1 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA Franklin Maduko Adanna Chukwuma Gevorg Minasyan Armineh Manookian Miguel Angel Saldarriaga Noel Ajay Tandon © 2021 The World Bank Group, 1818 H Street NW, Washington, DC 20433. This report was prepared by World Bank staff with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. This report was originally published in English by the World Bank (More Money for Health: Resource Mobilization for Universal Health Coverage in Armenia) in 2021. Where there are discrepancies, the English version will prevail. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning any territory’s legal status or the endorsement or acceptance of such boundaries. The material in this report is subject to copyright. The World Bank encourages the dissemination of its knowledge. Hence, this work may be reproduced, in whole or in part, for noncommercial purposes if full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. Design by Veronica Elena Gadea, GCSDE, The World Bank Group. MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA Franklin Maduko Adanna Chukwuma Gevorg Minasyan Armineh Manookian Miguel Angel Saldarriaga Noel Ajay Tandon 4 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA ABOUT THIS REPORT This report, More Money for Health: Resource Mobilization for Universal Health Coverage in Armenia, is part of the World Bank’s technical support towards Universal Health Coverage (UHC) in Armenia, which includes advisory services and analytics aimed at supporting the Government’s efforts to expand access to high-quality health care. The report explores potential sources for additional public spending on health in Armenia given the health financing transition. This technical assistance was supported by Gavi, The Vaccine Alliance; and The Global Fund to Fight AIDS, Tuberculosis, and Malaria. Modeling the Actuarial costing Projecting impact of tax of a unified revenues from options on growth, benefits package alternative tax poverty, financial that meets and non-tax protection, health population health sources and employment care needs Informing policies to Support for increase public Modeling strategic plan for financing for allocations of primary health health care public financing care financing, in the benefits organization, and package to regulation maximize health Support for Facilitating Reforms to Assessment of strategic plan the alignment Technical support public financial towards Universal align public for continuity of service financing for management in of care across delivery with Health Coverage the health in Armenia health with providers better health value sector Support for Assessment of regulating, strategic monitoring and purchasing in paying providers Knowledge exchanges on the health for better sector quality investing in Universal Health Coverage Convening Harvard-World policy and Study tours to Bank Global technical selected Flagship discussions on countries Course on reform options Health Reform 5 TABLE OF CONTENTS About this Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 About the Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Chapter 1. Why Mobilize Domestic Resources for Health? . . . . . . . . . . . . . . . . . . . . . 15 1.1 Economic Context. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 1.2 Health Systems Context. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.3 Purpose of this Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Chapter 2. Projecting Fiscal Space for Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Chapter 3. Macroeconomic Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Chapter 4. Reprioritization of Health in the Government Budget . . . . . . . . . . . . . . 28 Chapter 5. Taxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 5.1 Excise Tax. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 5.2 Payroll Tax. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 5.3 VAT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 5.4 Turnover tax. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Chapter 6. Sector-Specific External Funding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 Chapter 7. Efficiency Gains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 Chapter 8. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 6 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 Appendix 1A: Elasticity of Public Spending on Health to Nominal GDP. . . . . . . . . . . . . . . . . . . . . . . . . . 58 Appendix 1B: Elasticity of Public Spending on Health to Real GDP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Appendix 1C: Fiscal Space from Conducive Macroeconomic Environment. . . . . . . . . . . . . . . . . . . . . . . 60 Appendix 1D: Observed vs. Projected Government Spending on Health (2012-2017). . . . . . . . . . . . . 61 Appendix 2A: Elasticity Estimation by Deaton’s LA/AIDS Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 Appendix 2B: Product-Level Summary of Revenue from Excise Tax. . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 Appendix 3A: Regression of Total Health Expenditure on Foreign Transfers. . . . . . . . . . . . . . . . . . . . 66 Appendix 3B: Regression of Total Government Health Expenditure on Foreign Transfer . . . . . . . . . 66 Appendix 3C: Linear Fit of Total Government Health Expenditure (net of foreign transfers, millions AMD). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 Appendix 3D: Regression of Out-of-Pocket Health Expenditure on Foreign Transfer . . . . . . . . . . . . 67 Appendix 3E: Linear Fit of Total Out-of-Pocket Health Expenditure (net of foreign transfers, millions AMD). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 Appendix 4A: Armenia and Comparator Countries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 Appendix 4B: List and Description of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 Appendix 4C: Health Outcomes and Inputs for Armenia and Comparator Countries. . . . . . . . . . . . . 70 Appendix 4D: Analytical Description of DEA Programming and Savings Estimation. . . . . . . . . . . . . . 72 Appendix 4E: Potential Gains in Outputs with Efficiency Improvements. . . . . . . . . . . . . . . . . . . . . . . . . 74 Appendix 4F: Efficiency Scores and Potential Gains in Health System Performance. . . . . . . . . . . . . 75 Endnotes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 7 ACKNOWLEDGMENTS This report was supervised by Sylvie Bossoutrot (Country Manager, Armenia) and Tania Dmytraczenko (Practice Manager, Health, Nutrition, and Population Global Practice, Europe and Central Asia Region). The analysis benefited from the close engagement of the Ministry of Health, including the National Institute of Health, and the Statistical Committee of the Republic of Armenia. The team appreciates the Ministry of Finance, Ministry of Health, and the Ministry of Economy for participating in the discussion of preliminary findings of the analysis in February 2021. The authors are grateful to World Bank colleagues Arvind Nair (Senior Economist, Macroeconomics, Trade, and Investment Global Practice), Evgenij Nadov (Senior Economist and Program Leader, Macroeconomics, Trade, and Investment Global Practice), Israel Osorio-Rodarte (Economist, Macroeconomics, Trade, and Investment Global Practice), and Owen K. Smith (Senior Economist, Health, Nutrition, and Popul- ation Global Practice) for their insightful feedback on initial drafts of the report. Thanks also to colleagues in the Global Fund for reviewing and providing helpful suggestions on prior report drafts. The excellent editorial support from Richard A. B. Crabbe and Seemi Qaiser and the significant operational assistance from Marianna Koshkakaryan and Arpine Azaryan are appreciated. All errors and omissions are the authors’ responsibility. 8 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA ABOUT THE AUTHORS Franklin Maduko is a Lecturer in Economics at the University of Exeter Business School, where he teaches and conducts research in the fields of international trade, private sector development and public economics. He also works as a Consultant at the World Bank Group, where he supports the design, implementation and analysis of the Business Pulse Survey, and also provides technical assistance within the framework of the Universal Health Coverage agenda in Armenia. Franklin holds a Master of Science in Economics and Econometrics from the Catholic University of Louvain and a Doctor of Philosophy in Economics from Central European University. Adanna Chukwuma is a Senior Health Specialist in the Health, Nutrition, and Population Global Practice, where she leads the design, implementation, and evaluation of investment operations. She has over ten years of experience advising national reforms to improve access to high-quality health care, through service delivery organization, strategic purchasing, revenue mobilization, and demand generation, including in Sri Lanka, Sierra Leone, India, Moldova, Tajikistan, the South Caucasus Countries, and Romania. She has published on health care financing, access, and quality in peer-reviewed journals, including the Bulletin of the World Health Organization and Social Science and Medicine. Adanna obtained a medical degree from the University of Nigeria, a Master of Science in Global Health from the University of Oxford, and a Doctor of Science in Health Systems from Harvard University. Gevorg Minasyan is the Head of Special Studies Division at the Economic Research Department of the Central Bank of Armenia, where he coordinates the bank’s research projects on development, growth, and public policy. Previously, he served as an International Monetary Fund mission member for Israel, participating in macro-fiscal forecasting model development and capacity building. He has also worked at the World Bank Group as a Consultant, providing technical assistance within the framework of the Universal Health Coverage agenda. Gevorg completed the Data, Economics, and Development Policy Micro-Master’s degree of Massachusetts Institute of Technology and holds a Doctor of Philosophy from Yerevan State University. Armineh Manookian is the World Bank Country Economist for Armenia, in the Macroeconomics, Trade and Investment Global Practice, covering macroeconomic and fiscal issues, economic reporting, and macroeconomic projections. She is engaged in macroeconomic policy dialogue with the client. Armineh joined the World Bank in 2017 and prior to that, worked for more than 10 years in the International Monetary Fund’s Resident Representative office in Armenia as a Macroeconomist. Before moving to Armenia in 2005, Armineh worked with the Central Bank of Iran as a Senior Economist in the Research and Policy Department. She holds a Master of Public Administration in Economic Policy Management from Columbia University. About the Authors 9 Miguel Angel Saldarriaga Noel is the World Bank Country Economist for Nigeria in the Macroeconomics, Trade, and Investment Global Practice. Miguel has worked on operations and advisory services and analytics on inclusive growth and macro-fiscal issues in Guatemala, Honduras, Malawi, Panama, Tanzania, and Zimbabwe. Before joining the Bank, he worked at the Central Bank of Peru for eight years, where he was the Head of the Real Sector Department. Miguel holds a Master of Science in Economics from the London School of Economics and a Master of Science in Economics from Universidad del Pacifico. Ajay Tandon is a Lead Economist in the Health, Nutrition, and Population Global Practice. He works on several countries including Bhutan, India, Indonesia, Laos, and the Philippines. Between 2004 and 2007, he worked with the research department of the Asian Development Bank in Manila on issues related to human development. He also worked with the Evidence and Information for Policy department of the World Health Organization in Geneva from 1998 to 2003 and has held visiting research appointments at both Oxford University and Harvard University. Ajay has written several publications on the issues of statistical methodology, health systems efficiency, health financing, domestic resource mobilization, and universal health coverage. He received a Doctor of Philosophy in Economics from Virginia Tech University. 10 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA ACRONYMS ADB Asian Development Bank ARMSTAT Armenian National Statistical Office AMD Armenian Dram BBP Basic Benefits Package CBA Central Bank of Armenia CIT Corporate Income Tax DALY Disability-adjusted life year DEA Data envelopment analysis ECA Europe and Central Asia ES Executive Summary FEACN Foreign Economic Activity Commodity Nomenclature GDP Gross Domestic Product GFATM Global Fund to Fight AIDS, Tuberculosis and Malaria GGE General Government Expenditure ICT Information and communications technology LA/AIDS Linear approximated/almost ideal demand system MoH Ministry of Health MoF Ministry of Finance NCD Non-communicable disease OOP Out-of-pocket PIT Personal income tax PT Payroll tax SHA State Health Agency SHI Social Health Insurance SRC State Revenue Committee UHC Universal Health Coverage UMI Upper-Middle-Income USD United States Dollar VAT Value-Added Tax Acronyms 11 12 EXECUTIVE SUMMARY Armenia has made significant gains in population health, but faces challenges in ensuring health care access, due to financial barriers. As mortality caused by infectious diseases has fallen over the past two decades, the prevalence of noncommunicable diseases (NCDs) has increased. The NCD burden can be reduced via public health measures, such as tobacco control exposure, and access to high-quality health care. However, financial barriers to access are a significant challenge. Over 80 percent of current spending on health is paid out-of-pocket (OOP), far above the average in upper middle income (UMI) countries of 33 percent. High OOP is driven by the low coverage of essential care, including outpatient medicines and inpatient care, for most of the population. The Ministry of Health (MoH) is championing reforms to mobilize additional revenues to ensure access to essential care for the whole population. This report, developed following a request from the MoH, assesses potential sources to expand public financing for health in Armenia, to finance improvements in health coverage. This assessment identifies the potential for increased public spending in the sector without jeopardizing solvency or crowding out spending for important objectives in other sectors. Per a framework developed by Tandon and Cashin (2010), fiscal space for health is assessed under five pillars: macroeconomic conditions, reprioritization of the health sector in the budget, sector-specific revenues, external resources, and efficiency gains. The findings are expressed as additional annual spending relative to public spending on health in 2018 in the short (2021), medium (2023), and long (2031) term (Executive Summary (ES) Table 1) and provide a qualitative summary of the relative potential for revenue mobilization across pillars. The estimates were expressed in nominal terms and without future discounting. The 100 percent earmarking scenarios considered for all taxes were proposed by the MoH. Executive Summary 13 ES. TABLE 1: Overview of additional fiscal space for health across pillars ADDITIONAL FISCAL SPACE FOR HEALTH 2021 2023 2031 PILLAR (BILLIONS (% of GDP (BILLIONS (% of GDP (BILLIONS (% of GDP AMD) 2020) AMD) 2020) AMD) 2020) Baseline forecast 6.1 0.10 19.1 0.31 97.7 1.58 Macroeconomic conditions Downside forecast 3.1 0.05 14.3 0.23 78.6 1.27 Increase in the current ratio of Reprioritization public spending on health in 50.13 0.81 51.99 0.84 82.24 1.33 of health total government expenditure from 5.27% to 8.0% 100% earmarked for the Excise Tax 22.14 0.36 42.97 0.70 113.37 1.83 health sector Payroll Tax 100% earmarked for the 85.70 1.39 94.91 1.54 142.90 2.31 (6% increase) health sector 100% earmarked for the Value-Added Tax 49.21 0.80 53.02 0.86 71.45 1.16 health sector Increase in 100% earmarked for the 11.18 0.18 14.26 0.23 23.74 0.38 Turnover Tax health sector Reducing Thresh- 100% earmarked for the 41.12 0.67 45.60 0.74 68.92 1.11 old Turnover tax health sector External General health sector 12.8 0.21 3.2 0.05 N/A Resources Vertical programs 1.2 0.02 1.5 0.02 N/A Potential annual savings, in bil- lions AMD, considering adult Upper Lower mortality rates, if spending bound bound = N/A N/A efficiency were equivalent to = 0.6 0.01 the most efficient comparator Efficiency Gains Potential annual savings, in billions AMD, considering Upper Lower under-five mortality rates, bound bound = N/A N/A if spending efficiency were = 3.5 0.06 equivalent to the most effi- ciency comparator These results suggest that the additional revenue needed to support expanded coverage, which may be up to AMD 310 billion by 2031, cannot be fully financed under any single pillar, in the hypothetical scenarios considered. Nonetheless, the potential for revenue-raising by increasing the priority for the health sector in the national budget is significantly higher than additional fiscal space that may accrue to other pillars, outside payroll tax and excise tax by 2031. There is significant room to increase priority for public health spending. For instance, public spending on health per capita in Armenia is less than half the level seen in neighboring Georgia, despite similar income levels. This is in part because Georgia allocates above nine percent of the national budget to health, higher than Armenia (5.27 percent). A progressive increase in budgetary allocations to match the median in comparator countries of 8.0 percent could mobilize an additional AMD 50 billion in 2021, increasing the 2018 health budget by over 50 percent. 14 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA Decisions by the Ministry of Finance (MoF) to increase funding for the health sector will reflect multiple considerations, beyond revenue-raising potential from each pillar. These considerations may include the potential for the fiscal outlays to boost medium-term growth and yield future revenue, the compatibility of revenue-raising options within the Government’s tax reform strategy and debt management framework, and the degree of consensus in the Government and public over the objectives and design of the reforms to be supported through increased funding. The MoH built consensus on the reform agenda for increasing public spending on health by demonstrating the potential returns on these investments. In the short-term, improvements in efficiency of spending on the health sector can result in additional fiscal space for expanding coverage. Comparing Armenia with countries with the most efficient public health spending (as measured by child mortality) suggests that Armenia can save about AMD 3.5 billion each year. Efficiency improvements are largely within the realm of control of the MoH. These include scaling up centralized procurement of medical goods through the MoH; implementing a structured process for revising the Basic Benefits Package (BBP) that incorporates analysis of the cost-effectiveness of alternative technologies; investing in stronger primary care systems and improvements in coordination between primary care and hospitals; and strengthening the budget program manager function and increasing flexibility for re-allocation across activities within budget programs. Armenia faces crucial decisions. The country is confronted with underutilization of essential health care, that arises from gaps in financial protection, and contributes to the growing burden of NCDs. The evidence from global experience on expanding coverage for essential care is clear. Successful countries have mobilized pre-paid revenue via predominantly public sources, mobilized through compulsory contrib- utions, with subsidies for the poorest. This report, and other technical assistance from the World Bank, provides detailed analysis of the potential cost of expanding coverage, options for raising domestic revenue for UHC, implications of financing UHC reforms for macroeconomic outcomes, and strategies to ensure value for money in the health sector. The Government will ultimately decide whether to guarantee access to essential health care for every citizen and the financing options that are compatible with broader macroeconomic objectives. By shaping the health of the population and productivity of the workforce, these decisions will have significant implications for decades to come. 15 CHAPTER 1. WHY MOBILIZE DOMESTIC RESOURCES FOR HEALTH? 1.1 ECONOMIC CONTEXT Armenia is a landlocked country in the South Caucasus in the World Bank Europe and Central Asia (ECA) region. The country is bordered in the north by Georgia, in the south by Iran, in the east by Azerbaijan, and in the west by Turkey. Armenia is administratively subdivided into ten Marzes or regions and the capital city of Yerevan. In 2019, there was a total population of 2.9 million, 36.1 percent of whom resided in rural areas.1 Since 1990, the population has declined by 16 percent, and the share of the population that resides in urban areas has decreased from 69 percent to 64 percent.2 This decrease is primarily due to economic emigration and below-replacement fertility. Since 2018, Armenia has been classified by the World Bank as UMI.3 From 2017 to 2019, the Gross Domestic Product (GDP) had real annual growth rates ranging from 5.2 percent to 7.6 percent. Simultaneously, the proportion of the population living below the UMI poverty line fell from 41.5 percent to 37.0 percent.4 Between 2000 and 2008, the annual per capita economic growth rates averaged 12.6 percent, decelerating to 2.0 percent between 2009 and 2017, due to the 2008-2009 global financial crisis and the 2014-2015 Russian financial crisis (Figure 1).5 By 2019, the per capita income in Armenia had risen to USD 4,680.6 Between 2000 and 2017, total government expenditure varied between 20 percent and 29 percent of GDP and is below the UMI average.7 Over the same period, total government revenue, excluding grants, ranged from 16.5 to 22.5 percent of GDP, exceeding the UMI average of 16.3 percent in 2017.8 16 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA FIGURE 1: Trends in per capita GDP in Armenia, 2000-19 5000 4000 Per Capita GDP (in USD) 3000 2000 1000 0 2000 2004 2008 2012 2016 2020 Year Source: World Economic Outlook, 2019 The Armenian economy has been hard hit by the COVID-19 pandemic. By midyear in 2020, the economy had contracted by 5.7 percent year-on-year, due to falls in private consumption and investment, which were only partially offset by higher government spending and import compression. Construction and services were the most negatively affected sectors, while the financial, information and communications technology sectors were more resilient due to a relatively higher reliance on digital technology.9 Overall, the economy contracted by 8.0 percent in 2020, and the proportion of the population living below the UMI poverty line is projected to rise by up to 12.8 percentage points. The adverse economic and social impacts of the pandemic put recent gains in welfare at risk. The Government launched measures to mitigate these impacts at an estimated cost of 2.3 percent of GDP, increasing current spending by 19 percent in the first seven months of 2020. Capital spending also increased by 62 percent. However, revenue has fallen by six percent year-on-year.10 1.2 HEALTH SYSTEMS CONTEXT Armenia has made significant gains in population health but faces challenges in ensuring health care access. The average life expectancy at birth has risen from 68 years in 1990 to 75 years in 2018 (Table 1). Over the same period, the infant mortality rate fell from 41.7 deaths to 11 deaths per 1,000 live births.11 Armenia performs as well or better than the average UMI and ECA country on most population health outcomes. Between 2005 and 2018, the tuberculosis incidence rate fell from 92 to 31 cases Chapter 1. Why Mobilize Domestic Resources for Health? 17 per 100,000 population.12 In 2019, Armenia had a measles incidence rate of 2.38 per 1,000,000 population, below the average of 125.14 in the World Health Organization (WHO) Europe region.13 It is projected that life expectancy at birth will rise to 79.8 years among males and 84.4 years among females at the current trajectory.14 TABLE 1: Comparing health outcomes in Armenia and comparator groups UNDER-FIVE MATERNAL LIFE ADULT CHILDHOOD STUNTING MORTALITY MORTALITY COUNTRY EXPECTANCY SURVIVAL (PERCENT OF (PER 1,000 (PER 100,000 (YEARS) (PERCENT) CHILDREN UNDER FIVE) LIVE BIRTHS) LIVE BIRTHS) Armenia 75 12 81 26 9 ECA 74 13 79 18 10 UMI 74 19 78 65 14 Source: World Development Indicators; Global Health Expenditure Database As mortality due to infectious diseases has fallen, the prevalence of NCDs has increased. Currently, NCDs are responsible for 93 percent of deaths. In 2019, the leading causes of death were ischemic heart disease, stroke, lung cancer, diabetes, and chronic obstructive pulmonary diseases.15 Armenia also has a higher age-standardized Disability-Adjusted Life Year (DALY) rate per 100,000 for ischemic heart disease, stroke, and diabetes than the average country with comparable income per capita, education and fertility rates.16 The high burden of NCDs in Armenia is driven by aging, exposure to behavioral risk, and gaps in access to care.17 About 24.5 percent of Armenian adults are smokers and 11.5 percent are heavy, episodic drinkers, while 20.9 percent are obese.18 By 2030, one in ten residents will be 70 years old or older. The NCD burden can be reduced via public health measures, such as tobacco control to reduce risk exposure, and by improving access to high-quality health care. In Armenia, an estimated 1,396 deaths occur due to barriers to health care access annually.19 Only one in three individuals reports visiting a primary health care facility when sick.20 Financial barriers to access are a significant challenge. In 2018, 84.3 percent of current spending on health was paid OOP (Figure 2), far above the average in UMI countries (32.9 percent) and even higher than countries in conflict, such as Yemen and Afghanistan.21 As countries transition to UMI status, and donor financing reduces, it is expected that public financing for health will become an increasing share of total spending on health. However, Armenia has regressed in its health financing transition, as the OOP share of spending on health has increased at a faster rate than the increase in public spending on health as a share of GDP (Figure 2).22 18 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA FIGURE 2: Public or government spending on health as a share of GDP and OOP share of spending on health 2.1 90 1.9 84 OOP share of health spending (%) Public spending on health 1.7 78 Share of GDP (%) 1.5 72 1.3 66 1.1 60 .9 54 .7 48 2000 2003 2006 2009 2012 2015 2018 Public spending on health (left) OOP share (right) Source: Global Health Expenditure Database The high share of OOP in total health expenditure reflects the relatively low coverage through the public sector for health services. Armenia introduced a BBP that explicitly defined priority services and groups for coverage by the state (Box 1).23, 24 While primary and emergency services are universally covered, outpatient drugs, inpatient care, and other specialized services are covered for socially vulnerable groups and the public sector. Hence, for about 70 percent of the population, OOP payments are the main source of financing for health care. In 2013, 16 percent of Armenian households spent over 10 percent of total household expenditure on health care – catastrophic health expenditure – while an estimated 4.1 percent of households were pushed below the poverty line due to “impoverishing” health expenditure.25 Groups with lower coverage have been shown to have 36 percent lower rates of outpatient care use.26 Chapter 1. Why Mobilize Domestic Resources for Health? 19 BOX 1: Overview of health system context Governance: The MoH is responsible for oversight and regulation of the health sector in Armenia. In 1998, the State Health Agency (SHA) of the MoH was established as the single public payer responsible for purchasing services in the BBP on behalf of the Government. However, multiple private insurers are responsible for processing claims for government workers eligible for the Social Package under the BBP. With decentralization of administrative functions, regional authorities can make decisions on facility spending for services outside the BBP. Financing: The public share of total spending on health is low, declining from 17.4 percent in 2016 to 13.2 percent in 2018. This is below the average in UMI countries of 57 percent. External funding as a share of total spending on health has also declined from 6.5 percent in 2014 to 5.4 percent in 2018. OOP is the predominant source of financing for health. Public spending flows from the state budget to the MoH, and predominantly to the SHA to purchase health services. There are separate pools for the Ministry of Defence and the Police. There are also multiple voluntary health insurance schemes, which account for less than five percent of total spending on health. Service delivery: According to the "2019 Social Snapshot and Poverty report", Armenia had 124 hospitals and 501 primary care facilities. Between 1990 and 2019, the number of hospitals fell by 30 percent while the number of primary care facilities reduced by seven percent. Most health care facilities are publicly managed. However, dental facilities and some specialist hospitals in the capital city of Yerevan are privately managed. Armenia has a relative undersupply of skilled health workers. In 2015, the country had 45.1 physicians per 10,000 population, below the density in Georgia (47.8), a neighboring country. Public spending on health in Armenia is low at USD 53 per capita in 2018, far below the average in UMI countries of USD 268 or in ECA (excluding high-income countries) of USD 249.27 Per capita public spending on health rose from USD 13 in 2000 to a high of USD 63 in 2016, declining to USD 53 in 2018. A decomposition of changes in the levels of per capita public spending on health over the past decade indicates that the overall decline was driven by reductions in total government spending and the relatively low priority for health in the national budget (Figure 3). Illustratively, despite a similar per capita GDP and a decrease in total government spending, Georgia saw significant growth in public spending on health over the same period driven by a high priority for health in the national budget. 20 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA FIGURE 3: Decomposition of changes in levels and composition of per capita public spending on health, 2009-2018 12 US $18 US $125 Annual Growth Rate 2009-2018 (%) 8 US $83 US $32 4 US $252 US $1,143 US $748 US $362 US $302 US $194 US $895 US $42 US $55 US $111 0 US $41 US $53 US $109 US $340 US $277 US $620 US $187 US $840 US $20 US $44 US $57 US $7 US $844 US $168 -4 KAZ HRV KGZ ARM UKR RUS TUR HUN BLR EST UZB TKM GEO TJK Economic Growth Change in total government spending Reprioritization 12 US $18 US $125 Annual Growth Rate 2009-2018 (%) 8 US $83 US $32 4 US $252 US $1,143 US $748 US $362 US $302 US $194 US $895 US $42 US $55 US $111 0 US $109 US $340 US $277 US $620 US $187 US $840 US $20 US $44 US $57 US $7 US $844 US $41 US $53 US $168 -4 KAZ HRV KGZ ARM UKR RUS TUR HUN BLR EST UZB TKM GEO TJK Domestic budgetary External budgetary SHI Source: Global Health Expenditure Database and World Development Indicators Notes: Levels of per capita public health spending on health in 2009 and 2018 denoted in red and black text respectively. Chapter 1. Why Mobilize Domestic Resources for Health? 21 Donor financing for the health sector is declining with Armenia’s transition to UMI status. For instance, essential activities for vaccination and tuberculosis care that had received support from Gavi and the Global Fund, will increasingly be funded through the state budget.28 Hence, to improve population health, there is a need to ensure financial protection for access to essential services that contribute to the appropriate management of NCDs, while sustaining funding for services that had been supported through external funders, by increasing public spending. Faced with the need for surge capacity in the sector, the Government has mobilized additional funds. Between March and November 2020, the Government spent AMD 36.37 billion on health services for the COVID-19 response, mobilized from the state budget and private donors.29 However, given historical trends, it is unclear whether this increase in funding for health care in the pandemic will persist, given the low priority for health spending in the national budget. In 2015, Armenia became a signatory to the Sustainable Development Goals, including the target of making progress towards UHC. This commits the Government to working towards guaranteeing access to quality health care for all, without risk of financial difficulty. To translate this political commitment into policy, the MoH has developed a concept note for the introduction of Universal Health Insurance. The note includes a proposal to mobilize additional revenues to cover access to the BBP for the entire population through labor taxes. But there are diverging views on the potential sources of funding for UHC reforms and important lessons from global experience for mobilizing funding to expand coverage in Armenia.30, 31 In a recent modeling exercise, relative to 2020, the cumulative cost of subsidizing up to 95 percent of household expenditure for health care in Armenia, including for essential services for maternal and child health, NCDs, infectious diseases, and outpatient medicines, rises from AMD 270 billion in 2020 to AMD 310 billion in 2031, and reaches AMD 390 billion in 2050, depending on the fiscal option for revenue-raising.32 There is also an ongoing actuarial costing exercise of a unified BBP, led by the MoH and supported by the World Bank. Global experience from efforts to mobilize additional resources to finance health care points to the stylized fact that voluntary prepayment plays a negligible role in universal coverage due to adverse selection. The rationale is that individuals with worse health risks are more likely to volunteer to prepay for care, more likely to use these services, and less likely to be able to afford the high risk-rated premiums than individuals with better health risks. Hence, voluntary prepayment, through adverse selection, limits the potential for risk redistribution, and equitable access to care – that is, access is linked to need rather than ability to pay. Compulsory contributions have thus been essential in countries that have been successful at progressing towards UHC.33 The contribution base for revenue mobilization is also an important consideration, for which there are also lessons from global experience. Countries with large informal sectors have an insufficient revenue base for labor taxes. In high-income countries, such as France, aging populations and distortions in labor market decisions have also informed transitions away from funding social services through predominantly relying 22 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA on payroll contributions to taxing consumption and financial assets.34 Hence, delinking contributions from entitlement is now considered fundamental to ensuring progress towards UHC. Regardless of how funds are raised, efforts to ensure universal access to high-quality health care may also require pooling risk, strategically allocating funds to providers, and ensuring service delivery readiness.35, 36 1.3 PURPOSE OF THIS REPORT This report assesses potential sources to expand public financing for health in Armenia, to finance improvements in health coverage for NCDs, and sustain access to services that had been supported through external funding. This report was developed following a request from the MoH for support in policy discussions on the domestic resource mobilization for health, the final decisions on which are the domain of the MoF. The target audience includes senior policy makers and technical advisers in the MoH, Ministry of Economy, and MoF. This report complements other technical assistance provided by the World Bank team to estimate the implication of additional taxes on economic growth, employment, and inequality; undertake actuarial costing of a revised, uniform (across the population) BBP; and identify reforms to improve allocative efficiency in the health sector. The team projected fiscal space under different scenarios, as agreed with the MoH, and consistent with the 2019 Draft Strategy for the Implementation of Tax Reforms in the Republic of Armenia, and Tax Code Amendments. The rest of this report is organized as follows. In Chapter 2, the authors review the conceptual framework for this report. Chapter 3 discusses the potential for resource mobilization through economic growth. Chapter 4 explores the increase in resources that may result from reprioritizing the health sector in the state budget. Chapter 5 estimates revenues that may accrue due to changes in payroll, value-added, turnover and excise taxes. Chapter 6 projects potential revenues from external funding to the health sector, while Chapter 7 estimates the potential increases in health sector resources from improvements in efficiency of current spending, and Chapter 8 presents conclusions derived from the analysis of findings. 23 CHAPTER 2. PROJECTING FISCAL SPACE FOR HEALTH Fiscal space has been defined as the availability of budgetary room that allows a government to provide resources for a desired purpose without prejudice to the sustainability of the government’s financial position.37 This definition explicitly references sustainability because efforts to increase sectoral spending by the Government should ideally be consistent with servicing debt obligations, financing critical programs, and ensuring its solvency, an important consideration in Armenia.38 Therefore, fiscal space for health refers to the potential for increased spending in the sector without jeopardizing solvency or crowding out spending for important objectives in other sectors.39 The concept of fiscal space for health can be understood by examining an algebraic representation of the Government’s intertemporal budgetary constraint, where the left- hand-side represents total government spending, and the right-hand-side represents total government revenue.40 Where: • - government’s non-interest expenditure in time, . • - the proportion of the Government budget allocated to the health sector. • −1- non-discretionary debt interest payments. • - taxes, fees, seigniorage, and other government revenues. • - the Government's domestic and foreign borrowing net of use of deposits. • - grants. • - other sources of funds. 24 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA In a 2010 paper, Tandon and Cashin developed a conceptual framework for assessing fiscal space for health, drawing on a prior framework described by Peter Heller.41 Per the Tandon and Cashin framework, the potential for increasing fiscal space for health can be assessed under five pillars: macroeconomic conditions, reprioritization of the health sector in the budget, sector-specific revenues, external resources, and efficiency gains. These pillars are reflected in the algebraic representation above. For example, an increase in could result from favorable macroeconomic conditions, and lead to an increase in spending on the health sector if is constant. If, alternatively, increases, then spending on health will rise, all other things being equal. Efficiency gains would result from maximizing outputs for a given health budget, The framework provides a useful structure for quantitative assessments of fiscal space. However, it does not account for the interdependence between pillars. For example, improved macroeconomic conditions may lead to a fall in external resources and increase the share of the national budget allocated to the health sector or a reduction in sector-specific revenues. The combined implications of these changes are not explored.42 With this caveat, in this report, authors estimate potential fiscal space for health in Armenia under the five pillars in the short (2021), medium (2023), and long (2031) term. These timelines were defined in discussion with the MoH. Fiscal space for health is expressed as additional annual spending relative to public spending on health in 2018. Thereafter, a qualitative summary of the relative potential for revenue mobilization across pillars and their overall implications for policy is provided. 25 CHAPTER 3. MACROECONOMIC ENVIRONMENT The conduciveness of the macroeconomic environment for fiscal space expansion reflects multiple factors, including growth, spending, debt, and inflation. The shock of the COVID-19 pandemic and regional political tensions resulted in the contraction of real GDP in Armenia by 8.0 percent in 2020, high above the average of 4.9 percent in the South Caucasus. The Armenian economy is projected to rebound to 4.6 percent growth in 2021, with a downside scenario of 3.1 percent, which compares favorably with a projected rebound to 2.7 percent growth in 2021 in the South Caucasus, on average. However, the downside scenario predicts growth of only 1.8 percent in 2021. During the COVID-19 pandemic, current spending increased by 19 percent year-on-year, while capital spending rose by 62 percent. Revenue fell by six percent year-on-year, resulting in a deficit of 1.7 percent of projected annual GDP in the year-to-July. To finance the deficit, the Government increased domestic and external borrowing, invoking an escape clause in the fiscal rule allowing debt increases in crises. Hence, while in 2019, government debt as a percentage of GDP was 53.5 percent, it increased by 10 percentage points to 63.5 percent in 2020, and it was projected to increase slightly to 63.6 percent in 2021. Inflation averaged 0.8 percent in the year-to-August due to low aggregate demand and fall in food and oil prices. The policy rate was reduced to 4.25 percent by the Central Bank of Armenia (CBA) in September 2020, the lowest level since 2006, and was later increased to 5.25 percent in December 2020. The CBA does not publish public projections of its policy rate. The team assessed the additional government spending on the health sector, excluding external resources, that could be achieved from additional GDP growth in Armenia, adjusting for general government gross debt and demographic predictors of variation in spending on health. First, the elasticity of government spending on health to 26 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA GDP was estimated, using data expressed in nominal and real terms, to ascertain that the elasticity estimates are not sensitive to this choice. This elasticity reflects the percentage change in government spending on health in response to a given percentage change in GDP. For example, an elasticity of three would imply that an increase in GDP by one percent predicts an increase in government spending on health by three percent. The empirical specification is below: Where: • - the log total government spending on health, excluding donor funds, computed as the sum of transfers from domestic revenue to the health sector and social insurance contributions. • � - the log of GDP in nominal or real terms. • - the general government gross debt expressed as a percentage of GDP, which has implications for the capacity of the government to meet its financial obligations.43 • - the control variables expressed in log which consists of percentage of older adult population who are 65 years or older, the percentage of children under five years of age, average life-expectancy, and infant mortality rate. These variables are chosen consistent with the literature and control for demographic characteristics across time, which may partially explain the variation in spending on health.44 The theoretical possibility of reverse causality in this regression specification is acknowledged, that government spending on health could have a contemporaneous effect on GDP. However, current empirical evidence does not support such a relationship. The elasticity of government spending on health to GDP is presented in Appendix 1A and Appendix 1B.45 The analysis showed that the elasticity in nominal terms remains stable at 1.21 following adjustment for demographic controls and general government gross debt as a percentage of GDP. Hence, a one percent increase in nominal GDP predicts an increase in government health expenditure by 1.21 percent in nominal terms. The adjusted R-squared is between 96 percent and 97 percent in all the specifications. While authors also estimated elasticity in real terms, they drew on the elasticity from the nominal variables for the subsequent analysis as this offers a conservative estimate of fiscal space. The team projected fiscal space for health using the above elasticity estimate as well as baseline (6.60 percent) and downside (5.90 percent) scenarios of GDP growth forecasts up to 2022 from the World Bank. The baseline forecast assumed that recovery started mid-summer 2021, while the downside scenario assumed a prolonged outbreak of COVID-19 and that recovery will take longer. This forecast is as of 9th October 2020. It was assumed that GDP growth between 2023 and 2031 would be consistent with the above baseline and downside scenarios. The additional annual revenue was calculated as the product of elasticity, GDP growth in time t, and the total government health expenditure in period t-1 for each period. That is: Chapter 3. Macroeconomic Environment 27 Relative to 2018, authors forecast additional annual public spending on health of AMD 0.6 to 3 billion in 2021, AMD 10 to 15 billion in 2023, and AMD 67 to 83 billion in 2031 (Figure 4). Appendix 1C presents estimates for the additional annual and cumulative revenue from GDP growth for each year between 2021-2031. As a robustness check, authors examined if their methodology replicates historical patterns in public spending on health. They predicted public spending on health between 2012 and 2017 using the elasticities above and compared it to actual public spending on health in those years. In Appendix 1D, authors show that the average error margin between observed and projected public spending on health is approximately 0.48 percent of the average GDP, suggesting that their estimate is not overly optimistic. The analysis assumed that total government expenditure on health is positively correlated with economic growth. In the pre-pandemic context, this relationship is positive and robust as discussed in Tandon and Cashin (2010). However, following unprecedented shocks to the health sector and economy, as in the COVID-19 pande- mic, this relationship may be different. For example, the COVID-19 pandemic has led to declines in economic growth accompanied by increasing the government expenditure on health to cushion the effects of the shock in several countries. In Armenia, despite a decline in GDP growth in 2020, the Government increased its total government spending on health to finance COVID-19-related health care, and transfers to vulnerable households and firms. The implication for this analysis is an underestimation of the projected fiscal space from macroeconomic conditions. As such, our analysis may be considered a lower bound under this pillar. FIGURE 4: Projected public spending on health from GDP growth 83,256 Additional Fiscal Space for Health (Millions AMD) 80,000 66,749 60,000 40,000 20,000 14,756 10,494 3,264 650 0 2021 2023 2031 Baseline Forecast Downside Forecast Source: Authors 28 CHAPTER 4. REPRIORITIZATION OF HEALTH IN THE GOVERNMENT BUDGET In 2018, public spending on health in Armenia was approximately 5.27 percent of total government expenditure. Government expenditure on health was equivalent to 59 percent of the allocation to education, 23 percent of the allocation to social protection (including pensions), and 32 percent of the allocation to defense in 2018.46 While these differences may reflect in part the needs across sectors in budget proposals, they may also indicate the low priority for health in public spending. Public spending on health per capita in Armenia (USD 52) is less than 50 percent of the level in neighboring Georgia (123 USD). Also, public spending on health as a percentage of total government spending is low in Armenia, below the median of 9.20 percent and the mean of 9.23 percent, among comparator countries (Figure 5). These comparator countries, listed in Table 2, were selected in discussion with the MoH, and met one or more of the following criteria: located in ECA, similar life expectancy, and shared political history as post-Soviet states. Chapter 4. Reprioritization of Health in the Government Budget 29 FIGURE 5: Public spending on health as a percentage of total government expenditure – Armenia and comparators Armenia 5.27 Belarus 10.61 Croatia 12.35 Estonia 12.54 Georgia 10.32 Hungary 9.92 Kazakhstan 9.10 Kyrgyz Republic 8.39 Russian Federation 9.76 Tajikistan 6.14 Turkey 9.30 Turkmenistan 8.71 Ukraine 8.87 Uzbekistan 7.87 0 5 10 15 Government Health Expenditure as % of General Government Expenditure in 2018 Source: Global Health Expenditure Database TABLE 2: Selected characteristics of comparator countries GENERAL GOVERNMENT FINAL LIFE PER CAPITA GDP PUBLIC SPENDING TAX REVENUE COUNTRY CONSUMPTION EXPECTANCY (CONSTANT 2010 ON HEALTH PER (AS A SHARE EXPENDITURE (AS % (YEARS) USD) CAPITA (USD) OF GDP) OF GDP IN 2018) Armenia 11 75 4,407 52 21 Belarus 16 74 6,586 251 15 Croatia 20 78 15,971 844 22 Estonia 20 78 19,933 1,143 21 Georgia 13 74 4,734 123 22 Hungary 20 76 16,793 748 22 Kazakhstan 8 73 11,166 168 12 Kyrgyzstan 17 71 1,091 37 18 Russia 18 73 11,844 362 11 Tajikistan N/A 71 1,073 16 N/A Turkey 15 77 15,190 302 18 Turkmenistan N/A 68 7,648 83 N/A Ukraine 21 72 3,106 109 20 Uzbekistan 14 72 2,374 31 14 Source: World Development Indicators 30 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA The team projected the additional public resources for health that would result from an increase in the proportion of total government health expenditure in general government expenditure, from 5.27 percent to 8.0 percent – an increase in priority for health in the state budget. This illustrative target, selected in discussion with the MoH, while ambitious for Armenia, remains below the mean and median in the comparator countries. This projection excluded external sources. Using forecasts of total government general expenditure from 2019 to 2022 from the World Bank, the team computed the total government general expenditure as a share of the real GDP (27.22 percent in 2022).47 This share was then applied to the GDP forecast in 2023 and 2024 to compute the total government general expenditure in 2023 and 2024.48 Subsequently, the team computed the growth rate of total government general expenditure between 2023 to 2024 and used this rate to estimate the total government general expenditure from 2025 to 2031. Based on the above, total government health expenditure (excluding external sources) was projected using the current proportion of 5.27 percent of total government general expenditure and the proposed proportion of 8.0 percent. Increasing the percentage of total government spending allocated to health from 5.27 percent to 8.0 percent will amount to additional revenue for health of 50.13 billions AMD, 51.99 billions AMD, and 82.24 billions AMD in 2021, 2023, and 2031, respectively as presented below in Figure 6. FIGURE 6: Projected fiscal space from reprioritization of government expenditure 250,000 241,002 Government Health Spending (Millions AMD) 200,000 152,354 158,760 146,914 150,000 96,780 100,363 100,000 82,242 50,134 51,991 50,000 0 2021 2023 2031 Total Government Health Spending at 5.27% of GGE Total Government Health Spending at 8.0% of GGE Additional Government Health Spending from Reprioritization Source: Authors 31 CHAPTER 5. TAXATION In 2019, tax revenue as a share of GDP in Armenia was 21 percent, above the mean (17 percent) and median (18 percent) in the aforementioned selected comparator countries in Table 2. In the same year, taxes constituted 89.8 percent of general government revenue.49 Tax revenue has risen from 15 percent in 2002 due to improved tax compliance, with increased contributions from income, profit, value-added, and excise taxes.50 In 2019, taxes on income, profits, and capital gains brought in a total revenue of AMD 591.2 billion or 40 percent of tax revenue. Property tax revenue was AMD 25.8 billion or less than two percent of tax revenue, while taxes on goods and services brought in a total revenue of AMD 761.3 billion or 51 percent of tax revenue.51 A large informal sector limits the potential for raising revenue from payroll taxes, given the difficulty in verifying total individual income.52 At the same time, empirical research suggests that with tax revenue above 15 percent of GDP, Armenia has the potential to move to a higher growth path, in part due to increases in social spending.53 The key consideration is identifying feasible methods of mobilizing additional revenue. As low and middle-income countries modernize their tax systems, they have tended to expand broad-based consumption taxes, particularly Value-Added Tax (VAT), and institute excise taxes on energy, alcohol, and tobacco on top of the normal VAT.54 While additional revenue can be raised from strong payroll tax systems, high levels of tax exemption and evasion limit revenue-raising potential in low and middle-income countries. 32 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA FIGURE 7: How have low and middle-income countries raised tax revenue? TAX LEVELS AND COMPOSITION Trade Corporate income tax The composition of taxes in richer countries differs from that of poorer countries, with greater emphasis Personal income tax on broad-based consumption and excise taxes. Excise tax Consumption 30 25 Revenue (% of GDP) 20 15 10 5 0 1990-99 2010-16 1990-99 2010-16 1990-99 2010-16 1990-99 2010-16 High-income Upper-Middle Lower-Middle Low-income Income Income Source: International Monetary Fund World Data Set In 2019, the MoF developed a Draft Strategy for the Implementation of Tax Reforms in the Republic of Armenia, aimed at improving investment attractiveness and improving tax compliance.55 The tax code was amended in June 2019, reducing the diversity of alternative tax regimes, decreasing payroll taxes, and increasing taxes on consumption and wealth. A flat payroll tax of 23 percent was introduced in 2020 and will be reduced to 20 percent in 2023. The profit tax rate was reduced from 20 percent to 18 percent, and the number of tax regimes was lowered from five to three: regular regime, turnover tax regime, and microentrepreneurs. Tax exemptions were introduced for small businesses, while excise rates on tobacco and alcohol have been increased. The team estimated additional fiscal space for health from increases in taxes, include excise taxes on alcohol, tobacco, and sugar-sweetened beverages (SSB), payroll taxes, VAT, and turnover tax. The fiscal options were selected in discussion with the MoH. A recent review by the WHO summarizes important considerations for countries that aim to introduce earmarking for health, to reduce distortions and ensure optimal design. Box 2 highlights some of the arguments for and against earmarking taxes for health.56 Authors also reflected on tax policies in neighboring countries and aimed to be consistent with Armenia’s current tax policy framework and Draft Strategy shared by the MoF. However, the team has also included projections for payroll tax, a proposal put forward by the MoH to the MoF, which diverges from the Draft Strategy, to support the ongoing policy discus- sions. The estimates were expressed in nominal terms and without future discounting. Chapter 5. Taxation 33 BOX 2: Earmarking tax for health Earmarking is setting aside all or a percentage of revenue from a tax or group of taxes, or a percentage of expenditure, for a specific purpose. Earmarking for health has been documented in at least 80 countries (Figure 8). It is an increasingly important part of policy discussions on mobilizing revenue to finance health care and, in some cases, contributes to strategies for reducing the consumption of unhealthy products, as in tobacco taxation. Earmarking for health often reflects political priority. Hence, even for taxes on unhealthy product consumption, the proportion of additional revenue allocated to the health sector varies. For example, in Panama, 100 percent was allocated to tobacco control activities; in the Philippines, 85 percent was allocated to health programs including UHC under the National Health Insurance Programme and alternative livelihood programs for tobacco farmers; in Romania, the additional one percent of budgetary funds from excise tax on cigarettes was used to finance sports. There are arguments for and against earmarking tax revenue for health. Earmarking can: protect these resources from competing political priorities; link the taxes from service users to the benefits from the service; make policymakers and providers more accountable to taxpaying service users; bypass rigid budgetary systems if earmarked revenues are pooled in an extrabudgetary fund; and contribute to improving public health when taxes target unhealthy consumption and expenditure is directed to public health activities. On the other hand, earmarks can contribute to distortions in the economy, of which a key concern is increased informal employment following earmarking of payroll taxes for health care. Revenues may be more susceptible to economic expansion and downturns and limit the government’s ability to adjust spending to respond to shocks. Earmarks may introduce fragmentation into health financing, and the resources from earmarking may not be additional to existing financing, as this can be offset by reducing funds from the general budget. 34 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA FIGURE 8: Earmarking for health across the world At least 80 countries are using earmarking for health Countries use income or payroll Countries earmark all or a portion tax to fund health care for the of revenues from tobacco taxes. population or formal-sector workers in a public scheme. 35 62 Countries earmark all or a portion of revenues from taxes on alcohol sales. Countries earmark a portion of their 9 Value-Added Tax (VAT). 4 Countries earmark revenue from taxes on other goods that can negatively affect health (e.g., sugar- 10 sweetened beverages). Countries earmark general revenue for health causes. 5 2 Countries earmark all or a portion of revenue 8 generated from lotteries. Countries earmark a portion of 1 transfers from the national level or Country introduced an earmarked levy earmark revenue generated at the on foreign personal money transfers and subnational level for health spending. mobile phone company revenue. Source: World Health Organization; Joint Learning Network 5.1 EXCISE TAX Corrective taxes on alcohol, tobacco, and SSBs, levied on top of normal VAT can support reductions in the disease burden that results from their consumption and increase long-term workforce productivity.57 At the same time, they can also be an administratively feasible and efficient source of revenue in the short-term in low and middle-income countries. The evidence shows that these health taxes also have non- significant negative effects or net positive effects on overall employment.58 The team projected fiscal space from excise taxes on alcohol and tobacco arising from changes under Armenia’s new tax code and also considered a scenario with excise taxes on SSBs, which was part of the new tax code plan, but was postponed. Chapter 5. Taxation 35 This analysis used data from several sources. The first is the excise tax revenue data for all product categories where excise tax was applicable between the period 2011 to 2019. This dataset comes from the State Revenue Committee (SRC) and contains information on the domestic goods tax base, imported goods tax base, and the excise tax revenue from both types of goods. The products are classified at the 4-digit Foreign Economic Activity Commodity Nomenclature (FEACN), and in a few cases some products appeared in 10-digit FEACN. The second dataset is the household consumption survey data for Armenia which comes from Armenian National Statistical Office (ARMSTAT). This survey data spans between 2002 to 2018 and includes self-reported data on the quantity and cost of products consumed. While this dataset provides rich information on the consumption patterns across households, there is a high proportion of missing values (over 90 percent across products). This may result in biased elasticity estimates, consumption growth forecast, and a lower base for quantity of products consumed. More details on this are provided below. The third dataset is a product import dataset that spans between 2008 to 2019 with products expressed at different levels of disaggregation from 4- to 10-digit FEACN. This dataset includes information on the import quantity and value (in USD) for most products that are of interest to this assessment. This dataset was used to compute the import prices and estimate elasticities in cases where the household survey data is not applicable, such as the petrochemical and natural gas products as discussed below. There does not seem to be a comprehensive dataset of average domestic product prices. Therefore, the team relied on several sources for this data. First, the team used import prices, constructed from import data for products that were not manufactured in Armenia. Second, the team relied on the average retail prices for a few products that were included as part of a report published by the Statistical Office of Armenia.59 Third, for the remaining products, the team used the average prices from two online sources – “numbeo” and “sas.”60 There are a few products for which the team could not find the average domestic prices on both websites, so import prices were used as a proxy for average domestic prices. The average product prices are presented below in Table 3. The team estimated the price elasticity of demand using household consumption survey data and a variant of the linearized almost ideal demand (LA/AIDS) model developed in Deaton and Muellbauer (1980) (Appendix 2A).61, 62 For petrochemical and natural gas products,63 the elasticity was estimated using the product-import dataset, by regressing the log of quantity imported on the log of unit import prices. The elasticity estimates shown in Table 3 ranged from -0.22 (for beer) to -2.2 (for natural fruit juice). The estimates imply that households will reduce their consumption of natural fruit juice by about 22 percent, given a rise in price by 10 percent, but for beer, consumption will be reduced by only 2.2 percent despite the same 10 percent increase in the price for beer. 36 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA TABLE 3: Excise Taxes UNIT OF PRICE DOMESTIC PRICE IMPORTED EXCISE TAX RATE ESTIMATED PRODUCT DESCRIPTION PRODUCTS (AMD) (AMD) 2020 (AMD) ELASTICITY Beer Liter 354 496 130 -0.220 Grape wines Liter 2,350 6,856 150 -0.59 Vermouth & other Liter 3,230 3,230 1,000 -0.59 grape wines Other fermented Liter 1,771 1,771 270 -0.445 beverages Ethyl alcohol Liter 4,000 10,564 3,380 -0.667 Alcoholic beverages Liter 18,071 18,071 7,000 -0.445 Brandy Liter 17,500 51,317 6,000 -1.047 Whiskey, rum, other Liter 30,000 19,757 7,000 -1.047 alcoholic beverages Vodka Liter 1,885.5 3,942 800 -1.085 Ind. tobacco substitutes Kilograms 20,002 20,002 1,500 -0.783 Gasoline Kilograms 314 314 40 -0.521 Diesel Kilograms 350 350 13 -0.521 Lubricants Kilograms 504 504 500 -1.547 Compressed nat. gas 1000 m 3 80,000 80,000 34,000 -1.106 Petroleum gases & other Kilograms 1,350 1,350 1 -2.11 similar hydrocarbons Cigarettes (filtered and 1000 pcs 21,570 21,570 9,625 -0.97 unfiltered) Condensed milk, such as sugar & cacao 1 liter 637 637 50 -1.034 supplements Lemonade 1 liter 258 258 50 -0.709 Other non-alcoholic 1 liter 300 300 50 -1.702 drinks (colas, Pepsi, etc.) Natural fruit juice 1 liter 500 500 50 -2.204 Notes: a) Import prices were computed by dividing import values by the quantity and converting the prices to AMD at 482.5 AMD=1 USD. b) Domestic price of beer and wine were taken from https://www.numbeo.com/cost-of-living/coun- try_result.jsp?country=Armenia. c) Domestic price of Ethyl Alcohol, Brandy, Whiskey was taken from the online store https://en.sas.am/. d) Domestic price for Vodka was taken from https://www.armstat.am/file/article/gner_2019_2.pdf e) For products where authors could not find their domestic prices, the import prices were used. f) For sugar-sweetened beverages (SSB), the analysis used 50 AMD as the excise tax, as already agreed with the MoH. The prices for the SSB were computed from consumption data. Since the excise tax is the same for imported and domestic products of each commodity, the analysis was conducted at the product level without distinguishing between imported and domestic products. Doing this required an average product price in 2019. The team computed the product prices by taking the weighted average of foreign and domestic product prices, with weights constructed as the share of quantity Chapter 5. Taxation 37 of foreign or domestic products purchased.64 The applicable excise tax rate in 2019 was also computed for both imported and domestic products by dividing the excise tax revenue by quantity purchased – their tax base – and weighted-average applicable excise tax in 2019 was computed using weights similar to that used for prices. Next, the team calculated the year-on-year growth rate for each imported and domestic product from 2011 to 2019 using data from the SRC, and averaged it for each product.65 This growth rate was applied to project future consumption quantity for each imported and domestic product from 2020 to 2031.66 Both quantities were then summed up to obtain the projected consumption quantity of each product in the absence of the newly proposed excise taxes. The team calculated the percentage change in price from the proposed excise tax67, and multiplied it by the elasticity estimates; this gives the percentage change in quantity. This percentage change was used to adjust the projected consumption quantity. Finally, the new excise tax rate was applied on this adjusted quantity to compute the expected excise tax revenue between 2020 to 2031. Table 4 displays the projected excise tax revenue under the new excise tax code. The team also computed the excise tax revenue under the existing tax regime in 2019. Under the team’s assumptions, the new excise tax code will increase tax revenue, relative to the tax regime in 2019, by approximately AMD 22.1 billion, AMD 43 billion and AMD 113.4 billion in 2021, 2023, and 2031, respectively. This amounts to an increase in domestic government spending on health as percent of nominal GDP in 2017 from 1.37 percent (the latest year with this variable) to 1.58 percent, 1.78 percent and 2.44 percent in 2021, 2023 and 2031 respectively, if these funds are 100 percent earmarked for the health sector. TABLE 4: Additional revenue from excise tax changes PROJECTED REVENUE FROM NEW PROJECTED EXCISE TAX UNDER ADDITIONAL FISCAL SPACE YEAR EXCISE TAX (BILLIONS AMD) 2019 REGIME (BILLIONS AMD) (BILLIONS AMD) 2021 129.76 107.62 22.14 2023 190.91 147.94 42.97 2031 788.22 674.85 113.37 Source: Authors’ calculations Appendix 2B presents the projections at the product level between 2020 to 2031. The projected excise tax revenue on SSB products is relatively low compared to other products. This is due to the shortcomings in the household data where response rate for these products was very low. There is no pre-existing excise tax on the total consumption of the products under the SSB category in 2019. Thus, the consumption quantity for the baseline year (2019) and its forecast were understated. This implies that the projected excise tax may be considered a lower bound of the expected excise tax on SSB. 38 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA 5.2 PAYROLL TAX As noted above, in Armenia there is a flat payroll tax of 23 percent, levied on labor income. In assessing the fiscal space from increment in payroll tax, authors used industry-level data on the average industry wages, total number of employees within each industry, and growth forecasts for each industry, provided by the CBA.68 The authors projected fiscal space for a hypothetical scenario in which payroll taxes increase by six percent, in line with the health reform concept note put forward by the MoH. The baseline assessment assumed perfectly inelastic labor supply. Since an increase in payroll tax reduces the net wages, one may think that such increase would reduce the supply of labor.69 However, there are reasons why this type of reduction in wages will have no effect on labor supply. First, an increase in payroll tax is applied to all workers in all industries. Manning (2003) found a positive wage elasticity of labor supply (0.75 to 1.5) when examining the effect of changes in wages at the firm level.70 This analysis examined a policy that reduces wages for every worker irrespective of their industry. Secondly, if part of the additional revenue from the increase in payroll tax is either earmarked for or allocated to health care, this frees up resources that would have been used for health expenditure. Hence, an increase in payroll tax may be a reallocation of health expenditure especially for sick workers or workers with sick family members. Also, access to health care has been shown to improve worker productivity (Dudu et al 2021), thus, labor supply after adjusting for productivity is unlikely to decrease. However, authors also assessed fiscal space from increases in payroll tax in line with Manning (2003) assuming wage elasticity of labor supply of 0.75 and 1.5. The fiscal space estimates under these assumptions were lower than estimates that assumed a perfectly inelastic labor supply. For each industry, the team forecast the average industry wages per employee (using the industry growth rate forecast) and multiplied it by the number of employees in each industry. It was assumed that entry into the labor market equals the exit in each year such that the size of the labor market and the distribution of skills remains unchanged. For scenarios where the wage elasticity of labor supply was used, the number of employees was adjusted using this elasticity. The additional payroll tax rate on the total industry wages in each period was applied and calculated across industries. The results are presented in Table 5 under three scenarios. The first panel, the baseline scenario, is the case where labor supply is perfectly inelastic to wages for the six percent increase in payroll tax in the years 2021, 2023 and 2031. An increase in payroll tax by six percent yields additional AMD 85.70 billion, AMD 94.91 billion, and AMD 142.90 billion in 2021, 2023 and 2031 respectively. This tax revenue reduces by less than four percent if it is assumed that the wage elasticity of labor supply is 0.75, and by less than eight percent, assuming an elasticity of 1.5. The revenue with the six percent additional payroll tax amounts to an increase in the domestic government spending on health to GDP ratio Chapter 5. Taxation 39 in 2017 from 1.37 percent to 2.18 percent, 2.26 percent, and 2.72 percent in 2021, 2023 and 2031, if these funds are 100 percent earmarked for the health sector. TABLE 5: Additional fiscal space from payroll tax (billions AMD) YEAR PAYROLL TAX, 6% INCREASE Scenario 1: Perfectly Inelastic Wage Elasticity Labor Supply 2021 85.70 2023 94.91 2031 142.90 Scenario 2: Wage Elasticity of Labor Supply=0.75 2021 81.84 2023 90.64 2031 136.47 Scenario 3: Wage Elasticity of Labor Supply=1.5 2021 77.98 2023 86.37 2031 130.04 5.3 VAT VAT refers to a consumption tax levied incrementally on the price of a good or service at each stage of production, distribution, or at sale to the consumer. The new tax code which took effect in 2020 lifted VAT exemptions on most goods and services. The standard VAT rate is 20 percent for goods and services. Under this section, this analysis projected additional revenue for a hypothetical VAT increment from 20 percent to 21 percent. The authors used data for all sectors in 2019, classified using the 4-digit NACE 2.0 product classification, obtained from the SRC, and the consumption forecast from the CBA for this assessment, adjusted due to the COVID-19 pandemic. The sector-level data consists of three variables – the sector identifier, turnover, and VAT revenue from each sector in 2019. This data comes with some limitations. First, within a sector, authors were unable to identify products or services that were partially or fully exempted from VAT in 2019. Thus, it was assumed that all products within a sector had a 20 percent VAT in 2019. Hence, authors underestimated the tax revenue from this proposed increment in VAT rate by implicitly assuming that the tax base for products and services that were VAT-exempted in 2019 was not zero. Thus, the resulting tax revenue is a conservative estimate. The second limitation lies in product aggregation. Authors observed products that were aggregated at the 4-digit sector-level with no information on the specific products or services within a sector. This made it difficult to estimate the price elasticity 40 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA of demand. However, this limitation does not appear as a source of large upward bias given the marginal increase in prices which stands below one percent.71 The VAT tax base in 2019 was constructed by dividing the VAT revenue by 20 percent. This tax base is the implied sales revenue before VAT. The implied sales revenue was forecast using the consumption forecast from CBA for the years between 2020 to 2031. The forecast stands at 0.8 percent in 2020, 6.0 percent in 2021 and 3.8 percent in 2022. The forecast of 3.8 percent was applied in years between 2023 and 2031. Also, the current (20 percent) and hypothetical increased (21 percent) tax rate was applied to the VAT tax base and the additional revenue computed from the difference. Table 6 presents the projections for the years 2021, 2023 and 2031. The estimates imply an additional revenue of AMD 49.21 billion, AMD 53.02 billion and AMD 71.5 billion in 2021, 2023 and 2031, respectively. Overall, if the revenue from the proposed increment in VAT rate is used for health financing, this will translate to an increase in the domestic government spending on health to GDP ratio in 2017 from 1.37 percent to 1.84 percent, 1.87 percent and 2.04 percent in 2021, 2023 and 2031. TABLE 6: Additional fiscal space from hypothetical increase in VAT REVENUE AT 20% VAT REVENUE AT 21% VAT ADDITIONAL REVENUE YEAR (BILLIONS AMD) (BILLIONS AMD) (BILLIONS AMD) 2021 984.20 1,033.40 49.21 2023 1,060.41 1,113.45 53.02 2031 1,429.07 1,500.53 71.45 5.4 TURNOVER TAX Turnover tax, akin to VAT, is levied on an ad valorem basis, to a production process or stage. Unlike VAT, this tax is levied on intermediate and possibly, capital goods. In Armenia, it typically replaces the VAT or corporate income tax obligations for small and medium-sized enterprises (SMEs) below a specific revenue threshold. The turnover tax regime is reviewed in Table 7. Authors estimated the implications for fiscal space for health of two hypothetical measures to raise the turnover tax collection and to encourage businesses to opt for the regular regime. First, a hypothetical increase in the turnover tax rate may serve as an additional source of revenue and might nudge businesses to choose the regular regime over the turnover regime. The second, which is a proposal the government intended to implement in 2020 but postponed, aims at reducing the turnover regime threshold from the current AMD 115 million per annum to AMD 58.35 million. This will limit the number of firms eligible for the turnover tax regime and will likely increase the tax revenue because more firms will be subject to both VAT and profit tax. Once a firm falls below the threshold, 100 percent compliance is assumed. Chapter 5. Taxation 41 TABLE 7: Business activity and turnover tax TYPE OF ACTIVITY STATUTORY RATE (%) HYPOTHETICAL RATE (%) Income from commercial (buying and 5 6 selling) activities Income from commercial (buying and selling) 1.5 2.5 activities related to secondary raw materials Income from newspaper alienation by editors 1.5 2.5 Income from production activities 3.5 4.5 Rent payments, interest, royalties, real estate, 10 and other income from vehicle alienation 11 Income from notarial activities 10 11 Income from organization of lotteries 25 26 Income from public catering activities 6 7 Income from other activities if they are not part 20 21 of public catering Income from other activities 5 6 This table was taken from PWC report https://taxsummaries.pwc.com/armenia/corporate/other-taxes Fiscal space expansion was assessed from a hypothetical increment in the turnover tax. The statutory turnover tax in 2019 varied across business activities, ranging from 1.5 percent to 25 percent (Table 7). Authors therefore examined a hypothetical increase in the turnover tax by one percentage point for each business activity—e.g. from 1.5 percent to 2.5 percent—and evaluated its contribution to fiscal space in 2021, 2023 and 2031. The data for this assessment comes from the SRC. It spans from 2013 to 2019 and contains information on turnover tax rate for each business activity, number of firms whose main income lies in each of the business activity,72 total turnover, and total tax paid. Authors computed the average year-on-year growth rates of total turnover per firm, and the number of firms whose main income lies in each of the business activity. This average growth rate was then applied to forecast future values of turnover per firm, and the number of firms within each business activity from 2020 to 2031. The dataset has limitations. It is unbalanced for 58 percent of the business activities (7 out of 12) with different magnitudes of missing values across business activities. For example, information on “income from organization of lotteries” appeared only in 2018 and 2019. These limitations make it impossible to use average year-on-year growth rates in forecasting the future values of the variables of interest. To resolve this challenge, it was assumed that the growth rate for the total turnover per firm, and the number of firms in each business activity with less than four observations, is equal to the growth rate in the service sector.73, 74 The average turnover per firm in each year was calculated by dividing the total turnover in each business activity by the number of firms whose main income lies in the business 42 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA activity.75 The average year-on-year growth rate for turnover per firm and the number of firms within each business activity was then computed. For business activities that appeared fewer than four times in our panel, the growth rate of the service sector, which was provided by the CBA, was used. Using the average year-on-year growth rate or service sector growth rate (where applicable), this analysis forecast turnover per firm and number of firms in each business activity from 2020 to 2031. The forecast turnover per firm was multiplied by the number of firms in each business activity, and then the tax income for both the statutory and proposed tax rate was calculated, taking their difference and summing over the business activities to estimate the additional fiscal space for each year between 2020 to 2031. Table 8 shows the additional fiscal space from an increase in turnover tax rate for the years 2021, 2023 and 2031, corresponding to the short, medium and long-term, respectively. An increase in turnover tax is expected to increase fiscal space by approximately AMD 11 billion, AMD 14 billion and AMD 23 billion in 2021, 2023 and 2031, respectively, which amounts to an increase in the domestic government spending on health to GDP ratio in 2017 from 1.37 percent to 1.48 percent, 1.51 percent and 1.60 percent in 2021, 2023 and 2031, respectively, if these funds are 100 percent earmarked for the health sector. TABLE 8: Fiscal space contribution from an increase in turnover tax BASELINE TURNOVER TAX PROPOSED TURNOVER TAX ADDITIONAL FISCAL SPACE YEAR INCOME INCOME FROM TURNOVER TAX (BILLIONS AMD) (BILLIONS AMD) (BILLIONS AMD) 2021 58.09 69.27 11.18 2023 73.96 88.22 14.26 2031 124.50 148.24 23.74 Next, authors propose a reduction in the turnover tax regime threshold from the current AMD 115 million per annum to AMD 58.35 million. Firms with turnover above AMD 58.35 million will be subjected to VAT and profit tax on their sales. Hence, this will imply a reduction in turnover tax and increase in both VAT and profit tax. The aim is to assess VAT revenue arising from such firms and compare it with the revenue from turnover tax in the absence of this threshold. Profit tax is not considered given the difficulty estimating profits of establishments using available data. This analysis requires data on the VAT base for such firms, which is unavailable. Authors therefore used information on the turnover, which is typically larger than the VAT base,76 and relied on two scenarios in computing the VAT base. The first scenario assumed the VAT base to be 50 percent of the turnover tax base, while the second, assumed the VAT base to be 75 percent of the turnover base. In the data used for the VAT analysis, the average ratio of the VAT tax base to the turnover is 75 percent. Hence, Chapter 5. Taxation 43 the second scenario is the authors’ preferred scenario. The first scenario is a modest estimate of VAT base and provides a lower bound on the fiscal space that is available from this proposal. The dataset for this assessment contains information on the number of firms, their total turnover and the turnover tax paid, for all firms below and above this threshold (AMD 58.35 million) in each year between 2013 to 2019. It was assumed that the growth rate of total turnover is like that of the service sector growth rate, which was provided by the CBA. This implies that the growth rate of turnover tax revenue is the same as the service sector growth rate. Only the case for which the total turnover exceeds AMD 58.35 million was considered, before applying the service sector growth rate on the total turnover and turnover tax revenue to predict their future values. For each of the two scenarios (50 percent vs 75 percent), the VAT base is computed and multiplied by the VAT rate of 20 percent. Finally, the difference between the VAT revenue and the predicted turnover tax revenue, which is the additional fiscal space from this proposal, is calculated. The results are presented in Table 9. The first panel assumes the VAT base is 50 percent of the turnover. The estimates imply that this proposal offers additional tax revenue of over AMD 22.6 billion, AMD 25.1 billion, and AMD 37.9 billions AMD in 2021, 2023 and 2031, respectively, which translates to an increase in domestic government spending on health as percent of nominal GDP in 2017 from 1.37 percent (the latest year this variable was available) to 1.59 percent, 1.61 percent and 1.73 percent in 2021, 2023 and 2031, respectively, if these funds are 100 percent earmarked for the health sector. The second panel provides estimates for authors’ preferred assumption, that the VAT base is 75 percent of the turnover. The fiscal space from this scenario is AMD 41.1 billion, AMD 45.5 billion, AMD 68.9 billion in 2021, 2023 and 2031, respectively, which translates to an increase in domestic government spending on health as percent of nominal GDP in 2017 from 1.37 percent to 1.76 percent, 1.80 percent and 2.02 percent in 2021, 2023 and 2031, respectively, if these funds are 100 percent earmarked for the health sector. TABLE 9: Fiscal space from reducing the threshold of turnover tax YEAR FORECAST TURNOVER TAX VAT BASE ADDITIONAL FISCAL SPACE (BILLIONS AMD) (BILLIONS AMD) FROM 20% VAT (BILLIONS AMD) Panel I: VAT base is assumed to be 50% of Turnover 2021 14.33 36.97 22.64 2023 15.89 40.99 25.10 2031 24.02 61.96 37.94 Panel II: VAT base is assumed to be 75% of Turnover 2021 14.33 55.45 41.12 2023 15.89 61.49 45.60 2031 24.02 92.94 68.92 44 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA Table 10 presents a summary of the additional fiscal space from tax options, and the resulting domestic government spending on health, under a scenario where the additional funds are 100 percent earmarked for the health sector. TABLE 10: Summary of additional fiscal space for health from hypothetical scenarios TAX OPTIONS ADDITIONAL FISCAL SPACE (BILLIONS 2021 2023 2031 AMD) Excise Tax 100% earmarked for the health sector 22.14 42.97 113.37 Payroll Tax (6% increase) 100% earmarked for the health sector 85.70 94.91 142.90 VAT 100% earmarked for the health sector 49.21 53.02 71.45 Increase in Turnover Tax 100% earmarked for the health sector 11.18 14.26 23.74 Reducing Threshold 100% earmarked for the health sector 41.12 45.60 68.92 Turnover Tax 45 CHAPTER 6. SECTOR-SPECIFIC EXTERNAL FUNDING Under this pillar, authors assessed the potential for external funding to be a source of additional fiscal space for health in Armenia. External funding for the health sector comes in two forms. First, assistance provided directly to government to fund government planned expenditures, and secondly, as a transfer to non-governmental health organizations specifically aimed for health-related purposes. Authors refer to the former as on-budget support for health, while the latter is off-budget support. Figure 9 shows that the external financing for health as a proportion of current total health expenditure in Armenia is declining over time. This finding is consistent with what is typically observed in UMI economies (Figure 10). 46 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA FIGURE 9: Total health expenditure from external sources as a percentage of total health expenditure in Armenia percentage of Total Health Expenditure 20 Foreign transfers for health as 15 10 5 0 2000 2005 2010 2015 2020 Years (2000-2017) Source: WHO Database FIGURE 10: Change in composition of health financing with transition LOWER- UPPER- MIDDLE- MIDDLE- LOW-INCOME INCOME INCOME HIGH-INCOME 10,000 50 Public Health spending per capita (USD) Share of health expenditure (%) 2,500 40 500 30 100 20 25 10 5 0 500 995 3,895 12,056 35,000 100,000 GNI per capita, USD Health expenditure per capita (left axis) OOP share (right axis) External share (right axis) Source: World Bank Chapter 6. Sector-Specific External Funding 47 The aim is twofold. First, to assess the contribution of external assistance for health in Arme- nia and, second, to assess whether external assistance is fungible, that is whether external sources for health financing serve as a substitute for government health expenditure. For the fiscal space assessment, the focus is on external development lenders and donors’ commitments to spending on health in Armenia.77 The analysis is limited to lending from the World Bank, the Asian Development Bank (ADB), and donor funding from the Global Fund to Fight AIDS, Tuberculosis and Malaria (GFATM). These commitments have not been modified significantly, given the COVID-19 pandemic and Armenia’s status as an UMI post-transition country. Table 11 presents these lenders and donors commitments to the health sector in general and vertical programs in the medium-term (2021-2024), where the latter restricts funding to specific diseases or programs. There will be higher lender funding for the health sector in Armenia in 2021 and 2022 than in 2023. TABLE 11: Projected fiscal space from external funding for health YEAR WORLD BANK ADB GFATM TOTAL COMMITMENT TOTAL COMMITMENT: (MILLIONS AMD) (MILLIONS AMD) (MILLIONS AMD) FOR HEALTH SECTOR IN VERTICAL PROGRAMS GENERAL (MILLIONS AMD) (MILLIONS AMD) 2021 3,200 9,600 1,202.27 12,800 1,202.27 2022 3,200 9,600 1,546.75 12,800 1,546.75 2023 3,200 N/A 1,546.75 3,200 1,546.75 2024 N/A N/A 1,546.75 N/A 1,546.75 Total 9,600 19,200 5,842.52 28,800 5,842.52 Notes: ADB is Asian Development Bank, GFATM is Global Fund to Fight AIDS, Tuberculosis and Malaria, UHC is Universal Health Coverage, and HSP are vertical programs where there is less discretion over allocations by the government. The vertical programs are HIV/AIDS and Tuberculosis financed by GFATM. All the external financial assistance is denominated in US dollars. Projected financial assistance from ADB comes in the form of Results-Based Loans. To be consistent, US dollars was converted to Armenian dram at the exchange rate of 1 USD = 480 AMD. The authors assessed if external sources of finance are fungible. That is, do they replace government expenditure in the health sector? The analysis determined that the growth in total foreign transfers in Armenia is associated with a drop in total current health expenditure (net of foreign transfers), driven by direct foreign transfers.78 Also, the growth in total foreign transfers is positively associated with the growth in total government spending on health.79 Disaggregation of total foreign transfers into direct foreign transfers and transfers to government showed that direct foreign transfers are positively associated with total government health expenditure.80 The analysis determined the relationship between the OOP expenditure on health and the total foreign transfers for health. The growth in total foreign transfer is negatively related to the growth in total OOP health expenditure.81 By disaggregating total foreign transfers into foreign transfers to the government and direct foreign transfers, it was found that this negative relationship is primarily driven by the direct foreign transfers for health.82 In sum, external sources of finance do not replace government expenditure on the health sector, however they may serve as substitutes for OOP expenditure. 48 CHAPTER 7. EFFICIENCY GAINS The effect of increased spending on health system outcomes depends on the extent to which resources are used efficiently. In Armenia, there are indications of the potential for improving the efficiency of resource use. Figure 11 shows that while life expectancy at birth is above the median in Armenia, Turkey has attained an even higher life expectancy at birth than Armenia with lower total per capita health expenditure levels. The team assessed the efficiency of total and government spending on health in Armenia, relative to the comparator countries identified in Chapter 4 and listed in Appendix 4A. A data envelopment analysis (DEA)83 was used to develop scores of the relative technical efficiency of transforming inputs (health financing) into outputs (health outcomes) in the health sector, adjusting for relevant variables. The analysis aimed to estimate potential savings in health expenditure for the current level of health system performance (input-oriented model) and the potential increases in health system performance that might be attained within the current level of inputs (output- oriented model). Chapter 7. Efficiency Gains 49 FIGURE 11: Comparing life expectancy at birth vs total per capita spending on health 2017 79 Estonia Croatia 77 Turkey Life Expectancy at Birth Hungary 75 Armenia Belarus Georgia median Kazakhstan Russian Federation Uzbekistan Ukraine Moldova 71 Kyrgyz Republic Tajikistan Turkmenistan 67 0 median 500 1,000 1,500 Total Spending on Health Per Capita (constant 2018 USD) Source: World Bank Health system performance was measured in terms of eight indicators: life expectancy at birth, healthy life expectancy at birth, life expectancy at 60, healthy life expectancy at 60, health-adjusted life expectancy, amenable mortality rate, adult mortality rate, and the under-five mortality rate. Inputs considered were the total spending on health per person and the government spending on health per person, where the latter is important for examining the potential for improvements in the efficiency of public spending on health. In keeping with a similar analysis by Medeiros and Schwierz (2015),84 in some of the models, the analysis is adjusted for population-level predictors of health outcomes, including GDP per capita, prevalence of overweight, total alcohol consumption per capita, and prevalence of tobacco usage in 2016.85 An overview of the analytical description of the DEA linear programming approach, assuming variable returns to scale, is presented in Appendix 4D. A full description can be found in Medeiros and Schwierz (2015), Coelli et. al. (2005), and Afonso and Aubyn (2006).86, 87 The savings from improving efficiency were calculated by examining the extent to which spending on health would decrease if Armenia adopted the most efficient system and achieved the same level of health outcomes.88 A range of model specifications was tested to provide sensitivity checks for each input and output- oriented DEA model (Table 12). The team’s preferred model is Model 4, which adjusts for GDP per capita and population health risks, and approximates efficiency gains that accrue to improving efficiency of public spending on health services.89 50 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA TABLE 12: DEA model specifications OUTPUT VARIABLES INPUT VARIABLES Model 1 Model 2 Model 3 Model 4 (Adjusted Model 1) (Adjusted Model 2) Life-Expectancy At Birth Healthy Life-Expectancy Total Spending on Government Spending At Birth Health per Person on Health per Person Life-Expectancy At 60 + + Government 1. GDP per capita 1. GDP per capita in Healthy Life-Expectancy Total Spending Spending on in PPP PPP At 60 on Health per Health per 2. % Overweight 2. % Overweight Health-Adjusted Life Ex- Person population (i.e. with population (i.e. with Person pectancy BMI>=25 BMI>=25 3. Total alcohol con- 3. Total alcohol con- Amenable Mortality Rate sumption per capita sumption per capita Mortality Rate (Adults) 4. Tobacco 2016 4. Tobacco 2016 Mortality Rate (Under-five) Figure 12 illustrates the analytical output using the production function for the DEA model with life expectancy at birth as the outcome variable and total spending on health per person (2018 USD) as the input variable. The blue line is the production frontier (without bias correction), the red dashed line is the production frontier estimated by the bootstrap DEA (bias-corrected DEA), and the black dashed line is the 97.5 percent confidence interval of bias corrected values. Armenia is below the bias- corrected production frontier, an indication of inefficiency. The efficiency score would thus be below one (0.63). However, its efficiency is higher that several comparator countries, including Russia, Turkmenistan, Moldova, Hungary, among others. FIGURE 12: Estimated (bias-corrected) production frontier - an example 80 HR TR HU EE UZ HA AR TJ KG MD GE RM KZ TM Life Expectancy at Birth 60 DEA Boot CI97.5 40 20 0 0 500 1000 1500 Total Health Spending per Person (2018 USD) Source: Authors Chapter 7. Efficiency Gains 51 Table 13 shows efficiency scores for Armenia across a range of health outcomes.90, 91, 92 These scores can be interpreted as the fraction of observed inputs that can achieve the observed output if the health system were to be efficient.93 In the unadjusted models, Armenia falls below the production possibility frontier for every health outcome, with the opportunity to improve efficiency illustrated by the scores for under-five mortality (0.89) and adult mortality in the adjusted models (0.88). Adjusting for GDP per capita and population health risks, Armenia scores relatively higher on efficiency of public spending on health, when other outputs are considered, including amenable mortality, life-expectancy, and healthy life-expectancy. TABLE 13: Efficiency scores for bias-corrected DEA HEALTH INPUTS HEALTH INPUTS + ECONOMIC AND LIFESTYLE VARIABLES 1. Total Spending on 1. Government Spend- Health per Per- ing on Health per Total Government son (THS) Person (GHS) OUTPUTS 2. GDP per capita 2. GDP per capita Spending Spending (HEALTH OUTCOMES) on Health on Health in PPP in PPP 3. % Overweight pop- 3. % Overweight per Person per Person ulation (i.e., with population (i.e. with (2018 US (2018 US BMI>=25) BMI>=25) Dollars) Dollars) 4. Total alcohol con- 4. Total alcohol con- sumption per capita sumption per capita 5. Tobacco 2016 5. Tobacco 2016 Life-Expectancy M1 0.63 0.63 0.92 0.93 At Birth Healthy M2 Life-Expectancy 0.74 0.61 0.91 0.92 At Birth Life-Expectancy M3 0.51 0.68 0.93 0.93 At Birth At 60 Healthy M4 Life-Expectancy 0.65 0.69 0.92 0.92 At Birth At 60 Health-Adjusted M5 0.57 0.62 0.92 0.92 Life-Expectancy Amenable M6 0.44 0.66 0.93 0.92 Mortality Rate, % Mortality Rate M7 0.32 0.57 0.80 0.89 (Adult), % Mortality Rate M8 0.25 0.69 0.82 0.88 (Under-five), % In the models that adjusted for GDP per capita and population health risks, it is noted that within public spending on health, Armenia could have a life-expectancy at birth that is 1.31 years higher, a healthy life expectancy at birth that is 1.07 years higher, an adult mortality rate that is 12 percent lower, and an under-five mortality rate that is 22.58 percent lower, if the health system achieved the efficiency of the best-performing comparators. In 13 out of the 16 models, Armenia performed better than more than half of the countries included in the assessment.94 52 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA The estimated potential per capita savings that would accrue to improvements in efficiency of spending on health is presented in Table 14. In the models that adjusted for GDP per capita and population health risks, the potential per capita savings lies between USD 0.43 or AMD 206 (in per capita terms) when the outcome considered is adult mortality, and USD 35.44 or AMD 17,011 (in per capita terms) when the outcome considered is child mortality.95 This is equivalent to a health system savings of AMD 0.6 billion, if adult mortality is considered, and AMD 3.5 billion if under-five mortality is used to benchmark Armenia against comparators.96 TABLE 14: Potential per capita savings (in US dollars) from efficiency improvements HEALTH INPUTS HEALTH INPUTS + ECONOMIC AND LIFESTYLE VARIABLES 1. Total Spending on 1. Government Spend- Health per Per- ing on Health per Total son (THS) Person (GHS) Government 2. GDP per capita 2. GDP per capita OUTCOME VARIABLE Spending Spending in PPP in PPP on Health on Health 3. % Overweight 3. % Overweight per Person per Person population (i.e. with population (i.e. with (2018 US (2018 US BMI>=25) BMI>=25) Dollars) Dollars) 4. Total alcohol con- 4. Total alcohol con- sumption per capita sumption per capita 5. Tobacco 2016 5. Tobacco 2016 Life-Expectancy M1 68.78 7.24 67.28 2.69 At Birth Healthy M2 Life-Expectancy 18.18 8.22 275.20 17.44 At Birth Life-Expectancy M3 136.35 2.53 46.27 6.82 At Birth At 60 Healthy M4 Life-Expectancy 45.92 0.00 92.89 9.56 At Birth At 60 Health-Adjusted M5 89.85 6.79 45.43 7.06 Life-Expectancy Amenable M6 157.51 4.20 76.40 4.32 Mortality Rate, % Mortality Rate M7 199.81 0.00 103.36 0.43 (Adult), % Mortality Rate M8 250.71 7.96 334.81 35.44 (Under-five), % These findings on efficiency must be considered within the broader context of the very low public financing for health in Armenia compared to peer countries. Nevertheless, empirical studies of the Armenian health system reveal potential sources of inefficiency, including small-scale procurement, budget program fragmentation, suboptimal allocations in the benefits package, the imbalance between prevention and curative care in spending, and high OOP health expenditure. Previous reports on public financial management, strategic purchasing, and benefits package design discussed these factors.97, 98, 99 In partnership with the MoH, there is also an ongoing assessment of Chapter 7. Efficiency Gains 53 service delivery that will explore opportunities to improve the efficiency of resource use and address gaps in coordination across service delivery levels. The following are comments on selected opportunities to improve the Armenian health system efficiency: 1. Decentralized procurement: As per procurement of medical goods, smaller facilities often lack the skills to develop adequate technical specifications for procured goods and the small size of the market may not attract potential bidders.100 An analysis of the 2016 decentralized purchase of 2,147 medicines, revealed that variation in the prices of drugs was 42 percent on average, with an estimated AMD 1.3 billion potential savings lost. 2. Budget program fragmentation: While Armenia has adopted program-based budgeting, in practice, there is a limited role for program managers and a lack of flexibility for reallocation of the budget. Budget programs are broken down into numerous activities and the redistribution of funding requires revisions of facility budgets over the course of the year, with lengthy approval procedures. 3. Suboptimal benefit package allocations: Allocations to health services within the BBP are not systematically informed by assessments of relative cost-effectiveness of alternative interventions. A recent study estimated that within the current funding envelope for the BBP, an additional 10,600 to 31,400 DALYs can be averted, relative to the baseline of 120,000 DALYs averted given current allocations within the BBP.101 4. Suboptimal preventive-curative care mix: Armenia allocates a relatively high and growing amount of spending on health to curative care. Between 2014 and 2018, the proportion of current financing for health allocated to curative care rose from 57 percent to 59 percent, above the average level in the European Union (53.4 percent). Spending on preventive care is less than two percent, below the EU average of 2.7 percent. There is no comparable database for UMI and ECA countries. Fund flows reflect the hospital-centric patterns of health care use, with 66 percent of hospital visits not being preceded by a referral from a primary care physician.102, 103 5. High OOP expenditures: The difference in potential per capita savings from efficiency gains in the models considering total health expenditure and public health expenditure provides insight into the health system waste attributable to private spending, majority of which is OOP.104 For example, when adult mortality is considered, the potential per capita savings for public health expenditure of USD 0.43 or AMD 206.4 is dwarfed by the per capita savings for total health expenditure of USD 103 or AMD 49,440. Financing health care predominantly via OOP is inefficient due to market failures in health service delivery. For example, information asymmetry confers discretionary power on health providers contributing to induced demand in line with the provider (rather than patients) interests. In emergencies, patients often have little time or ability to shop around for care of the best quality and least cost, and the inherent uncertainty involved in translating better care to improved health reduces opportunities to learn from experience.105 54 CHAPTER 8. CONCLUSIONS Given the COVID-19 pandemic, regional conflict, and projected economic contraction, the Government of Armenia is confronted with significant pressure to increase public spending for a range of priorities, including in defense, health care, education, jobs, and social protection, in the short to medium-term. This analysis presents estimates of the potential for revenue mobilization for UHC from a range of hypothetical scenarios, that can inform the dialogue between the MoF and MoH on financing improvements in health coverage. The estimates are summarized in Table 15 and Figure 13 below. These results suggest the additional revenue needed to support expanded coverage, which may be up to AMD 310 billion by 2031, cannot be fully financed under any single pillar, according to the hypothetical scenarios considered. Chapter 8. Conclusions 55 TABLE 15: Overview of additional fiscal space across pillars ADDITIONAL FISCAL SPACE FOR HEALTH 2021 2023 2031 PILLAR (BILLIONS (% of GDP (BILLIONS (% of GDP (BILLIONS (% of GDP AMD) 2020) AMD) 2020) AMD) 2020) Macroeco- Baseline forecast 6.1 0.10 19.1 0.31 97.7 1.58 nomic conditions Downside forecast 3.1 0.05 14.3 0.23 78.6 1.27 Increase in the current ratio of Reprioritiza- public spending on health in 50.13 0.81 51.99 0.84 82.24 1.33 tion of health total government expenditure from 5.27% to 8.0% 100% earmarked for the Excise Tax 22.14 0.36 42.97 0.70 113.37 1.83 health sector Payroll Tax 100% earmarked for the 85.70 1.39 94.91 1.54 142.90 2.31 (6% increase) health sector Value-Added 100% earmarked for the 49.21 0.80 53.02 0.86 71.45 1.16 Tax health sector Increase in 100% earmarked for the 11.18 0.18 14.26 0.23 23.74 0.38 Turnover Tax health sector Reducing 100% earmarked for the Threshold 41.12 0.67 45.60 0.74 68.92 1.11 health sector Turnover tax External General health sector 12.8 0.21 3.2 0.05 N/A Resources Vertical programs 1.2 0.02 1.5 0.02 N/A Potential annual savings, in billions AMD, considering adult Upper Lower mortality rates, if spending ef- bound = bound = N/A N/A ficiency were equivalent to the 0.6 0.01 most efficient comparator Efficiency Gains Potential annual savings, in billions AMD, considering Upper Lower under-five mortality rates, bound = bound = N/A N/A if spending efficiency were 3.5 0.06 equivalent to the most effi- ciency comparator 56 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA FIGURE 13: Summary of fiscal space projections Macro conditions (baseline forecast) Macro conditions (downside forecast) 150 97.7 100 78.6 50 19.1 6.1 14.3 3.1 0 2021 2023 2031 2021 2023 2031 Projected Revenue (in Billions AMD) Reprioritization of health Excise Tax 150 113.37 82.24 100 50.13 51.99 42.97 50 22.14 0 2021 2023 2031 2021 2023 2031 Payroll Tax 142.9 VAT 150 85.7 94.91 100 71.45 49.21 53.02 50 0 2021 2023 2031 2021 2023 2031 Increase in Turnover Tax Reducing Threshold Turnover Tax 150 68.92 100 41.12 45.6 23.74 11.8 14.26 50 2021 2023 2031 2021 2023 2031 0 Source: Authors The potential for raising revenue by increasing the priority for the health sector in the national budget is significantly higher than additional fiscal space that may accrue to other pillars, outside payroll taxes. Advocating for increased health sector allocations should be considered given the relatively higher priority for health among peer countries. As stated previously, public spending on health per capita in Armenia is less than half the level seen in neighboring Georgia, despite similar income levels in both countries. This in part because Georgia, allocates above nine percent of the national budget to health, higher than Armenia (5.27 percent). A progressive increase in budget- ary allocations from 5.27 percent to 8.0 percent — which stands below the mean and median in comparator countries — could mobilize an additional AMD 50 billion in 2021, increasing the 2018 health budget by over 50 percent. Beyond the revenue generating potential for the health sector, the choice of tax policies to finance UHC, including the optimal tax rate, may reflect the implications of such policies for the wider economy and social issues. Hence, at the request of the MoH, the World Bank has modeled the macroeconomic impact of changes in tax policy to finance UHC in a related report by Dudu et al (2021).106 Results from this study suggest that Chapter 8. Conclusions 57 increasing the direct taxes on non-wage income is better than increasing indirect taxes, as the former are less distortionary and introduces smaller allocative inefficiencies than indirect taxes. It further suggests that a broader tax base will yield a higher positive impact on economic growth. For instance, spreading the burden of financing UHC throughout the economy through increased VAT for all commodities has better implica- tions for economic growth than payroll tax paid by employers on behalf of employees. Notwithstanding, the report buttresses that the choice of tax policy should also consider the potential to raise significant funding to finance UHC. In the Armenian context, any decision by the MoF to increase funding for the health sec- tor will also likely reflect compatibility of revenue-raising options with the Government’s tax reform strategy and debt management framework, the degree of consensus over the objectives and design of the reforms to be supported through increased funding, and the degree to which additional spending can be efficiently spent within the health sector. Therefore, the MoH has taken a leadership role in building consensus for the reform agenda through public consultations and negotiations within the Government. Following feedback on the initial reform draft, the MoH has also commissioned a range of studies to demonstrate the potential returns to UHC investments. As noted above, some of these studies have been supported by the World Bank. Whereas fiscal policy is within the decision space of the MoF, a range of improvements in efficiency spending are within the span of control of the MoH. Theoretically, improving the efficiency of spending on health should strengthen negotiations between MoH and MoF on raising sectoral allocations to health, by demonstrating the sectors absorptive capacity. Hence, the team has identified options for improving efficiency, including scaling up centralized procurement of medical goods through the MoH, investing in stronger primary care systems, and improvements in coordination between primary care and hospitals. However, in practice, there is mixed evidence on whether the resulting cost-savings from implementing efficiency improvements are redeployed within health budgets.107 Furthermore, some efficiency reforms will require a leadership role for the MoF, as is the case for public financial management reforms. Armenia faces crucial decisions. The country is confronted with underutilization of essential health care, that arises from gaps in financial protection, and contributes to the growing burden of NCDs. The evidence from global experience on expanding coverage for essential care is clear. Successful countries have mobilized pre-paid revenue via predominantly public sources, mobilized through compulsory contributions, with subsidies for the poorest. This report, and other technical assistance from the World Bank, provides detailed analysis of the potential cost of expanding coverage, options for raising domestic revenue for UHC, implications of financing UHC reforms for macroeconomic outcomes, and strategies to ensure value for money in the health sector. The Government will ultimately decide whether to guarantee access to essential health care for every citizen and the financing options that are compatible with broader macroeconomic objectives. By shaping the health of the population and productivity of the workforce, these decisions will have significant implications for the decades to come. 58 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA APPENDIX APPENDIX 1A: Elasticity of public spending on health to nominal GDP LOG TOTAL GOVT SPENDING ON HEALTH DEPENDENT VARIABLE (WITHOUT DONOR FUNDS) (1) (2) (3) 1.17*** 1.21* 1.21*** Log Nominal GDP (0.06) (0.60) (0.05) -0.006*** Govt. Gross Debt (% of GDP) (0.02) 0.742 Log % Population >65 years (1.85) -1.91 Log % Population <5 years (2.04) -15.82 Log Average Life Expectancy (10.55) -0.88 Log Infant Mortality Rate (1.11) -9.13*** 58.73 -10.04*** Constant (1.63) (50.34) (1.51) No. of Observations 19 16 19 Adj. R-squared 0.959 0.973 0.971 Notes: This is a country-level regression for Armenia between 2000 to 2018. The table reports the result for the elasticity of government spending on health (without foreign donor financing) to nominal GDP. Column (2) shows the estimate when controlled for percentage of population over 65 and Under-five years, life expectancy and mortality rate. Column (1) reports the unconditional elasticity (i.e. does not consider government’s debt), while column (3) considers government debt. Log of total government spending on health is cointegrated with the log of GDP. ***, **, * implies significance at the 1, 5, and 10 percentage level. Appendix 59 APPENDIX 1B: Elasticity of public spending on health to real GDP LOG TOTAL GOVT SPENDING ON HEALTH DEPENDENT VARIABLE (WITHOUT DONOR FUNDS) (1) (2) (3) 1.29*** 1.22* 1.32*** Log Real GDP (0.09) (0.64) (0.08) -0.005* Govt. Gross Debt (% of GDP) (0.024) 0.712 Log % Population >65 years (1.94) -0.163 Log % Population <5 years (1.99) -15.82 Log Average Life Expectancy (10.56) -1.04 Log Infant Mortality Rate (0.84) -12.86*** 58.85 -13.59*** Constant (2.64) (50.29) (2.32) No. of Observations 19 16 19 Adj. R-squared 0.942 0.947 0.85 Notes: This is a country-level regression for Armenia between 2000 to 2018. The table reports the result for the elasticity of government spending on health (without foreign donor financing) to real GDP. Column (2) shows the estimate when controlled for percentage of population over 65 and Under-five years, life expectancy and mortality rate. Column (1) reports the unconditional elasticity (i.e. does not consider government’s debt), while column (3) considers government debt. Log of total government spending on health is cointegrated with the log of GDP. ***, **, * implies significance at the 1, 5 and 10 percentage level. 60 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA APPENDIX 1C: Fiscal space from conducive macroeconomic environment ADDITIONAL ANNUAL GOVT CUMULATIVE ADDITIONAL TOTAL GOVT. HEALTH SPENDING ON HEALTH FROM ANNUAL GOVT SPENDING ON REAL GDP EXPENDITURE YEAR GDP GROWTH HEALTH FROM GDP GROWTH GROWTH (IN REAL TERMS, (REAL TERMS, RELATIVE TO (REAL TERMS, REL. TO 2018, MILLIONS AMD) 2018, MILLIONS AMD) MILLIONS AMD) BASELINE SCENARIO 2021 4.60 69,189.84 3,264.22 3,264.22 2022 6.60 74,715.34 8,789.72 12,053.95 2023 6.60 80,682.11 14,756.49 26,810.44 2024 6.60 87,125.38 21,199.76 48,010.20 2025 6.60 94,083.21 28,157.59 76,167.79 2026 6.60 101,596.70 35,671.07 111,838.87 2027 6.60 109,710.20 43,784.58 155,623.45 2028 6.60 118,471.66 52,546.04 208,169.48 2029 6.60 127,932.80 62,007.18 270,176.66 2030 6.60 138,149.50 72,223.88 342,400.53 2031 6.60 149,182.11 83,256.49 425,657.03 DOWNSIDE SCENARIO 2021 3.10 66,575.14 649.52 649.52 2022 5.90 71,327.95 5,402.33 6,051.85 2023 5.90 76,420.05 10,494.43 16,546.27 2024 5.90 81,875.68 15,950.06 32,496.33 2025 5.90 87,720.79 21,795.17 54,291.50 2026 5.90 93,983.18 28,057.56 82,349.06 2027 5.90 100,692.64 34,767.02 117,116.09 2028 5.90 107,881.09 41,955.47 159,071.56 2029 5.90 115,582.73 49,657.11 208,728.66 2030 5.90 123,834.18 57,908.56 266,637.22 2031 5.90 132,674.70 66,749.09 333,386.31 Appendix 61 APPENDIX 1D: Observed vs. projected government spending on health (2012-2017) ACTUAL PROJECTED DIFFERENCE NOMINAL GDP YEAR (BILLIONS AMD) (BILLIONS AMD) (BILLIONS AMD) (BILLIONS AMD) 2012 67.33 68.89 1.56 4,266 2013 62.67 71.73 9.06 4,556 2014 72.69 74.96 2.27 4,829 2015 81.21 77.96 -3.25 5,044 2016 82.97 78.15 -4.82 5,067 2017 76.37 85.48 9.11 5,564 Total 443.24 457.17 13.93 29,326 62 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA APPENDIX 2A: Elasticity estimation by Deaton’s LA/AIDS model Consider a group of products, say 5 types of alcoholic beverages, beer, red wine, vodka, whiskey, vermouth wine, denoted by , , , , . Here, number of products () is 5 (i.e. =5). To estimate the elasticity, the following variables are required: • is the expenditure share of the h good in the group of products where {, , , , } (this can be constructed from the household data) • � is the nominal price of the h good (can be constructed from the household data) • is log of the total expenditure for each product group • is the random or error term; and • is the log of price index The price index here will be computed as (i.e. Stone Price Index) (1) Note: Some papers use because of simultaneity problems in the above specification. Next, estimate a system of equation for each product group. In this example—alcoholic beverages, the form of the estimating equation is: (2) Specifically, there will be 5 set of equations that will be estimated jointly using seemingly unrelated regression model (SUR). The system of equations takes the form: (3) (4) (5) (6) (7) Appendix 63 Having obtained the coefficients, computing the elasticities is straightforward. There are two types of elasticities that can be computed namely the compensated elasticity and uncompensated elasticities. This can be done as follows: 1. Compensated elasticities (): - own price elasticity 2. Uncompensated elasticities – own price elasticity For cross-price elasticities, for each product pair within the same product group (say {, }), this will be: 1. Compensated elasticities (): - cross price elasticity 2. Uncompensated elasticities cross price elasticity 64 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA APPENDIX 2B: Product-level summary of revenue from excise tax PROJECTED PROJECTED PROJECTED PROJECTED PROJECTED PROJECT REVENUE REVENUE REVENUE REVENUE REVENUE REVENU DESCRIPTION FROM EXCISE FROM EXCISE FROM EXCISE FROM EXCISE FROM EXCISE FROM EX TAX IN 2020 TAX IN 2021 TAX IN 2022 TAX IN 2023 TAX IN 2024 TAX IN 2 Beer 1.93E+09 1.99E+09 2.06E+09 2.13E+09 2.14E+09 2.15E+0 Grape wines 2.82E+08 2.69E+08 2.56E+08 2.44E+08 2.26E+08 2.09E+ Vermouth and other 3.00E+07 2.79E+07 2.60E+07 2.42E+07 2.20E+07 2.00E+ grape wines Other fermented beverages 3.02E+07 3.25E+07 3.51E+07 3.78E+07 3.97E+07 4.16E+ Ethyl alcohol 1.07E+09 1.45E+09 1.84E+09 2.09E+09 2.36E+09 2.66E+ Alcoholic beverages 4.00E+09 4.79E+09 5.68E+09 6.65E+09 6.06E+09 5.52E+ Brandy 6.48E+08 5.72E+08 5.04E+08 4.43E+08 3.87E+08 3.38E+ Whiskey, rum and other 4.60E+08 4.71E+08 4.82E+08 4.93E+08 4.95E+08 4.96E+ alcoholic beverages Vodka 6.01E+09 6.72E+09 7.51E+09 8.38E+09 9.14E+09 9.98E+ Industrial tobacco 7.75E+08 1.18E+09 1.80E+09 2.75E+09 4.08E+09 6.06E+ substitutes Gasoline 2.67E+08 1.97E+08 1.45E+08 1.06E+08 7.61E+07 5.45E+ Diesel 2.47E+08 2.18E+08 1.93E+08 1.70E+08 1.47E+08 1.26E+0 Lubricants 3.71E+09 3.56E+09 3.42E+09 3.27E+09 3.07E+09 2.87E+ Petroleum gases and other 1.98E+07 2.14E+07 2.32E+07 2.51E+07 2.65E+07 2.79E+ similar hydrocarbons Compressed natural gas 1.99E+10 2.37E+10 2.80E+10 3.31E+10 3.87E+10 4.52E+ Cigarettes 6.48E+10 8.46E+10 1.07E+11 1.31E+11 1.58E+11 1.90E+ Condensed milk, including sugar and cacao 2.06E+04 2.26E+04 2.47E+04 2.70E+04 2.88E+04 3.07E+ supplements Lemonade 9.56E+04 9.51E+04 9.45E+04 9.39E+04 9.11E+04 8.84E+ Natural fruit juices 8.31E+04 9.51E+04 1.09E+05 1.24E+05 1.39E+05 1.55E+0 Other non alcoholic drinks (colas. Coca-Cola, Pepsi, 4.97E+04 5.27E+04 5.58E+04 5.90E+04 6.19E+04 6.48E+ etc.) Total Revenue 1.04E+11 1.30E+11 1.59E+11 1.91E+11 2.25E+11 2.66E+ Appendix 65 OJECTED PROJECTED PROJECTED PROJECTED PROJECTED PROJECTED PROJECTED PROJECTED VENUE REVENUE REVENUE REVENUE REVENUE REVENUE REVENUE REVENUE M EXCISE FROM EXCISE FROM EXCISE FROM EXCISE FROM EXCISE FROM EXCISE FROM EXCISE FROM EXCISE IN 2024 TAX IN 2025 TAX IN 2026 TAX IN 2027 TAX IN 2028 TAX IN 2029 TAX IN 2030 TAX IN 2031 4E+09 2.15E+09 2.17E+09 2.18E+09 2.19E+09 2.21E+09 2.22E+09 2.23E+09 6E+08 2.09E+08 1.94E+08 1.80E+08 1.66E+08 1.54E+08 1.43E+08 1.32E+08 0E+07 2.00E+07 1.82E+07 1.65E+07 1.50E+07 1.37E+07 1.24E+07 1.13E+07 97E+07 4.16E+07 4.37E+07 4.59E+07 4.81E+07 5.05E+07 5.30E+07 5.57E+07 6E+09 2.66E+09 2.99E+09 3.37E+09 3.80E+09 4.28E+09 4.82E+09 5.44E+09 06E+09 5.52E+09 5.03E+09 4.58E+09 4.18E+09 3.81E+09 3.47E+09 3.16E+09 87E+08 3.38E+08 2.95E+08 2.58E+08 2.25E+08 1.96E+08 1.72E+08 1.50E+08 95E+08 4.96E+08 4.98E+08 4.99E+08 5.01E+08 5.03E+08 5.04E+08 5.06E+08 4E+09 9.98E+09 1.09E+10 1.19E+10 1.29E+10 1.41E+10 1.54E+10 1.68E+10 08E+09 6.06E+09 8.99E+09 1.34E+10 1.98E+10 2.94E+10 4.37E+10 6.48E+10 61E+07 5.45E+07 3.90E+07 2.80E+07 2.00E+07 1.43E+07 1.03E+07 7.36E+06 7E+08 1.26E+08 1.08E+08 9.30E+07 7.99E+07 6.87E+07 5.90E+07 5.07E+07 07E+09 2.87E+09 2.69E+09 2.52E+09 2.36E+09 2.22E+09 2.08E+09 1.95E+09 65E+07 2.79E+07 2.93E+07 3.09E+07 3.25E+07 3.42E+07 3.60E+07 3.79E+07 87E+10 4.52E+10 5.28E+10 6.16E+10 7.19E+10 8.40E+10 9.80E+10 1.14E+11 58E+11 1.90E+11 2.29E+11 2.75E+11 3.31E+11 3.99E+11 4.80E+11 5.78E+11 8E+04 3.07E+04 3.28E+04 3.49E+04 3.72E+04 3.97E+04 4.23E+04 4.51E+04 1E+04 8.84E+04 8.58E+04 8.33E+04 8.08E+04 7.85E+04 7.62E+04 7.39E+04 9E+05 1.55E+05 1.73E+05 1.94E+05 2.17E+05 2.42E+05 2.71E+05 3.03E+05 9E+04 6.48E+04 6.79E+04 7.12E+04 7.46E+04 7.82E+04 8.19E+04 8.58E+04 25E+11 2.66E+11 3.15E+11 3.76E+11 4.50E+11 5.40E+11 6.51E+11 7.88E+11 66 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA APPENDIX 3A: Regression of total health expenditure on foreign transfers DEPENDENT VARIABLE LOG TOTAL SPENDING ON HEALTH NET OF FOREIGN TRANSFERS (1) (2) (3) (4) Log Foreign Transfers 0.180*** 0.118** to Govt. (0.055) (0.044) Log of Direct Foreign -0.248** -0.135 Transfers (0.090) (0.096) -0.303*** Log Total Foreign Transfers (0.099) 1.208*** 1.525*** 1.318*** 1.569*** Log Nominal GDP (0.146) (0.095) (0.120) (0.090) -23.512*** -29.281*** -25.060*** -30.006*** Constant (3.924) (3.115) (3.355) (2.962) Number of Observations 18 18 18 18 R-Squared 0.965 0.963 0.968 0.965 APPENDIX 3B: Regression of total government health expenditure on foreign transfer DEPENDENT VARIABLE LOG TOTAL GOVERNMENT SPENDING ON HEALTH (1) (2) (3) (4) Log Foreign Transfers -0.057 0.022 to Govt. (0.042 (0.037) Log of Direct Foreign 0.159** 0.180** Transfers (0.047) (0.068) 0.187*** Log Total Foreign Transfers (0.050) 1.271*** 1.213*** 1.175*** 1.186*** Log Nominal GDP (0.095) (0.045) (0.079) (0.044) -6.547*** -25.622*** -24.841*** -25.115*** Constant (1.118) (1.309) (3.355) (1.252) Number of Observations 18 18 18 18 R-Squared 0.978 0.984 0.968 0.984 Appendix 67 APPENDIX 3C: Linear fit of total government health expenditure (net of foreign transfers, Millions AMD) 54,000 Fitted Total Govt. Health Expenditure Net Foreign Transfers 52,000 50,000 48,000 46,000 5,000 10,000 15,000 20,000 Total Foreign Transfers for health in Armenia Source: Authors APPENDIX 3D: Regression of out-of-pocket health expenditure on foreign transfer DEPENDENT VARIABLE LOG HOUSEHOLD OUT-OF-POCKET PAYMENT (1) (2) (3) (4) Log Foreign Transfers 0.245*** 0.140** to Govt. (0.081) (0.060) Log of Direct Foreign -0.353** -0.218 Transfers (0.118) (0.124) -0.433*** Log Total Foreign Transfers (0.129) 1.155*** 1.626*** 1.378*** 1.688*** Log Nominal GDP (0.197) (0.125) (0.170) (0.121) -5.451** -31.476*** -26.429*** -33.499*** Constant (2.339) (4.024) (4.615) (3.828) Number of Observations 18 18 18 18 R-Squared 0.965 0.947 0.952 0.965 68 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA APPENDIX 3E: Linear fit of total out-of-pocket health expenditure (net of foreign transfers, Millions AMD 300,000 Fitted Out-of-Pocket Health Expenditure Net Foreign Transfers 250,000 200,000 150,000 100,000 5,000 10,000 15,000 20,000 Total Foreign Transfers for health in Armenia Source: Authors APPENDIX 4A: Armenia and comparator countries COUNTRY 2-LETTER ABBREVIATION Armenia AR Belarus BY Croatia HR Estonia EE Georgia GE Hungary HU Moldova MD Russia RU Kazakhstan KZ Kyrgyzstan KG Tajikistan TJ Turkey TR Turkmenistan TM Ukraine UA Uzbekistan UZ Source: https://www.iban.com/country-codes Appendix 69 APPENDIX 4B: List and description of variables VARIABLE DESCRIPTIONS Wbname Country name year Years Life expectancy at birth (years): This is the length of time in years that a newborn Life_who can expect to live on average if current death rates do not change. Mortality rate, Under-five (per 1,000): The under-five mortality rate is the proba- bility of a child born in a specified year, dying before reaching the age of five, if subject to current age-specific mortality rates, expressed as rate per 1,000 live births. This rate was transformed as: under5 1000 mortalityunder5 = mortalityunder5 Alcohol Total alcohol consumption per capita (liters of pure alcohol, projected estimate) GDP pc PPP (current international dollars) -World Bank - https://data.worldbank. gdppc_ppp_WB org/indicator/NY.GDP.PCAP.PP.CD?end=2016&start=1990 Healthy life expectancy (HALE) at birth (years)-(WHO): This is the average num- hale_who ber of years that a person can expect to live in full health by accounting for years lived in less than full health due to disease or injury. hale_ihme Health-Adjusted Life Expectancy, Years (IHME): This is defined as above. the_per_cap_ihme Total Spending on Health per Person (2018 USD) ghes_per_cap_ihme Government Spending on Health per Person (2018 USD) Prevalence of current tobacco use by country. Collected from WHO - http://apps. who.int/gho/data/view.main.GSWCAH20v Data missing for Tajikistan and Turkmenistan. For Tajikistan, data was imputed from a world bank report: http://documents. tobacco_2016 worldbank.org/curated/en/357221561130314918/pdf/Tajikistan-Overview-of-Tobac- co-Use-Tobacco-Control-Legislation-and-Taxation.pdf For Turkmenistan, data was imputed from: http://www.euro.who.int/__data/as- sets/pdf_file/0010/337438/Tobacco-Control-Fact-Sheet-Turkmenistan.pdf?ua=1 Amenable Death in 2015 per 100,000 - From http://ghdx.healthdata.org/record/ ihme-data/gbd-2015-healthcare-access-and-quality-index-1990-2015 Amenable_ per_100k_2015 100000 Amenable — per — 100k — inv = Amenable — per — 100k — inv abrev 2-digit Country abbreviation hale_60 Healthy life-expectancy at 60 life_60_who Life-expectancy at 60 overweight_who Prevalence of overweight among adults, BMI >= 25 (age-standardized) (%) Mortality rate, adult, (per 1,000 adults). Transformed as: mortality_adult 1000 mortalityadult = mortalityadult 70 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA APPENDIX 4C: Health outcomes and inputs for Armenia and comparator countries AR BY HR EE GE HU KZ HEALTH OUTCOMES Life expectancy at birth 74.84 74.16 78.29 77.76 72.58 75.98 71.14 (years) Healthy life expectancy at 66.30 65.45 68.98 68.19 64.92 66.85 63.36 birth (years) Health-Adjusted Life 65.60 63.85 67.21 66.89 63.82 65.48 62.50 Expectancy (years) Life expectancy at 60 (years) 19.60 19.60 21.50 22.20 18.30 20.10 17.70 Healthy life-expectancy at 15.10 15.10 16.30 16.90 14.60 15.10 13.90 60 (years) Mortality rate, adult, 116.00 161.00 88.00 119.00 160.00 126.00 181.00 (per 1,000 adults) Mortality rate, under-five 13.70 3.80 4.80 3.00 10.20 4.80 11.00 (per 1,000) Amenable Death in 2015 181.20 247.50 104.90 112.50 246.90 113.60 280.00 per 100,000 HEALTH INPUTS Total Spending on Health 365 354 939 1392 319 1029 295 per person (2018 USD) Govt. Spending on Health 58 216 729 1051 109 679 181 per person (2018 USD) ECONOMIC AND LIFESTYLE VARIABLES Gross Domestic Product, PPP 8808.57 18098.42 24511.89 30913.21 10480.97 27171.16 25314.87 Prevalence of current 26.80 28.30 37.10 31.90 30.40 30.80 25.10 tobacco use Total alcohol consumption 5.50 11.20 8.90 11.60 9.80 11.40 7.70 per capita (liters) Prevalence over weight among adults, BMI>=25 (age - 54.40 59.40 59.60 55.80 54.20 61.60 53.60 standardized) (%) Appendix 71 HU KZ KG MD RU TJ TR TM UA UZ HEALTH OUTCOMES 75.98 71.14 71.40 71.50 71.87 70.76 76.38 68.16 72.50 72.33 66.85 63.36 63.54 63.57 63.46 63.47 66.02 61.38 64.03 64.54 65.48 62.50 63.28 62.63 61.91 61.20 67.62 61.53 60.30 61.39 20.10 17.70 18.00 17.60 19.40 17.70 21.20 17.40 19.10 18.40 15.10 13.90 14.20 13.70 14.90 14.30 15.50 13.90 14.70 14.50 126.00 181.00 162.00 167.00 203.00 123.00 104.00 191.00 180.00 131.00 4.80 11.00 21.10 16.20 8.00 36.90 11.90 49.00 9.20 24.20 113.60 280.00 340.00 205.30 214.70 313.60 107.40 402.80 250.50 326.00 HEALTH INPUTS 1029 295 79 204 574 53 445 511 171 76 679 181 32 102 333 15 346 108 74 36 ECONOMIC AND LIFESTYLE VARIABLES 27171.16 25314.87 3567.81 6424.37 24072.28 3006.51 26150.40 16895.48 8289.71 7723.94 30.80 25.10 27.10 25.30 40.90 6.30 27.60 9.40 30.50 13.00 11.40 7.70 6.20 15.20 11.70 3.30 2.00 5.40 8.60 2.70 61.60 53.60 48.30 51.80 57.10 45.30 66.80 51.80 58.40 48.20 72 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA APPENDIX 4D: Analytical description of DEA programming and savings estimation Suppose there are inputs and outputs for decision-making units (DMUs) – here DMUs are the countries in our dataset. For the – h DMU, is a vector of outputs, while is a vector of inputs. The output-oriented DEA model is specified as the following mathematical programming problem for the – h DMU: Where is an m by 1 vector of output weights and υ is a by 1 vector of input weights. This optimization involves finding weights ( and υ) such that output of the – h DMU () is maximized subject to the ratio of weighted-output to weighted-input less than or equal to 1 ('/υ' ≤ 1) for all DMU’s in our dataset. Using the duality in linear programming, an equivalent minimization problem is derived in the form below: Where is the k by n vector of inputs and is the m by n vector of outputs. is a scalar and is a n by 1 vector of constants. The value of obtained is the efficiency score of the – h DMU and satisfies ≤ 1, with the value 1 indicating a point on the frontier, thus the most efficient country. This programming problem is solved n times, once for each country in our dataset. A value implies that country (i.e., – h DMU) can reduce its inputs by 1 – without changing its output. The model presented above is the constant return-to-scale (CRS) model that assumes that all DMUs operate at an optimal level. But this may not always be the case. If all DMUs do not operate at the optimal scale, the technical efficiency scores obtained using the CRS DEA model might be obscured due to scale efficiencies. Several authors suggest the variable returns-to-scale (VRS) DEA model - a slight modification of the CRS model - which eliminates these scale efficiencies if it exists, and is equivalent to the CRS model if it does not exist.108 To compute potential savings of improving efficiency, authors relied on the work by De Cos and Moral-Benito (2011), documented in Medeiros and Schwierz (2015). Potential savings in this context is derived by exploring the counterfactual scenario of how much Appendix 73 in nominal terms could expenditure on health care be decreased in a country (DMUs) if it adopts the most efficient system, while keeping the same health outcome? A cross- sectional regression of the form is estimated below: Where of country is the health outcome. is the estimated efficiency level in the production of the health outcome ; is per capita health care expenditure; and Z is a matrix with other socioeconomic characteristics that can affect life expectancy. The equation is estimated in logarithmic form. Potential savings in health care expenditure to GDP ratio is computed using the estimated coefficients from the regression, as shown below: Where ∂1 and ∂2 are the estimated coefficients for the efficiency and expenditure variables, respectively. 74 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA APPENDIX 4E: Potential gains in outputs with efficiency improvements HEALTH INPUTS HEALTH INPUTS + ECONOMIC AND LIFESTYLE VARIABLES 1. Total Spending on 1. Government Spend- Health per Per- ing on Health per son (THS) Person (GHS) Total Government 2. GDP per capita 2. GDP per capita OUTPUT Spending on Spending on in PPP in PPP Health per Health per 3. % Overweight 3. % Overweight Person Person population (i.e. with population (i.e. with (2018 USD) (2018 USD) BMI>=25) BMI>=25) 4. Total alcohol con- 4. Total alcohol con- sumption per capita sumption per capita 5. Tobacco 2016 5. Tobacco 2016 Life Expectancy 1.13 1.31 1.14 1.31 M1 At Birth Healthy Life Ex- 0.31 1.09 0.87 1.07 M2 pectancy At Birth Life Expectancy 1.32 0.61 0.55 0.61 M3 At 60 Healthy Life Ex- 0.29 0.29 0.28 0.35 M4 pectancy At 60 Health-Adjusted 1.73 1.42 1.22 1.41 M5 Life Expectancy Amenable Mor- 59.07 16.47 13.83 17.05 M6 tality Rate, % Mortality Rate 14.76 9.82 18.10 12.10 M7 (Adult), % Mortality Rate 348.90 33.82 57.06 22.58 M8 (Under-five), % Appendix 75 APPENDIX 4F: Efficiency scores and potential gains in health system performance MODEL 1 OUTCOME: LIFE EXPECTANCY AT BIRTH INPUT: TOTAL SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains 12 Potential Gains in Life Expectancy at Birth 10 0.8 9.12 8 0.6 TJ UZ Efficiency TR HR 5.65 6 AR KG 0.4 BY 3.99 4 EE UA 3.14 2.69 2.82 HU 1.98 0.2 GE MD 2 1.67 1.83 1 1.14 1.22 1.38 KZ 0.75 0.96 RU TM 0 TR HR UZ AR EE UA BY KG TJ MD GE HU KZ RU TM Output: Life Expectancy at Birth Input: Total Health Spending per Person Countries MODEL 2 OUTCOME: LIFE EXPECTANCY AT BIRTH INPUT: GOVERNMENT SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains 12 Potential Gains in Life Expectancy at Birth 10 0.8 8 7.41 0.6 Efficiency TJ TR 6 AR HR UZ 4.75 4.91 0.4 KG 4.09 4 3.3 EE 2.99 3 UA 1.87 1.92 1.99 0.2 HU 2 1.07 1.27 1.31 1.4 GE BY MD 0.48 KZ TM RU 0 TR HR UZ AR EE KG BY TJ GE HU UA MD KZ RU TM Output: Life Expectancy at Birth Input: Government Health Spending per Person Countries 76 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA MODEL 3 OUTCOME: HEALTHY LIFE EXPECTANCY AT BIRTH INPUT: TOTAL SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains 1.0 12 Potential Gains in Life Expectancy at Birth 10 0.8 8 0.6 TJ AR Efficiency UZ 5.93 6 HR TR 0.4 KG BY 4.18 4 EE 2.79 2.8 GE HU 2.08 0.2 UA 1.67 1.31 1.38 1.4 1.46 2 MD 0.75 0.8 0.94 1.07 KZ 0.31 RU TM 0 AR UZ HR TR BY EE GE TJ UA KG MD HU KZ RU TM Output: Healthy Life Expectancy at Birth Input: Total Health Spending per Person Countries MODEL 4 OUTCOME: HEALTHY LIFE EXPECTANCY AT BIRTH INPUT: GOVERNMENT SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains Potential Gains in Healthy Life Expectancy at Birth 12 10 0.8 8 Efficiency 0.6 TJ 5.56 6 AR HR UZ 4.35 0.4 3.8 4 3.38 3.08 KG 2.75 EE 1.84 1.84 1.88 2.02 0.2 1.56 1.71 2 GE UA HU BY MD 0.93 1.08 1.08 TR TM KZ 0.0 RU 0 HR AR UZ EE TJ BY TR KG GE HU UA MD KZ RU TM Output: Healthy Life Expectancy at Birth Input: Government Health Spending per Person Countries Appendix 77 MODEL 5 OUTCOME: LIFE-EXPECTANCY AT AGE 60 INPUT: TOTAL SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains 1.0 12 Potential Gains in Life Expectancy at age 60 10 0.8 8 0.6 TJ UA UZ Efficiency TR 6 EE KG 0.4 HR 4.31 AR BY 4 2.63 2.05 2.17 2.22 2.34 0.2 RU 2 HU 1.22 1.33 MD 0.93 1.02 GE 0.56 0.57 0.64 0.69 KZ 0.33 TM 0 UA TR UZ HR EE KG TJ BY AR MD HU GE RU KZ TM Output: Life Expectancy at Age 60 Input: Total Health Spending per Person Countries MODEL 6 OUTCOME: LIFE-EXPECTANCY AT AGE 60 INPUT: GOVERNMENT SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains 1.0 12 Potential Gains in Life Expectancy at age 60 10 0.8 8 0.6 TJ Efficiency AR TR UZ 6 EE 0.4 HR KG UA 4 2.83 2.5 2.71 2.16 1.82 1.97 0.2 2 GE 1 1.15 BY 0.88 0.97 HU MD RU TM 0.47 0.6 0.63 0.66 0.67 KZ 0 TR AR UZ HR EE KG TJ UA BY GE HU RU MD TM KZ Output: Life Expectancy at Age 60 Input: Government Health Spending per Person Countries 78 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA MODEL 7 OUTCOME: HEALTHY LIFE-EXPECTANCY AT AGE 60 INPUT: TOTAL SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains Potential Gains in Healthy Life Expectancy at age 60 12 10 0.8 8 Efficiency 0.6 TJ TR UZ UA 6 HR AR BY EE 0.4 KG 4 RU 1.85 0.2 GE 1.29 1.56 2 HU MD 0.95 1.26 0.46 0.53 0.53 0.66 KZ 0.16 0.18 0.22 0.24 0.26 0.3 TM 0 TR UA HR UZ BY AR EE KG TJ GE RU MD KZ HU TM Output: Healthy Life Expectancy at Age 60 Input: Total Health Spending per Person Countries MODEL 8 OUTCOME: HEALTHY LIFE-EXPECTANCY AT AGE 60 INPUT: GOVERNMENT SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains Potential Gains in Healthy Life Expectancy at age 60 1.0 12 10 0.8 8 0.6 Efficiency AR TJ 6 HR EE TR 0.4 UZ 4 UA KG 0.2 1.6 1.31 1.4 1.51 2 GE 0.7 0.79 0.4 0.49 0.52 0.63 0.65 BY MD RU 0.22 0.22 0.29 0.37 TM HU KZ 0 TR HR AR BY UZ EE TJ UA KG GE RU HU TM KZ MD Output: Healthy Life Expectancy at Age 60 Input: Government Health Spending per Person Countries Appendix 79 MODEL 9 OUTCOME: HEALTH-ADJUSTED LIFE EXPECTANCY INPUT: TOTAL SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains Potential Gains in Health-Adjusted Life Expectancy 12 10 0.8 8 6.93 6.44 0.6 Efficiency TJ KG 6 TR 4.78 AR UZ 0.4 3.89 4 3.3 2.65 2.75 2.85 2.86 2.09 HR GE 1.72 0.2 BY 2 MD 1.21 1.23 UA 0.94 1.06 EE HU KZ RU TM 0 HR TR EE KG AR TJ HU MD UZ GE BY KZ UA RU TM Output: Health-Adjusted Life Expectancy Input: Total Health Spending per Person Countries MODEL 10 OUTCOME: HEALTH-ADJUSTED LIFE-EXPECTANCY INPUT: GOVERNMENT SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains Potential Gains in Health-Adjusted Life Expectancy 12 10 0.8 8 6.39 6.54 0.6 Efficiency KG TJ 6 AR TR 4.99 4.53 0.4 3.89 3.31 3.47 4 2.7 2.81 UZ 2.17 0.2 HR GE 1.32 1.43 2 MD 0.98 1.05 1.18 EE BY KZ TM UA HU RU 0 TR HR KG EE AR TJ GE HU UZ BY MD KZ TM UA RU Output: Health-Adjusted Life Expectancy Input: Government Health Spending per Person Countries 80 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA MODEL 11 OUTCOME: AMENABLE MORTALITY RATE INPUT: TOTAL SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains Potential Improvements in Amenable Mortality rate, % 350 316.95 300 0.8 250 200 Efficiency 0.6 TJ TR HR MD 150 120.41 0.4 KG UA UZ 108.32 111.32 100 AR 94.28 GE HU 58.89 0.2 50 BY 35.31 EE KZ RU 27.44 29.11 34.72 14.96 16.74 21.31 8.22 13.72 TM 0 HR TR EE HU MD TJ UZ UA KG AR GE KZ BY RU TM Output: Amenable Mortality rate (%) Input: Total Health Spending per Person Countries MODEL 12 OUTCOME: AMENABLE MORTALITY RATE INPUT: GOVERNMENT SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains Potential Improvements in Amenable Mortality rate, % 200 169.27 150 0.8 124.47 117.53 106.5 Efficiency 0.6 TJ 100 AR HR TR 0.4 59.59 60.08 60.98 65.36 50 HU KG MD 36 UA UZ 0.2 27.47 EE GE 16.31 17.34 19.59 10.51 15.25 BY KZ RU TM 0 HR TR AR EE HU TJ MD KG UZ UA GE BY KZ RU TM Output: Amenable Mortality rate (%) Input: Government Health Spending per Person Countries Appendix 81 MODEL 13 OUTCOME: MORTALITY RATE (ADULTS) INPUT: TOTAL SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains 200 Potential Improvements in Mortality Rate (Adults), % 0.8 150 121.2 0.6 Efficiency 102.01 100 HR TJ TR 74.26 0.4 KG UZ 67.5 61.64 55.37 56.42 58.55 50.5 51.26 50 AR 0.2 MD UA 22.29 25.14 KZ 13.58 14.78 BY GE 6.83 EE RU TM 0.0 HU 0 TR HR AR UZ TJ KG EE GE MD BY HU UA KZ TM RU Output: Mortality Rate (Adults) Input: Total Health Spending per Person Countries MODEL 14 OUTCOME: MORTALITY RATE (ADULTS) INPUT: GOVERNMENT SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains 200 Potential Improvements in Mortality Rate (Adults), % 0.8 150 0.6 111.57 Efficiency AR HR TJ 100 TR 0.4 79 74.8 68.87 56.4 57.55 58.07 KG 49.98 51.37 52.92 50 0.2 UZ 22.86 24.71 UA 13.06 GE MD TM 9.14 9.7 BY KZ 0.0 EE HU RU 0 TR AR HR UZ TJ GE EE KG MD HU BY UA KZ TM RU Output: Mortality Rate (Adults) Input: Government Health Spending per Person Countries 82 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA MODEL 15 OUTCOME: MORTALITY RATE (UNDER-FIVE) INPUT: TOTAL SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains 2000 Potential Improvements in Mortality Rate (Under 5), % 1.0 1497.79 1500 0.8 KG Efficiency 0.6 UA 1000 TJ UZ BY EE 0.4 500 GE KZ 349.39 286.65 0.2 MD AR RU 194.65 HR HU TR 163.83 180.47 189.4 74.27 78.84 24.15 24.88 25.26 30.71 36.41 46.78 TM 0 KG EE BY UA UZ TJ HR HU RU MD KZ GE TR AR TM Output: Mortality Rate (Under 5) Input: Total Health Spending per Person Countries MODEL 16 OUTCOME: MORTALITY RATE (UNDER-FIVE) INPUT: GOVERNMENT SPENDING ON HEALTH PER PERSON Bias-corrected DEA and Confidence Interval (95% CI) Potential Gains 800 Potential Improvements in Mortality Rate (Under 5), % 696.06 600 0.8 UA Efficiency 0.6 KG 400 AR TJ BY EE UZ 272.77 0.4 GE 191.56 200 MD 150.52 151.39 0.2 KZ RU 54.59 64.5 67.05 67.3 HR HU 33.69 36.17 TR 13.66 18.32 22.15 22.88 TM 0 UA EE BY KG AR TJ UZ HU GE HR RU MD KZ TR TM Output: Mortality Rate (Under 5) Input: Government Health Spending per Person Countries ENDNOTES 83 ENDNOTES 1  Statistical Yearbook of Armenia. (2020). Accessed at: https://armstat.am/file/article/ nasel_01.04.2020.pdf. 2  Chukwuma, A., Gurazada, S., Jain, M., Tsaturyan, M., Khcheyan, M. (2020). FinHealth Armenia: Reforming Public Financial Management to Improve Health Service Delivery. Washington, DC: World Bank. https://openknowledge.worldbank.org/handle/10986/34747. 3  World Bank Data Team. New country classifications by income level: 2018-2019. https://blogs. worldbank.org/opendata/new-country-classifications-income-level-2018-2019. 4  Fraser, N., Chukwuma, A., Koshkakaryan, M., Yengibaryan, L., Hou, X., Wilkinson, T. D. (2021). Reforming the Basic Benefits Package in Armenia: Modeling Insights from the Health Interventions Prioritization Tool. Washington, DC: World Bank. 5  IMF. (2019). World Economic Outlook. Domestic Resource Mobilization Report. Joint Learning Network for Universal Health Coverage. 6  World Bank. (2020). World Development Indicators. 7  IMF. World Economic Outlook (2019) and IMF World Revenue Longitudinal Data (2019); JLN UHC DRM Report. 8  World Bank. (2020). World Development Indicators. Accessed at: https://data.worldbank.org/ indicator/GC.REV.XGRT.GD.ZS?end=2017&start=1972. 9  World Bank. (2020). Europe and Central Asia Economic Update, Fall 2020: COVID-19 and Human Capital. Washington, DC: World Bank. https://openknowledge.worldbank.org/ handle/10986/34518. 10  World Bank. Europe and Central Asia Economic Update, Fall 2020: COVID-19 and Human Capital. 11  World Bank. World Development Indicators. 12  World Bank. Incidence of tuberculosis (per 100, 000 people). Accessed at: https://data. worldbank.org/indicator/SH.TBS.INCD?locations=AM&view=chart. 13  World Health Organization. Measles and Rubella Surveillance Data. Accessed at: https://www. who.int/immunization/monitoring_surveillance/burden/vpd/surveillance_type/active/ measles_monthlydata/en/. 14  Institute of Health Metrics and Evaluation. Life expectancy at birth, 1990–2100. Forecasted data based on Global Burden of Disease 2017 results. See related publication: https://doi. org/10.1016/S0140-6736(20)30677-2. 15  Institute of Health Metrics and Evaluation. Armenia. Accessed at: http://www.healthdata.org/ armenia. 16  GBD 2019 Diseases and Injuries Collaborators. "Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019." The Lancet 396.10258 (2020): 1204-1222. 17  Andreasyan, D., Bazarchyan, A., Torosyan, A., Sargsyan, S. Prevalence of tobacco smoking in Armenia, STEPs survey. Accessed at: https://doi.org/10.18332/tid/94872. 18  World Bank. Europe and Central Asia Economic Update, Fall 2020: COVID-19 and Human Capital. 19  Kruk, M., Gage, A., Joseph, N.T., Danaei, G., García-Saisó, S., Salomon, J.A. "Mortality due to low-quality health systems in the universal health coverage era: a systematic analysis of amenable deaths in 137 countries." The Lancet 392.10160 (2018): 2203-2212. 84 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA 20  Statistical Committee of the Republic of Armenia. (2019). Social Snapshot and Poverty. 21  World Bank. Out-of-pocket expenditure (% of current health expenditure). Accessed at: https:// data.worldbank.org/indicator/SH.XPD.OOPC.CH.ZS. 22  Chukwuma, A., Cain, J., Tandon, A. (2021). Government Budgetary Spending on Health in Armenia. Domestic Resource Mobilization Collaborative. Joint Learning Network for Universal Health Coverage. 23  Chukwuma, A., Meessen, B., Lylozian, H., Gong, E., Ghazaryan, E. (2020). Strategic Purchasing for Better Health in Armenia. Washington, DC: World Bank. Retrieved on February 2, 2021 from https://openknowledge.worldbank.org/handle/10986/34491. 24  Fraser, N., et al. Reforming the Basic Benefits Package in Armenia: Modeling Insights from the Health Interventions Prioritization Tool. 25  World Health Organization. (2020). Global Health Observatory. 26  Angel-Urdinola, D.F., Jain, F. (2006). Do Subsidized Health Programs in Armenia Increase Utilization among the Poor? Policy Research Working Paper, No. 4017. Washington, DC: World Bank. https://openknowledge.worldbank.org/handle/10986/9277. 27  World Bank. World Development Indicators. 28  World Bank. (2018). Assessing Fiscal Space for Armenia in the Context of Immunization Financing. Unpublished manuscript. 29  Republic of Armenia Ministry of Health. COVID-19 response financial report. Accessed at: https://www.moh.am/#1/3381. 30  Kutzin, J., Yip, W., Cashin, C. "Alternative financing strategies for universal health coverage." World Scientific Handbook of Global Health Economics and Public Policy: Volume 1: The Economics of Health and Health Systems (2016): 267-309. 31  Abdo, Y., Savedoff, W.D., Hsiao, W.C., Kutzin, J., Soucat, S., Tandon, A., Wagstaff, A., Yip, W. “The Case Against Labor-Tax-Financed Social Health Insurance for Low- And Low- Middle-Income Countries.” Health Affairs (Millwood) (2020) 39(5):892-897. https://www. healthaffairs.org/doi/10.1377/hlthaff.2019.00874. 32  Dudu, H., Chukwuma, A., Manookian, A., Aghazaryan, A., Muhammad, Z. (2021). Macroeconomic Effects of Financing Universal Health Coverage in Armenia. World Bank, Washington, DC: World Bank. https://openknowledge.worldbank.org/ handle/10986/35688 33  Kutzin, J., et al. "Alternative financing strategies for universal health coverage." 34  Kutzin, J., et al. "Alternative financing strategies for universal health coverage." 35  Chukwuma, A., et al. Strategic Purchasing for Better Health in Armenia. 36  Chukwuma, A., Lylozian, H., Gong, E. “Challenges and Opportunities for Purchasing High- Quality Health Care: Lessons from Armenia.” Health Systems & Reform (2021) 7:1 e1898186. https://doi.org/10.1080/23288604.2021.1898186. 37  Heller, P. S. Understanding fiscal space. (2005) 5(4). Washington, DC: International Monetary Fund. 38  Heller, P. S. Understanding fiscal space. 39  Tandon, A., Cashin, C. (2010). Assessing public expenditure on health from a fiscal space perspective (English). Health, Nutrition and Population (HNP) discussion paper. Washington, D.C.: World Bank Group. 40  Tandon, A., et al. Assessing public expenditure on health from a fiscal space perspective (English). ENDNOTES 85 41  Heller, P. “The prospects of creating ‘fiscal space’ for the health sector.” Health Policy and Planning (2006) 21(2):75–79. https://academic.oup.com/heapol/ article/21/2/75/554947#8806263. 42  Tandon, A., et al. Assessing public expenditure on health from a fiscal space perspective (English). 43  Kose, M.A., Kurlat, S., Ohnsorge, F., Sugawara, N. (2017). A cross-country database of fiscal space. Policy research working paper 8175. Washington, DC: World Bank. 44  Bustamante, A V., Shimoga, S.V. “Comparing the Elasticity of Income to Health Expenditure in Middle-Income and High-Income Countries: The Role of Financial Protection.” International Journal of Health Policy and Management (2018) 7.3: 255. 45  There are 19 observations in the elasticity estimation due to data availability (2000 to 2018). In addition, there is a trade-off between using longer versus shorter periods if the elasticity is inconsistent over time. Despite these challenges, the standard errors are relatively small, and our estimates were significant at the 1 and 10 percent levels. 46  IMF. Government Finance Statistics, By Country. Accessed at: https://data.imf. org/?sk=89418059-d5c0-4330-8c41-dbc2d8f90f46&sId=1437488721405. 47  This is the COVID-19 adjusted forecast provided by the World Bank and does not reflect the latest adjustments in mid-October 2020. 48  The forecast was done using the “downside” real GDP growth forecast provided by the World Bank: -7.9% in 2020, 3.1% in 2021, 5.9% in 2022, 2023 and 2024. 49  IMF. Government Finance Statistics. Accessed at: https://data.imf.org/?sk=89418059-d5c0- 4330-8c41-dbc2d8f90f46. 50  Chukwuma, A., et al. Strategic Purchasing for Better Health in Armenia. 51  IMF. Government Finance Statistics. Accessed at: https://data.imf.org/?sk=89418059-d5c0- 4330-8c41-dbc2d8f90f46. 52  Coady, D. “Creating fiscal space.” Finance and Development. (December 2018) 55(4). International Monetary Fund. 53  Gaspar, V., Jaramillo, L., Wingender, P. (2016). Tax Capacity and Growth: Is There a Tipping Point? IMF Working Paper 16/234. Washington, DC: International Monetary Fund. 54  Coady, D. “Creating fiscal space.” 55  Ministry of Finance of the Republic of Armenia. (2019) Strategy for the Implementation of Tax Reforms in the Republic Armenia. Draft. 56  Cashin, C., Sparkes, S., Bloom, D. (2017). Earmarking for health: from theory to practice. No. WHO/HIS/HGF/HFWorkingPaper/17.5. Geneva: World Health Organization. 57  Coady, David. “Creating fiscal space.” 58  Pan-American Health Organization. (2021). Health taxes. Accessed at: https://www.paho.org/ en/topics/health-taxes. 59  See https://www.armstat.am/file/article/gner_2019_2.pdf 60  Numbeo is one of the world’s largest databases that provides information on product prices in over 9789 cities, while “sas” is an online grocery shop. Accessed here: https://www. numbeo.com/cost-of-living/country_result.jsp?country=Armenia and https://en.sas.am/. The analysis used the prices for products in “numbeo.” For products, where the prices are not in “numbeo,” the average product price in “sas” was used. SAS displays prices of several brands of each product. Prices were randomly selected from the first two pages of a product and the average computed. This method obviously introduced some bias in the analysis and even makes replication difficult because of prices changes. The direction of bias depends on whether the average price is over- or under- estimated. 86 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA 61  The LA/AIDS approximates the QUAIDS under the assumption of a linear Engel curve, and was chosen over other methods due to its simplicity in estimation. A description of estimating equations is provided in Appendix 2A. 62  Deaton, A., Muellbauer, J. (1980) Economics and Consumer Behavior. Cambridge: Cambridge University Press. doi:10.1017/CBO9780511805653. 63  This includes gasoline, diesel, lubricants, compressed natural gas, petroleum gases, and other similar hydrocarbons. 64  Assuming a product say B, produced domestically and imported denoted as Bd and Bf, let their prices be Pd and Pf, respectively. The weighted average product price is constructed as 65  For compressed gas, there is an abnormal increase in growth rate by 32,042% between 2018 to 2019. This results to a very high average growth rate. A growth rate of 16.76% used was the growth rate of compressed natural gas between 2017 to 2018. 66  Due to the pandemic, this computed average growth rate might be higher than the expected growth rate for 2020. Most existing forecasts, such as done by the IMF and the Central Bank of Armenia shows the normalization of economic activities and even higher growth projections from 2021. This may introduce acknowledge the bias in the analysis. 67  Let PET be the proposed excise tax rate, CET be the applicable excise tax rate in 2019, and P the price of the product. Authors used the equation: 100 to compute the percentage change in prices due to the excise tax. 68  The industry wage data and total number of employees within each industry comes from the State Revenue Commission (SRC). This data includes employees who are formally registered as taxpayers and does not include informal employees. The data for formally registered taxpayers was used here, as it captures the potential revenue from increase in payroll taxes better. 69  Manning, A. (2003). Monopsony in motion: Imperfect competition in labor markets. Princeton University Press. 70  Manning, A. Monopsony in motion: Imperfect competition in labor markets. 71  The actual increase in price is 0.83%. The argument here is that a less than 1% increase in prices, which is below the average inflation rate in Armenia, may be negligible—with no real effect on demand. Authors are agnostic about the accuracy of this assumption. 72  Each firm indicates its own main activity according to the NACE classification when it registers as a taxpayer. 73  The growth forecast in the service sector was provided by the CBA and they are 1.4% in 2020, 7.9% in 2021 and 5.3% in 2022 and other preceding years. 74  The business activity “income from commercial (buying and selling) activities related with secondary raw materials” has a very high year-on-year high growth rate (3266% between 2015 to 2016, 34% between 2016 to 2017, 41% between 2017 to 2018, and 68% between 2018 to 2019), with average year-on-year growth rate of 852%. Applying such growth rate for periods 2020 to 2031, will generate an unrealistic amount of tax revenue, which clearly overstates the fiscal space. To resolve this issue, it was assumed that the growth rate of the service sector applies to this business activity. 75  Authors acknowledge that the turnover is computed from business activity within a firm, but as each firm is classified based on the activity which its main income comes from, this variable is the best approximation of turnover per firm. 76  This is confirmed from the data used for the VAT analysis. In that data, authors noted the total sector turnover and the VAT revenue. The VAT tax base was calculated by dividing the VAT revenue by 0.2. Comparing the VAT base to the turnover, in over 84% of cases the turnover is larger than the VAT base. ENDNOTES 87 77  The commitments in this report were made before the COVID-19 pandemic. 78  A linear regression of the historical log of total current health expenditure (net of foreign transfers) was fitted on the log of total foreign transfers in Armenia. Details are provided in Appendix 3A. 79  Details in Appendix 3B. 80  The linear fit of the regression of the log of total government health expenditure on the log of total foreign transfers is presented in Appendix 3C, evidence against the fungibility of external assistance for health. 81  The linear fit of the regression of the log of total OOP health expenditure on the log of total foreign transfers is presented in Appendix 3E. 82  See Appendix 3D. 83  The DEA is a non-parametric linear programming technique in which deviations between observed values and an estimated production possibility frontier are attributed to inefficiency. 84  Medeiros, J.,Schwierz, C. (2015). Efficiency estimates of health care systems. No. 549. Directorate General for Economic and Financial Affairs (DG ECFIN), European Commission. 85  The variables are listed and described in Appendix 4B; the full dataset is provided in Appendix 4C. 86  Afonso, A., Aubyn, M. “Cross-country efficiency of secondary education provision: A semi- parametric analysis with non-discretionary inputs.” Economic Modelling (2006) 23(3):476- 491. 87  Coelli, T.J., Rao, D.S.P., O'Donnell, C.J., Battese, G.E. (2005). An introduction to efficiency and productivity analysis. Springer Science & Business Media. 88  The approach is also described in Appendix 4D. 89  Model 4 was chosen as the preferred specification as it aligns with the objective of the analysis which is to reflect on the efficiency of government expenditure on health. 90  While endogeneity from reverse causality – health outcomes leading to GDP growth – may be a potential source of bias, existing evidence does not support this notion. For example, Lange and Vollmer (2017) provide a detailed review of existing papers that finds both weak and strong effect of economic growth on health outcomes, while Ashraf and Weil (2008) finds that better health outcomes lead to modest GDP per capita growth— after many years (up to 30 years). 91  Lange, S., Vollmer, S. "The effect of economic development on population health: a review of the empirical evidence." British Medical Bulletin (2017) 121.1: 47-60. 92  Ashraf, Q.H., Lester, A., Weil, D.N. "When does improving health raise GDP?" NBER Macroeconomics Annual (2008) 23.1: 157-204. 93  That is (1-efficiency scores) input can be saved while achieving the observed level of output. 94  Appendix 4E has estimates of the potential improvements in health outcomes that could be attained at the current level of government spending on health, if the system achieved the efficiency of the best-performing comparators. The efficiency scores and potential health gains for all the countries are summarized in Appendix 4F. 95  Conversion rate: USD 1 = AMD 480. 96  Per the United Nations World Population Prospects, in 2019, the total population was 2.9 million of whom 208,000 were under-five. 97  Chukwuma, A., et al. FinHealth Armenia: Reforming Public Financial Management to Improve Health Service Delivery. 88 MORE MONEY FOR HEALTH: RESOURCE MOBILIZATION FOR UNIVERSAL HEALTH COVERAGE IN ARMENIA 98  Chukwuma, A., et al. Strategic Purchasing for Better Health in Armenia. 99  Fraser, N., et al. Reforming the Basic Benefits Package in Armenia: Modeling Insights from the Health Interventions Prioritization Tool. 100  Chukwuma, A., et al. FinHealth Armenia: Reforming Public Financial Management to Improve Health Service Delivery. 101  Fraser, N., et al. Reforming the Basic Benefits Package in Armenia: Modeling Insights from the Health Interventions Prioritization Tool. 102  Davtyan, N., Bazarchyan, A., Aghazaryan, A., Hovhannisyan, L. (2018). National Health Accounts of Armenia. National Institute of Health after S. Avdalbekyan. Ministry of Health of the Republic of Armenia. 103  Eurostat. (2020). Healthcare expenditure by function. Accessed at: https://ec.europa.eu/ eurostat/statistics-explained/index.php/Healthcare_expenditure_statistics#Healthcare_ expenditure_by_function. 104  In 2018, voluntary health insurance constituted 1.2% of current health spending in Armenia. 105  Culyer, A. J., Newhouse, J.P. eds. (2000). Handbook of health economics. Elsevier. 106  Dudu, Hasan, et al. Macroeconomic Effects of Financing Universal Health Coverage in Armenia. 107  Barroy, H., Gupta, S. (2020). From Overall Fiscal Space to Budgetary Space for Health: Connecting Public Financial Management to Resource Mobilization in the Era of COVID-19. 2020. Accessed at: https://www.cgdev.org/publication/overall-fiscal-space-budgetary- space-health-connecting-public-financial-management. 108  See Coelli, Timothy J., et al. An introduction to efficiency and productivity analysis for a full treatment of this extension. ENDNOTES 89