of these recommenda- a fiscally sustainable way), improve targeting, and tions is briefly discussed here, and in detail in Chap- consider strengthening conditionality on education ter 6. and health. Promote competition, market contestability, and In conclusion, addressing poverty in Madagascar improvement of the business climate: Madagas- requires a multi-faceted approach that focuses car needs to improve its business environment by on improving the conditions for broad-based and removing barriers to entry in key sectors of the sustained growth. This means promoting competi- economy. This can be achieved by improving and tion, market contestability, and improving the busi- enforcing the Competition Law, creating a Compe- ness climate, as well as investing in infrastructure tition Council and regulatory bodies that are inde- to improve connectivity and access to energy and pendent, simplifying and digitizing administrative digital services. Boosting agricultural productivity rules and procedures. and protecting against reoccurring risks with index insurance and investing in and protecting the qual- Improve connectivity and access to energy and ity of human capital are also crucial steps. By imple- digital services: Madagascar needs to invest in its menting these recommendations, Madagascar can infrastructure to improve the quality of its road create an environment that enables private sector network and increase access to electricity and dig- growth, employment creation, and poverty reduc- ital services. This can be achieved by improving tion that will benefit the entire population. Detailed the public investment execution process, increas- strategies for implementing these recommenda- ing resources for road maintenance, strengthen- tions are discussed in Chapter 6.  ing logistical support infrastructure, and improving rail and air competitiveness. It is also important to ensure the financial sustainability of the national electricity company (JIRAMA) and promote open and competitive private investments in digital infra- structure by reducing licensing costs. Boost agricultural productivity and protect against reoccurring risks. Rice productivity is in decline, and approximately just 20 percent of domestically grown rice is marketed. To address this issue, the country needs to deepen current investments in rural connectivity, expand rice-farming areas, and strengthen land tenure security. It is also impor- tant to reduce trade distortions (customs duties exemptions/export restrictions) and encourage exports of high-value products like vacuum-packed vanilla, lychees, ylang-ylang, and livestock. It is also important to protect rice farmers against the reoccurring risks of climate shocks with insurance. However, these tools must be used in a market friendly manner by indirectly protecting the farm- ers through an aggregator to reduce the often-high cost associated with supplying insurance to small- holder farmers. 13 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Chapter 1 Two decades of poverty stagnation against a modest growth performance Key findings In 2022, 75.2 percent of the national population tricity (55 percent), clean drinking water (53 percent) was poor (79.9 percent for rural and 55.5 percent for and education (50 percent). The multidimensional urban areas). This is a slight (non-statistically signif- poverty headcount ratio dropped from 76 percent icant) increase from the 72.9 percent estimated for in 2008 to 67 percent in 2018 but then increased 2012. Rural poverty decreased from 80.6 percent in slightly between 2018 and 2021, when the share 2012 to 79.9 percent in 2022 (a statistically insignif- of electricity and adequate nutrition depriva- icant change), while average consumption among tions increased after the country suffered multiple the rural poor increased by 1-percent per year.9 At shocks. The deterioration of nutrition occurred in the same time, poverty significantly increased in rural and urban areas, returning close to its 2008 urban areas, from 42.2 percent in 2012 to 55.5 per- level, despite a steady improvement in stunting cent in 2022. The largest increases in urban poverty among children under 5. took place in secondary cities (from 46 to 61.1 per- cent), while in the capital poverty increased slightly Multiple deprivations explain persistently high from 33.3 to 34.8 percent (not a statistically signif- monetary and multidimensional poverty. First, most icant change). people are employed in subsistence low-produc- tivity agriculture, which is uneconomic due to lack National inequality declined over the past decade of inputs, infrastructure and favorable institutions. as urban welfare deteriorated and rural welfare In this sector, 90 percent of households are poor. slightly improved. In contrast to the 2005-2012 Second, a slow accumulation of human capital has period, inequality fell during the last decade, despite prevented people from escaping poverty through the little variation in poverty nationally. This decline more productive and higher paying employment. in inequality is the result of a deterioration in wel- Child vulnerability is extremely high, with high mal- fare along the urban income distribution, which has nutrition among children (39.8 percent stunting), brought urban incomes closer to rural ones. At the child labor and high rates of early marriages and same time, urban inequality increased slightly, and teenage pregnancies, all of which reinforce the rural inequality fell significantly, thanks to a signifi- intergenerational transmission poverty. Other fac- cant reduction in rural extreme poverty. tors that predict household poverty include house- hold size, illiteracy of the household head, lack Significant spatial variation in poverty and ine- of ownership of land or livestock, and absence of quality persists, with Southern provinces showing electricity, water, sanitation, paved roads, transport, vastly more acute poverty and deprivation levels. internet, and cellphone networks. Finally, repeated Southern regions continue to bear a higher propor- weather shocks, including floods and droughts, tion of poor and extreme poor households. They also destroy infrastructure, crops and livestock, and the face the highest rates of multidimensional poverty. recent external shocks (COVID-19 and the Ukraine Conversely, inequality tends to be higher in certain invasion) have affected prices and urban labor mar- northern regions. Finally, small-area estimations kets, diminished employment and earnings oppor- reveal that some areas in the high-plateaus suffer tunities. from “hidden hunger” reflected in higher-than-av- erage stunting rates despite apparent high food Malagasy household heads express a sense of security. constant struggle to make ends meet and lack of aspirations for the future, citing multiple causes of Multidimensional poverty stood at 69 percent in poverty in their society. The primary factors iden- 2021, among the highest globally. Madagascar tified by respondents include the lack of jobs (43.5 is the 8th poorest country in the world in multidi- percent), inflated cost of living (13.1 percent), limited mensional headcount. Deprivations are highest for access to land (9.1 percent), low salaries (7.8 percent), clean cooking fuel (69 percent), improved sanita- and insufficient education (6.2 percent). The scarcity tion (68 percent), safe housing (62 percent), elec- of employment opportunities severely hampers 9 Approximately 80 percent of Madagascar’s population dwells in rural areas. 14 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment their ability to generate sustainable income, while lytical tools including survey-to-survey imputation the soaring cost of living makes it increasingly dif- to address lack of comparability of consumption ficult to afford basic necessities. Limited access to data between 2012 and 2022 (Jarotschkin, Vincent, land restricts engagement in productive activities, et al., 2023; Jarotschkin, Yoshida, et al., 2023).11 low salaries leave them unable to cover essential Latest small-area poverty estimation methods to expenses, and insufficient education undermines estimate poverty levels at the commune level are their prospects for better employment. These first- also presented. All of the above provide a more hand insights underline the complex nature of pov- nuanced and granular understanding of the spatial erty and highlight the priorities of poor households and temporal trends in poverty and inequality in the to address the challenges they face. country, enabling the prioritization of interventions areas and programs to alleviate poverty and ine- 1. Poverty against broader political quality. and economic trends The analysis suggests that while poverty remains Since the start of the 21st century, significant eco- predominantly a rural phenomenon, urban areas nomic and political shocks have trapped Madagas- outside of the capital city have also been slipping car’s population in poverty.10 Madagascar contin- quickly. Urban poverty, outside the capital city, has ues to have one of the highest poverty rates in the been on the rise in recent years. In line with declin- world, particularly in rural areas, where 80 percent ing average consumption levels, which can be of the population resides. Persistent high poverty attributed to a range of factors including political has deteriorated not only the economy but is also instability, low private investment, the COVID-19 risking its very social fabric, with malnutrition, poor pandemic and rising food and fuel prices, limited health, low education attainment, and social cohe- employment opportunities, and inadequate social sion being worse in areas with high poverty. safety nets. The increase in urban poverty is particu- larly concerning, with many people migrating from The past decade was volatile. While the political rural areas to cities in search of better economic instability of 2009–2013 was followed by a rela- opportunities, a trend that will only last with the tively calmer period, the economy swung between immensely high rural poverty rates. Thus, tackling a surge in mining investments, the reengagement of poverty in both rural and urban areas will require development partners and increased foreign aid on different approaches as the analysis speaks to, but the positive side, and multitude of geopolitical (fol- a focus on both is paramount for achieving inclusive lowing Russia-Ukraine-war-induced price shocks), and sustainable economic growth in Madagascar. climatic (cyclones and droughts), and health crisis (COVID-19 pandemic) shocks on the negative side. 2. While the number of poor At the same time, the lack of a detailed survey on increased by 50 percent in 10 monetary welfare did not allow to monitor poverty years, inequality decreased for more closely, but evidence on living conditions, from the wrong reason the Census of 2018, and MICS 2018 showed little progress regarding human capital, access to basic In the last two decades, poverty in Madagascar services, and employment conditions. The analy- has fluctuated above 70% of the population. The sis presented in this chapter provides an update to updated national poverty line is estimated at MGA complement this evidence, assessing the evolution 1,477,565 ($335.81)/person/year. This poverty line of poverty and inequality across time and space – represents the cost of basic needs: covering the using the 2022 EPM & 2012 ENSOMD - linking it to cost of a basic consumption basket to reach 2133 the larger macro trends as well as (lack of progress kcal intake per day, and a small amount of additional on multidimensional dimensions of welfare using a funds for non-food items, such as shelter. 75.2 per- variety of sources, including DHS 2008/9, Housing cent of Malagasy do not see their cost of basic needs and Population Census of 2018, MICS 2018, DHS covered, meaning they could not afford this basket 2021. in 2022 and are therefore considered poor. This is higher than the previous estimate of 72.9 percent in The objective of this chapter is to provide a pro- 2012 (Table 2), 71.7 percent in 2010, 73.2 percent in file of poverty and inequality in Madagascar and an 2005, and 70.8 percent in 2001.12 In the capital city, indication of recent trends. This report uses various there is no difference in the poverty rate between poverty indicators such as poverty incidence, depth 2012 and 2022, while in other urban areas, driven by and severity of poverty, multidimensional poverty, secondary cities, poverty has been increasing nota- and subjective poverty, concepts of vulnerability, bly, from 42.2 percent (based on imputed expendi- the human opportunity index, and advanced ana- tures) in 2012 to 55.5 percent in 2022. Ever so 10 Osborne et al. 2016. 11 Consumption data from the ENSOMD series until 2012 are not comparable with those of EPM 2022. The EPM 2022 consumption data contain 15 an updated survey instrument collecting far more food consumption items than the ENSOMD series and recording consumptions in local non- standard units for the first time. To overcome the incomparability in consumption data from the ENSOMD 2012 and EPM 2022 surveys, a survey- to-survey methodology was used to impute household expenditures data in the ENSOMD 2022 comparable to the EPM 2022 data. (Jarotschkin, Vincent, et al., 2023; Jarotschkin, Yoshida, et al., 2023.) The imputed expenditure data are used to estimate poverty rates for 2012 in table 2. 12 Poverty lines were set at MGA 525,000 person/year in 2012, MGA 381,791 person/year in 2010, MGA 289,169 person/year in 2005, and MGA 192,733 person/year in 2001. TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Table 2: Poverty incidence steadily increased in holds. Then, from 2010 to 2012, households across the last 20 years the distribution suffered a decline in real consump- Poverty headcount 2001 tion, with minimal gains confined to a handful of 2005 201013 201214 2022 ratio middle-class families. As a result, between 2001 to National (%) 70.8 73.2 71.7 72.9 75.2 2012, real consumption increased overall for house- Rural (%) 77.7 79.6 80.1 80.6 79.9 holds in the bottom half of the consumption dis- All urban (%) 34.1 40.8 29.8 42.2 55.5 tribution and declined for wealthier and predom- Capital city (%) 33.3 34.8 inantly urban households. The latter group was Other urban without 46.0 61.1 especially affected by macroeconomic instability capital city (%) due to political unrest between 2009 and 2012, the Source: World Bank analysis based on ENSMOD, 2001, 2005, effects of which persisted into the following decade. 2010, 2012; and EPM, 2022. Notes: Poverty line estimated from 2010 EPM survey and adjusted for inflation in each year. The poor urban population grew significantly between 2012 and 2022. Although the absolute minorly, poverty decreased in rural areas during the number of rural poor in Madagascar remains six same reference period. Overall, the period between times as large as that of the urban poor, the poor 2012 to 2022 has been marked by very limited pro- urban population rose twice as fast as the poor rural gress in rural areas, keeping high poverty prevalence one during this time (Figure 11). The increase in the a continued problem while underlining the need for number of urban poor is explained by higher fertility a renewed focus on urban areas outside of the cap- rates among the poor, economic shocks and income ital city where poverty has been importantly rising losses, and – in some cases – the migration of rural (Figure 11). poor to the cities. Meanwhile, in rural areas, the non- poor population grew slightly faster than the poor Average consumption levels among the extreme population. poor rose from 2001 to 2005. Consumption growth was positive for households in the bottom 40 per- Considering substantial population growth in the cent of the consumption distribution during this past decade, the number of poor has grown sub- time, while tapering off for the middle class. Con- stantially more than the poverty rate. The recent versely, the average consumption rate declined for Population Census was a big reveal that the pop- the top 60 percent of the distribution – indicating ulation had grown by a lot more than initially pre- that wealthier households, mainly based in urban sumed. De facto, between the two censuses (1993 areas, reduced their consumption during this period. and 2018), the population had doubled. At a 72.9 This finding is consistent with the observed increase percent poverty rate as of 2012, 15.4 million of in urban poverty between 2001 and 2005. people were considered poor whereas in 2022, the number of poor now towers at 21.1 million – 50 Urban poverty dropped between 2005 and 2010 percent higher. This brings home that for a coun- but deteriorated thereafter. During this time, con- try growing as fast as Madagascar, poverty reduc- sumption levels increased in households above the tion strategies need to be bold to have a chance of 60th percentile and declined among poorer house- reducing the number of poor in the medium term. Figure 11: Urban poor population grew significantly in 2012-2022, but the poor remain predominantly rural Population growth rate (%) Absolute Value population poor non-poor 30 71.0% 25 millions 20 15 10 43.7% 40.9% 5 36.9% 38% - 37% 30% 24.9% population population poor non-poor poor non-poor 0.0% 2012 2022 Urban Rural URBAN RURAL TOTAL Source: World Bank analysis based on 2012 and 2022 EPM data. 13 Pre-2012 numbers are not comparable with post-2012 numbers. 16 14 Revised from the numbers reported in Osborne et al. 2016, to ensure data comparability over time (Jarotschkin, Vincent, et al., 2023). Urban estimate is the weighted average of capital city and other urban areas. TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Among its peers, Madagascar remains one of the economic fallouts have affected Madagascar in the poorest countries. Madagascar’s poverty rate past two decades. And most recently, in addition measured at the international poverty line of $2.15 to home-made crises, Madagascar has been hit by (in 2017 PPPs) per day per capita is 50.9 percent.15 different exogenous crisis, ranging from world-wide As shown in Figure 12, Madagascar remains poorer pandemic (COVID-19), geo-political rifts, and cli- than the neighboring countries and an international mate (droughts and cyclones). All of the above have comparator, like Bangladesh. However, there are a made a dent into any modest progress few caveats on this international poverty rate. First, since the consumption data of the EPM 2022 are The pandemic had an adverse effect on house- not comparable to those of the ENSOMD 2012, holds headed by salaried employees or business this number should not be compared to the pov- owners. The COVID-19 pandemic induced a GDP erty rate previously reported for 2012.16 Second, contraction of 7.1 percent year-on-year, which was the 2022 number should be seen as preliminary followed by an increased price of rice, the staple because the PPP for Madagascar is currently under food that accounts for more than 50 percent of the review and the international poverty estimate could caloric intake of the average Malagasy and a series change. Third, if the current 2017 PPP conversion of disasters including a major drought in the south. rate is used, the international poverty line is MGA According to the World Bank (2021), “…the COVID- 995,932.8 per person per year, covering just about 19 shock reversed more than a decade of modest two thirds of the national poverty line that was esti- gains in poverty reduction.” The World Bank’s High mated to cover the cost of basic needs (consisting Frequency Phone Survey shows that 77 percent of of 2133 kcal per day and small amount of additional Malagasy households experienced a drop in their funds for non-food items). As a result, the interna- business income in June 2020 (with similar percent- tional poverty line would fall substantially short of ages across rural and urban areas – 79 percent and the WHO-recommended 2133 kcal daily food intake. 73 percent, respectively). By this point, 8 percent Therefore, beyond international comparisons, the of the working Malagasy population had lost their poverty estimates measured at the national poverty jobs. The loss in employment and income translated line are more appropriate as a Subdued economic growth with into decreased consumption measure for tracking poverty frequent crises flattened and reversed and a 16 percent increase in over time and policy making short episodes of progress in poverty food insecurity across the in Madagascar. reduction with differential impacts in population. rural and urban areas Subdued economic growth with frequent crises More recently, high inflation prompted by Rus- flattened and reversed short episodes of poverty sia’s invasion of Ukraine, is also estimated to have progress. First, economic growth in Madagascar reduced consumption. The invasion that began has been too slow to create opportunities for its in February 2022 interrupted the post-pandemic people. While economic growth averaged 3.5 per- economic recovery by triggering increases in the cent per annum in the past decade, it was barely prices of food, fuel, and fertilizers. The latter two are surpassed population growth. In addition, different farming inputs and their inflated prices increase the cost of food production for rural farming house- Figure 12: Madagascar remains among the holds, which ultimately leads to food inflation. poorest countries globally Urban households, which are less involved in sub- 60 sistence farming and tend to buy most of their food, 50.9* 52 are especially affected. Inflation has reached double 50 44.9 digits at 11.4 percent today and as a result urban 42.2 poverty is expected to rise further. 40 30 Inclement weather exacerbated poverty trends and disrupted economic growth. Madagascar’s 20 geography makes it vulnerable to climate shocks, 13.5 and the past decade has been riddled with natu- 10 ral disasters. Between 2015 and 2022, the country experienced two severe droughts that plunged 1.3 0 Bangladesh Madagascar Rwanda Tanzania Uganda million people into food insecurity and triggered (2016) (2022) (2016) (2018) (2019) rural urban migrations. In addition, the country was Source: World Development Indicators (last updated: hit by two to three cyclones per year. This caused 16/09/2022) & EPM2022. repeated flooding, made 2.4 million people poor, * Subject to change after forthcoming review of PPP. and resulted in losses equivalent to 6 percent of 15 Purchasing Power Parity. 17 16 The sources of non-comparability are discussed in the footnote 2. TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment GDP, leading affected households to reduce their Figure 13: Poverty Gap and Poverty Gap square consumption by more than 30 percent (Keller & have declined Mulangu, 2023). a. Poverty depth – poverty gap, 2001 - 2022 50 Extreme poverty and severity of poverty 40 slightly improved in rural areas but 30 worsened in urban areas 20 10 Some improvements in rural areas are noted, yet 0 ending extreme poverty remains a tall order with 2001 2005 2010 2012 2022 much of the population living below the food pov- erty line, and urban populations slipping quickly. Nati onal Rural Urban At the national level, extreme poverty has remained b. Poverty severity – poverty gap squared, 2012 - 2022 quasi-unchanged, with more than 1 in 2 Malagasies Urban living in below the food poverty line, which repre- sents the costs of a basic food consumption basket Rural covering 2,133 kcal per person per day or MGA 1,005,974 (Table 3). In rural areas, some improve- Nati onal ments have been noted, with an almost four-per- centage point decrease. Yet, 56.7 percent of Mala- 0 10 20 30 40 gasies in rural areas continue not being able to fulfill 2012 2022 their basic daily caloric needs. Concurrently, in urban areas, 3 in 10 Malagasies live below the extreme to the poverty line has been 37 percent. The poverty poverty line as of 2022 but marking a dramatic slip gap square index (poverty severity) is more sensitive of roughly 1 in 10 urban households below the food to incomes of those furthest away from the pov- poverty line. Importantly, extreme poverty remained erty line. And poverty severity has been decreas- the same in the capital city, with negative changes ing at the national level, driven predominantly by in urban areas attributable to populations residing decreases in rural areas (Figure 13b). Simultaneously, in the country’s other big cities, but also its second- in urban areas poverty severity has been picking up. ary cities. In rural areas, poverty has been stagnant and high, as has been poverty depth. Yet, poverty severity in Table 3: Evolution of extreme urban poverty is rural areas has been decreasing. While stable over most significant in secondary cities time, the alarming levels of poverty in rural areas Capital Other All command attention as slipping much further is National Rural city urban urban beyond tragic. Simultaneously, the improvements 2012 17 52.6% 15.6% 24.9% 22.1% 60.2% among the poorest of the poor have been a wel- 2022 51.8% 13.3% 35.7% 31.0% 56.7% come development. Concurrently, poverty in urban difference 0.9% 2.3% -10.8% -8.8% 3.5% areas has been increasing, as have its depth and significance *** *** *** severity, imposing renewed focus on urban areas Source: World Bank analysis based on 2012 ENSMOD and particularly outside the capital city for interventions 2022 EPM data. to stop this fast-paced negative trend. Driven by improvements in rural areas, poverty depth has remained unchanged, yet poverty sever- Inequality trends: regional variations over ity has decreased nationally, while poverty depth the past 20 years suggest reducing rural and severity have been sharply increasing in urban poverty is key areas. The poverty gap reflects the intensity of pov- erty in a country, and at national level, the depth of Consumption growth and contraction have been poverty has not changed. The depth of poverty in spatially unequal, following a reverse and repeat urban areas has been alarmingly increasing in the pattern during the past 20 years. Between 2001 past decade: while in 2012, the consumption of and 2005, consumption growth was positive for the average poor urban household experienced a households in the bottom 40 percent of the con- shortfall of 14 percent of the poverty line (Figure sumption distribution during this time, while taper- 13a). This shortfall has now almost doubled in 2022. ing off for the middle class. Conversely, growth in In rural areas, the depth of poverty has decreased the average consumption rate was negative for the slightly but continues to be almost twice as high as top 60 percent of the distribution – indicating that in urban areas overall. As of 2022, the shortfall of wealthier households, mainly based in urban areas, average consumption in rural areas in comparison diminished their consumption during this period. 17 Revised from the numbers reported in Osborne et al. 2016, to ensure data comparability over time. Urban estimate is the weighted average of 18 capital city and other urban areas. TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Figure 14: Consumption growth was higher among the poorest households between 2001-2022 2001 to 2012 2012 to 2022 10 1.5 8 1 0.5 6 0 4 Growth rate, % -0.5 2 -1 0 -1.5 –2 –2 –4 -2.5 –6 –3 1 2 3 4 5 6 7 8 9 10 –8 Decile of consumption –10 National Annualized consumption growth rate (percent) 0 20 40 60 80 100 Rural Annualized consumption growth rate (percent) Expenditure percentile Urban Annualized consumption growth rate (percent) Source: World Bank analysis based on ENSMOD, 2001, 2005, 2010, 2012; and EPM, 2022. Table 4: National inequality slightly declined over the period 2012-2022 (Gini index and shares of total consumption at the extremes of the consumption distribution in %, 2005–22) National Rural Urban Lowest Top Lowest Lowest Top Gini Gini Top quintile Gini quintile quintile quintile quintile quintile 2005 38.9 6.9 46.6 35.4 7.6 43.5 39.2 6.5 45.9 2010 42.7 6.0 49.7 37.9 6.9 45.4 38.6 6.6 45.4 2012 41.0 5.9 47.6 37.3 6.5 44.3 38.4 6.3 45.1 2022 36.7 7.2 44.6 33.9 6.4 47.2 39.9 6.4 47.2 Source: World Bank estimates based on 2005, 2010, 2012 ENSMOD and 2022 EPM data. Between 2005 and 2010 consumption levels by region. Within-group inequality refers to the ine- increased among households above the 60th per- quality that exists within either urban or rural areas. centile and declined among poorer households. Yet In this case, the within-group inequality for either following political unrest between 2009 and 2012 – urban or rural areas was 22. This suggests that there with persistent effect in the years after- consump- is a relatively high level of inequality within either tion declined for wealthier and predominantly urban urban or rural areas. The between-group inequality households and increased overall for households in for urban and rural areas was 8.3. This indicates that the bottom half of the consumption distribution there is a moderate level of inequality between urban overall (Figure 14). Between 2012 and 2020, a sim- and rural areas, with urban areas having higher levels ilar pattern transpired, particularly heightened by of income than rural areas. The overlap component exogenous crisis but also homemade ones. of 6.4 indicates the degree of overlap between the income distributions of urban and rural areas. Inequality broadly declined during the past two decades. Starting from a value of 46.9 in 2001, Mad- Regional inequality reflects the heterogeneity in agascar’s Gini index dropped to 38.9 in 2005, rose socio-economic conditions across the country. to 42.7 in 2010, then steadily diminished to its cur- Table 5 provides within/between-group inequality rent value of 36.7 (Table 4). The consumption shares across urban and rural areas, and also across regions. of the lower quintile of the consumption distribu- tion (mainly comprised of rural households) rose, Table 5: Inequality is largest within broader while the consumption share of the upper quintile geographic areas and between regions (mostly urban households) decreased, leading to a Total 36.8 decline in overall inequality. However, while inequal- Urban / rural ity within the rural population decreased, inequality Within-group inequality 22.0 within the urban population increased. Between-group inequality 8.3 Overlap 6.4 The decrease in national inequality can be decom- Region posed into to an increase in urban inequality, a Within-group inequality 2.4 decrease in rural inequality and a slight decline in Between-group inequality 13.8 inequality between rural and urban incomes. Table 5 provides a breakdown of the Gini index in 2022 Overlap 20.6 by geography, first by urban/rural status and then Source: Author’s calculations based on 2022 EPM data. 19 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Within-group inequality by region was relatively low Figure 15: Madagascar's inequality level remains at 2.4, indicating relatively equal income distribu- below peer countries' tion. Between-group inequality was much higher at 13.8, indicating that there is significant income Bangladesh, 2016 inequality between different regions of Madagas- Madagascar, 2022 car, which is not surprising given the heterogeneity in geography, urbanization, and vulnerability across Malawi, 2019 regions. Tanzania, 2018 Inequality remains lower than among aspirational Uganda, 2019 peer countries and other countries with similar poverty rates (Figure 15). For example, Madagas- Sub-Saharan Africa, 2019 car has lower inequality than Tanzania, Uganda, and Rwanda, and similar to Malawi and Niger, which Rwanda, 2016 also have high poverty rates. However, the relatively lower inequality is the result of consumption falling 0.0 10.0 20.0 30.0 40.0 50.0 among urban households, together with slightly Gini index better consumption among the rural poor. Source: World Bank Staff estimates based on 2022 EPM data and WDI Online. Consumption gaps between urban and rural cent) or other urban areas (43 percent). Yet the gap households widen when moving from lower to between secondary cities, with those agglomera- higher deciles. Figure 16 illustrates inequality pat- tions that count between 5k-100k twice as many terns by comparing consumption across consump- poor as the capital city, counting almost 67 percent tion deciles. Except among the poorest, households among the poor, and are thus substantially more in the capital city and other urban areas consume comparable to rural areas than bigger cities. Simul- more than their rural counterparts. Even though taneously, poverty rates across regions also show they constitute most of the population, rural house- relevant variation. Western regions and particu- holds only consume 34 percent of overall food and larly those in the Grand South display some of the non-food resources in Madagascar, while urban highest rates, with Androy region having the near households account for 66 percent of consumption. totality of its population living below the national poverty line, followed by three other regions where Poverty and inequality across space more than 9 in 10 households are poor. Partially, this is explained by the high rural population shares in Monetary poverty in rural areas touches more than these regions, with Androy, Vatovavy, and Atsimo 8 in 10 households, with the Grand South showing Atsinanana having more than 90 percent of its pop- some of the highest rates, but with secondary cities ulation mired in poverty. Yet other factors matter counting as many as 7 in 10 households among the also: these regions are far from the capital city and poor. Poverty is twice as prevalent in rural areas (80 its relatively superior infrastructure and are often hit percent), compared with the capital city (35 per- by natural disasters such as locust infestations and Figure 16: Urban-rural consumption gap widens for richer deciles Average household consumption by decile 7 000 000 6 000 000 5 000 000 MGA 4 000 000 Rural 3 000 000 All Urban 2 000 000 Capital City 1 000 000 - 1 2 3 4 5 6 7 8 9 10 Deciles of consumption Source: World Bank Staff Estimates based on 2022 EPM data. 20 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment droughts. Northern regions benefit from greater These statistics suggest that the people of Androy economic activity, including in tourism and vanilla are among the most economically vulnerable, lack- production, and the incidence of poverty is there- ing in almost everything, and require cross-sec- fore noticeably lower. The eastern side of the coun- tional interventions to move the needle on poverty try has become an area of high poverty as cyclones there. Analamanga region, where the capital city of often make landfall in the east, leaving trails of Antananarivo is located, has a low poverty head- destruction and loss of assets among households. count rate of 53.7 percent. However, the high Gini Further local spatial heterogeneity is revealed by index of 40 percent indicates that income inequality the small areas estimates poverty map (Map 3) is more pronounced than elsewhere, implying that following Corral et al. (2022). In Atsimo Andrefana, a significant proportion at the bottom end of the one of the poorest regions, there are pockets of distribution of the population still struggles to make lower poverty rates in some communes. Simultane- ends meet. (Figure 17b) The Anosy region has the ously, some districts in wealthier regions display on highest inequality with a Gini index of 42, but this average lower poverty rates, such as in Betsiboka, is expected due to its mining activities and enclave Alaotra Mangoro or in Atsinanana. Lagging regions economy, which have rendered a small share of the with particularly high poverty rates and depth will population substantially better off. require targeted and tailored interventions as they lack in basically everything. 3. Multidimensional poverty: Urban areas are slipping on nutrition Different regions face different challenges: Androy in the South has the highest poverty headcount Multidimensional poverty declined significantly rate, while Analamanga region, displays a lower in the 2008-2018 decade and increased slightly poverty rate but higher levels of inequality. Androy’s between 2018 and 2021 after the country suffered poverty rate is a staggering 95 percent (Figure 17a). multiple shocks.18 The multidimensional poverty Additionally, the poverty gap in Androy is the highest headcount ratio dropped from 76 percent in 2008 in the country, with the average shortfall of the poor to 67 percent in 2018 (Figure 18), an 8-percentage from the poverty line at 50.6 percent. The squared point decline. However, multidimensional poverty poverty gap in Androy is also the highest among all then increased slightly between 2018 and 2021 regions, indicating poverty severity at 31.7 percent. (Figure 18). Figure 17: Poverty and Inequality varies widely across regions a. Poverty rate b. Inequality Poverty rates 2021/2 GINI 2021/2 ANDROY 95% Madagascar 0.37 Rural 0.34 VATOVAVY 92% Urban 0.39 AMORON I MANIA 88% Antananarivo Capitale 0.38 BONGOLAVA 87% Autres Grands Centres… 0.37 BETSIBOKA 84% Centres Urbains… 0.37 ANDROY 0.30 ANOSY 83% VATOVAVY 0.26 ATSIMO ATSINANANA 83% AMORON I MANIA 0.31 HAUTE MATSIATRA 83% BONGOLAVA 0.25 83% BETSIBOKA 0.27 ALAOTRA MANGORO ANOSY 0.44 IHOROMBE 83% ATSIMO ATSINANANA 0.28 MENABE 81% HAUTE MATSIATRA 0.31 ATSIMO ANDREFANA 81% ALAOTRA MANGORO 0.30 75% IHOROMBE 0.30 ITASY MENABE 0.30 SOFIA 74% ATSIMO ANDREFANA 0.39 VAKINANKARATRA 74% ITASY 0.29 SAVA 73% SOFIA 0.27 VAKINANKARATRA 0.33 DIANA 70% SAVA 0.27 BOENY 67% DIANA 0.29 ATSINANANA 67% BOENY 0.28 MELAKY 66% ATSINANANA 0.29 MELAKY 0.30 ANALANJIROFO 62% ANALANJIROFO 0.29 ANALAMANGA 56% ANALAMANGA 0.35 Source: EPM 2022. 18 See definition in Annex 1. 21 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Map 3: There is a large North-South poverty divide at the commune level Poverty (based on national poverty line) Extreme poverty (based on food poverty line) Percentage of population in poverty Percentage of population in food poverty at commune level at commune level 89.79 to 99.27 70.70 to 95.48 82.52 to 89.79 58.40 to 70.70 75.34 to 82.52 48.88 to 58.40 63.87 to 75.34 36.40 to 48.88 14.74 to 63.87 4.53 to 36.40 Missing Missing Source: Author’s calculations based on 2022 EPM data. Between 2018 and 2021, most areas of the mul- access to improved water sources, child mortality, tidimensional poverty index remained stable, and school attendance. Peer countries outpaced except for the nutrition indicator, which deterio- Madagascar’s progress on multidimensional living rated. Between 2008 and 2018, there were large standards indicators such as access to electricity of improvements in several living standards indica- access to clean water and improved housing met- tors, including sanitation, cooking fuel, housing, rics by important margins. And while Madagascar’s access to electricity, adequate nutrition and years of advances in nutrition standards were noticeable, all schooling.19 Deprivation rates in all these indicators peer countries were able to improve on this metric decreased by 7-10 percentage points in the 2008- more substantially. By 2021, the effect of the pan- 2018 period. The smallest improvements were in demic and several weather shocks was reflected in an increased share of electricity and adequate Figure 18: Multidimensional poverty fell nutrition deprivations. The latter in particular fell significantly since 2008, despite the recent back to its 2008 level. The deterioration of nutri- shocks tion is observed in rural and urban areas where it 78% 76% returned close to its 2008 level (Figure 19). Impor- 76% tantly, this deterioration was not driven by stunt- 74% ing among children under 5, which has improved 72% (declined) steadily. 70% 69% 67% Deprivations among the poor remain a predom- 68% inantly rural phenomenon, but with urban areas 66% also slipping on nutrition deprivation while not 64% recovering on other indicators. Overall level of dep- 62% rivations in rural areas are two to three times higher Madagascar than those observed in urban areas (Figure 20). In 2008/9 2018 2021 2021, deprivation in nutrition among rural house- Note: Multidimensional poverty captures acute deprivations holds are 39 percent, compared to 23 percent in in health, education, and living standards that Malagasy urban areas. Water deprivations are experienced people can face simultaneously in a given household. Source. OPHI. Madagascar corresponds to 2008, 2018, 2021 with by 2 out 10 individuals in urban areas, compared to harmonized scores for 2018 & unharmonized score for 2021.20 6 in 10 in rural areas. However, between 2008 and A household is considered deprived in adequate nutrition if it has a child under 5 whose height-for-age or weight-for-age is more than two 19 22 standard deviations below the median, or if it has a teenager with BMI-for-age that is under two standard deviations below the median, or if it has adults with a BMI below 17. OPHI (2022). 20 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Figure 19: Nutrition deprivation increased between 2018 and 2021 2008 2018 2021 2008 2018 2021 Child 100% mortality 80% 80% Cooking fuel Nutrition 72% 75% 76% 60% 69% 67% 69% 60% 59% 60% 58% 67% 40% 56% 56% 65% School 49% 59% Sanitation 20% attendance 53% 54% 36% 48% 49% 52% 40% 49% 0% 27% 33% 26% Years of 20% 26% Electricity 25% schooling 6% 6% 0% 5% Housing Assets lity ion nce ling ssets Water using tricity ation g fuel orta Nutrit ttenda schoo A Ho Elec it San Cookin dm ol a ears of Chil h o Water Sc Y Source: OPHI & Authors’ own calculations. Figure 20: Indicators of deprivation in Madagascar over time in rural (left) and urban (right) areas 2008 2018 2021 2008 2018 100% 100% 81% 82% 83% 80% 76% 75% 76% 80% 69% 67% 76% 66% 63% 72% 63% 60% 57% 56% 68% 60% 60% 62% 55% 56% 60% 40% 39% 40% 38% 38% 31% 32% 33% 36% 25% 25% 24% 28% 23% 24% 32% 34% 29% 20% 33% 20% 28% 29% 20% 17% 12% 19% 18% 10% 22% 22% 20% 16% 18% 15% 7% 6% 4% 11% 0% 6% 0% 3% 3% 8% lity ion nce ling ssets Water using tricity ation g fuel lity ion nce ling ssets Water using tricity ation g fuel orta Nutrit ttenda schoo A Ho Elec it San Cookin orta Nutrit ttenda schoo A Ho Elec it San Cookin dm a o f dm a o f Chil o o l r s Chil o o l r s Sch Yea Sch Yea Source. Censored headcount ratios, OPHI & authors’ calculations based on DHS 2021. 2018 levels of deprivation on several indicators have in Madagascar living amid multiple deprivations. been increasing in urban areas. Among the poor in According to the 2018 Census, 74 percent of the rural areas, deprivations in electricity and cooking Malagasy population is considered multidimension- also decreased furthermore. Mostly, urban areas ally poor in 2018, experiencing deprivations in more never fully recovered back to 2008 levels. Simul- than three dimensions.21,22 This translates into 19.04 taneously, the small wins booked among the poor million poor across the Malagasy territory. Overall, in rural areas between 2008 and 2018 have not these census-based calculations mirror the results been entirely eroded. Yet, this may not be the end using Demographic and Health Survey (DHS) infor- of it in either rural or urban areas. Nutrition tends mation. The share of the population experiencing to adjust more quickly than other markers as cut- deprivations across either of the multidimensional ting food consumption is one of the most common poverty dimensions is generally higher than 50 per- coping strategies among the poor. These negative cent. Whereas access to some of the indicators is adjustments tend to further translate and show experienced by the near totality of the Malagasy into other deprivations over time, including health population, such as access to improved sources of and education markers, but also asset decreases cooking fuel (99 percent), improved sanitation (88 as households continue to engage in both human percent), and improved water (84 percent). In addi- capital and asset depletion to weather particularly tion, around half of the population lives in severe long-lasting shocks. poverty, experiencing more than 50 percent of weighted deprivations. And 15 percent of the Mala- Multidimensional poverty and its correlates gasy population is considered vulnerable and at risk of falling into multidimensional poverty, cementing Multidimensional poverty is very high across Mal- ubiquity of multidimensional poverty in the country. agasy territory, with more than 7 out of 10 people 21 Small disparities in the multidimensional poverty rate between INSTAT’s published report and the one presented in this report are due to minor 23 differences in calculation regarding education and child mortality. 22 For the calculation of the multidimensional poverty index using the Census 2018 information, nutrition and child mortality are not included. Nutrition was not collected as part of the census and child mortality requires additional information not available to the WB team. Please see information in Annex 1 on calculation and comparison with DHS/ MICS calculations. TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment The largest overlap of different multidimensional Figure 21: Multidimensional poverty is mostly rural poverty dimensions is to be found in rural areas 90 as well as in the country’s Southern and West- 80 % of multidimensionally poor ern regions. The urban-rural divide runs deep in 70 84,3 Madagascar: the population living in rural areas 60 being almost three times as likely of being mul- population 50 tidimensionally poor, when compared to those 40 living in urban areas.23 To put this into perspec- 30 tive: 1.54 million of poor reside in urban areas, 17.5 million of poor reside in rural areas. Moreover, 20 31,2 there is a strong relationship of the incidence of 10 poverty, and the intensity of poverty across dif- 0 ferent communes. More than 90 percent of the Urban Rural population faces multidimensional poverty in Source: Authors’ calculations based on 2018 Madagascar the regions of Vatovavy Fitovinany (92 percent), Census. Atsimo Atsinanana (95 percent), Melaky (92 per- Map 4: Multidimensional poverty is higher in the cent) and Androy (95 percent; Figure 21 and Map 4). South % of population 0-25 Multidimensional poverty is higher among house- 25-50 holds with a younger head. The incidence of pov- 50-75 erty is 80 percent among the population living in 75-80 80-85 a household with a head aged between 15 and 24 85-90 years, compared to 74 percent or less for those 90-95 living with older heads. With an economy structured 95-100 around primary sector activities, population growth will mean that already scarce resources will be divided among more individuals, leaving continu- ously fewer resources for the next young generations. Multidimensional poverty is furthermore higher among households with 7 or more members. Indi- viduals living in larger households are more likely to be multidimensionally poor. The incidence of pov- erty is 71 percent among individuals living in house- holds with one to six members (Table A1.5). And the poverty rate is 9 percentage points higher among those living in households with 7 or more members. Poor households have more household members because they tend to have a larger number of chil- dren aged 0-10: on average 1.5 among poor house- holds compared to 0.9 for non-poor ones. With more dependents, these households are more vul- Note: The solid white line outlines the boundaries of regions. nerable to shocks that may arise due to poor health, Source: Authors’ calculations based on 2018 Madagascar Census. death, or other environmental circumstances. Larger households also show other levels of multidimen- ferent pull factors, such as looking for employment sional poverty not directly considered in the multi- and following studies for example, to push factors, dimensional poverty calculation, such as number of such as droughts, insecurity, other family or health household members per room. Larger households reasons.24 According to these definitions, a very tend to have more members per room. diverse association between migration and poverty unfolds. Individuals who came because they are The association between migration and poverty in looking for employment in a given district, display a Madagascar is heterogeneous in function of initial multidimensional poverty rate of 39 percent, which reasons for migration. According to the Census, 13 is still higher than the average urban resident. When percent of the population was born in one district looking at those having left due to family reasons, and now resides in another. Among these 3.4 million we witness that 42 percent live in poverty among migrants, the reasons for migration were heterog- those individuals. Yet when zooming in on those enous (Table A1.4 in the Annex 1), ranging from dif- having left their initial place of residence due to Unless otherwise indicated, this report uses multidimensional poverty headcount ratios from the 2018 Census. 23 24 One third did not respond to their reasons for migration. 24 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment classic push factors, such as insecurity or drought Figure 22: A quarter of Madagascar's population occurrences, flooding, the poverty rates among is illiterate those individuals equal those in rural areas, at a rate 45 of 82 percent. 40 39 Iliteracy rate, adult total (% of people aged 15+) 35 Literacy status is strongly associated with multidi- 30 27 25 mensional poverty. Literacy is highly correlated with 25 23 22 23 education, a direct component of the multidimen- 19 20 sional poverty measure. But literacy measures an 15 additional concept that captures capacity to read 10 and write, irrespective of school attendance. With 5 extremely high poverty rates across the country, it 0 comes as no surprise that even among households 18 20 20 2015 01 8 01 5 01 8 with a literate head, the poverty rates are as high , 20 , 20 , 20 , ,2 ,2 ,2 DG C D M A A A as 66 percent. Yet, when looking at households in M LI BG KH RW TZ UG which the head is illiterate, in any of the languages Note: BGD Bangladesh, KHM Cambodia, MDG Madagascar, RWA Rwanda, TZA Tanzania, UGA Uganda, LIC average of spoken in Madagascar, the share of those in poverty low-income countries. In all cases the graph presents the is at a frightening 97 percent. Simultaneously, a rate latest available year. N/A not available. Source: World Development Indicators. of 23 percent illiterate population is high in compar- ison to other countries (Figure 22). Figure 23: Multidimensional poverty is concentrated in the primary sector Primary sector workers are also more likely to be 100 4 3 90 20 3 multidimensionally poor. Malagasy households % of employed population whose head is employed in the primary sector, face 80 a higher poverty toll with 89 percent. Meanwhile, 70 31 households with a head engaged in the non-pri- 60 aged 15+ mary sector are less likely to live in poverty. When 50 14 90 splitting further up by sectors, it is apparent that 40 among the non-poor, employment in the primary 30 sector is much less prevalent. The lion’s share of 20 35 non-poor employment is taken up by the second- 10 ary (14 percent) and tertiary sector (31 percent), 0 Non-poor Poor as well as other (20 percent). Whereas among the Primary sector Secondary sector Tertiary sector Other poor, employment in these sectors is minimal, with Source: Author’s calculation based on the 2018 Census. a combined share of 6 percent in total (Figure 23). tidimensionally poor and have access to electricity The Human Opportunity Index (Figure 24). The HOI looks at how some groups of people face different chances in life because of their Inequalities experienced early on in life may have conditions at birth. If access to opportunities dif- detrimental impacts on outcomes later on life. The fers according to certain conditions, the HOI goes Human Opportunity Index (HOI) measures how down. In Madagascar, among children aged 6-16, individual circumstances (i.e., characteristics – such the largest registered inequality is whether chil- as place of residence, gender, and education of dren live in a home that is not multidimensionally the household head – that should not determine poor. For instance, among children between 6 and access to basic goods and services) can affect a 16 the probability of living in a non-multidimen- child’s access to basic opportunities such as water, sionally-poor household is not equally distributed; education, electricity, and sanitation. The index is instead, some children face a higher probability of a measure of the coverage rate of an opportunity, poverty. Hence, while the overall multidimensional discounted by inequality in its distribution across cir- poverty rate is 22 percent for this group, their HOI is cumstances groups. And it delivers a measure that 15 percent, a 7 percentage-point penalty. Moreover, reflects how far a society is from universal access 33 percent of children live in homes with access to to an essential good or service, and how equitably electricity. Yet, due to its uneven distribution along access is distributed across individuals with differ- circumstances, the HOI is 28, after the application ent initial circumstances.25 of a 5-percentage points penalty factor. The largest inequalities observed among children Spatial and inter-generational factors tend to in Madagascar is to live in homes that are not mul- explain the largest variation in the inequality of The methodology and the variables used as circumstances and opportunities for Madagascar. 25 25 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Figure 24: The Human Opportunity Index for children aged 6-16 is below the coverage rate 100 90 80 70 60 50 40 30 20 10 0 Improved Improved Electricity Literacy Ever Attending Attending Having Having Not being water sanitation (any attending primary secondary primary secondary poor language) school school school education education HOI Coverage Source: Authors’ calculations based on 2018 Madagascar Census. opportunities among children (Figure 25). Location did not, particularly in disadvantaged regions and characteristics, such as residing in a rural or urban rural areas, will be important. In addition, laying a area, or in a given region, head of household char- particular focus on children in school whose parents acteristics, such as gender or existing literacy in any are not literate will also be important to improve lit- language, but also socio-economic status, reflected eracy rates among younger generations. by the asset index quintile of a household explain the largest share of variation across opportunities. 4. Characteristics of the poor, Location factors explain up to 54 percent of dispar- poverty vulnerability, and ities in access to improved sanitation, 37 percent subjective poverty regarding access to improved water, and 36 of not being poor, followed by 36 and 33 percent in having Household characteristics and rural literacy in any language or having ever attended community features predict monetary school respectively. Socio-economic factors as poverty the most proxied by the asset index, capture 51 percent in disparities regarding access to electricity, 45 per- It is important to understand which characteristics cent in disparities in the opportunity of not being are on average associated with monetary poverty. multidimensionally poor, as well as 49 percent of For Madagascar, spatial dimensions play a big role. having some primary schooling. Characteristics of But other socio-demographic correlates also matter the head of household, such as their own literacy and understanding which ones can help tailor inter- and sex, explain 41 and 40 percent in explaining dis- ventions that target people versus places. parities observed among children in literacy in any language and ever having attended any schooling. Education, household size, age, and formal work As a result, improving income-generating opportu- are on average associated with lower poverty nities but also intervening to ensure children do not (Figure 26). Given the pervasive nature of poverty in end up not going to school because their parents the country, it is little surprising that poverty attains Figure 25: Location and assets explain most disparities in access to opportunities for children aged 6-16 100 90 80 70 60 50 40 30 20 10 0 Improved Improved Electricity Literacy Ever Attending Attending Having Having Not being water sanitation (any attending primary secondary primary secondary poor language) school school school education education Individual Location Household Head Socio-economic Source: Authors’ calculations based on 2018 Madagascar Census. 26 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Figure 26: Larger households with children and agricultural work are associated with higher poverty 95 85 % of population 75 65 55 45 35 25 15 5 Man Woman 0-14 years 15-65 years 65 and older Primary certificate Secondary certificate Tertiary certificate Without education Agricultural worker Domestic chores Non-farm enterprise (nonwage) Not working Wage worker HH head Man HH head Woman Non-migrant Migrant HH size: 1 person HH size: 2-5 persons HH size: 6-10 persons HH size: 11-20 persons Single Married Free Union Divorce/Separate Widower Not Concerned Sex of Migration Sex Age range Education Labor status household status Household size Marital status head Source: EPM2022. Note: Dashed line is poverty rate. individuals from households with different char- of the household grows older. The second signifi- acteristics similarly. Gender, or the gender of the cant variable is household size: each extra house- head of household make very limited difference and hold member lowers consumption per capita by observed poverty rates are almost identical across at least 35 percent, depending on location, and these groups. Children are poorer, as are households increases the likelihood of poverty in the household with more individuals, pointing to limited returns to by more than 60 percent. Third, number of adult additional children even in this highly (subsistence) females is positively correlated with consumption agriculture reliant country. While access to good per capita in the household (in secondary cities, the work has been one of the main drivers of poverty boost amounts to 2.3 percent). Fourth, number of reduction around the world, work makes only very adult household members has the opposite effect, limited difference in Madagascar as formal or good negatively correlated with consumption levels – for jobs are very limited. Nonetheless, while wage work example, in rural areas, each extra elderly person is in Madagascar is rare, this group’s poverty rate is 10 correlated with a 5.5 percent consumption reduc- percentage points lower than the national average, tion and increases the likelihood of poverty in the at 69 percent. The mainstay work of Malagasies, household by 9.5 percent. Fifth, the consumption which is in agriculture, the poverty rate is above the levels of female-headed households in rural areas national average at 87 percent. Education on the are 2.8 percent higher than those of male-headed other hand is strongly associated with lower rates of households, and it lowers the likelihood of poverty poverty: among households with a head in access to by 11 percent. The sixth significant characteristic is primary education, the poverty rate is substantially the marital status of the head of household, but the lower. Heads with a head of household with a sec- effect depends on location: in major cities, house- ondary certificate or even tertiary education have holds with a married head have a 10 percent lower an average poverty rate of 45.8 and 16.6 percent consumption, in secondary cities, their consump- respectively. Moreover, in Madagascar, migrants tion is 4 percent higher and the likelihood of pov- are on average wealthier. Importantly, this category erty drops by 8.5 percent. Finally, households with captures anybody who has been born in one place educated heads consume more and are less likely to and is now living in another. As a result, it does not be poor, but the effect is not statistically significant, only encompass those individuals who have been other things equal. forcibly displaced against their will but also (pre- dominantly) those who may have self-selected into Certain indicators of living standards are strongly new locations to find jobs and better education. As related to household consumption. For instance, such, it is not entirely surprising that the observed in secondary cities, households who live in larger average poverty rate among migrants is 61 percent homes tend to consume more, with each extra room and well below the national average. corresponding to a 1 percent increase in consump- tion. In rural areas, using electricity is a costly exer- Using simple regression design confirms these cise, leading to a 3.9 percent drop in consumption results. Results are presented in Table 6 and A2 (in and a 12 percent increase in the likelihood of pov- the appendix). In major urban centers, consump- erty. Electricity is expensive in rural areas because tion decreases at an accelerating rate as the head demand is low. However, communities where an 27 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment electricity network is available tend to host more households practice agriculture and should benefit affluent households with higher consumption levels from high rice prices—it is explained by the limited and there is a lower likelihood of being poor in areas market participation of the majority of farmers, who without electricity. Sanitation is another key wel- only sell a small share of their production. The pro- fare indicator. In rural areas, households without ceeds from sales of crops and livestock contribute toilet facilities have lower levels of consumption to greater consumption and a lower probability than those with them. In secondary cities and rural of poverty among households in secondary cities. areas, households in communities with piped water Moreover, in communities where agricultural inputs networks have higher consumption levels and are such as fertilizers and pesticides are available, con- less likely to be poor. In major urban centers, being sumption decreases as the distance to the source of connected to the piped water network corresponds the inputs increases. Then in agricultural communi- to 31.2 percent less consumption and a 66 percent ties, which comprise 80 percent of the population, increase in the probability of poverty (likely due to the effect of salaries differs by the gender of the the cost of water rates). Other significant infrastruc- worker. The salary of a male farm worker is asso- tural variables with a positive effect on consumption ciated with lower consumption in households and include access to paved roads, frequently available higher poverty levels. The opposite is true for the transport, and cellular telephone networks: house- salary of female farm workers. The average salary holds in communities where these utilities are avail- for a male farm worker in the major urban centers is able tend to exhibit higher levels of consumption MGA 1,186; in secondary cities it is MGA 5,096; and and a lower probability of poverty. in rural areas it is MGA 4,539. The average salary for a female farm worker in the major urban centers is Other important determinants of household wel- MGA 1,167; in secondary cities it is MGA 4,539; and fare relate to agriculture. Households that own land in rural areas it is MGA 4,232. Though lower than tend to consume more and are less likely to be poor, their male counterparts, female salaries tend to and the intensity of the effect increases with the size be associated with higher consumption and lower of the plot. On the other hand, the likelihood of rural household poverty. This is consistent with literature poverty grows with the price of rice. While this find- which shows that female workers often come from ing might appear counter-intuitive—since most rural wealthier households (Karoly and Burtless, 1995). Table 6: Demographic and agricultural variables drive poverty Ordinary Least Squares PROBIT Secondary Secondary Major Urban Major Urban VARIABLES Urban Rural Urban Rural Centers Centers Centers Centers Dependent variable: Poor Age of HH head -0.078 0.026 -0.006 0.102 0.032 0.050 -0.056 -0.027 -0.020 -0.118 -0.070 -0.056 Age of HH head Squared 0.0115* -0.002 0.001 -0.016 -0.006 -0.005 -0.007 -0.003 -0.002 -0.014 -0.009 -0.007 HH Size -0.390*** -0.345*** -0.364*** 0.624*** 0.692*** 0.747*** -0.044 -0.023 -0.016 -0.096 -0.056 -0.055 # of Adult Male in HH -0.040 -0.002 0.006 0.083 0.010 0.011 -0.027 -0.014 -0.009 -0.059 -0.034 -0.028 # of Adult Female in HH 0.018 0.0234* 0.012 -0.069 -0.012 -0.022 -0.025 -0.014 -0.011 -0.059 -0.034 -0.030 # of Children in HH -0.003 0.004 -0.006 0.032 -0.007 -0.003 -0.012 -0.006 -0.005 -0.028 -0.015 -0.013 # of Elderly in the HH 0.043 0.028 -0.0546*** -0.151 -0.044 0.0948* -0.046 -0.028 -0.018 -0.119 -0.063 -0.054 HH head is female -0.050 -0.013 0.0297* 0.069 0.057 -0.110** -0.044 -0.026 -0.018 -0.103 -0.058 -0.051 HH Head is Married -0.0995** 0.0404* -0.010 0.112 -0.0845* 0.016 -0.040 -0.021 -0.015 -0.091 -0.050 -0.043 Years of School of HH head 0.008 0.000 0.001 -0.009 0.000 0.001 -0.005 -0.003 -0.002 -0.011 -0.006 -0.005 HH Head is an employee -0.035 0.000 0.025 0.054 0.040 -0.076 -0.048 -0.025 -0.018 -0.109 -0.063 -0.052 28 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment HH Head is in Business -0.049 -0.003 0.010 0.144 -0.011 -0.032 -0.046 -0.024 -0.017 -0.109 -0.058 -0.050 # of Rooms in the HH -0.016 0.0107* -0.004 0.024 -0.021 -0.007 -0.015 -0.006 -0.004 -0.035 -0.014 -0.013 HH has a water pipe -0.142 -0.031 -0.014 0.180 0.084 -0.016 -0.143 -0.047 -0.038 -0.240 -0.126 -0.108 HH has Electricity 0.010 0.032 -0.0386** -0.057 0.041 0.124*** -0.042 -0.025 -0.017 -0.098 -0.060 -0.048 HH has Internet 0.274 0.027 -0.034 -0.177 -0.219 0.224 -0.181 -0.084 -0.065 -0.447 -0.232 -0.216 HH has No Toilet -0.033 0.007 -0.0285* 0.025 0.006 0.035 -0.044 -0.023 -0.017 -0.095 -0.058 -0.047 HH has a Flush Toilet 0.032 -0.018 0.0474* -0.116 -0.184* -0.152** -0.079 -0.037 -0.025 -0.209 -0.096 -0.073 HH is Accessible -0.031 0.025 -0.020 0.034 -0.001 0.031 -0.038 -0.021 -0.015 -0.088 -0.049 -0.042 Land size 0.000959** 0.000 -0.000343*** -0.00222* 0.000 -0.00112** 0.000 0.000 0.000 -0.001 0.000 -0.001 Price of rice 0.000 0.000 0.000 0.000 0.000 2.30e-07*** 0.000 0.000 0.000 0.000 0.000 0.000 Livestock Sales Revenue 0.000 8.15e-09** 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Crop Sales revenue 0.000 3.05e-10*** 0.000 0.000 -2.82e-08* 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Travel Time to nearest city 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Commune has Paved Road 0.089 0.023 0.0619*** -0.264 0.036 -0.0958* -0.070 -0.026 -0.020 -0.161 -0.067 -0.054 Commune has Frqnt Transport 0.070 0.0547** 0.0659*** -0.243** -0.148** -0.175*** -0.058 -0.026 -0.018 -0.119 -0.063 -0.049 Commune has Electricity Ntwk 0.300*** 0.214*** 0.165*** -0.531*** -0.453*** -0.290*** -0.079 -0.027 -0.025 -0.203 -0.069 -0.063 Commune has Pipe Water Ntwk -0.312*** 0.0535** 0.0808*** 0.661*** -0.039 -0.148*** -0.075 -0.024 -0.019 -0.184 -0.063 -0.050 Commune has telecom Ntwk -0.183 0.003 0.0893*** 0.465* 0.027 -0.191*** -0.123 -0.038 -0.017 -0.249 -0.095 -0.052 Price of Cereals in Commune -0.024 -0.0538** 0.0646*** 0.278 0.048 -0.036 -0.094 -0.026 -0.022 -0.204 -0.069 -0.056 Org fertilizer avg price -0.189* 0.018 -0.106*** 0.152 0.009 0.125* -0.110 -0.028 -0.027 -0.245 -0.067 -0.073 Non-org fertilizer avg price 0.317*** -0.140*** -0.004 -0.373 0.253*** -0.013 -0.119 -0.032 -0.028 -0.255 -0.072 -0.074 Pesticide avg price -0.255*** -0.028 -0.0988*** 0.422** 0.057 0.225*** -0.088 -0.023 -0.019 -0.183 -0.057 -0.056 Water flow all year -0.004 0.005 0.022 0.040 0.075 -0.058 -0.114 -0.026 -0.015 -0.286 -0.063 -0.044 Avg salary for male -0.00103*** 0.000 -4.30e-06** 0.00230*** 0.000 0.000 0.000 0.000 0.000 -0.001 0.000 0.000 Avg salary for female 0.00102*** 7.72e-06*** 1.45e-05*** -0.00233*** -1.46e-05* -2.67e-05*** 0.000 0.000 0.000 -0.001 0.000 0.000 Constant 15.36*** 14.42*** 14.55*** -1.516*** -0.232 -0.346** -0.153 -0.079 -0.053 -0.364 -0.194 -0.154 Observations 2291 5567 8664 2291 5567 8664 R-squared 0.157 0.293 0.247 Source: Author’s calculations based on 2022 EPM data Notes: Robust standard errors in below coefficients. Significance level: *** p<0.01, ** p<0.05, * p<0.1. 29 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Vulnerability that assesses not only today’s poverty status but also a household’s probability to fall below a certain Vulnerability combines the concepts of poverty, poverty threshold needs to be assessed. For policy exposure to risks, and risk management in order to design, particularly in a low-income context like predict the probability that a household or individ- Madagascar, it makes sense to focus on a short time ual will be poor in the future. Vulnerability is a for- horizon. As such, following Günther and Harttgen ward-looking measure that reflects the probability (2009), in this analysis, we define a household as of poverty in the future, and thus closely linked to vulnerable if their predicted probability to fall under coping and consumption smoothing capacity, if it is the poverty line within coming two years is greater built on a consumption measure (Klasen and Waibel than 50 percent. 2015). Looking into the future is more uncertain than evaluating poverty in the cross-section. Optimally, Vulnerability rates in 2022 are higher than poverty panel data is employed to capture welfare dynam- rates, reaching quasi-totality in rural areas overall ics and exposure to shocks of a given household. In and steep rates in regions with high share of rural many countries, panel data is not available, rendering populations. Albeit high poverty rates, vulnerability alternative strategies necessary (Chaudhuri, Jalan, rates are even higher since even among the non- and Suryahadi 2002; Günther and Harttgen 2009). poor Malagasy population, there are those who are Following Chaudhuri, Jalan, and Suryahadi (2002), vulnerable to falling into poverty in the near future. the next section estimates vulnerability in four steps: The vulnerability numbers are staggering, in par- 1. In the first step, the main correlates of the house- ticular in rural areas where the quasi-totality of indi- hold’s consumption level are identified to assess the viduals is either poor or prone to falling into poverty. strength of the relationships between different char- It is further notable that in urban areas, where pov- acteristics and household welfare, and household erty has been on the rise between 2012, an even consumption is regressed on a set of independent larger share, another 20 percent, of individuals are variables which include household composition and currently non-poor but prone to fall into poverty in demographics, livelihoods, and regional and geo- the following two years, which is true for the largest graphic control variables. Secondly, the relationship part of the population. Given the high poverty rates between the household characteristics and the risk and high shares of rural populations living in each of of welfare shocks is estimated, and the variation in Madagascar’s region, the vast majority of the Mala- household consumption that is not explained by the gasy population is vulnerable, illustrating the enor- estimation model in step 1 includes the household’s mous needs for intervention to heave populations risk of shocks. This variation is used to test which out of structural poverty but also shield those that characteristics are associated with the risk of welfare are at risk of falling into poverty in the near future. shocks. Thirdly, and based on step 1 and 2, a house- hold’s future level of consumption and variation of Vulnerability remains much higher in rural areas consumption is predicted. Ultimately, we determine but particularly in secondary cities is catching up, households’ probability of falling into at any stage given more comparable consumption levels. Figure over the next 2 years. Those with a probability of 27, Map 5 and Figure 28 show the cumulative pro- over 50 percent are classified as vulnerable (follow- portion of the rural and urban populations, split into ing Günther and Harttgen 2009). Importantly, this capital city, other grand urban centers, and second- approach is based on several assumptions about the ary urban conglomerations, against consumption distribution of risks. Employing cross-sectional data levels relative to the poverty line. 80 percent of the means that we observe households only during one population living in rural areas display the largest period and thus assume that household’s variation share of the population with consumption levels of consumption is constant over time. As a result, below the poverty line. Yet, among the 11.6 percent large but rare shocks that do not occur in every year of the population residing in secondary urban areas, go unaccounted for. Other important assumptions the situation is catching up quickly. Here, a large include the absence of measurement error in con- share of the population has been consuming close sumption reports, and assumptions on the distribu- to the poverty line and is more similar to the rural tion of risks and the validity of ordinary least square populations in terms of consumption levels than to estimates (see Klasen and Povel 2013) for a more the other big urban areas, let alone the capital city. detailed discussion). Vulnerability focuses on poverty dynamics, com- In addition to rural-urban variation, some groups bining concepts of poverty, and the risk in the near are more prone to be vulnerable, underlining in future. Given high consumption volatility of house- particular the role of education and demographic holds in risky environments over time, a concept transition. Among households where the head 30 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment declares no education at all are the most likely to be Figure 27: Rural vulnerability is significantly higher vulnerable to poverty. Among this group, the uncon- 100 ditional vulnerability rate is 98 percent (Figure 28), 90 96 94 98 where heads of household report no education at 80 88 all. It is important to note that the situation among 70 78 % of population younger cohorts has been improving, but particular 60 emphasis on high quality educatoin remains key. 8 50 57 40 percent of households have a head with lower sec- 30 ondary education. Among this group, vulnerability 20 is already considerably lower at 80 percent. Among 10 those, whose heads of household attained higher 0 secondary education or more, the vulnerability rate 2012 2022 Overall Urban Rural drops to 56 percent (who constitute another 6.5 Source: Author’s calculations based on 2022 EPM data. percent among Malagasy’s population). 56 percent of Malagasy households report 5 or more house- Map 5: Vulnerability is greatest in the South holds. Among this group, 97 percent are vulnerable. % of population Demographic dividends have been widely discussed 95-100 90-95 in other SSA ocutnries and it remains important to 85-90 keep the dialogue going to propel next generations 80-85 to becomore more aware of the benefits of smaller 75-80 70-75 family sizes and give reproductive control mech- anisms to families and women alrady requesting them today. Gender gaps are strongly related to poverty Despite having higher educational attainment than men, women have lower access to paid employ- ment. School completion rates are higher among females than males. EPM 2022 data reveals that the secondary school completion rate for females (37.1 percent) is higher than that for males (36.8 percent), and in junior high school, females score higher in reading exercises than males. Even so, the absolute number of employed females (5.4 million) is lower than that of males (6 million). Moreover, women have very different types of employment than men. Fewer than 30 percent of women are either wage employees or employers, compared to 41 percent Figure 28: Rural, large households with no education are the most vulnerable 100 90 80 % of population 70 60 50 40 30 20 10 0 Antananarivo Capital Other big urban centers Secondary urban centers Rural 4 or less 5 or more Female Male None CEPE (Primary) BEPC (Lower Secondary) CAP or higher Employed Not employed Owned Rent Other Electricity Other/None Flush toilet Latrines Other/No toilet Location Household Sex Education Employment Dwelling Lighting Toilet size of the head of the head of the head source Overall Source: Author’s calculations based on 2022 EPM data. 31 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment of men. In contrast, women are overrepresented Figure 29: Women are relatively more likely to among family contributors, subsistence farmers work as family workers and in subsistence farming and the self-employed (Figure 29). Looking at mul- Male tivariate correlations, regression results from Table Family Contribution 6 show that, controlling for demographic charac- 5% Employer teristics including household size, education of the 6% household head, access to basic services, and use of agricultural inputs, rural households headed by women have higher consumption, as well as house- holds that have more female adults (especially in Employee secondary urban centers). The salary of female agri- 35% Subsistence cultural workers is also associated with higher con- farm sumption and lower probability of poverty. How- 23% ever, compared to household size, which is strongly linked to fertility, the magnitude of the correlation on gender is relatively small. Self-employed 31% Males are dominant in most employment sectors, except for the professional, crafts, services, and sales sectors (Figure 30). Among those who report Female a specific occupation (mostly wage employees), Employer women and men have similar shares of employment 3% among professionals, clerical and support workers, craft, and trade workers, but men are overrepre- Family sented among skilled agricultural workers, plant Contribution and machine operators, managers, and unskilled 14% Subsistence workers, whereas women are more likely than men farm to work in sales and services. This is consistent with 32% the hypothesis that women who are wage employ- ees are also more likely to be urban, skilled workers. Employee 24% Women tend to work fewer hours than their male Self-employed counterparts (Table 7). Among women, close to 27% 70 percent report working less than 35 hours per week, with a majority (45 percent) working between 25 and 34 hours. By contrast, 56.4 percent of men Source: Author’s calculations based on 2022 EPM data. report working less than 35 hours on average. This Figure 30: Services tend to employ relatively more women Occupation employment shares (%) Skilled agricultural forestry and fishery workers 24.3 31.7 Elementary occupation 23.2 27.0 Craft and related trade workers 9.1 8.8 Service and sale workers 8.6 4.3 Professionals 2.6 2.4 Plant and machine operators and assemblers 0.4 2.2 Technician and associate professionals 1.2 1.3 Clerical support workers 0.9 0.9 Manager 0.3 0.6 Armed Force occupation 0.1 0.3 0.0 5.0 10.0 15.0 20.0 25.0 30.0 Female Male Source: Author’s calculations based on 2022 EPM data. 32 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Table 7: Women tend to work fewer hours than men Self-reported perceptions of poverty and (employment shares by hours of work per week, %) well-being reveal most people “struggle” Male Female Globally, subjective poverty results reconcile with <25 19.8 24.5 observed monetary poverty measures. Asked how 25-34 36.6 44.9 their household ranked on the welfare scale rela- 35-39 8.3 6.8 tive to society at large, 62.6 percent of household 40-48 20.2 13.6 heads said that they were struggling economically, while 9.7 percent perceived themselves as poor 49-59 8.2 5.4 (Figure 32). This is consistent with the 2022 EPM >=60 6.9 4.8 data, which shows that 75.2 percent of households Total 100.0 100.0 are below the national poverty line. Subjectively, Source: Author’s calculations based on 2022 EPM data. another 27 percent thought their poverty status was average. The rich constitute 0.56 percent of the reflects the gender gap in time-use among adults sample, and the highest category of “richer” house- linked to care activities and housework. Likewise, holds constitute 0.06 percent of the sample. female wages are lower than male wages. As seen in Figure 30 above, men have a higher proportion of Despite stark observed spatial differences in mon- employees (35 percent) than women (24 percent), etary welfare and overall living standards, the vast while women are more likely to work in subsistence majority of Malagasies deem themselves strug- farming, which brings low returns. gling. To a large extent, this result speaks to impor- tant peer effects of what people believe to make a Among the working population 72.1 percent have decent living. Residents in rural areas declare most earnings below the poverty line. Figure 31 shows frequently that they either struggle or are poor, the type of employment of workers whose earnings which reconciles with rural poverty being the most are below the poverty line. Interestingly, there is rel- prevalent phenomenon in the country. Residents atively little variation in the share of working poor in urban areas closely follow in the same pattern. across types of employment, showing how low Yet, residents in Tana more often than in secondary earnings are for the vast majority of workers, even cities perceive themselves to be struggling. In part, those who are relatively better off. For instance, this comes back to the anchoring effect whereby among subsistence farmers, over 80 percent are even richer residents when surrounded by other working poor, but among employers (the category better off residents will level their minimum needs with the smallest share of working poor) the share is at a higher level than someone who is objectively still 57 percent for men and 46 percent for women. and observably poorer. If that was not the case, one Also noteworthy is the fact that working poverty would have expected more variation in the distribu- among the self-employed is extremely high, which tion between locations rather than the vast majority reflects the lack of paid employment opportunities. of Malagasies declaring to be struggling (Figure 33). Figure 31: Most workers earn poverty wages, regardless of type of employment Share of working poor by employment status (%) Subsistence farm Family Contribution Domestic Self-employed Trainee Employee Employer 0 20 40 60 80 100 Female Male Source: Author’s calculations based on 2022 EPM data. 33 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Strikingly, the minimum expenditure that the aver- Figure 32: Most people report to be struggling age household in each self-reported category con- financially sidered necessary for a decent living is below the Subjective Household Welfare poverty line. The mean minimum required amount Richer Rich was below the food poverty line of MGA1.15 million 0.06% 0.56% for all categories, except for the rich, who needed slightly more in order to have a decent living. This Poor implies that the majority of households do not 9.73% have enough income to afford basic necessities for Average a decent standard of living and are struggling to 27.08% afford even basic food. Elaborating on the main causes of poverty in their society, household heads cited a lack of jobs (43.5 percent), the inflated cost of living (13.1 percent), a Struggling lack of land (9.1 percent), low salaries (7.8 percent), 62.57% and insufficient education (6.2 percent). Such perceptions were similar across the consumption distribution (Figure 34). Among the main perceived Richer Rich Average Struggling Poor causes of poverty, the highest proportion of households that suffered from a lack of jobs was in Minimum monthly amount for a decent living (MGA) the fifth quintile. Cost of living affected a greater 1 400 000 proportion of households in the second quintile. 1 200 000 1 158 564 The lack of land was equally problematic for those in 1 000 000 the first and fourth quintiles. Low salaries affected more households in the second quintile. Insufficient 800 000 670 265 education was a more prevalent problem among 600 000 535 000 471 645 households in the third quintile. Lack of livestock 381 071 400 000 was a greater problem among households in the fourth quintile. Poor roads affected households 200 000 in the third and fifth quintiles more than others. - Droughts and floods affected more households in Poor Rich Average Richer the fifth quintile. Struggling Source: Author’s calculations based on 2022 EPM data. Figure 33: Perceptions of financial difficulty are higher in rural areas 60.0% 40.0% 20.0% 0.0% Plus riche Riche Moyen En di culté Pauvre Votre HH sur une échelle de bien-être Antananarivo Capitale Autres Grands Centres Urbains Centres Urbains Secondaires Rural Source: Author’s calculations based on 2022 EPM data. 34 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Figure 34: Lack of jobs is the main perceived cause of poverty Main perceived causes of poverty, by consumption quintile 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Q1 Q2 Q3 Q4 Q5 Lack of j obs Insufficient education Lack of l and Lack of l ivestock herds Lack of roads Lack of water, pasture Droughts & floods Corruption Cost of li ving Poor agricultural sal es Laziness Low salary l evel Others Source: Author’s calculations based on 2022 EPM data. 5. Conclusion High poverty prevalence remains predominantly a urban areas having higher levels of consumption rural phenomenon but with urban areas slipping than rural areas remain. As a result, Madagascar’s quickly, warranting renewed focus as well. In 2022, low Gini, including compared to its peer countries, 75.2 percent of the Malagasy population could not is little reason for joy and is a result of a rising urban afford the minimum amount required to cover basic poor population. needs, which is estimated at MGA 1,477,565 person/ year. Rural poverty remained very high, while urban Poverty across Madagascar is heterogene- poverty increased significantly from 42.2 percent in ous across, with some regions lacking in almost 2012 to 55.5 percent in 2022, with the increase more everything while others displaying higher levels of severe outside the capital. The COVID-19 pandemic, inequality, but also household characteristics. Pov- high prices of staple food, natural disasters, and a erty rates are more comparable between rural areas soaring population of urban poor, particularly out- (80 percent) and secondary cities, counting as many side the capital city, contributed to the increase in as 7 in 10 households among the poor, and twice as poverty. The pandemic caused a decline in business high, compared with the capital city (35 percent) or income, loss of employment, decreased consump- other urban areas (43 percent). Yet the gap between tion, and an increase in food insecurity. This adds to secondary cities, with those agglomerations that the evidence that recent crises really affected the count between 5k-100k twice as many poor as the urban population in terms of consumption losses. capital city, counting almost 67 percent among the At the same time, it needs to be borne in mind that poor, and are thus substantially more comparable to rural populations, have very limited coping strat- rural areas than bigger cities. Western regions and egies as they start off at higher poverty levels to particularly those in the Grand South display some begin with. of the highest poverty rates. Androy in the South has the highest poverty headcount rate. High rural Inequality fell between 2012 and 2022, with the population shares and other factors, such as dis- Gini index dropping from 38.2 in 2012 to its cur- tance from capital city (and its relatively superior rent value of 36.7, mainly owed to impoverishment infrastructure), but also their exposure to natu- of urban population. The consumption shares of ral disasters (locust infestations and droughts) are the lower quintile of the consumption distribution some of the common culprits of extremely high (mainly comprised of rural households) rose, while poverty prevalence. Northern regions benefit from the consumption share of the upper quintile (mostly greater economic activity, including in tourism and urban households) decreased, leading to a decline vanilla production, and the incidence of poverty in overall inequality. Inequality between urban and is therefore noticeably lower. The eastern side of rural areas overall are retained, and moderate levels the country has become an area of high poverty of inequality between urban and rural areas, with as cyclones often make landfall in the east, leav- 35 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment ing trails of destruction and loss of assets among 16.6 percent respectively, compared to the national households. Simultaneously, Analamanga region, average of 75.2 percent. Multidimensional poverty displays a lower poverty rate but higher levels of is higher among households with a younger head inequality. These regional disparities suggest that (80 percent among those living with 80 percent targeted interventions are needed to address the with older heads of household). Multidimensional specific challenges faced by different regions in poverty is furthermore 9 percentage points higher Madagascar. among those living in households with 7 or more members, with 7 or more members. Among 3.4 mil- High poverty is not only monetary but also mul- lion migrants, the reasons for migration are heter- tidimensional, affecting more than two-thirds of ogenous ranging from different pull factors, such as its population. Madagascar is the eighth poorest looking for employment and following studies for country in multidimensional poverty, at 67 percent example, to push factors, such as droughts, inse- nationally with 77.3 percent of its rural population curity, other family or health reasons. Those coming and 41.2 percent of its urban population affected. to cities for employment, on average display higher In addition, 45.5 percent of the population suffers poverty rates than the average observed for urban from severe multidimensional poverty, i.e., they are areas, at 39 percent. When zooming in on those deprived in 50 percent or more of the 10 indicators. having left their initial place of residence due to According to the Population Census more than 7 classic push factors, such as insecurity or drought out of 10 people in Madagascar living amid multi- occurrences, flooding, the poverty rates among ple multidimensional deprivations, with the larg- those individuals equal almost everyone, at a rate est overlap of different multidimensional poverty of 82 percent. Literacy status is another important dimensions found in the country’s Southern and correlate of multidimensional poverty: its poverty Western regions and being twice as high in rural prevalence among heads of household illiterate in areas (81.4 percent), compared to urban ones (31.2 any of Madagascar’s language is a startling 97 per- percent). However, between 2008 and 2018, dep- cent. Employment in Madagascar does not guar- rivation rates in most multidimensional metrics had antee escape from (multidimensional) poverty, with decreased by 7-10 percentage points. All of the 76 percent of the employed living in poverty. This is above had come to a hard stop in 2021, following particularly striking when compared to the unem- the compound crises that hit Madagascar. Between ployed, those searching for work or other of whom 2018 and 2021, urban areas never fully recovered are less likely to be living in multidimensional poverty, back to 2008 levels, all while small wins booked shedding an important light on quality of employ- among the poor in rural areas between 2008 and ment in Madagascar. Meanwhile, households with 2018 have not been entirely eroded. Yet, this may a head engaged in the non-primary sector are less not be the end of it in either rural or urban areas, likely to live in poverty, while heads of household as coping strategies pertaining to selling assets can employed in the primary sector face a higher pov- unfold over many years and importantly decrease erty toll with 89 percent. We furthermore learn that households’ resilience to weather future and par- inequalities experienced early on in life may have ticularly long-lasting shocks, resulting in further detrimental impacts on outcomes later life. And in in slips on multidimensional poverty metrics. Madagascar, location factors explain up to 54 per- cent of disparities in access to improved sanitation, Monetary and multidimensional poverty prev- 37 percent regarding access to improved water, and alence importantly varies by household charac- 36 of not being poor, followed by 36 and 33 per- teristics, all while spatial and inter-generational cent in having literacy in any language or having factors tend to explain the largest variation in the ever attended school respectively. Socio-economic inequality of opportunities among children. Mon- factors, as proxied by the asset index, capture 51 etary poverty prevalence is on average associated percent in disparities regarding access to electricity, with education, household size, age, and formal 45 percent in disparities in the opportunity of not work. Children are poorer, as are households with being multidimensionally poor, as well as 49 per- more individuals, pointing to limited returns to cent of having some primary schooling. Investing in additional children even in this highly (subsistence) stronger human capital accumulation and creating agriculture reliant country. While access to good higher-quality jobs for the Malgache population will work has been one of the main drivers of poverty be key. reduction around the world, work makes only very limited difference in Madagascar as formal or good Vulnerability remains much higher in rural areas but jobs are very limited. Heads with a head of house- particularly in secondary cities is catching up, with hold with a secondary certificate or even tertiary more alike consumption levels. Vulnerability rates in education have an average poverty rate of 45.8 and 2022 are higher than poverty rates, reaching qua- 36 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment si-totality in rural areas overall and steep rates in lose policy focus either. To address these challenges, regions with high share of rural populations. Vul- a high-level policy implication would be to prioritize nerability focuses on poverty dynamics, combining targeted interventions and investments in human concepts of poverty, and the risk in the near future. and physical capital development, job creation, and Vulnerability is a forward-looking measure closely social protection programs. linked to coping and consumption smoothing capacity. Vulnerability remains much higher in rural areas but particularly in secondary cities is catching up, with more alike consumption levels. Here, a large share of the population has been consuming close to the poverty line and is more similar to the rural populations in terms of consumption levels than to the other big urban areas, let alone the capital city. If secondary cities are to become a vehicle for trans- formation, it will be important to diligently monitor progress on monetary and multidimensional met- rics in these agglomerations as different interven- tions are being rolled out. Eliciting subjective poverty prevalence, the major- ity of Malagasies deem themselves struggling, while a lack of jobs is cited as the main cause of poverty by 43.5 percent of household heads. Despite stark observed spatial differences in mon- etary welfare and overall living standards, the vast majority of Malagasies deem themselves struggling, with 62.6 percent of household heads stated that they were struggling economically, and 9.7 percent perceived themselves as poor. Elaborating on the main causes of poverty in their society, 43.5 percent of household heads cited a lack of jobs, 13.1 percent the inflated cost of living, 9.1 percent a lack of land, 7.8 percent low salaries, and 6.2 percent insufficient education In conclusion, Madagascar is facing significant and persistent challenges in poverty and inequal- ity. Many structural and cyclical factors have made matters worse for the Malagasy population, with differential impacts across rural and urban areas, but also across the country’s overall territory. Mad- agascar’s face of poverty is not only monetary but also multidimensional, affecting more than two- thirds of its population. The regional disparities suggest that targeted interventions are needed to address the specific challenges faced by differ- ent regions in Madagascar. In addition, the coun- try needs to importantly accelerate human capital accumulation by bettering different metrics, includ- ing improving educational outcomes. Major slips among the urban population into poverty is a cause for concern, indicating overall poor economic activ- ity and the lack of good jobs. Particularly with poor populations migrating from rural to urban areas in search for better opportunities, this issue warrants attention. Given the very high poverty prevalence in rural areas, it is also clear that this agenda cannot 37 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment This note proposes tests to identify possible errors Annex 1 in the conversion table and also possible correc- tions. This note includes two approaches – one uses 1. GDP Changes due to crises unit values, and the other uses calorie values. Table A1.1: Countries which experienced a decline (i) Detection of outliers using unit values in income The unit value is the ratio of expenditure to quan- GDP per capita – GDP per capita – Country 1960 2020 tity. To detect outliers, we calculate unit values in a (Constant 2015 US$) (Constant 2015 US$) standard unit. To do this, we convert consumption HTI – Haiti 1,714.10 1,322.80 from a nonstandard unit to a standard one. We then COD – Democratic estimate a unit value by dividing an expenditure of 1,254.70 487.40 Republic of Congo this item by its quantity in the standard unit. Even in NER – Niger 744.00 519.70 MDG – Madagascar 818.30 433.80 a developing country where integration of markets CAF – Central African is limited in rural and remote areas, it is difficult to 583.60 375.20 think a unit value is more than ten times bigger or Republic BDI – Burundi 290.70 263.40 less than the national median unit value. So, if we CAF – Central African 583.60 375.20 find such an outlier, it is likely that the unit value Republic is wrong because the conversion rate of the non- BDI – Burundi 290.70 263.40 standard unit to the standard unit is wrong. Source: World Bank, World Development Indicators Database. 2. Estimating poverty line Below, we demonstrate this idea mathematically. Details about the approach used to estimate the We calculate the unit values, (unit value of consumption aggregate and the poverty line are household i, item j) by dividing the expenditure for listed in the background paper entitled “Inequality this item, , by the quantity in a nonstandard unit, and Poverty in Madagascar: Estimates based on , after converted to a standard unit using the 2021-22 EPM”. The report provides details about original conversion rate, . the data cleaning exercise, the consumption aggre- gation approach, and the determination of the pov- erty line. 3. Detection of errors in the conversion Suppose the true (unobserved) unit value is rate of nonstandard units to standard (unit value of household i, item j), which can be cal- units culated with the true conversion rate : Household surveys in Africa use nonstandard units when collecting consumption data since standard units like litter and gram are not used in many rural areas and remote places. However, if consumption is The ratio of the unit values based on the original reported in a nonstandard unit, we need to convert conversion rate to the true unit value can be pre- them to a standard unit to estimate calorie intake sented as a ratio of the conversion rates: from consumption because the calorie conversion table is not available for nonstandard units. Since calorie conversion is essential when estimating food poverty lines, this conversion of nonstandard units to standard units needs to be accurate. However, This equation shows that if the original conversion the number of nonstandard units in a typical African rate overestimates the true conversion rate, the unit country is large. Worse, in some countries, the con- value is also overestimated. For example, if the orig- version rate of a nonstandard unit to a standard unit inal conversion rate is 100 times larger than the true can differ by region. As a result, in Madagascar, the rate, then the unit value based on the original rate is conversion table includes more than 4000 different 100 times bigger than the true unit value. Therefore, rates. Even if the conversion table is carefully devel- if we see too large a unit value, it can be because the oped, the probability of having errors in the conver- original conversion rate is wrong. sion table is likely to be non-negligible. 38 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment However, in reality, we do not see the true unit value unit conversion rate. But it is sometimes useful to and conversion rate. So, we cannot directly see the show possible corrections of the conversion rate, magnitude of error in the conversion rate. Instead, which reduces the diversion from the median unit we construct the ratio of the unit value based on value. We recommend the following correction. the original conversion rate to the median unit value. This ratio can be shown as a multiplication of Let refer to the median unit value of item j the ratio of the original conversion rate to the among outliers detected in the above rule 1. Then, we modify the nonstandard unit conversion rate true one with the ratio of the true unit value to the median unit value (see equation below). from . If this new conversion rate is used, the diversion from the median unit value declines to the diversion from the median unit values among outliers, which is . where refers to the median unit value of item j. (ii) Detection of outliers using calorie intake per day per capita As stated above, even though market integration is limited in rural and remote areas, the second com- We can do a similar analysis using each item's cal- ponent should be between 0.1 and 10. But, if there orie intake per day per capita. The idea is that if a is a mistake in the conversion rate, the first com- calorie intake from an item is too big or too low and ponent is not one; thus, the full effect could be far such outliers are concentrated in specific nonstand- more than 100 times less or bigger than the median ard units, there is a high likelihood that the outliers value. So, whenever we see the ratio of unit value are created due to errors in the nonstandard unit to the median unit value is more than ten times or conversion rates. Below, we present this idea more less than 0.1 than the median value, we suspect the formally. possibility of error in the conversion rate. Calorie intake per day per capita based on the origi- Lastly, an outlier can happen due to a reporting nal NSU conversion table can be written as: error of a specific sample household, not due to an error in the unit conversion rate. But if such outli- ers frequently happen when a specific nonstandard unit is used, we suspect that it is not due to report- ing errors but a systematic error like an error in the where aj refers to the calorie conversion table for nonstandard conversion rate. Therefore, if outliers in food item j in a standard unit, like gram. Calorie unit values are detected for more than 50 percent intake per day per capita based on the true non- of observations using a specific nonstandard unit, standard unit conversion rate can be written as: we recommend the nonstandard unit conversion rate be reviewed. Identification of potentially problematic nonstandard unit conversion rates (Rule 1) If the original NSU conversion rate is over-esti- mated, the calorie intake per day per capita will be Assuming 10 > > 0.1, we identify the unit under-estimated: values are outliers if > 10 or < 0.1. If such outliers are detected for more than 50 percent of observations using a specific non- standard unit, we recommend that the non- Again, since the true unit conversion rate is standard unit conversion rate be reviewed. unknown, we compare the calorie intake per capita calculated after converting a nonstandard unit with the median value. If the calorie intake per day per A possible correction of the nonstandard unit con- capita is much higher than the median of the food version rate based on Rule 1 item, the difference is unlikely to be explained by actual quantity differences but likely to be explained If we see clear outliers, we should contact the at least partly by the underestimation of the con- National Statistics Office to review the nonstandard version rate. 39 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Here is the formula tics and found that the corrections have a minimum impact of them. 4. Iterative poverty line estimation is the true difference, which is the same as the When we estimate a food poverty line, we need ratio of quantity in the standard unit. The above to determine a basket of food items and prices for equation shows that the ratio will be over-estimated them. To select the food basket, Ravallion (1998) if the NSU conversion rate is under-estimated >1). recommends we define a reference group and use their average consumption pattern to estimate a A possible correction of the nonstandard unit con- cost to consumer a calorie threshold, which is usually version rate based on Rule 2 around 2200 kcal per day per capita. To do this, for Madagascar, Edo, Mancini, and Vecchi (2022) esti- Identification of potentially problematic mate the cost per calorie from the reference group’s nonstandard unit conversion rates (Rule 2) food consumption and expenditures and multiply it with 2200 to estimate the food poverty line. Math- We identify outliers if +3*standard ematically, we can represent the derivation of the deviation of or -3*the standard food poverty line by Edo et al. (2022) as follows: deviation of . If such outliers are detected for more than 50 percent of observations of 1. Calculate average daily calorie intake per capita a specific nonstandard unit, we recommend of a reference group that the nonstandard unit conversion rate be reviewed. Like before, if we see clear outliers, we should con- where refers to daily consumption of household tact the National Statistics Office to review the non- ir’s item j per capita, and aj refers to a calorie conver- standard unit conversion rate. But it is sometimes sion coefficient for item j. useful to show possible corrections of the con- version rate, which reduces the diversion from the 2. Calculate the average daily household food median daily calorie intake per capita from item j. expenditure per capita of this reference group We recommend the following correction. Let refer to the median daily calorie intake per capita of item j among outliers detected in the above Rule 2. Then, we modify the nonstandard unit where refers to the price or unit value of item j faced by household ir. conversion rate from to = . If this new conversion rate is used, the diversion from 3. Calculate the cost per calorie for this reference the median daily calorie intake per capita declines group to the diversion from the median daily calorie intake per capita among outliers, which is . (iii) Applications to the Madagascar 4. Calculate a food poverty line by multiplying the 2021/22 data cost per calorie with a calorie threshold (here, 2200 kcal per day per capita) We applied the above two rules to the Madagas- car 2021/22 data. Good news is that we identified a relatively limited number of nonstandard unit con- version rates that produce many outliers. We iden- The food poverty line calculated with the reference tified 38 and 3 nonstandard unit conversion rates group (Gr) differs by the welfare level of households that might need corrections based on Rules 1 and in the reference group. This is because (i) there are 2, respectively. These potentially problematic non- some quality differences in each item, (ii) richer standard units and possible corrections presented households tend to buy higher quality of goods above were shared with the Madagascar NSO. Also, within the same item, and (iii) richer households we estimated the impact of possible corrections on tend to buy more of items whose price per calorie is the poverty headcount rates and other key statis- high than poorer households. 40 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment For example, suppose 80 percent of the reference holds should be consistent in that the basket and groups G1 and G2 have the same households but the prices of the food poverty line should reflect those rest of G1 are significantly richer than the rest of G2. whose food expenditures are close to the food pov- If the unmatched households in G1 pay higher prices erty line. For example, suppose no household in than the unmatched households in G2 to some item j’: the reference group is poor, then the food poverty line reflects the prices and consumption patterns of households who are significantly richer than the poor. Suppose all households in the reference group Then, even though the consumption patterns are are poor, then the food poverty line reflects the identical, the cost per calorie of the reference group prices and consumption patterns of those whose 1 is higher than that of the reference group 2. food consumptions tend to be lower than the food poverty line. Ideally, we would like to set the refer- ence group that includes households who spends for food items a similar amount to the food poverty line. Also, richer households’ consumption patterns can differ systematically from those of poorer house- However, there is no guarantee that this matching holds. For example, richer households tend to con- takes place when estimating a food poverty line. If sume more for food items, whose prices per calorie the matching does not happen, Ravallion recom- are high, and less for those whose prices per calorie mends the following iterative process. If the food are low, than poorer households. If so, as the ref- poverty line is bigger than food expenditures of all erence group becomes richer, the cost per calorie households in the reference group, then we move becomes higher. up the reference group (see Figure 35). Since raising the reference group usually increases the food pov- For example, suppose for items j and j’ such that erty line and the poverty rate. If the food poverty line is lower than food expenditures of all house- holds in the reference group, then we can move Figure 35: An example for the iterative process if the reference group is too low Otherwise, FPL1 G1 If so, the cost per calorie of G1 is higher than the cost per calorie of G2. FPL1 FPL2 These two conditions are not necessarily satisfied, G2 but it is often the case that the food poverty line becomes higher as the reference group becomes richer even though the calorie threshold is fixed. Figure 36: An example for the iterative process if the reference group is too high Ravallion (1998)’s iterative poverty line FPL1 approach G1 G1 The food poverty line differs by the welfare level of the reference group. If the reference group is raised, the food poverty line also tends to rise. The relation- ship is not certain, but it is observed in many coun- FPL1 FPL2 tries and surveys. This means that the food poverty line is not uniquely set by the calorie threshold but G2 affected by the calorie threshold and the reference group. Note. G1 is the original reference group and FPL1 is the original food poverty line. G2 is the adjusted reference group and FPL2 Ravallion recommends that the selection of the ref- is the new food poverty line. The axis shows the size of food erence group and the identification of poor house- household expenditures per capita. 41 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment down the reference group, which reduces the food Table A1.2: Average daily calorie intake per capita poverty line (see Figure 36). Unlike the contraction and cost per calorie for each decile of annual food mapping theory, there is no guarantee that this expenditures per capita iterative process converges to the reference group mean whose median household spends the food poverty Calorie per cost annual food Decile day per per FPL expenditure line as their food expenditures, but we usually see capita calorie per capita a reasonably good correspondence between the 1 655 1.16 990,411 245,312 food poverty line and the reference group. 2 1,045 1.10 990,411 384,102 3 1,312 1.08 990,411 480,597 Facilitate the convergence 4 1,477 1.12 990,411 562,944 5 1,671 1.14 990,411 644,141 To facilitate this convergence, we can calculate the 6 1,834 1.20 990,411 738,493 average daily calorie intake per capita by deciles 7 2,010 1.27 990,411 856,188 of household expenditure per capita (or per adult 8 2,213 1.35 990,411 1,003,194 equivalence) and see what decile’s average daily cal- 9 2,547 1.49 990,411 1,270,753 orie intake per capita is close to the calorie threshold 10 3,575 2.12 990,411 2,324,365 and use this group as the reference group. By con- struction, this group’s average calorie intake per day Source. Author’s calculations using the Madagascar 2021-22 EPM data. per capita should be close to the calorie threshold and the food poverty line, therefore, become close to the average household expenditure per capita of Consistency between the food poverty line the select decile. and the upper poverty line Here is a special case if a reference group is the Lanjouw and Lanjouw (2001) show that under 8th decile and their average daily calorie intake per normal conditions, the poverty headcount rate capita is 2200 kcal. In this case, is the average measured by comparing food expenditure per food expenditure per capita of the 8th decile, say capita with a food poverty line is close to that Madagascar Ariary 1 million, and is 2200. There- measured by comparing household expenditure fore, the food poverty line is the mean of the ref- per capita with an upper poverty line. To obtain this erence group, which is Madagascar ariary 1 million. result, the upper poverty line should be estimated from the food poverty line using the cost of basic needs methodology (Ravallion, 1998.) We now test whether this is indeed the case. In this way, the food poverty line is guaranteed to We set the calorie threshold at 2200 kcal per day be inside the reference group. In reality, the mean per capita. If we estimate the poverty headcount daily calorie intake per capita of any decile is not rate using food expenditures per capita and the exactly the same as the threshold. For example, in food poverty line, it is 74.9 percent. If we estimate the case of Madagascar 2021/22 data, Table A1.2 it using household expenditures per capita and lists the average daily calorie intake for each decile the upper poverty line, it is 74.6 percent. Therefore, of household food expenditure per capita. Except Lanjouw and Lanjouw (2001)’s theorem holds in this for the first three deciles, we can see the cost per dataset. calorie increases as households become richer.26 5. Survey to survey imputation If the calorie threshold is set at 2200 kcal per day approach per capita, since it is higher than the mean of the 7th decile and lower than the mean of 8th decile, we can The national statistics office of Madagascar collected choose the 7th and 8th deciles. Or since this Mada- a household budget survey, L’Enquête Nationale sur gascar 2020/21 data’s sample is large and the cal- le Suivi des indicateurs des Objectifs du Millénaire orie threshold is close to the mean of the 8th decile, pour le Développement (ENSOMD), in 2012 and a we can simply choose the 8th decile as the reference household poverty survey (EPM) in 2022. If raw con- group. If the former approach is taken, the food sumption data from ENSOMD 2012 and EPM 2022 poverty line is estimated to be Madagascar Ariary and an official poverty line of the upper poverty line 990,411, which is almost identical to the mean of in 2022, the national poverty rate declined from annual food expenditure per capita of the 8th decile, 99.1 percent to 79.7 percent. However, EPM 2022 Madagascar Ariary 1,003,194. used an improved questionnaire for collecting con- sumption expenditure data by largely increasing the The deviations from this in the first three deciles might reflect that the quality of data in the first two deciles might not be good. 26 42 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment number of consumption items. Beegle et al. (2010) the mean squared error and the absolute show that increasing the number of items tends to difference between the actual and imputed expand household expenditures. As a result, the poverty rate in the testing dataset. consumption data from ENSOMD 2012 are not directly comparable to those from EPM 2022. IV. Repeat the above exercise by changing the selection of folds for the training and testing To restore the comparability of consumption data datasets. between ENSOMD 2012 and EPM 2022 data, the household expenditures in ENSOMD 2012 were V. Average 1o mean squared errors and absolute imputed using the SWIFT (Survey of Wellbeing via differences. These numbers are identified as the Instant and Frequent Tracking) methodology. The mean squared errors and absolute differences SWIFT methodology imputes household expendi- between actual and imputed poverty rates tures in ENSOMD 2012 using the urban, capital corresponding to a threshold p-value of 0.5 city, and rural imputation models trained in EPM percent for the stepwise regression. 2022 data. According to Yoshida et al. (2023), the imputed household expenditures in ENSOMD 2012 VI. Repeat the above three exercises (3 – 5) by data are comparable to raw consumption expendi- increasing the threshold p-value by 0.5 percent tures in EPM 2022 data. at a time until the p-value reaches 10 percent. For this Madagascar study, we trained a model for VII. Select an optimal p-value by looking at the each stratum (capital city, other urban areas, and distributions of the mean squared errors and rural areas) using EPM 2022 data and applied the absolute differences between actual and stratum-specific model to ENSOMD 2012 data to imputed poverty rates impute household expenditures comparable to those of EPM 2022 and estimated poverty and inequality VIII. Repeat steps 1 to 7 to identify the optimal statistics of 2012. Below is the list of concrete steps. p-value for all three strata First, we prepared common variables in ENSOMD Third, using the optimal p-value for each stratum, 2012 and EPM 2022. Since the questionnaires of we conducted a stepwise regression analysis to ENSOMD 2012 and EPM 2022 are different, we con- finalize a SWIFT model for the stratum. structed common variables by matching the defi- nitions of variables. The common variables include Fourth, using the SWIFT model, we ran multiple household socio-demographic and economic char- imputations to impute household expenditures 20 acteristics (e.g., household size, age of the head of times in ENSOMD 2012 and estimated the point household) and fast-changing variables such as estimates and standard errors of poverty and ine- food and nonfood consumption based on their pre- quality statistics using the imputed expenditures. dictive power of actual consumption expenditure in the dataset. Several quality checks and other simulations are performed to ensure that the simulated poverty Second, we conducted cross-validation to search rate in the appended data differs by less than one for the optimal p-value for the stepwise regressions percentage point from the absolute poverty rate to minimize the risk of overfitting and other small in the household survey data. We also checked the sample biases. The cross-validation includes the fol- adjusted R2 values of the final models, guaran- lowing steps: teed that the poverty correlates have the expected signs, and dropped variables whose coefficients are I. Randomly split data from one stratum in the unreasonable. The model is also validated by com- EPM 2022 dataset into ten sub-samples with paring the kernel density of household consump- an equal sample size. Each subsample is called tion for the actual and imputed series. a “fold.” 6. Small areas poverty estimation II. Construct a training dataset using nine folds approach and a testing dataset with the remaining fold. A sample size of household surveys is typically too III. Run a stepwise regression in the training small to produce reliable estimates below a certain dataset with a threshold p-value of 0. 5 geographical level. The 2022 EPM household survey percent, impute household expenditures in the in Madagascar is only representative at the regional testing dataset using the model, and estimate level and thus producing reliable district-level pov- 43 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment erty rates from it requires supplementing the survey 7. Multidimensional poverty index with census data using small area estimation. The district-level poverty rates produced for this report The Multidimensional Poverty Index (MPI) identifies are computed based on the Empirical Best Predic- multiple deprivations experienced by households tion (EBP) approach using a variant of the R EMDI in three dimensions: education, health and living package.27 The team first identified and harmonized conditions. The health and education dimensions a common set of household-level variables that exist are based on two indicators each, while six indica- both in the 2022 EPM household survey and 2018 tors are applied to the living conditions dimension. Population Census. To carry out the estimation, the The MPI methodology requires to group variable team first estimate a prediction model using the modalities per dimension (Table A1.3). 2022 EPM survey that relates per capita consump- tion to a subset of common variables. The estimated All the indicators necessary to calculate the MPI for model parameters were then used to simulate wel- Madagascar are taken from the census. Each indi- fare in the census one hundred times. These simu- vidual household is assigned a score based on the lated consumption aggregates from the census are number of deprivations faced per household for then used to compute district-level poverty rates. each of the component indicators. The maximum score is 100, each dimension is weighted equally The set of indicators used in the prediction model of (33.33 percent). The health dimension is based on per capita consumption – which was selected using two indicators, each of these components has a a Least Absolute Selection and Shrinkage Operator value of 1/6 or 16.66 percent. Drinking water has (LASSO) to minimize BIC - included the following: been used as an indicator of health in the absence of information on nutrition. Education comprises two • Indicators for five of the six provinces indicators, each component of which has a value of 16.66 percent. The standards of living dimension is • An urban dummy based on five indicators, each component is equal to 1/15, or 6.66 percent. • Household size To determine multidimensional poverty, the depri- • The number of persons in the household above vations of each household are summed to obtain 64 and the dependency ratio the level of deprivation per household. The thresh- old value of 33.33 percent, which corresponds to • The number of adults employed one third of the weighted indicators, is used to distinguish between the poor and the non-poor. A • Age and sex of the head household (and each person in it) is considered to be multidimensionally poor if its level of deprivation • The marital status of the head is equal to, or greater than 33.33 percent. A house- hold with a deprivation level between 20 percent • Educational attainment of the head and 33.33 percent is vulnerable to multidimensional poverty or is at risk of finding itself in such a situa- • Characteristics of the house, including wall tion. Households with a level of deprivation greater material, type of roof, and type of floor than or equal to 50 percent are in a situation of extreme multidimensional poverty. • The type of energy used for cooking The poverty ratio, H, represents the proportion of • The type of energy used for lighting households in multidimensional poverty: • Ownership of household assets including radio, TV, video player, stove, refrigerator or freezer, washing machine, sewing machine, computer, where q is the number of households in multidimen- internet equipment, car, air conditioner or fan, sional poverty and n is the total number of house- motorcycle, landline telephone, mobile phone, holds. and bicycle. Poverty intensity (A) reflects the proportion of com- The marginal R2 of this model was 0.57 and the ponent-weighted indicators in which, on average, conditional R2 was 0.62. The district-level pov- poor people experience deprivation. The higher the erty rates and standard errors associated with the level of deprivation (the value of A), the more intense resulting estimates are reported in Map 3. the poverty. For poor households only, the sum of 27 Molina and Rao 2010. Household consumption was log-transformed. Sampling weights are applied in the consumption prediction model following 44 Guadarrama et al. (2018), as described in Skarke and Kreutzmann (2020). Household size was used as additional weights to aggregate poverty headcount estimates across households, using the EMDI plus package available at: https://github.com/SSA-Statistical-Team-Projects TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Table A1.3: Variables of the MPI Modalities of the Indicator Variable Modalities of the questionnaire variable variable 7. Protected well; 8. Unprotected wells; 9. Protected spring water; 10. Unprotected spring water; 1: No access to 11. Surface water; drinking water 12. Rainwater; 13. Tanker truck; No access to drinking 14. Water seller; Source of water supply 15. Bottled water; water 16. Other 1. Running water at home; 2. Individual faucet in the course; 3. Common faucet in the course; 0: Otherwise 4. Public fountain/pump terminal; 5. Drilling; 6. Human motor pump wells 1. Charcoal; 2. Firewood; 1: Unclean cooking 3. Oil; method Cooking fuel Type of cooking fuel 6. Dung; 8. Other 4. Gas; 0: Otherwise 5. Electricity 1. No installation/In nature; 1: No access to 5. Latrine with wooden platform, earth; adequate sanitation 6. Pit; facilities Type of sanitation 7. Other Sanitation facilities 2. Latrine with chair (flush or not); 3. Latrine without chair (flush or not); 0: Otherwise 4. Toilet with smooth concrete platform, porcelain, fiberglass 2. Kerosene lamp/Petromax; 3. Candle; 1: No electricity Electricity Light source 4. Tallow/Seeds; 8. Other 0: Otherwise 1. Electricity (Grid, group, solar, wind) 1.Bare soil/ Earth/ Sand; 2. Stem/Leaf/Bamboo; 1: Rudimentary 3. Mat; materials 4. Rudimentary board; Type of floor 8. Other 5. Parquet/Waxed wood; 0: Otherwise 6. Cement; 7. Vinyl, Tile, Carpet 4. Bozaka; 1: Rudimentary 5. Stem/Leaf; materials 6. Recovered material; Type of roof 8. Other Housing 1. Tile; 0: Otherwise 2. Sheet metal; 3. Cement/Fibro-cement 2. Clay/Uncooked brick; 3. Stem/Leaf; 4. Sheet metal; 1: Rudimentary 5. Board; materials Type of wall 6. Bozaka; 7. Recovered material; 8. Other 0. Cinder block, stone; 0: Otherwise 1. Baked brick 45 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Table A1.4: MPI dimensions and indicators MPI dimension Indicators weight Health No access to drinking water 1/6 = 16.66% (Weight=1/3) One or more children under the age of 18 died in the household 1/6 = 16.66% No individual aged 11 and above completed five years of 1/6 = 16.66% Education schooling in the household (Weight=1/3) At least one school-age child (6-10 years) does not attend to 1/6 = 16.66% school in the household Unclean cooking fuel 1/15 = 16.66% No access to adequate sanitation facilities 1/15 = 16.66% Living No electricity 1/15 = 16.66% standards (Weight=1/3) Dwelling with a floor or roof or wall made of rudimentary materials 1/15 = 16.66% The household does not own a car, van, or similar motor vehicle, but owns at most one of the following goods: bicycle, motorcycle, 1/15 = 16.66% radio, refrigerator, telephone or television Total 100% the deprivation level will be related to the total number of poor households. where c corresponds to the level of deprivation suf- fered. The MPI value is the product of two measures, the multidimensional poverty ratio (H) and the poverty intensity (A). IPM = H * A The MPI reflects the level of deprivation of the poor supported by the entire population. IPM = 46 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment 8. Additional tables and graphs Table A1.5: Poverty by characteristic of the household head Contribution to poor Characteristic Poverty rate (%) Population share (%) population (million) Overall 74.1 19.0 100 Area Urban 31.2 1.5 19.3 Rural 84.4 17.5 80.7 Household size 1-3 members 71.4 4.1 22.3 4-6 members 71.6 9.0 48.9 7 or more members 80.5 5.9 28.8 Gender Female 73.8 3.6 19.0 Male 74.2 15.4 81.0 Age group 15-24 years 79.7 1.8 8.7 25-54 years 74.1 13.7 72.1 55-65 years 70.7 2.4 13.1 66 years or more 73.4 1.1 6.0 Disabled Yes 82.5 0.1 0.6 No 74.1 18.9 99.4 Open defecation Yes 95.7 10.2 41.4 No 58.9 8.8 58.6 Ever attended school Yes 66.6 12.8 74.9 No 96.4 6.2 25.1 Literacy status Illiterate 96.7 6.5 26.0 Literate 66.2 12.6 74.0 Formal education None 96.4 6.2 25.1 Some 66.6 12.8 74.9 Educational level Primary 83.5 9.3 60.2 Secondary 45.8 2.8 33.3 Technical or higher 16.6 0.2 6.5 Employment status Employed 75.8 17.7 90.9 Unemployed 51.1 0.1 0.2 Searching for work 58.7 0.1 0.6 Other 57.4 1.2 8.4 Employment sector Primary 89.2 15.9 76.4 Non-primary 32.5 1.8 23.6 Migration status Migrant 43.6 2.0 17.8 Non-migrant 80.7 17.0 82.2 47 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Migration reason Employment 41.4 0.85 62.3 Studies / apprenticeship 10.9 0.02 5.9 Famine/ floods 82.2 0.02 0.8 Insecurity 67.1 0.01 0.6 Family 50.4 0.44 26.4 Health 43.8 0.01 0.5 Housing 20.0 0.01 1.3 Other 29.2 0.02 2.3 Migration reason and location Urban areas Employment 17.8 0.164 28.1 Studies / apprenticeship 7.1 0.012 5.0 Famine/ floods 50.0 0.002 0.1 Insecurity 30.2 0.001 0.1 Family 21.6 0.062 8.7 Health 25.3 0.002 0.2 Housing 15.4 0.001 0.2 Other 22.9 0.003 0.3 Rural areas Employment 60.8 0.68 34.2 Studies / apprenticeship 32.1 0.01 0.9 Famine/ floods 86.6 0.02 0.7 Insecurity 76.8 0.01 0.4 Family 64.4 0.37 17.7 Health 53.7 0.01 0.3 Housing 21.0 0.01 1.1 Other 30.3 0.02 2.0 Marital status Single/never married 71.0 1.0 5.6 Married/co-habiting 74.1 15.4 81.0 Separated/divorced 78.2 1.5 7.4 Widowed 71.8 1.1 6.0 Marital status and sex Women Single/never married 75.8 0.73 3.7 Married/co-habiting 69.9 0.78 4.4 Separated/divorced 77.5 1.20 6.1 Widowed 71.3 0.89 4.9 Men Single/never married 61.7 0.30 1.9 Married/co-habiting 74.4 14.63 76.6 Separated/divorced 81.4 0.27 1.3 Widowed 73.8 0.22 1.1 Marital status, sex and location Women in urban areas Single/never married 38.7 0.09 0.9 Married/co-habiting 29.9 0.09 1.1 Separated/divorced 38.2 0.13 1.4 Widowed 30.5 0.09 1.2 48 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Men in urban areas Single/never married 31.5 0.06 0.7 Married/co-habiting 30.0 1.04 13.5 Separated/divorced 39.2 0.02 0.2 Widowed 29.0 0.02 0.2 Women in rural areas Single/never married 87.8 0.64 2.8 Married/co-habiting 84.2 0.69 3.2 Separated/divorced 89.0 1.07 4.7 Widowed 84.6 0.80 3.7 Men in rural areas Single/never married 79.2 0.24 1.2 Married/co-habiting 83.8 13.60 63.2 Separated/divorced 90.5 0.25 1.1 Widowed 84.8 0.20 0.9 Source: Authors’ calculations based on 2018 Madagascar Census. Table A1.6: Stunting across national, rural, and urban areas in 2018 and 2021 Stunting National Rural Urban 2018 41.6 42.6 37.7 2021 39.8 40.5 35.5 Source: Demographic and Health Surveys (2018, 2021). Table A1.7: Full regression analysis—Determinants of household welfare Observations PROBIT Secondary Secondary Major Urban Major Urban VARIABLES Urban Rural Urban Rural Centers Centers Centers Centers Dependent variable Poor Age of HH head -0.078 0.026 -0.006 0.102 0.032 0.050 -0.056 -0.027 -0.020 -0.118 -0.070 -0.056 Age of HH head 0.0115* -0.002 0.001 -0.016 -0.006 -0.005 Squared -0.007 -0.003 -0.002 -0.014 -0.009 -0.007 HH Size -0.390*** -0.345*** -0.364*** 0.624*** 0.692*** 0.747*** -0.044 -0.023 -0.016 -0.096 -0.056 -0.055 # of Adult Male in HH -0.040 -0.002 0.006 0.083 0.010 0.011 -0.027 -0.014 -0.009 -0.059 -0.034 -0.028 # of Adult Female in 0.018 0.0234* 0.012 -0.069 -0.012 -0.022 HH -0.025 -0.014 -0.011 -0.059 -0.034 -0.030 # of Children in HH -0.003 0.004 -0.006 0.032 -0.007 -0.003 -0.012 -0.006 -0.005 -0.028 -0.015 -0.013 # of Elderly in the HH 0.043 0.028 -0.0546*** -0.151 -0.044 0.0948* -0.046 -0.028 -0.018 -0.119 -0.063 -0.054 HH head is female -0.050 -0.013 0.0297* 0.069 0.057 -0.110** -0.044 -0.026 -0.018 -0.103 -0.058 -0.051 HH Head is Married -0.0995** 0.0404* -0.010 0.112 -0.0845* 0.016 -0.040 -0.021 -0.015 -0.091 -0.050 -0.043 49 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Years of School of HH 0.008 0.000 0.001 -0.009 0.000 0.001 head -0.005 -0.003 -0.002 -0.011 -0.006 -0.005 HH Head is an -0.035 0.000 0.025 0.054 0.040 -0.076 employee -0.048 -0.025 -0.018 -0.109 -0.063 -0.052 HH Head is in Business -0.049 -0.003 0.010 0.144 -0.011 -0.032 -0.046 -0.024 -0.017 -0.109 -0.058 -0.050 # of Rooms in the HH -0.016 0.0107* -0.004 0.024 -0.021 -0.007 -0.015 -0.006 -0.004 -0.035 -0.014 -0.013 HH has a water pipe -0.142 -0.031 -0.014 0.180 0.084 -0.016 -0.143 -0.047 -0.038 -0.240 -0.126 -0.108 HH has Electricity 0.010 0.032 -0.0386** -0.057 0.041 0.124*** -0.042 -0.025 -0.017 -0.098 -0.060 -0.048 HH has Internet 0.274 0.027 -0.034 -0.177 -0.219 0.224 -0.181 -0.084 -0.065 -0.447 -0.232 -0.216 HH has No Toilet -0.033 0.007 -0.0285* 0.025 0.006 0.035 -0.044 -0.023 -0.017 -0.095 -0.058 -0.047 HH has a Flush Toilet 0.032 -0.018 0.0474* -0.116 -0.184* -0.152** -0.079 -0.037 -0.025 -0.209 -0.096 -0.073 HH is Accessible -0.031 0.025 -0.020 0.034 -0.001 0.031 -0.038 -0.021 -0.015 -0.088 -0.049 -0.042 # of farm asset owned 0.000 0.000 0.000 0.000 0.000 1.37e-10** 0.000 0.000 0.000 0.000 0.000 0.000 Land size 0.000959** 0.000 -0.000343*** -0.00222* 0.000 0.00112** 0.000 0.000 0.000 -0.001 0.000 -0.001 Price of rice 0.000 0.000 0.000 0.000 0.000 2.30e-07*** 0.000 0.000 0.000 0.000 0.000 0.000 Livestock Sales 0.000 8.15e-09** 0.000 0.000 0.000 0.000 Revenue 0.000 0.000 0.000 0.000 0.000 0.000 Crop Sales revenue 0.000 3.05e-10*** 0.000 0.000 -2.82e-08* 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Travel Time to nearest 0.000 0.000 0.000 0.000 0.000 0.000 city 0.000 0.000 0.000 0.000 0.000 0.000 Commune has Paved 0.089 0.023 0.0619*** -0.264 0.036 -0.0958* Road -0.070 -0.026 -0.020 -0.161 -0.067 -0.054 Commune has Frqnt 0.070 0.0547** 0.0659*** -0.243** -0.148** -0.175*** Transport -0.058 -0.026 -0.018 -0.119 -0.063 -0.049 Commune has 0.300*** 0.214*** 0.165*** -0.531*** -0.453*** -0.290*** Electricity Ntwk -0.079 -0.027 -0.025 -0.203 -0.069 -0.063 Commune has Pipe -0.312*** 0.0535** 0.0808*** 0.661*** -0.039 -0.148*** Water Ntwk -0.075 -0.024 -0.019 -0.184 -0.063 -0.050 Commune has telecom -0.183 0.003 0.0893*** 0.465* 0.027 -0.191*** Ntwk -0.123 -0.038 -0.017 -0.249 -0.095 -0.052 Price of Cereals in -0.024 -0.0538** 0.0646*** 0.278 0.048 -0.036 Commune -0.094 -0.026 -0.022 -0.204 -0.069 -0.056 Org fertilizer avg price -0.189* 0.018 -0.106*** 0.152 0.009 0.125* -0.110 -0.028 -0.027 -0.245 -0.067 -0.073 50 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment Non-org fertilizer avg 0.317*** -0.140*** -0.004 -0.373 0.253*** -0.013 price -0.119 -0.032 -0.028 -0.255 -0.072 -0.074 Pesticide avg price -0.255*** -0.028 -0.0988*** 0.422** 0.057 0.225*** -0.088 -0.023 -0.019 -0.183 -0.057 -0.056 Water flow all year -0.004 0.005 0.022 0.040 0.075 -0.058 -0.114 -0.026 -0.015 -0.286 -0.063 -0.044 Avg salary for male -0.00103*** 0.000 -4.30e-06** 0.00230*** 0.000 0.000 0.000 0.000 0.000 -0.001 0.000 0.000 Avg salary for female 0.00102*** 7.72e-06*** 1.45e-05*** -0.00233*** -1.46e-05* -2.67e-05*** 0.000 0.000 0.000 -0.001 0.000 0.000 Vakinankaratra -0.356*** -0.120*** 0.752*** 0.130* -0.095 -0.027 -0.269 -0.069 Itasy -0.138** -0.107** 0.160 -0.011 -0.054 -0.047 -0.120 -0.129 Bongolava -0.0992* -0.227*** -0.064 0.620*** -0.060 -0.032 -0.139 -0.102 Haute Matsiatra 0.111 -0.274*** -0.273*** -0.108 0.036 0.416*** -0.084 -0.084 -0.048 -0.200 -0.204 -0.150 Amoron I Mania -0.210*** -0.382*** 0.256* 0.707*** -0.057 -0.040 -0.135 -0.140 Vatovavy -0.0934* -0.223*** 0.030 0.586*** -0.049 -0.030 -0.115 -0.111 Ihorombe 0.114** -0.127*** -0.413*** 0.216* -0.056 -0.041 -0.135 -0.114 Atsimo Atsinanana -0.015 -0.033 -0.036 0.136 -0.073 -0.045 -0.159 -0.133 Atsinanana -0.045 -0.055 0.188*** 0.130 0.221 -0.159 -0.080 -0.063 -0.039 -0.189 -0.196 -0.109 Analanjirofo 0.306*** 0.233*** -0.697*** -0.372*** -0.047 -0.038 -0.117 -0.103 Alaotra Mangoro 0.142*** -0.260*** -0.270** 0.625*** -0.051 -0.040 -0.117 -0.133 Boeny -0.074 0.214*** 0.205*** 0.152 -0.449*** -0.129 -0.082 -0.063 -0.052 -0.189 -0.166 -0.147 Sofia 0.256*** 0.017 -0.514*** 0.061 -0.045 -0.035 -0.112 -0.098 Betsiboka -0.028 -0.115*** -0.036 0.433*** -0.050 -0.043 -0.131 -0.140 Melaky 0.384*** 0.0893** -0.811*** -0.169 -0.059 -0.042 -0.141 -0.106 Atsimo Andrefana -0.606*** -0.260*** 0.833*** 0.190** -0.063 -0.040 -0.162 -0.094 Androy -0.484*** -0.401*** 0.540*** 0.652*** -0.060 -0.055 -0.176 -0.204 Anosy -0.365*** -0.187*** 0.495*** -0.044 -0.051 -0.066 -0.126 -0.130 Menabe 0.030 -0.0525* -0.157 0.136 -0.043 -0.030 -0.106 -0.084 Diana 0.047 0.070 0.033 -0.158 -0.246** 0.089 -0.089 -0.049 -0.038 -0.210 -0.120 -0.104 Sava 0.322*** 0.052 -0.722*** 0.069 -0.046 -0.037 -0.115 -0.105 51 TOC Chapter 1 Madagascar Two decades of poverty stagnation against a modest growth performance Poverty and Equity Assessment 51o.Atsimo Andrefana - - Constant 15.36*** 14.42*** 14.55*** -1.516*** -0.232 -0.346** -0.153 -0.079 -0.053 -0.364 -0.194 -0.154 Observations 2291 5567 8664s 2291 5567 8664 R-squared 0.157 0.293 0.247 Source: Author’s calculations based on 2022 EPM data. 52 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment Chapter 2 Low agricultural productivity and market access trap rural households in a poverty cycle Key findings Agriculture is the backbone of Madagascar’s rural Limited market access and increasing price vari- economy. With more than 80 percent of the pop- ability have also compromised profitability in the ulation involved in farming, agriculture is the main agricultural sector. Large rice price differences source of livelihoods in rural and some urban areas. across the regions show weak market integration, However, agriculture only constitutes 25 percent mostly due to prohibitively high transport costs of Madagascar’s GDP. Subsistence farming is prev- made worse by dilapidated roads. Only 11.4 percent alent and explains why 90 percent of agriculture of the rural population has access to good road net- workers are poor. Only 27 percent of farmers prac- works, as a result, only 61.2 percent of farmers who tice commercial agriculture and 17 percent produce sell their production travel to market to sell them cash crops. Smallholder farmers (who on average and fetch higher prices. Unsold produce is stored cultivate on less than 1 ha. of land) produce 45 to 90 for future sales, to consume at home or use as seed. percent of the commercialized vegetables, peanuts, However, half of this unsold produce is either stored sugar cane, fruits, vanilla, coffee and spices (some in suboptimal conditions (on house roofs) or not of which rank among the best worldwide). Still, rice stored at all, leading to post harvest losses of up to accounts for 70 percent of total agricultural pro- 35 percent. duction, 40 percent of cultivated land, and all of the irrigated land. The average national annual rice 1. The link between agricultural yield is approximately 2.5 t/ha, similar to other East productivity and poverty African countries, but very low compared with the major rice-producing regions in Asia. Farming, fishing, and forestry are the foundations of Madagascar’s economy. Agriculture is the pri- Low agricultural productivity is a key contributor mary source of livelihood for about 80 percent of to the stubbornly high rural poverty rate. Agricul- the country’s population. Seventy-eight percent of tural labor productivity in Madagascar is significantly households in the country practice agriculture, with lower than the Sub-Saharan African average and has 71 percent raising livestock and 18 percent involved dropped by 31 percent since 1991. The low quantity in fishing. Mixed crop-livestock systems are the and quality of inputs broadly explain low productiv- most common type of farming (practiced by 61 per- ity. First, mechanical inputs such as irrigation systems cent of farmers), followed by crop-livestock-fish (10 and tractors are unaffordable for most farmers: only percent) and mixed crop-fish systems (1.5 percent) an estimated 5 to 10 percent use tractors and power (INSTAT, 2021a). Subsistence farming is prevalent, tillers, irrigation accounts for 40 percent of cultivated with only 27 percent of farmers practicing com- areas, but technology is mostly obsolete. Second, mercial agriculture and 17 percent focusing on cash chemical inputs such as improved seeds and fertiliz- crops. However, smallholder farmers are responsible ers are not used by most farmers, yet they are signif- for the majority of commercial agricultural produc- icant determinants of yield levels. Only 7.4 percent of tion in the country: they largely sell their crops to farmers use fertilizers, 4.5 percent use pesticides, and the market except for cash crops such as vegeta- 11.9 percent use improved seed types. Third, land and bles, peanuts, sugar cane, vanilla, coffee, fruits, and skilled agricultural workers are limited. Production spices—with the latter four crops ranking among the growth in recent years was not driven by increased country’s top agricultural exports – are largely sold land productivity but by farmland expansion through to traders. Only 2.1 percent of households practice slash and burn. Inequitable distribution of land and beekeeping, but the share rises to 10.6 percent in weak ownership rights exacerbate the problem. At the Androy region. Small-scale inland fishing is the the same time, few agricultural workers have formal most common type of fishing, practiced by 10.9 education or technical training. Finally, farmers lack percent of households, followed by fish-farming in access to finance (to purchase better inputs) and ponds (4.8 percent) and rice-fish farming (3.8 per- insurance, which further block the adoption of more cent).28 productive technologies. The latter two types of fishing are starting to gain ground in the regions of Vakinankaratra, Itasy, and Amoron'i Mania. 28 53 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment Although agriculture is a major contributor to Table 8: Multidimensional poverty is highly the economy, sustained growth in the sector has correlated with agricultural activity proved to be elusive. Agriculture accounts for 70 percent of total employment, but its shares of total Poverty Population Number Activity Group of poor rate (%) (Million) exports (37 percent) and GDP (25 percent) are rel- (Million) atively small. Agricultural labor productivity is sig- Rice 84.9 18.4 15.7 nificantly lower than the Sub-Saharan African (SSA) Other cereals 86.8 12.8 11.1 average and has dropped by 31 percent since 1991. Tubers 87.5 15.7 13.8 Despite the abundance of productive land and Legumes 86.4 10.5 9.1 Agriculture Industrial water resources and significant untapped potential, crops 86.7 7.8 6.8 the agricultural sector has failed to drive structural Cash crops 86.9 4.6 4.0 economic transformation and growth in Madagas- Other fruit & 84.3 8.0 6.7 car to the same extent as elsewhere in Africa and vegetables the wider developing world. Low productivity, vul- Cattle 87.2 10.6 9.2 nerability to shocks, poor post-harvest techniques, Small 92.4 2.4 2.2 ruminants limited storage and logistical capabilities, ecosys- tem degradation, and underdeveloped road net- Livestock Pigs 81.3 6.4 5.2 Poultry 82.2 16.8 13.8 works that impede farmers’ access to markets are Bees 88.2 0.6 0.6 among the critical bottlenecks hindering the sec- Silkworms 87.2 0.1 0.1 tor’s performance and growth prospects. Inland small- 89.8 3.0 2.7 scale fishing Rice is the main crop grown in Madagascar, Small-scale 85.0 0.6 0.5 accounting for 70 percent of total agricultural marine fishing production, 40 percent of cultivated land, and all Fishing Marine 84.6 0.1 0.0 aquaculture the irrigated land. High annual rainfall and abun- Fish farming in dant freshwater resources provide favorable con- ponds 82.4 1.5 1.2 ditions for water-intensive rice production in Fish farming in 83.0 0.2 0.1 various regions, including the rainfed central high- cages lands, the irrigated and terraced lowlands, and the Source: Authors’ calculations based on 2018 Madagascar semi-flooded coastal zones. About 80 percent of Census. agricultural households in Madagascar grow rice, average deprivation rate as measured by multidi- largely for subsistence, on small holdings, and rely mensional poverty rate among those working in the on family labor. Most such households, however, primary sector at 88.7 percent, way higher than the suffer from chronic food insecurity, as low produc- average population rate. Irrespective of horticul- tivity prevents them from diversifying their diet. ture, livestock, or fishing, poverty rates among pop- Despite the high availability of water, rice-farming ulations engaging in these workstreams are very productivity in Madagascar is on par with the levels high. As a result, almost unsurprisingly, there is very of other SSA countries, significantly lower than the little variation in poverty rates across the specific world average, and even lower compared with East activities people engage in. Poverty is high among Asia. Small-scale schemes account for 86 percent those who engage in farming cash crops (86.9 per- of land under controlled water management, while cent), and almost equally high among those who medium-scale schemes account for 13 percent are engaged in rice farming (84.9 percent) or other and large-scale schemes for 0.4 percent. Insuffi- fruits and vegetables (84.3 percent). From a policy cient maintenance and weak management capacity perspective, this speaks to very low returns in these among user groups often plague irrigation schemes, sectors as well as potential shortcomings of value resulting in unreliable water supply, insecure land chains whereby producers obtain substantially less tenure, limited extension and input supply net- than middlemen who might be the ones selling works, poor connectivity to downstream markets, products to markets, the next distributors. and low farm gate prices, which discourage farmers from using improved, productivity-enhancing tech- The link between agriculture and poverty in rural nologies such as high-yielding seeds, fertilizers, and Madagascar is complex and multifaceted, with a machinery. The rice sector’s performance has been wide range of factors contributing to the chal- hindered by stagnant yields and structural deficien- lenges faced by rural communities. Understand- cies, as well as recurrent negative shocks. ing these challenges is important for implementing effective interventions to improve the livelihoods of Albeit some variation, all primary sector occupa- rural communities and reduce poverty. This chapter tions are highly affected by poverty (Table 8). The aims to explore the link by examining the obstacles 54 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment that farmers face and the interventions that can be Figure 37: Agricultural productivity is below peer implemented to address these challenges. We aim countries and declined in the last 30 years to provide insights into the complexities of the issue Land productivity and offer recommendations how to improve the (in constant 2004-2006 US$ per hectare) livelihoods of rural communities and break the stub- 3000 bornly high poverty levels in rural areas. Agriculture 2500 is the backbone of livelihoods in Madagascar. 2000 Increasing agricultural productivity and crop prices 1500 can dramatically improve welfare for rural house- holds. Rice price has an important link to rural pov- 1000 erty, as demonstrated by the strong correlation 500 shown in Table A1.7 in the previous chapter. The table illustrates that the price of rice is strongly con- - 1990 2000 2010 2015 nected to the poverty status of rural households. Additionally, simulations conducted as part of a Madagascar Rwanda Tanzania technical report for an agricultural project in Mad- Uganda Bangladesh Cambodia agascar have indicated that even small changes in Labor productivity productivity and prices could result in substantial (in constant 2004-2006 US$) increases in production and reductions in imports, 900 ultimately leading to improvements in welfare. For 800 instance, a 10 percent increase in rice productivity 700 can lead to a corresponding 10 percent increase in 600 rice production (equivalent to 287 thousand tons of 500 milled rice) if markets function optimally and prices 400 remain at import parity levels. Due to these reasons, 300 rice value chain policies in Madagascar are often 200 seen as strategic for poverty reduction. 100 - Increasing agricultural productivity is key to reduc- 1990 2000 2010 2015 ing poverty and promoting economic growth in Madagascar Rwanda Tanzania developing countries. Higher productivity expands Uganda Bangladesh Cambodia the availability of food for both consumption and sale. This enables farming households to save money Source: IFPRI, 2019. and allocate more labor to other industries, push- ing up their income. In turn, rising incomes in rural or if they know about the latter, they may lack the areas boost demand for a wider range of products, financial means to acquire the necessary inputs or leading to growth in processing, packaging, trans- to shoulder the high initial cost of their implemen- portation, trading, and other non-farming activities. tation. Moreover, farmers in Madagascar often face Ultimately, urbanization, non-farming activities, and a range of market failures in relation to inputs, out- non-agricultural income become major drivers of puts, finance, and insurance, which exacerbate the development, with both urban and rural consumers cost and risk of adopting more productive technol- driving demand for agricultural products through ogies. value chains that connect rural areas to cities and towns. Limited use of inputs and poor technological adoption among smallholder farmers are the main Agricultural productivity in Madagascar is lower reasons for low Total Factor Productivity (TFP) in than in comparator countries (Figure 37). While land Malagasy agriculture. We take a closer look at pro- productivity (i.e., total output relative to total agri- duction efficiency and output growth to under- cultural area) has remained low but steady over the stand Madagascar’s agricultural productivity deficit. past 30 years, labor productivity (i.e., total output Specifically, we calculate TFP—an indicator of how relative to the number of persons economically efficiently agricultural land, labor, capital, and mate- active in agriculture) has been deteriorating sig- rials (agricultural inputs) are used to produce crops nificantly. Multiple factors might explain this trend: and/or livestock (agricultural output)—as the ratio of for example, farmers may use ineffective or out- total agricultural output to total production inputs. dated production practices as they lack knowledge When more output is produced from a constant of more productive methods and technologies; number of resources (i.e., resources are used more 55 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment Figure 38: While TFP and outputs grew util 2010, mate change also pose significant threats to agri- they fell thereafter cultural productivity and livelihoods in Madagascar. 5.0 Deforestation, soil erosion, and declining soil fertility 4.0 are major environmental challenges that negatively 3.0 impact crop yields and limit the ability of farmers 2.0 to earn a decent income. The effects of climate Growth (%) 1.0 change, such as increased frequency and sever- ity of droughts in the South and floods caused by 0.0 repeated cyclones on both the East the Center, also -1.0 have a significant impact on agricultural productiv- -2.0 ity, rural livelihoods, and poverty. -3.0 1991-2000 2001-2010 2011-2015 2. Agricultural inputs and low TFP growth (%) Output growth (%) agricultural productivity Source: IFPRI, 2019. The equity and efficiency of land use efficiently) TFP increases (Figure 38), shows that the growth rate of agricultural TFP in Madagascar Two land property rights regimes operate in Mad- had remained steady between 1991 and 2010, then agascar: an official but highly circumscribed titling dropped until 2015—a time when the country was system, and a more common unofficial system. in the midst of a deep political crisis. Output grew, In the latter system, access to land (with the pos- especially between 2001 and 2010, but it then fell sible exception of village commons) is controlled between 2011 and 2015. Madagascar's agricultural not by the community but largely by individuals. In productivity deficit is due to an inefficient use of areas with greater commercial activity, such as Lac resources, as indicated by the decline in the TFP Alaotra, land can even be sold to outsiders without growth rate and output growth during the period approval from traditional authorities. In this system, 1991–2015, particularly after 2010. however, land ownership claims continue to draw their legitimacy from communal institutions, some Agricultural productivity is higher in certain areas of which were invented or adapted for precisely this of Madagascar that enjoy access to irrigation and purpose (Jacoby and Minten 2006). proximity to markets. The green revolution in Asia and Latin America relied heavily on a combination Access to land is unequally distributed across of improved seeds, fertilizers, irrigation, and farm- gender and age, and in the wake of population ing education that allowed for higher yields of crops growth is increasingly coming at the expense of such as wheat, rice, and maize. In Madagascar too, deforestation. As a finite resource amid a growing yields benefit when inputs are readily available. For population, land is shared among fewer households example, in Alaotra-Mangoro—the country’s sec- (Map 6). Households headed by women are less ond-most important rice-producing region, with likely to own land than those headed by men, with a more than 136,000 hectares of land dedicated to larger gap for younger cohorts. This provides indica- this crop – annual rice yields reach up to 4 tons per tive evidence of unfair inheritance rights (Figure 39), hectare (t/ha), versus the national average of 2.5 t/ and the overall lower availability of land to younger ha. Access to functional irrigation and proximity to generations. Importantly, this measure observes the both Antananarivo and Toamasina, the two larg- extensive and not the intensive margin, hence not est urban centers in the country, have been key to showing by how much the average plot size may achieving higher yields.29 have decreased but simply that fewer households are reporting acccess when compared to 2008. Other structural constraints are also at the root of Nonetheless, in some regions, higher access rates low agricultural productivity. The country's rural are reported than previously, hinting at important population is also characterized by high levels of deforestation to increase access to land (for farm- illiteracy and limited access to education, health- ing). Atsimo Adrenafana as well as Melaky, but also care, and other basic services. These factors create Sava, have a higher share of households reporting a vicious cycle of poverty, where poor health, lack access to land in 2018 compared to 2008. Simul- of education, and limited opportunities perpetu- taneously, particularly in areas of Atsimo Adre- ate economic hardship and further exacerbate the nafana appear to show important forest cover loss. challenges faced by rural communities. In addition What’s clear from the forest cover loss map is that to the socio-economic challenges faced by rural even in areas that report decreases in access to land communities, environmental degradation and cli- (i.e. Analanjirofo), latter are much smaller than one 29 Source: Ministère de l’agriculture et l’élevage, Division Statistique Agricole. 56 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment Map 6: Agriculture land use is widespread and is increasingly unsustainable Land ownership by regions (households where at least one member owns land usable for agriculture) 2008 2018 Forest Cover Lost Value 0% 100% 0% 100% -0.21 to 0.00 -1.82 to -0.21 Diana Diana 60% 49% -3.37 to -1.82 Sava Sava -6.79 to -3.37 75% 83% -18.94 to -6.79 Sofia Sofia 86% 80% Boeny Boeny Analanjirofo Analanjirofo 63% 58% 88% 82% Betsiboka Alaotra- Betsiboka Alaotra- Melaky 84% Mangoro Melaky 76% Mangoro 84% 74% 84% 74% Analamanga Analamanga Bongolava 56% Atsinanana Bongolava 57% Atsinanana 89% 75% 79% 72% Itasy Itasy 91% 90% Vakinankaratra Vakinankaratra 95% 81% Menabe Menabe 78% Amoron’i Mania 75% Amoron’i Mania 89% 85% Vatovavy Vatovavy Matsiatra Fitovinany Matsiatra Fitovinany Ambony 86% Ambony 80% 85% 85% Ihorombe Ihorombe 87% 79% Atsimo- Atsimo- Atsino Atsino Andrefana Andrefana Atsinanana Anosy Atsinanana Anosy 72% 79% 92% 87% 93% 70% Androy Androy 94% 86% Source: Author’s calculations based on 2018 Census data. Figure 39: The gender gap in agricultural land would expect in the wake of population growth, and ownership diminishes with age this is likely due to further encroachment in those Women Men regions also. Southern regions of Toliara and Fia- own don’t own own don’t own narantsoa are at the forefront to bear the brunt of 100% drought risks and are more exposed to floods due to long coastlines. Given the importance of forest 40% 50% for livelihoods, including protection from soil ero- 75% sion, the short-term solution to encroach on land is 72% 82% creating untenable living conditions for the future in 50% an already harsh environment. 25% The EPM survey for 2021–22 shows that land culti- vated by households is primarily owned by its users 0% and largely allocated to growing rice. More than urban rural urban rural half of arable land is used to grow rice (Figure 40), with the remaining 45 percent largely dedicated to women men tubers, cash crops, and other cereals. Approximately 80% 77% 74% 72 percent of the land allocated to rice and other 70% 69% 63% 63% cereals is owned by those who farm it, 20 percent 60% 54% is used by farmers at no cost, and the remainder is rented. Similarly, 66 percent of the land used to 43% 40% grow legumes, tubers, and vegetables is owned by those who farm it, 20 percent is on a free loan, and 7 percent is rented. The land most likely to be owned 20% by its farmers is that used to grow cash crops (such as vanilla) and fruits, often for export. Only 13 per- 0% cent of such land is on a free loan, and 2 percent is 18-24 24-35 35-45 45-49 rented. Source: Author’s calculations based on the 2018 Census data. 57 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment More than 70 percent of farmers across the con- Figure 40: Rice occupies over half of arable land sumption distribution own their land, 22.7 percent use it on a free loan, and 6.3 percent are tenants. 3% In the first quintile of the consumption distribution, 6% 74 percent of landholders own their plots, 5 percent pay rent, and 21 percent use borrowed land without 8% paying for it. The shares are similar among landhold- ing households in the richest quintile: 77 percent are owners, about 17 percent use borrowed land 9% for free, and 6 percent pay rent. The average land 55% area held is smallest among households in the first quintile and tends to grow among those in higher quintiles (Table 9). 17% The correlation between employment status and land area cultivated varies across the consump- tion distribution. Notably, households in the richest Rice Tubers Cash crops Other cereals quintile have less land on average than those in the Other crops Legumes Vegetables Fruits fourth, as members of the former are more likely to have formal jobs in the industrial or services sectors. Source: Author’s calculations based on 2021–22 EPM data. Indeed, Figure 41, that welfare tends to increase with employment. Many households in the third and Table 9: Most agricultural households own their fourth quintiles have self-employed heads, usually land active in farming—thus, they make major use of Land Tenancy, Owned Free land for agriculture. On the other hand, households Quintile area (%) loan (%) sharecropping in the fifth quintile have the highest percentage of (hectare) or other (%) heads in salaried employment. Most of the poor Poorest 0.20 74.1 21.0 4.9 work in agriculture and produce their own food, 2 0.27 65.2 27.8 7.0 while wealthier households are comparatively less 3 0.30 71.6 21.7 6.8 active in this sector—and when they operate in it, 4 0.50 66.8 26.3 6.9 they tend to engage in commercial farming of cash Richest 0.35 77.3 16.7 6.0 crops and rice. Average 0.30 70.6 23.0 6.4 Labor requirements by crop Source: Author’s calculations based on 2021–22 EPM data. To diversify their risk exposure, households grow a Figure 41: Self-employed household heads tend range of crops that entail varying degrees of labor to cultivate larger land areas, relative to wage intensity. Figure 42 shows the number of laborers employed heads, except for the richest 20 percent involved in key farming tasks across various crops, broken down into family and non-family members 1800 in agricultural enterprises. Preparing the land for 1600 Land area (Hectares) planting takes almost as much effort as harvesting, 1400 while intermediary tasks such as weeding, applying 1200 fertilizers, and spraying chemicals are less labor-in- 1000 tensive. In most cases, families tend to their crops with minimal involvement of non-family members. 800 600 Rice requires both the highest number of workers 400 overall, and the largest amount of non-family labor. 200 The number of non-family members involved in rice 0 farming is high across all production phases; hence, 1 2 3 4 5 rice is mainly grown by wealthier households who can afford paying for non-family labor (Figure 42). Consumption quintile Although households across income quintiles seem wage_employment self_employment to produce a variety of agricultural products such as rice, tubers, legumes, vegetables, and fruits, there all_unemployed are relatively few differences in crop diversification Source: Author’s calculations based on 2021–22 EPM data. 58 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment Figure 42: Rice utilizes more labor than any other crop 5.0 4.6 4.4 4.5 Number of laborers 4.0 3.5 3.0 3.0 2.5 2.6 2.5 1.9 2.0 2.0 1.6 1.4 1.5 1.0 0.5 - Family Non-family Total Family Non-family Total Family Non-family Total Soil preparation Weeding, etc. Harvest Other cereals Tubers Legumes Vegetables Fruits Other crops Cash crops Rice Source: Author’s calculations based on 2021–22 EPM data. Table 10: Agricultural households diversify production across crops Other Other Cash Quintile Rice Tubers Legumes Vegetables Fruits Total cereals crops crops Poorest 85.0 67.2 77.3 79.2 91.3 70.3 69.5 71.8 79.2 2 88.1 81.3 79.3 89.3 83.6 95.0 66.2 80.6 84.1 3 89.3 83.8 80.9 87.4 93.2 76.4 82.5 86.8 86.1 4 90.4 85.3 82.2 88.3 89.6 74.9 89.8 80.3 86.8 Richest 90.9 86.5 83.8 89.3 91.4 91.5 81.8 82.3 87.9 Average 89.0 81.2 80.5 87.1 90.0 82.5 78.1 81.7 85.0 Source: Author’s calculations based on 2021–22 EPM data. between income groups (Table 10). This suggests Fewer than 12 percent of households report using that there may not be significant barriers to entry in improved or certified seeds. Only 15.1 percent use agriculture based solely on income levels (Table 10). organic fertilizer products, mainly manure, while 3.4 percent resort to inorganic fertilizers. A mere 17 Fertilizer, chemicals, and improved seeds percent apply chemicals to their crops (12.5 percent use pesticides; 2.6 percent, herbicides; and 1.2 per- Malagasy farmers make little use of modern inputs cent, fungicides). Notably, only 12 percent of farm- such as chemical fertilizers, pesticides, and certi- ing households use new seeds to start their crops fied seeds. Imports of chemical fertilizers (mainly (Table 11). NPK and urea) increased by about 6 percent per year between 2016 and 2020, from 42,000 tons Table 11: Fertilizer and other input use is minimal to 58,000 tons per year.30 Almost all domestically across the welfare distribution (% households) made inorganic fertilizer (about 130,000 tons of Improved ammonium sulfate per year) is produced by the Quintile Fertilizer Pesticides or certified Total Ambatovy mining company for export.31 Another seeds firm, GUANOMAD, sells about 2,500 tons of Poorest 5.5 2.4 10.6 7.6 guano on the domestic market and 1,500 tons to 2 6.2 2.8 10.6 7.9 international buyers every year.32 According to the National Strategy for Fertilizer Use Development 3 7.1 4.5 11.9 9.1 (MINAE, 2006), the government targeted dou- 4 8.2 6.0 12.6 10.1 bling rice production through the intensive use of up to 125,000 tons of chemical fertilizers by 2030. Richest 10.0 6.7 14.0 11.5 Annual consumption of chemical fertilizers across all Weighted 7.4 4.5 11.9 9.2 crops almost tripled between 2006 and 2020, from average 20,000 to 58,000 tons, but remains well below Source: Author’s calculations based on 2021–22 EPM data. target.33 30 COMTRADE, ITC, retrieved October 2021. 59 31 Ambatovy website, retrieved August 2021. 32 The Malagasy government, in collaboration with Morocco, also has started to build a 200,000-ton chemical fertilizer plant in Alaotra (NPK and Super-Phosphate), but the factory was not yet operational as of the end of 2021. 33 COMTRADE, retrieved November 2021. TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment There is a positive correlation between household Table 12: Most households produce their own income and use of fertilizers and pesticides, sug- agricultural inputs gesting that cost may be the key constraint to their Improved wider adoption. Moreover, supply-side research Quintile Fertilizer Pesticides or certified Total reveals that 40 percent of inputs are self-produced, seeds which explains the strikingly low proportion of Poorest 5.5 2.4 10.6 7.6 households using improved or certified seeds. Few 2 6.2 2.8 10.6 7.9 farmers buy new seeds, while the rest use those 3 7.1 4.5 11.9 9.1 stored from their previous harvest. Approximately 4 8.2 6.0 12.6 10.1 29 percent of households use inputs sourced on Richest 10.0 6.7 14.0 11.5 the market, while 10 percent rely on inputs supplied Weighted by other farming households. Table 12 shows that average 7.4 4.5 11.9 9.2 self-production of inputs is a function of poverty: Source: Author’s calculations based on 2021–22 EPM data. households in the lower quintiles of the income dis- tribution are more likely to produce their own inputs, ducer is from it, SFA helps to identify local potential and less likely to buy them on the formal market. and efficiency levels to construct the typology. 3. Agricultural potential and This approach allows the comparison between esti- efficiency: the productivity gap mated agricultural potential and efficiency levels under current conditions and hypothetical invest- Stochastic frontier analysis (SFA) is used to esti- ment scenarios, which can be used to calculate the mate productivity, agricultural potential gap, and agricultural profit gains linked to each case. These farm efficiency following Maruyama et al. (2018). results can be extrapolated at the regional level for The approach utilizes SFA to estimate smallhold- the whole country combined them with GIS data on ers’ agricultural potential under optimal conditions local agroecological conditions, topography, and and compares it with their current performance to road infrastructure to construct a typology to assess assess their efficiency levels. SFA allows the econo- where investments in agriculture would be more metric exploration of the notion that, given fixed effective in bringing rural households out of poverty, local agroecological and economic conditions in a and how different types of investments can increase region and the occurrence of random shocks that rural household income through an increase in the affect agricultural production, the decisions made profitability of smallholder agriculture. A schematic by farmers and policymakers translate into higher or illustration of this approach is illustrated in Figure 43. lower production, revenues, and profits. Inefficiency is then defined as the loss incurred by operating The core component of the typology construction away from an ideal production frontier. By estimat- generates region- and district-specific agricul- ing where this frontier lies, and how far each pro- tural potential and efficiency estimates (Map 7). Figure 43: A conceptual framework to understand macro and micro drivers of agricultural efficiency Prediction of changes in the agricultural potential and e ciency levels between the baseline scenario and hypothetical scenarios: Example 1: Rural road improvement program Potential: Maximum production given external that converts all environmental factors tertiary and feeder roads in rural areas into secondary grade roads Biophysical environment: Climate, topography, E ciency: Ability to attain that potential through rainfall, etc. optimal decision-making at all levels Example 2: Rural electrificaiton expansion program Policymaker: Farmer: Economic environment: • Provision of extension services and trainings • Use of best practices that brings universal Market prices, travel time • Price information systems • Investment decisions access to electricity to to markets • Access to credit and crop insurance • Technology adoption all households in rural areas Source: Maruyama et al., 2018. 60 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment The agricultural potential map generated by this Map 7: Northern and Eastern regions have the component shows differences in crop production highest agricultural revenue potential potential across different regions of Madagascar. (a) Regions The estimated potential is low in regions such as Farm potential at region level 4,291,798 to 7,777,114 Anosy, Androy, Melaky, and Diana, which are vul- 3,772,830 to 4,291,798 nerable to droughts. On the other hand, regions in 3,486,774 to 3,772,030 the east—such as Atsinanana, Atsimo Atsinanana, 3,245,672 to 3,486,774 and Sava—have high agricultural potential due to 1,734,899 to 3,245,672 (b) Districts frequent tropical storms and higher rainfall volume. Missing Farm potential at ditrict level The Boeny and Itasy regions also have high poten- 4,338,604 to 7,777,114 tial, thanks to their soil type and a dry tropical cli- 3.783,204 to 4,338,604 3,505,423 to 3,703,204 mate that is suitable for jujube trees, baobabs, and 3,121,413 to 3,505,423 mango trees (Map 7).34 The Sava region has the 1,521,656 to 3,121,413 highest potential farm revenue, with a value of MGA Missing 7,777,114 per year. This may be due to the region's favorable climate and soil conditions for grow- ing high-value crops such as vanilla, coffee, and cocoa. Other regions with relatively high farm rev- enue potential include Atsinanana (MGA 5,576,601), Menabe (MGA 4,139,082), Boeny (MGA 4,415,828), and Analanjirofo (MGA 4,016,485). On the other hand, some regions have much lower farm revenue potential compared to the rest. For example, the Anosy region has the lowest potential farm revenue with a value of MGA 1,734,899 per year. This may be due to the region's arid climate and limited arable land. Melaky (MGA 1,973,229) and Anosy are other regions with relatively low farm revenue potential. Overall, the differences in farm revenue potential across the regions can be attributed to a range of factors, including climate, soil conditions, available infrastructure, access to markets, and the types of crops grown. Additional differences are further noted at the district level to help better target inter- ventions. Unlocking agricultural potential depends on the region's potential and requires different interven- Source: Author’s calculations based on 2021–22 EPM data. tions. A one-size-fits-all approach is unlikely to be effective. Policymakers need to tailor their inter- on their unique potential and needs. This can help ventions to suit the specific needs of each region. to ensure that resources are used effectively, and In areas with high agricultural potential, the focus that agricultural development is sustainable and should be on providing farmers with the essential inclusive, leading to long-term growth and poverty resources they need to maximize productivity, such reduction. as quality inputs, access to credit, and reliable exten- sion services. Investments in these areas are likely to 4. Agricultural productivity: the produce relatively quick and tangible results, leading case of rice to increased yields, incomes, and economic growth. In contrast, low potential areas may require more Most farmers across Madagascar grow rice, the substantial investments in infrastructure, such as country’s staple food. Rice is grown by 80 percent irrigation systems and roads, to support agricultural of agricultural households; it is cultivated over an development. These areas may also require longer- area of 1.3 million hectares—79 percent of its pro- term investments in research and development to duction is irrigated, 8.4 percent is rainfed, and 13 identify suitable crops and technologies that can percent comes from slash-and burn systems. The thrive in the local environment. Therefore, targeted average national annual yield is approximately 2.5 policies are needed to ensure that investments are t/ha, similar to other East African countries, but directed towards the most appropriate areas, based very low compared with the major rice-producing The farm revenue potential values provided are in Malagasy ariary (MGA) per year. 34 61 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment regions in Asia. Yields are highest in the irrigated National rice production has been falling since lowlands (between 3.9 and 4.3 t/ha), and much 2009. Both rice production and consumption lower in the rainfed uplands (from 1.8 to 2.7 t/ha). increased between 2000 and 2010 (Figure 44). Alaotra-Mangoro, Sofia, and Vakinankaratra are the Abundant domestic output meant that imports most productive regions. were low during that period and decreased from 2005 to 2010. However, a major structural break Smallholder farmers keep less than 1 hectare of rice in the 2010 farming season resulted in falling pro- fields on average. Rice is typically grown following duction and consumption from 2011 to 2016. In traditional practices, either in rotation with maize, 2017, when demand for rice exceeded the domes- vegetables, or legumes, or in double cropping tic supply, imports soared and have remained rela- where irrigation systems are available. Only an esti- tively high since, with the exception of 2020 when mated 5 to 10 percent of households use agricul- the pandemic adversely affected trade. Although tural machinery such as tractors and power tillers, domestic production increased in 2020, it has not mostly in Alaotra, while zebu cattle are commonly recovered to the levels of the previous decade. used for tillage in other regions. Better-off farming households tend to hire labor for transplanting and Both poverty rates and rice yields are highly cor- harvest. related with access to markets in rural areas. Panel (a) of Figure 45 shows a significant negative cor- Production is essentially aimed at local markets. relation between the Rural Access Index (RAI) and Domestic consumption of rice per capita ranks poverty at the district level, while panel (b) Figure among the highest in the world, with an average of 45 illustrates that rice yields decrease as the dis- about 100 kg per year (2012-17). Rice accounts for tance from markets increases. Market access incen- about 50 percent of the total caloric intake of the tivizes the adoption of improved farm management average Malagasy. Supply chains are short, typically practices and technology, which are necessary for including only one or two intermediaries—collectors increasing yields. Improving agricultural produc- and wholesalers. Small or medium-sized process- tivity and market access in rural areas is essential ing units exist at farm level, and rice collectors use for reducing poverty and food insecurity in Mada- large hulling machines. The main products include gascar. Achieving these goals requires investment white rice, parboiled or converted rice with greater in modern farming techniques and equipment, nutritional value, husked and polished rice for luxury improved infrastructure, and policies that support consumption and export, husked rice for local con- local farmers and promote fair trade.35 sumption, and dry grains ground into flour or sem- olina. Rice bran, broken rice, and straws are fed to Use of inputs in rice production pigs and poultry, while rice husks are used to manu- facture organic fertilizers and as fuel. The allocation of land to rice production varies by region. Most domestic production is concentrated Figure 44: Domestic rice production systematically exceeds consumption, but production has fallen since 2009 5 4 3 Millions 2 1 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 0 0 0 0 0 0 0 0 0 0 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 RICE PRODUCTION (TONS) RICE CONSUMPTION (TONS) RICE IMPORTS (TONS) Sources: World Bank, World Development Indicators Database; United States Department of Agriculture. 35 The RAI measures the proportion of the rural population who live within 2 km of an all-season road. 62 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment Figure 45: Rice yields, poverty and connectivity are highly correlated (a) Poverty and Rural Access (b) Yields and Distance from Markets 1.0 4.5 4 0.8 3.5 3 0.6 Yield in t/ha RAI (0 to 1) 2.5 2 0.4 1.5 0.2 1 0.5 0.0 0 0.0 0.2 0.4 0.6 0.8 1.0 0 20 40 60 Poverty headcount (0 to 1) Distance to market in km Source: Author’s calculations with World Bank (2021). in the Central Highlands, particularly around the significant determinant of rice yields. Notably, the capital city of Antananarivo. This region is known for utilization of improved seed for rice cultivation was its fertile soils and high rainfall, ideal for rice cultiva- four times greater than for maize. The government tion. In coastal and southern regions of the country, of Madagascar, in collaboration with international where conditions are not as favorable, less land is organizations and local NGOs, has implemented typically allocated to rice. According to online data several programs aimed at increasing the adoption from FAOSTAT, approximately 1.6 million ha of land were consistently used for rice production between Figure 46: Improved rice seed use remains low 2019 and 2021, which equals a little higher than 50 Rice Maize percent of total arable land (3 million ha). 5 020 4 978 4 900 4 884 Fertilizer use is highest in rice cultivation. In 2020, more than 50,000 out of the 58,000 tons of fer- SEED (TONS) tilizer used in Madagascar were dedicated to rice production, with deployment by rice farmers up from 27,553 tons in 2015. Nitrogen-based fertiliz- ers accounted for 80 percent of total consump- 1 236 1 150 1 134 tion (Table 13), followed by phosphate- and pot- 915 ash-based products (approx. 10 percent each). Demand increased at similar rates for all types of fertilizer between 2015 and 2020. 2015 2016 2017 2018 YEAR The use of improved rice seed grew slowly between Source: Ministère de l’agriculture et l’élevage, Division 2015 and 2018 (Figure 46). The quality of seed is a Statistique Agricole. Table 13: Nitrogen-based fertilizer use has doubled since 2015 Parameter/sub-parameter (Tons) 2015 2016 2017 2018 2019 2020 1. Total fertilizer consumption (N+P, N+P+K) in Tn, Fc 27,553 42,092 47,446 49,818 54,576 50,279 1a. Consumption of nitrogen-based fertilizers (N, 21,267 33,463 37,957 39,854 44,039 40,076 FAOSTAT code 3102) in Tn, Fc1 1b. Consumption of fertilizers made from phosphate (P, 2,735 4,312 4,745 4,982 4,982 4,982 FAOSTAT code 3103) in Tn, Fc2 1c. Consumption of fertilizers made from potash (N, 3,550 4,317 4,745 4,982 5,555 5,221 FAOSTAT code 3104) in Tn, Fc3 Source : Ministère de l’agriculture et l’élevage, Division Statistique Agricole. 63 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment of improved seed varieties. These programs have to rice in Ihorombe than in Sofia, production in the focused on developing and distributing high-yield- latter region is significantly higher. This is a con- ing and disease-resistant rice seeds to farmers, par- sequence of regional differences in yields that are ticularly those in remote and disadvantaged areas. related to factors such as quality of soil and seed, However, yields remain low in these areas which access to irrigation, and climate. Notably, certain implies that there is still a need for further invest- high-production regions only have average yields ment and support in the rice sector, particularly in (Figure 47) indicating significant room to increase the areas of research and development, seed pro- their output. duction, and distribution. Rice yields have been stagnating around an aver- Rice yields by region age of less than 2.5 t/ha. Production growth in recent years was mostly driven by increased alloca- Rice production levels vary significantly across the tion of land. On the other hand, yield growth has regions. Across the country, rice production was been held back by the poor state of irrigation infra- low in 2017 and increased in subsequent years. structure and damaged by repeated cyclones and Between 2017 and 2020, the top producing region siltation (as most irrigation schemes lack functional was Alaotra-Mangoro, followed by Vakinakaratra sediment extraction at the head of the main canal). and Sofia, in the highlands. On the contrary, pro- duction was lowest in Androy, Ihorombe, and Bet- Rice marketing siboka (Figure 47). Rice production is highly seasonal, reflecting the Rice production tends to be higher where more inability of the sector to prices. The vast majority land is allocated to it, as in Alaotra-Mangoro and of rice farmers rely on a single harvest, usually in Vakinakaratra. However, this is not the case every- May and June, and sell most of their output imme- where. For instance, although more land is allocated diately (Minten et al., 2006; Barrett, 1996, 1997). Figure 47: Rice production in highland regions is significantly larger than elsewhere (average 2012 – 2015) 600000 4 Annual yield (tons per hectare) 3.5 500000 Annual production (tons) 3 400000 2.5 300000 2 1.5 200000 1 100000 0.5 0 0 nk oro ra la fia Bo a y la sy fo to a te en y at e ng tra a I’ na M ia in ky an a A a ts sy or a be re y na en au M av nd o g Va Sav av ts an an Ih bok M ab an at iro na Ita Be no ts a on ia A ndr fa na So an om Bo sia na g A el v ol A an ar D M nj ki an i m A in Va M ra A or A o ot o im m im la H A ts A ts A A Regions Source: Ministère de l’agriculture et l’élevage, Division Statistique Agricole. 64 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment Water access and temperature prevent second or Figure 48: Rice prices bottom up at mid-year third cropping seasons, which are common in other (a) Local and imported rice, 2005-20 rice-growing countries.36 The seasonal nature of rice 105 production and sales is reflected in various atten- 100 dant costs—e.g., storage—which prompt substantial seasonal fluctuations in prices (Figure 48). On aver- 95 (January = 100) Seasonal index age, between 2005 and 2020, seasonal rice prices 90 on the local market oscillated by 20 percent—peak- 85 ing in January and bottoming out in June, when most rice is sold (Minten et al., 2006). 80 75 More concerning, prices also vary on a regional 70 basis. Figure 49 shows how local rice prices across various regions compare with the price in the major Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec rice-producing area of Alaotra-Mangoro. Nota- bly, average rice prices are comparatively higher in Local Imported areas more distant from this major producing hub— including in southern regions such as Anosy (+18 (b) All rice (various years) percent), Androy (+26 percent), and Atsimo Atsi- 105 nanana (+15 percent), and in northern regions such as Diana (+24 percent) and Sava (+22 percent). This 100 implies that no arbitrage opportunities are being 95 (January = 100) Seasonal index exploited because of artificial market restrictions. 90 Large price differences across regions are con- 85 sistent with previous analyses showing a lack of 80 market integration in Madagascar. For example, using community-level rice price data, Moser et al. 75 (2009) estimated that the integration of markets 70 across regions was poor, mostly due to prohibitively Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec high transportation costs. The analysis showed that while markets were well integrated at the sub-re- gional level, rural market fragmentation was signif- 2005-10 2011-15 2016-20 icant at the regional and national levels. This pat- Source: IFPRI, 2022. tern has become more apparent in the past decade (Figure 50). Figure 49: Price variation is substantial across regions, reflecting lack of arbitrage Regional rice price differentials rose substantially between 2005 and 2020, as deteriorating road conditions contributed to weakening market inte- gration. Dorosh et al. (2022) analyze price differen- tials between various regions and Alaotra-Mangoro over time, splitting price data from the Observatoire du Riz (OdR) into five-year intervals. Such differen- tials increased over time, indicating that marketing costs across the country increased, in percentage terms, relative to the same costs in Alaotra-Mangoro (Figure 50a). For example, while rice prices in Itasy were 3.9 percent higher than in Alaotra between 2005–10, the difference increased to 6.8 percent from 2011–15, and to 13.5 percent between 2016– 20. A largely consistent pattern emerges when con- sidering absolute differences in price levels (Figure 50b). Among the regions analyzed, only one expe- Note: Difference in regional rice prices relative to price in rienced a decline in relative prices from 2016–20 Alaotra-Mangoro, 2005-2020. The scale (e.g., 0.1) on the x-axis indicates a 10 percent difference with the price in compared with 2005–10, while absolute differences Alaotra-Mangoro. declined in three regions. Source: IFPRI, 2022. Functional irrigated areas exist in Madagascar where multiple rice crops are harvested each year, but they only account for a small share of the 36 65 total land allocated to rice. TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment Figure 50: Regional price variation has widened in the last 5-10 years (a) Log (prices) (b) Price levels (in MGA/kg) Note: Difference in regional rice prices relative to price in Alaotra-Mangoro from 2005–10, 2011–15, and 2016–20 Source: IFPRI, 2022. and ports. However, their quality is generally poor 5. Marketing and post-harvest due to past underinvestment and under-main- management tenance. Only 11.4 percent of the rural population has access to good road networks (Map 8). This is Market access and marketed surplus lower than the previous estimate in 2006, which was 25 percent. Rural accessibility differs signif- Market access is a major constraint for agrarian icantly between semi-urban areas and the rest of communities in Madagascar. In Madagascar, lim- the country. RAI is estimated at greater than 80 ited transport connectivity is a common constraint percent in Toamasina, Mahajanga and Antsiranana across all sectors. The country possesses important Districts. In most rural districts, rural accessibility is transport infrastructure, including roads, railways less than 5 percent. Map 8: Rural access and transport costs are correlated Rural Access Index Transport costs to major markets (US$/ton) Note: The higher the RAI value, the greater the proportion of the rural population who live within 2 km of an all-season road. Source: World Bank, Madagascar Urban Transport Report, 2021. 66 TOC Chapter 2 Madagascar Low agricultural productivity and market access trap rural households in a poverty cycle Poverty and Equity Assessment Table 14: Most sales, except for cash crops, take place through markets Households/ Sales (%) Market Cooperatives Intermediaries Others Individuals Rice 61.0 21.3 1.3 15.9 0.4 Other cereals 59.5 12.0 1.1 27.1 0.3 Tubers 71.3 16.1 0.5 11.9 0.3 Legumes 74.3 9.6 0.7 14.5 0.9 Vegetables 76.8 10.6 1.4 10.9 0.4 Fruits 67.0 14.0 1.5 14.2 3.3 Other crops 45.6 25.3 5.5 21.4 2.2 Cash crops 21.8 17.8 6.0 53.8 0.7 Average 61.2 17.2 1.7 19.4 0.6 Source: Author’s calculations based on 2021–22 EPM data. Table 15: Farmers lack basic storage solutions for most of their crops Other Vegeta- Other Cash Storage type (%) Rice Tubers Legumes Fruits Total cereals bles crops crops Granary (inside 3.8 5.9 13.0 2.3 11.8 12.7 9.0 3.6 7.1 house) Granary 13.9 7.5 5.5 7.6 9.2 7.3 3.9 13.8 9.9 (outside house) Store 7.1 1.8 2.5 2.8 1.1 3.7 4.2 8.6 4.7 Shed 4.8 6.1 3.7 5.5 2.1 7.0 2.6 2.5 4.4 Roof of house 26.4 38.6 18.3 35.4 10.7 5.3 7.2 4.8 23.2 No storage 25.7 26.4 48.6 33.5 58.7 57.0 62.7 50.9 37.1 Other 18.2 13.8 8.4 13.0 6.5 7.0 10.4 15.8 13.6 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Source: Author’s calculations based on 2021–22 EPM data. Sales outlets vary across crop types. Vegetables, tions on food security and spoilage. In fact, early legumes, fruits, and cereals such as rice are gener- result from a seed system assessment in the South ally sold through markets, while cash crops such as of Madagascar estimated that 35 percent of har- vanilla are often sold to third-party intermediaries vest is spoiled due to poor storage.37 Spoilage is a (Table 14). Direct sales to households or individuals major problem when farmers do not have access are another major channel, while less than 3 percent to appropriate storage infrastructure to keep their of all fruits and vegetables are sold to cooperatives produce fresh for a longer period of time. This results and other buyers. in significant losses and reduces the availability of food for both the farmers and consumers. More- Storage dynamics over, limited storage infrastructure can also affect food security, particularly in regions where food Most agricultural produce in Madagascar is not production is the primary source of income for many stored properly. Unsold produce should ideally be households. Inadequate storage facilities make it stored for household consumption, or for replanting difficult for farmers to preserve their produce and in the following season. However, as shown in Table sell them at a later time when prices are higher. This 15, over half of unsold produce is either stored in can lead to a surplus of produce during peak harvest suboptimal conditions on house roofs, or not stored season, which lowers prices and reduces the income at all. More affluent households can store their of farmers. Additionally, limited storage infrastruc- unsold produce in granaries, placed either inside ture can also lead to a lack of diversity in the types the house (often used for tubers, fruits, and vege- of crops that farmers grow. When farmers do not tables), our outside (mainly used for rice, cash crops, have access to appropriate storage methods, they and vegetables). may be forced to focus on crops that can be sold quickly, rather than those that require longer-term Limited storage capacity will lead to spoilage, food storage. This can limit the variety of food available insecurity, and low food diversity. The limited stor- in the market, which in turn can affect the overall age capacity for farmers can have serious implica- nutritional value of people's diets. 37 The is a result of a presentation by Seed System at the Madagascar World Bank Office in June 8th as part of an ongoing study of emergency seed 67 system funded by the agricultural practice group. TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Chapter 3 As urban markets fail, urban poverty rises Key findings Urban poverty increased from 42.2 percent in 2012 Against the backdrop of failing labor markets, to 55.5 percent in 2022. This represents a 31.5 per- returns to education are declining. The coun- cent increase in a decade. In the capital city of Anta- try has made progress in primary and secondary nanarivo, poverty increased minimally, from 33.3 school enrollment rates, but net secondary com- to 34.8 percent, but in secondary cities it rose from pletion rates still remain below the low-income 46 percent in 2012 to 61.1 percent in 2022. While country average. Education is positively correlated rural consumption grew mostly among the poor, it with earnings and the returns to education, espe- dropped for virtually all urban households, particularly cially post-secondary, continue to be high, espe- those in middle deciles. Going further back, urban cially for men. However, regression analysis shows poverty increased in 2001-2005, 2010-2012, and that returns to education are significantly higher 2012-2022, which follow closely the 2002, 2009, and for older cohorts than for younger ones. This sug- 2020 crises. This confirms that urban areas are vul- gests that education is less likely to bring people nerable to political and external crises, whereas rural out of poverty fast, either because education has areas are more sensitive to weather-related shocks. lost quality or because there are so few jobs in the market that most educated people end up in low Even before the pandemic, urban markets suffered paid employment. a steady deterioration which ultimately resulted in higher monetary and multidimensional poverty. Lack of opportunities and low aspirations are In the last decade, declining economic opportuni- intertwined factors contributing to urban pov- ties and an unfriendly business environment eroded erty. Insufficient investment in education, health- private investment, firm productivity and economic care and urban infrastructure limits human capital growth. Market concentration further exacerbated development and economic activity and increases the situation as dominant firms maximized rents and poverty traps. Low real income reduces access to avoided competition. Market capture in key sectors finance and investment opportunities. The domi- such as telecommunications, petroleum and agricul- nance of the elite in political and economic spheres tural exports ended up increasing prices and worsen- creates a sense of hopelessness among the urban ing the quality of services for consumers. Ultimately, population, leading to diminished aspirations. Low urban households saw increasing deprivation in edu- education and weak social cohesion further hinder cation and living standards, partly due to inadequate individuals' ability to demand government services infrastructure and services, and to lack of employ- and participate in their communities, exacerbating ment opportunities. the poverty trap. The disproportionate impact of the pandemic on 1. Defining poverty in an urban urban households further increased urban poverty. context Despite a relatively low number of COVID-19 cases, the country experienced a deep recession, causing a Urban living conditions had been deteriorating significant contraction in GDP and income per capita. in the last decade and worsened further with the This resulted in an estimated 2.4 million people fall- COVID-19 pandemic. In 2020, GDP contracted by ing below the international poverty line. The gov- 7.2 percent and average income per capita fell by 9.8 ernment’s border closures and restrictions on public percent, with a disproportionate impact on urban gatherings helped control the spread of the virus but households. Job losses in sectors such as trans- negatively impacted urban households, particularly port, tourism, and trade, delayed salary payments, those dependent on trade, transport, hospitality, as and declining revenues in the informal sector hurt well as informal labor. The measures affected live- urban workers disproportionately. However, urban lihoods and the gradual recovery of incomes was centers had been suffering even before the pan- observed only after the reopening of the country's demic (Jarotschkin, 2023; OPHI, 2022; UNDP and borders in early 2022. OPHI, 2022). A deteriorating business environment, 68 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment where few firms dominate key sectors, discouraged out of 15 main assets across Madagascar.39 Own- private investment and depressed firm productiv- ership is even more limited in rural areas, where the ity and job creation. Since the early 2000s, high majority of households own less than two assets. taxes and government bureaucracy have impeded Overall, radio is the most commonly owned asset growth in the business sector in Madagascar (Cadot (61 percent of households), followed by mobile and Nasir, 2001). phones (47 percent), TV (22 percent), and bicycle (19 percent). Some spatial variation, even among Urban poverty rose significantly between 2012 rural areas is noticeable, with fewest assets owned and 2022, particularly in secondary cities, where it in Androy region, Toliara province (Figure 51). In jumped from 46 percent to 61 percent. This trend parallel, men-headed households own dispropor- stemmed from an overall drop in income among tionately more assets than women-headed ones. urban households, which was especially acute Inequities along gender and space particularly for those in the middle of the income distribu- hold for mobile phone ownership. The latter has tion. While declining consumption levels in urban been linked to higher important productivity gains, areas boosted urban inequality, rural consumption due to improved information on prices, and better increased, resulting in a slight reduction of rural ine- connectedness to relevant markets. Yet, mobile quality.38 phone ownership was very low in Madagascar in 2008 overall, except for the capital city. The share The increase in urban poverty has significant impli- of households with access to at least one mobile cations as the share of urban population is pro- phone improved across different regions in 2018. jected to increase further. According to the official Yet, the spatial pattern with more Southern regions definition of an urban area (INSTAT, 2021c) urbaniza- being disadvantaged in terms of access rates, both tion has been relatively slow in the country between in 2008 as well as in 2018, prevailed. At least 4 in 1975 and 2018, going from 16 to 19 percent. How- 10 households had a mobile phone in the central ever, by other accounts, the urban population could and northern regions, whereas 2-3 households in be close to 40 percent (Box 1). Indeed, the World 10 reported access to at least one mobile phone. Urbanization Report (United Nations, 2018) sug- In general, urban households were more likely than gests a definite upward trend where the population rural ones to own a mobile phone with households residing in urban areas increased from 7.8 in 1950 to headed by men more often having a mobile phone 37 percent in 2022. Against a backdrop of poten- than those headed by women (Table 16). And finally, tially faster urbanization than in other Sub-Saharan younger generations in the 18-24 age bracket African countries, urban poverty is set to become a reported lower access rates than older individuals pressing issue in Madagascar. (24 through 49). Asset ownership, and mobile phone ownership in particular, is inequitably distributed across space and gender. On average, households own only two Box 1: Urban and Rural definitions for poverty analysis Several definitions of what constitutes an urban area are used for analysis in Madagascar. The Institut National de la Statistique (INSTAT), which carries out the National Population and Housing Census, uses commune boundaries as reference,40 and defines as urban any commune with a population of at least 20,000. According to INSTAT, as of 2018, 19 percent of Madagascar’s 25.7 million inhabitants were urban residents. However, the Ministry of Territorial Development and Land Services (MTDLS), which oversees urban development, adopts a more granular definition of urban area in line with the EU-OECD standard. Using satellite imagery, the MTDLS defines cities as agglomerations character- ized by a continuity of buildings, with a population of at least 5,000, and density of at least 1,500 inhabitants per km2. Based on this definition, Madagascar’s estimated urbanization rate in 2018 was 30 percent. Finally, the World Bank’s World Development Indicators (WDI) define urban areas based on the criteria of the UN World Urbanization Prospects and estimate the country’s urbanization rate at 37 percent. These projections are based on the 1993 Census, which is considered less accurate. The poverty assessment uses the INSTAT definition, as it informed the construction of the sample for the 2022 EPM survey. For this reason, our estimates should be considered a lower boundary. 38 As discussed in Chapter 1, rural inequality decreased because the increase in consumption during this period was highest among the poorest rural 69 households and much lower in upper deciles. 39 List of main assets: Bicycle, computer, radio, sewing machine, mobile phone. 40 The Ministry of the Interior and Decentralization oversees the definition of the administrative status of communes. TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Figure 51: Asset ownership fell in urban and rural areas Asset Ownership (% Households) 70% 60% 50% 40% 30% 20% 10% 0% TV Radio Cellphone Fridge Computer Bike Motorbike Car Boat Oxcart 2012/3 Rural 2021/2 Rural 2012/3 Urban 2021/2 Urban Source: Author’s calculations using 2021/2 EPM data. Table 16: Asset ownership is heavily urban and male Sewing Head Mobile (%) Computer (%) Radio (%) Bicycle (%) machine (%) Overall Men 48.2 4.7 64.1 4.8 21.0 Women 43.9 3.7 50.0 5.4 11.0 Urban Men 77.2 16.4 77.3 10.3 23.5 Women 73.3 11.3 66.3 10.0 13.2 Rural Men 41.0 1.9 60.9 3.5 20.4 Women 34.0 1.2 44.5 3.8 10.3 Antananarivo Men 57.8 9.6 77.6 6.3 28.2 Women 53.8 9.1 68.1 7.2 11.7 Fianarantsoa Men 35.6 2.1 59.6 3.3 11.0 Women 33.7 2.0 45.9 3.9 6.5 Toamasina Men 43.7 3.5 63.8 5.6 22.7 Women 44.8 3.5 50.3 6.6 13.0 Mahajanga Men 47.8 2.6 59.9 3.4 15.3 Women 43.9 2.1 48.9 4.2 9.4 Toliara Men 38.7 1.7 41.6 2.8 17.2 Women 30.8 1.1 28.8 3.0 10.6 Antsiranana Men 66.1 3.9 71.5 6.8 28.3 Women 61.5 2.9 58.7 7.9 16.3 Source: Authors’ calculations based on 2018 Madagascar Census. growing concern, especially in medium and small Staggering changes in well-being and the cities. Urban transport services are mostly “infor- role of secondary cities mal” and loosely regulated.42 Solid waste manage- ment remains underdeveloped as most cities do not Deteriorating access to basic services and infra- have a sufficient supply of landfills, creating serious structure undermine the potential and well-being environmental and public health concerns.43 Access in urban communities. Urban areas have better to direct water, sanitation and electricity remains access to transportation and communication net- concentrated in the capital city and the six large works, including roads and telecommunication ser- cities (Ansiranana, Antsirabe, Toliara, Mahajanga, vices.41 This comparative advantage is the reason Toamasina, and Fianarantsoa).44 Both medium- and why the urban areas are often considered growth small-size cities are deprived of these amenities, poles since their superior amenities attract key making them comparable to rural areas (Table 17). industry around which linked industries develop Moreover, asset ownership declined in urban areas mainly through direct and indirect effects. How- in the last decade, possibly as households used their ever, deteriorating access to basic services are a assets to cope with income losses (Figure 51). 41 For instance, only 10 percent of the rural population has access to electricity, compared with 50 percent of the urban population. 42 Transport services are often overconcentrated where the demand is high, increasing traffic congestion in urban areas, while less populated, often 70 poor, areas or remote suburban areas are left unserved (Iimi, 2022; Iimi et al., 2022). Urban residents use different types of transport modes, such, minibus, also called “taxibe”, tuktuk and rickshaw bicycle (Iimi, 2022). These transport services are largely “informal” and loosely regulated. Because of limited urban space for road and the informality of the sector, their operations are generally inefficient and unsafe, though perhaps cheap. In Antananarivo where the people’s mobility is most constrained, half of the residents still just walk, not using any public transportation. An average trip in Antananarivo takes 46 minutes one way, twice longer than those in other secondary cities where people commute 15-25 minutes. Because of the lack of efficient and reliable transportation, people are currently missing opportunities to be paid better, which are concentrated in the center of the city. TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Table 17: There are wide gaps in access to services between large and small cities Capital City Major Cities Secondary Cities Rural Towns Electricity in dwelling* 33.64 31.19 31.22 30.79 Piped water in dwelling 5.03 4.01 3.69 3.04 Paved roads in community 86.47 73.09 41.98 15.27 Regular daily transport in community 89.19 70.30 47.59 21.16 Source: Author’s calculations using 2021/2 EPM data. *Connection to electricity includes direct connection, connection via neighbor, and solar panels. Although economic activity is concentrated in are coastal cities (70 percent of medium size cities). urban areas, most employment remains informal Soalala, Morondava, and Toliary on the west coast and precarious. While agriculture contributes to are projected to be amongst the total proportion approximately 25 percent of Madagascar’s GDP of built-up areas exposed to sea level rise. Maha- while employing more than 70 percent of the popu- janga, Toamasina and Manakara are also located lation (World Bank Group, 2022) urban areas benefit within coastal erosion zones and the latter is highly from a wider range of economic activities—includ- exposed to cyclone. Unfortunately, exposure to sea- ing manufacturing, construction, and services—that level rise will increase if global climate mitigation in principle offer more employment opportunities efforts fail to substantially reduce greenhouse gas and higher incomes. Indeed, cities with more than emissions. As Iimi (2019) shows, climate disasters 100,000 inhabitants host most private sector sal- can also adversely affect the economic outcomes in aried employment, as they are home to most large Antananarivo in a significant way. firms and MSMEs. Not surprisingly, about 75 per- cent of the country’s GDP is estimated to come from In contrast to rural poverty, which is sensitive to cities (IOE, 2023). Still, the urban labor market is weather shocks, urban poverty is sensitive to polit- largely informal, with only 12 percent of the country’s ical and external shocks. The urban poverty rate urban labor force being formally employed, often in rose from 42 percent in 2012 to a staggering 56 low-wage jobs that offer little social protection.45 percent in 2022, and even more in secondary cities, The unemployment rate is low, and the labor market where it soared from 46 percent to 61 percent participation rate (75.4 percent) is on par with com- (Figure 52). Urban poverty increased between 2001 parator countries—indicating that a significant pro- and 2005, from 2010 to 2012, and from 2012 to portion of the working-age population is econom- 2022, which corresponds closely to the 2002, 2009, ically active. However, most jobs offer low pay, and and the 2020 crises. In a way this means that urban poverty rates among low earners remain very high. areas are vulnerable to crises, but unlike rural areas that are vulnerable to weather shocks, urban areas Moreover, cities are vulnerable to natural hazards are vulnerable to political and most recently pan- that can lead to massive damages and the reversal demic-related crises. The impact of these shocks of developmental gains and coastal cities are espe- is amplified by poor urban planning and delayed cially exposed. Thirty percent of urban communes policy implementation. Figure 52: Secondary cities drive the expansion of urban poverty, 2001–22 Poverty headcount (%) 80 70 60 50 40 30 20 10 0 2001 2005 2010 2012 2022 National (%) All Urban (%) Capital City (%) Secondary Cities (%) Source: Author’s calculations using 2021/2 EPM data. 43 Antananarivo has only one landfill, Andralanitra, where 3,000 waste pickers live and was created in 1960. 71 44 Throughout this chapter we will distinguish major urban center from secondary urban centers. Major urban centers often include the capital cities with the six cities with more than 100,000 inhabitants and secondary cities refers to both medium- and small-size cities between 5,000 and 100,000 inhabitants. There circumstance in the chapter when Major urban centers are separated from the capital city and explicitly mentioned. 45 We define formally employed if an employee is registered with the national pension system CNAPS. TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Figure 53: Urban consumption fell throughout the distribution Annualized consumption growth rate (%) across consumption deciles 1.50 1.00 0.50 0.00 -0.50 -1.00 -1.50 -2.00 -2.50 -3.00 1 2 3 4 5 6 7 8 9 10 National Urban Source:Author’s calculations based on 2021–22 EPM data. The rise in urban poverty stems from a broad decline How do urban poor households distinguish in income among urban households. Between 2012 themselves from rural poor households? and 2022, consumption growth at the national level was low, and income gains were concentrated Urban poverty is linked to both household-level among the very poor in rural areas and those in the characteristics and community-level access to ser- wealthiest decile. While consumption grew in rural vices. A random forest analysis (Box 2) of poverty areas in a way that generally benefited the poor, it predictors is presented in ranking each indicator by dropped among urban households (Figure 53). This importance, from top to bottom (Figure 54). For all indicates that income losses were felt most acutely urban areas, household size and revenue from crop by urban households, particularly those in the mid- sales are the most important household-specific dle-income bracket, leading to an increase in the predictors of poverty, while access to paved roads, prevalence of urban poverty. piped water, the electricity grid, and mobile network coverage are crucial community-specific predictors. Urban poverty has become increasingly deep and As stated in Chapter 1, women’s casual wage rate in severe. Between 2012 and 2022, the urban poverty the community is an important predictor of urban gap increased from 14.6 to 20.6 (i.e., the average poverty. poor urban household lives on 20.6 percent less than the amount equivalent to the poverty line), and the urban poverty gap square went from 6.9 to Box 2: Predicting poverty through Random 10.1, implying increased poverty severity (Table 18). Forest Analysis Decreasing asset ownership in urban areas reflects the greater depth and severity of poverty. In fact, Using the 2021/2022 EPM data, the estimation urban households owned fewer televisions, radios, distinguishes between household-specific fridges, computers, bikes, and cars in 2022 than in characteristics that can be used for instance 2012 (Figure 51). Other factors such as the use of to inform the design of proxy means testing more advanced technical gadgets like smartphones for urban social assistance programs; and might have contributed to the observed changes, community-specific characteristics, whose but that applies to a very small percentage of the patterns can help prioritize investment policies. urban subgroup. The analysis encompasses all urban areas, and separate assessments for major and secondary Table 18: Poverty gap and poverty gap squared cities—with the latter emerging as hotbeds increased in urban areas of urban poverty. It is important to note 2012 2022 2012 2022 that the figures reported are not regression Poverty gap Squared poverty gap coefficients but express the relative importance of each indicator. All indicators are statistically National 33.4 32.5 18.7 17.3 significant determinants of urban poverty, but Urban 14.6 20.6 6.9 10.1 some have a greater effect/importance than Rural 38.1 35.3 21.7 19.0 others. Source: Author’s calculations based on 2021–22 EPM data. 72 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Figure 54: Demographic characteristics, infrastructure and economic opportunity are the main predictors of urban poverty Major Urban Centers Secondary Urban Centers All Urban Centers Head_Age_Sq Head_Married Hhd_Size Head Age Hh_Rooms Crop_Sales Hh_Adult Female Hh_Adult_Male Cty_Paved_Road Cty_Paved_Road Hh_Land Hhd_Size Hhd_Size Cty_Female_Salary Crop_Sales HH_Flush Toilet Cty_Frqnt_Transport Cty_PipeWater Service Cty_Paved_Road Cty_Electric_Network Cty_Electric_Network Cty_Electric_Network Livestock Sales Cty_Cellular_Network Cty Cellular_Network Cty_Cell_Network - 0.2 0.4 0.6 0.8 1.0 - 0.2 0.4 0.6 0.8 1.0 - 0.2 0.4 0.6 0.8 1.0 Importance Importance Importance Source: Author’s calculations based on 2021–22 EPM data. The poor who live in major cities have different fea- findings, a typical poor urban household has a large tures from those in secondary urban centers. Key number of members, low revenue from agricultural predictors of poverty in major urban centers—such activities, and limited access to basic utilities such as Antananarivo, Toamasina, Antsirabe, Tolear, Fia- as paved roads, water, and electricity. Higher female narantsoa, Mahajanga, and Antsiranana—include casual wage rate is also an important predictor household characteristics such as the age of the of urban poverty on average, highlighting the link household head, the number of adult males in the between women’s empowerment and urban pov- household, land size, household size, and whether erty. These results can help guide policymakers in the house has a flushing toilet (Figure 54). In con- designing more effective poverty alleviation pro- trast, the key household-specific predictors of grams and investment policies for different types of poverty in secondary urban centers (encompass- urban centers. ing small and medium-sized cities with a popula- tion between 5,000 and 100,000) are the marital Urban labor market dynamics and their status of the household head, the number of rooms poverty impact in the household, the number of adult women in the household, household size, and revenues from At the national level, most of the working popu- the sale of crop and livestock. The major communi- lation is in the informal sector, earning low wages. ty-specific predictors of poverty related to access Figure 55 shows that out of the estimated popu- to key infrastructure (e.g., transportation networks, lation of Madagascar (27.5 million), 2.3 percent are the electricity grid, and mobile cellular networks) child laborers, 3.2 percent are employed elders and were similar across all urban centers. beyond the age of 60 and 39.8 percent are workers aged 15 to 59 years. Among the 10,931,240 individ- In conclusion, household size, revenue from crop uals in the working age population who are working, sales, access to paved roads, piped water and the only 760,000 are registered in the national pen- electricity grid, and mobile network coverage are sion fund. This implies that only 7 percent of work- crucial predictors of urban poverty across all urban ers are formal. The rest are in the informal sector areas. Major and secondary cities exhibit different and generally underpaid. Indeed, an analysis of household-specific predictors but share similar the 2021/22EPM data shows that 99.8 percent of community-level characteristics. Based on these employed individuals are working poor.46 The work- An individual is classified as working poor if he or she is (a) employed and (b) living in a household with per capita consumption or income below 73 46 the poverty line. The working poverty rate is the proportion of working poor in total employment: working poverty rate = number of employed persons living in a household with per capita consumption or income below the poverty line/total employment * 100 (Pietschmann et al., 2016). TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Figure 55: Over half the population 6+ years old needs to work Source: Author’s calculations using 2021/2 EPM data. ing poverty rate is highest in the agriculture and hours was 30 percent, and 11.7 percent of work- mining sectors at 100 percent and lowest at 91.8 ers had two or more jobs. The Gini index for earn- percent for technical and scientific professionals. ings was 63.8, indicating high income inequality in 2021/2. Likewise, the urban labor market features high rates of informality and a significant gender gap in Low earners in urban areas are more likely to work employment opportunities. According to the World long hours than those in rural areas. Table 20 pre- Bank’s urbanization review (World Bank, 2021a) sents a comprehensive breakdown of the proportion as of 2021, trade accounts for the largest share of of low earners engaged in full-time or more work, urban employment, followed by services—such as transportation, restaurants, and retail—and manu- Table 19: Labor market outcomes show most facturing, which is dominated by textile. Unfortu- people work for very low wages nately, since wages in manufacturing such as textile Urban unemployment rate 8.2 are kept low, most urban poor do not make a decent Employment-to-working-age- living from urban employment since most jobs are 73.8 population ratio considered unskilled (Nicita, 2006). Working age population as a fraction 53.9 of total population The average urban worker earns significantly more than their rural counterpart, but wages remain low, Child labor rate 10.2 particularly in the informal sector. According to a Median earnings (MGA) 150,000 2020 report by the International Labor Organiza- tion, the median monthly wage for urban workers Median hourly earnings 1,226.6 in Madagascar was around MGA 200,000 (approx- Working poverty (% employed living imately US$48), compared with around MGA 99.8 in poor HH) 100,000 (approximately US$24) for rural workers Poverty rates among low earners 72.6 (ILO, 2020). However, most urban workers in Mad- agascar earn less than the national minimum wage, Poverty rate among the unemployed 61.8 currently set at MGA 200,000 per month. Share of low earners who have low 2.7 earnings due to short hours Data from the 2021/22 EPM survey shows a rel- Share of low earners who work long atively tight labor market, with the urban unem- 30 hours ployment rate at 8.2 percent. The employ- Share of non-low earners who escape ment-to-working-age-population ratio was 73.8 8.6 low earnings due to long hours percent, indicating that a large proportion of the Share of workers with two or more working-age population was employed (Table 19). 11.7 jobs However, most wage earners are considered low Share of workers with formal wage earners as they earn wage below the poverty 17.3 contract line after adjusting for the dependency rate —with Gini index for earnings 63.8 the poverty rate among low-earners reaching 72.6 percent. The share of low earners who worked long Source: Author’s calculations based on 2021–22 EPM data. 74 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Table 20: Close to half of urban low earners work Additionally, men (35.4 percent) tend to work longer full-time or more hours than women (23.9 percent), while workers Low earners who work full-time of aged between 25-54 years (32.7 percent) exhibit 29.9 the highest proportion of long-hour work, followed more (% low earners) Gender by those aged 55-64 years (26.3 percent) and 15-24 Male 35.4 years (25.0 percent). These outcomes suggest that there is a greater demand for low-skilled workers in Female 23.9 urban regions, mainly in the retail, hospitality, and Age group transportation sectors. 15-24 25.0 25-54 32.7 Poverty among the working-age population is 55-64 26.3 more prevalent in rural than in urban areas. Table 21 presents the poverty rate of the working-age Area of residence population in in 2021/2, disaggregated by employ- Urban 47.5 ment status and across urban and rural areas. The Rural 26.1 poverty rate for the entire working-age population Source: Author’s calculations using 2021/2 EPM data. is 69.3 percent, but it is much higher in rural areas (73.5 percent) than in urban areas (50.2 percent). Table 21: Poverty among the working-age population is higher in rural areas Among the working-age population, the poverty rate for the employed is higher than for the unem- Poverty rate of working-age population (%) ployed or inactive. In urban areas, the poverty rate Employed among the employed stands at 50.2 percent, versus Urban 50.2 45.2 percent and 44.0 percent, respectively, for the Rural 73.5 unemployed and inactive. This is explained in part by Total 69.3 the long working hours of employed people. In rural areas, the poverty rate among the employed stands Unemployed at 73.5 percent, which is a higher than among the Urban 45.2 unemployed and almost on par with inactive pop- Rural 68.7 ulation. These statistics suggest that while employ- Total 58.9 ment opportunities remain critical to reducing pov- Inactive erty, working people in Madagascar are so severely Urban 44.0 underpaid that they are often worse-off than those who do not work, especially in urban areas—i.e., the Rural 74.0 cost of supplying labor frequently exceeds the ben- Total 65.7 efit. Total working age population Urban 48.0 2. Key drivers of increasing urban Rural 73.5 poverty Total 68.1 Several factors explain the rise in urban poverty. Total population First, a long-standing decline in firm productivity Urban 52.9 has reduced opportunities in urban areas, which Rural 77.6 in turn suffer from lower returns to higher educa- Total 72.9 tion than rural areas. The COVID-19 pandemic has exacerbated this issue in the short term. Second, a Source: Author’s calculations using 2021/2 EPM Data. small elite controls a large share of the economy and fiercely resists competition, with a resulting drop in segmented by gender, age, and residential location. private investment to an all-time low of 17 percent The number of hours worked by low earners holds of GDP as of 2021. Such “elite capture” has created significance because it sheds light on the reasons a hostile environment for entrepreneurship and behind the persistent poverty among “wage earn- private sector growth, further narrowing the path ers”. The hours worked, along with an already low to prosperity for urban residents. Finally, a lack of income, contribute to their low standard of living. opportunities and limited social mobility have con- Interestingly, the statistics reveal a higher percent- tributed to forming low aspirations among many age of low earners working long hours in urban urban households, which can potentially aggravate areas (47.5 percent) than in rural areas (26.1 percent). urban poverty. 75 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment The disproportionate impact of the access to basic services, household employment, pandemic on urban households and the subjective well-being of households. The survey aimed to track the same households over Despite a comparatively modest number of COVID- time, with selected respondents (typically the house- 19 cases, the pandemic triggered a recession about hold heads) periodically completing a phone inter- three times deeper in Madagascar than in the rest view. Four rounds were conducted in the country. In of Sub-Saharan Africa. Export revenues and private the first round, 1,240 households were interviewed investment collapsed in 2020, as GDP contracted in June 2020. In August 2020, 1,580 households by 7.2 percent and income per capita by 9.8 percent. were surveyed—different from those surveyed in As a result, an additional 2.4 million people are esti- the first round. In the third and fourth rounds, 1,580 mated to have fallen below the international pov- and 1,345 households were interviewed in Novem- erty line in 2020, bringing the poverty rate to an all- ber 2020 and May 2021, respectively. The results of time high of 80.7 percent. Rising poverty reflected the survey are representative at the national level job losses in sectors such as transportation, tour- for households with access to a telephone. ism, and trade, delayed payments of salaries (part of which is often remitted to poor relatives in rural The onset of the pandemic caused a significant areas), and declining revenues in the informal sector. increase in unemployment worldwide, and Mada- gascar was no exception. According to the United The first three COVID-19 cases, all imported, were Nations, “by 2020, the global unemployment rate confirmed on March 20, 2020, prompting the dec- reached 6.5 percent, up 1.1 percentage points from laration of a state of emergency the next day. The the previous year.47 The number of people unem- state of emergency was lifted on October 18, 2020, ployed worldwide increased by 33 million, reaching after the end of the first wave which had reached 220 million.” In Madagascar, the unemployment rate its peak in July/August. The state of emergency rose from 2.5 to 12.2 percent in 2020, starting with was reinstated in early April 2021 due to the second the declaration of the first state of emergency and wave. It was lifted again on September 4, 2021. the associated restrictive measures. In June 2020 The epidemiological situation worsened during the alone, 7 percent of workers interviewed lost their second wave, as the Beta variant of the virus hit jobs. The loss of jobs was greater in urban areas Madagascar in March–April 2021—resulting in over than in rural areas. However, the rate of job loss, 600 new daily cases on average and peaking with which was high at the beginning of the pandemic, 854 new cases on April 14, 2021. The daily number tapered off with time (Table 22). of confirmed deaths dropped from a peak of 12.86 in December 2021 to 0.14 as of August 2022. Table 22: Urban households were disproportion- ately affected by pandemic-related job losses Major containment measures helped curb the spread Households that of COVID-19 but had a negative impact on the eco- experienced job June August November May nomic welfare of many households. The state of loss due to the 2020 2020 2020 2021 pandemic (%) emergency entailed the closure of borders, churches, Urban 15.3 8.5 0.9 0.8 and mosques; the prohibition of public gatherings; Rural 6.3 3.2 0.1 0.3 mandatory mask-wearing and handwashing; and All 7.0 4.4 0.3 0.5 curfews. While most measures were lifted along with Source: High Frequency Phone Survey 2020–21. the state of emergency, borders remained closed until March 2022. Unfortunately, the stringent meas- Malagasy households experienced a protracted ures affected livelihoods, especially among urban period of income losses, probably linked to the households reliant on activity in tourism, transport, closure of borders and its effects on the tourism and hospitality, as well as on informal labor. industry—especially hotels, restaurants, and trans- port. In both June and August 2020, a large major- To monitor changes in livelihoods, the government ity of households experienced a decline in income launched a series of High-Frequency Phone Sur- (Table 23). Although incomes had largely stabi- veys (HFPS). The HFPS were part of a wider initia- lized by November 2020, more than 30 percent tive launched by the World Bank and several coun- of respondents still reported a decrease in income tries to track changes in household welfare as the that month—likely due to the continued closure of pandemic evolved. The purpose was to provide a the country’s borders and its impact on tourism. real-time understanding of the effects of the pan- Incomes started to rise for a minority of households demic that could inform interventions and policy after November 2020, pointing to a recovery in responses. The HFPS aimed to obtain information economic activity. The country eventually reopened on dimensions including knowledge of COVID-19, its borders in early 2022. https://unstats.un.org/sdgs/report/2021/goal-08/#:~:text=COVID%2D19%20has%20led%20to,33%20million%2C%20reaching%20 47 76 220%20million. TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Table 23: Two-thirds of households lost income in pandemic. Layoffs were a survival mechanism for 2020 firms. Another coping mechanism, mainly adopted by large firms, was changing the product mix struc- Percentage of June August November May households (%) 2020 2020 2020 2021 ture towards mostly digital solutions. Most firms Increase 3.5 2.0 6.5 4.9 desired support from government in the form of Remained money transfers and tax reductions but only 33 per- the same 27.8 29.2 58.8 57.1 cent benefited. As a result, most resorted to letting Decrease 67.4 65.4 33.8 36.5 employees go and adjusting loan repayment terms. N/A 1.3 3.4 0.9 1.5 Source: High Frequency Phone Survey 2020–21. Deteriorating business environment leads to lost competitiveness The Business Pulse Survey (World Bank, 2022) shows that COVID-19 adversely affected business Urban poverty was already worsening before the activity. First, there was a drastic and significant pandemic. Although multidimensional poverty— drop in demand. Between May 2019 and 2020 an which encompasses 10 indicators across three cat- average firm lost 56 percent of sales. In turn, sales egories (health, education, and living standards)— revenues declined significantly (Figure 56). Second, decreased at the national level between 2008 and there were negative supply side effects such as 2018, it rose in urban areas when considering edu- input shortages, which were experienced by 10 cation and living standards. A potential reason for percent of firms. The most severely affected were this trend was the growth in urban population that smaller firms in the transport, construction, ware- was not accompanied by supply for adequate urban housing, and tourism industries. Up to 25 percent of infrastructure and services, accompanied by a 70 workers in these industries lost their jobs during the percent increase in the number of urban poor. Fac- tors that contributed to this trend include the high Figure 56: Micro firms were most affected by the fertility rate among the poor, a series of economic pandemic shocks that reduced incomes, and—in some cases— Change in revenues by firm size (%) the migration of rural poor towards cities, many of whom were fleeing frequent droughts in the South or repeated cyclones in the West. More fundamentally, urban poverty has been on the rise due to an ever more acute shortage of eco- nomic opportunities, as declining private invest- ment led to a deterioration in firm productivity. Pri- vate investment as a share of GDP has been steadily dwindling—from more than 25 percent in 2010 to less than 17 percent in 2021. This drop appears con- nected to a negative trend in firm productivity. Rea- sons for the decline over the past decade include inadequate infrastructure, limited access to finance, and weak institutions. While access to infrastructure such as electricity, water, and sanitation decreased, Change in revenues by sector (%) access to finance has remained exorbitant as real interest rates have remained above 30 percent over the past decade—well above the rates in compar- ator countries (Figure 57). This has resulted in low economic growth, which has not been sufficient to create the jobs and opportunities needed to lift the growing number of rural-urban migrants out of poverty. Increasing market concentration exacerbated the decline in firm productivity across manufacturing and services, and therefore the rise in urban pov- erty. An analysis of surveys of formal firms con- ducted in 2009, 2013, and 2022 shows that both Source: Madagascar Business Pulse Survey – 2021. values added per worker and Total Factor Produc- 77 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment tivity (TFP) deteriorated significantly over the past power to set wages below a level that adequately decade (Figure 58). The negative trend in produc- reflects the productivity of the labor force (Deb et tivity started before the pandemic but was exacer- al., 2022). Workers struggle to find alternative jobs bated by the lockdowns and other measures imple- (due to frictions in mobility across geographical mented to control the spread of COVID-19. This locations and industries) and are effectively captive. forced many formal firms to shut down or reduce In turn, low wages depress local demand. Sectoral their operations—with a likely increase in informality. dominance by one or few firms does not necessar- For example, the disruption of supply chains affected ily distort the wider labor market—especially if the commercial agriculture more than subsistence agri- sector is small relative to the size of the economy— culture, while turbulence in global demand mostly but it is more likely to do so in small, underdeveloped impacted trade-related sectors featuring higher economies such as Madagascar. Even when labor productivity and better-paying jobs. However, the markets are competitive, excessive market power underlying reason for lower private investment was can have a detrimental impact on society through the consolidation of key sectors, which constrains its effect on prices: high price markups can gener- the competitiveness of firms whose owners have ate rents for dominant firms, which then have an limited access to the country’s elites. Such consol- incentive to maintain their market power by setting idation creates an environment that favors a small barriers to entry—including by means of corruption. group of firms and makes it difficult for others to In addition, market capture exacerbates inequality compete, resulting in lower private investment and (Feldman et al., 2021; Khan and Vaheesan, 2016; overall productivity. Commanor and Smiley, 1975), with potential long- term consequences on political, social, and eco- Lack of competition can distort wages and prices, nomic stability. Overall, market capture by a handful reducing the potential for growth and hurting the of firms can have significant impacts on local econ- poor disproportionately. Dominant firms have the omies and communities. In Madagascar, for exam- Figure 57: Access to finance is hampered by exceedingly high interest rates 60.0 50.0 Real Interest Rate (%) 40.0 30.0 20.0 10.0 - (10.0) (20.0) 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 Bangladesh Madagascar Rwanda Tanzania Uganda Source: World Bank, World Development Indicators Database. Figure 58: Productivity among formal firms suffered a dramatic decline since 2009 Source: World Bank Enterprise Surveys, 2009, 2013, and 2022. 78 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment ple, the concentration of market power among a enrollment in primary school is commendable and few agricultural exporters curtails the bargaining generally falls in line or above those of peers, with a power of producers of key export products, such as rate of 96 percent. And when looking at enrollment vanilla and lychees. rate increases among the cohort of 12–18-year-olds, progress when compared to the previous cohort is Most sectors in Madagascar feature a small easily traceable across most districts in Madagascar number of dominant firms. The telecommunica- (Figure 60). tions industry is comprised of four major operators: Telma, Orange, Airtel, and Gulfsat (Blueline). Telma, Moreover, important progress has been made on the former telecoms monopolist, has a strong pres- secondary schooling enrollment also, increasing ence in all segments of the broadband value chain, overall enrollment rates from 17 to 30 percent. Yet, controlling international connectivity (along with when compared to its peers, net secondary school- Orange), the national backbone and backhaul net- ing rates remain below LIC average of 34 percent. works, and more than 60 percent of the last-mile And when looking at a map, it is apparent that pre- network. Similarly, four major players control petro- dominantly regions in the Center and North-East leum import and distribution, with Jovena—majority have benefitted from recent surges in secondary owned by Telma—holding the largest market share. schooling enrollment. In all other regions, progress Two companies—Galana Refinery and Petroleum across cohorts regarding secondary schooling has Logistics Company—dominate petroleum storage been limited (Figure 60). and transportation. The Groupement des Exporta- teurs de Litchi (GEL) sells more than 90 percent of Literacy is higher in urban areas and positively cor- Madagascar’s lychees to two European importers. related with average shares of population in sec- In the vanilla industry, Symrise holds a 70 percent ondary schooling. Figure 61 shows that districts market share in vanilla extract, and four export com- with higher education show higher levels literacy panies—most of them owned by a single conglom- overall. On the bright side, it is noteworthy that even erate—market about half of the country’s vanilla regions with the lowest levels of literacy have been beans. showing increases across cohorts. Proportionally, some of the largest improvements in literacy rates Declining returns to higher education in have been witnessed in regions with lowest total urban areas rates. At the same time, it’s startling to see how very low literacy and educational attainment are among Madagascar has made commendable improve- older cohorts, particularly Southern regions where ments in primary enrollment over time, accruing to literacy levels are as low as 30 percent among the cohorts across all of Madagascar’s regions. Mada- 26–64-year-old population. Nonetheless, with the gascar has made a lot of progress over a 14-15-year overall expansion in primary and secondary school- period in terms of increasing primary and second- ing, one might expect more improvements than ary enrollment rates (Figure 59). Madagascar’s net witnessed. Figure 59: School enrollment increased over time 97 96 95 70 67 100 90 86 School enrollment, secondary 90 82 81 60 School enrollment, primary 81 78 80 50 70 62 44 60 40 (% net) (% net) 34 50 30 30 27 40 30 20 17 20 10 10 0 0 0217 01 16 02 17 02 17 02 17 04 18 4 18 4 18 20 , 20 , 20 C, 20 , 20 , 20 , 20 A, 20 , 20 A, 20 20 20 00 , 20 00 , 20 , C G , , ,2 C ,2 G M M L I L I G D A W TZ A TZ GD GD LI C L I G D KH KH MD M RW R B B M D M Note: BGD Bangladesh, KHM Cambodia, MDG Madagascar, RWA Rwanda, TZA Tanzania, UGA Uganda, LIC average of low- income countries. In all cases the graph presents the latest available year. N/A not available. Source: World Development Indicators. 79 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Figure 60: Primary and secondary education have increased in most regions Population aged 12-18 Population aged 19-25 Population aged 26-64 90 Population with primary 80 70 education (%) 60 50 40 30 20 10 0 nk ga ra au ong sy t a vy I M a vi a im ho ny in be in na la ala ana M ofo ro y ts ia a nd ky na y M sy e na va en ro M lav tr i ok ab Fi an Be of at go Ita no A ela Sa na A ana fa ia n A om to oro sia nd ra njir ib A an Bo en ki ma S o ar D re an A M r A a to Va ala va n ts ts na I B n n te A o ot o m im A H ts A ts A A Va Secondary Population aged 19-25 Population aged 26-64 70 Population with secondary 60 education (%) 50 40 30 20 10 0 nk ga ra au ong sy t a vy I M a vi a im ho ny in be in na la ala ana M ofo ro y ts ia a nd ky na y M sy e na va en ro M lav tr i ok ab Fi an Be of at go no Ita la Sa na A ana fa ia n A om to oro sia nd ra njir ib Bo A an en ki ma S o Me o ar D re an A r A a to Va ala va n ts ts na A I B n n te A ot o m im A H ts A ts A A Va Source: Authors’ calculations based on 2018 Madagascar Census. Figure 61: Malagasy literacy increases with educational access; less so for French and English Malagasy French English 80 80 80 Population without formal education (%) Population without formal education (%) Population without formal education (%) 70 70 70 Rural Rural Rural 60 Urban 60 Urban 60 Urban 50 50 50 40 40 40 30 30 30 20 20 20 10 10 10 0 0 0 0 50 100 0 50 100 0 50 100 Malagasy literacy rate (%) French literacy rate (%) English literacy rate (%) Source: Authors’ calculations based on 2018 Madagascar Census. 80 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment More than 28 percent of Malagasies over the age of illustrates that returns to education are also gen- three have no formal education, and 44.4 percent erally higher for individuals from wealthier house- have only completed primary education, according holds, and for those living in households whose to the 2018 Census. Only 2.8 percent of the popula- heads are themselves more educated. Moreover, tion has university-level education. A gender gap in overall returns are higher in urban areas than rural. education exists, as men tend to be more educated Considering the total effect of education on earn- than women, but it is not very significant. Educa- ings (i.e., the combined effect of primary education tion levels are generally higher in urban locations. and more than four years of higher education), each Approximately 32 percent of the rural population year of schooling is associated with a 4.9 percent has no formal education compared to 13 percent increase in earnings at the national level and 5.6 in in cities. Of those living in the countryside only 47.3 urban areas, but only a 4.2 percent increase in rural percent have completed primary education, versus areas. In general, it is apparent that tertiary edu- 32.3 percent of city dwellers. Only 1.1 percent of cation makes a difference but its value in terms of the rural population areas have earned a university wage increase is decreasing as older people with degree, compared with 9.8 percent in urban areas. tertiary education earn more than younger with tertiary education while accounting for experience Years of schooling are positively correlated with through age. This means that that value of tertiary income, but the magnitude of the effect varies education is decreasing. The few jobs available in across education levels. Across all age groups, the market are being taken by the older and more individuals with higher levels of education tend to experienced workers. earn higher wages then their less-educated coun- terparts (Figure 62). However, the degree of remu- Returns to education in urban areas decreased neration by education level varies considerably across all education levels. Return on education in depending on employment type. For instance, edu- urban areas (Model 4 in Table 24) is unambiguously cation did matter for subsistence farmers but its higher for urban areas. This is the case for all educa- effect on wages is only significant up to the primary tion levels when using no education or incomplete school level. On the contrary, education really mat- primary education as the reference point. When ters in formal employment as salaries are highest for we interact education with age group to proxy for those with tertiary or university level training. These change in returns to education overtime, we note findings imply that while investing in education at that the interaction term of tertiary education com- all levels can lead to increased earnings potential, pletion with age group 50 – 64 is the only statis- higher education beyond the secondary level offers tically significant variable which is also lower than the best chance of achieving financial prosperity in for the entire sample under model 3. The remain- the long term. ing interaction terms are not statistically different than the reference interaction term – no education Returns to education tend to increase with the between 15 and 29 years old, implying a reduction level of education and are often higher for men in returns to education across both age categories than women. Table 24 shows the wage premium and education level in urban areas. associated with additional years of education. It Figure 62: Earnings barely increase with education levels Non-agriculture, 2021 25000 20000 15000 10000 5000 0 No education or Primary complete Secondary incomplete Secondary complete Tertiary/post primary incomplete secondary Employee Domestic Employer Self-employed Family Contribution Trainee Subsistence farm Source: Author’s calculations using 2021/2 EPM data. 81 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Table 24: Returns to education are higher among older cohorts Dependent variable: hourly earnings Population R2 adjusted 0.2391 0.2395 0.2415 0.2422 Dependents variables Model 1 Model 2 Model 3 Model 4 (Urban) Sex Male reference Female -0.1906*** -0.1634*** -0.1864*** -0.1710*** Household head No reference Head 0.1608*** 0.1615*** 0.1641*** 0.2256*** Level of education completed No education or primary incomplete reference Primary complete 0.2320*** 0.2311*** 0.1832*** 0.2051*** Secondary incomplete 0.4324*** 0.4307*** 0.3044*** 0.3960*** Secondary complete 0.5959*** 0.5937*** 0.4361*** 0.5701*** Tertiary/post secondary 1.2060*** 1.2033*** 1.0148*** 1.0159*** Age groups 15-29 reference 30-49 0.1900*** 0.1958*** 0.1167*** 0.2174*** 50-64 0.1075*** 0.1639*** 0.0023 0.1648** Household head Rural reference Urban sample Urban 0.1852*** 0.1861*** 0.1836*** Only Employment status Employee 0.0081428 0.0072 0.0075 -0.0072 Domestic -0.5417** -0.5479** -0.5546** 0.5220 Employer 0.2790*** 0.2772*** 0.2810*** 0.3376** Self-employed -0.0328 -0.0340 -0.0291 0.0257 Family contribution -0.3405*** -0.3434*** -0.3463*** -0.1408 Trainee -1.4360*** -1.4410*** -1.3901*** -1.3315*** Subsistence farm 0.1270* 0.1257* 0.1248* -0.1647 Employment sector Agriculture reference Industry 0.4098*** 0.4117*** 0.4100*** 0.4912*** Service 0.4137*** 0.4150*** 0.4141*** 0.4667*** Others -0.1568*** -0.1578*** -0.1559*** -0.1085** Constant 11.3440*** 11.3330*** 11.3907*** 11.4107*** Interactions between age and sex 30-49#Female -0.0141 50-64#Female -0.1346*** Not defined Not defined 30-49#Men 0.0141 50-64#Men 0.1346*** Interactions between age and education 15-29#No education or primary incomplete reference 30-49#Primary complete 0.0815** 0.0292 30-49#Secondary incomplete 0.1914*** 0.0835 30-49#Secondary complete 0.2425*** 0.1121 30-49#Tertiary/post-secondary Not defined 0.1938*** 0.1085 50-64#Primary complete 0.0626 -0.0339 50-64#Secondary incomplete 0.2402*** -0.0028 50-64#Secondary complete 0.3373*** 0.0321 50-64#Tertiary/post-secondary 0.5534*** 0.4096*** Source: Author’s calculations using 2021/2 EPM data. 82 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Low aspirations due to lack of opportunities Low literacy rates and weak social cohesion are other critical contributors to lack of aspiration and A series of focus groups in six urban communes urban poverty. Without adequate literacy, individu- suggests several perceived root causes and sub- als may not have the skills or knowledge necessary causes of poverty in those areas. Using a six-step to demand government services, thus limiting their methodology, 75 problem trees identify key bind- ability to participate in their community and advo- ing constraints and their respective root-causes cate for themselves. Moreover, weak social cohe- and sub-causes (Mulangu, 2023). Through a system sion can undermine community mobilization and dynamics approach, we then analyzed the relation- collective action, reducing the power of individuals ships between causes and sub-causes across prob- to demand change. These factors help explain the lem trees. Eight self-reinforcing and interconnected inadequacy of public investment in infrastructure, loops create a poverty trap that inhibits urban which exacerbates vulnerability to climate events households from improving their welfare (Figure 63). and shocks, and hinders the accumulation of assets that could prevent households from falling into Political and economic capture by the elite poverty. depresses aspirations among urban households. Economic literature highlights that the aspirations Finally, scarce investment in education, health- of individuals are influenced by their social environ- care, and sanitation hinders the availability of pro- ment and their expectations about future opportu- ductive human capital for the economy. By con- nities. When individuals observe that the elite con- straining economic activity, the shortage of human trol a majority of resources and opportunities, they capital lowers the return to education, which can may become discouraged and lower their aspira- exacerbate the lack of aspiration among the pop- tions, preventing their utility function from experi- ulation. Tackling these challenges requires address- encing the distinct jump predicted by literature (La ing the root causes of inequality, particularly elite Ferrara and Novak, 2022). This is particularly true capture, and enhancing community participation in for those living in poverty, who may not have access demanding public services from the government. to the same information or networks as the elite. In Empowering individuals to advocate for their needs, Madagascar, the dominant economic position of a and addressing the constraints that perpetuate the handful of families of aristocratic origin has created urban poverty trap, can help create a more equita- a sense of hopelessness and despondency among ble and prosperous future for those living in Mada- the urban population, which in turn contributes to gascar’s cities. perpetuating the poverty trap. Figure 63: Poverty reinforces low aspirations Source: Mulangu, 2023. 83 TOC Chapter 3 Madagascar As urban markets fail, urban poverty rises Poverty and Equity Assessment Figure 64: Madagascar has one of the lowest rates of financial inclusion Financial Institution account (% age +15) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Bo ne a… M Le blic er . ar ba na s e bi in ga o oz B bia rit ne Bu Co , T a o, pia am u e rk mo he am c a rit a m . s Sa T er ia w n ut Ke ia te a o Ca Lib ire A ia ng a C nia ig li d i ut Gh ini Su n an Ndan e A a G da e r p er Zi tsw gal ng hio ia a os M Afr a M aso Iv a G G eon . ra za n Za Rep aw L p iu N Ma Se fric Le bw bi ine au ic h an qu h ny U oth d' d Es abo Co RwTog h da o, da ha ha an oo m er ib n n Co Et er am en ra Re Si em asc Si Re ig N fri at in r m a au o o n n a ra u al F ut Su a g So th D u So So Co M So M ic fr lA b- Su ra nt Ce Source: Global Findex Data. Low real income has also contributed to low finan- cial inclusion and independence. Historically, Mad- agascar’s financial inclusion rate has been one of the lowest in Sub-Saharan Africa. The Global Findex Database (2021) shows that in 2017, 10 percent of people aged 15 years and above had an account in a financial institution (Figure 64). This group has since increased to 14 percent, out of which 28 percent use the internet or a mobile phone to make payments, purchases and transfer funds to other individuals. This resonates with Monnier (2021) who credits the increased financial inclusion to the introduction of mobile money platforms. Now 24 percent of the population have made or received a digital payment. In the group that remains unbanked, the main rea- sons cited are insufficient funds (68 percent), long distance to banks (52 percent), lack of necessary documentation (50 percent) and high expenses of financial services (47 percent). 84 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Chapter 4 Low human capital limits options for escaping poverty Key findings According to the World Bank's Human Capital increased risks for adolescent girls, reduced educa- Index (HCI) and other data sources, Madagascar tional opportunities, and limited competitiveness in faces significant challenges in terms of human the job market. Although female school enrollment capital development. A child born in Madagascar has improved over the past two decades, the pri- just before the pandemic was projected to be 39 mary completion rate remains low, and learning out- percent as productive as they could be with com- comes are poor. Gender disparities exist in attend- plete education and good health. This placed Mad- ance rates and literacy skills across urban and rural agascar slightly below the average for Sub-Saharan areas. Child labor, particularly among boys, further Africa (SSA) and just above the average for Low hampers education participation and outcomes, Income Countries (LICs). However, when consider- with a significant proportion of children involved in ing recent data on school enrollment and child sur- economic activities, mainly in agriculture. vival, the adjusted HCI drops to 34 percent. More- over, scores indicate that students in Madagascar Child mortality and infant mortality have also been perform below average compared to other regions. decreasing, but the rates remain higher than those There is also a significant disparity in HCI between of peer countries and the overall LIC average. While the richest and poorest children, with the gap in child mortality levels have improved compared to future productivity being larger than the global the 2000s, the gains have been smaller than in average. These findings highlight the pressing need countries like Cambodia, Rwanda, and Uganda. for investment in education and health to improve These countries started with similar or higher levels Madagascar's human capital. but have achieved substantially lower child mortal- ity rates than Madagascar. Infant mortality, specif- High fertility rates in low-income households ically deaths occurring before the first year of life, contribute to chronic poverty by overburdening remains a significant issue across peer countries, households, hindering educational attainment, although it is still lower in Madagascar compared to and limiting income generation opportunities. the overall LIC average. However, research suggests that providing essen- tial healthcare, education, and family planning ser- Caloric intake is low. In rural areas, the poorest vices can lead to a rapid decrease in fertility rates. decile has an individual average intake of 707 kcal/ By improving access to these services, women can day, while individuals in the richest decile consume make informed decisions about when and how many 3,827 kcal/day, indicating a significant disparity in children to have, resulting in improved maternal nutrition across socioeconomic levels. A similar pat- and child health outcomes, increased educational tern is observed in urban areas, although the dis- attainment, and reduced poverty rates. Although parity is less pronounced. The average calorie intake Madagascar has made some progress in enhancing for the country is 2,112 kcal/day in rural areas and access to healthcare, education, and family plan- 2,227 kcal/day in urban areas. However, focusing ning, further investment is necessary to overcome solely on caloric intake fails to account for hidden challenges such as limited resources, cultural and hunger, which refers to a chronic lack of vital micro- social barriers, and lack of awareness. nutrients. Combining data from household and health surveys, areas with high hidden hunger are Despite the decline in stunting prevalence over identified, primarily in the high plateau region with the past decade due to investments in health and some pockets in the south. nutrition, stunting remains a significant issue. However, regions where health and nutrition pro- In comparison to its peer countries, Madagascar jects were implemented have experienced a faster stands out with one of the highest child marriage decline in stunting, indicating the effectiveness of rates, a pressing issue that demands attention. donor-financed programs. The country also faces Shockingly, 40 percent of girls aged 20-24 in the high adolescent fertility rates, which contribute to country have already been married before the age 85 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment of 18. The prevalence of child marriages is most expect to complete 6.1 years of school by their 18th pronounced in the poorest regions, where vulner- birthday.48 The most recent harmonized test scores able communities struggle to escape the cycle of (PASEC 2015) showed that students in Madagas- poverty. Although higher rates of child marriage are car scored on average 351/625 points, significantly found in rural areas, it is impossible to ignore the below SSA and LIC average and lower than earlier alarmingly high rates in urban areas as well. Dis- tests. As for health indicators, across Madagascar, turbingly, there is a strong correlation between the 80 percent of 15-year-olds are expected to survive prevalence of child marriage and other detrimental until age 60 (a proxy for health risks), while only 58 practices like child labor and high numbers of out- percent of children are growing free of malnutrition. of-school children. The districts with the highest occurrence of child marriage are also the ones with Regarding complementary HCI indicators, Mada- high illiteracy rates, highlighting the interconnec- gascar stands out for lower-than-average school tion between these issues and the urgent need for enrollment and worrisome youth-related outcomes effective strategies to combat them. If left unad- (Figure 66). While basic child health indicators seem dressed, intergenerational decision-making perpet- to be at their predicted level for its GDP per capita uates the vicious cycle of poverty, ensnaring future or even slightly better, educational outcomes are generations. still low and they are deteriorating.49 Besides having significantly lower predicted years of schooling than 1. Human capital development and the SSA region and any other region, in 2019, 95 poverty percent of 10-year-olds are defined as “learning poor,” compared to 78 percent in SSA and 89 per- Human capital is lagging cent among LICs.50 Moreover, the adolescent fer- tility rate (births per 1000 women aged 15-19) was According to the World Bank’s Human Capital 143 in 2021 (DHS). This is higher than both the aver- Index (World Bank, 2020), a child born in Mad- age for SSA (93) and LICs (95). agascar just before the pandemic would be 39 percent as productive as an adult as she could be Figure 66: Expected years of school in if she enjoyed complete education and full health Madagascar are well below Sub-Saharan Africa's (Figure 65). This is just below the average for the and South Asia's average SSA region (40 percent) and just above the average South Asia (10.8) for LICs (38 percent). However, adjusting for recent Sub-Saharan Africa (8.3) data on school enrollment (MICS 2018) and child survival (DHS 2021) yields an HCI of just 34 percent. Madagascar (6.1) In terms of the different components of the HCI, 95 Note: Expected years of school at age 18 among children percent of children born in Madagascar survive to enrolled by age 4. age 5, and children who start school at age 4 can Source: Human Capital Index, World Bank Figure 65: A child born in Madagascar can only reach 39 percent if their potential productivity as an adult Human Capital Index 2020 1.0 Productivity Relative to Benchmark 0.8 0.6 0.4 Madagascar 0.2 6 8 10 12 Log GDP per capita at PPP The revised expected years of school calculated with the latest data are 6.1 years on average, compared to the official HCI estimate of 8.4 years. 86 48 The large drop in expected achievement is due to a much lower primary enrollment rate than previously estimated. The neonatal mortality rate was 20 per 1,000 live births in 2020, below both the regional average of 25 and the LIC group average of 26. In 2018, 49 79 percent of children 0-23 months had adequate meal frequency, above the regional share of 44 percent and the LIC share of 43 percent. Learning poverty means being unable to read and understand a simple text by age 10. This indicator brings together schooling and learning 50 indicators: it begins with the share of children who haven’t achieved minimum reading proficiency (as measured in schools) and is adjusted by the proportion of children who are out of school (and are assumed not able to read proficiently). TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Finally, Madagascar shows a large gap in HCI (2018) Table 25: The incidence of health shocks more between the poorest and the richest children. The than doubled in 2012-2022 ratio in HCI between the richest and poorest 20 Health Shocks 2005 2010 2012 2022 percent of the population is 1.44, whereas globally Illness of adult this ratio is 1.35. In 2018, the gap in future produc- household member 6.54 2.47 2.35 5.03 tivity between a child born in the poorest quintile Illness of another household member 4.37 1.41 1.16 2.46 and one born in the richest quintile was 18 percent- age points. This gap is larger than the typical gap Death of adult household member 1.25 0.64 0.87 4.79 across the 50 countries (15 percentage points). Death of another household member 1.58 0.9 1.74 1.85 The Human Capital Index (HCI) measures the Source: World Bank analysis based on 2005, 2010, 2012 human capital that a child born today can expect ENSMOD and 2022 EPM data. to attain by age 18, given the risks to poor health and poor education that prevail in the country to 2012 but spiked thereafter. Most of the cases where she lives. The HCI quantitatively illustrates of illness (59.7 percent) and death (52.9 percent) the key stages in this trajectory and their conse- were in urban centers. This indicates the need for quences for the productivity of the next gener- greater investment in healthcare for the bene- ation of workers, with three components: Com- fit of household welfare, especially in urban areas. ponent 1: Survival. This component is measured with survival to age 5. Component 2: School. This The MICS (2018) data for Madagascar shows component of the index combines information that fertility is high in poor households (Figure on the quantity and quality of education meas- 67), which often contributes to chronic poverty. ured as the number of years of school a child can High fertility rates in low-income households can expect to obtain by age 18 given the prevailing lead to overburdened households, lower educa- pattern of enrollment rates and harmonized test tional attainment, and reduced opportunities for scores from major international student achieve- income generation, all of which perpetuate a cycle ment testing programs. Component 3: Health. of poverty. However, research suggests that fertility Two proxies for the overall health environment rates can fall rapidly in low-income countries when are used: Adult survival rates, measured as the essential health care, education, and family plan- share of 15-year-olds who survive until age 60, and healthy growth among children under age Figure 67: Fertility among the poor is much higher, 5, measured using stunting rates. The compo- including among adolescents nents of the HCI are combined into a single index Pregnant women: ages 15-19(%) by first converting them into contributions to productivity. Multiplying these contributions to Richest quintile productivity gives the overall HCI. 4th quintile Source: Human Capital Index, World Bank. https:// datacatalog.worldbank.org/search/dataset/0038030/ 3rd quintile human-capital-index 2nd quintile Human capital and poverty are tightly Poorest quintile linked 0 20 40 60 Investment in human capital has the poten- Fertility rate by asset quintile, 2018 tial to increase consumption and reduce poverty among Malagasy households. We consider varia- Poorest asset quintile bles pertaining to health and education to assess 6.6 human capital in Madagascar. The previous Pov- 2nd asset quintile erty Assessment (World Bank 2016) found that 5.1 the time necessary to reach a health facility had 3rd asset quintile a strong positive correlation with extreme pov- 4.3 erty in the country. The report also showed that 4th asset quintile health shocks, such as sudden illness and death of 3.6 household members, were among the key deter- Richest asset quintile minants of decreased household consumption 2.7 between 2005 and 2010. Table 25 shows that the incidence of health shocks reduced from 2005 Source: MICS 2018. 87 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Figure 68: Stunting remains high compared to peer countries Incidence of stunting (%) among children under the age of 5. 2010 and 2020 50 45 40 35 30 25 20 15 10 5 0 Uganda Tanzania Rwanda Bangladwch Madagascar 2010 2020 Source: Latest Demographic and Health Surveys and MICS for each country. ning services are made wisely available. Improving Figure 69: Teenage pregnancy is strongly related access to health care and education, along with to education providing family planning services, can help break the cycle of poverty by enabling women to choose when to have children and how many children to have. This can lead to improved maternal and child health outcomes, increased educational attainment and income generation opportunities, and reduced poverty rates. In Madagascar, efforts to improve access to health care, education, and family plan- ning services have been made, but significant chal- lenges remain, including limited infrastructure and resources, cultural and social barriers, and lack of awareness about the benefits of family planning. However, continued investment in these areas can help improve the well-being of the people of Mad- agascar and contribute to sustained economic growth and development. The prevalence of stunting has gone down over the past 10 years but remains high. Figure 68 shows 60 that stunting is common in Madagascar, although its prevalence has been declining over the past 10 50 years due to major investments in health and nutri- tion. A study by the World Bank’s Independent 40 Evaluation Group (Independent Evaluation Group, 2022) found that regions where health and nutrition 30 projects were implemented tended to have high ini- tial stunting rates but experienced a faster decline 20 in stunting than the country as a whole—suggesting that donor-financed health and nutrition programs 10 were both well-targeted and effective.51 0 The high fertility rate among adolescents is of con- y y n ry y ar ar ar io ia cern in Madagascar, although it has been decreas- d at d im rt on on uc Te Pr ing (Figure 69). Almost a third (31 percent) of girls c c ed se se o er er between the ages of 15 and 19 have children or are N pp w Lo pregnant, while 2018 Census data indicates that U the share reaches 26 percent among girls between Source: MICS 2018. 51 https://ieg.worldbankgroup.org/evaluations/world-bank-group-madagascar/chapter-5-world-bank-group-support-fostering-development 88 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Figure 70: Primary school completion improved considerably between 2000-2010 and slightly declined thereafter 80.0 70.0 60.0 Primary completion rate, total 50.0 (% of relevant age group) 40.0 Primary completion rate, male (% of relevant age group) 30.0 20.0 Primary completion rate, female (% of relevant age 10.0 group) - 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Source: World Bank, World Development Indicators Database Online. the ages of 12 and 17. The adolescent fertility rate in currently attends. The data shows the percentage Madagascar is high at 108 births per 1,000 women of both males and females currently attending in aged 15 to 19, compared to SSA with an average of all areas, with urban areas having the highest per- 101 births and the global average of 42 per 1,000.52 centage of currently attending males at 62.8 per- Early childbirth exacerbates the risk of death among cent and currently attending females at 59.6 per- adolescent girls, and considerably reduces their cur- cent. Furthermore, the data suggests that there rent educational opportunities and future competi- is a significant difference in attendance between tiveness in the marketplace. rural and urban areas, with rural areas having higher percentages in the ‘never attends’ and ‘frequently Female school enrolment has risen over the past attends’ categories compared to urban areas. The two decades. Primary enrolment is almost universal, national level, on the other hand, seems to have a and educational attainment and literacy rates have more even distribution of attendance across all cat- improved. However, the primary completion rate egories. Additionally, the data shows that females remains low, although it has increased since 2001 have slightly lower attendance rates compared to (Figure 70).53 Moreover, learning outcomes are still males across the ‘currently attends’ category in all poor: among children enrolled in fifth grade, only 6 areas, except for the frequently attends category out of 10 have basic reading skills, and fewer than 2 where females have a slightly higher percentage out of 10 have basic math skills. Though more girls than males. finish primary school than boys, poverty precludes females from proceeding to secondary school and Literacy skills and reading comprehension skills often they end up in unpaid employment and early increase as students progress through school. family formation. As illustrated in Table 27, the percentage of stu- dents who correctly read 90 percent of the words Educational outcomes vary by location across in a story, answer comprehension questions cor- gender. Table 26 presents data on school attend- rectly, and demonstrate basic reading skills tends ance of males and females in urban, rural, and to increase as they advance from CP2-CE to CM2 national areas. The attendance is categorized into and then to junior high school. For example, in CM2, three types: never attends, frequently attends, and 88.4 percent of male students and 92.3 percent of Table 26: Attendance is much lower in rural areas, but the gender gap is small (attendance rates of population aged 3 to 15, by location) Urban Rural National Male Female Male Female Male Female Never attends 10.3 9.4 29.1 28.6 25.7 25.0 Frequently attends 27.0 31.0 29.4 31.0 28.9 31.0 Currently attends 62.8 59.6 41.5 40.3 45.3 43.9 Source: Madagascar, INSTAT (RGPH 2018). World Development Indicators, World Bank. https://data.worldbank.org/indicator/SP.ADO.TFRT?locations=MG-ZG 52 89 Discrepancies in primary completion rates between the 2018 Census and the World Bank estimates are due to the fact that World Bank estimates 53 are projected from survey data. TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Table 27: Fewer than 6 in 10 children finish primary education with basic reading skills Male Female Percentage Percentage who Percentage Percentage who who answered the Percentage who answered the Percentage correctly comprehension questions having correctly comprehension questions having % of % of read 90% of correctly demon- read 90% of correctly demon- children children the words strated ba- the words strated ba- aged 7-14 aged 7-14 of a story in Three Two sic reading of a story in Three Two sic reading Malagasy or literals inferential skills Malagasy or literals inferential skills French French CP2-CE 37.2 23.2 15.5 12.0 33.5 39.6 23.7 18.4 14.4 31.8 CP2 23.8 14.9 9.0 7.7 17.9 20.7 14.2 8.9 7.4 16.9 CE 52.6 32.8 23.0 16.9 15.6 61.1 34.6 29.2 22.4 14.8 CM1 83.4 63.1 46.6 41.4 12.9 77.2 54.5 45.2 36.6 11.9 CM2 88.4 68.7 67.1 57.2 10.3 92.3 76.2 62.6 55.9 11.3 Junior High School 96.5 88.3 82.4 79.0 9.8 96.9 89.4 85.2 80.5 13.3 Source: MICS Madagascar (2018). Table 28: Over one-third of children under 11 are engaged in economic activity Percentage of children Percentage of children aged 12 to 14 involved in aged 15 to 17 involved in economic activities for: economic activities for: % of children aged 5-11 % of % of % of engaged in children Less than 14 hours or children Less than 43 hours children economic aged 14 hours more aged 43 hours or more aged activities for at least one hour 5-11 12-14 15-17 Total 35.8 100.0 32.4 31.5 100.0 56.7 10.3 100.0 Male 38.3 50.4 29.6 38.0 49.0 57.2 14.8 53.2 Female 33.4 49.6 35.1 25.2 51.0 56.1 5.2 46.9 Urban 23.1 19.0 24.7 18.9 21.0 40.9 5.1 25.9 Rural 38.8 81.0 34.4 34.9 79.0 62.3 12.1 74.1 Source: MICS Madagascar (2018). female students correctly read 90 percent of the absenteeism and climate disasters. Marchetta et al. words in a story, compared to 96.5 percent and 96.9 (2019) found that cyclones reduce the probability of percent, respectively, in junior high school. Similarly, attending school, reduce average French and Math the percentage of students who answered com- test scores, and increase the probability of young prehension questions correctly and demonstrated people, especially girls, entering the work force. Girls basic reading skills generally increased from CP2-CE in Madagascar, especially in the regions of Sofia and to junior high school. It is worth noting that there Atsimo-Atsinanana, contribute notably to agricul- are still some gender disparities in literacy and read- tural work. ing comprehension, but the disparities vary across grades and subjects. 2. Health and sanitation The prevalence of child labor contributes to low- Overall life expectancy at birth has been improving ering education participation and outcomes, over the past 20 years. Life expectancy among the especially among boys (Table 28). Approximately Malagasy population is 67 years. This is above the one-third of children aged 5-11 and more than 60 overall LIC average of 64 years, as well as expectancy percent of those aged 12-14 work, with the propor- in Tanzania (age of 66) and Uganda (age of 64). Over tion of boys involved in economic activities higher time, life expectancy has improved in Madagascar than that of girls. This is especially common in rural from 59 years in 2000 but the pace of improvements areas, where 62.3 percent of children (mostly boys) is below that of Rwanda, Tanzania and Uganda work between one and 43 hours a week or more. (Figure 71 and Figure 72). In sum, while life expec- An intensive use of child labor, mainly in agricul- tancy has been improving in Madagascar, it’s persis- ture, hinders human capital development by con- tently been at a slower pace than in other countries. tributing to poor educational outcomes, high teen- age pregnancy rates, and malnourishment among Child mortality has been decreasing slowly, remain- children (Bau et al. 2020). Other issues that com- ing at higher rates than peers but below overall LIC promise educational outcomes include teacher averages, similarly as infant mortality. Child mor- 90 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Figure 71: Life expectancy at birth has risen Figure 72: Life expectancy level is above the LIC steadily but slower than among peers average 75 80 Life expectancy at birth, total (years) Life expectancy at birth, total (years) 70 75 73 70 69 65 70 67 66 60 64 64 65 55 60 50 55 45 50 40 45 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 40 20 20 02 0 20 20 20 20 MDG BGD KHM LIC , 20 , 20 ,2 , 20 , 20 , 20 , 20 DG LI C BG D KH M A TZ A UG A RWA TZA UGA M RW Note: BGD Bangladesh, KHM Cambodia, MDG Madagascar, RWA Rwanda, TZA Tanzania, UGA Uganda, LIC average of low- income countries. In all cases the graph presents the latest available year. N/A not available. Source: World Development Indicators. tality levels, since the 2000s, have consequently provision in Madagascar, compared to some of the remained lower, when compared to overall LIC levels, peer countries, where the total number of births per which is encouraging (Figure 73 and Figure 74). Yet, woman has fallen more since 1960, such as Bangla- when compared to select peer countries, gains have desh, Cambodia and Rwanda. Taking the number of been smaller in the past 20 years. Notably, Cambo- hospital beds per 1000 people, Madagascar remains dia, Rwanda, Uganda had started out with higher below peer countries' and LIC averages with 0.2 or equal levels, and now experience substantially hospital beds per 1000 people, which is less than lower child mortality rates than Madagascar, at 25 half the hospital bed coverage than these countries to 43 child deaths per 1000 live births. Particularly (Figure 75).54 At the current birth rate, Madagascar’s for infant mortality, deaths occurring before the first population could reach 36 million by 2030, further year of life remain highest across peer countries, at increasing pressure on health services.55 36 per 1000 live births, yet remaining at lower levels that the overall LIC average (47 percent). Access to sexual and reproductive health services has been improving slowly, and there are signs it Madagascar’s population has doubled in 15 years, declined since the pandemic. Contraceptive use creating important strains on access to health ser- prevalence increased between 2009 and 2018, vices. Madagascar’s population doubled between when 39 percent of women 15-49 reported using 1993, the time of the last Census, and 2018 to 26 modern contraceptive methods, against 29 per- million, according to the 2018 Census. Overall, pop- cent in 2009. Yet, 16 percent of women reported ulation growth has increased pressure on service unmet needs for contraception, which explains in Figure 73: Child mortality has declined at a slower Figure 74: Infant mortality is higher than among pace than among SSA peers peer countries, but below LIC average 200 50 47 180 45 Mortality rate, under-5 (per 1,000 live births) 160 40 (per 1,000 live births) Mortality rate, infant 140 36 35 35 120 32 30 100 30 25 24 80 22 60 20 40 15 20 10 0 5 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 0 20 20 20 20 02 0 02 0 02 0 BGD KHM LIC MDG , 20 , 20 , 20 , 20 ,2 ,2 ,2 D G L IC G D M A Z A G A RWA TZA UGA M B KH RW T U Note: BGD Bangladesh, KHM Cambodia, MDG Madagascar, RWA Rwanda, TZA Tanzania, UGA Uganda, LIC average of low- income countries. In all cases the graph presents the latest available year. N/A not available. Source: World Development Indicators. 54 Reference years differ. Except for Bangladesh and Cambodia, reference years date further back. 91 55 https://data.worldbank.org/indicator/SP.POP.TOTL?locations=MG TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Figure 75: Madagascar has less than half the part why 14 percent of women still had unwanted hospital beds than its peer countries and LIC births (Figure 76). According to the Global Financing average Facility for Women, Children and Adolescents (GFF 1.8 2020), the mortality rate of children under 5 is esti- 1.6 1.6 mated at 60 per 1,000 live births without disruption Hospital beds (per 1,000 people) of services and 65 per 1,000 live births with disrup- 1.4 tion of services. Updated models suggest that large 1.2 service disruptions during the pandemic could have 1.0 0.9 left as many as 81,600 women without access to 0.8 facility-based deliveries, and 645,100 fewer women 0.8 0.7 0.7 receiving family planning services. The results of 0.6 0.5 these disruptions would be lethal, with an 18 and 12 percent increase in infant and maternal mortality 0.4 0.2 respectively. 0.2 0.0 Health facilities are more prevalent in richer regions 01 0 00 6 01 6 01 6 00 7 01 0 01 0 (Figure 77). It is well established that access to ,2 ,2 2 2 2 2 2 G D, M, A, A, A, health facilities has tremendous implications for MD LIC BG KH RW TZ UG human capital and productivity (Thomson et al. Note: BGD Bangladesh, KHM Cambodia, MDG Madagascar, RWA Rwanda, TZA Tanzania, UGA Uganda, LIC average of 2009). Arguably, health service provision is always low-income countries. In all cases the graph presents the important but particularly so for more vulnerable latest available year. N/A not available. individuals such as children under five and pregnant Source: World Development Indicators. Figure 76: Modern contraceptive use has increased, but 16 percent of women still lack access to any contraception Source: Contraceptive prevalence and unmet need for contraception from World Development Indicators, World Bank. Wanted fertility from DHS 2008 and MICS 2018. Figure 77: Richer regions have slightly more health facilities Public and private health facilities by region 300 250 Public Private 200 150 100 50 0 fia va sy na y sy y y e be ka va a o ga a ia ro na ny a y en ro ak ab an tr an an of an go Ita no Sa o la ia So bo fa om an ra nd jir el Bo ib an an en in go D iM re A an ka m m M n ts v or A in in M la nd to n la A an ' -M Be on Ih ts ts Bo na na Fi A a n A A or ra tr A o- ki A - o- sia ot m Va vy im im A la at va ts A ts M A to A Va Source: Authors’ calculations based on 2018 Madagascar Census. 92 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Map 9: The South has the lowest number of health Map 10: The South and the Capital have the lowest facilities per children aged 0-5 number of health facilities per women aged 15-49 No. of facilities per No. of facilities per 100,000 children aged 0-5 100,000 women aged 15-49 0-35 0-25 35-60 25-40 60-85 40-55 85-160 55-80 160-280 80-140 Note: The solid white line outlines the boundaries of regions. Source: Authors’ calculations based on 2018 Madagascar Census. women. The 2018 Census enumerated public and limited access to treatments due to costs even in private health facilities, which appear to be more public facilities, shortages of supplies, absenteeism frequent in richer and more populated regions. of staff, long waits, and poor infrastructure. Lack of While this information does not contain quality or access appears more acutely in the Grand South. size indicators, it is a rough measure of access. Additionally, provinces particularly in the South- ern but also Eastern regions reported in 2018 that However, when looking at facilities per capita, access to service delivery deteriorated, when com- even the capital shows limited access. Regions like pared to previous years. This is corroborated by the Androy, Ihorombe, Bongolava and Vakinankaratra recent Health Service Delivery Index reports (World are least-serviced in terms of number of facilities Bank 2017, 2024). per 0–5-year-old children (Map 9) while Anala- manga shows very limited number of facilities per Figure 78: Madagascar has lower access to 15–49-year-old women (Map 10), though it is possi- drinking water than the average LIC ble that the latter also has larger institutions of higher 70 People using safely managed drinking quality. Similarly, districts right around the capital 59 water services (% of population) 60 city appear heavily under-serviced by public insti- 50 tutions, while displaying strong coverage by private 40 institutions. This could be pointing toward impor- 29 30 28 tant inequalities along the income distribution with 21 excellent services for wealthier individuals seeking 20 17 12 out private institutions, and important under-provi- 10 sion of the less well-off and poor with limited access 0 to public facilities, with important bottlenecks. 20 0 0 0 0 A 0 20 02 02 02 02 N/ 02 , C ,2 ,2 ,2 ,2 Z A, ,2 DG L I BG D KH M A T UG A Health outcomes are affected by low quality of M RW health services and low safe water and sanitation Note: BGD Bangladesh, KHM Cambodia, MDG Madagascar, coverage. The Afrobarometer 2015 (Afrobarome- RWA Rwanda, TZA Tanzania, UGA Uganda, LIC average of low-income countries. In all cases the graph presents the ter and Transparency International 2015) reported latest available year. N/A not available. that a large share of the Malagasy population has Source: World Development Indicators. 93 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Map 11: 4.4 million households lack access to Map 12: 4.7 million households lack access to improved sources of water improved sanitation % of households % of households 0-15 0-10 15-30 10-20 30-45 20-30 45-60 30-40 60-75 40-50 75-90 50-60 Note: The solid white line outlines the boundaries of regions. Source: Authors’ calculations based on 2018 Madagascar Census. Few areas of Madagascar have good access to safe Figure 79: Access to improved water and water. Other key services linked to health outcomes sanitation are highly correlated in rural areas include water and sanitation. Access to improved 50 water and sanitation has been linked to improved 45 nutrition and lower stunting due to less exposure Access to improved sanitation fetching water or sharing toilets (Skoufias, Vihna 40 and Sato 2019). The proportion of households with 35 (%) of households access to clean drinking water in Madagascar is sig- 30 nificantly lower than the average in LICs and in the 25 mid-range of its peer countries (Figure 78). In 44 districts, fewer than 15 percent of households have 20 access to safe water and overall, about 4.4 million 15 households across the country lack access to safe 10 water (Map 11). 5 Access to improved sanitation is concentrated in 0 central Madagascar and select coastal regions. At 0 20 40 60 80 Access to improved water source national level, 4.72 million households lack access to (%) of households improved sanitation (Map 12). And in most districts, Urban Rural fewer than 10 percent of households have access Source: Authors’ calculations based on 2018 Madagascar to improved sanitation. The districts with the lowest Census. access are in Fianarantsoa and Toliara. In the capi- tal city, 1.66 million out of the total 2.37 million resi- dents lack access to sanitation. Access to improved sources of water is positively correlated with access to improved sanitation, with the pattern being more salient in rural areas (Figure 79). 94 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment 3. Food security Table 29: Urban average calorie consumption is higher than rural Food security, defined as the availability, access, Distribution of calorie intake (kcal/person/day), 2021-22 and utilization of sufficient, safe, and nutritious PCER deciles Rural Urban Madagascar food, is a critical concern in Madagascar. One of the key indicators of food security is the caloric con- Poorest 707 669 696 sumption of a population. Table 29 showcases the 2 1,120 1,051 1,098 distribution of calorie intake in urban and rural areas 3 1,380 1,211 1,330 of Madagascar, shedding light on the disparities 4 1,596 1,415 1,535 across the country. Analyzing caloric consumption 5 1,793 1,540 1,700 patterns is vital for understanding the challenges 6 1,983 1,709 1,868 and opportunities related to food security. By exam- ining the variations in calorie intake across different 7 2,203 1,776 2,010 deciles, policymakers and organizations can identify 8 2,487 2,046 2,253 areas with inadequate nutrition and implement tar- 9 2,859 2,309 2,543 geted interventions to ensure equitable access to Richest 3,827 3,451 3,572 nourishing food, ultimately striving to achieve sus- Madagascar 2,112 2,227 2,167 tainable food security for all. Average caloric intake differs widely across the dis- calorie intake of 707 kcal per person per day, against tribution, but less between rural and urban areas. 669 kcal per person per day in urban areas. Within Table 29 provides information on the distribution of the richest decile, rural individuals consume 3,827 calorie intake in rural and urban areas of Madagas- kcal and urban individuals consume 3,451 kcal per car for the year 2021-22, as well as the overall figures person per day. While rural consumption is slightly for the country. The data is categorized according higher for rural individuals within each decile, rural to the national per capita real consumption (PCER) individuals are concentrated in poorer deciles, thus deciles, which divide the population into ten equal the average rural consumption is lower in rural (2,112 groups based on their total consumption. Within kcal) than in urban areas (2,227 kcal). the poorest decile, rural individuals have an average Map 13: Hidden hunger is largely located in the high plateau with a few pockets in the south . . . Percentage of stunted population Percentage of stunted population Communes with high stunting at commune level at commune level and low poverty rates 0.484 to 0.598 89.79 to 99.27 Yes 0.429 to 0.484 82.52 to 89.79 No 0.379 to 0.429 75.34 to 82.52 0.317 to 0.379 63.87 to 75.34 0.146 to 0.317 14.74 to 63.87 Missing Missing Source: Authors’ estimation using 2021/2 EPM data. 95 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Hidden hunger, which may occur despite sufficient Figure 80: Child marriage is prevalent in caloric intake, is prevalent in the High Plateau with Madagascar ... pockets in the South. Caloric intake is not a suffi- 60 (% of women ages 20-24) 51 Women who were first cient indicator of food security as individuals may 50 married by age 18 lack essential vitamins and minerals, a concept 40 40 34 called hidden hunger. Hidden hunger refers to the 30 31 chronic lack of vital micronutrients, such as iron, zinc, 19 20 iodine, vitamin A, and vitamin D, in an individual's 10 6 diet which can lead to a wide range of long-term 0 health complications. Combining small area poverty 8 /A 9 4 20 16 6 estimates from the 2022 household survey (EPM) 01 ,N 01 01 20 20 01 ,2 ,2 ,2 A, A, ,2 G LIC D M TZ A and 2018 Census with stunting data from DHS pro- MD BG KH RW UG vides a picture of households with hidden hunger, Note: BGD Bangladesh, KHM Cambodia, MDG Madagascar, that is, where enough caloric intake per capita and RWA Rwanda, TZA Tanzania, UGA Uganda, LIC average of stunting (a result of micronutrient deficiency) are low-income countries. In all cases the graph presents the latest available year. N/A not available. present (Map 13). Hidden hunger is largely pres- Source: World Development Indicators. ent in the high plateau with a few pockets in the south. Communes in red have high hidden hunger. Figure 81: ... it is more common rural areas They are located between the regions of Anala- 40 manga and Vakinankaratra with a few communes in Girl marriages as % of total 35 Alaotra-Managoro and Atsinanana in the East. 30 4. Child marriage is a key 25 marriages determinant of intergenerational 20 poverty 15 10 Madagascar’s child marriage rate is among the 5 highest in the world. In comparison to its peer 0 countries, Madagascar has one of the highest child Urban Rural marriage rates with 40 percent of girls aged 20-24 having been married by the age of 18 (Figure 80). Map 14: Child marriage is higher in the West and The only country with even higher rates is Bangla- South desh where as many as 51 percent are married in % of child marriages among women 10-20 the same age bracket. Up to 37 percent of young 20-30 women in SSA are married before the age of 18 30-40 (UNICEF 2020) and one in four young women have 40-50 their first child by this age. Among these high rates 50-60 overall, Madagascar is at the misfortunate helm 60-70 of it, with 40 percent of girls at the age of 15-19 being child-married. Unlike women, adolescent girls are still developing cognitively and biologically, and must navigate the formation of agency through norms of age, in addition to gender. On all accounts, child marriage is not only interfering with a young woman’s life, it also puts her in life-threatening sit- uation of childbearing which has a higher morbidity rate among its youngest mothers. The highest shares of child marriages among women are in the poorest regions especially in rural areas. In the 2018 Census, we observe child mar- riages across all age cohorts. As of 2018, 37 percent of all marriages in rural areas were the result of child marriages (Figure 81) and 21 percent of all mar- riages in Madagascar involved girls under the age of 18. The highest shares are in the Southern regions Note: The solid white line outlines the boundaries of regions. Source: Authors’ calculations based on 2018 Madagascar and Western cone of Melaky. In those areas, child Census. 96 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment marriages are as high as 60-70 percent (Map 14). Table 30: Poverty is higher for households with The tragedy can be barely overstated and creates child marriage, 2021-22 clear alarm bells for policy with dire need for inter- Contri- vention to break this inter-generationally persistent Child bution Popula- Poverty practice. Location married to poor tion rate (%) women popula- share (%) tion (%) There is a strong correlation between multidi- Overall mensional poverty prevalence and child marriage Yes 45.4 18.8 4.3 Urban occurrences across districts. Figure 82 plots child No 25.2 10.5 15.2 Yes 90.2 37.5 30.4 marriage rates against multidimensional poverty Rural No 80.0 33.2 50.1 rates at the district level. There is a clear positive Province correlation between them. In fact, taking the poor- Yes 67.9 7.3 7.0 Antananarivo est districts together, a slightly higher multidimen- No 49.9 5.4 22.5 Yes 94.1 10.2 7.4 sional poverty is associated with a disproportion- Fianarantsoa No 84.7 9.2 14.3 ately larger rate of child marriage. Child marriage Yes 83.1 9.0 4.3 Toamasina and poverty are linked in various ways. Poor families No 69.9 7.6 10.2 make difficult choices for survival, including marry- Yes 87.3 9.4 5.4 Mahajanga No 71.8 7.8 6.5 ing off girls as young as possible, either to receive a Yes 91.9 10.0 7.7 Toliara dowry or to reduce the number of dependents at No 80.1 8.7 7.6 home. In turn, young girls who marry early lack edu- Yes 79.7 8.6 2.9 Antsiranana No 63.4 6.9 4.2 cation and work experience. They start childbear- Region ing and -rearing early, which increases their risk of Yes 44.9 1.3 2.8 Analamanga maternal mortality, and they are more likely to have No 28.8 0.8 12.1 stunted children (Efevbera et al, 2017) who in turn Yes 85.5 2.4 2.3 Vakinankaratra No 76.7 2.2 6.0 will be less productive as adults. Geographically, Yes 79.7 2.3 1.0 there is a strong overlap between multidimensional Itasy No 71.7 2.0 2.6 poverty map and the prevalence of child marriage at Bongolava Yes 82.0 2.3 0.9 No 72.4 2.1 1.7 the district level (Map 4 and Map 14), with the poor- Yes 90.1 2.6 1.6 Haute est areas in the country displaying the highest rates Matsiatra No 78.8 2.2 4.5 of child marriage. Households with child marriage Amoron I Yes 90.2 2.6 0.9 Mania No 81.6 2.3 2.6 are overrepresented among the poor in both urban Vatovavy Yes 95.6 2.7 1.8 and rural areas (Table 30). Nationally, child-married Fitovinany No 89.9 2.5 3.9 households represent 56.3 percent of the multidi- Ihorombe Yes 93.3 2.6 0.8 mensionally poor population, while they represent No 77.9 2.2 0.8 Atsimo Yes 97.5 2.8 2.3 34.7 percent of overall population. In urban areas Atsinanana No 92.9 2.6 2.4 45.4 percent of households with spouses married as Atsinanana Yes 83.1 2.4 1.8 a child are in multidimensional poverty and in rural No 64.6 1.8 3.9 Yes 81.2 2.3 1.3 areas 90.2 percent. Analanjirofo No 72.5 2.1 2.7 Alaotra Yes 85.1 2.4 1.3 Figure 82: Child marriage is higher in districts with Mangoro No 73.7 2.1 3.6 Yes 82.7 2.3 1.4 high multidimensional poverty Boeny No 58.9 1.7 2.3 Yes 86.9 2.5 2.7 45 Sofia Child marriage (% of total marriages) No 77.0 2.2 2.8 40 Yes 90.7 2.6 0.6 Betsiboka No 79.6 2.3 0.9 35 Yes 95.5 2.7 0.6 Melaky 30 No 87.2 2.5 0.5 Atsimo Yes 90.4 2.6 3.5 25 Andrefana No 76.0 2.2 3.2 Yes 96.5 2.7 1.6 20 Androy No 92.5 2.6 1.4 15 Yes 93.1 2.6 1.5 Anosy No 81.1 2.3 1.5 10 Yes 88.7 2.5 1.1 Menabe 5 No 75.9 2.2 1.5 Yes 80.5 2.3 1.4 0 Diana No 57.3 1.6 1.9 0 20 40 60 80 100 Yes 78.9 2.2 1.5 Sava Poverty rate (% of population) No 68.3 1.9 2.4 Source: Authors’ calculations based on 2018 Madagascar Source: Authors’ calculations based on 2018 Madagascar Census. Census. 97 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Figure 83: Child labor and child marriage are Table 31: Geography, literacy and education correlated across districts of head of household are associated with not 50 attending school Child marriage (% of total marriages) Dependent variable: not attending school 40 Not attend- Not attending Variable ing primary secondary 30 school school Total population in the Antananarivo 0.274*** -0.018 commune 20 Toamasina Urban population -0.046*** -0.082*** Toliara Fianarantsoa Household member is female -0.021*** 0.0131*** 10 Mahajanga Age of household member -0.015*** 0.082*** Antsiranana 0 Household head is female 0.004*** -0.01 0 5 10 15 20 Age of head of household -0.000*** -0.003*** Child labor (% children under 15 working) Literacy rate of head of -0.078*** -0.071*** Source: Authors’ calculations based on 2018 Madagascar household (any language) Census. Education of the head: None Reference group There is also a positive correlation between child Education of the head: Pre or -0.152*** -0.063*** marriage and child labor across districts. Certain Primary districts in Toliara, Fianarantsoa, and in Mahajanga Education of the head: -0.202*** -0.167*** are among the most affected by the triple chal- Secondary lenges of high out-of-school, child labor, and child Education of the head: -0.179*** -0.188*** Technical or Higher marriage rates, risking their future human capital (Figure 83). In 6 districts, child labor counts for more Household size 0.001 -0.004*** than 10 percent of the employed population and 36 Room density (no. people per 0.011*** 0.008*** percent of marriages are underage. These districts room) are predominantly rural have higher average house- Household has access to -0.036*** -0.042*** improved water sources hold sizes and people work mostly in agriculture and livestock rearing. Household has access to -0.038*** -0.070*** electricity Improved wall materials -0.013*** -0.044*** In addition, districts with high child marriage have lower educational attainment and literacy. Child Improved roof materials -0.091*** -0.102*** brides are more likely to be illiterate and to marry Improved floor materials -0.004 -0.000*** an illiterate spouse. In the literature, education has Observations 4,141,640 4,444,680 been heralded as an important tool against child R-squared 0.19 0.43 marriage, with multiple potential mechanisms. Secondary schooling helps to delay marriage and Source: Jarotschkin (2023). increases economic opportunities and agency for young women. Parental education is also more likely to reduce child marriage. In the absence of parental education data to be linked to child marriage, dis- trict level measures of education are plotted against child marriage (Figure 84 and Figure 85). Districts with higher literacy and number of secondary schools have lower child marriage. When looking at simultaneous correlates of child marriage, second- ary education, higher education of male spouse, rural area, the sex and age of the household head, and household size are the most significant. In addi- tion, looking at the simultaneous correlates of not attending primary or secondary school (Table 31), education of the head of household is a key deter- minant. At the same time, because child marriages often involve non-educated spouses and brides, the likelihood that their children will have lower edu- cation increases. 98 TOC Chapter 4 Madagascar Low human capital limits options for escaping poverty Poverty and Equity Assessment Figure 84: Child marriage is higher in districts with Figure 85: Child marriage is higher in districts with low literacy fewer secondary schools 70 50 45 Child marriage (% of total marriages) 60 40 Child marriage for girls under 18 35 (% of total marriages) 50 30 40 25 20 30 15 10 20 5 R! = 0.5547 10 0 10 20 30 40 50 60 70 80 90 100 0 2 4 6 Total number of secondary schools Literacy rate (%) (per 10,000 population) Child marriage of girls and education levels of population across districts 99 TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Chapter 5 High vulnerability to shocks drives short term poverty but has long term consequences Key findings Madagascar's vulnerability to recurrent shocks, the 68 percent of respondents that had experi- including climate disasters, political instability, and enced an increase in the price of major food items global crises, negatively impacts economic output, usually consumed by their household, 46 percent household well-being, and poverty levels. For cen- reduced their consumption to cope with the shock. turies Madagascar has been ravaged by cyclones and droughts. The nation is also exposed to the Madagascar canhelp vulnerable households better effects of global crises such as the COVID pandemic cope with repeated shocks and reduce their chances and rising commodity prices. In addition to these falling further into poverty. Climate-smart invest- exogenous shocks, within its borders, Madagascar ments and index insurance are two ways to enhance has experienced political instability multiple times. agricultural productivity and reduce vulnerability to Climate shocks damage infrastructure while political shocks which can be implemented quickly and at and economic shocks disrupt productivity. Regard- scale. These policies must be accompanied by pro- less of type, at a national level, shocks suppress ductive and adaptive safety nets.. economic output and growth. At the household level, property, livestock, and crops are damaged; 1. An overview of recent shocks and the value of real income declines; consumption challenges to recovery decreases and poverty increases. For example, pov- erty increased by 2.5 percent during the COVID Madagascar has been vulnerable to shocks since pandemic. Food insecurity also increases during before it became an independent country, with such times, especially in the Grand Sud region. droughts in the south causing acute food insecu- rity since the late 1800s. In addition, since gain- Drought, irregular rains, tropical storms, and high ing independence in 1960, Madagascar has expe- food prices are the most severe systemic shocks rienced several episodes of political unrest, often faced by households. Each had an impact on instigated by power struggles among elites. The household wealth (income and assets), agricultural latest such crisis, in 2009, resulted in cuts to exter- output (crops and livestock) and food (purchases nal aid and a period of isolation from donor partners. and stock). Sixty-eight percent of households that Climate change has exacerbated Madagascar’s vul- experienced food inflation deemed it the most nerability to weather shocks, increasing the severity severe shock they had suffered and those affected of cyclones, locust infestations, and droughts. The by it have not returned to their pre-shock consump- country has historically struggled to recover from tion levels. such episodes. Notably, Madagascar is the only country globally whose GDP per capita in 2020 was Apart from shocks which affect whole communities lower than in 1960 without experiencing a civil war households also suffer idiosyncratic shocks that in that interval (Figure 86). Its poverty rate is among increase poverty. In 2022, households experienced the highest in the world. illness (5.6 percent), death (4.7 percent), theft (3.7 percent), divorce (3 percent), job loss (2.8 percent), The country was hit by multiple disasters in recent income loss (1 percent) and bankruptcy (0.8 percent) years, including the recession caused by the and other forms of misfortune. The loss of jobs and COVID-19 pandemic and five storms and tropical salaries had the most adverse effect on households. cyclones that damaged food crops, livestock, and In the absence of employment, people sell assets, infrastructure. The cyclones affected over two- crops, and livestock in order to meet their needs. thirds of the country and caused direct damages of Many households often lack an immediate coping more than US$658 million, equivalent to 4.8 per- strategy, and when they have one it frequently con- cent of GDP. In addition, Russia’s invasion of Ukraine sists of buying cheaper food, soliciting help from has impacted the supply and pricing of essential relatives and friends, or using their savings. Out of goods such as fuel, edible oils, and grains. While 100 TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Figure 86: Shocks have repeatedly set back GDP growth in Madagascar a) GDP per capita… in Madagascar around major shocks b) … in 2020 vs in 1960, select countries Note: in b), orange dots indicate countries where GDP per capita in 2020 was lower than in 1960. For example, GDP per capita in Haiti was US$1,714 in 1960 and US$1,323 in 2020.56 Source: Authors based on WDI. such shocks disproportionately affected vulnerable 1.64 million out of 2.7 million people—faced acute households, inflationary pressures are felt by all res- food insecurity and required emergency assistance idents, and particularly by those whose food basket between October and December 2021. This included primarily comprises imported goods. over 500,000 people facing emergency (IPC Phase 4), and an additional 28,000 people facing famine Prolonged droughts have caused a severe famine conditions (IPC Phase 5). At least 405,000 children in southern Madagascar. The Grand Sud area of under the age of five suffered from acute malnour- the country experienced the worst drought in a ishment, with 110,000 of them in need of urgent decade during the 2019-2020 agricultural season, lifesaving assistance.58 resulting in the loss of up to 60 percent of the crop yield in three of the area’s most populated dis- 2. Types of shocks and coping tricts.57 Specifically, maize, cassava and rice yields mechanisms declined (ACAPS, 2022). According to a November 2021 report by the Integrated Food Security Phase Households in Madagascar face idiosyncratic Classification (IPC 2021), an estimated 37 percent shocks that are either social or economic in nature of the population in the 10 hardest-hit districts—or (Figure 87). Social shocks, which affect incomes indi- Figure 87: Illness and death are the most frequent idiosyncratic shocks (Self-reported) Serious illness or accident of household member Death of household member Idiosyncratic Shocks Theft of money, property, crops or livestock (Dahalo) Divorce or separation Loss of salaried employment of member of household Significant loss of salary income Bankruptcy of an HH ENA Significant loss of o -farm income from HH End of regular transfers from other households 0 1 2 3 4 5 6 Households A ected (%) Source : Author’s calculations based on Enquête Permanente Auprès des Ménages (EPM) data, 2022. 56 See Table A1 in the Appendix for a full contrast of GDP in the countries that experienced a decline. 57 The Grand Sud area is the southern part of Madagascar, consisting of Androy, Anosy and Atsimo-Andrefana regions. It is very vulnerable to 101 climate shocks because of its geographic position in proximity to the Indian Ocean. According to Radi (2023), “The livelihood of 95 percent of the population in the Grand Sud depends on agriculture, raising livestock, and fishing. https://kujenga-amani.ssrc.org/2023/01/13/the-impact- of-climate-change-on-the-food-crisis-in-the-grand-sud-of-madagascar/. The Grand Sud witnessed a long drought period that began in 2019 and continued until January 2022, which led to crop failure. The droughts caused the suffering of more than 1.6 million people in the Grand Sud at the beginning of 2022 as a result of high food insecurity levels.” 58 https://www.wfp.org/news/malnutrition-among-children-expected-quadruple-southern-madagascar-drought-worsens-warn-unicef TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment rectly, include illness/accident or death of a house- Approximately 8 percent of these households did hold member, and divorce/separation. Economic not see a change in the size of their agricultural pro- shocks have a more direct impact on incomes and duction or livestock, but 32 percent experienced a include: theft of money, property, livestock (Dahalo) decline in purchases and stocks of food. Food pur- or crops; loss of salaried employment; significant chases and stocks also decreased significantly for loss of salary income; bankruptcy of a household households that lost a significant part of their salary enterprise; loss of off-farm income; and the end and those that suffered theft of money, property of of regular transfers from other households in their agricultural output. social network.59 As households face various shocks, they adopt More households experienced social shocks than several strategies to cope with them. Table 32 economic shocks. The combined share of house- shows different ways in which households attempt holds affected by social shocks was 13.15 percent of to adapt to shocks. Many households resort to the population, of which 5.56 percent suffered from buying cheaper food, soliciting help from relatives the serious illness/accident of a household member, and friends, or using their savings. A much smaller 4.64 percent from the death of a household member, group of households change their consumption and 2.95 percent went through a divorce or separa- habits, having acknowledged that recovering from tion. On the other hand, 9.42 percent of households the shock might take time. Less frequent but costlier experienced direct economic shocks: 3.68 percent coping mechanisms include selling property, assets, lost money, property, livestock, or crops to theft; and food, or inducing children in the household to 2.81 percent had a member who lost a salaried job; work or get married. Worryingly, certain households 0.97 percent lost salary income; 0.79 percent had a cut healthcare and education expenditures to cope household enterprise that went bankrupt; 0.59 per- with shocks, which can have negative long-term cent lost non-agricultural income; and 0.58 percent effects on human capital. stopped receiving transfers from other households. Systemic shocks affect more people than idiosyn- Loss of salaried employment and the signifi- cratic shocks. Only 1.25 percent of households in cant loss of salary income had the most profound Madagascar suffered from systemic social shocks: adverse impact on households. Theft caused a loss 1.02 percent were affected by armed conflict, vio- of income, assets, crops, and livestock for 83, 87, 64, lence, or insecurity; and 0.23 percent experienced and 57 percent of the affected households, respec- farmer-herder conflict. Conversely, systemic eco- tively, while fewer than 15 percent of the households nomic shocks affected 28.5 percent of the sampled that experienced such theft were not materially households, led by high food prices (affecting 24.62 affected by it. The shock with the largest adverse percent of households) ahead of low prices for agri- impact on households was loss of salaried employ- cultural produce (2.51 percent) and high prices for ment. Households where a member lost a job expe- agricultural inputs (1.37 percent). Natural hazards rienced a decline in assets, crops and livestock. In affected 42.93 percent of households, broken down these households, food purchases decreased by as follows: drought, 17.68 percent; tropical storms, 85 percent and food stocks by 62 percent. Among 10.24 percent; irregular rains, 9.29 percent; floods households that lost salary income, there was no 2.56 percent; crop and animal diseases, 1.86 per- significant change in the structure of their asset cent; locust and pest attacks, 0.67 percent; fires, portfolios, but the quantity of assets decreased. 0.51 percent; and landslides, 0.12 percent. Table 32: Households have few mechanisms to cope with idiosyncratic shocks (Self-reported) Percentage of households that used each coping mechanism Seek help Change Mobilize Buy cheaper from relatives consumption No strategy Other savings food / friends habits Salary loss 8.33 13.64 8.33 21.21 15.91 32.58 Salaried job loss 11.98 13.54 9.11 20.31 16.67 28.39 Illness/accident 11.56 18.79 6.18 9.07 21.42 32.98 Death 8.86 11.99 6.62 10.25 37.85 24.61 Divorce/separation 7.18 16.83 6.68 8.17 37.62 23.52 Theft 4.57 7.55 8.15 7.55 40.76 31.42 Source: Author’s calculations based on 2022 EPM data. ‘Dahalo’ is a native term referring to cattle rustling / theft. The New Humanitarian (2012) explains the cultural context. https://www. 59 102 thenewhumanitarian.org/analysis/2012/07/18/madagascar-s-unforgiving-bandit-lands TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment High food prices are the most common systemic tural produce, while 67 and 35 percent said that their economic shock, and those affected pre-shock con- assets and livestock numbers had fallen. Approxi- sumption levels. Sixty-eight percent of households mately 25 percent reported no change in assets and that experienced food inflation deemed it the most livestock—less than 10 percent had no change in severe shock they had suffered, while 25 percent income, and a meagre 1 percent (probably made up ranked it second and 7 percent ranked it third. Low of the few farmers who use irrigation systems) saw prices of agricultural output affected 2.51 percent of no change in their agriculture production. It is there- households, and none recovered. Fifty-nine percent fore not surprising that food purchases and stocks of them considered this the most severe shock they decreased for many of these households. Similarly, had experienced, while 29 percent ranked it second tropical storms resulted in lower income for 92 per- and 12 percent ranked it third. Finally, inflation of cent of households, a smaller livestock herd for 35 agricultural inputs affected 1.36 percent of house- percent of households, and a decrease in assets holds. None of them recovered and 34 percent of and agricultural output for nearly 75 percent. In the them deemed it the most severe shock they had group affected by tropical storms, most households experienced, with 41 percent ranking it second and reduced food purchases and stocked less food for 25 percent ranking it third. future consumption. Among households affected by drought, 94 percent saw a decrease in income, Droughts are the most common natural hazard. 88 percent harvested less agricultural produce, 70 Among households affected by drought, 83 percent percent had fewer assets, and 41 percent suffered said it was the most severe shock they had experi- a decrease in the number of livestock they owned. enced, while 13 percent ranked it second and 4 per- Food purchases and stocks were also adversely cent ranked it third (Figure 88). This last subgroup affected. Indeed, food inflation emerged as the recovered after five months, whereas the others most severe systemic shock. never recovered. Tropical storms adversely affected 10.24 percent households, which never recovered. In Households are ill-equipped to cope with systemic this group, 71 percent of households considered it shocks. Table 33 presents the main coping mech- the most severe shock they had suffered, while 23 anisms by which households adjust to four major percent ranked it second and 5 percent ranked it shocks—food inflation, drought, tropical storms, and third. The last significant natural hazard was irreg- irregular rains. It is apparent that households are ular rains, which affected 9.28 percent of the pop- better able to cope with food inflation and drought ulation. None of these households recovered and than with tropical storms and irregular rains. More- 56 percent described this as the most severe shock over, households have more room for maneuver they had experienced. Thirty-six percent ranked this to adjust to food inflation than to the other three shock second while 7 percent ranked it third. shocks. The most common coping strategies involve reducing food consumption or using savings, fol- Drought, irregular rains, tropical storms, and high lowed by soliciting help from relatives and friends. In food prices are the most severe systemic shocks. many cases, however, households lack any strategy Among households affected by irregular rains, 90 to deal with shocks. percent reported a decrease in income and agricul- Figure 88: Food price and weather shocks are the most common aggregate shocks (Self-reported) High food prices Drought Tropical storms Irregular rains Floods Systemic Shocks Lower prices of agricultural products High prices of agricultural inputs Armed conflict/Violence/Insecurity High rate of animal diseases High rate of crop diseases Locust attacks or other pests Fires Farmer/herder conflict Landslide 0 5 10 15 20 25 Households A ected (%) Source: Author’s calculations based on 2022 EPM data. 103 TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Table 33: Households have few mechanisms to cope with aggregate shocks (Self-reported) Coping mechanisms High food prices Drought Tropical storms Irregular rains Mobilizing savings 1.51 1.05 0.59 0.75 Help from relatives or friends 1 1.23 0.38 0.36 Government/State aid 0.13 0.47 0.11 0.1 Help from religious organizations or 0.08 0.45 0.07 0.18 NGOs Marry off children 0.01 0.07 0.01 - Change in consumption habits 5.21 1.98 0.77 1.01 Buy cheaper food 8.5 2.3 1.37 1.16 Employed household members took 0.05 0.01 0.03 0.01 additional jobs Members >=15 years inactive or 0.04 0.03 0.04 0.04 unemployed took jobs Children <15 years were made to work 0.1 0.12 0.04 0.04 Children taken out of school 0.11 0.06 0.01 0.04 Migration of household members 0.07 0.07 0.1 0.01 Reduced health/education expenditure 0.35 0.15 0.07 0.13 Obtaining a loan / credit 0.1 0.08 0.07 0.03 Sale of agricultural assets 0.04 0.17 0.01 0.07 Sale of household durable assets 0.23 0.15 0.04 0.04 Sale of land/buildings/houses 0.05 0.11 0.01 0.01 Sale of food stock 0.18 0.14 0.1 0.1 Increased fishing activities 0.04 0.01 0.01 0.01 Sale of livestock 0.31 0.7 0.12 0.28 Entrusting children to other 0.05 0.02 0.03 - households Engagement in spiritual activities 0.18 0.08 0.07 0.06 Practice of off-season cultivation 0.29 1.13 0.37 0.86 No strategy 5.29 6.34 5.5 3.77 Source: Author’s calculations based on 2022 EPM data. To summarize, households in Madagascar are par- income or consumption level and the frequency or ticularly vulnerable to food inflation and inclem- intensity of shocks. The three recent systemic shocks ent weather, which are among the worst systemic include the COVID-19 pandemic and the recession shocks they face. On the other hand, salary loss and it created; food and fuel price increases ignited by job loss, while infrequent, are the most devastat- Russia’s invasion of Ukraine; and weather patterns ing idiosyncratic shocks. Unfortunately, households which resulted in cyclones in parts of the country. often struggle to recover from these shocks due to the lack of effective coping mechanisms available to COVID-19 and its effect on labor, them. This highlights the pressing need for better consumption, services and poverty support and resources to help households weather the many systemic and idiosyncratic shocks they The COVID-19 pandemic started in early 2020, may face. with the first three cases confirmed on March 20. This prompted the declaration of a state of emer- 3. Systemic shocks and their effects gency the next day which was lifted on October on welfare in urban and rural 18, 2020, after the first wave of the pandemic had settings peaked in July/August. The state of emergency was reinstated in early April 2021 and lifted again on Systemic shocks are important drivers of poverty. September 4, 2021. The epidemiological situation As systemic shocks are significantly more frequent worsened during the second wave from March– and touch a large share of the population, the dis- April 2021, resulting in over 600 new daily cases cussion below focuses on the three most recent and and peaking on April 14, 2021, with 854 new cases disruptive shocks that affected the country. While (Figure 89). The daily number of confirmed deaths no conclusions are drawn regarding causal effects, dwindled from a peak of 12.86 in December 2021 to the analysis highlights the correlations between 0.14 as of August 2022. 104 TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Figure 89: Several COVID-19 waves hit Madagascar in 2020/2021 but the country's borders fully opened only in late 2022 Note: COVID-19 cases in Madagascar during the pandemic (daily new confirmed cases). Source: Johns Hopkins University CSSE COVID-19 Data. Major containment measures to curb the spread of gascar was not exempt from this trend. According COVID-19 had a negative impact on welfare. The to the United Nations, “by 2020, the global unem- state of emergency entailed the closure of bor- ployment rate reached 6.5 percent, up 1.1 percent- ders, churches, and mosques; the prohibition of age points from the previous year.61 The number public gatherings; mandatory wearing of masks of people unemployed worldwide increased by 33 and handwashing; and curfews. Although most million, reaching 220 million.” In Madagascar, the measures were lifted when the state of emergency unemployment rate increased from 2.5 to 12.2 per- ended, borders remained closed until March 2022. cent in 2020, starting from the declaration of the The stringent measures affected livelihoods, espe- first state of emergency. In June 2020 alone, 7 cially among urban households, dependent on tour- percent (15 percent in urban areas) of interviewed ism, transport, hospitality, restaurants, and informal workers lost their jobs. The loss of jobs was greater labor. in urban areas than in rural areas, but the rate of job loss tapered off with time (Table 34). To monitor changes in livelihoods, the government launched a series of high-frequency phone sur- Table 34: Urban job losses due to the pandemic veys (HFPS). The HFPS were part of a wider initia- far outnumber rural job losses tive by the World Bank and participating countries June August November May 2021 to track household welfare changes as the pan- 2020 2020 2020 demic evolved. The purpose was to provide a con- Urban 15.3 8.5 0.9 0.8 temporaneous understanding of the effects of the Rural 6.3 3.2 0.1 0.3 pandemic and inform policy responses. The HFPS All 7 4.4 0.3 0.5 aimed to obtain information on various dimensions, Source: Authors based on 2020/1 HFPS data. including knowledge of COVID-19, access to basic services, household employment, and the subjec- tive well-being of households.60 The results of the Many individuals lost their income due to job loss surveys are representative at the national level for or salary reduction. In both June and August 2020, households with access to a telephone. a majority of households experienced a decline in income. But by November 2020, household Labor income incomes were stable and increased for a minority of households after, suggesting a recovery in eco- The onset of the pandemic caused a significant nomic activity (Table 34). increase in unemployment worldwide, and Mada- 60 In Madagascar, four rounds were conducted. In the first round, 1,240 households were interviewed by telephone in June 2020. In the second 105 round, additional respondents were added to reach 1,580 households in August 2020. In the third and fourth rounds 1,580 and 1,345 households were interviewed in November 2020 and May 2021, respectively. 61 https://unstats.un.org/sdgs/report/2021/goal-08/ TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Figure 90: Perception of food insecurity worsened during the pandemic early stages 45% 34.30% 35.70% 40% 35% 30% 25.40% 23.10% 25% 20% 15% 10% 4.20% 5.10% 3.30% 3.30% 5% 0% -5% Mar-20 Apr-20 Jun-20 Aug-20 Sep-20 Nov-20 Jan-21 Feb-21 Apr-21 Jun-21 Jul-21 Households under food insecurity Households under food Security Source: World Bank, High Frequency Phone Survey, 2020 and 2021. Effect on consumption Table 35: Access to health services remained high and increased during the pandemic As labor incomes declined, household budgets June November May 2021 May 2021 became more constrained, and lifestyles changed. 2020 2020 For example, expenditure on healthcare and educa- Urban 82.35 83.50 89.40 0.8 tion lessened while households adopted a survival Rural 90.10 81.30 97.50 0.3 mode and spent a larger share of their incomes on All 88.50 81.80 96 0.5 food. Notably, food security deteriorated for many Source: Author’s calculations based on 2022 EPM data. at the height of the pandemic, especially among households reliant on daily wages from informal Table 36: Access to education services was inter- activities. The situation improved by November rupted only briefly in 2020 2020 as the pandemic subsided. To illustrate this June August November May 2021 point, 4.2 percent of households claimed to be in 2020 2020 2020 a situation of food insecurity in June 2020 due to Urban 18.75 32.60 0 0 COVID restrictions. This is an addition to the people Rural 24.60 42.40 0 0 already in chronic food insecurity condition. By May All 23.90 39.90 0 0 2021, 3.3 percent of households said they were in a Source: Author’s calculations based on 2022 EPM data. situation of food insecurity (Figure 90). Service disruption Poverty Medical services were not significantly disrupted The COVID-19 pandemic increased the poverty rate by the pandemic. The rapid response by the gov- by more than 2.5 percentage points between 2019 ernment to the medical emergency meant that and 2020. As a result of the economic downturn most households that needed medical services (and caused by the pandemic, 81.93 percent of house- could afford to pay) reported having access to them, holds were estimated to be poor in 2020 based on in both urban and rural areas. Moreover, this access the (World Bank, 2022), from 21.4 million people in increased as the pandemic evolved (Table 35). 2019 to 22.7 million in 2020. Some of the increase in poverty was transitory—by 2021, the poverty Access to education was initially reduced, but rate had declined to 2.12 percentage points above schools quickly reopened. Access to education was its pre-pandemic level. The severity of poverty and severely disrupted during the second wave, with 40 inequality also worsened due to the pandemic. percent of households reporting that their children were not going to school—a sharp increase from The households that became poor because of the June 2020, when only 24 percent of households pandemic have different characteristics from those made that claim. By November 2020 schools had that were already poor. The newly poor households fully reopened, and all households reported that had younger household heads, were more likely to their children were back to school (Table 36). live in urban areas, had fewer members and were better educated. Furthermore, the newly poor were more reliant on the services sector for employment, and their household heads more likely to participate 106 TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment in the labor force. This data underscores that pov- cent. Households hit by one cyclone tended to be erty increased during the pandemic due to employ- middle-class, but comparing households hit by two ment losses. cyclones reveals another marked difference in the frequency of cyclone impact between the rich and Cyclones and their effect on household the poor: approximately 81 percent of households in consumption levels and welfare the bottom 10 percent, and more than 50 percent of those in the bottom half of the consumption dis- In addition to the pandemic, households were hit by tribution were hit by cyclones twice, compared with several tropical storms and cyclones. In 2022 alone, only 46.5 percent of households in the top decile and five cyclones—Ana, Batsirai, Dumako, Emnati, and less than 48 percent of those in the top half of the Gombi—damaged food crops, livestock, and infra- distribution. An even starker trend emerges among structure, with many households directly affected households hit by three cyclones: approximately 30 by one or more of them (Map 15). percent of households in the bottom decile of the consumption distribution experienced this ordeal, Poor households were more likely to be hit by mul- versus only 3 percent of those across the rest of the tiple cyclones. The storms had different impacts on distribution. The location of the dwellings of poor households across consumption and income distri- households contributes to their poverty, by exacer- butions (Figure 91). Combining geo-referenced data bating their vulnerability to cyclones. The observed on cyclone trajectories with household location data pattern indicates that wealthy households tend to from the EPM survey, households that lived along live in areas less vulnerable to the physical effects of the cyclones’ paths, the number of times they were cyclones and tropical storms. On the contrary, the hit by cyclones, and their consumption levels could data strongly suggests that poor households are be traced. More than 60 percent of households in more likely to live in parts of the country especially the top 20 percent of the consumption distribution vulnerable to cyclones—usually remote areas with were not hit by any cyclones in 2022, versus only minimal access to infrastructure. 9.7 percent of households in the bottom 20 per- Map 15: Recent cyclones have affected the East of the island more significantly Spatial distribution of total economic damages, 2022 Agricultural damage, 2022 Source: 2022 Madagascar Global Rapid Damage Estimation (GRADE) Report. 107 TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Figure 91: Poor households were disproportionately affected by cyclones in 2022 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1 2 3 4 5 6 7 8 9 10 All 0 1 2 or more Note: Geographical trajectory of cyclones cross-referenced with household location data. Source: Author’s calculations based on 2022 EPM data. This report estimates the economic impact of the ability of poverty rose by 27 percent. The reduction 2022 cyclones on households. Data collection for in expenditure also compromised food security: on the 2020 EPM survey was interrupted due to the average, an affected household consumed 409 pandemic and could only resume in 2022. This fewer calories per day than before the cyclones, interruption provided two survey rounds instead of and the range of foods it consumed shrank by nine one and, by coincidence, between the two rounds items (from 30 to 21). Madagascar was hit by five cyclones. To estimate the economic impact of the cyclones on affected Cyclones also had a proportionally larger nega- households, affected clusters were matched with tive impact on richer households. Figure 92 shows non-affected clusters, using propensity score the aggregated impact of the cyclones on aver- matching to make the treatment and control groups age household expenditure at each decile, both in equal in terms of geographical characteristics. Sub- percentage terms and in absolute terms measured sequently, difference-in-differences estimation in local currency (MGA). The impact was negative techniques were used to estimate the impact of the across the distribution with an increasing trend, i.e., cyclones by household quantile. smaller for those in the poorer deciles than for those in the richer. Average monthly household expendi- Cyclones appear to have had a devastating impact ture decreased by between 12 percent (MGA 76,764) on expenditure, food consumption and poverty. and 48 percent (MGA 3,138,807), bringing down The total expenditure of an affected household inequality for the wrong reason. decreased by 34 percent on average, while its prob- Figure 92: Controlling for household characteristics, richer households suffered larger losses 10% 20% 30% 40% 50% 60% 70% 80% 90% 0% -10% Percentage change -12% -12% in consumption -20% -30% -29% -27% -32% -33% -40% -35% -50% -48% -51% -60% Deciles Note: Impact of cyclones on average household expenditure, by decile. Bold values indicate significance at the 10 percent level, non-bold values and hollow bars indicate an insignificant coefficient. Source: Author’s calculations based on 2022 EPM data. 108 TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Extreme weather events and cyclones also led to Table 37: A small percentage of households was negative impacts on human capital. Using panel not able to buy cooking oil or rice even before the data from the Young Lives project, Marchetta et price al. (2019) investigate how temperature and pre- Households Households Households cipitation variations affect educational attainment needing item needing item Item needing item and able to and unable to and labor force participation. The findings suggest (%) buy it (%) buy it (%) that extreme weather events, such as droughts and Cooking Oil 88.7 82.5 6.25 floods, significantly disrupt schooling and increase Rice 81.6 76.9 4.65 the likelihood of engaging in work activities among Sugar 73.1 69.9 3.16 young adults in Madagascar. These weather shocks Meat 62.5 60.2 2.25 not only impede educational progress but also Beans 56.5 55.8 0.66 lead to an early transition into the labor market, Fish 54.4 51.7 2.67 potentially compromising long-term human capital Bread 41.6 41.2 0.42 development and socioeconomic outcomes for this Potatoes 41.0 40.5 0.51 population. The study highlights the importance of considering climate factors in policies aimed at Eggs 39.8 39.5 0.32 improving educational and employment opportu- Pasta 36.6 36.5 0.12 nities for young adults in developing countries like Lentils 18.3 17.5 0.84 Madagascar. Butter 14.9 14.5 0.38 Flour 14.7 14.0 0.72 How the Ukraine invasion affected Cooking Gas 2.4 2.0 0.37 (LPG) inflation and food-scarcity Source: Author’s calculations based on 2022 High Frequency Phone Survey data. The economy has been adversely affected by the ongoing Russian invasion of Ukraine. Both Russia tion rate and contribute to 64 percent of the infla- and Ukraine are essential grain exporters to Africa, tion basket. In 2022, inflation increased particularly and many countries—including Madagascar—have rapidly in July when the government increased fuel been experiencing war-induced food scarcity and prices by nearly 40 percent. This led to a short term price hikes that are expected to exacerbate poverty transportation cost increase which in turn affected (Figure 93). Fertilizer costs have also been rising, food prices later in the year. This situation was exac- as Russia is the world’s largest exporter of nitro- erbated by deteriorating road infrastructure due to gen-based fertilizers and the second largest for frequent rains. phosphorous- and potassium-based fertilizers. Respondents to the HFPS reported cutting con- The inflation rate has been on the rise and is largely sumption of the most frequently purchased food driven by food products. The annual rate of infla- items. Respondents to the HFPS conducted in June tion has doubled since 2020, reaching an average 2022 listed the most frequently needed and pur- of 8.2 percent in 2022. Inflation has largely been chased items over the previous 30 days (Table 37). driven by food items which experienced a 9.5 infla- At the top of the list are cooking oil (89 percent), Figure 93: Food prices have increased sharply since 2022 Monthly inflation rate, 2022 12.0 10.0 8.0 6.0 4.0 2.0 - er r s il ai in t t e e e e ie ille oû vr ar br br br br M Ju i nv vr A M m o em em A Ju Fé ct Ja e pt ov éc O Se D N Ensemble Riz PPN Produits alimentaires et boissons non alcoolisés Source: INSTAT, April 2023. 109 TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Figure 94: Price increase is the main reason for ing oil, and 70 percent of those who could not buy not being able to buy basic staples meat. The second most common barrier was that 90% households did not have sufficient money to pur- chase those items irrespective of inflation, showing 79% 75% a deterioration in the purchasing power of house- 70% holds. 49% Coping mechanisms against rising prices 33% 30% Reducing consumption is the most frequent coping strategy against price increases. Among the 68 15% percent of respondents that reported an increase in the price of major food items usually consumed by their household, 46 percent reduced their food Rice Cooking oil Sugar Meat consumption, a quarter reduced non-food con- sumption, and one in five relied on savings. Sixteen Price Increase No money percent did nothing to cope with the shock (Figure Source: Author’s calculations based on 2022 High Frequency Phone Survey data. 95). Similarly, out of the 46 percent of respondents who reported an increase in the price of major non- rice (82 percent), sugar (73 percent), and meat (63 food items consumed by their household, 45 per- percent). Out of 89 percent of households that cent reduced non-food consumption, twenty-two reported needing cooking oil, 6 percent did not percent opted to reduce their food consumption, 14 have the ability to buy it. For rice this was 82 per- percent did not enact any coping strategy, and 13 cent needing and 5 percent unable to buy. Such percent relied on savings (Figure 96). gaps in purchasing ability were likely due to pricing, as inflation rates were higher for the goods most in Projected changes in the poverty rate demand. Estimates of the effect of Russia’s invasion of Price increases were the most common reason why Ukraine show a limited short-term impact on pov- households were unable to buy key items (Figure erty. Direct evidence of the impact of the Ukraine 94). During the 30 days prior to the survey prices had invasion is limited due to a lack of data. A combi- risen by 10 percent on average, and flour, potatoes, nation of macro level data (World Bank 2022) with and fish increased by more than 50 percent (see the EPM and HFPS conducted in 2022 was done to Table A5.2 in Annex 5). Rising prices were reported estimate changes in consumption resulting from the as the main obstacle by 90 percent of households price hikes and subsequent impacts on poverty.62 unable to buy sugar, 79 percent of those unable to Madagascar’s poverty rate in 2024 is expected to buy rice, 75 percent of those unable to buy cook- rise by less than 0.2 percentage points relative to Figure 95: Reducing food consumption is the most common coping mechanism against food price increases Reduced food consumption 46% Reduced non-food Consumption 25% Relied on savings 20% Coping Mechanism Did nothing 16% Engaged in additional income generating activities 10% Borrowed from friends & family 9% Credited purchases 6% Sale of ASSETS (AG AND NO-AG) 3% Received assistance from Friends & family 2% Took a loan from a financial institution 1% Sold harvest in advance 1% Agricultural produce was used as food 1% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Households adopting mechanism (%) Source: Authors based on 2022 High Frequency Phone Survey data. 62 The World Bank regularly prepares projections of key national accounts statistics for 185 countries, using a growth projection model called 110 MFMOD. The Macro Poverty Outlook published in the Spring 2022 (World Bank 2022) incorporated the impact of Russia’s invasion of Ukraine and the resulting inflationary pressures, and well as projected inflation from 2022 to 2024. TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Figure 96: Reducing non-food consumption is the most frequent way of dealing with non-food price shocks Reduced non-food Consumption 45% Reduced food consumption 22% Did nothing 14% Relied on savings 13% Coping Mechanism Borrowed from friends & family 9% Engaged in additional income generating activities 6% Credited purchases 5% Sale of ASSETS (AG AND NO-AG) 3% Received assistance from Friends & family 3% Sold harvest in advance 2% Took a loan from a financial institution 2% Reduced expenses 1% Delayed payment obligations 1% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Households adopting mechanism (%) Source: Author’s calculations based on 2022 EPM data. Figure 97: Poverty is expected to increase further due to inflation Source: Wu and Yoshida, 2022. the pre-war scenario (Figure 97). The country’s nom- inal GDP per capita is projected to outgrow inflation, and thus the growth rate of real GDP per capita is anticipated to remain positive in 2023 and 2024. A slightly more pessimistic scenario could materialize if food prices increase faster. Wu and Yoshida (2022) estimate that if food inflation rises twice as fast as non-food inflation due to the invasion of Ukraine, poverty in 2024 will be 0.6 percentage points higher than in the pre-war scenario. However, if the opposite occurs (i.e., if non-food inflation rises twice as fast as food inflation), the uptick in the country’s poverty rate in 2024 will only amount to 0.2 per- centage points relative to the pre-war scenario. 111 TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Annex 2 Table A2.1: Only 6 countries have suffered a secular income decline since independence GDP per capita – 1960 GDP per capita – 2020 Country (Constant 2015 US$) (Constant 2015 US$) HTI – Haiti 1,714.10 1,322.80 COD – Democratic Republic of Congo 1,254.70 487.40 NER – Niger 744.00 519.70 MDG – Madagascar 818.30 433.80 CAF – Central African Republic 583.60 375.20 BDI – Burundi 290.70 263.40 Source: World Bank, World Development Indicators Database. Table A2.2: Median variation in prices for food and cooking items Median unit price Median unit price Unit of Change in median Item (in MGA) 30 days (in MGA) at time of measurement unit price (%) before survey survey Kg 2,300 2,400 4.3 Packet 100,000 100,000 0 Rice Tin/Kapoaka 650 650 0 Gram 6 6 0 Kg 8,333.33 8,333.33 0 Cooking Gas (LPG) Cylinder (gas) 40,000 40,000 0 Liter 9,000 10,000 11.1 Cooking Oil Packet 500 500 0 Kg 12,000 14,000 16.7 Meat Gram 12 12 0 Kg 10,000 10,000 0 Packet 2,000 3,000 50 Fish Batch 2,700 3,000 11.1 Gram 10 10 0 Kg 4,000 4,000 0 Packet 1,200 1,200 0 Sugar Tin/Kapoaka 1,200 1,500 25 Gram 4 4.8 20 Kg 2,000 2,200 10 Packet 1,000 1,500 50 Potatoes Batch 500 500 0 Tin/Kapoaka 900 900 0 Gram 2 2 0 Kg 6,000 6,000 0 Butter Packet 4,000 4,000 0 Gram 16 16 0 Kg 3,000 3,500 16.7 Pasta Packet 1,200 1,200 0 Kg 1,000 1,000 0 Bread Packet 500 600 20 Kg 3,400 3,500 2.9 Lentils Tin/Kapoaka 800 1,000 25 Kg 3,800 4,000 5.3 Packet 1,500 1,500 0 Flour Tin/Kapoaka 1,200 1,200 0 Gram 5 8 60 Packet 600 650 8.3 Eggs Batch 600 700 16.7 Kg 4,100 4,800 17.1 Beans Packet 1,200 1,500 25 Tin/Kapoaka 1,000 1,200 20 Source: Author’s calculations based on 2022 EPM data. 112 TOC Chapter 5 Madagascar High vulnerability to shocks drives short term poverty but has long term consequences Poverty and Equity Assessment Table A2.3: Food inflation coping mechanisms adopted by households, by province (%) Antanana- Fianarant- Antsira- Toamasina Mahajanga Toliara Total rivo soa nana Reduced food 49 38 44 50 50 51 46 consumption Reduced non-food 30 13 28 31 24 23 25 Consumption Relied on savings 28 23 7 21 18 11 20 Did nothing 14 20 18 16 11 18 16 Additional income 5 14 6 10 14 10 10 generating activities Borrowed from friends 11 5 14 8 6 11 9 & family Credited purchases 6 8 4 5 7 3 6 Sale of assets 4 1 3 6 4 1 3 Assisted by friends & 3 3 1 2 4 0 2 family Took a loan from a 0 1 1 3 4 0 1 financial institution Sold harvest in advance 0 2 2 1 0 1 1 Agricultural produce 0 2 0 0 0 0 1 was used as food Source: Author’s calculations based on 2022 EPM data. Table A2.4: General inflation coping mechanisms adopted by households, by province (%) Antanana- Fianarant- Antsira- Toamasina Mahajanga Toliara Total rivo soa nana Reduced non-food 42 41 49 48 56 43 45 Consumption Reduced food 28 17 23 11 19 27 22 consumption Did nothing 11 19 15 11 10 24 14 Relied on savings 18 12 2 24 10 12 13 Borrowed from friends 11 4 7 11 8 10 9 & family Additional income 5 11 5 2 8 4 6 generating activities Credited purchases 7 7 6 5 0 2 5 Sale of assets (AG AND 3 3 2 5 2 0 3 NO-AG) Assisted by friends & 3 5 1 6 0 2 3 family Sold harvest in advance 0 3 0 8 2 6 2 Took a loan from a 1 1 4 3 2 4 2 financial institution Reduced expenses 0 1 0 3 2 0 1 Delayed payment 1 0 0 0 2 0 1 obligations Source: Author’s calculations based on 2022 EPM data. 113 TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment Chapter 6 Charting a course for change Madagascar's domestic and global challenges became opaquer and more centralized, indicating make reducing poverty a daunting task, but the intensifying budget management challenges and cost of inaction is too big to overlook. Domestic illustrating the impact of increasingly cumbersome challenges include political instability, elite capture, budget authorization procedures. high population growth, and resource depletion. Combined with climate change, growing volatility As the latest Systematic Country Diagnostic points of prices and other global challenges, these obsta- out, constraints to growth and poverty reduction cles deepen the country's poverty trap. This con- are closely interlinked (World Bank Group, 2022). text increases the urgency for effective and con- These constraints can be grouped in four catego- sensus-driven public policy interventions to lay the ries: (i) weak governance; (ii) low investment in phys- foundation for sustainable development and resil- ical (i.e., infrastructure), human and natural capital, ience. Better infrastructure, support to job creation (iii) low productivity and stagnant structural social and agricultural productivity, and stronger human and economic change; (iv) high and rising vulner- capital investment and assistance can play a crucial ability to shocks. Each of these constraints pre- role in reducing poverty. To be effective, such poli- vents broad-based development and reinforces cies need a better quality of governance. the others, so that single-sector approaches are not enough to make a real impact on growth and To this end, the government launched the "Plan poverty reduction. The resulting priorities from the Emergence Madagascar" (PEM). The PEM is struc- analysis follow in a similar organizing framework: tured around three pillars: social, economic, and enhance economic opportunities, improve service environmental to attract investment and address delivery, build resilience, and improve governance long-standing development issues. The PEM com- (Box 3). prises 13 commitments (“Velirano”) to grow real GDP by 8 percent per year over the next 17 years This chapter focuses on additional policy recom- and reduce poverty from 75 to 35 percent by mendations to complement the priorities outlined 2040. Despite relying on optimistic macroeco- in the SCD Update. The chapter outlines policy rec- nomic assumptions, the PEM is well-structured, ommendations supported by economic literature with a clear strategy to mobilize public and private and evidence from other low-income countries to resources for investment, precise result indicators, tackle the barriers to reduce poverty in rural and and measures to enhance dialogue with develop- urban areas, considering their drivers in each case. ment partners, the private sector, and civil society. 1. Improving agricultural Implementing the PEM requires a radical departure productivity, market connectivity from the historical performance of government and resilience policies in the country. To start, government spend- ing reached a high of 16.7 percent of GDP in 2020 Agricultural productivity in Madagascar is lower (it then fell to 14.2 percent in 2021), remaining below than in comparator countries, due to the scarce most aspirational peers, with fiscal space structur- use of inputs and deteriorating irrigation infra- ally constrained by a small revenue base, spending structure. Annual yields of rice, the staple crop, do inefficiencies, limited capacity, and large transfers not exceed 3 t/ha even in fertile highland regions. to poorly managed SOEs (World Bank, 2023). As a While extreme weather events such as floods and result, investments to improve connectivity infra- droughts pose a recurring challenge, the limited structure and essential public services have been utilization of inputs—e.g., fertilizers, chemicals, and insufficient to boost growth. In recent years, insti- improved seeds—is a key issue. Less than 12 percent tutional and governance weaknesses have inten- of farmers purchase new seeds for planting, and sified. For instance, budget execution rates, which less than 17 percent apply pesticides, herbicides, are structurally low, fell further during the COVID- or fungicides. As a result, both yields and produce 19 crisis, while the selection of investment projects quality tend to be low. 114 TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment Box 3: Policy priorities from the Systematic Country Diagnostic Update To reduce poverty, improve living standards, and catch up with peers, Madagascar’s government needs to maintain a 7-percent annual GDP growth rate. This is achievable by implementing holistic policy reforms grouped under the following three objectives: (i) mobilizing productivity-enhancing investments; (ii) boosting human capital and resilience to shocks of the poor; and (iii) strengthening citizen engagement. Mobilize productivity-enhancing investments to accelerate economic and spatial transformation. Attracting private investment requires improved market access and contestability of incumbent companies. This entails reducing barriers to business creation, prohibiting cartels, ensuring greater consistency of rules applying to private investments across various sectors and special regimes, and easing processes for land acquisition and transfer. Connecting undeserved rural areas, supporting agglomeration around secondary cities, and developing effective industrial poles are key priorities to accelerate spatial transformation, and will require renewed efforts to support private investments in the power, water, ICT and transport sectors, alongside much needed reforms of public entities operating in these sectors. Boost human capital and resilience to shocks of the poor to improve their welfare. Improving health and learning outcomes will require the mobilization of additional domestic resources and reforms. Considerable efficiency gains in education and health sectors are possible in the short term, includ- ing through reforms in human resources, public financial management, results-based financing, and decentralization. Greater access to prenatal care, assisted deliveries, and effective nutrition programs will be key to reduce the incidence of child and maternal mortality as well as to reduce high stunting rates, while improved access of adolescent girls to secondary education will contribute to reduce high fertility rates in underserved areas. Strengthening social protection systems will ensure greater resilience to shocks and can help stimulate better nutrition outcomes and increase the demand for health, education, and water services, so as to improve human development outcomes. Strengthen engagement of citizens to deliver better policy outcomes and limit state capture. Making the state more responsive to the needs of citizens, especially the poor, requires regulatory reforms to increase the separation of private interests from public policies, stronger mandates and independence of anti-corruption institutions, the judiciary, and the court of accounts, better enforce- ment of rules on asset declarations, asset recovery, and whistleblower protection, and international tax transparency. Greater autonomy and control of decentralized entities are also critical to improve service delivery and reinforce the link between public policies and citizens. Finally, community-driven solutions can help deliver better infrastructure and social services, particularly in more remote areas of the country where the state presence is limited. Build resilience and maintain macroeconomic stability. To build resilience to future shocks the coun- try needs scaled up interventions in nutrition, safety net programs and connectivity infrastructures. Given the limited fiscal resources, selectivity, planning, and execution of public investment projects need to be improved, together with reforms to accelerate domestic resource mobilization and to ensure prudent debt management and an independent monetary policy. Source: World Bank Group (2022). Access to markets complicates the low productivity Many farmers lack appropriate infrastructure to challenge. The market for cash crops is highly pro- store their harvest, leading to significant losses tected and only accessible to 20 percent of farm- from spoilage. Over half of unsold produce is either ers, while the rest are relegated to the horticulture stored on roofs or set aside for immediate use. Only market. The market for rice is relatively accessible, 10 percent of farmers have access to more secure but prices vary significantly across regions, largely storage methods, such as storehouses and sheds. due to high transport costs made worse by the This then exposes farmers to high spoilage, food deteriorating transport infrastructure. insecurity, and poor diet due to lack of production diversity. 115 TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment Transforming agriculture and food systems in as discussed in Chapter 2. For areas with high Madagascar hinges on three key priorities: potential, the most efficient policy is to undertake low-cost investment, including access to inputs 1. intensifying and diversifying food production, such as fertilizers, improved seeds, finance, insur- ance, and extensions services. These low-cost 2. improving rural connectivity and the efficiency of investments should help farmers tap into existing domestic markets, and potential by increasing their efficiency levels. For low potential areas, the most efficient policy is to invest 3. making agriculture more financially resilient to in high-cost infrastructure that alters the produc- climate change. tion possibilities’ frontier. These investments include feeder roads, irrigation schemes, and developing Effective interventions require a targeted spa- agricultural value chains. tial approach. The analysis for (1) and (2) considers the untapped agricultural potential of each region Map 16 provides an overview of the regions where described in Section 3 of Chapter 2, to identify low-cost and high-cost investments could be pri- an optimal combination of investments in feeder oritized. Betsiboka, which features both high crop roads and irrigation projects by region. A stochas- potential and high poverty is a typical region where tic frontier analysis predicts the agricultural poten- low-cost investment could reduce poverty (Map 16, tial of a region under optimal conditions compared panel c). Investments into improved agricultural effi- to its current performance. The difference is the ciency in this region are expected to yield the high- “untapped potential”—defined as the upper limit to est return in terms of poverty reduction. The sec- realistic returns from development programs and ond-highest level of priority would be investments policies in that region. in the regions of Sofia, Menabe, Alaotra-Mangoro, Bongolava, Amonon’i Mania, Haute Matsiatra, Overlaying crop potential with poverty rates Menabe, and Androy where the correlation between reveals the regions where agricultural investment crop potential and poverty rate is average. For the can make the greatest impact on poverty reduc- case of high-cost investments, priority reigions tion. Agricultural investment can be grouped into include Melaky, Ihorombe, Anosy, Atsimo-Atsina- low-cost and high-cost infrastructure investment, nana, and Atsimo-Andrefana. Map 16: Some regions with high poverty also have untapped agricultural potential (a) Farm potential (b) Multidimensional poverty (c) Farm potential and poverty Farm potential at region level Multidimensional poverty rate → 4.291,798 to 7,777,114 94.06 to 97.82 3,772,030 to 4,291,798 88.92 to 94.06 3,486,774 to 3,772 830 83.23 to 88.92 Diana 3,245,672 to 3,486,774 76.25 to 83.23 Sava → 1,734,899 to 3,245,672 6.46 to 76.25 Missing Missing Sofia Boeny Analanjirofo Betsiboka Melaky Alaotra- 84% Mangoro Analamanga Bongolava Atsinanana Itasy Vakinankaratra Menabe Amoron’i Mania Vatovavy Matsiatra Fitovinany Ambony Ihorombe Atsimo- Atsino Andrefana Atsinanana Anosy Androy Source: Authors’ estimation. 116 TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment Intensifying and diversifying food Improving rural connectivity and the production entails better irrigation efficiency of domestic markets is crucial systems Gains in yield and enhancements to land use meth- The rehabilitation, effective operation and main- ods will only be sustainable if appropriate market tenance of irrigation infrastructure is necessary structures incentivize them. Feeder road invest- for water availability and management. This is crit- ments need to be targeted strategically to ensure ical for maintaining food security, particularly by that markets offer a sufficient return on investment. increasing rice output. Additionally, promoting the Moreover, improved transportation can help make a use of improved inputs—such as climate-resilient greater variety of produce available to consumers, seeds and fertilizers—and conservation practices can which in turn can result in healthier diets. unlock significant yield gains. Small-scale, water-ef- ficient irrigation technologies can also boost food Starting from data on farm revenue, the impact of production beyond rice. In dryland areas, low-cost investments in feeder-road infrastructure on travel water harvesting technologies—such as sand dams time to markets is estimated. Unlike the previous and water reservoirs—can capture and store rainfall analysis, farm potential as opposed to crop poten- water for human consumption and productive use. tial is the outcome of interest, since farmers also produce livestock that they would like to sell. Farm To prioritize irrigation investment, low agricul- potential captures the potential holistic impact tural potential and high poverty areas are overlaid. of market access. The basic assumption is that The analysis estimates first the effects of expand- the relationship between farm revenue and travel ing access to irrigation (via river diversion) on crop time is negative—i.e., longer travel time to mar- potential. The potential gains are then correlated kets result in lower revenues, either because they with poverty rates in the regions with low poten- cannot effectively reach markets, or because they tial, which results in priority regions for irrigation must bear high transportation costs. However, the investment including Melaky, Atsimo-Andrefana, assumption of a negative relationship may not hold Ihorombe, Anosy, and Atsimo-Atsinanana (Map 17). if a significant portion of farms primarily produce This exercise can also be done at the district level. for self-consumption, or if only a limited number of Map 17: Large gains in agricultural productivity can be obtained through improved irrigation Increase in crop production potential Increase in production due to irriga- Increase in production due to irriga- from irrigation at region level tion and poverty at the region level tion and poverty at the district level Increase in crop production potential → → from irrigation at region level High Missing Missing Diana Sava → → Low Sofia Boeny Analanjirofo Betsiboka Alaotra- Mangoro Melaky 84% Analamanga Bongolava Atsinanana Itasy Vakinankaratra Menabe Amoron’i Mania Vatovavy Matsiatra Fitovinany Ambony Ihorombe Atsimo- Atsino Andrefana Atsinanana Anosy Androy Source: Authors’ estimation. 117 TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment markets offer price premiums. Taking farm poten- and improving market access. Therefore, targeted tial, travel time, and poverty rates allows to iden- investments in these regions can play a vital role in tify regions where road infrastructure investments promoting agricultural development and poverty would improve market access and reduce poverty reduction. the most (Map 18). The highest-priority region is Betsiboka, followed by all regions in the South Five climate-smart agriculture packages were (except Anosy), Menabe (known for its isolation), identified through expert consultation and data and Sofia (the country’s second-largest rice-pro- analysis.63 The first package includes measures that ducing region, which is known for its limited access can improve access to quality seed for diverse crops. to markets). This involves establishing decentralized variety testing networks involving the rice research center To conclude, investing in irrigation and rural road FOFIFA and partners such as NGOs and develop- infrastructure will have the largest impact on agri- ment organizations, supporting local seed producer cultural output and poverty reduction in Besti- groups, and reviewing FOFIFA's budget allocations boka, Menabe, Androy, Atsimo-Andrefana, and and facilities for seed production and storage to Atsimo-Atsinanana. Priority regions for investment meet varietal demand. Second, improving access in irrigation are Melaky, Bestiboka, and Androy, fol- to irrigation during drought events can be done by lowed by Atsimo-Andrefana, Menabe, Bongolava, introducing (solar) pumps for nurseries to ensure Haut Matsiatra, Ihorombe, and Atsimo-Atsinanana, timely preparation of rice seedlings, installing drip with availability of surface water a critical determi- irrigation for off-season vegetable farming, and nant. Similarly, the priority regions for investment in developing an early warning system on rain anom- road infrastructure are Betsiboka, Sofia, Menabe, alies to anticipate irrigation needs. Third, crop nutri- Androy, Atsimo-Andrefana, and Atsimo-Atsina- tion management can be improved through tailored nana. A combined investment in irrigation and road extension programs that include decision-support infrastructure in Bestiboka, Menabe, Androy, Atsi- tools for a more effective use of fertilizers and pre- mo-Andrefana, and Atsimo-Atsinanana can lead to cision agriculture. Post-harvest losses and distress significant poverty reduction by increasing yields sales can be reduced by establishing community Map 18: Regions with high poverty and high development potential should be prioritized for feeder roads investment Farm potential at region level Farm potential and travel time to Farm potential, travel time to cities, cities at the region level and poverty rates at the region level Farm potential at region level → → Missing 3,937,049 to 7,057,775 3,694,292 to 3,937,049 3,153,998 to 3,694,292 Diana 2,888,220 to 3,153,998 → Sava → 1,570,174 to 2,888,220 Missing Sofia Boeny Analanjirofo Betsiboka Alaotra- Mangoro Melaky 84% Analamanga Bongolava Atsinanana Itasy Vakinankaratra Menabe Amoron’i Mania Vatovavy Matsiatra Fitovinany Ambony Farm potential Ihorombe 3e+06 Atsimo- Atsino Andrefana Atsinanana Anosy 5e+06 7e+06 Androy Source: Authors estimation. The process began by collecting information on climate risks and their impacts on agricultural value chains, as well as coping mechanisms and 63 118 supporting interventions used by value chain actors in the Lake Alaotra and Sofia regions. This was done through expert interviews and literature review. A consultative workshop was then organized to prioritize climate risks and coping mechanisms across value chains. Experts identified and prioritized barriers that prevent the wider adoption of key coping mechanisms using a problem tree exercise. Barriers were grouped by themes for the intervention packages. Effective and feasible interventions were then prioritized among the interventions compiled based on interviews and literature review. TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment granaries that provide credit to farmers using crops remotely sensed satellite-based indices which offer as collateral and training farmers on storage tech- a high degree of geographical resolution. A common niques to reduce losses and maintain crop quality. example is Normalized Difference Vegetation Index Lastly, improved market access and product value (NDVI). NDVI is a measure of vegetation greenness for producer groups can be achieved by support- and vigor, which can be used to estimate crop yields ing farmers' organizations in improving product and monitor drought conditions. To be effective, quality and uniformity. Assistance can also be given index insurance contracts will have to be tailored to to exporters to identify opportunities for contract regional nuances in terms of crops grown, cultiva- farming and to develop farmers' capacities and tion methods, timing and duration of growing sea- infrastructure to facilitate contract agreements. sons, and exposure to weather risk. Promoting index insurance to shield Second, the prevalence of poverty among Mal- agrarian households from recurring shocks agasy farmers makes it necessary to reduce the and improve their access to finance costs associated with the provision of index insur- ance. Evidence shows that the rate farmers are Index insurance is an innovative tool to help willing to pay for insurance is often lower than the farmers cope with weather related risks. Agricul- actuarily fair premium rate. Mobile banking, which tural index insurance is a type of insurance that is has revolutionized financial services for the poor in designed to protect farmers and agricultural pro- Madagascar and throughout the developing world ducers from the risks associated with weather over the past decade, offers a model for the low- events, such as drought, flood, or excessive rainfall, cost provision of insurance products. However, the that can adversely affect crop yields and incomes. coverage of mobile networks and mobile money It is a financial tool that helps farmers manage risk remains limited, and logistical issues around the and protect their livelihoods. In agricultural index collection of premiums and distribution of payouts insurance, a farmer pays a premium to an insurance need to be resolved. Thus, donors and international company, which provides coverage against losses in agencies are encouraged to experiment with mobile crop yields due to specific weather events, such as marketing of agricultural insurance products as part drought or excessive rainfall, based on an index that of their future index insurance initiatives. measures the severity of those events. The index can be based on several variables, including rain- Third, donors and international agencies need to fall, temperature, and other weather-related factors reassess how index insurance can best be deployed that can impact crop yields. to help poor farmers. It is becoming apparent that index insurance is universally of little value to farm- Index insurance can help enhance the resilience of ers, primarily due to irreducible basis risk (i.e., the agrarian communities to repeated natural shocks. If failure of index insurance to cover all losses that the index value falls below a certain threshold, indi- may be experienced by the insured). Mishra et al. cating a significant reduction in crop yields due to a (2021) show that index insurance is better suited to specific weather event, the farmer receives a payout addressing the portfolio risk borne by agents in the from the insurance company to compensate for the agricultural marketing chain—including lenders, pro- loss. This payout can be used to cover the cost of cessors, and nucleus farmers—who enter into con- inputs, such as seeds and fertilizer, or other expenses tractual agreements with large numbers of farmers associated with farming. The advantage of agricul- simultaneously. By mitigating the risk associated tural index insurance is that it is often more afforda- with lending and contracting, index insurance has ble and accessible to farmers than traditional crop the potential to expand access to credit and con- insurance. It also eliminates the need for time-con- tract marketing for smallholders, at reduced interest suming and costly claims processing, as payouts are rates and on improved terms of trade in the face of automatically triggered by the index value. growing climate risks. To make index insurance viable in Madagascar, sev- No clear answer has yet emerged as to how lenders eral issues need to be addressed: and processors can effectively adopt index insur- ance, while maximizing benefits for poor farmers. First, the development of new and improved index Contingent credit—whereby payouts from index insurance contracts that provide better risk pro- insurance contracts go directly to the lender for tection is paramount. A pilot rainfall insurance the purpose of retiring the farmers’ outstanding program is underway in Madagascar. Based on debt—is a promising approach but has not been lessons learned from other African countries, we fully explored (Mishra et al, 2021). Questions per- recommend index insurance contracts that rely on sist as to how lenders should adapt loan recovery, 119 TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment debt restructuring, and portfolio risk management ter and President for all expenditure commitments practices to contingent credit, and the use of index above MGA 200 million and all transfers contrib- insurance to support production marketing con- uted to further delays in the budget execution. tracts has received limited attention. Madagas- car’s ongoing pilot rainfall insurance program can The country receives a significant amount of Offi- help pave the way for more efficient uses of index cial Development Assistance (ODA), with social insurance by working closely with multiple lenders sectors being the largest destination of external and processors, and by assessing, through practical financing. Although the share of ODA allocated to experimentation, the benefits of alternative lending social sectors declined from 60 percent in 2014 to and contracting practices for poor farmers. 44 percent in 2019, health still dominates the share of ODA allocation, while education stabilized and 2. Raising the productivity and social assistance programs increased significantly quality of services of urban areas since 2016 (Razafindravonona, 2023). However, compared to Sub-Saharan countries, Madagascar Investing in more and better human capital received less than half of ODA, when adjusted by population size. The recent decline in returns to education com- bined with the slow growth in good quality jobs The education sector faces several supply-side suggest that human capital investment may not challenges, including workforce management, be economically attractive. In light of the still rapid teacher effectiveness, expenditure management, population growth, building more and better human and school management. Teachers' low academic capital is the only option to ensure that future gen- qualifications and regional imbalances in teacher erations will be able to live better and more produc- deployment create disparities in educational out- tive lives and exit poverty sustainably. The country comes between rural and urban areas. Inadequate needs to invest more, but especially to improve the school management, teacher absenteeism, and low efficiency of its human development sectors so that job satisfaction among public primary school per- resources invested deliver good quality outcomes. sonnel further worsen the situation. Corruption in To achieve this, improving governance in education the education sector, including recruitment, is on and health sectors is key. the rise despite interventions to curb such prac- tices (AMD International, 2021; World Bank, 2021b, Madagascar has made progress in increasing the 2021a). budget allocation to social sectors, but it remains below that of its peers in Sub-Saharan Africa and Most national education is provided by contract Low-Income Countries. In 2020, social sector teachers, especially in preschool and primary. spending accounted for 5.1 percent of the coun- Despite the importance of the resources allocated try's GDP, up from 4.4 percent in 2017. Despite this to national education, an analysis of the 2018- increase, Madagascar still lags SSA and LIC averages 2019 data shows that communities and parents of in terms of percentage of GDP allocated to social students are responsible for a large proportion of spending (World Bank, 2023). Social sectors are the community teachers (FRAM). These teachers rep- second largest recipients of government spending resent nearly 78.4 percent of teachers in preschool, in Madagascar, with administrative functions taking 21 percent in primary school, 32 percent in middle up the largest share at 49.8 percent. The education school and 27 percent in high school.64 In addition, sector is the largest recipient of social spending, the input allocation for public schools remains insuf- accounting for 73 percent of the total, followed by ficient. For instance, pupils have less than one seat health at 21 percent, nutrition at 3 percent, water per pupil (0.24 in primary school, 0.4 in junior high, and sanitation at 2 percent, and social assistance at 0.46 in high school) and textbook allocations are 1 percent. also insufficient for students (UNICEF, 2018). Although the budget allocation to social sectors Finally, the school calendar is not aligned with the has increased in recent years, the execution rate agricultural (rice) season. The school calendar fol- of social spending has declined. The actual budget lows a five bimester system, September-October; execution stagnated at 3.6 percent of GDP over the November-December; January-February; March- period between 2017 and 2020. The execution rates April; and May-June. This calendar is not designed of spending in social sectors declined, from 82.3 to align with the agricultural and the cyclones’ sea- percent in 2017 to 72.7 percent in 2020, indicating sons, and the Ministry of National Education has growing challenges in budget implementation. The failed to bring it in line with Madagascar’s weather, exercise of “warrant” authority by the Prime Minis- instead maintaining the alignment to the French Civil servants represent only 8 percent in primary school, 23 percent in middle school and 36 percent in high school. Subsidized FRAMs 64 120 account for 7.8 percent of preschool teachers, 35 percent of primary teachers, 6 percent of middle school teachers and 1 percent of high school teachers. (UNICEF, 2018) TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment school calendar. There is evidence that aligning the observance of the national guidelines for manag- school calendar with the agricultural season can ing critical health conditions. To ensure access to increase school attendance and educational attain- necessary medical supplies, equipment, and drugs, ment in rural areas where agriculture is a major eco- the government needs to strengthen SALAMA (the nomic activity, for instance in Malawi (Allen, 2022). medical procurement agency), include equipment in the material accounts in inventory measurement, To improve the education system the government and conduct frequent inventory checks. The sector needs to implement deep governance reforms and also needs to improve the efficiency and transpar- invest in teacher quality and fair remuneration. ency of health sector financing. Finally, promoting First, the Ministry of National Education needs to community-based health promotion and education significantly raise the skills of teachers, especially at programs, strengthening community health com- public primary schools and in key subjects such as mittees, and improving communication and feed- mathematics and languages. Using cost-efficient back mechanisms between health providers and technologies can boost the efficacy of teacher train- communities are essential to improve health service ing (Quota et al., 2022). Deploying more qualified delivery. teachers and paying them salary premia for working in rural areas and ensuring adequate teacher remu- Improving conditions for investment and neration for FRAM teachers would reduce the dis- job creation parity between rural and urban educational attain- ment. Governance failures in education also require As discussed in Chapter 3, urban poverty has been urgent attention. For instance, establishing an rising in part because the labor market is unable to independent regulatory body to ensure the quality, provide decent employment conditions to most of efficacy, and efficiency of educational institutions, the urban workers. While issues of low-quality labor expanding digital payments of teacher salaries, supply are real, the main problem is the low creation ensuring clear hiring procedures, job descriptions, of wage employment and the low salaries prevalent and improving training for school directors. in the labor market. Inadequate infrastructure, lim- ited access to finance, and weak institutions have all Deficiencies in the quality of health services also contributed to maintaining low private investment, contribute to lower levels of human capital. For low productivity and low levels of competition in instance, the Service Delivery Indicators (SDI) sur- the private sector, which favor rent extraction (from veys show that the country performed worse than consumers and workers) by powerful firm owners. other African countries on most indicators, despite Major reforms to improve the business environment having a similar outpatient caseload (World Bank, would increase private investment and allow more 2017) and (World Bank forthcoming 2023). More- firms to enter the market and compete for jobs and over, health workers' absence rates more than dou- customers, potentially raising wages and lowering bled between 2016-2021, exceeding 40 percent in prices. 2021.65 Diagnostic accuracy also declined from 29.4 percent of the health conditions correctly diag- Improving the investment climate is crucial to nosed by health providers in 2016 to 23.2 percent in boost private investment, especially against slow- 2021.66 Availability of drugs and functioning equip- ing global growth. As a frontier market, Madagascar ment also declined over the period (from 42.5 per- needs to make visible improvements in the invest- cent in 2016 to 23.5 percent in 2021 for hospitals) as ment climate to retain existing investors and attract well as essential drugs and vaccines for under-five a wider pool of investors (World Bank Group, 2021). children (from 52.6 percent in 2016 to 37.3 percent Policy predictability, particularly of trade policy, in 2021) and running water (from 53.4 percent in investment rules, and access to land are crucial to 2016 to 38.5 percent in 2021). increase private investment, which has been fall- ing for the last decade. For instance, trade restric- The Ministry of National Health needs to address tions reduce investors’ appetite for long-term pro- health worker absenteeism, declining diagnostic jects, and trade policies protecting incumbent firms and treatment accuracy of health providers, and create disincentives for new entrants and increase lack of essential drugs and supplies. One strat- costs and prices, hurting consumers. The multiplic- egy to reduce health worker absenteeism is to ity of laws and regulations governing investment, improve supervision, management, motivation, and coupled with their uneven application, are often a accountability mechanisms. To improve the quality source of concern in the private sector. Adopting of the workforce, pre-service training needs to be an investment law that defines the different forms improved and should include a uniform qualification of investments, lists the activities reserved for local examination, and for practitioners, monitoring their investment, and harmonizes the various laws and The absenteeism rate was highest at the hospital level, where nearly half of the health workers were absent on a typical day. 65 121 Nurses and midwives showed a significant decline in diagnostic accuracy. Treatment accuracy of health professionals also showed significant 66 deterioration, with lower treatment accuracy scores for all patient conditions except for malaria with anemia. Adherence to clinical guidelines (percentage of clinical cases treated as expected) dropped from 29.4 percent to 25.2 percent. The management of maternal and neonatal complications, the last indicator of health providers' competence, was relatively constant, dropping slightly by less than 1 percent during the period. TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment regulations would be a step in the right direction. on unreliable daily incomes and has created a new Streamlining the processes for obtaining construc- class of urban poor. Some of those affected have tion permits, connecting to utilities, paying taxes, since recovered from the shock, but many remain and assessing credit risk would boost investment in poverty. Repeated cyclones have underscored and encourage job creation. the long-standing neglect of infrastructure in the country, especially in areas where poor households Promoting inclusive and sustainable growth tend to live. However, the ongoing inflationary pres- requires policies that reduce the concentration of sure has had a limited impact on poor households, market power and promote a level playing field mainly because they have been able to lessen their for all firms. The concentration of market power in dependence on the items most affected by inflation a few firms can limit competitiveness and produc- thus far. tivity, resulting in economic growth that is neither inclusive nor sustainable. The proposed measures Madagascar’s investment in social assistance is include building on what is already working, lever- among the lowest in the world. The country only aging experiences from other countries, implement- allocated an estimated only 0.3 percent of GDP to ing well-designed performance-based measures, social safety net spending, compared to an aver- making tax expenditure statements transparent, age of 1.2 percent in sub-Saharan Africa (UNICEF, applying open contracting principles, clarifying 2018, 2023; World Bank, 2023) and barely above considerations in determining dominance, prohib- the 0.26 percent in 2019. Administrative data and iting the abuse of monopoly power, and introduc- population figures from the latest census indicate ing penalties for most anti-competitive practices. that Madagascar’s main social assistance programs Specifically, adopting a freedom of information law, cover approximately 5.3 percent of households and ratifying the OECD/Global Forum rules on fiscal a similar share of the poor. Furthermore, per capita transparency and information exchange, launching social assistance spending in Madagascar is among an open-data platform at the Ministry of Finance, the lowest in the East and Southern African Region, and digitizing the Official Journal would build con- comparable to countries such as Burundi, South fidence among the public and the donor commu- Sudan, and Zimbabwe (UNICEF, 2023) The Fonds nity about the government’s ability to manage d'Intervention pour le Développement (FID), which resources efficiently and effectively. Additionally, is responsible for social assistance expenditure, the Competition Council needs to increase moni- is almost entirely funded by public development toring and sanctioning of anti-competitive behav- aid, with a limited contribution of 16 percent by ior, estimate the quantitative impacts of a lack of the state.67 This dependence on external financing competition, and communicate results to the public causes significant variations in spending from one and policymakers. Enhancing the deterrence effect year to the next, and a low level of prioritization of of the law on anti-competitive behavior requires social assistance. building a credible threat of enforcement action with sufficient penalties, developing a framework Madagascar has two cash transfer programs. The for settlements, and implementing safeguards for first – Transfert Monétaire pour le Développement its independence. Finally, regulatory agencies need Humain (TMDH) – is a conditional cash transfer pro- to improve appointment procedures to enhance gram introduced in 2016 to support primary school independence, seek international expertise in some attendance. TMDH transfers are equal to about cases, and collaborate with the Competition Council 6 percent of the income that corresponds to the on sector-specific issues. national poverty line, and to about 8 percent of the extreme poverty line. The second program – Argent 3. Increasing resilience to shocks Contre Travail Productif (ACTP) – provides income and strengthening basic safety support through paid work, over limited periods, to nets poor workers in select districts. ACTP transfers are equal to about 8.4 percent of the income that cor- Malagasy households are vulnerable to both sys- responds to the national poverty line, and to 12 per- temic and idiosyncratic shocks, but the former pose cent of the extreme poverty line. the greatest challenge. The repeated occurrence of systemic shocks—linked to both the country’s Social safety net coverage remains low. Figure 98 history and the effects of climate change—ham- shows the percentage of individuals in Madagas- pers the coping ability of households, which thus car who are covered by social assistance programs, seldom recover from such shocks. The pandem- broken down by their poverty status (extreme poor, ic-induced recession has exposed the vulnerability moderately poor, non-poor), and location (urban or of some urban dwellers whose livelihoods depend rural). Overall, 5.3 percent of the population is cov- Despite the existence of coordination structures provided for by law, none have been established, leaving social protection technically 67 122 coordinated by the Social Protection Working Group (SPWG). The SPWG is responsible for establishing a general policy framework for social protection for households in vulnerable situations, but its implementation remains incomplete. TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment Figure 98: Coverage of safety nets is still very low Well-targeted social assistance can reduce poverty Cash transfer and public works coverage rates by up to 7.2 percentage points, depending on (% population), 2020-2021 its coverage of the poor. There may be a need to 10.0 enhance the poverty reducing impact of the exist- ing social assistance by changing its scope, improv- ing targeting, and behaviorally enhancing its inter- Urban; 9.3 ventions.69 A simulation of the impact of expanding 5.4 the coverage of TMDH by targeting 30 percent of 5.2 4.4 National; the poorest households via a proxy means-testing Rural; 4.3 5.3 (PMT) formula and offering each household MGA 4.1 4.4 50,000 per month between 2023 and 2030 shows 4.0 2.6 that the cost of the program would be equal to 2.16 1.7 percent of GDP in 2025, decreasing to 1.91 percent by 2030. The revised program would cover 1.8 mil- Urban Rural National lion households on average, with the value of the Total Extreme poor transfer corresponding to 37 percent of GDP per Moderate poor Non-poor capita. Source: Author’s calculations based on 2022 EPM data. Social assistance programs also need to change ered by social assistance programs. Looking at the their scope, with a focus on graduating beneficiar- poverty level, coverage is highest for the extreme ies out of poverty. Graduation programs aim to lift poor at 5.4 percent, followed by 4.4 percent of mod- people out of extreme poverty via financial support, erate poor and 2.6 percent of the non-poor pop- skills training, and other forms of assistance over a ulation. Finally, social assistance is skewed toward fixed period. Unlike traditional social assistance pro- urban households as coverage is higher in urban grams, which typically provide ongoing support to areas (9.3 percent) than in rural areas (4.4 percent).68 vulnerable populations, graduation programs are designed to provide temporary support to individ- As a result, it has limited poverty impact, though it uals and households living in extreme poverty, with makes a difference for beneficiary households. The the goal of helping them become self-sufficient and simulation results from the 2021–22 EPM data sug- self-reliant in the long term. gest that social assistance has a very small impact on poverty and inequality reduction (0.2 percentage Cash transfer programs can improve social-cogni- points lower and the Gini index 0.6 points lower) tive development in children by conditioning them (Table 38). This can be attributed to the extremely with behavior enhancing interventions. Datta et al. low coverage of social assistance programs. To (2021) examined the effects of coupling interven- improve the impact of social assistance on poverty tions based on behavioral science with cash trans- reduction in Madagascar, the government needs fers on early childhood socio-cognitive develop- to prioritize scaling up the coverage of these pro- ment and household-level outcomes for children grams. This can be achieved through increased in Madagascar. The study employs a multi-arm budget allocation, better targeting of beneficiaries, cluster-randomized trial, where three behavio- and more efficient implementation of social assis- ral interventions are layered onto a child-focused tance programs. cash transfer program. These interventions include a Mother Leaders group and associated activities, Table 38: The current social assistance system has which are augmented with a self-affirmation or marginal impact on poverty a plan-making nudge. The findings suggest that Poverty incorporating behavioral interventions into social Gini assistance programs can enhance their effective- headcount Post-transfer Indicators 75.2 36.8 ness in improving household-level outcomes and Indicators without listed child development. In the case of Madagascar, poli- transfer cymakers and practitioners could consider incorpo- All social assistance 75.4 37.4 rating behavioral interventions, such as the Mother Cash transfers, allowances, 75.4 37.1 Leaders group and associated activities, to com- last resort programs plement existing cash transfer programs as such Food and in-kind transfers 75.3 37.2 interventions may help promote positive parenting Household, food transfers 75.2 37.1 behaviors and improve children's socio-cognitive Source: Author’s simulations based on 2021/2 EPM data using development, including nutrition. ADEPT. The period covered in the survey coincides with an expansion in urban coverage in response to COVID-19. 68 123 A World Bank's credit is of US$250 million is expected to help Madagascar expand its social safety net coverage and the establishment of a 69 registry of beneficiaries, benefiting at least 3 million people, or 13 percent of extremely poor households, in the 23 regions. TOC Chapter 6 Madagascar Charting a course for change Poverty and Equity Assessment Graduation programs conditioned with behavio- ers and stakeholders with real-time poverty data, rally enhanced interventions have the potential facilitating timely and targeted policy interventions. to lift people out of extreme poverty and improve their long-term economic prospects. To enhance To strengthen the poverty monitoring system, sig- the cash transfer-based social assistance pro- nificant emphasis should be placed on regular data grams in Madagascar, the government can intro- collection and dissemination. Conducting monthly duce behaviorally enhanced interventions such as data collection enables the tracking of changes in Mother Leaders groups and associated activities, living conditions, resilience, and coping strategies, self-affirmation, or plan-making nudges to promote thereby identifying vulnerabilities and evaluat- desirable parenting behaviors, increase interactions ing intervention effectiveness. Developing a web- with children, and improve food security. To intro- based platform or mobile application would provide duce poverty graduation elements, the govern- stakeholders with real-time access to poverty and ment should carefully target the most vulnerable resilience indicators. Continuous training programs populations, collaborate with different government for enumerators and data analysts, along with agencies, local communities, and other stakehold- public awareness campaigns, will enhance capac- ers. In addition, monitor and evaluate all programs ity building and community engagement, ensuring to ensure effectiveness and identify any challenges. the sustainability of the system and maximizing its Programs should be scaled up with careful plan- impact on policy formulation and evaluation. By ning, adequate resources, and strong political com- implementing these measures, Madagascar can mitment. By following these steps, the government gain valuable insights into poverty dynamics and of Madagascar can effectively enhance its social effectively design targeted interventions to improve assistance programs and lift people out of extreme the well-being of its most vulnerable populations. poverty while improving their long-term economic prospects. 4. Establishing a Real-Time Poverty Monitoring System Adapted to Monitor and Measure Impact of Climatic Disasters Madagascar, particularly its South and East regions, is confronted with the recurring challenges of nat- ural disasters and economic shocks, which dis- proportionately affect the poor and vulnerable populations. To address this issue effectively, it is imperative to develop a comprehensive poverty monitoring system tailored specifically to these areas. The system should encompass a Rapid and Frequent Monitoring System (RFMS)70 in collabora- tion with key stakeholders from government agen- cies, international organizations, local communities, and academic institutions. By hiring and training local enumerators, costs associated with data col- lection can be minimized, while ensuring the collec- tion of accurate and culturally sensitive information. The integration of the Survey of Well-Being through Instant and Frequent Tracking (SWIFT) into the RFMS questionnaire is essential for estimating household expenditure/income, poverty levels, and related statistics. SWIFT utilizes machine learning and multiple imputation techniques, enabling rapid and accurate poverty assessments. Furthermore, by incorporating the SWIFT Plus approach, the system can effectively capture sudden changes in poverty resulting from economic shocks, such as the impact of drought. This integration empowers policymak- Catholic Relief Services is implementing a RFMS in Beloha and Tsihombe (in the Androy; both fully representative district-wide) and Ampanihy in 70 124 Antsimo Adrefana (not the entire district, but representative across a portion of it) that can be scaled up to other key districts while integrating a SWIFT module to measure poverty. TOC Madagascar Executive summary Poverty and Equity Assessment References ACAPS. 2022. “MADAGASCAR: Food insecu- Comanor, W. S., and Smiley, R. H. 1975. "Monopoly rity crisis in the Grand Sud regions.” Thematic and the Distribution of Wealth." The Quarterly Report, March 2022. Journal of Economics, vol. 89, no. 2, pp. 177–194. https://doi.org/10.2307/1884423. Afrobarometer and Transparency International. 2015. "People and Corruption: Africa Survey Cornell University, INSEAD, and WIPO. 2020. 2015 – Global Corruption Barometer." Afroba- "The Global Innovation Index 2020: Who Will rometer, www.afrobarometer.org. Finance Innovation?" https://www.wipo.int/ edocs/pubdocs/en/wipo_pub_gii_2020.pdf. Allen, J. E. 2022. "Double-Booked: Effects of Overlap between School and Farming Corral, P., I. Molina, A. Cojocaru and S. Segovia. Calendars on Education and Child Labor." 2022. "Guidelines to Small Area Estimation University of Michigan, https://drive.google. for Poverty Mapping." World Bank, http://hdl. com/file/d/1fO59p9vGSpQehKoEf-Vigup- handle.net/10986/37728. 9JUlcE1h6/view. Datta, S., J. Martin, C. MacLeod, L. Rawlings and A. AMD International. 2021. "Enquête sur le Suivi Vermehren. 2021. "Do Behavioral Interventions des Dépenses Publiques dans le Secteur de Enhance the Effects of Cash on Early Childhood l'Éducation au Niveau Primaire à Madagascar Development and Its Determinants? Evidence (Enquête PETS)." from a Cluster-Randomized Trial in Mada- gascar." World Bank Policy Research Working Barrett, C.B. 1996. "Urban Bias in Price Risk: the Paper 9747. Geography of Food Price Distribution in Low-income Countries." Journal of Develop- Deb, S., J. Eeckhout, A. Patel and L. Warren. 2022. ment Studies, vol. 23, no. 6, pp. 193-215. "What Drives Wage Stagnation: Monopsony or Monopoly?" BSE Working Paper 1361. Barrett, C.B. 1997. "Food Marketing Liberalization and Trader Entry: Evidence from Madagascar." Dorosh, P., B. Minten, JC. Randrianarison and D. World Development, vol. 25, no. 5, pp. 763-777. Stifel. 2022. "Recent Development in Madagas- car’s Rice Sector and Policy Options." Unpub- Bau, N., M. Rotemberg, M. Shah, and B. Steinberg. lished background paper, World Bank. 2020. "Human Capital Investment in the Presence of Child Labor." NBER Working Paper Edo, M., G. Mancini and G. Vecchi. 2022. "Inequality No. 27241, May 2020, Revised November 2023. and Poverty in Madagascar: Estimates based on 2020, and 2021-22 EPM." Unpublished Cadot, O., and Nasir, J. 2001. "Incentives and background paper, World Bank. Obstacles to Growth: Lessons from Manu- facturing Case Studies in Madagascar (RPED Feldman, M., F. Guy, S. Iammarino and C. Ioramash- Paper No. 117)." https://www.researchgate.net/ vili. 2021. "Gathering round Big Tech: how the publication/265238860. market for acquisitions reinforces regional inequalities in the US." Kenan Institute of Chaudhuri, S., Jalan, J., and Suryahadi, A. 2002. Private Enterprise Research Paper No. 21-01 "Assessing Household Vulnerability to Poverty https://ssrn.com/abstract=3845674. from Cross-sectional Data: A Methodology and Estimates from Indonesia." Discussion paper Global Financing Facility. 2020. "Country Briefs: 0102-42, Columbia University. Preserve Essential Health Services during the Covid-19 Pandemic." https://www.globalfinanc- 125 TOC Madagascar Executive summary Poverty and Equity Assessment ingfacility.org/sites/gff_new/files/documents/ Jarotschkin, A. 2023. "Non-Monetary Poverty in Madagascar-Covid-Brief_GFF-FR.pdf. Madagascar-Challenges and Policy Priorities for Poverty Reduction." Unpublished back- Günther, I. and K. Harttgen. 2009. "Estimating ground paper, World Bank. Households Vulnerability to Idiosyncratic and Covariate Shocks: A Novel Method Applied in Jarotschkin, A., Vincent, R. C., Aron, D. V., and Madagascar." World Development, vol. 37, no. 7, Yoshida, N. 2023. "Restoring the Compara- pp. 1222-1234. bility of Poverty Estimates over Time for Mada- gascar." Unpublished background paper, World Iimi, A. 2022. "Estimating the Demand for Informal Bank. Public Transport Evidence from Antananarivo, Madagascar." Policy Research Working Paper Jarotschkin, A., Yoshida, N., Vecchi, G., and Mancini, No. 10006, http://www.worldbank.org/prwp. G. 2023. "Inequality and Poverty in Mada- gascar: Estimates based on 2021-22 EPM." Iimi, A., A. H. Rakotoarisoa, G. F. Raserijaona, Z. S. El Unpublished background paper, World Bank. Nakat, S. J. Rabary, and D. Wolde Woldearegay. 2022. "Toward Efficient, Sustainable and Safe Karoly, L. and G. Burtless. 1995. "Demographic Urban Transport in Madagascar: Antananarivo Change, Rising Earnings Inequality, and the and Other Major Cities - Synthesis Report." Distribution of Personal Well-Being, 1959- 1989." Demography, vol. 32, no. 3, pp. 379–405. ILO. 2020. "Madagascar Decent Work Country Profile." Keller, M. 2022. "The Socioeconomic Impact of Industrial Mining in Madagascar." Unpublished Independent Evaluation Group. 2022. "The World background paper, World Bank. Bank Group in Madagascar, Fiscal Years 2007-21 Country Program Evaluation." www. Keller, M. and F. M. Mulangu. 2023. "Four in a Row: worldbank.org. The Economic Costs of Extreme Weather Events caused by Climate Change in Mada- INSTAT. 2021a. "Caractéristiques Économiques de gascar." Unpublished background paper, World la Population." Bank. INSTAT. 2021b. "Compétences Linguistiques et Khan, L. and S. Vaheesan. 2017. "Market Power Scolarisation à Madagascar." and Inequality: The Antitrust Counterrevo- lution and Its Discontents." 11 Harvard Law INSTAT. 2021c. "État et Structure de la Population and Policy Review 235. https://ssrn.com/ à Madagascar." abstract=2769132. INSTAT. 2021d. "Mesure et Cartographie de la Klasen, S. and F. Povel. 2013. "Defining and Measuring Pauvreté Non Monétaire des Ménages et de la Vulnerability: State of the Art and New Population à Madagascar." Proposals." In: Klasen, S., Waibel, H. (eds) Vulner- ability to Poverty. Palgrave Macmillan, London. IOE. 2023. "Analysis of Business Environment https://doi.org/10.1057/9780230306622_2. in Least Developed Countries - Mada- gascar." https://data.worldbank.org/indicator/ Klasen, S. and H. Waibel. 2015. "Vulnerability to SL.UEM.1524.ZS?locations=MG. Poverty in South-East Asia: Drivers, Measure- ment, Responses, and Policy Issues." World Dev. IPC. 2021. “MADAGASCAR [GRAND SOUTH]: Jul;71:1-3. doi: 10.1016/j.worlddev.2014.01.007. Food Security and Nutrition Snapshot.” https:// www.ipcinfo.org/fileadmin/user_upload/ La Ferrara E. and V. Novak. 2022. "Valuing Hope: A ipcinfo/docs/IPC_Madagascar_Food_Security_ model of Aspirations." Working Paper. Nutrition_Snapshot_English.pdf November 2021. Lanjouw, J. O. and P. Lanjouw. 2001. "How to Compare Apples and Oranges: Poverty Meas- Jacoby, H. and B. Minten. 2006. "Land Titles, urement Based on Different Definitions of Investment, and Agricultural Productivity in Consumption." Review of Income and Wealth, Madagascar: A Poverty and Social Impact vol. 47, pp. 25-42. Analysis." World Bank, Washington, DC. 126 TOC Madagascar Executive summary Poverty and Equity Assessment Lybbert, T. J. and B. Wydick. 2016. "Hope as Osborne,T., N. Belghith, C. Bi, A. Thiebaud, L. Aspirations, Agency, and Pathways: Poverty Mcbride, M. Jodlowski. 2016. “Shifting fortunes Dynamics and Microfinance in Oaxaca, Mexico." and enduring poverty in Madagascar: recent In C. B. Barrett, M. R. Carter, and J. Chavas (Eds.), findings” Washington, D.C.: World Bank. The Economics of Poverty Traps, pp. 153–177. http://documents.worldbank.org/curated/ University of Chicago Press. en/413071489776943644/Shifting-for- tunes-and-enduring-poverty-in-Madagas- Marchetta, F., D.E. Sahn and L. Tiberti. 2019. "The car-recent-findings. role of weather on schooling and work of young adults in Madagascar." American Journal of Agri- Pietschmann, I., Kapsos, S., Bourmpoula, E., Sajaia, cultural Economics, vol. 101, no. 4, pp. 1203–27. Z., and Lokshin, M. 2016. "Key Labor Market Indicators: Analysis with Household Data." Maruyama, E., Scollard, P., Elias, M., Mulangu, F., World Bank and ILO. and Seck., A. 2018. "Frontier Analysis and Agri- cultural Typologies." SEF-Discussion Papers on Psacharopoulos, G., and Patrinos, H. A. 2018. Development Policy No. 251. "Returns to Investment in Education A Decennial Review of the Global Literature." MINAE. 2006. "Stratégie Nationale du Développe- Policy Research Working Paper No. 8402. ment de l’Utilisation des Engrais." Ministère de http://econ.worldbank.org. l’Agriculture, Elevage, et Pêche. Quota, M. B. N., Cobo Romani, J. C., Patil, A. S., Minten, B., P. Dorosh, M.H. Dabat, O. Jenn-Treyer, and Wilichowski, T. M. 2022. "Effective Teacher J. Magnay and J. Razafintsalama. 2006. "Rice Professional Development Using Technology: markets in Madagascar in disarray: Policy Technology-Based Strategies from across options for increased efficiency and price stabi- the Globe to Enhance Teaching Practices A lization." Africa Region Working Paper Series Guidance Note." www.worldbank.org. No. 101, World Bank, Washington, D.C. Ravallion, M. 1998. "Poverty Lines in Theory and Mishra, K., A. Gallenstein, M. Miranda, A. Sam, P. Practice." LSMS Working Paper Number 133. Toledo, and F.M. Mulangu. 2021. "Insured Loans and Credit Access: Evidence from A Rand- Razafindrakoto, M., Roubaud, F., and Wachsberger, omized Field Experiment in Northern Ghana." J.-M. 2020."The Malagasy Mystery through American Journal of Agricultural Economics, the Lens of Economic Growth and Develop- vol. 103, no. 3, pp. 923-943. ment Theories." Puzzle and Paradox: A Political Economy of Madagascar, Cambridge University Monnier. 2021. "Madagascar Mobile Money Boom." Press. http://ebookcentral.proquest.com/lib/ https://www.ifc.org/en/stories/2023/mada- jointlib-ebooks/detail.action?docID=6121077. gascars-mobile-money-boom Razafindravonona, J. 2023."Situation de l’Aide Moser, C., C. Barrett and B. Minten. 2009. "Spatial Publique au Développement à Madagascar." integration at multiple scales: Rice markets in Unpublished background paper, World Bank. Madagascar." Agricultural Economics, vol. 40, no. 3, pp. 281-294. Serneels, P., and Dercon, S. 2021. "Aspirations, Poverty, and Education. Evidence from India." Mulangu, F.M. 2023. "Exploring Poverty Dynamics Journal of Development Studies, vol. 57, no. 1, in Madagascar: Insights from Focus Group pp. 163–183. https://doi.org/10.1080/002203 Analysis on Key Constraints and Drivers." 88.2020.1806242. Unpublished background paper, World Bank. Skoufias, E., K. Vinha and R. Sato. 2019. "All Hands on Nicita, A. 2006. "Export Led Growth, Pro-Poor Or Deck: Reducing Stunting through Multisectoral Not? Evidence From Madagascar’s Textile And Efforts in Sub-Saharan Africa." Africa Develop- Apparel Industry." Policy Research Working ment Forum Series, AFD and the World Bank. Paper No. 3841. https://doi.org/10.1596/978-1-4648-1396-2. OPHI, O. P. and H. D. I. 2022. "Global MPI Country Thomson H, Thomas S, Sellstrom E, Petticrew Briefing 2022: Madagascar (Sub-Saharan M. 2009. "The Health Impacts of Housing Africa)." www.ophi.org.uk. Improvement: A Systematic Review of Inter- 127 TOC Madagascar Executive summary Poverty and Equity Assessment vention Studies from 1887 to 2007." American World Bank. 2022. "How COVID-19 is affecting Journal of Public Health, doi: 10.2105/ firms in Madagascar." Business Pulse Survey. AJPH.2008.143909, World Bank. 2022. “Macro Poverty Outlook: Coun- UNDP and OPHI. 2022. "Global Multidimensional try-by-country Analysis and Projections for Poverty Index 2022 - Unpacking deprivation the Developing World.” Washington DC: World bundles to reduce multidimensional poverty." Bank. Spring 2022. http://hdr.undp.org and https://ophi.org.uk/ multidimensional-poverty-index/. World Bank. 2023. "Madagascar Public Expendi- ture and Institutional Review: Boosting Infra- UNICEF. 2018. "Analyse du Budget des Secteurs structure and Social Service Delivery." Sociaux, 2014-2018." World Bank. 2024. (forthcoming). “Madagascar : UNICEF. 2023. "Budget Brief - Protection Sociale." Indicateurs de Prestation de Services de Santé 2021.” Banque mondiale. United Nations. 2018. "World Urbanization Prospects - The 2018 Revision." Wu, H and Yoshida, N. 2022. "Poverty Projec- tions to Estimate the Impact of the Ukraine World Bank Group. 2021. "Creating Markets in War Using Data from Macro Poverty Outlook Madagascar - Country Private Sector Diag- (MPO)." Unpublished, World Bank. nostic.". https://worldbankgroup.sharepoint. com/sites/DNRINT/Shared%20Documents/ Forms/AllItems.aspx?id=%2Fsites%2FD- NRINT%2FShared%20Documents%2FI- DU%2D25e9936b%2Df718%2D4ac8%2D b86c%2D351ed9d5317e%2Epdf&parent=%2F- sites%2FDNRINT%2FShared%20Documents. World Bank Group. 2022. "Madagascar System- atic Country Diagnostic Update." https:// d o c u m e n t s 1 .w o r l d b a n k . o r g / c u r a t e d / en/551231652117328109/pdf/Mada- g a s c a r-Sy s t e m a t i c- C o u n t r y- D i a g n o s - tic-The-Urgency-of-Reforms-Structur- al-Transformation-and-Better-Govern- ance-at-the-Heart-of-the-Strategy-to-R- educe-Poverty.pdf. World Bank. 2017."Madagascar: Indicateurs de Prestation de Services - Education." Indica- teurs de Prestation de Services En Education et Sante (SDI), World Bank. 2021a. "Madagascar Urbanization Review." World Bank. 2021b. "Management, Transparency, and Accountability in the Education Sector in Madagascar- Key findings and recommenda- tions from an empirical study of public officials, school directors and teachers." World Bank. 2021c. "Republic of Madagascar Education Service Delivery Indicators-Round I and II." 128 Contact: Diana Styvanley Madagascar 1 Rue Andriamifidy BP 4140 Antananarivo 101, Madagascar Telephone : +261-320-50-01-27 E-mail : dstyvanley@worldbank.org