Health Systems & Reform ISSN: 2328-8604 (Print) 2328-8620 (Online) Journal homepage: http://www.tandfonline.com/loi/khsr20 What Are Governments Spending on Health in East and Southern Africa? Moritz Piatti-Fünfkirchen, Magnus Lindelow & Katelyn Yoo To cite this article: Moritz Piatti-Fünfkirchen, Magnus Lindelow & Katelyn Yoo (2018) What Are Governments Spending on Health in East and Southern Africa?, Health Systems & Reform, 4:4, 284-299, DOI: 10.1080/23288604.2018.1510287 To link to this article: https://doi.org/10.1080/23288604.2018.1510287 Published with license by Taylor & Francis Group, LLC© 2018 International Bank for Reconstruction and Development / The World Bank Accepted author version posted online: 25 Sep 2018. Published online: 30 Oct 2018. Submit your article to this journal Article views: 576 View Crossmark data Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=khsr20 Health Systems & Reform, 4(4):284–299, 2018 Published with license by Taylor & Francis Group, LLC ISSN: 2328-8604 print / 2328-8620 online DOI: 10.1080/23288604.2018.1510287 Research Article What Are Governments Spending on Health in East and Southern Africa? Moritz Piatti-Fünfkirchen*, Magnus Lindelow, and Katelyn Yoo Health Nutrition and Population Global Practice, World Bank Group, Washington, DC, USA CONTENTS Abstract—Progress toward universal health care (UHC) in Africa Introduction will require sustained increases in public spending on health and Definitions, Data, and Methods reduced reliance on out-of-pocket financing. This article reviews Muddying the Waters: Issues in Analyzing Government Health trends and patterns of government spending in the East and Expenditures Southern Africa regions and points out methodological challenges Conclusion with interpreting data from the World Health Organization’s (WHO) Global Health Expenditure Database (GHED) and other sources. References Government expenditure for health has increased for most coun- tries, albeit at a slower rate than gross domestic product (GDP). In most countries there has been a prioritization away from health in government budgets, putting the onus on the private sector and donors to fill the gap. Donor support is important in the region but reliance on external spending is not consistent with countries’ stated ambitions of universal health coverage. A number of methodological challenges with estimating health expenditures are identified. Capturing health expenditures ade- quately across agencies and levels of decentralization can be challenging, and off-budget funds and arrears are evasive. Measurement error can be significant because actual expenditure information can be hard to come by and is often dated and unreliable. Furthermore, how external financing is captured will affect government health expenditure estimates. These factors have contributed to differences in expenditure estimates between the WHO GHED and country-specific public expenditure reviews and complicate interpretation. The article concludes that it is critical to strengthen national data capacity and international efforts to promote quality and consistency of data. The GHED is an invaluable resource for monitoring and benchmarking health Keywords: Abuja target, Africa, expenditure measurement, health expendi- expenditures. It is best used in combination with deep dive coun- ture, health financing try expenditure assessments. Received 30 April 2018; revised 18 July 2018; accepted 6 August 2018 *Correspondence to: Moritz Piatti-Fünfkirchen; Email: mpiatti@worldbank. org Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/khsr. INTRODUCTION © 2018 International Bank for Reconstruction and Development / The World Bank This is an Open Access article distributed under the terms of the Creative Commons Progress toward universal health coverage (UHC) and the Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits Sustainable Development Goals (SDGs) will not be unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. achieved without adequate resources. According to a recent estimate, an additional 371 billion USD will be needed per 284 Piatti-Fünfkirchen et al.: What Are Governments Spending on Health 285 year for low-income and middle-income countries to reach that applies across all countries and that factors such as the nature the health-related SDG targets. This would represent an of the health challenges, policy objectives in the health sector, increase in health spending as a share of gross domestic health system efficiency, fiscal capacity, and competing demands product (GDP) from a current mean of 5.6% to a mean of on public resources need to be considered.9 Along similar lines, 7.5%, with significant financing gaps, in particular in low- Jowett et al. make the point that there is wide variation in coverage income countries.1 and outcomes at any level of government health spending, sug- These recent estimates follow in the vein of similar efforts gesting that, although the level of government spending is impor- during the Millennium Development Goals era to cost the tant, other factors also come into play in determining health achievement of targets and identify financing gaps. For exam- system performance.10 ple, the Commission on Macroeconomics and Health devel- Even if there is no “magic” target for government health oped estimates for the cost and financing needs to expand spending, robust data and monitoring of government health coverage of a limited set of priority services.2 More recently, spending are critical. Levels of government spending on health the High-Level Task Force on Innovative Financing for —in absolute terms and as a share of GDP and overall budget— Health Systems estimated that a total of 54 USD per capita indicate government commitment to health. These data can be was needed by 2015 to achieve the health Millennium used to benchmark with peer countries and monitor changes over Development Goals for low-income countries, which implied time. Moreover, in the context of sub-Saharan Africa (SSA), an increase of 39 USD relative to the contemporary level of where an explicit target for government health spending has spending of 25 USD per capita.3 been established, data are needed to assess progress toward this Rallying calls for increased spending on health inevitably target. This is particularly important given concerns that develop- raises the question of how incremental spending can be ment assistance for health displaces government spending,11,12 financed. In most health systems in low- and middle-income longstanding concerns about the sustainability of development countries (LMICs), out-of-pocket spending to pay for ser- assistance for health in Africa, and recent commitments under the vices, pharmaceuticals, and other health care costs makes Addis Ababa Action Agenda to increase domestic resource mobi- up a large share of spending—estimated at 40% of total lization to finance government services.13 current health spending in 2015.4 It is well established that Against this backdrop, this article aims to provide an reliance on out-of-pocket spending to finance health systems overview of levels of government spending on health in is associated with barriers to access and higher prevalence of East and Southern Africa (ESA). It starts by outlining key impoverishing and catastrophic health expenditures.5,6 issues relating to the definition of government spending and Hence, progress toward UHC in LMICs will require a transi- then presents summary data from the World Health tion in financing, away from out-of-pocket payments toward Organization’s (WHO) Global Health Expenditure Database mandatory prepayment (taxes, other government revenues, (GHED), which has long been the primary source of com- and statutory health insurance) as the main form of domestic parative analysis of government spending.4 The article dis- financing for health care. cusses challenges that arise with interpreting the data. The Several efforts have been made to estimate levels of govern- authors draw on public expenditure reviews for selected ment spending required to make considerable progress toward countries to explore methodological and data issues in the UHC. The 2010 World Health Report6 suggests that broad-based analysis of government health expenditures. access to a core set of services and effective financial protection requires government spending on health in the region of around DEFINITIONS, DATA, AND METHODS 5%. McIntyre and Metheus reach a similar conclusion based on updated analysis of patterns of spending, service coverage, and This article uses the revised System of Health Accounts 2011 outcomes.7 This is a long way from current levels of domestic (SHA 2011) framework for concepts and definitions and government spending in LMICs (1.4% and 2.5% of GDP in 2015, draws primarily on the WHO GHED to analyze spending respectively). Some targets have also been established for the trends and patterns in the ESA region.1 An overview of the share of government budgets dedicated to health. Most notable, data is provided in the Data Annex. The GHED is a data set African heads of state committed in 2001, through the Abuja that provides health expenditure estimates for 190 countries Declaration, to allocate at least 15% of their annual budget to from the year 2000 onwards and follows the SHA 2011 improve the health sector.8 health expenditure accounting framework.4 As defined in The analytic foundations for government health expenditure SHA 2011, health expenditures are considered to include targets have been questioned on numerous grounds. Savedoff “all activities with the primary purpose of improving, main- noted that there is no “right” or “optimal” level of health spending taining and preventing the deterioration of the health status of 286 Health Systems & Reform, Vol. 4 (2018), No. 4 persons and mitigating the consequences of ill-health through Trends and Patterns in Health Expenditures the application of qualified health knowledge” (p.52).14 Domestic Government Health Expenditure Estimates on financing sources and arrangements (i.e., In the ESA region, GGHE-D averaged 2.3% of GDP in financing schemes2) are used to review trends and patterns 2013–2015. This is higher than that of peers in other parts of in government financing and expenditures on health in SSA (the average for SSA is 1.8%) but lower than the proposed ESA, including the role of development assistance for target of 5%.27,28,5 Government schemes are by far the largest health. Domestic general government health expenditures share of GGHE-D, making up on average 97% of total public (GGHE-D) are considered to include transfers from gov- health expenditures in ESA (90.1% in SSA). Social health insur- ernment domestic revenue (FS.1), social insurance contri- ance schemes make up 2.4% and 9.9% of GGHE-D, respectively. butions (FS.3), and compulsory prepayment (FS.4).15 GGHE-D is positively correlated with per capita GDP, Transfers from government domestic revenue include intra- albeit with significant variation across countries (Figure 1). governmental transfers in terms of budgetary allocations, In terms of trends over time, GGHE-D has decreased as a transfers by government on behalf of specific groups (e.g., share of GDP in many countries in the ESA region, falling government may buy voluntary insurance covering the from 2.6% to 2.3% of GDP between 2000–2002 and copayment for the poor), subsidies (e.g., subsidies for 2013–2015. The decrease was more pronounced in low- compulsory health insurance schemes managed by private income countries, where the share of spending in GDP fell companies), and other transfers (e.g., in-kind transfers and from 2.1% to 1.6% of GDP. transfers to nonprofit institutions). Social insurance contri- Despite the decline in the share of domestically financed butions are receipts3 that secure entitlement to social health government health spending in GDP and rapid population insurance benefits. Any contributions or subsidies from growth, total per capita GGHE-D has increased in most government on behalf of specific groups are excluded to countries over the last 15 years. On average in the ESA avoid double counting. Compulsory private insurance pre- region per capita GGHE-D has increased from 60 USD in miums are payments that have been mandated by govern- 2000–2002 to 97 USD in 2013–2015 (62% increase, com- ment and secure entitlement to benefits.14 Estimates of pared to a 54% increase in SSA, from 30 USD to 60 USD). GGHE-D therefore do not include support from external Both of these are, however, significantly below global per sources, such as direct foreign transfers (FS.7) and trans- capita GGHE-D increases, which almost tripled from 271 fers distributed by government from foreign origin (FS.2). USD to 769 USD over the same time period. The rate of These are categorized separately from GGHE-D and con- increase in per capita GGHE-D varies significantly by level sidered as overseas development assistance for health. This of income. Low-income countries in ESA have seen only a article uses estimates of recurrent expenditure4 for the modest increase in per capita GGHE-D allocations from 8.7 analysis because they are more consistent over time and USD in 2000–2002 to 10.3 USD in 2013–2015. Health drive the provision of services today, whereas capital expenditures in countries at higher levels of income have expenditures finance the accumulation of assets required increased at a significantly faster rate than that of lower for future service delivery. The ratio of capital to recurrent income countries, which already allocated less toward health expenditures in the SSA region is on average 1:8, with (Figure 2). Thus, there is a widening gap in per capita health capital expenditures accounting for 1.7% of general gov- expenditures across countries in the region. ernment expenditures (GGEs) and recurrent expenditures accounting for about 6.8% of GGE. Throughout the article, Domestic Resource Mobilization and Prioritization of Health average figures over a three-year period, rather than latest Expenditures year available, are reported to minimize annual fluctuation. Government spending on health reflects both overall domestic resource mobilization and the degree of prioriti- The article also draws on a repository of health sector zation to health.29 Countries’ abilities to collect revenue public expenditure reviews (PERs),16–26 which are field- tends to increase with the level of income as the economy based in-depth assessments of the health expenditure pro- gets formalized. Revenue collection in ESA, excluding file of any given country, to reflect on GHED expenditure grants, is around 23.3% of GDP, which is higher than the estimates and comment on what is driving some of the SSA average of 21.7% but below the global average of differences observed. There are guidelines and good prac- 24.9%. Low- and lower-middle-income countries in the tice notes on PER methodology but, contrary to the region collect comparatively less. Madagascar, Ethiopia, SHA11, there is no single unique methodology that is and Burundi, for example, collect 10.2%, 10.9%, and followed in all countries. 13.1% of GDP in revenue, respectively (Figure 3). Piatti-Fünfkirchen et al.: What Are Governments Spending on Health 287 FIGURE 1. GGHE-D as a percentage of GDP by Level of Income, Average 2013–2015. Source: WHO GHED, based on SHA 2011; World Development Indicators (WDI). FIGURE 2. Level and Change of GGHE-D per Capita. Source: WHO GHED, based on SHA 2011; World Development Indicators (WDI). 288 Health Systems & Reform, Vol. 4 (2018), No. 4 Similarly, budgetary allocations are positively associated with External Financing for Health level of income, albeit with a lot of variation across the region. The reduction in GGHE-D as a share of GDP and government With regards to international benchmarks, Swaziland is the only budgets discussed in the previous section has in part been com- country in ESA that has met the Abuja target of 15% of general pensated for by the availability of donor financing. Development government expenditure in 2013–2015, with South Africa reach- assistance is an important source for financing recurrent health ing close to 14%. The majority of countries, however, fall sig- expenditures in the region and constitutes at times 30% of total nificantly behind, as shown in Figure 4. recurrent expenditures or more in some low-income countries. On average it has increased from 22.6% to 25.5% of total recurrent Further, recent years have seen a prioritization away from expenditures between 2005–2007 and 2013–2015. In per capita health in many countries. The share of government expenditure terms, development assistance for health was 17.0 USD in ESA in on health is estimated to have decreased from 8.4% to 6.7% in the 2013–2015 (compared to 14.9 USD in SSA as a whole). There is, region between 2000 and 2015 (Figure 5). This trend appears to be however, large variation in ESA across countries, with Swaziland more pronounced for low-income countries and is unique to SSA and Botswana receiving 35 USD to 45 USD per capita and other and the East Asia Pacific region. In conclusion, per capita GGHE- countries such as Angola, South Sudan, and Ethiopia receiving D is increasing in absolute terms, though not evenly across the less than 5 USD per capita (Figure 6). HIV/AIDS is likely to be an region, and has not kept pace with economic growth. important factor driving these differences with high levels of Domestically financed government expenditures on health fall external support to high prevalence countries in Southern Africa. well short of Abuja targets in most countries and of resource Especially in low-income countries, donor assistance plays requirements to reach the health SDGs. a significant role and has often matched or exceeded FIGURE 3. Revenue Performance (Excluding Grants) by Level of Income. Note: Revenue as a percentage of GDP used for latest year available; GDP per capita average of 2013–2015. Source: WHO GHED, based on SHA 2011; World Development Indicators (WDI). Piatti-Fünfkirchen et al.: What Are Governments Spending on Health 289 FIGURE 4. Budget Allocations to Health by Level of Income. Note: Revenue as a percentage of GDP used for latest year available; GDP per capita average of 2013–2015. Source: WHO GHED, based on SHA 2011; World Development Indicators (WDI). government contributions to health (Figure 7). Lu et al. esti- country processes. There does not appear to be a clear asso- mate the extent of additionality of donor funds in the health ciation between countries’ levels of income and the modality sector and find that for every dollar of development assis- through which aid is provided. tance to government, the government implicitly reduces spending from its own sources by 46 cents.30 In ESA, how- MUDDYING THE WATERS: ISSUES IN ANALYZING ever, an increase in external financing has tended to be GOVERNMENT HEALTH EXPENDITURES associated with increases in government spending (reflected by countries in the northeastern quadrant of Figure 8). Only The SHA 2011 provides a thorough basis for conceptualizing in a few outlier countries such as Uganda and the Comoros financial flows in a health system, and the WHO GHED has has there been a shift away from health following increased introduced significant changes to better capture financing donor inflows. On the downside, there are a number of sources and schemes. The GHED is hence an invaluable countries where both donor support and government alloca- source of data for country-specific and cross-country analysis tions have significantly reduced, as shown in the southwes- of health expenditures. However, in undertaking such analy- tern quadrant in Figure 8. sis, it is important to be aware of a number of conceptual and The modality through which development assistance is data issues that can have implications for the interpretation of provided has shifted significantly toward greater use of coun- findings. This has been discussed to some extent in the try systems. The share of development assistance routed literature. Witter et al., for example, point to issues such as through government channels6 has increased from 34.3% in the difficulty in measurement, capturing actual expenditures 2005–2007 to 41.3% in 2013–2015. This trend is precipitated as opposed to budgets, differentiating between domestic and by donor countries signing up to the Paris accord and the domestically managed funds, and conceptual challenges of Accra agreement for action in the mid-2000s, emphasizing budget support. They also argue that discretionary and non- the importance of alignment and donor harmonization with discretionary funding should be treated differently.31 290 Health Systems & Reform, Vol. 4 (2018), No. 4 FIGURE 5. Change in Domestic Financing for Health. Source: WHO GHED, based on SHA 2011. 50.0 45.0 40.0 35.0 USD, current 30.0 25.0 20.0 15.0 10.0 5.0 0.0 Rwanda Malawi Mozambique Zambia Lesotho Zimbabwe Uganda Mauritius Seychelles Madagascar Namibia Angola Swaziland Botswana Tanzania Kenya Burundi Comoros Ethiopia South Africa Eritrea South Sudan FIGURE 6. Per Capita Development Assistance for Health, All Sources 2013–2015. Source: GHED based on SHA 2011. To explore these issues, this section considers three questions: and undermine comparability across countries and over time? What is missing from government health expenditure estimates? How is external financing for health accounted for? Although What are the key measurement issues that may affect estimates there is specific guidance on how these questions should be Piatti-Fünfkirchen et al.: What Are Governments Spending on Health 291 FIGURE 7. Relative Importance of Domestic and Donor Financing for Health. Source: WHO GHED, based on SHA 2011. addressed, data limitations and country-specific complexities can, however, be difficult to capture in countries where financial may hamper a consistent approach in data collection and compi- management information systems are insufficiently deployed and lation. To illustrate the resulting challenges in presenting and expenditures are transacted manually. Extrabudgetary funds7 can contrasting country data on health expenditures, this section be difficult to capture, because they are funds that are outside the also compares data from the WHO GHED with those from parameters and controls of conventional budgetary rules and recently completed PERs from selected countries. procedures.32 Social security funds tend to fall into this category. The SHA 2011 makes clear provisions for this, but adequate What Is Missing from Government Expenditure capture may be difficult because they are placed outside the Estimates? budget and not subject to regular financial management processes including reporting. Again, the potential omission of health Although the SHA 2011 clearly defines scope of health expen- expenditures by local government or extrabudgetary funds can diture as “all activities with the primary purpose of improving, result in expenditures being underestimated. Similarly, tax expen- maintaining and preventing the deterioration of the health status ditures, or the deduction from the income tax base of payments of persons and mitigating the consequences of ill-health through for certain medical expenses, can constitute a sizable share of the application of qualified health knowledge” (p.52).14 this GGHE-D and lead to estimation bias if they are insufficiently definition is often difficult to operationalize. Health expenditures captured.33 In some middle-income countries such as South can originate from multiple agencies, including ministries of Africa and Namibia, these could potentially play a key role but finance, defense, education, and agriculture; HIV/AIDS commis- are inherently difficult to monitor and capture. sions; and others. However, in many countries, the budget struc- ture is not sufficiently detailed to clearly identify expenditures on health by these agencies. As a result, estimates of government health expenditures risk underestimating total spending by What Are the Key Measurement Issues That May Bias excluding expenditures by some agencies—such as hospitals Estimates? operated by other ministries. There is clear guidance in the SHA 2011 that expenditure esti- Similarly, estimates of government health spending should mates should be used based on actual expenditure data, rather than include health expenditures at the local government level. These budget or budget release data. These are, however, not always 292 Health Systems & Reform, Vol. 4 (2018), No. 4 FIGURE 8. Change in GGHE-D and Donor Support for Health, 2005–2007 to 2013–2015. Source: WHO GHED, based on SHA 2011. made readily available and often only become available with expenditures may differ from the transfer received as a result significant delay and sometimes only in print format. The WHO of low absorption capacity or a diversion of funds to other GHED does not always clearly state whether estimates were uses. Thus, as noted in the Zambia PER, using central gov- drawn from allocations and outturns. This can significantly bias ernment transfers as a proxy for local government expendi- estimates. A recent study estimates that the proportion of unspent tures will likely be inaccurate, with a tendency toward health budget ranges from 10% to 30% of authorized allocations overestimation of government spending on health.25 in African countries, with some outliers getting close to 60% Arrears in the health sector can be significant and are (Democratic Republic of Congo) unspent.29 important to include because goods and services have been Further, in decentralized systems with low accounting capa- delivered even though payment has not yet been made. As city, government spending is likely to be inadequately captured. shown in the Seychelles PER, they are particularly proble- Recent evaluation of the financial management systems in matic if denominated in a foreign currency and exposed to Zambia and Malawi found that transactions are often captured exchange rate fluctuations.23 However, because such commit- outside the system.34,35 If not recorded adequately, this can lead ments usually happen outside the treasury system, it can be to an underestimation of total expenditures or give the wrong difficult to monitor or capture arrears.36 If arrears payments impression of low budget execution. Furthermore, in countries are captured, this happens at a later stage than the actual such as Tanzania, the chart of accounts is inconsistent across transaction, and often they cannot easily be mapped to their levels of government. This complicates the aggregation of original functional classification. expenditure estimates, especially if not survey based. Central government reporting on expenditures can be prone to error in countries where spending authority has How Is External Financing for Health Accounted For? been decentralized. Central governments frequently provide As we have seen, donor support constitutes a significant transfers to local governments and consider these as share of recurrent health expenditures, especially in low- expenses. However, at the local government level, actual and lower-middle-income countries. Conceptualizing the Piatti-Fünfkirchen et al.: What Are Governments Spending on Health 293 role of donor support and capturing it adequately can, how- expenditures. In Tanzania, for example, the donor health sector ever, be challenging. basket fund for service delivery at the district level is channeled The first question is whether and what part of externally through the development budget line despite there being explicit financed health expenditures should be considered “public” or instruction that these funds may not be used for capital expendi- “government” expenditure. One criterion used for determining tures. If development and recurrent accounts serve an adminis- this is whether external financing is “on budget” or “distributed by trative rather than a functional purpose, this can undermine government,” although this does not ensure alignment with gov- accurate expenditure reporting.24 ernment priorities. For example, donor support can be off budget but on plan, meaning that donors support comprehensive plans but are outside government budgetary processes. They may, how- Comparing Expenditure Estimates to Public Expenditure ever, be fully aligned with government priorities and the govern- Reviews ment is expected to assume these donor-funded expenses over time. Such expenses could arguably be considered government Public expenditure reviews provide an alternative, in-depth expenditure, despite not being on budget. In some countries, assessment of government expenditures on health, and esti- donors do channel their funds through the budget but continue mates of government expenditures frequently deviate from ring-fencing their funds; they also use external processes for the GHED. A review of expenditure estimates from the nine execution, accounting, and reporting. Putting vertical project countries for which both GHED and PER data were available funds on budget does not necessarily guarantee government own- suggests that much of this disconnect can be explained by ership and commitment to sustainability. Until recently, GHED differences in scope, accuracy of measurement, and account- estimates of government spending included some externally ing for the role of donor financing determine.8 financed expenditures, although not consistently, resulting in Comparing recurrent expenditures estimates from the WHO higher estimates of government spending as a share of GDP GHED with those available from PERs in the region shows than reported in this article. This was a source of confusion that close alignment in some countries but significant discrepancies was addressed by SHA 2011, which provides a clear picture of the in others (Figures 9 and 10). Discrepancies do not appear to level of domestically financed government spending. It is, how- follow a clear pattern. In Kenya, Madagascar, and Zimbabwe, ever, much harder to determine what share of external financing GHED estimates are considerably higher than PER estimates, can reasonably be considered part of government spending. An for example. On the other hand, Lesotho is a clear outlier, with assessment of this would likely need to be country specific and PER estimates being significantly higher than those of GHED based on in-depth analysis and dialogue. estimates. Estimates for Malawi, the Seychelles, Mauritius, Capturing donor support is likely to suffer from measurement Tanzania, and Zambia are within a percentage point. error in instances when the GHED draws on Organization for Where is there alignment in expenditure estimates, and Economic Cooperation and Development Development what drives the differences? Not all WHO GHED estimates Assistance Committee (OECD DAC) data. This is based on are backed up by field-based health accounts exercises, and at donor reporting and does not reflect recipient country inputs. times there is a considerable time lapse between the NHA and Because development partners include administrative and other extrapolated years in the GHED. This is likely to impact the expenses in reporting to the OECD, this is likely to produce a accuracy of estimates. Lesotho, for example, does not have an higher estimate than that produced through a survey-based NHA, and the PER recurrent expenditure estimate as a share National Health Accounts (NHA). Mixing spending estimates of GGE is 4 percentage points higher. It is unclear what the at the GHED from NHA and OECD DAC estimates undermine basis for Lesotho estimates in the GHED are, given the the value of the data because the interpreter would have to refer absence of a survey to project from. Important expenditure individually to metadata to assess where the estimates come from items such as the financing of nongovernmental organiza- and how reliable they are likely to be. tions who receive subsidies and deliver services may thus Lastly, it is unclear how general budget support to the treasury have alluded the GHED. The 2007 NHA for Madagascar is captured because it is not possible to determine what source a precedes the 2013 estimate by six years, and the expenditure transaction has been financed with. Sector budget support can estimate exceeds PER estimates by 6.1 percentage points. usually be captured, country systems permitting, and the source The Madagascar PER finds that the health budget has been of expenditures can be mapped confidently. Furthermore, sector contracting, which an extrapolation based on previous years budget support is often channeled through the capital account to would not have been able to capture. Donor fund contribu- allow for earmarking, regardless of the purpose of funds. This tions were estimated to have increased by 300% in Lesotho complicates clear differentiation between recurrent and capital since 2005–2007, an outlier in the region. This is likely to be 294 Health Systems & Reform, Vol. 4 (2018), No. 4 6% 5% 4% 3% 2% 1% 0% Zambia 2015 Kenya 2013 Seychelles 2011 Zimbabwe 2013 Mauritius 2011 Tanzania 2012 Lesotho 2015 Madagascar 2013 Malawi 2012 GHE as % of GDP PER GHED FIGURE 9. Comparing GHED to PER Recurrent Expenditure Estimates: GHE as a Percentage of GDP. Source: WHO GHED, based on SHA 2011 and Various World Bank PERs. 12% 10% 8% 6% 4% 2% 0% Seychelles 2011 Zambia 2015 Lesotho 2015 Zimbabwe 2013 Kenya 2013 Malawi 2012 Mauritius 2011 Tanzania 2012 Madagascar 2013 GHE as % of GGE PER GHED FIGURE 10. Comparing GHED to PER Recurrent Expenditure Estimates: GHE as a Percentage of GGE. Source: WHO GHED, based on SHA 2011 and Various World Bank PERs. overestimated by drawing on OECD DAC data because no from equally poor data. The Malawi PER, for example, only survey-based NHA was available. uses budget and not expenditure data. The Malawi 2011 Differences in scope are likely to bias expenditure estimates Public Expenditure and Financial Accountability (PEFA) in the Kenya, Madagascar, and Zimbabwe PERs downwards assessment finds that health sector expenditure was repeat- because they only account for Ministry of Health expenditures edly and significantly below the budget. Budget perfor- and thus explain some of the disconnect with the GHED. For mance was 8.0%, 0.2%, and 10.0% below target between example, the GHED tries to capture health-related cross-sec- 2009 and 2011,37 and expenditure estimates based on bud- toral expenditures including nutrition, which are not captured gets only in the Malawi PER are thus likely to be over- in these PERs. Other relevant agency expenditures including estimated in that range. the military, AIDS commissions, and subsidies going directly In Tanzania, the PER is cognizant of differentiating between to extrabudgetary funds are also outside the scope and there- the functional allocation of recurrent and capital expenditures. fore are not captured. Some PERs, like Zambia and the This required a forensic review of the budget documentation, Seychelles, have taken a more inclusive approach attempting which is outside the scope of the GHED. Because expenditures to capture expenditures from these agencies and the scope is were remapped to recurrent expenditures, this may in part thus more closely aligned with the GHED. explain the higher PER estimate. Similarly, in Zambia, it A number of PERs note the difficulty in obtaining reli- required a detailed review to identify local government expen- able expenditure data. Though expenditure estimates in ditures that were off the books, which may explain the larger Malawi appear to be similar, they are likely to be drawn PER estimates than that provided through the GHED. Piatti-Fünfkirchen et al.: What Are Governments Spending on Health 295 CONCLUSION efforts to strengthen national data capacity and international efforts to promote quality and consistency of data. Per capita expenditure from domestic sources on recurrent Finally, although this article focused on the level of govern- health activities varies widely across countries in the East and ment spending on health, it is important to remember that spend- Southern Africa region and is on average 97 USD, signifi- ing is ultimately a means to achieve results in the health sector. cantly above SSA averages. Per capita expenditure has grown From this perspective, questions of how effective spending is in across the region, though at a faster rate in higher-income achieving better results is just as important as overall levels of countries. Thus, there is a widening gap in per capita health spending. This points to the importance of in-depth analytical expenditures. Though absolute spending on health has work to inform policy. The GHED is an invaluable resource for increased, this has been at a slower rate than GDP, and monitoring and benchmarking health expenditures. For advocacy domestically financed government spending on health as a and policy dialogue, it is most effectively used in combination share of GDP is low and declining. with in-depth country assessments. Countries’ available budgets for health are a function of their capacities to raise resources and their willingness to NOTES prioritize health in the budget. Though overall per capita GGHE-D is increasing, it has not kept pace with GDP [a] Countries in the ESA subregion follow the United Nations Development growth, reflecting an apparent prioritization away from Program definition and include Angola, Botswana, Burundi, Comoros, health. Reaching the Abuja target of allocating 15% of total Eritrea, Ethiopia, Kenya, Lesotho, Madagascar, Malawi, Mauritius, Mozambique, Namibia, Rwanda, Seychelles, South Africa, Somalia, government expenditure to health has for many countries South Sudan, Swaziland, Tanzania, Uganda, Zambia, and Zimbabwe. become a distant and unrealistic goal. Domestically financed Analysis is done for countries for which data were available. government expenditures on health fall well short of Abuja [b] Health care financing schemes are perceived as the main “building targets in most countries and of resource requirements to blocks” of the structure of a country’s health financing system: they reach the health SDGs. are the main types of financing arrangements through which people can Donor funds play a key role in the region and make up a get access to health care; for example, government schemes, social insurance, and voluntary insurance. significant share of total recurrent health expenditures, at times surpassing government contributions. The availability of donor [c] These could be from employers on behalf of their employees or from the funds appears to be closely associated with whether government employees, the self-employed, or the nonemployed on their own behalf. is prioritizing health. In contexts where governments have [d] Recurrent expenditures are defined as the “final consumption expendi- increased spending, there has been increased donor activity and, ture of resident units on health care goods and services, including the conversely, where government health expenditure has contracted, health care goods and services provided directly to individual persons as well as collective health care services.”14 The SHA 2011 explicitly so has donor expenditure. Donor dependence is high for many discourages the aggregation of recurrent and capital expenditures. The low- and lower-middle-income countries and is not consistent with 2018 GHED data release explicitly differentiates between recurrent their stated ambitions of progress toward UHC. expenditures and capital formation for the first time. The GHED for 2000–2015 is the first attempt to capture [e] The WHO has never formally adopted 5% of national income as a recom- comprehensive health expenditure data based on the SHA 2011, mended level of health spending. Instead, the WHO committed itself to which has brought about significant improvements from previous monitoring the number of countries where at least 5% of Gross National attempts. If done correctly, there should be no differences in Product (GNP) is spent on health. expenditure estimates between GHED and PER estimates, though [f] These are transfers from foreign origin distributed through the general they serve different purposes and at times draw on different data. government and captured by the government budget. This article, however, highlighted important conceptual and data [g] Expenditures that are not included in the annual budget or are not subject issues with the GHED, in particular with regards to scope, mea- to the same general level of reporting, regulation, or audit as other public surement, and the role of external financing, which may drive finance items. some of the observed deviations in expenditure estimates. These issues are important because they undermine the credibility and [h] Metadata from the GHED are not sufficiently detailed to allow for a granular comparison of GHED and PER methods that would shed effectiveness of the data for advocacy and policy dialogue. Given light on what drives the observed differences. their consistent methodological application, they lend themselves best for trend and cross-country analysis. More transparency on data and methods in how the individual country GHED estimates DISCLOSURE OF POTENTIAL CONFLICTS OF are derived and shortcomings would increase credibility and help INTEREST the analyst in interpretation. It will remain important to continue No potential conflict of interest was reported by the authors. 296 Health Systems & Reform, Vol. 4 (2018), No. 4 REFERENCES 18. World Bank. Tanzania health sector public expenditure review 2018. Washington (DC): World Bank; Forthcoming. 1. Stenberg K, Hanssen O, Edejer -T-T-T, Bertram M, Brindley C, 19. World Bank. Madagascar public expenditure review 2014. Meshreky A, Rosen JE, Stover J, Verboom P, Sanders R, et al. Washington (DC): World Bank; 2015. Health Sector Background Financing transformative health systems towards achievement of Paper. the health Sustainable Development Goals: a model for projected 20. World Bank. Malawi public expenditure review. Washington resource needs in 67 low-income and middle-income countries. (DC): World Bank; 2013. Report No. 79865 - MW. Lancet Glob Health. 2017;5(9):e875–e887. doi:10.1016/S2214- 21. World Bank. Mauritius policy notes: building analytical capa- 109X(17)30263-2. city to raise public sector efficiency. Washington (DC): World 2. World Health Organization. Macroeconomics and health: Bank; 2013. investing in health for economic development: report of the 22. World Bank. Mozambique public expenditure review. commission on Macroeconomics and Health. Geneva Addressing the challenges of today, seizing the opportunities (Switzerland): World Health Organization; 2001. of tomorrow. Washington (DC): World Bank; 2014. 3. Taskforce on Innovative International Financing for Health 23. World Bank. Seychelles. Programmatic public expenditure review Systems. Constraints to scaling up and costs: working group policy notes - health education & investment management. 1 report. Geneva: International Health Partnership; 2009. Washington (DC): World Bank; 2014. 4. World Health Organization. Global Health Expenditure 24. World Bank. United Republic of Tanzania. Public expenditure Database (GHED). Geneva (Switzerland): International Health review 2011. Washington (DC): World Bank; 2012 Forthcoming. Partnership; 2018. 25. World Bank. Zambia health sector public expenditure review 5. Xu K, Evans DB, Kawabata K, Zeramdini R, Klavus J, Murray 2018. Washington (DC): World Bank; Forthcoming. CJ. Household catastrophic health expenditure: a multicountry 26. World Bank. Health public expenditure review Zimbabwe. analysis. Lancet. 2003;362(9378):111–117. doi:10.1016/ Washington (DC): World Bank; 2015. S0140-6736(03)13861-5. 27. World Health Organization. Global strategy for health for all by 6. World Health Organization. World Health Report, 2010: health the year 2000. Geneva (Switzerland): World Health systems financing the path to universal coverage. Geneva Organization; 1981. (Switzerland): World Health Organization; 2010. 28. World Health Organization. How much should countries spend 7. Meheus F, McIntyre D. Fiscal space for domestic funding of on health. Geneva (Switzerland): World Health Organization; health and other social services. Health Econ Policy Law. 2003. Discussion Paper No3. 2017;12(2):159–177. doi:10.1017/S1744133116000438. 29. World Health Organization. Public financing for health in 8. African Union. Abuja declaration on HIV/AIDS, tuberculosis and Africa: from Abuja to the SDGs. 2016. other related infectious diseases. OAU/SPS/Abuja/3. 2001;27. 30. Lu C, Schneider MT, Gubbins P, Leach-Kemon K, Jamison D, 9. Savedoff WD. What should a country spend on health care? Murray CJ. Public financing of health in developing countries: Health Aff. 2007;26(4):962–970. doi:10.1377/hlthaff.26.4.962. a cross-national systematic analysis. Lancet. 2010;375 10. Jowett M, Brunal MP, Flores G, Cylus J. Spending targets for (9723):1375–1387. doi:10.1016/S0140-6736(10)60233-4. health: no magic number. Geneva (Switzerland): World Health 31. Witter S, Jones A, Ensor T. How to (or not to) … measure Organization; 2016. performance against the Abuja target for public health expen- 11. Dieleman JL, Hanlon M. Measuring the displacement and diture. Health Policy Plan. 2014 Jul;29(4):450–455. replacement of government health expenditure. Health Econ. doi:10.1093/heapol/czt031. 2014;23(2):129–140. doi:10.1002/hec.3016. 32. Radev D, Allen R. Managing and controlling extrabudgetary 12. Morrissey O. Aid and government fiscal behavior: assessing funds (EPub). Washington (DC): International Monetary Fund; recent evidence. World Dev. 2015;69:98–105. doi:10.1016/j. 2006. worlddev.2013.12.008. 33. Lowry S. Health-related tax expenditures: overview and analysis. 13. UN General Assembly. Addis ababa action agenda of the Third Washington (DC): Congressional Research Service; 2016. CRS International Conference on financing for development. 2015. Report. Addis Ababa Action Agenda, Resolution A/RES/69/313, 34. World Bank. Project performance assessment report. evaluation of undocs. org/A/RES/69/313 [2 June 2016]. the Zambia public sector management program support project. 14. OECD Eurostat and WHO. A system of health accounts 2011: Washington (DC): World Bank; 2016. Report No. 106280. organisation for economic co-operation and development. 2017. 35. World Bank. Project performance assessment report. Evaluation of 15. World Health Organization. Technical brief on the indicators the Malawi financial management, transparency, and accountability published on the World Health Organization’s global health project. Washington (DC): World Bank; 2016. p. 103060. expenditure database. World Health Organ. 2017. p. 1–16. 36. Flynn PS, Pessoa M. Prevention and management of govern- 16. World Bank. Ethiopia public expenditure review. Washington ment expenditure arrears. Washington (DC): International (DC): World Bank; 2013. p. ACS14541. Monetary Fund; 2014. 17. World Bank. Laying the foundation for a robust health care 37. PEFA. Public finance management assessment for Malawi system in Kenya. Kenya public expenditure review. Washington based on the public expenditure and financial accountability (DC): World Bank; 2014. framework. 2011. Piatti-Fünfkirchen et al.: What Are Governments Spending on Health 297 DATA ANNEX Country Indicator Average 2000–2002 Average 2013–2015 Botswana GGHE-D as % GGE 9.77 8.57 GGHE-D as % GDP 4.07 3.02 GGHE-D per capita in USD 129.71 210.14 External health expenditure per capita in USD 1.42 36.98 External health expenditure as % GGE N/A 1.51 GDP in current USD per capita 3,177.47 6,989.50 Burundi GGHE-D as % GGE 6.43 6.87 GGHE-D as % GDP 1.59 2.02 GGHE-D per capita in USD 2.07 5.84 External health expenditure per capita in USD N/A 11.40 External health expenditure as % GGE N/A 12.97 GDP in current USD per capita 130.71 286.33 Comoros GGHE-D as % GGE 7.54 3.80 GGHE-D as % GDP 1.53 1.02 GGHE-D per capita in USD 6.19 8.22 External health expenditure per capita in USD 1.86 6.45 External health expenditure as % GGE N/A 2.98 GDP in current USD per capita 401.68 804.95 Eritrea GGHE-D as % GGE 2.81 1.60 GGHE-D as % GDP 1.54 0.65 GGHE-D per capita in USD 3.20 5.53 External health expenditure per capita in USD 2.79 7.64 External health expenditure as % GGE N/A 2.25 GDP in current USD per capita 208.37 840.33 Ethiopia GGHE-D as % GGE 8.09 5.42 GGHE-D as % GDP 1.97 0.97 GGHE-D per capita in USD 2.29 5.31 External health expenditure per capita in USD 0.85 4.55 External health expenditure as % GGE N/A 4.63 GDP in current USD per capita 116.51 550.19 Kenya GGHE-D as % GGE 9.31 6.75 GGHE-D as % GDP 2.07 1.80 GGHE-D per capita in USD 8.27 23.36 External health expenditure per capita in USD 1.00 14.74 External health expenditure as % GGE N/A 4.22 GDP in current USD per capita 400.53 1,301.78 Lesotho GGHE-D as % GGE 8.20 8.03 GGHE-D as % GDP 3.27 4.15 GGHE-D per capita in USD 14.65 48.37 External health expenditure per capita in USD N/A 32.07 External health expenditure as % GGE N/A 5.25 GDP in current USD per capita 449.12 1,175.30 Madagascar GGHE-D as % GGE 13.56 13.27 GGHE-D as % GDP 2.33 1.98 GGHE-D per capita in USD 6.11 8.61 External health expenditure per capita in USD 2.70 5.22 External health expenditure as % GGE N/A 8.00 GDP in current USD per capita 262.26 438.66 (Continued on next page ) 298 Health Systems & Reform, Vol. 4 (2018), No. 4 Country Indicator Average 2000–2002 Average 2013–2015 Malawi GGHE-D as % GGE 7.38 9.37 GGHE-D as % GDP 1.81 2.32 GGHE-D per capita in USD 3.31 8.19 External health expenditure per capita in USD 3.22 22.10 External health expenditure as % GGE N/A 25.26 GDP in current USD per capita 197.00 351.34 Mauritius GGHE-D as % GGE 7.05 9.48 GGHE-D as % GDP 1.69 2.34 GGHE-D per capita in USD 66.62 224.14 External health expenditure per capita in USD N/A 7.75 External health expenditure as % GGE N/A 0.33 GDP in current USD per capita 3,942.48 9,559.11 Mozambique GGHE-D as % GGE 14.47 2.88 GGHE-D as % GDP 3.34 1.11 GGHE-D per capita in USD 8.56 6.72 External health expenditure per capita in USD N/A 21.42 External health expenditure as % GGE N/A 9.78 GDP in current USD per capita 257.26 585.97 Namibia GGHE-D as % GGE 21.00 13.28 GGHE-D as % GDP 6.91 5.33 GGHE-D per capita in USD 119.29 277.01 External health expenditure per capita in USD 8.38 32.86 External health expenditure as % GGE 1.58 GDP in current USD per capita 1,724.51 5,215.73 Rwanda GGHE-D as % GGE 5.42 7.38 GGHE-D as % GDP 1.23 1.99 GGHE-D per capita in USD 2.51 14.30 External health expenditure per capita in USD 3.74 23.74 External health expenditure as % GGE 12.29 GDP in current USD per capita 204.58 718.96 Seychelles GGHE-D as % GGE 7.10 9.42 GGHE-D as % GDP 3.73 3.27 GGHE-D per capita in USD 290.45 470.05 External health expenditure per capita in USD N/A 8.10 External health expenditure as % GGE 0.16 GDP in current USD per capita 7,798.59 14,384.90 South Africa GGHE-D as % GGE 10.40 14.03 GGHE-D as % GDP 2.58 4.29 GGHE-D per capita in USD 69.64 271.05 External health expenditure per capita in USD 2.99 11.60 External health expenditure as % GGE 0.60 GDP in current USD per capita 2,685.79 6,330.78 Swaziland GGHE-D as % GGE 10.11 15.07 GGHE-D as % GDP 2.57 4.50 GGHE-D per capita in USD 37.86 156.65 External health expenditure per capita in USD 1.20 43.49 External health expenditure as % GGE 4.17 GDP in current USD per capita 1,474.97 3,478.07 (Continued on next page ) Piatti-Fünfkirchen et al.: What Are Governments Spending on Health 299 Country Indicator Average 2000–2002 Average 2013–2015 Uganda GGHE-D as % GGE 9.63 6.77 GGHE-D as % GDP 2.11 1.14 GGHE-D per capita in USD 4.98 7.84 External health expenditure per capita in USD 4.92 18.50 External health expenditure as % GGE 16.04 GDP in current USD per capita 236.25 684.10 Zambia GGHE-D as % GGE 7.59 6.41 GGHE-D as % GDP 2.14 1.70 GGHE-D per capita in USD 6.99 26.50 External health expenditure per capita in USD 7.93 21.31 External health expenditure as % GGE 5.13 GDP in current USD per capita 326.75 1,586.26 Source: WHO GHED, based on SHA 2011.